UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Structural and dynamic analysis of oligosaccharide binding by CBDn1 Johnson, Philip Edward 1997

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_1998-271714.pdf [ 18.57MB ]
Metadata
JSON: 831-1.0061647.json
JSON-LD: 831-1.0061647-ld.json
RDF/XML (Pretty): 831-1.0061647-rdf.xml
RDF/JSON: 831-1.0061647-rdf.json
Turtle: 831-1.0061647-turtle.txt
N-Triples: 831-1.0061647-rdf-ntriples.txt
Original Record: 831-1.0061647-source.json
Full Text
831-1.0061647-fulltext.txt
Citation
831-1.0061647.ris

Full Text

STRUCTURAL AND DYNAMIC ANALYSIS OF OLIGOSACCHARIDE BINDING BY C B D N I by PHILIP EDWARD JOHNSON B.Sc, Simon Fraser University, 1991  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES DEPARTMENT OF CHEMISTRY  We accept this thesis as conforming to the required standard  THEVNIVERSITY OF BRITISH COLUMBIA November 1997 © Philip Edward Johnson, 1997  In  presenting  degree freely  at  this  the  available  copying  of  department publication  of  in  University  of  for  this or  thesis  this  British  reference  thesis by  partial  for  his thesis  and  for  her  of  T h e U n i v e r s i t y o f British Vancouver, Canada  Date  DE-6 (2/88)  Columbia,  I  Columbia  I  further  purposes  gain  the  shall  requirements  agree  that  agree  may  representatives.  financial  permission.  Department  of  study.  scholarly  or  fulfilment  It not  be  that  the  Library  permission  granted  is  by  understood be  for  allowed  an  advanced  shall for  the that  without  head  make  it  extensive of  my  copying  or  my  written  11  Abstract  This thesis describes the biophysical and structural characterization of CBDNI, the aminoterminal cellulose-binding domain (CBD) from the Cellulomonas  fimi  (3-1,4-glucanase CenC.  Heteronuclear multidimensional nuclear magnetic resonance (NMR) methods were used to determine the three dimensional structure of CBDNI in the presence of saturating amounts of cellotetraose. It was found that CBDNI is composed of 10 (3-strands, folded into two antiparallel p-sheets with the topology of a jelly-roll p-sandwich. The dominant feature of the  CBDN I  structure is a cleft which runs along the length of one face of the molecule. On the basis of perturbations of the NMR spectrum of CBDNI due to the addition of sugar, it was shown that CBDNI binds soluble cellooligosaccharides within the cleft. Intermolecular nuclear Overhauser enhancements (NOEs) between residues in the cleft and the bound sugar confirmed that this face is responsible for ligand binding. Association constants, determined from the dependence of the amide 'H and N chemical shifts upon added sugar, were found to increase l5  with increasing sugar length, reaching a maximum at a ligand length of five glucose units. This corresponds approximately to the length of the binding cleft. The binding cleft is composed of a strip of hydrophobic residues, flanked by hydrophilic residues. It is proposed that the pyranose rings of the sugar lie over the hydrophobic residues while the polar side-chains are involved in hydrogen bonding interactions with the equatorial hydroxyl groups of the glucose rings.  CBDNI  binds a calcium ion at a site opposite the oligosaccharide binding face. The binding affinity of oligosaccharides is unaffected by the presence of calcium. CBDNI  binds nitroxide spin-labelled oligosaccharides in multiple orientations. In one  orientation the spin-label group lies near residue alanine 18, in the other near glycine 86. This ability to bind the same ligand in several orientations indicates that different hydrogen bonding combinations between the sugar and protein must occur. This is likely related to the motional disorder observed for residues present in the binding face as determined by NMR relaxation methods.  iii  Table of Contents  Abstract  ii  Table of Contents  iii  List of Tables  vii  List of Figures  viii  Abbreviations  xii  Acknowledgments  xiv  Chapter 1- Cellulose Binding Domains: Structures and Binding Domains  1  Introduction  1  CBDs that bind crystalline cellulose : Structures Identification of binding face (i) CBDcBHI (ii) CBDcex Mechanism for binding crystalline cellulose Structural considerations of CBDs binding to crystalline cellulose  5 5 6 6 8 9 10  CBDs that do not bind crystalline cellulose Structure Binding specificity and thermodynamics of C B D N I  13 13 16  Summary  17  Chapter 2- Cellooligosaccharide Binding by C B D N I  18  Abstract  18  Introduction Background Ligand binding studied by NMR  19 19 19  Experimental methods Expression and Purification of C B D N I Characterization of C B D N I C D and FTIR spectroscopy NMR Spectroscopy Titration of C B D N I with Cellooligosaccharides Monitored by N M R N M R studies of reduced C B D N I  21 21 26 26 27 28 29  iv Results CD and FTIR spectroscopy C B D N I Binds Soluble Cellooligosaccharides Binding Stoichiometry Association Constants Determined by NMR Spectroscopy Identification of the Cellooligosaccharide Binding Site Effect of Cellotetraose Binding on the Aromatic Residues in C B D N I Effect of High Concentrations of Cellotetraose Reduction of the disulphide in C B D N I results in the protein unfolding  29 29 31 33 37 38 43 46 46  Discussion  48 48  CBDNI  Binds Soluble Cellooligosaccharides  Binding Site is Formed by a 5-Stranded (3-sheet Comparison to Other CBDs Stability of C B D N I  Chapter 3- Structure Determination of CBDNI  48 49 50  52  Abstract  52  Introduction Resonance assignment of proteins Structure calculations Experimental methods Sample Preparation N M R Spectroscopy Amide Hydrogen Exchange Structure Calculations  53 53 55 60 60 61 65 65  Results Main Chain Resonance Assignments Aliphatic Side-Chain Resonance Assignments Aromatic Side-Chain Assignments Stereospecific Assignments and Side-Chain Torsion Angle Restraints Secondary Structure Determination Tertiary Structure  66 67 67 71 72 77 82  Discussion Structure Binding Mechanism Comparison to Other Family IV CBDs Comparison to CBDs that bind crystalline cellulose  92 92 93 95 95  Structural Similarity with 1,3-1,4-|3-Glucanase  97  CBDNI  Chapter 4- Calcium Binding by CBD i N  101  Abstract  101  Introduction Background  101 101  V  Experimental Methods Sample preparation NMR spectroscopy Determination of binding constants Analysis of binding kinetics Structure calculations of calcium-loaded  102 102 103 104 105 105  CBDNI  Results  106 binds metal ions 106 Calcium equilibrium binding association constant and binding stoichiometry .... 106 Effect of calcium on cellooligosaccharide affinity 112 Kinetics of calcium binding 114 Identification of calcium binding site 115 CBDNI  Discussion Location of the metal binding site and structure of calcium-loaded Binding affinity and kinetics A structural role for calcium binding by C B D N l Chapter 5- Dynamic Analysis of Ligand Binding  CBDNI  121 121 124 126 128  Abstract  128  Introduction Background N relaxation and dynamics Methyl group deuteron relaxation and dynamics  129 129 130 134  Experimental Methods Sample preparation for N relaxation experiments Preparation of fractionally deuterated C B D N I N relaxation experiments Methyl group deuteron relaxation experiments Data processing Resonance assignment of the calcium-loaded C B D N I Ultracentrifugation and viscosity measurements  136 136 136 137 139 140 142 143  Results N T i , T and NOE values Estimation of the overall correlation time, T T1/T2 analysis Model-free analysis of backbone N relaxation data Ultracentrifuge and viscosity measurements Resonance assignment of calcium-loaded C B D N I Methyl deuteron relaxation and analysis  144 144 150 151 153 157 157 159  Discussion Comments on the model-free analysis of N relaxation Global'trends Effect of calcium binding Effect of cellooligosaccharide binding on the dynamics of C B D N I  169 169 170 173 176  1 5  l 5  15  l 5  2  m  l 5  l 5  vi i) Backbone N relaxation data ii) Methyl containing deuteron relaxation Changes in t with viscosity  176 178 181  1 5  m  Chapter 6- Structural Analysis of Ligand Binding  183  Abstract  183  Introduction Background Theory on the use of spin labels  184 184 184  Experimental methods Spin labelled cellooligosaccharides Preparation of C B D N I samples and titration of nitroxide-labelled cellooligosaccharides NMR spectroscopy of nitroxide-labelled cellooligosaccharide-CBDNi complexes Data analysis and calculation of A(Rj) and A(R ) Titration of mutant C B D N I protein samples monitored by NMR  187 187  Results Binding of the nitroxide-labelled cellooligosaccharides to C B D N I C B D N I binds nitroxide-labelled cellooligosaccharides in two distinct orientations Measurement of ARi and AR values  191 191  2  2  Calculation of x values and electron-proton distances Cellooligosaccharide binding of C B D N I variants  187 188 189 190  192 197 201 204  c  Discussion Binding affinity of the nitroxide-labelled cellooligosaccharides Measurement of the relaxation parameters AR| and AR and calculation of  207 207  2  T and r values Ligand binding by C B D N I mutants Changes in the line-width of binding face residues in the C B D N I mutants Implications for the association of C B D N I with cellooligosaccharides c  207 211 212 214  Appendix  220  Bibliography  227  vii  List o f Tables Table 1.1  Biochemical data for CBDs of known structure  5  Table 2.1  Association constants for the binding of cellooligosaccharides by C B D N I  38  Table 3.1  Acquisition and processing parameters for NMR experiments recorded on C B D N I for use in the structure determination process  Table 3.2  Coupling constants and % 1 assignments for valine residues in C B D N I  Table' 3.3  Coupling constants and %l assignments for isoleucine and threonine residues in C B D N I  S  Table 3.4  Structural statistics and rms differences  Table 5.1  Description of samples used in N relaxation experiments  Table 5.2  Summary of model-free parameters  Table 5.3  Acquisition and processing parameters for NMR experiments used to  74 75 85  15  assign calcium-loaded  62  138 142  143  CBDNI  Table 5.4  Number of residues per sample for which N relaxation data was obtained  147  Table 5.5 Table 5.6  Values for the overall correlation time (x ) determined for C B D N I Summary of the number of resonances occurring for each spectral density model used to fit N Ti, T and 'H- N NOE data for the different C B D N I samples  151  15  m  15  15  2  Table 5.7  Average S is values calculated for each methyl type in C B D N I  Table 5.8  Comparison of the spectral density models used to fit the experimental data with original and introduced minimum errors  Table 6.1 Table 6.2 Table 6.3  2  ax  153 166 170  Association constants for C B D N I binding TEMPO-labelled cellooligosaccharides  192  Order of disappearance of residues in 'H- N HSQC spectra induced by TEMPO-labelled cellotetraose binding  195  I5  Order of disappearance of residues in 'H- N HSQC spectra induced by TEMPO-labelled cellotriose binding 15  196  viii List of Figures  Figure 1.1  Cellulases produced by Cellulomonas fimi  Figure 1.2  Ribbon diagrams of CBDs of known structure  Figure 1.3  Surface representation of CBDs  7  Figure 1.4  Schematic view of the cross section of the structure of cellulose  11  Figure 1.5  Widths of the binding faces of the CBDs that bind to crystalline cellulose  Figure 2.1  Plasmid map of pTugNln  Figure 2.2  SDS PAGE gel showing the overexpression of C B D N I  Figure 2.3  SDS PAGE gel showing the purification of C B D N I on Avicel  25  Figure 2.4  FTIR spectrum of C B D N I  30  Figure 2.5  CD spectra of C B D N I  32  Figure 2.6  Titration of C B D N I with cellotetraose monitored by 'H- N HSQC spectra  Figure 2.7  Overlayed fits for C B D N I binding cellotetraose  35  Figure 2.8  Association constants for C B D N I binding cellooligosaccharides  36  Figure 2.9  Pattern of N and H chemical shift perturbation with cellotetraose binding ... .40  Figure 2.10  Schematic fold of C B D N I  •  42  Figure 2.11  Aromatic region of DQF-COSY spectra of C B D N I in the presence and absence of cellotetraose  44  Figure 2.12  2 4  22 24  15  I5  14  34  N  Histogram summarizing the perturbation of aromatic ring 'H chemical shifts with cellotetraose binding  45  'H- N HSQC spectra of native C B D N I and C B D N I denatured by the reduction of the disulphide and by urea  47  Figure 3.1  Flowchart of the structure determination process  54  Figure 3.2  Three-dimensional heteronuclear experiments used to assign the resonances of C B D N I  56  HNCACB and CBCA(CO)NH spectra demonstrating the sequential assignment of C B D N I  57  Figure 2.13  Figure 3.3  15  ix Figure 3.4  (H)C(CO)NH-TOCSY and H(CCO)NH-TOCSY spectra demonstrating the assignment side-chain resonances of C B D N I  58  Figure 3.5  Labelled H- N HSQC spectrum of cellotetraose-bound C B D N I  68  Figure 3.6 Figure 3.7  Spectra of the side-chain NH assignment process Portion of the constant time 'H- C HSQC spectrum and H- C HSQC spectrum of C B D N I fractionally 10% C labelled  70  l  [5  2  13  ]  I3  73  13  Figure 3.8  Strip plot of the long range C- C coupling experiment  Figure 3.9  Summary of data used to determine the secondary structure of C B D N I  78  Figure 3.10  Cross strand NOEs between the (3-strands of sheets A and B in C B D N I  80  Figure 3.11  Strip plot of the N HSQC-NOESY spectrum of cellotetraose-bound CBD i ..81  Figure 3.12  Schematic diagram showing the fold of C B D N I  Figure 3.13  C traces of the ensemble of 25 structures of C B D N I  Figure 3.14  Structural statistics for C B D N I  Figure 3.15  Ramachandran plot of the mean minimized C B D N I structure  Figure 3.16  Superposition of the side-chains of residues that lie in the hydrophobic core of C B D N I 89 Diagram of C B D N I showing the side-chains present on the oligosaccharide binding face 91  Figure 3.17  l3  76  13  15  N  83  a  86 8  7  88  Figure 3.18  Alignment of the amino acid sequences of family IV CBDs  96  Figure 3.19  Ribbon diagrams of C B D N I and the hybrid  98  Figure 3.20  Superposition of C B D N I and the hybrid 1,3-1,4-P-glucanase showing the catalytic residues of the glucanase and the corresponding residues in C B D N I  1,3-1,4-fJ-glucanase  B a c i l l u s  B a c i l l u s  100  Figure 4.1  Screen for metal binding using one-dimensional 'H NMR  107  Figure 4.2  Titration of C B D N I with calcium monitored by one-dimensional 1H NMR  108  Figure 4.3  Titration of C B D N I with calcium monitored by H- N HSQC spectra  109  Figure 4.4  Determination of K values for calcium-binding by C B D N I  Ill  Figure 4.5  Determination of K values for cellopentaose binding by calcium-loaded C B D N I  1  15  A  A  113  X  Figure 4.6  Comparison of experimental and simulated lineshapes to obtain the k ff value for calcium association/dissociation 0  116  Figure 4.7  Pattern of N and H chemical shift perturbation with calcium binding  Figure 4.8  C worm diagram of C B D N I showing the location of residues that experience the most significant changes in chemical shift with calcium binding 118  Figure 4.9  Calcium titration of C B D N I monitored by 'H- C HSQC spectra  Figure 4.10  Structure of Bacillus macerans 1,3-1,4-p-glucanase showing the location of the calcium and ligating groups  15  N  117  a  120  I3  123  Figure 4.11  Structures of C B D N I calculated with and without a calcium ion  Figure 5.1  Plots of Ti, T , and heteronuclear NOE vs. i  Figure 5.2  Portion of spectra used to measure T values  Figure 5.3  Two-parameter monoexponential curve fitting to obtain Tj and T values  146  Figure 5.4  15  N T i, T and H- N NOE data for calcium-bound CBD i in the absence and presence of cellopentaose  148  Figure 5.5  125  132  m  2  145  2  2  1  15  2  15  N  N Ti, T and !H- N NOE data for C B D N I in the absence and 15  2  presence of calcium  149  Figure 5.6  l5  N Ti/T values for C B D N I  152  Figure 5.7  Model-free parameters for calcium-bound C B D N I in the absence and  2  presence of cellopentaose  155  Figure 5.8  Model-free parameters for C B D N I in the absence and presence of calcium  156  Figure 5.9  Values of viscosity and tm plotted against protein concentration  Figure 5.10 Figure 5.11  'H- N HSQC spectrum of calcium-loaded cellooligosaccharide-free C B D N I • • 160 Spectrum illustrating the N-H correlations of residues following  158  15  15  N  methyl containing amino acids  161  Figure 5.12  'H- C HSQC spectrum of calcium-loaded cellooligosaccharide-free C B D N I ... 162  Figure 5.13  Cellopentaose titration of calcium-loaded C B D N I monitored by 'H- C HSQC spectra  I3  13  163  Figure 5.14  Portion of spectra used to measure the decay rates of I C D magnetization  Figure 5.15  Examples of the decay of I C D , I C D and I C magnetization  165  Figure 5.16  Ti(D) and Ti (D) values of methyl groups  167  Z  Z  p  Z  Z  z  z  y  Z  Z  Z  Z  164  xi 168  Figure 5.17  Order parameters for methyl groups  Figure 5.18  Comparison between rms deviation of the ensemble of 25 structures and 'H- N NOE values  172  Structure of C B D N I showing methyl groups present in the oligosaccharide-binding face  179  15  Figure 5.19  Figure 6.1  "H- N HSQC spectra of C B D N I bound to TEMPO-labelled cellotetraose  193  Figure 6.2  Schematic view of the structure of C B D N I  194  Figure 6.3  ARj, AR and ALW values resulting from the binding of TEMPO-labelled cellotriose  198  AR\, AR and ALW values resulting from the binding of TEMPO-labelled cellotetraose  199  Figure 6.4 Figure 6.5  15  2  2  C worm diagram of C B D N I showing the location of residues with the largest a  ARi, AR and ALW values with TEMPO-labelled cellotetraose binding 2  200  Figure 6.6  i  Figure 6.7  Values for the electron-proton distance (r)  Figure 6.8  'H-^N HSQC spectra of wild-type C B D N I and the variants Y19A and Y85A ..205  c  values for the electron-proton vector  202 203  Figure 6.9  1  H- N HSQC spectra from the titration of Y85A with cellopentaose  206  Figure 6.10  Possible orientations of TEMPO-labelled cellotriose bound to C B D N I  209  Figure 6.11  Model of the cellopentaose-CBDNi complex  Figure 6.12  Orientations of cellotetraose showing the location of oxygen atoms  15  211 215  Xll  Abbreviations  BMCC  bacterial microcrystalline cellulose.  CBD  cellulose-binding domain.  CBDCBHI  cellulose-binding domain from Trichoderma reesei cellobiohydrolase 1.  CBDcex  cellulose-binding domain from Cellulomonas fimi xylanase-glucanase Cex.  CBDcip  cellulose-binding domain from Clostridium thermocellum scaffoldin subunit Cip B.  CBDE4  cellulose-binding domain from Thermomonospora fusca endo/exocellulase E4.  CBDEGZ  cellulose-binding domain from Erwinia chrysanthemi cellulase EGZ.  CBDNI  N-terminal cellulose-binding domain from Cellulomonas fimi p-l-4-glucanase CenC.  CBDN2  cellulose-binding domain from Cellulomonas fimi (3-1,4-glucanase CenC following C B D N I in sequence.  CBDNIN2  the tandem cellulose-binding domains from Cellulomonas fimi [3-1,4-glucanase CenC.  CD  circular dichroism.  CT-HSQC  constant-time heteronuclear single quantum correlation.  DQF-COSY  double quantum filtered correlation spectroscopy.  DSS  2,2-dimethyl-2-silapentane-5-sulphonic acid, sodium salt.  DTNB  5,5'dithio-bis(2-nitrobenzoic acid).  DTT  dithiothreitol.  EDTA  ethylenediamine tetraacetate.  FITC  fluorescein isothiocyanate.  FTIR  fourier transform infra red.  HMQC  heteronuclear multiple quantum correlation.  HSQC  heteronuclear single quantum correlation.  INEPT  insensitive nuclei enhanced by polarization transfer  IPTG  isopropyl p-D-thiogalactopyranoside.  J  coupling constant.  NMR  nuclear magnetic resonance.  NOE  nuclear Overhauser effect.  NOESY  nuclear Overhauser effect spectroscopy.  PASA  phosphoric acid-swollen Avicel.  PASC  phosphoric acid-swollen cellulose.  PH*  the observed pH meter reading without correction for isotope effects.  ppm  parts per million.  rms  root mean square.  rmsd  root mean square deviation.  SDS  sodium dodecylsulphate.  sw  spectral width.  TEMPO  2,2,6,6-tetramethylpyrrolidine-l-oxyl.  TOCSY  total correlation spectroscopy.  xiv Acknowledgments  I would like to thank my supervisor, Lawrence Mcintosh, for all his great ideas and help during the last four and a half years. Many thanks also go to Peter Tomme who first showed me around a biochemistry lab, and provided invaluable assistance throughout my doctoral studies. I also acknowledge the the help of the other people I have collaborated with on this project; Tony Warren, Doug Kilburn, Jeff Kormos, Louise Creagh, Chip Haynes, Steve Withers and Lloyd McKenzie. Many thanks go to Lewis Kay for providing his great NMR expertise and most of the pulse sequences I used. I also thank past and present members of the Mcintosh lab for the help they have given me; especially Logan Donaldson for his computer expertise and willingness to rock early in the morning, Manish Joshi and Koman Joe for their help on the C B D N I project, and Emmanuel Brun for enlightening discussions about CBDs. For helping to make the past five years so enjoyable I thank Claire, and for encouragement through all of my years at university I thank my Mom. Most of all I would like to acknowledge my wife, Anne, for all the support and encouragement she has given me.  i  Chapter 1 Cellulose-Binding Domains: Structures and Binding Mechanisms  Introduction  Enzymes involved in the degradation of cellulose generally have a modular architecture with binding to the substrate mediated by one or more cellulose-binding domains (CBDs) (figure 1.1). Currently, over 180 different putative CBD sequences have been identified and classified into 13 families based on their sequence similarities (Tomme et ah, 1995; Dr. Peter Tomme, pers. comm.). Only a small fraction of these sequences have been shown experimentally to bind cellulose, but the binding ability of representatives from each of these families have been demonstrated. Although chemically homogeneous, cellulose is a structurally complex and heterogeneous substrate and poses a challenge for enzymes to hydrolyze. In nature, cellulose exists in several crystalline forms, as well as in disordered or "amorphous" states. Paralleling this heterogeneity, CBDs from various families exhibit a range of specificities and affinities toward the various allomorphs of cellulose. Most notably, C B D N I binds only amorphous cellulose (and soluble (31,4-glucans, including cellooligosaccharides), whereas the other CBDs of known structure bind preferentially to crystalline cellulose. Calorimetric studies reveal that the binding of C B D N I to amorphous cellulose is enthalpically-driven (Tomme et al, 1996a), whereas the association of CBDcex with crystalline cellulose is favored entropically (Creagh et al, 1996). Cellulose is composed of repeating units of (3-1,4-linked glucose and is the primary structural component of plant cell walls. It is the most abundant natural polymer on earth, with an estimated  Chapter 1-CBDs: Structures and Binding Mechanisms  CenA CenB  1EM0E  CenC CenD  w^//Amm  CbhA CbhB Cex XylD catalytic domain | linker  •  • • • • • •  • v_---_--._---_. • • • • • •  CBD II Y///k Fn3 module  CBD III  CBD IV other  Figure 1.1. Cellulases produced by Cellulomonas fimi. Each cellulase has a multi-domain architecture. Roman numeral after the CBDs indicates which family that it to belongs to.  3  Chapter 1-CBDs: Structures and Binding Mechanisms  4 x IO kg biosynthesized and degraded annually (Coughlan, 1990). As a result of the abundance 12  and low cost of cellulose, a number of potential biotechnology applications for CBDs have been devised. CBDs have been used as affinity tags for protein purification in batch, on a column or in an aqueous two phase system (Ong et al, 1989; Tomme et al, 1996b). CBD fusion proteins can also be used to immobilize proteins on a cellulose matrix with the activity of the heterogeneous protein often comparable to that of native protein. The binding of CBDs to cellulose also represents an excellent model system to study protein carbohydrate interactions. A number of structures of lectins, both free and complexed with oligosaccharide are known (Rini, 1995). But, in general, these only interact with a mono or disaccharide. In contrast CBDs interact with multiple sugar subunits of the oligosaccharide and thus possess a much larger binding site. The CBD my research has focused on is the family IV N-terminal CBD from fimi  Cellulomonas  (3-1,4-glucanase CenC (CBDNI) (figure 1.1). CenC has a family 9 catalytic domain, and is a  semi-processive enzyme with both endo- and exoglucanase activity. This enzyme works with inversion of configuration at the anomeric carbon and is most active on soluble sugars, though it also degrades Avicel and BMCC (Tomme et al, 1996c). A previous study has shown that C B D N I has the unique ability to bind only phosphoric acid-swollen cellulose, and not crystalline cellulose (Coutinho et al, 1992). The goal of my studies was to determine the three-dimensional structure of C B D N I to explain its binding selectivity. When I started this research in 1993, the only published structure of a cellulose-binding domain was the 36 residue family I CBD from the fungus (CBDCBHI)  Trichoderma reesei cellobiohydrolase  I  (Kraulis et al, 1989) (figure 1.2). A preliminary structure of the family II CBD from  the Cellulomonas  fimi  xylanase/(3-l,4-glucanase Cex (CBDcex) (Xu etal, 1995) was also known  (figure 1.2). Over the course of this thesis, structures of three other CBDs have been determined. These are the family III CBDs from Clostridium (Tormo  et al,  1996) and Thermomonospora  thermocellum  fusca  scaffoldin subunit Cip B (CBDrjip)  endo/exocellulase E4. (CBDE4) (Sakon  1997), and the family V CBD from plant pathogenic bacterium Erwinia  chrysanthemi  et al,  cellulase  Chapter 1-CBDs: Structures and Binding  Mechanisms  Figure 1.2. Ribbon diagrams offiveCBDs with known structure. Shown in green are the aromatic rings and hydrophobic residues on the binding faces of the CBDs. Selected hydrophilic residues are in red. Figure drawn using Molscript (Kraulis, 1991) and Raster3D (Merritt & Murphy, 1994).  4  Chapter 1-CBDs: Structures and Binding Mechanisms  EGZ  5  (CBDEGZ) (Brun et al, 1997) (figure 1.2). Complementing these structural studies are  several detailed biochemical and functional studies, including the thermodynamics of CBDcellulose interactions (Creagh et al, 1996; Tomme et al., 1996a). A summary of the biochemical and structural data for these CBDs is given in table 1.1. The aim of this chapter is to introduce to the reader the field of CBDs. This will serve as a background to my work, presented in chapters 2 to 6. Table 1.1. Biochemical data for CBDs of known structure. CBDcip CBDcBHI CBDcex Family I II III # Residues 36 110 155 Size (kDa) 3.7 18.2 11.1 Calculated pi 6.75 8.26 5.05 I Location* C C 1 # Disulfides 2 0 Fold wedge p-barrel P-jelly roll Binding Site flat face flat face flat face / cleft Ca binding no no yes cellulose Parental cellobiohydrolase xylanase / (3integrating Enzyme 1,4-glucanase protein (nonhydrolytic) 2  2+  a  CBD i IV 152 15.4 3.24 N 1 [3-jelly roll cleft yes (3-1,4glucanase N  CBDEGZ  V 62 6.5 5.21 C 1 boot shaped flat face no (3-1,4glucanase  C: carboxy terminus, N: amino terminus, I: internal.  CBDs that bind crystalline cellulose  Structures  The three-dimensional structures of representative members of four families of CBDs that bind crystalline cellulose have been determined by NMR or x-ray crystallography (figure 1.2). Each of these CBDs shares a common feature of being composed entirely of anti-parallel (3-strands connected by turns and loops. The three [3-strands of C B D C B H I fold into a wedge shape, while  Chapter  1-CBDs:  Structures  and Binding  6  Mechanisms  CBDcex has a nine-stranded p-barrel fold. CBDrjip is a jelly-roll P-sandwich, and C B D E G Z is a boot-shaped molecule composed of a large loop region and five p-strands that form two p-sheets. The dominant structural feature of these CBDs is the presence of three solvent-exposed aromatic rings aligned along a flat face of each molecule (figure 1.2, 1.3). In C B D C B H L these are three tyrosine side-chains (Tyr5, Tyr31 and Tyr32), while CBDcex has three tryptophan residues solvent-exposed (Trp 17, Trp54 and Trp72). For CBDcip, these residues are two tryptophans and a histidine (Trpll8, His57 and Tyr67), and in C B D E G Z they are two tryptophans and a tyrosine (Trp 18, Trp43 and Tyr44). A common theme in protein-carbohydrate interactions is the stacking of aromatic side-chains on sugar rings (Quiocho, 1986, 1989; Vyas, 1991). Although all four structures were determined in the absence of cellulose or cellulose-derivatives, evidence from sitedirected mutagenesis and cellohexaose-binding studies strongly indicates that these aromatic residues, and others nearby, mediate the association of CBDs with crystalline cellulose.  Identification  (i)  of binding  CBDCBHI:  face  Extensive studies of the effects of amino acid substitutions on the binding of  the family I C B D C B H I to cellulose have been reported (Reinikainen  et al,  1992, 1995; Linder  et  al, 1995a, b). These studies show that the three solvent-exposed tyrosines (Tyr5, Tyr31, Tyr32), and a polar residue (Gln34) are important for retaining binding affinity. Mutating these residues to alanines greatly reduces binding to bacterial micro-crystalline cellulose (BMCC). The side-chains of these four residues form a strip across one face of C B D C B H I (Figure 1.3). In contrast, the mutation of Asn29 to alanine reduced BMCC binding slightly (Linder et al, 1995a). This residue is found on the binding face, but off to the side of the strip formed by the tyrosines and Gln34 (figure 1.3). Changing Prol6 to Arg, present on the "rough" face of the protein, opposite to the tyrosine-containing face, has only a small effect on binding (Reinikainen et al, 1992). The structures of all three tyrosine-to-alanine mutant C B D C B H I proteins have been determined by NMR spectroscopy (Mattinen et al, 1997a). The three-dimensional structures of the Y31A and Y32A variants are very similar to that of wild type. These are also well defined, having a low rms  Chapter 1-CBDs: Structures and Binding  7  Mechanisms  GRASP (Nicholls et al., 1991) surface representation looking down onto the binding face or cleft of each CBD. The CBDs that bind crystalline cellulose, CBD , CBD , CBD and CBD , all have flat binding faces with three exposed aromatic residues. CBD does not bind crystalline cellulose and has a binding groove. Shown in yellow are the aromatic and hydrophobic residues (W, L, V, A, Y, F, H, I, P, M, C) with the hydrophilic residues in red (N, D, E, Q, R, S, T, K). Selected residues are labelled to orient the reader. The different CBDs are not drawn to scale relative to each other in this figure. Figure 1.3.  CBHI  cip  EGZ  Cex  N]  Chapter 1-CBDs: Structures and Binding Mechanisms  8  deviation amongst the ensemble of final structures. This indicates the elimination of binding by these mutations results directly from the removal of the tyrosine, and not indirectly by causing structural changes in the protein. In contrast, the structure of the Y5A variant is significantly altered from that of wild type. The ensemble of structures for this mutant have a higher rms deviation, especially in the N-terminal region near position 5. Consequently, it is difficult to separate the structural and functional roles of Tyr5 in C B D C B H L except to state that its presence is required for cellulose binding. Chemical shift perturbations of several other mutant CBDcBHI proteins relative to the wild-type have been reported (Linder et al, 1995a). The variants Q34A and N29A show little change in their chemical shifts, indicating that the structural effects of these mutations are minimal. In the case of the P16R variant, changes in chemical shift were significant, suggestive of conformational changes in this variant. However, any such structural changes have no effect on binding. The importance of Tyr31 in ligand binding was also shown when the association of various T. reesei CBDs with cellohexaose was studied (Mattinen et al, 1997b). The NMR line widths of  cellohexaose increased when bound weakly by the CBDs from cellobiohydrolase I and II and endoglucanase I. For the Y31A variant of C B D C B H I , no increase in the NMR line width was observed, indicating that this protein does not appreciably bind cellohexaose. (ii) CBDcex The involvement of the three solvent-exposed tryptophans of CBDcex in ligand :  binding is supported by chemical shift data, a chemical modification study, and by data for mutant forms of a related family II CBD. The ' H and N e l  15  8 1  chemical shifts of the indole rings of  Trp54 and Trp72 are perturbed upon the addition of cellohexaose to CBDcex (Xu et al, 1996). Both of these residues are solvent-exposed (figure 1.3). Additionally, when the equivalent of two of the three solvent-exposed tryptophans of CBDcex are oxidized by N-bromosuccinimide (NBS), binding to BMCC is essentially eliminated (Bray et al, 1996). Finally, the mutation of Trp68 in the family II CBD from C. fimi endoglucanase A (CBDcenA) to alanine reduces binding by 30% (Din et al, 1994). From sequence alignments (Tomme et al, 1995), this residue corresponds to Trp72 ofCBDcex-  Chapter 1-CBDs: Structures and Binding Mechanisms  Together, evidence for  CBDCBHI  9  and CBDcex indicates that only residues on the face  containing the solvent-exposed aromatics are involved in cellulose-binding. Although experimental evidence is not available for CBDcip and CBDEGZ> it can be safely assumed that similar results would be obtained.  Mechanism for binding crystalline cellulose  Thermodynamic data defining the mechanism of binding of CBDcex to insoluble BMCC were reported by Creagh et al. (1996). The dominant thermodynamic driving force of binding is a net increase in entropy (AS°). This is accompanied by a small decrease in enthalpy (AH°), and a large negative AC . The increase in entropy, as well as the negative AC , indicates a dehydration or p  p  hydrophobic effect taking place during cellulose binding. A structural explanation for the increase in entropy is the exclusion of water when the three indole rings on the binding surface of CBDcex contact cellulose. This results in a smaller net number of ordered water molecules present in the system, as there is less net polar surface area present to hydrate. Some hydrogen bond formation between the CBD and cellulose can be envisioned to account for the small negative enthalpy of binding. An increase in entropy as the driving force for complex formation is different from the interactions of most other carbohydratebinding proteins (Vyas, 1991). Though it remains to be experimentally proven, the presence of three exposed aromatic rings in the family I, III and V CBDs indicates that they likely share a similar binding mechanism with CBDcexIt is expected that a tryptophan side-chain, being larger than a tyrosine, would cover more surface area of cellulose, and thereby displace more water molecules upon binding. This in turn would result in a larger increase in entropy and a higher affinity for cellulose. In support of this argument, the family I CBD from Endoglucanase 1 of T.  reesei (CBDEGI)  has a significantly  higher affinity for crystalline (tunicate) cellulose than does CBDCBHI- A major difference between these two CBDs is the presence of a tryptophan residue at position 5 in C B D E G L  a s  compared to a  tyrosine in C B D C B H I - The Y5W variant of C B D C B H I has substantially higher binding affinity for  Chapter 1-CBDs: Structures and Binding Mechanisms  10  crystalline cellulose over wild-type, although not equivalent to C B D E G I (Linder et al, 1995b). It would be possible to test further the effect of creating a larger surface area of dehydration by also looking at the binding of a Y31W variant of C B D C B H I - A double C B D C B H I variant, Y5W and Y31W, could also be created to see if an additive effect on binding affinity is observed. In addition to the tryptophans, there are a number of hydrophilic residues on the binding face of CBDcex which could be involved in the hydrogen bond formation to cellulose, as suggested by the thermodynamic data. These include Asnl5, Gln52, Gln83 and Asn87 (figure 1.3). The hydrophilic character of these residues tends to be conserved among family II CBDs (Tomme et al, 1995). Other CBDs also contain hydrophilic residues that punctuate the hydrophobic character of the binding face (figure 1.3). In some cases, these residues are thought to form intramolecular salt bridges, for example R112 and D56 in CBDcip and K16 and D17 in C B D E G Z - These hydrophilic residues could also stabilise the orientation of the aromatic rings by forming hydrogen bonds to atoms on the aromatic ring. None of the CBD structures determined at present contain a phenylalanine among the exposed aromatic rings on the binding face. From the sequence alignments (Tomme et al, 1995) of families I, II, III and V, very few phenylalanines align with residues corresponding to their solvent-exposed aromatics. This suggests that an electronegative atom on the aromatic ring is important, enabling hydrogen bond formation either to the cellulose or to another residue on the binding face. Alternatively, a phenylalanine may have a detrimental effect on the solubility of the protein due to its hydrophobic character.  Structural considerations of CBDs binding to crystalline cellulose  Cellulose is an insoluble polymer of P-l,4-linked glucopyranosyl units. Through interchain hydrogen bonds, cellulose strands of 100 to over 10 000 glucose units form fibers. These in turn wrap into bundles to form the main structural component of plant cell walls. Highly ordered regions of cellulose are referred to as crystalline cellulose, whereas regions of the cellulose crystal lattice that have been disrupted are termed amorphous cellulose. Cellulose that CBDs encounter in  Chapter 1-CBDs: Structures and Binding Mechanisms  11  nature exist in a spectrum of states ranging from highly crystalline to amorphous, and also contain other compounds, such as lignin and xylan. A number of possible allomorphs of crystalline cellulose have been identified. In all forms, alternate glucose residues are rotated 180° relative to each other, making cellobiose the repeating subunit.  The allomorphs differ primarily in the arrangement of hydrogen bonds between  individual cellulose chains and sheets. It is proposed that neighboring chains can be oriented parallel (cellulose I) or antiparallel (cellulose II) to each other. In all forms, cellulose chains are hydrogen bonded together to form sheets that are held together by van der Waals interactions. The sheets of cellulose are staggered, resulting in a diamond arrangement with only the top layer, the (0,2,0) face, having its pyranose rings completely exposed (figure 1.4). A recent crystal structure of (3-D-cellotetraose has been solved (Gessler et al, 1994; Gessler et al, 1995), providing an atomic resolution model of the cellulose II polymer.  Figure 1.4. Schematic view of the cross section of the structure of crystalline cellulose I according to Gardner and Blackwell (1974). This cross section shows the strands of cellulose coming out of the page.  Chapter 1-CBDs: Structures and Binding Mechanisms  12  A commonly used form of cellulose is bacterial microcrystalline cellulose (BMCC). This is a highly crystalline form of cellulose I prepared from Acetobacter xylinum (Gilkes et al, 1992). Cellulose that has significant amorphous character can be generated by treatment with phosphoric acid. This is referred to as phosphoric acid-swollen cellulose (PASC), or phosphoric acid-swollen avicel (PASA), depending on the starting material. A limitation of this compound is the inability to measure directly the amount of residual crystalline regions present. Without this measure, caution should be used in interpreting binding studies using PASC, as the .CBD may interact selectively with residual crystalline regions and not in the disordered, or amorphous regions. Other types of cellulose such as Avicel, filter paper and cotton are used for the purification of CBDs, but are not very useful for quantitative binding studies due to their varied composition. It remains to be established where CBDs bind on the cellulose crystal. The dehydration mechanism of binding, suggested by thermodynamic data, only indicates that water is displaced upon binding. Reinikainen et al. (1995) and Tormo et al. (1996) both propose that CBDs bind to the (0,2,0) face of cellulose (figure 1.4). The rings of the exposed aromatic side-chains of the CBDs are spaced to directly overlap the pyranose rings of a single cellulose chain. In a perfect cellulose crystal, this face is only one chain wide. Since the binding faces of the CBDs are not concave, this implies a CBD molecule would only make contact with one strand of cellulose and only residues in line with the aromatic rings would be able to interact with this strand. Residues off to the side of the binding face would likely not be able to make contact with cellulose without significant conformational change on binding. In the binding model outlined in Tormo et al. (1996), CBDs are proposed to interact with three adjacent cellulose chains. Hydrophilic residues off to the side of the strip of aromatic residues act as anchor points for hydrogen bonding to cellulose. This idea of anchor points has yet to be supported by studies where the proposed anchor residues are mutated and the effect on binding measured. As mentioned earlier, the mutation N29A in C B D C B H I s a small, but significant, na  effect on the BMCC binding ability of C B D C B H I (Linder et al, 1995a). Asn29 is one of the proposed anchor points for C B D C B H I (Tormo et al, 1996).  Chapter 1-CBDs: Structures and Binding Mechanisms  13  It is consistent with a binding mechanism involving the displacement of water, for CBDs to bind the cellulose crystal in regions other than where a flat planar array of cellulose strands is present. Binding could occur on any of the cellulose crystal faces shown in figure 1.4. Here the CBD could be aligned along the cellulose chain or perpendicular to it. As shown in figure 1.5 the binding faces of the CBDs are not all completely flat.  CBDCBHI.  CBDcex d C B D E G Z all have an  binding faces where hydrophilic residues lie above and or below the plane formed by the linear arrangement of aromatic rings. This slope of the binding site could enable the CBD to simultaneously interact with cellulose chains on multiple sheets. The aromatic rings on the binding face of the CBD could still align with the partially exposed pyranose rings of cellulose. Although smallest in molecular weight, C B D C B H I has a binding face just as wide, if not wider, than those of CBDcex or C B D E G Z - It is clear from figure 1.5 that the binding face of CBDcip is wider and flatter than the other CBDs. The width of the binding face of CBDcip may hinder its binding to regions of the cellulose crystal that are not also fairly flat. Experiments on CBDcex labelled with fluorescein isothiocyanate (FITC) show that at low protein concentrations FITC-CBDcex binds uniformly onto BMCC (Creagh et al, 1996). At saturating concentrations FITC-CBDcex does concentrate in one area of BMCC. This is attributed to a region of high surface area, not a region of preferential binding. It has also been shown that FITC-CBDcex rapidly diffuses across the cellulose surface without becoming unbound from the cellulose crystal (Jervis etal, 1997).  CBDs that do not bind crystalline cellulose  Structure  The family IV CBDs C B D N I ,  CBDN2,  and the native tandem  CBDNIN2  are unique among  CBDs in that they bind only to PASC and soluble (3-1,4 linked glucans, and not to crystalline cellulose (chapter 2; Coutinho et al, 1992; Tomme et al, 1996a, Johnson et al, 1996a). The  C h a p t e r  1 - C B D s :  S t r u c t u r e s  a n d  B i n d i n g  M e c h a n i s m s  Figure 1.5. Head-on view of the putative binding faces for the four CBDs that bind crystalline cellulose. These views are rotated 90° compared to those shown in figure 1.2. Shown are ribbon diagrams of the backbone with aromatic and hydrophobic residues on the putative binding faces shown in green, hydrophilic residues are shown in red. Superimposed on this is the solvent accessible surface area produced by GRASP.  14  15  Chapter 1-CBDs: Structures and Binding Mechanisms  structure of C B D N I is composed of two P-sheets each containing 5 p-strands. These two sheets then form a jelly-roll sandwich structure (figure 1.2) (chapter 3; Johnson et al, 1996b). In contrast with the CBDs that bind crystalline cellulose, the binding site of C B D N I is a cleft, or groove, that runs along one face of the molecule. The structure of C B D N I was determined in the presence of cellotetraose, but this ligand was not included in the structure calculations. Nevertheless, binding is known to occur in this cleft based on the perturbation of chemical shifts upon the addition of cellooligosaccharides (chapter 2; Johnson et al, 1996a) and the detection of NOEs between the protein and unassigned protons on the sugar (chapter 3; Johnson et al, 1996b). This presence of a groove readily explains the binding selectivity of this CBD. The cleft prevents the protein from interacting with the flat, rigid surface of crystalline cellulose. Instead, single cellulose chains, as might be encountered in the amorphous regions of cellulose, can fit into the binding cleft. A preliminary structure of C B D N 2 shows this protein has a very strong structural similarity with  CBDNI,  an  d that its binding site is also a cleft (Dr. E. Brun, personal  communication). The structure of C B D N I resembles that of C B D c i p , as both proteins have a jelly-roll fold. The structure of C B D c i p also has a groove running along one P-sheet face of the molecule. This region of the protein is not thought to be the site where binding to crystalline cellulose occurs. Instead it has been suggested that this tyrosine rich cleft might be involved in protein-protein interaction between C B D c i p and another module of the cellulosome complex (Tormo et al, 1996). This face could also be involved in protein-carbohydrate interactions as either a second binding site for cellulose, or for interaction with a glycosylated region of another protein (Tormo et al, 1996). CBDNI  and C B D c i p also share the ability to bind calcium ions, although not at similar  locations in the structure. In both proteins the calcium-binding site is located away from the cellulose-binding sites. Calcium increases the stability of C B D N I , but does not alter its ability to bind cellooligosaccharides (chapter 4). A calorimetric study shows that C B D N I is a marginally stable protein, having a maximum stability of 33 kj mol' at 1 °C and pH 6.1 (Creagh et al, 1997). 1  CBDNI  can be reversibly  Chapter 1-CBDs: Structures and Binding Mechanisms  16  denatured by either thermal or chemical means. Such reversibility may be common among CBDs, many of which are purified by binding to Avicel, with a denaturing concentration of guanidinium hydrochloride used to elute the protein. Another structural feature of C B D N I is that it contains a single disulphide bond. Reduction of this disulphide results in complete unfolding of the protein (chapter 2; Creagh et al, 1997). As shown in table 1.1 disulphides are commonly found in CBDs. However the reduction of the disulphide bond in C B D E G Z does not cause unfolding of the protein, and reduced C B D E G Z retains its cellulose-binding ability (Dr. E. Brun, personal communication).  Binding specificity and thermodynamics of CBD'NI  The binding of C B D N I to soluble cellooligosaccharides provides an opportunity to characterize the interaction of this domain with well defined model ligands. In parallel with the structural studies of C B D N J , a detailed calorimetric analysis of the binding of C B D N I to soluble ligands and PASC was conducted (Tomme et al, 1996a).  CBDNI  was found to bind long polymeric  substrates, such as hydroxyethyl cellulose and barley- and oat-p-glucan, with equal affinity to that of cellohexaose and cellopentaose. For shorter substrates, affinity decreases with chain length. This maximum binding affinity for cellopentaose correlates well with the binding cleft being approximately the length of five glucose rings. Oligosaccharide binding by C B D N I is accompanied by a favorable enthalpic change (AH°), compensated in part by a decrease in entropy (AS°). This implies that a predominance of polar interactions, such as hydrogen bonding and van der Waals interactions, provide the primary driving force for binding. This is similar to other carbohydrate-binding proteins (Quiocho, 1986, 1989), and in sharp contrast with that of CBDcex where an increase in entropy is the thermodynamic driving force for the binding of BMCC. This proposed binding mechanism is supported by the structural features of the binding cleft of CBDNI  (Figures 1.2, 1.3). The weak van der Waals interactions take place between the strip of  hydrophobic residues (Vall7, Tyrl9, Val48, Leu77, Tyr85, Alal26) that span all five strands of the binding face, and the relatively non-polar pyranose rings of the glucose sugars. On both sides  Chapter 1-CBDs: Structures and Binding Mechanisms  17  of this hydrophobic strip are residues (Asn50, Glnl24, Glnl28, Asn81, Arg75, Asp90) which can form hydrogen bonds to the equatorial hydroxyl groups on the sugars.  Summary  Two distinct structural classes of CBDs exist. CBDs that bind crystalline cellulose have a binding face that contains three solvent-exposed aromatic rings. The proposed binding mechanism for these CBDs involves a displacement of the ordered water molecules from the surfaces of the protein and cellulose, thereby increasing the entropy of the system. For the family IV CBDs that do not bind crystaline cellulose, their structures are characterized by the presence of a binding groove. Single sugar chains fit into this groove and binding is mainly driven by hydrogen bond formation. This thesis describes my studies on the family IV cellulose-binding domain C B D N I • In chapter 2, the means of expressing and purifying C B D N I is outlined, and binding to cellooligosaccharides characterized. Chapter 3 details the structure of C B D N I , while chapter 4 examines the calcium binding properties of this protein. The dynamics of apo, calcium-bound and cellooligosaccharidebound  CBDNI,  as studied by  1  5  N  and  2  H  relaxation, are presented in chapter 5. Chapter 6  examines how C B D N I binds oligosaccharides.  18  Chapter 2 Cellooligosaccharide Binding by  C  B  D  N  I  Abstract  The N-terminal cellulose-binding domain (CBDNI) from Cellulomonas fimi (3-1-4glucanase CenC binds amorphous but not crystalline cellulose. To investigate the structural and thermodynamic bases of cellulose binding, NMR spectroscopy was used in parallel with calorimetry (Tomme et ai, 1996a) to characterize the interaction of soluble cellooligosaccharides with C B D N I - Association constants, determined from the dependence of the amide H and N l  15  chemical shifts of C B D N I upon added sugar, increase from 180 ± 60 M~l for cellotriose to 4200 ± 720 M"l for cellotetraose, 34000 ± 7600 M for cellopentaose, and an estimate of 50000 M' for _1  1  cellohexaose. This implies that the C B D N I cellulose-binding site spans approximately five glucosyl units. Based on the observed patterns of amide chemical shift changes, the cellooligosaccharides bind along a five-stranded (3-sheet that forms a concave face of the jelly-roll (3-sandwich structure of C B D N I - This (3-sheet contains a strip of hydrophobic side-chains flanked on both sides by polar residues. NMR measurements demonstrate that tyrosine, but not tryptophan, side-chains may be involved in oligosaccharide binding. These results lead to a model in which  CBDNI  interacts with soluble cellooligosaccharides, and by inference, single  polysaccharide chains in regions of amorphous cellulose, primarily through hydrogen bonding to the equatorial hydroxyl groups of the pyranose rings, van der Waals stacking of the sugar rings against the apolar side-chains may augment binding.  CBDNI  stands in marked contrast to  previously characterized CBDs that absorb to crystalline cellulose via a flat binding surface dominated by exposed aromatic rings.  Chapter 2-Cellooligosaccharide Binding by CBD^j  19  Introduction  Background  An absolute requirement for an NMR oriented study of protein structure is the availability of milligram quantities of material. For a protein the size of C B D N I , 152 residues or 15 kDa, the production of isotopically N and C labelled protein is also essential. This chapter outlines the 15  L 3  method for expressing and purifying samples of C B D N I that I used throughout my thesis work. Included in this chapter is the initial biophysical characterization of C B D N I performed using circular dichroism and fourier transform infrared spectroscopy. Work related to the stability of CBDNI  to thermal and chemical denaturation is also presented. Very early in this study I found that C B D N I binds short chains of p-l,4-linked glucose  monomers called cellooligosaccharides. These sugars, which are chemically equivalent to cellulose, provide a well defined substrate for investigating sugar-binding by C B D N i • The studies outlined in this chapter were performed before the resonance assignments of C B D N I were known. However, in this chapter, the results are interpreted with the knowledge of both the assignments and structure, which will be presented in chapter 3. Most of the work discussed in this chapter has been published (Johnson et ai, 1996a). With the exception of the cloning and difference ultraviolet spectroscopy measurements, I performed all the work that led to this publication.  Ligand binding studied by NMR  Throughout this thesis, NMR spectroscopy is used to measure association binding constants (K ). In this chapter, the binding of a series of cellooligosaccharides to C B D N I is a  investigated. In chapter 4 the association constants of calcium with CBDNI  CBDNI  and calcium-loaded  with cellopentaose are determined. In chapters 5 and 6, results of C B D N I binding to spin-  labelled cellooligosaccharides and the binding of C B D N I variants with cellooligosaccharides is presented. In all these studies, the same equations and strategy for determining equilibrium association binding constants was followed, and is described here.  Chapter 2-Cellooligosaccharide Binding by CBD^j  20  For binding events monitored by NMR spectral changes that occur in the fast exchange limit of the NMR timescale, the fraction of protein in the bound form at each point (i) in the titration (fbi) is calculated using the observed 'H or N chemical shift (5j) compared to that of the free (8f) 15  and fully bound forms (8b) by equation (2.1):  '-"f^sT  < 2 I )  8f is determined from the spectrum of the protein without added ligand and 8b obtained from the fitting routine. For a single binding site, the association constant K for the binding of free protein (Pf) and free ligand (Lf) to yield the bound complex (PL) is given by equation (2.2): a  K  a  = ^ L _  (2.2)  [P ] [L ] f  f  By definition, for each titration point, the fraction of protein in the bound form is given by equation (2.3), the total protein concentration [Pt] by equation (2.4), and the total ligand concentration [Lt] by equation (2.5). f  b  =  ^  L  (  2  . ) 3  [P ] + [PL] f  [P] = [P ] + [PL]  (2.4)  [LJ = [L ] + [PL]  (2.5)  t  f  f  Rearranging (2.4) and (2-5) using (2.2), followed by substitution into (2.3) results in equation (2.6). f  _ KatLJi -K f [P l l+ KJLJ.-KJJPJ a  bl  bi  t  6  Equation (2.6) is solved for fbi using the real root of the quadratic equation, allowing equation (2.1) to be recast in terms of the experimental parameters [P]j, [L]j, 8f, and 8j and the variables t  t  K and 8b- The latter two terms are determined by the fitting routine. An initial estimate for 8b is a  supplied from the 'H or N chemical shift at the final titration point, and the concentration of 15  protein at each point (i) is corrected for dilution resulting from the addition of ligand. For every study presented in this thesis the programme PLOTDATA (TRIUMF, UBC) was used for the nonlinear least squared fitting of the chemical shift data versus total ligand concentration.  21  Chapter 2-Cellooligosaccharide Binding by CBDNI  In all the quantitative studies of ligand association by C B D N I , binding is monitored by changes in the two-dimensional 'H- N 15  HSQC  spectrum of uniformly N-labelled ,5  CBDNI.  However, the method presented here can be applied to any measurable property of the system that changes in an incremental manner with binding. For binding processes studied by NMR it is necessary that association between the ligand and protein is in the fast exchange limit on the NMR chemical shift timescale to apply this formalism. For processes that are in the slow exchange limit, fbi can be estimated from peak intensities of the free and bound forms of the protein.  Experimental methods  Expression and Purification of  CBDNI  The gene encoding the 152 residue C B D N I from Cellulomonas fimi (3-1,4-glucanase CenC was obtained from Dr. Peter Tomme (Department of Microbiology, UBC). To achieve efficient overexpression in minimal media, Dr. Tomme recloned the gene encoding C B D N I from the pTZ vector used in previous studies (Coutinho et al, 1992) into the high expression vector pTug (figure 2.1) (Graham et al, 1995). This expression vector encodes a gene for kanamycin resistance. The gene encoding C B D N I  is fused to the Cex leader peptide to transport the  expressed protein into the periplasm of E. coli. The gene fragment encoding  CBDNI  used  throughout this thesis is termed pTugNln. The vector pTugNln was expressed in E. coli JM101 cells (Yanish-Perron et al, 1983). Unlabelled protein was produced in liquid tryptone / yeast extract / phosphate medium (TYP) (Sambrook, 1989). Biosynthetically N labeled protein was prepared using M9 media (Miller, l5  1972) containing lg/L 99% NH Cl (Cambridge Isotopes) and 1 g/L of 99% N labeled Isogro l5  15  4  (Isotec Inc.) as the sole sources of nitrogen. The Isogro (algal extract) supplement is necessary for efficient growth of E. coli JM101 and protein production in M9 media. The C B D N I samples with selectively deuterated aromatic rings were obtained from a synthetic medium containing 100 mg/L  Chapter 2-Cellooligosaccharide Binding by CBD/^j  22  of L-8i,e2,Cl,2>'l2-[ H5]tryptophan and 100 mg/L of either L-di^ei^C^Hslphetiylalanine or r  2  8i,2,£l,2-[ H4]tyrosine (Cambridge Isotope Laboratories and Isotec Inc.) (Mcintosh et al., 1990; 2  Mcintosh & Dahlquist 1990). One litre bacterial cultures were grown in 1.8 L Fernbach flasks at 30 ° C to an A600 of 0.6, induced with 0.5 mM IPTG, and incubated with shaking for approximately 24 hours. The leader peptide, which is cleaved upon translocation, targets the expressed C B D N I to the periplasm of E. CBDNI  coli.  However, over this extended incubation period,  leaks out into the culture supernatant (Ong et al, 1993). To increase yield when preparing labelled protein, the cell pellet was also subjected to  osmotic shock by resuspending the cells in 30 mM Tris-HCl (pH 8.0) buffer containing 20% sucrose. After 15 min, the cells were centrifuged, resuspended in 5 mM magnesium sulphate at 4  23  Chapter 2-Cellooligosaccharide Binding by CBDNI  °C, stored on ice for a further 15 min, and finally repelleted. The supernatants from this procedure were combined with the culture supernatant for purification. Figure 2.2 is an SDS gel showing the expression of C B D N I , and also the presence of C B D N I in the osmotic shock fractions. CBDNI  is initially purified based on its affinity to a commercial type of cellulose called  Avicel. Avicel is heterogeneous, containing regions of both crystalline and amorphous cellulose. To each litre of culture supernatant 35 g of dry Avicel PH-101 (Fluka Chemika) is added, and the resulting slurry is adjusted to 1 M sodium chloride and 50 mM potassium phosphate at pH 7. After standing at 4 °C for 4 hours, the Avicel is collected by vacuum filtration on a Whatman GF/A glass filter and washed with 250 mL of 1 M sodium chloride, 50 mM potassium phosphate buffer at pH 7.  CBDNI  is eluted from Avicel with 250 mL of distilled water. To increase the recovery of  labelled protein, the initial filtrate and the salt wash solution were recombined with the Avicel and left overnight at 4 °C. The above procedure was repeated and the two protein fractions combined. Figure 2.3 shows fractions from the first purification procedure on Avicel. This figure shows that some protein remains in the flow through and salt wash, and not all the protein is eluted from Avicel, the amount lost here is reduced with a second binding to Avicel. It is also seen that C B D N I in the water elution fraction is very pure at this stage. CBDNI  is purified further by anion exchange FPLC at pH 5 and then pH 7 using  Sepharose (Pharmacia). In both cases, C B D N I  wa  Q  s eluted with a gradient of 0-1 M sodium  chloride in 50 mM potassium phosphate buffer. Often C B D N I eluted from ion exchange columns in two highly overlapped, but distinct, peaks. Electrospray mass spectroscopy showed that both peaks have the same molecular mass. Also, both peaks bind cellooligosaccharides with equal affinity. I am unable to explain the cause of this difference in mobility on the ion exchange column. Finally, the purified C B D N I was de-salted, exchanged into sample buffer, and concentrated by ultrafiltration using non-cellulose membranes (Filtron Technology Corp., Northborough, MA.). The yields of the unlabelled, N labelled, and selectively deuterated C B D N I samples were 15  approximately 80, 25, and 30 mg/Lof culture supernatant, respectively.  Chapter 2-Cellooligosaccharide Binding by CBD^]  / ^ ^  f ^  <f c  ^  c  > ^  ^  c  #  f ^ ^  ^ f  *  **  f ^  °* $  ^  j  29.0  18.4  CBD N l 14.3 6.2  Plitofe,  Figure 2.2. SDS P A G E gel showing the overexpression of C B D . Most of the protein leaks out into the culture supernatant. Protein in the periplasm can be recovered by osmotic shock (sucrose and M g S Q fractions). N 1  4  25  Chapter 2-Cellooligosaccharide Binding by CBDj^i  Figure 2.3.  SDS P A G E gel showing the purification of C B D  N )  on Avicel.  26  Chapter 2-Cellooligosaccharide Binding by CBDNI Characterization of CBDN  J  Protein concentrations were measured using absorption spectroscopy.  The molar  absorptivity £280 of C B D N I was determined to be 21370 M" cm (or 1.39 mL mg" cm ) using 1  -1  1  -1  the method of Edelhoch (1967), as reviewed by Gill & von Hippel (1989) and Pace (Pace et al, 1995). The small contribution of a disulfide was included in the calculations. This molar absorptivity value, which differs from that originally published by Coutinho et al. (1992), was confirmed by quantitative amino acid analysis. The average molar absorptivity determined from three amino acid analyses, each run in duplicate, agreed within 4% of that obtained using the Edelhoch method. The molecular mass of the unlabelled  CBDNI  is 15425.3 ± 0.9 Da as measured by  electrospray mass spectroscopy. This is in excellent agreement with the expected value of 15425.8 Da based on the sequence of C B D N I , after post-translational cleavage of the secretory leader peptide and corrected for the presence of a disulfide bond. The disulfide bond was identified by the lack of reactivity of C B D N I with 5,5'-dithiobis-(2-nitrobenzoic acid), and subsequently confirmed by NMR  L 3  CP  chemical shifts (Wishart & Sykes, 1994). Using Edman degradation,  the ten N-terminal residues of C B D N I expressed in E. coli were confirmed to be ASPIGEGTFD. This matches that of native CenC from C. fimi. Based on sedimentation equilibrium measurements, C B D N I (0.19 mM) in 50 mM sodium chloride, 50 mM potassium phosphate, 0.02% sodium azide, pH 5.9 at 20 °C is a monomeric protein with an apparent mass of 14130 Da (assuming a partial specific volume of 0.73 cm/gm; 3  Chervenka, 1969). No evidence of higher-order association was detected under the conditions employed for the ultracentrifugation analysis. This conclusion is supported qualitatively by the relatively narrow linewidths observed in the NMR spectra of C B D N I -  CD and FTIR spectroscopy  Circular dichroism (CD) spectroscopy was performed on 0.2 mg mL" samples of C B D N l 1  in a buffer of 50 mM sodium chloride, 50 mM potassium phosphate (pH* 5.9), 0.02% sodium  27  Chapter 2-Cellooligosaccharide Binding by CBDNI  azide in the absence and presence of a 20 fold molar amount of cellohexaose, or a 10 fold molar amount of barley-fj-glucan. Spectra were acquired with a Jasco J-730 CD spectropolarimeter using a 0.1 cm pathlength jacketed quartz cell. The spectra were solvent subtracted and processed using Jasco software. The midpoint temperature at which the tertiary structure of C B D N I unfolds was obtained by monitoring the ellipticity of C B D N I at 204 nm as the temperature increased from 35 °C to 60 °C at a rate of 1 °C/minute. Fourier transform infrared (FTIR) spectra of C B D N I using a ca. 1 mM sample of C B D N I in a buffer of 50 mM sodium chloride, 50 mM potassium phosphate (pH* 5.9), 0.02% sodium azide were acquired with a Perkin-Elmer Model 2000 instrument. 6 U.L of protein solution was needed to fill the calcium fluoride cell of 6 u\m pathlength used. A spectral range of 2200-1200 cm" 1  and a spectral resolution of 2 cm was used, 1000 scans were signal averaged for each -1  spectrum. Post processing of the spectra involved subtracting the spectral contributions of the buffer and the water vapor, then baseline flattening.  NMR Spectroscopy  The  CBDNI  samples were exchanged into 50 mM sodium chloride, 50 mM potassium  phosphate (pH* 5.9), 0.02% sodium azide, 10% D2O/90% H 0. NMR spectra were recorded on 2  a Varian Unity 500 MHz spectrometer equipped with a pulse field gradient triple-resonance probe. All spectra were collected at 35 °C and processed using FELIX v2.30 (Biosym Technologies). H ]  chemical shifts were referenced to an internal standard of DSS at 0.00 ppm, and N chemical 15  shifts to external 2.9 M NH C1 in 1 M HC1 at 24.93 ppm (Levy & Lichter, 1979). 15  4  'H- N HSQC spectra were recorded using the enhanced sensitivity pulsed field gradient I5  experiment of Kay et al. (1992). A selective water flip back pulse was incorporated to ensure minimum perturbation of the water magnetization (Zhang et al, 1994). The assignments of the amide 'H and N resonances of 15  CBDNI  were obtained using triple resonance correlation  experiments (Bax & Grzesiek, 1993). The 'H resonances from the aromatic rings were identified from DQF-COSY, TOCSY, and NOESY spectra recorded with samples of 1.9 mM unlabelled, 2.0  Chapter 2-Cellooligosaccharide Binding by  mM (81,2, £i, -[ H ]-Tyr and 2  2  4  81,  e , Cl,2, T|2-[ H ]-Trp)-labelled, and 1.9 mM (81,2,61,2^2  2  5  [ H5]-Phe and 8i,£2,Cl,2>'l2-[H5]-Trp)-labelled r  2  28  CBDNI  2  CBDNI,  alone and in the presence of 80 mM  cellotetraose. These assignments were confirmed using the (H(3)C(3(CyC8)H8 and (H(3)C(3(CyC8C£)H£ experiments to connect the ring 'H resonances to the corresponding Cp 13  resonance of each aromatic residue in uniformly C / N enriched C B D N I (Yamakazi et al, 13  i5  1993). The protein samples were lyophilized from the above sample buffer and resolubilized in 99% D 0 for several of these measurements. The NMR assignments and structural analysis of 2  CBDNI  are described in chapter 3.  Titration ofCBD^i with Cellooligosaccharides Monitored by NMR  The binding of soluble cellooligosaccharides to  CBDNI  at 35 °C and pH* 5.9 was  measured quantitatively using >H- N NMR spectroscopy. Stock weight per volume solutions of 15  cellotriose, cellotetraose, cellopentaose, and cellohexaose (Seikagaku Corp.) were prepared in the identical buffer used for the  CBDNI-  The initial concentration of protein was 0.5 mM, except in  the case of the titration with cellotriose for which a 0.18 mM sample was used. Aliquots of the sugar solution were added directly to the protein in a NMR tube and mixed using 1 mm inner diameter PE-160 polyethylene tubing attached to a Gilson Pipetman. To avoid excessive dilution of the  CBDNI,  additions of greater than approximately 1 mg of sugar were made by removing the  protein from the NMR tube and dissolving the solid oligosaccharide directly in the sample solution. For each titration, 8 to 10 tH-^N  HSQC  spectra were recorded consecutively with increasing  concentrations of sugar. The spectra were measured in 1.5 hours with 1024 and 128 complex points obtained in the *H and N dimensions (spectral widths 6000 Hz and 1450 Hz), and 15  processed with zero filling to a final digital resolution of 2.93 and 1.42 Hz/point, respectively. Equilibrium association constants were determined by non-linear least squares fitting of the chemical shift data versus sugar concentration to the Langmuir isotherm describing the binding of one ligand molecule to a single protein site as outlined in the introduction to this chapter. The  Chapter 2-Cellooligosaccharide  Binding by CBD^J  29  programme PLOTDATA (TRIUMF, U B C ) was used for the analyses. In the case of cellohexaose, an estimate of K was made by simulation of the binding isotherm. a  NMR studies of reduced  CBDNJ  A 1.6 mM sample of uniformly N labelled C B D N I s split into two portions. To one 15  wa  urea was added and its H - N HSQC spectrum was recorded (unfolded oxidized). The l  15  concentration of urea was 6.9 M. as measured by refractive index. A 100 fold excess of D,Ldithiothreitol (DTT) was added and another 'H- N HSQC spectrum was obtained (unfolded 15  reduced). With the other portion of C B D N I , ^ - ^ N HSQC spectrum was acquired (folded a  oxidized) then a 100 fold excess of DTT was added and a final 'H- N HSQC spectrum was 15  obtained (unfolded reduced). As the reduction of a disulphide is via the thiolate anion of DTT, this reaction is slow at pH 6. To overcome this kinetic barrier the samples with added DTT were unfolded by heating to 90 °C. Upon cooling H- N HSQC spectra of the urea-free reduced J  15  sample was recorded. The validity of this approach was confirmed by repeating these experiments at pH 8 using a sample of unlabelled C B D N I • At pH 8 heating was not necessary to rapidly obtain reduced C B D N I -  Results  CD and FTIR spectroscopy  The stretching and bending vibrations found in the Amide I region (1700-1620 cm ) of an -1  FTIR spectrum of a protein contain features that are indicative of secondary structure content (Surewicz et al, 1993). Figure 2.4 shows the FTIR spectrum and the second derivative of the spectrum for the Amide I region for C B D N I - The peak at 1633 cm" is indicative of (3-sheet 1  conformation. The absence of peaks at 1658 and 1550 cm which are indicative of the presence of -1  a-helical conformation indicates that  CBDNI  does not contain helical regions.  Chapter 2-Cellooligosaccharide  Binding by  CBDNI  2.4. FTIR spectrum of C B D . In the second derivative (lower trace) of the Amide I region of the FTIR spectrum (upper trace) the peak at 1633 cnr arises from p-sheet conformation. The absence of peaks at 1658 and 1550 cm which are indicative of the presence of a-helical conformation indicates that C B D does not contain helical regions. Figure  N ]  1  -1  N 1  Chapter 2-Cellooligosaccharide  31  Binding by CBDNI  The circular dichroism effect observed for an amide chromophore in the far UV (190-250 nm) depends on its secondary structure (Johnson, 1988). The CD spectra of C B D N Ifreeand cellohexaose and barley-(3-glucan-bound are shown in figure 2.5. The presence of the observed trough at 220 nm is diagnostic of a protein that contains entirely (3-sheets. The CD effect of ahelical conformations, local minima at 208 nm and 222 nm, is much stronger than for p-sheets. The absence of these two local minima strongly suggest the absence of ot-helices in C B D N I • Also, the close similarity of the CD spectra of both free and cellohexaose and barley-f3-glucan-bound CBDNI  indicates there is\very little change in secondary-structure content upon sugar binding. Figure 2.5 also shows CD melts of C B D N I free, as well as bound to cellohexaose and  barley-p-glucan. Free protein thermally unfolds at 47.5 °C, cellohexaose-bound protein unfolds at 50.7 °C and barley-(3-glucan-bound C B D N I unfolds at 50.6 °C. The higher unfolding temperature of the latter two indicates both cellohexaose and barley-0-glucan bind C B D N I stabilising the folded structure. A similar increase in the unfolding temperature of C B D N I when cellohexaose-bound was seen in one-dimensional NMR spectra (data not shown). Cooling of the thermally-denatured protein results in an identical CD spectrum to that found before denaturation. There is little difference in the CD spectrum and denaturation temperature of C B D N  1  unfolded. This indicates that the thermal unfolding of  CBDNI  is a reversible process.  Thermodynamic characterisation of the stability of  and the effect of oligosaccharide  CBDNI  that has been repeatedly  binding is given in Creagh et al. (1997).  CBD^i  Binds Soluble  Cellooligosaccharides  The binding of C B D N I to cellotriose, cellotetraose, cellopentaose, and cellohexaose was detected initially by the observation of numerous changes in the 'H-NMR spectrum of the protein resulting from the addition of these soluble sugars (data not shown). In contrast, the spectrum of CBDNI  remained unperturbed in the presence of cellobiose, indicating that the protein does not  bind this disaccharide appreciably. Cellooligosaccharides longer than cellohexaose were not  Chapter 2-Cellooligosaccharide Binding by CBD^i  05 C  Q O  20 18 16 14 12 10 8 6 4 2 0 -2  32  i—i—i—i—i—r Cellohexaose-bound barley-p-glucan-bound 4  Unbound  200  240  220 Wavelength (nm) i  1  1  .  •  Cellohexaose-bound  •  Barley-p-glucan-bound" Unbound  \  40  60  80  Temperature (C)  Figure 2.5. (Top) Far U V C D spectra of free and oligosaccharide-bound CBDNI- The broad trough at 220 nm indicates the presence of [3-sheet conformation. The close similarity of the C D spectra of both free and cellohexaose and barley-p-glucan-bound C B D i indicates that there is very little change in secondary-structure content upon sugar binding. (Bottom) C D melts of C B D N I free, as well as bound to cellohexaose and barley-P-glucan. Free protein thermally unfolds at 47.5 ° C , cellohexaose-bound protein unfolds at 50.7 ° C and barley-p-glucan-bound C B D N I unfolds at 50.6 ° C . The higher unfolding temperature of the latter two indicates that both cellohexaose and barley-P-glucan bind C B D N I , stabilising the folded structure. N  Chapter 2-Cellooligosaccharide  Binding by CBDNI  33  investigated as these compounds are not commercially available and have limited solubility in aqueous buffers. The interaction of C B D N I with cellotriose, cellotetraose, cellopentaose, and cellohexaose was quantified using two-dimensional 'H- N correlation spectroscopy. Figure 2.6 shows an I5  overlay of ten !H- N HSQC spectra recorded as uniformly N enriched C B D N I was titrated with l5  15  cellotetraose. It is readily seen that many peaks, arising from the backbone amide groups of the protein, show significant changes in both 'H and N chemical shifts with the progressive addition 15  of the sugar. In the case of each cellooligosaccharide investigated, the free and bound forms of the protein are in fast exchange on the NMR timescale, resulting in the observation of population weighted average chemical shifts throughout the titration series.  Binding Stoichiometry  Cellotriose, cellotetraose, cellopentaose, and cellohexaose bind to  CBDNI  with a  stoichiometry of one sugar molecule per protein molecule. This conclusion is supported by the following evidence. First, as exemplified in figure 2.7 for cellotetraose, all the amides with 'H and N chemical shifts that are perturbed by sugar binding show co-incident titration curves. This 15  discounts the possibility of multiple binding sites on the CBD with differing affinities for the cellooligosaccharides. Given that C B D N I is monomeric in solution, it is unlikely to have two or more distinct binding sites with equal affinities for the sugar ligands. Therefore the simplest interpretation of the co-incident titration curves is that cellooligosaccharides bind to C B D N I at a single site and that each perturbed ^H-^N group reports the same association event. Second, plots of  CBDNI  amide chemical shifts versus added cellopentaose or cellohexaose show a plateau at  approximately equal concentrations of total sugar and protein, indicating a binding stoichiometry of 1:1 (figure 2.8). Cellotetraose and cellotriose do not exhibit such pronounced titration end points due to their lower affinities for C B D N I • However, all four sugars cause similar changes in the 'Hl5  N spectrum of the labelled protein, strongly suggesting that each binds to C B D N I with the same  stoichiometry and at the same site. Third, the titration curves measured for the four  Chapter 2-Cellooligosaccharide Binding by  eTe  9T0  9] 2  CBDNI  HI  :  eTi  a! A  (ppm)  Figure 2.6. Cellotetraose binds to CBD N 1 . A portion of ten 'H- N HSQC spectra 15  of uniformly N labelled protein in the presence 0, 0.04, 0.08, 0.16, 0.36, 0.51, 0.76, 1.26, 2.24, and 10.5 mM total cellotetraose are overlayed. The arrows indicate the directions in which the amide ^ - ^ N peaks shift with added sugar. Although all H- N resonances have been assigned, the peaks in the crowded region of the spectrum are not labelled for clarity. Q80 designates the NH of Gln80. The sample buffer was 50 mM sodium chloride, 50 mM potassium phosphate (pH* 5.9), and 0.02% sodium azide in 10% DO/90% H 0 at 35° C. 15  L  15  e2  2  2  15  e2  Chapter 2-Cellooligosaccharide Binding by CBD^j  Figure 2.7. Cellotetraose binds to CBD N 1 at a single site. The co-incident plots of the  normalized H chemical shift changes for residues Tyrl9, Gly44, Thr87, and Glyl30 versus total added cellotetraose demonstrate that each amide group in CBD monitors the same binding event. The solid lines represent the titration isotherms obtained by fitting the observed data points (0) to the equation describing the association of cellotetraose and CBD | to form a 1:1 protein-sugar complex. For clarity, the overlapping data and fits for the four residues are not labelled individually. N  N1  N  Chapter 2-Cellooligosaccharide Binding by CBD^j  Figure 2.8. The association constants of C B D ^ i for soluble cellooligosaccharides were determined from titration curves monitored by i J T - ^ N N M R spectroscopy. The amide H chemical shift of Tyr 19 is plotted as a function of the total concentration of added cellohexaose (circle), cellopentaose (square), cellotetraose (diamond), and cellotriose (star). The solid lines represent the best fit of the experimental data to the equilibrium equation describing binding to a single protein site, or in the case of cellohexaose, a simulated titration curve based on an estimation of the association constant. The arrow marking the plateau i n the titration curves for cellopentaose and cellohexaose falls at the point where the total sugar concentration equals the total protein concentration (-0.5 m M C B D ) indicating a 1:1 binding stoichiometry. N  N l  37  Chapter 2-Cellooligosaccharide Binding by CBDNI  cellooligosaccharides are adequately fit to the binding isotherm describing the simple equilibrium expressed by equation (2.2) (figures 2.7 and 2.8). Finally, this conclusion is confirmed by isothermal titration microcalorimetry (Tomme et al., 1996a).  Association Constants Determined by NMR Spectroscopy  The association constants (K ) describing the interactions of cellotriose, cellotetraose, and a  cellopentaose to C B D N I were determined by non-linear least squares fitting of the chemical shift titration data to the binding isotherm for a protein with a single ligand recognition site. Figure 2.8 shows an example of the analysis of data measured for Tyrl9, as well as a comparison of the binding curves for each cellooligosaccharide. Two K values were determined independently for a  each amide showing a significant chemical shift change upon sugar binding, one using the data for the amide H and the other for the N nucleus. In this manner, association constants were N  15  determined from titration curves measured for 10 to 16 amides in the protein. These included Vall7, Tyrl9, Val34, Tyr43, Gly44, Val45, Gly46, Leu49, Asn81, Gly82, Thr87, Alal26, Glyl30, Leul39, Leul41, and Alal45. The average association constants for the three cellooligosaccharides were calculated using the values obtained individually from the analyses of the H and N data recorded for each amide (Table 2.1). In the case of cellohexaose, binding to N  15  the C B D N I was sufficiently tight that, given the concentration of protein required for NMR analysis, it was not possible to determine accurately the K . Therefore an estimate of the upper a  limit of the association constant was obtained by a visual comparison of simulated titration curves with the observed NMR data (figure 2.8 and Table 2.1). The association constants of C B D N I for the cellooligosaccharides measured by NMR agree with those determined independently by isothermal titration calorimetry under similar experimental conditions (Tomme et al, 1996a). Due to the different concentration windows necessary for each method, NMR provided a good measure of the relatively weak binding of cellotriose, whereas calorimetry yielded a more accurate value of the association constant of cellohexaose with the CBD.  38  Chapter 2-Cellooligosaccharide Binding by CBDj^j  Table 2.1. Association Constants K for the Binding of Soluble Cellooligosaccharides to CBD i a  N  Ligand ^a(M-') K (M ) Cellotriose 180+60 not determined Cellotetraose 4200 ± 720 3200 ± 500 Cellopentaose 34000 ± 7600 21000 ±3000 Cellohexaose (50000) 22000 ± 4000 Data obtained at 35 °C and pH* 5.9 in 50 mM sodium chloride, 50 mM potassium phosphate, 0.02% sodium azide and 10% D2O/90% H2O. The reported K values are the average of the those determined from the H and N chemical shift perturbations of Vall7, Tyrl9, Val34, Tyr43, Gly44, Val45, Gly46, Leu49, Asn81, Gly82, Thr87, Alal26, Glyl30, Leul39, Leu 141, and Alal45. The error range is one standard deviation. The K reported for cellohexaose is an estimate of the upper limit of the association constant based upon simulations of the titration curves. Values determined by isothermal titration calorimetry at 35 °C and pH 7.0 in 50 mM potassium phosphate, 0.02% sodium azide (Tomme etal, 1996a). a  A b  a  a  a  L  15  &  b  Identification of the Cellooligosaccharide Binding Site  In addition to providing a means for quantitating the association of the cellooligosaccharides to C B D N 1 > NMR also yields information regarding the location and structure of the binding site. Perturbations of the resonances of main chain H and N nuclei upon sugar N  15  binding may arise due to the direct interaction of an amide with the ligand, or indirectly due to conformational changes resulting from the formation of the protein-sugar complex. These conformational changes are likely subtle as circular dichroism spectra reveal that the global secondary structure of C B D N I is unaffected in the presence of saturating quantities of cellotetraose. Bearing this in mind, structural insights into the binding site are provided by the pattern of NMR chemical shift changes observed with sugar binding. Figure 2.9 summarizes the changes of the H and N chemical shifts of each amide in C B D N I due to the binding of cellotetraose. Similar N  15  39  Chapter 2-Cellooligosaccharide Binding by CBDNI  shift perturbations result from the binding of cellotriose, cellopentaose and cellohexaose (not shown). As will be shown in chapter 3  CBDNI  is composed of 10 (3-strands, folded into two anti-  parallel [3-sheets identified as A and B (figure 2.9). As illustrated schematically in figure 2.10, the global topology of C B D N I is that of a jelly-roll p-sandwich. Residues with exposed side-chains that form P-sheet A are identified on this diagram. The average change in the absolute value of the H chemical shifts of all resolved amides in N  CBDNI 15  due to cellotetraose binding was 0.04 ppm with a standard deviation of 0.05 ppm. In the  N dimension, the average absolute change in shift was 0.3 ppm, with a standard deviation of 0.4  ppm. Of the 24 residues with a perturbation in amide H or N chemical shift greater than one N  15  standard deviation above the average, 18 lie within or immediately adjacent to the p-sheet of CBDNI  composed of strands A1-A5 (figures 2.9 and 2.10). This strongly indicates that the  binding site for the cellooligosaccharides lies on the face of the CBD formed by these strands. Furthermore, residues showing pronounced chemical shift changes upon the addition of sugar are located in all five of the P-strands that form this sheet. Therefore, the cellooligosaccharides are likely to bind across, and not parallel to, strands A1-A5. This conclusion is supported by the observation of intermolecular proton NOE interactions between bound cellotetraose and protein side-chains located within P-sheet A in C B D N I (chapter 3) and by the attenuation of the intensities of residues in this sheet by the binding of nitroxide-labelled cellooligosaccharides (chapter 6). A limited number of amides located outside of p-strands A1-A5 also show changes in H  N  or N chemical shift upon sugar binding that are greater than one standard deviation from the l5  average. All 6 of these amides either flank the Cys33-Cysl40 disulfide bond (Val34, Leu 139, Leu 141, and Alal45), or lie in p-strand B3, which is adjacent to the strands containing this cysteine group (Thr67 and Ala68). As will be discussed in chapter 4 many of these perturbations arise from a contaminating amount of calcium present in the cellotetraose used in this study. However, some residues near the disulphide bond are still perturbed by cellopentaose binding in the fully calcium-bound C B D N l • It is possible that sugar-binding indirectly influences a structural  40  Chapter 2-Cellooligosaccharide Binding by CBD^j  A  A l  BI  10  B2  20  30  A § p | l ] § E § @ @ @ @ G P©G(W®g)©G T D G P L D T S T Ch ange in  A2  40  50  C © A V P A G S A Q©30©©V©N  0.3-,  Shift (ppm)  1.8  A3  60  A4  70  80  90  100  V A I E E G T T Y T L R Y T A @ @ S TD V T V R A L @ G Q ® G A P @ 3 © V L D T S P A L T S E P R Q Change in Shift (ppm)  0.3  -mr  B4  A5  110 T E T  Change in Shift (DDm)  F T A S A T  120 Y  P A T  P A A D D  ^  B5  130 P E G @ I  A @ Q L @ G  140 F S A D A W T ( L ) C ©  150 D§[V](§)[L]D S  E  V E  Chapter  2-Cellooligosaccharide  Binding  by  CBDNI  Figure 2.9. (A) (previous  page) Patterns of NMR chemical shift perturbations due to sugar binding demonstrate that the cellooligosaccharide-recognition site of CBD is formed by a five-stranded anti-parallel (3-sheet. The absolute values of the differences between the H and N chemical shifts of the main chain amides in the free and cellotetraose-bound forms of CBD are indicated as positive and negative numbers, respectively. Blank spaces identify residues for which no difference could be determined. The locations of the ten (3-strands in the protein are shown by the arrows above the amino acid sequence. Two P-sheets identified in CBD are composed of strands A1-A5 (open) and B1-B5 (solid). Circled residues have a change in H and N chemical shift upon binding greater than one standard deviation from the mean absolute value change observed for all measured residues. Based on the pattern of chemical shift perturbations, the oligosaccharide-binding site lies across the face of the protein formed by strands A1-A5. Residues not detected in the 'H- N HSQC spectrum of CBD in the absence of added cellotetraose are boxed. N]  N  15  N]  N]  N  15  15  N]  (B) (this page) C worm diagram of CBD with residues that experience the largest change in chemical shift with cellopentaose binding coloured red. The top panel shows P-sheet A, the binding face. The lower panel is a 90° rotation from the view shown in the top panel, the binding cleft is seen at the top of the structure in this bottom view. The amino and carboxyl termini are denoted by the labels N and C, respectively. This figure was made using the programme GRASP (Nicholls et al., 1991). a  N1  Chapter 2-Cellooligosaccharide Binding by CBD^j  Figure 2.10. Schematic representation of the jelly-roll (3-sandwich fold of C B D . Based on the observed patterns of chemical shift perturbations resulting from sugar binding, the recognition site for the cellooligosaccharides lies across the face of the N ]  protein formed by (3-strands A1-A5 (open arrows). Residues in these (3-strands that have exposed side-chains are labelled. Cys33 and Cysl40 in strands B2 and B5, respectively, form a disulfide bridge. The loops connecting the (3-strands are not drawn to scale.  Chapter 2-Cellooligosaccharide Binding by CBD^j  43  feature of the disulfide bond, such as disulfide isomerism, resulting in the observed chemical shift perturbations. The changes in chemical shifts of remaining amides located in the (3-sheets and loops of the protein are less than one standard deviation from the average. It is likely that these chemical shift differences reflect subtle structural perturbations associated with ligand binding that are propagated through the core of the protein.  Effect of Cellotetraose Binding on the Aromatic Residues in  CBDNI  To investigate the possible roles that the aromatic side-chains play in sugar binding, the 'H resonances of the phenylalanine, tyrosine, and tryptophan residues in C B D N I were assigned in the absence and presence of saturating amounts of cellotetraose. Samples of the protein in which either tryptophan and tyrosine or tryptophan and phenylalanine were uniformly deuterated at all ring positions were prepared to simplify the NMR spectra of  CBDNI  and to identify  unambiguously the 'H resonances from the phenylalanine or tyrosine residues, respectively. Near complete aromatic 'H assignments were obtained for both the unbound and bound forms of CBDNI  (figure 2.11). An exception is Phel06, for which no resonances are observed in the  homonuclear spectra of the protein. This residue, as well as Tyrl9 and Tyr85, which show extensive line broadening, may undergo conformational averaging on a timescale of intermediate exchange due to partially hindered ring flipping (Wuthrich, 1986). The differences between the aromatic 'H chemical shifts of the free and cellotetraose-bound forms of the protein are summarized in figure 2.12. The average change in the absolute values of the 'H chemical shifts for all aromatic side-chains was 0.07 ppm with a standard deviation of 0.09 ppm. Of the assigned aromatic residues, only Phe9, Trpl6 and Tyrl9 show above average ring chemical shift changes upon binding the oligosaccharide. In addition, the lineshapes of the aromatic proton resonances of Tyr85 are clearly altered in the present of cellotetraose (figure 2.11). As illustrated in figure 2.9, Trpl6, Tyrl9, and Tyr85 are located within (3-sheet A and also exhibit significant changes in amide K and N chemical shifts upon sugar binding. This provides [  15  Chapter 2-Cellooligosaccharide Binding by  44  CBDNI  Figure 2.11. Aromatic region of the DQF-COSY spectrum of [ H ]-Phe- [ H ]-Trp-labelled C B D in the (a) absence and (b) presence of 80 mM cellotetraose. Only the resonances from the six tyrosine residues are detected because the aromatic rings of the tryptophan and phenylalanine residues in the protein were biosynthetically deuterated and the amide H signals eliminated by reversibly unfolding the protein in 99% D 0 buffer. 2  2  5  5  N 1  N  2  Chapter 2-Cellooligosaccharide Binding by CBD^j  0.35  F9 W16 Y19 Y43 Y60 Y64 Y112F127 F132 W137  Aromatic Residue Histogram summarizing the perturbations of the H chemical shifts of the aromatic rings of the tryptophan, tyrosine, and phenylalanine residues in C B D N I due to the binding of cellotetraose. The data represent the average of the absolute value of the chemical shift changes observed for all protons in the aromatic sidechain of each residue. The average change for all assigned aromatic rings is 0.07 ppm with a standard deviation of 0.09 ppm. The chemical shifts of Tyr85 could not be confidently determined due to linebroadening and are not included in the figure. The resonances of Phel06 were not observed in the iH-NMR spectra of C B D . Figure 2.12.  l  N 1  46  Chapter 2-Cellooligosaccharide Binding by CBDNI  further support for the identification of the sugar binding site in C B D N I - Although Phe9 and Trp 16 show pronounced NMR chemical shift changes upon sugar binding, structural studies reveal that the side-chains of these residues are located within the interior of C B D N I - Additionally, these residues are located close to the calcium-binding site in C B D N I (chapter 4). Their displayed chemical shift changed probably arise from the binding of the calcium present in the cellotetraose used in this study, and not due to cellotetraose itself. This leaves Tyrl9 and Tyr85 as the only two residues with exposed aromatic rings that are likely to be involved in the binding of cellooligosaccharides.  Effect of High Concentrations of Cellotetraose  In the absence of added cellotetraose, resonances from the amides of 13 residues are not observed in the ' H - ^ N  HSQC  spectrum of C B D N I - These residues are located at the N-terminus  of the protein and in p-strands B2 and B5, next to the disulfide bond (figure 2.9). The H and N  15  N resonances from the amides in question are observed in the presence of high concentrations of  cellotetraose. As will be discussed in chapter 4, the appearance of these amide resonances is due to a calcium contamination in the cellotetraose used in this study.  CBDNI  tightly binds calcium at a  site located on the opposite side of the protein from the binding face. The affinity of C B D N I for oligosaccharides is unaffected by the binding of calcium. Instead, calcium stabilises the tertiary structure of C B D N I - The presence of calcium in the cellotetraose used in this study therefore does not affect any of the results presented in this chapter.  Reduction of the disulphide in CBDNI CBDNI  results in the protein unfolding  contains two cysteine residues located at positions 33 and 140. As shown by the  lack of reactivity to 5,5'-dithiobis-(2-nitrobenzoic acid), by electrospray mass spectroscopy and by structural data (chapter 3), these residues form a disulphide bond in the native protein. The effect of reducing this disulphide to the free sulphydryls was studied by ' H - N 15  Figure 2.13 shows spectra of oxidised and reduced  CBDNI  HSQC  spectroscopy.  in the presence and absence of a  47  Chapter 2-Cellooligosaccharide Binding by CBDj^j  oxidized  •  5  °  o  a  reduced  Co  S ^ e r o ^ f l f i  0  'H(ppm) "  'H(ppm) '  »  oxidized 7M urea  VB a°  0  -  O  reduced 7M urea  D  o  •  7 1  Figure 2.13. 'H-  H (ppm)  'H(ppm)  ^ spectra at 35°C and pH6.0 of native, oxidised C B D N I , protein denatured by reduction, addition of 7 M urea, and reduction in the presence of 7 M urea. 1  A  N  D  Chapter 2-Cellooligosaccharide Binding by CBDNI  48  denaturing amount of urea. It is clearly seen in this figure that the only condition in which the resonances are well dispersed in the 'H dimension, indicating that the protein is folded, is in the absence of urea and when the disulphide is present. There is no evidence of folded reduced protein being present. Exactly the same results were found with spectra collected at 7 °C. All three forms of denatured protein remain soluble in these experiments, even at the high concentrations used (1.6 mM). Upon exposure to air, the reduced sample reoxidises and completely refolds. Similarly with the urea denatured forms, exchanging the protein into buffer without urea results in refolding of the protein. There are slight differences for each of the three forms of denatured C B D N I , probably reflecting differences in solution conditions, and the presence of a disulphide in the urea denatured C B D N i •  Discussion  CBDNI  Binds Soluble Cellooligosaccharides  I have shown using heteronuclear NMR spectroscopy that  CBDNI  binds soluble  cellooligosaccharides to form 1:1 protein-carbohydrate complexes. From quantitative titration measurements, it was found that the affinity of C B D N I for cellooligosaccharides increases in the order cellotriose < cellotetraose < cellopentaose ~ cellohexaose. The basis for the differences in the affinities for these sugars likely reflects the size of the  CBDNI  binding site. If this site is the  same length as cellopentaose, then cellopentaose and cellohexaose would be expected to bind approximately equally well, while shorter ligands would exhibit diminished affinity. As will be shown in chapter 3, the binding site is approximately the same length as cellopentaose.  Binding Site is Formed by a 5-Stranded p-sheet  The chemical shift of a nucleus is very sensitive to changes in its local environment, thus providing an avenue for identifying the binding site of a ligand on a protein (Otting, 1993). When  Chapter 2-Cellooligosaccharide Binding by CBD^j  49  mapped onto the sequence of C B D N I , it is clear that most amides with chemical shifts perturbed significantly due to oligosaccharide binding are located within one of the two (3-sheets of this jellyroll (3-sandwich protein (figure 2.9). Furthermore, changes in chemical shifts greater than one standard deviation above average occur for amides in each of the five strands of this sheet. I therefore conclude that the binding site for cellooligosaccharides lies across the face of C B D N I that is composed of strands A1-A5 (figure 2.10). With the exception of the anomalous behaviour observed for a few amides in the presence of high concentrations of cellotetraose (resulting from calcium binding, chapter 4), the effects of all four cellooligosaccharides on the tH- N HSQC spectra of C B D N I are very similar. That is, 15  saturating amounts of cellohexaose and cellotriose shift approximately equally the amide H and N  15  N resonances of the protein. This implies that the four sugars bind C B D N I at the same location  and that binding results in a common perturbation of the backbone structure of the CBD. A detailed discussion of the binding mechanism of C B D N I with cellooligosaccharides in relation to the side-chains present on the binding face is given in chapter 3.  Comparison to Other CBDs CBDNI  has the distinctive feature of binding solely to phosphoric-acid swollen cellulose  (PASC), and not crystalline cellulose (Coutinho et ai, 1992). In contrast, all other known CBDs bind both crystalline cellulose and PASC (chapter 1; Tomme et al, 1995). The binding of cellooligosaccharides by  CBDNI  reflects its binding specificity for cellulose. PASC is a  heterogeneous disordered form of cellulose. Its surface contains large areas where the interstrand hydrogen bonds are broken, resulting in regions that resemble the single chain cellooligosaccharides used in this study. The binding of cellooligosaccharides by two other CBDs has also been studied. Evidence for the role of the tryptophan residues of CBDcex in sugar binding was obtained using NMR spectroscopy to monitor the effects of cellohexaose on the spectrum of the protein (Xu et al, 1995). When four equivalents of this ligand were added to a uniformly N-labelled sample of 15  Chapter 2-Cellooligosaccharide Binding by  CBDcex* the indole H  el  and N 1 5  e l  50  CBDNI  resonances of two exposed tryptophans were perturbed  slightly. Although this supported the role of the surface tryptophans in cellulose binding, it also demonstrated that CBDcex interacts only weakly with soluble cellooligosaccharides. Recently, a study of the interaction between T. reesei CBDs and cellohexaose was published (Mattinen et al, 1997b). No mention was made of changes in the chemical shift of resonances of the protein. Instead binding was monitored by observing changes in the line width of resonances of cellohexaose. A dissociation constant of 350 ± 90 pJVI, corresponding to a K of -3000 M , was -1  d  found for  CBDCBHI  binding cellohexaose.  In contrast to the studies with CBDcex d C B D C B H I an  > CBDNI  tightly binds cellohexaose,  and the effect of ligand binding on the NMR spectrum of the protein is extensive. As shown in figure 2.9, approximately 40% of the main chain amides in C B D N I  a r e  the soluble sugars. The weak binding to cellohexaose by CBDcex d an  perturbed upon addition of CBDCBHI  indicates this  cellooligosaccharide is a poor model for crystalline cellulose. These CBDs probably need a less flexible substrate or have to interact with multiple chains of cellulose to obtain tight binding. In contrast, cellohexaose closely resembles the natural binding substrate for C B D N I d is bound an  tightly.  Stability of CBDj^i  From the CD melt data it was shown that the thermal unfolding of C B D N I is a completely reversible process. Though C B D N I has a relatively low denaturation temperature, 47 °C, it increased with the addition of both cellohexaose and barley-p-glucan. This indicates that these substrates are bound by C B D N I and oligosaccharide-binding stabilises the folded structure of CBDNI-  These results provided the ground work for a detailed calorimetric study of the stability  of C B D N I , as presented in Creagh et al. (1997). The fact that C B D N I can be unfolded simply by reducing its single disulphide is surprising. Similar unfolding of proteins upon reduction has been observed before, but generally with proteins that contain multiple disulphides such as bovine pancreatic trypsin inhibitor (BPTI), hen lysozyme  Chapter 2-Cellooligosaccharide Binding by CBD^j  51  and ribonuclease (Creighton, 1993). In general, disulphides act to decrease the entropy of an unfolded protein so a smaller entropy penalty of folding is encountered. The greater the number of residues in the loop created by a disulphide, the greater the decrease in entropy. For C B D N I the disulphide forms a loop of 106 residues out of a total of 152. The results of Creagh et al. (1997) show C B D N I has a AG of unfolding of 32 kJ mol- at 7 °C. This implies the disulphide imparts at 1  least this much free energy to the stability of the protein. This work was performed on CBDNI  CBDNI  not calcium bound. As will be shown in chapter 4  binds a calcium ion and calcium binding stabilises the tertiary structure of the protein  (L.Creagh & C. Haynes, personal communication). It is possible that reduction of the disulphide in calcium-bound C B D N I might not result in unfolding of the protein.  52  Chapter 3 Structure Determination of  C  B  D  N  I  Abstract  Multidimensional heteronuclear NMR spectroscopy was used to determine the tertiary structure of the 152 amino acid N-terminal cellulose-binding domain from Cellulomonas 1,4-glucanase CenC (CBDNI).  CBDNI  fimi  (3-  was studied in the presence of saturating concentrations of  cellotetraose, but due to spectral overlap, the oligosaccharide was not included in the structure calculations. A total of 1705 interproton NOE , 56 <|), 88 \)/, 42 %1, 9 %2 dihedral angle and 88 hydrogen bond restraints were used to calculate 25 final structures. These structures have an rmsd from the average of 0.79 ± 0.11 A for all backbone atoms excluding disordered termini, and 0.44 ± 0.05 A for residues with regular secondary structures.  CBDNI  is composed of 10 (3-strands,  folded into 2 anti-parallel (3-sheets with the topology of a jelly-roll [3-sandwich. The strands forming the face of the protein previously determined by chemical shift perturbations to be responsible for cellooligosaccharide binding (chapter 2; Johnson et al, 1996a) are shorter than those forming the opposite side of the protein. This results in a 5-stranded binding cleft, containing a central strip of hydrophobic residues that is flanked on both sides by polar hydrogenbonding groups. The presence of this cleft provides a structural explanation for the unique selectivity of C B D N I for amorphous cellulose and other soluble oligosaccharides, and the lack of binding to crystalline cellulose. The tertiary structure of C B D N I is strikingly similar to that of the bacterial 1,3-1,4-p-glucanases, as well as other sugar-binding proteins with jelly-roll folds.  Chapter 3-Structure Determination of  53  CBDNI  Introduction  This introductory section is designed to familiarise the reader with some of the procedures used in the process of resonance assignment and structure calculation of C B D N I (figure 3.1). Specific details on these NMR experiments will be given in the experimental methods section. The work presented in this chapter has been published (Johnson et al., 1996b). With the exception of the resonances assignment of the aromatic rings, and the preparation of some of the samples that aided in the resonance assignment, I performed all the work presented in this publication.  Resonance assignment of proteins  The development of multi-dimensional triple-resonance spectroscopy in the late 1980's and early 1990's helped revolutionize the field of protein NMR (Bax & Grzesiek, 1993). Providing it is possible to obtain a C / N labelled sample of about 1 mM that is stable for at least three 13  I5  weeks, it is a routine matter to assign the backbone and side-chain resonances of proteins up about -20 kDa. Recently, with the use of H, C and N labelling, this molecular weight limit has 2  l3  15  been extended significantly (Sattler & Fesik, 1996; Kay & Gardner, 1997). The key to the success of the triple-resonance experiments is the exclusive use of throughbond couplings to connect resonances from nuclei within adjacent amino acids. These experiments exploit the large coupling constants between C and N nuclei and their directly attached protons 13  15  for efficient magnetization transfer. Previously, with the use of homonuclear experiments, potentially ambiguous through-space NOE connections were necessary to link the spin-systems for each amino acid (Wuthrich 1986). A second development that has improved the sensitivity of experiments, and also helped to remove artifacts, is the incorporation of pulsed field gradients into pulse sequences (Keeler et al., 1994; Kay, 1995). These benefits result from pulsed field gradients reducing the need for phasecycling, selecting for transfer pathways that do not suffer from sensitivity loss, and improved water suppression.  Chapter 3-Structure Determination of  CBDNI  preparation of C / N labelled protein 1 3  1 5  I  assign backbone resonances assign aromatic side-chain resonances  assign aliphatic side-chain resonances  hydrogen-deuterium exchange  I  stereospecifically assign methyl and HP resonances complete assignments obtained  analyze J experiments  assign N O E S Y spectra  I  I  distance restraints  torsion angle restraints  hydrogen bond restraints  final structure ensemble Figure 3.1. Flowchart outlining the steps in the structure determination process used for C B D . N 1  Chapter 3-Structure Determination of CBD^j  55  An especially effective method to assign the backbone amide chemical shifts is the combination of HNCACB (Wittekind & Mueller, 1993) and CBCA(CO)NH (Grzesiek & Bax, 1992) experiments. The HNCACB experiment correlates the N and H resonances of an amide l5  N  with its own C and CP nuclei, as well as those of the preceding residue. In contrast, the a  CBCA(CO)NH experiment correlates the N and H resonances of an amide exclusively with the 15  N  C and CP nuclei of the previous residue (figure 3.2). Also, for the HNCACB experiment, when a  transformed, the signals due to the C are 180° out-of-phase with respect to the CP signals. As a  shown in figure 3.3 it is possible from inspecting the dataset for peaks at the appropriate frequency to connect the spin systems together. This results in the assignment of the backbone N and H 15  N  as well as the side-chain C and CP resonances. a  Near complete assignment of the remaining side-chain 'H and C resonances can be 13  obtained from the H(CCO)NH-TOCSY and (H)C(CO)NH-TOCSY experiments, figure 3.4 (Logan et ai, 1992; Montelione et al., 1992; Grzesiek etal., 1993a). These experiments connect the side-chain 'H or C resonances with the N and H resonances of the following residue. To 13  15  N  assign residues that precede a proline, and to correlate resonances from directly bonded 'H and 13  C nuclei HCCH-TOCSY (Bax et al., 1990; Kay et ai, 1993) and HCCH-COSY (Kay et al,  1990; Ikura et al., 1991) experiments are also run. Once the protein has been assigned as completely as possible the process of obtaining restraints for structure calculations begins.  Structure calculations  The determination of a three dimensional structure of a protein by the NMR method relies on the assignment of a large number of relatively short range, 5 A or less, structural restraints. These distance restraints are based on the observation of nuclear Overhauser effects (NOEs) between pairs of protons. NOEs are assigned from two, three and four-dimensional NOESY experiments. The distances derived from NOEs are not exact, but rather used as a range with the lower limit being the van der Waals radii of the two protons (1.8 A) and the upper limit either 2.9 A, 3.5 A or 5 A for strong medium and weak NOEs, respectively.  56  Chapter 3-Structure Determination ofCBD^j  H—C—H  H-tcT-H  H  -T-cVl—H  H-(C4-H  4A H  H  f-C-JN H  H  H —C—H  H—C—+1  H—{Sl—H  H —C—H  O  U-cI II  H  H  O  H  H  —Iii—H  i  r  H  H  CBCACO(CA)HA  H  C — C-  IH  I II  H  O  H  H—C—H  HUC-m  H—C—H  IU-cI II  —C-rN  C-  I rH II  1 11  O  O  H  HNCO  H—C—H  H  H—d —N  I I  H—C—H  I I  H  O  H—C— H  H—C—H  — H  CBCA(CO)NH  HNCACB  i T  H  —(i  O  H  (H)C(CO)NH-TOCSY  H —C—H  [Hj  O  H  H  O  H(CCO)NH-TOCSY  H _ C _ H  N—C—C—N—C—C-  I rh II I I ..  H  H  O  H  H O  HCCH-TOCSY  (H(3)Cp(CyC8)H6  H  H  &  H  H  il  HCCH-COSY  (Hp)Cp(CyC5Ce)He  Figure 3.2. Three dimensional heteronuclear experiments used to assign the resonances of C B D . Boxed atoms are detected in the experiments. N1  Chapter 3-Structure Determination ofCBD^j  oe  e  ON  ©  fa H < oa < E>Figure 3.3 Sequential assignments of proteins can be obtained by the identification of common C a and C P resonances in the H N C A C B (left) and C B C A ( C O ) N H (right) experiments. Shown here are strips through the amide resonances of F106 to Y l 12 of C B D N I . The C B C A ( C O ) N H experiment provides only inter-residue correlations between 15N(i) and HN(i) resonances and C a ( i - l ) and CP(i-l), while both inter-residue Ccc(i-l) and CP(i-l), and intra-residue Ca(i) and CP(i), correlations to 15N(i) and HN(i) resonances are identified in the H N C A C B spectrum. Intra-residue C a and C P correlations are indicated. C a correlations (green) in the H N C A C B spectrum appear 180 out-of -phase with respect to the C P correlations (red).  Chapter 3-Structure Determination  58  O/CBDNJ  B  O O CM  o no E  a  CL  |P3 **  O  @  P  0  c  0  a  -Q L  o ^ • E ^ a a o LT)  0 O  o o  CD  CD  O j .  U.  I -  <  (fi  <  I— > -  c o h .  oo  a)  o  O  O  O  i-  T  O LtZ  H  <  CO  <  h-  -  OJ  i-i>-  Figure 3.4. The assignment of side-chain resonances can be obtained by analysis of (H)C(CO)NH-TOCSY and H(CCO)NH-TOCSY spectra. The (H)C(CO)NH-TOCSY experiment (A) provides correlations between the 15N(i) and HN(i) resonances and the carbon resonances of the previous residue (i-l), while the H(CCO)NH-TOCSY experiment (B) correlates 15N(i) and HN(i) resonances with the proton resonances of the previous residue (i-l). Shown here are strips through the amide resonances of F106 to Y l 12 of CBDNI. Peaks below the dotted line in B indicate they appear alised in this spectrum. Peaks marked (*) are examples of spectral overlap of different residues with similar 15N and H N frequency. Peaks marked (**) are due to residues that are not overlapped but have such strong intensities that they appear in the strips of nearby residues.  IT  59  Chapter 3-Structure Determination of CBDNI  Torsion angle restraints are used for protein structure calculations. Values of the various dihedral angles are deduced by measuring three-bond proton-proton, nitrogen-carbon and carboncarbon coupling constants (Bax et al, 1994). Backbone phi ((j)) as well as side-chain chil (%1) and chi2 (%2) angles are the most amenable dihedral angles for this process. Hydrogen bond restraints are also very useful. These restraints are deduced from hydrogen-deuterium exchange experiments that identify slowly exchanging amide protons. After the secondary structure of the proton has been elucidated, these amides are paired with an oxygen atom. Hydrogen bonds are included as restraints between O and N, and O and H atoms. N  Initially, it is not possible to unambiguously assign most of the peaks present in the NOESY spectra. This is due to chemical shift degeneracy that is not resolved even in fourdimensional experiments. A starting set of dihedral and distance restraints derived from a limited number of unambiguously assigned NOEs are used to calculate initial structures. Based on these preliminary structures, additional restraints are assigned in a reiterative fashion, discarding possible assignments for NOEs that are far apart in the preliminary structures. This is a manual process that takes many months. Distance geometry (DG) is a common method for generating a starting structure. The DG algorithm chooses exact distance restraints at random from the experimental range used as input and fits them to a mathematical solution while maintaining normal bond lengths in the molecule. Following distance geometry the structures contain many sub-optimal bond lengths and angles and therefore need to be corrected. This regularisation is done by simulated annealing (Nilges et al, 1988). Simulated annealing (SA) is a term used to describe a minimization / molecular dynamics schedule where the contributions due to van der Waals interactions, bond lengths, angles and peptide-bond planarity as well as the experimentally determined distance and dihedral angle restraints are included with different weightings. Each of these factors contributes an energy term to a target function that the system tries to minimize. Using this method it is possible to obtain different structures, all of which satisfy the experimental data. Therefore a large number of structures are calculated with a subset, or  Chapter 3-Structure Determination of CBD^j  60  ensemble, of low energy ones that satisfy the experimental restraints as well as having regular geometry chosen to represent the structure of the protein.  Experimental methods  Sample Preparation.  Samples of uniformly (-99%) N and C / N labelled 15  13  15  CBDNI  were produced by  expression of the plasmid pTugNln in E. coli JM101 cells, as described previously (Chapter 2; Johnson et al, 1996a). In the case of the C/ N labelled protein, the growth media contained 2 13  15  g/L of [ C6]-glucose and 1 g/L of C / N Isogro algal extract (Isotec, Inc.). Since the JM101 13  l3  l5  cells containing pTugNln grew poorly in minimal media,  CBDNI  non-randomly fractionally C l3  labelled at a level of 10% (Neri et al, 1989) was produced by initially growing bacteria in M9 media (Miller, 1972) supplemented with 1 g/L unlabelled Isogro. At an ODgoo °f LO, the cells were spun down, washed with M9 media, and resuspended in M9 media containing 0.3 g 99% [ C6]-glucose and 2.7 g unlabelled glucose as the sole carbon sources. IPTG (0.5 mM) was then 13  added to induce expression of the gene encoding C B D N I - The cells were grown at 30 °C for an additional 16 hours, and the secreted protein was purified as outlined previously (Chapter 2; Johnson et al, 1996a). The amides of Tyr, Leu, Asp/Asn, and Val were selectively [oc- N] 15  enriched using the protocol of Mcintosh and Dahlquist (1990).  CBDNI  samples with selectively  deuterated aromatic rings were obtained from a synthetic media containing 100 mg/L of LS|,£2 Cl,2Jl2-[ H5] 2  5  oi,2>£l,2-[ H4] 2  tryptophan and 100 mg/L of either L-5i,2,£l,2,C-[ H5] phenylalanine or 2  tyrosine (Cambridge Isotope Laboratories and Isotec Inc.) (Mcintosh et al, 1990;  Mcintosh & Dahlquist 1990). Samples of C B D N I for NMR analysis were exchanged into 50 mM sodium chloride, 50 mM potassium phosphate (pH* 5.9), 0.02% sodium azide, 10% D2O/90% H 0 using 2  ultrafiltration through a cellulose free membrane (Filtron). Samples in deuterated buffer were  Chapter 3-Structure Determination of CBDNI  61  obtained by twice lyophilizing the C B D N I , and redissolving in an equivalent amount of 99.9% D2O.  Typical protein concentrations were 2 mM as determined by £280 21370 M" cm" . With =  the exception of the selectively N labelled proteins, C B D N I samples contained up to a 40 fold 15  molar excess of cellotetraose (Seikagaku Corp.) to facilitate the observation of all the resonances of CBDNI  (Chapter 2). As will be discussed in chapter 4, the appearance of these signals is due to a  calcium contaminant in the cellotetraose. It was unknown at the time the work presented in this chapter was performed that C B D N I bound calcium.  NMR Spectroscopy.  NMR spectra were recorded on a Varian Unity 500 MHz spectrometer equipped with a triple resonance probe and a pulsed field gradient accessory. 'H chemical shifts were referenced to an internal standard of DSS at 0.00 ppm, C chemical shifts were referenced to an external DSS i3  standard at 0.00 ppm, and N was referenced to external 2.9 M NH C1 in 1 M HC1 at 24.93 15  15  4  ppm (Levy & Lichter, 1979). This latter reference yields N chemical shifts 1.6 ppm greater than l5  those obtained using liquid NH3 (Wishart et al, 1995). All spectra were collected at 35 °C and analysed using a combination of FELIX v2.30 (Biosym Technologies; San Diego, Ca.), NMRPipe (Delaglio et al, 1995), and PIPP (Garrett et al, 1991). Experiments with ' H detection were recorded using the enhanced sensitivity pulsed N  field gradient approach of Kay et al. (1992) and Muhandiram and Kay (1994). Selective water flip back pulse was incorporated to minimize the perturbation of the bulk water magnetization (Grzesiek & Bax, 1993; Zhang et al, 1994). Quadrature detection was accomplished using the States-TPPI method (Marion et al, 1989a). The initial delays in most of the indirectly detected dimensions were set to l/(2*sw), resulting in a 180° first order phase shift across the transformed spectrum and the inversion of aliased peaks (Bax et al, 1991). A summary of the data collection and processing parameters for the NMR experiments used to determine the structure of C B D N I is given in Table 3.1.  62  Chapter 3-Structure Determination ofCBD^i ffi  ffi  n> ntd  4^  ffi Cd  (-1  O  ui  to 3  n ffi  § O  0  n  5) GO  ^ ffi  *ffi  eg n  1°  8 3  g-ffi -5 U  2  to  3  s n  3 ffi  g  n  O  GO  i-l  o  ffi  3  ffi  >  CB  3  oo  O  O  O  CT  TD CO  00  00  00  H  g-ffi  o  O  n  3  n  C L  o o  CD  UJ  [O  U  —  'L 'L  O  Ui  on IO o I—  OO  O N  1—'  o  o o  Ui O  o Ul  "  h  T  H  Oi  Ul  O N  i—'  4^  ,  t  4^.  O N  Ul  o o  o  ^  ffi  ^2  s  4^-  U  'L 'LTi  4^  )£  W  «  'L tTi ^ffi  to  4  -  00  v  ,  t  O N  o o  o Ul o o o o  Ul  w  M  «  ^*ffi  O N  4^ Ul  Ul O  W  O N  4^ Ul  2 g E to  O N  Ul  00  H -  O N  4^ Ul  o Ui o o o o  to  ||  (O  ffi Q  to  Ul  l  ffi Q  ffi o  o  o  ^  ^  O N  O N  to  Ul  o o • o o  T  l  O to  un  00 o Ul  o o o  O N  Ul  00 o Ul  o o o  *T*  Oil  z2  o  B 3 >  4^  H-TD  O  OJ  4^  4^  ^  4^  O N  ||  to  O N  1—>  O N  Ul  4^  on O  o on o O  31  £1  4^-  Ui  ffi £L  o o  Q-l  CB , X TD CB  B CB 3 CB  o 4^  to  OJ  o  4^  vj  4^ t o 4^  U l  vl  N O  I—>  n. TD  o  vi  TD -  -  ON ^  ^  ^ 'Co 'NB  i-'  o  -a  « J N J  o,  N<"  O N  U l  p  .  O N  N O  to P  GO TD  a* N  GO  Go  oo cr a* o oo oo  O O  TD N<"  "—'  ^1 Ul  ^1  O N  O N  on -O  TD CD" -  TD -  c/i  GO  oo cr a* o o /oo -N oo s-' © m > 9 ..o 9 T3 O  P  O N  4^ t O ON —J  N O  O  TD  4^  4^.  CL  GO  to  4^  N O  p  , . GO GO N O  ^1  P  C L  o  CT  O N  4^  ^—-  w  ao  GO CD*  ^1  to  4^  y  4^ •^1  OJ O N  ^  OJ  -o  b  OJ  O  cr a* t o to  P aOO s-^.  TD  oo o  4^. t o  O N  O N  P  •~J o n  •^J  N O  C L  P TD  to on O N  I— tO  oo oo  to Ofl  ON  S' a.  g  9?  1  ^ a  O  3  p  TD O  O  S ^  ^  to •-  Cd  8-1  D z  CTQ  3  to  3  O  2.  TD O  ^  3  TD  Ui  era  id. O  -—-  to — t o on t o on O N 00 O N  t o i — ' on on t o i — ' O N 00 tO  M Ul  N  M  U l  O N  O0  tO  H  o cr  cn  2-  CD  i-l  O  3 CD  ^<  OJ  N O  °  N O N O  N O  N O N O  4^  N O N O  CB  era CD  -  oo^  ^  * V  N O  H-  0O O  P. 3  o  ^ ?  N O  N O  OO N O  o  to  4^  o  to  4^  O O  o  to  to  4^  4^  4^  to  o  to  4^  4^  O O H -  tO  N O  4^  O N  o 4^  00 p <;. 3 C_  o o T D y, c  4^  O  oo  3 q 3' qq  PJ  «  r  N O  00  N O  CT  N O N O  to  N O N O  to  o  3  . C_  O  oo  oo  3 W 3-  era N O N O  to  $  C L  CD  3  P5 N<  Cd P5  td P5 N O N O  to  P N<  X  N O N O  2  2  3  to  Cd  O.  CO PJ  C L  N O  p  cr  T D  N c CD  CB  2  U l  3  ^ tO  o  n  0Q  3  4^ t o  C L  CL  TD "  O  P3  C  K  3  70  ^° &  o N O  00  o  3 g oo" 3  1  c3 C L  CB  Chapter 3-Structure Determination of A,  O  2  O ol O o w  O n -< o  n W  o n i  o o  S- o  z  o b O 3 O cn O o>  n n  IO  a  P  Co  63  CBDNy  o  GO to  >  to LU  o  Z Q  K nS3  ON Ul O O  b  TD  ^~  1—*  4^ CO Ul ON 0 4^  4^ t o ON p s ] Ul NO  si ON 4^ -J  -<  ON Ul O O  4^ ON si  O  ^  «  4^  4^  O J  O J  b b 4^  4^  LU  iL 'L II_ ffi n s  Ul . . >—' W tO  to ^ 00  4^ Ul O 0 to Ul 0 O N  4^ O N  -J  4^ O J L ^  4^ 0 Ul 0  Ul J i H O to o  NO ON  4^ Ul O 0 to 0 Ul  -1^ 0 Ul 0  O N  O J  4^  , 1—*  O N  4^  Ul  o  s)  to ^3  s]  10  ^1  o_  Ul  M  to  to  00 00  a  O  c i-t  1-1 EU  P  ^ NO NO O J  f» NO  S  NO  4^  vo 4^  Ul  O N  1—I  4^  1—»  O N  I  O  3  3  o  O  T3  00  U J  TD  13  ON  O  to N  si  a—  o  ui to to to ui ui O  O N  O N  2,0  to to to Ul Ul Ul O N  O N  O N  2-  L<-^<  «  H Ul U I  O N  ON Ul  o o  ON  Ul ON  Ui  ,0 V  ON Ul O O  Ul 0  i-. . Ul  5* E»  NO NO O J  »-t  E»  p l  < CD  CD  P  ^< CD  NO NO O  1—.  tO  O J  Ul o  Ul O J  1—. 00 4^ Ul 0  4^ O N O ON O N • s i Ul so  0 Ul 0  "  g "^  w  ^  TD" TD  S-J  ^ T3 O  TD  cn  cn  NO NO O O  cn  O J  tO  O  ON Ul 0 0  1—1  t— Ul  4^ Ul 0  4i. t o ON P s ) Ul NO  Cu TD  ON O si TD  Ul  o  TD  o  cn  TD  cr NO  NO  o  o • 00 o  TD  O  O  to ^- to Ul to Ul  tO H-> Ul t o ON 00  LTT  O N  to  to  ui to ui  ON 00 ON  « p X  l-l  to  to  to  O N  O N  ui to ui O N  I—I  O cn OQ CD  CD  1—f P ,  Ea  IzZ  o  00  to  ui to ui 00  O N  o c  NJ  II II 1 Ln  -  LU  "DO  o  to to to Ul Ul Ul O N  o  so  cr ^ Q o o  TD  Ul ON  ||  Z fi  O J  P  II  •GO  tO  4^ >— ON 4^ Ui 0 Ul Ui O 0  o  10  'L Ti  ^  4^  b  Ul NO 4^  io ^  T3  >  00  'L L T i  to  z  >  tO  OJ  ON  Ul 4^ 0 0 Ul Ul 0 0 0  ^1  1  f  O J  £  O J  2  c tr.  CD  H  W  ^  4^ Ul  »o  z  o F  H O O oo  SL?  O  T3  c o sr a fo t i  00  3  K)  O J  CD  ^ Z ffi  cr S-°*  o  M  4^.  3 O  i  cr  o  Ul  to  'L ~ 'L 33 (-j K  si  s j  ^  OO O 00 Ul O N 4^ 0  T3 T3  — 13 T3 V3  cr "  -<  LU  ^ b£ OJ ^ ON on 0 0 Is) S  to  00  Q  2:  4^ O O O  Ul  "CO  II  Ul _ . t. o1 n r t "Tr *—'k .CO (0 0 + ^  O J  Ti_  II II  O  OJ tO  LU  II  |s  z a  o w o > o o  O W n > n o n  cn  S° ft) fJQ  CD'  K  CD  CD  X  P  rr,  P  2  NO vO NO NO O J 0  P O N  p  j  p  o  S  3  f  3  o CD  to  » NO g L NO "  ON OO ON  N CD  cn  p NO NO O J  CD'  7?  9P C d p xNO NO  to  Chapter 3-Structure Determination of BJ  4L-  T3  o  4^  t o II  i  *"l  OQ  I  2  CD l  n  O  ao  3l 8* 2.f 3' fS  ffi GO  q  GO ra CD S GO CD 3?  3  P  ^  fa  CD - i 3 T3  ca ? r-t-  T3 C L 3" o  O t  5 ffi o  o  5  S  O  on oo  3 3S 3 ' &3  OQ  ffi  r  w  fa  C  O 3  IO  i—i  o  OQ  °  3  O  3  ai  GO  22 ">3 3 o'  CD  o  II  0  II l l _  ffi n  ^7  OJ OJ  ll  'L ll  1—K  o7-  ll 'L Ti  ffi ffi n n ffi n n ffi ffi n  on oo to  O  on on ^ t o 2 oo  &  -  -  O N  ^  O N  O N  4^  on . ^ g  to °  !  to* ll__  ffi ffi  ffi ffi  ffi  to O J  i-—*  to  00  i o _k  rt-  to  II  O J  o  o  9  -•^ t o  on oo H -  IO  (O O  o  on on on O J  to  O N  O N  00  o on on O i— on  O o o  1  o  to o  oo oo on on o o o O N  O o o o  o o  o o  O N  on on O o o  o o  o  o  O N  O N  on on  o o  o o  O N  O N  on on  o o  o o  O N  4^  on O  o o  o o  O N  on O  o  f  32  oo  o  3  4^  O  N O N O  3  4^ O N  —J  4^  O J  oo  4^  b  O N  N O  4^  N O  4i. O O  b  4^  O J  to  O  O N  O N  O N  O N  N O  N O  N O  N O  IO  O  ^  &*  O N  ^ ,  - J  N O  O N  4^  4i»  4^  4^  O N  O N  O N  O N  ^1  4^O N  O J  y  O N  o  ^1  fa oo  's £ TS  3 3 O JL P  X 2. cr  vf TJ  II T i _  (—*  to ll_ Ti  (a  r—t-fa o r+i, " on p; CD  ffi  3 CD  |  3*  n ro ro  . ? .  ET"  O O  <o  tri  o  2  3  w H•  4^  on on O J  lh  O  •^j GO  3  GO  i  p"  3 i-H  O  GO  GO «  •co " D O  o  P  3- ° 2  P  !— 4-T  ffi  i 1  ° 2  m  3  O  on ffi  O  n  —]  on  -I  1 / 1  O  — OJ  CD  O  1  " '  s § g 2 J2. 3  l?>  Z  i  C L | r  a ^6 =  fc!  poo nD. on  3  OQ  o  O J  S.3  "3  CD  on  l §  n  ffi GO  o o  o n> 3  r>  O  O  OJ  3  n  CD O 3"  _.  CD O  CD  TJ  —  oo  Oq  OO  o  3 o  fi> 3  CD O 3"  CT"OQ hS*  Oq P  TJ  CD  3  Ol  oo  3 5  4  on O  o  n  o  < 3  3 §  Q -*  r  II  3 - CD CK)  O J  64  CBDNI  TJ  oq  ,  _3 3 TS T J  3  O  O  CTQ  O  TS  TJ  O  O  CD 3" CD  , T J "O ^  on ^  o  —  O Q  O ,  O  00  C L  O  oo  TJ T J  O TJ  o  O  9 ^ 9  o  v-S o o  T3 O  W  !=  IO  n  -  IO  to  T3 T3 T ) O  oT  C L  oo  2- 2-  ^-~> GO N O cr o _ T3  O  3  o ^  o  00  o  TJ  *<*  C L  0Q  o  O  —  TJ  N<"  GO  ca •Q  0 J3  5.  3 P  CD  GO _  C L  <—t- /*—\  2 c-t-  o> -n  o  o  4^  4^  KJ t o  o  o  to t o 4*.  4^  N H on t o t o O N  O O  00  ^  ol  O N  O N  to 00  tO  H -  IO  on i o on O N oo O N  io to  o  o  4^  4^  00  oo  to to  o  4>-  O O  o  4^ 00  o  o  to  IO 4^  4^  3 ' CD 3 CD CD <*  ffi 2"! N i—  3>  CD  3-ai  ST Si 3n  < 3  8*  oo  1 3'  OQ  l-t  N CD CD  CD - •  ca  O  CD  CD  S CD  Q  a  T3 0O  O 03  CD  o c CD  P5  p  3 0Q CD •-I  HC-  P  N  CD  3  5» N O  8« S=  N O N O  N O  O J  O O  cr  N O  to  N O N O  N O N O  4^  N O  00  N O  N O  O O  N O  O O  oo  Chapter 3-Structure Determination of CBDN1  65  Amide Hydrogen Exchange.  Amide hydrogen exchange rates were determined by recording a series of sensitivityenhanced gradient lH- N HSQC spectra at 10, 29, 59, 95, 131, 266, 451, 688, 962, 5377, 15  10790, and 25211 min after dissolving lyophilized C B D N I in D2O. To minimize the quantity of residual H2O, the uniformly N-labelled protein was lyophilized twice, being resuspended in D2O l5  after the first freeze-drying step. The buffer concentration and pH were held constant by maintaining the sample volume.  Structure Calculations.  All structure calculations were performed using X-PLOR 3.1 (Briinger, 1992). Distance restraints from NOE experiments were tabulated with in-house programmes available from http://otter.biochem.ubc.ca/www/nmrtools.html.  Initially, a preliminary fold for C B D N I was  calculated following the DGSA protocol using NOE restraints involved in p-strand pairing, unambiguous NOE restraints identified from 3 and 4D heteronuclear NOESY spectra, and dihedral angle restraints. This preliminary structure was used in a reiterative fashion to assign additional NOE interactions. A total of 1705 NOE derived distance restraints were used in the final calculation of an ensemble of 60 structures. This data set was comprised of 711 nontrivial intraresidue, 411 sequential, 90 short-range (1 < I i-j I < 4), and 463 long range (I i-j I > 4) distance restraints. Interproton distances were assigned to three strengths following a square-well potential energy function: weak 1.8 - 5.0 A, medium 1.8 - 3.5 A, and strong 1.8 - 2.9 A. A correction of 0.5 A was added to the upper bounds of restraints involving methyls (Clore et al, 1987). The distance ranges for the N NOESY-HSQC were calibrated using the intensities of the 'HNj-'H'V 15  1 which should be strong (-2.2 A) in regions of P-strand conformation, and cross strand 'H ia  'H j NOEs, which in anti-parallel p-sheets should have a medium intensity (-3.2 A). For the N  simultaneous 3D C/ N NOESY-HSQC spectra (Pascal etal, 1994), these sequential NOEs, as 13  15  well as the strong cross-strand 'H^-tH ^ NOEs (-2.3 A), were used to calibrate the intensity 0  Chapter 3-Structure Determination  ofCBDNI  66  ranges. For the 2D NOESY spectra involving Phe and Tyr aromatic rings, the 'H^- 'H NOE was e  used for calibration. In the case of the 4D C- C NOESY, a 150 msec mixing time was used, 13  13  resulting in extensive spin diffusion. As a result of this complication, all NOEs extracted from this experiment were classified as weak. In addition, 88 hydrogen bond restraints (44 hydrogen bonds), 56 (|>-angle restraints, 88 \|/angle restraints, 42 %1-angle restraints, and 9 %2-angle restraints were included in structure calculations. Hydrogen bonds, deduced from patterns of amide hydrogen exchange and NOE o  interactions involving main chain protons, were restrained to 2.5 - 3.5 A between O and N atoms, and 1.5 - 2.5 A between H and O atoms. The (j) torsion angles were restrained to 60° ± 30° for J N  < 5.5 Hz, -120° + 30° for 8 Hz < J < 9 Hz, and -140° ± 20° for J > 9 Hz. These J N-Hcx 3  H  couplings were determined from a !H- N HMQC-J spectrum (Kay & Bax, 1990) using software 15  provided by Lewis Kay, as described by Foreman-Kay et al. (1990). The \|/ angles were restrained to 120° ± 100° or -30° ± 110° based on the ratio of H«i.i-'H i and 'Haj-'HN; NOE l  N  intensities (Gagne et al, 1994). The %1 and %2 angles were restrained to ± 30° from their assigned rotamer values.  Results  For this current study, all spectra of C B D N I were obtained in the presence of a 40 fold excess of cellotetraose (typically 80 mM) in order to obtain complete resonance assignments, as well as to investigate the effects of ligand binding. At 35 °C, the free and cellotetraose-bound forms of C B D N I are in fast exchange on the NMR timescale, resulting in the observation of population weighted average chemical shifts. With the concentration of cellotetraose used, C B D N I is fully saturated with this ligand (K = 4200 ± 720 M ; Chapter 2; Johnson et al, 1996a). _1  a  However, the cellotetraose was not specifically included in the structure calculations due to the  67  Chapter 3-Structure Determination of CBDN j  degeneracy of its NMR spectrum, which prevented the unambiguous assignment of intermolecular NOEs.  Main Chain Resonance Assignments.  The ' H and N resonances from the main chain amides of C B D N I were assigned using N  15  the combination of HNCACB (Wittekind & Mueller, 1993) and CBCA(CO)NH (Grzesiek & Bax, 1992) experiments to correlate the NH; and C°y CPj, C°Vi/ CPi_i and NHi and C ;_ 15  13  13  13  13  l5  13  a  |/ CPj.| resonances, respectively (figure 3.3). Selectively [a- N]-labelled Tyr, Leu, Val, and 13  15  Asp/Asn C B D N I provided amino acid-specific starting points for this assignment procedure. A modified version of the CBCACO(CA)HA experiment that detects only the H« and C / CP 1  13  a  l3  resonances of residues preceding prolines was also extremely useful for providing unambiguous reference points (Olejniczak & Fesik, 1994; Lewis Kay, pers. comm.). The assignments of the resonances from the remaining backbone residues were obtained using the HNCO (Ikura et al, 1990; Muhandiram & Kay, 1994) and CBCACO(CA)HA experiments (Kay, 1993). Figure 3.5 shows the assigned H- N HSQC spectrum of C B D N I - With the exception of the 13 amides not ]  15  detected in the absence of added cellotetraose, near complete assignments of the resonances from the main chain ' H and N nuclei of uncomplexed C B D N I were also obtained by following the N  15  progressive titration of the protein with cellotetraose (chapter 2; Johnson et al, 1996a).  Aliphatic Side-Chain Resonance Assignments.  Virtually complete side-chain assignments of the resonances from ^H, C, and N nuclei 13  15  in C B D N I were obtained using a combination of N TOCSY-HSQC (Marion et al, 1989c), 15  H(CCO)NH-TOCSY, (H)C(CO)NH-TOCSY (Logan etal, 1992; Montelione etal, 1992; Grzesiek etal, 1993a), HCCH-TOCSY (Bax etal, 1990; Kay etal, 1993), and HCCH-COSY (Kay et al, 1990; Ikura et al, 1991) experiments (figure 3.4). The HCCH-COSY and HCCH-  •er 3-Structure Determination of  CBDNI  o  G15  ©  o  ®G82  G30® @  T29,  G  S133  T138  G57<g> VI70  "  ©QG23  o  0  .7® D143  ®  I  ®  V  4  S  „  ' W16 indole ®  2 ®  10. 0  •o CM  •*4 •  ID  L 9 S  r»L^-a„  El04 <§> R63 ®  a a  o  g  V48® OJ28 ©  E  Q101.7N50S  80: „ D120  D142 I Q124 . ®T107 • I Y43  A  <g) V34  E55, Y60.D117  O - O  A68.  ro  ® V102  9.0  HJ  7. 0  8.0  (ppm)  O E  8. 7  8.4  8. I  HI (ppm)  Figure 3.5. Sensitivity-enhanced gradient 1H-15N HSQC spectrum of  at 35° C and pH* 5.9 showing the assignments of the resonances from backbone amide and tryptophan indole l5T>jelH groups, and those of the observable signals from the side-chain 15N§H2 and !5N£H2 of asparagine and glutamine, respectively. The aliased peak from Thr8 is denoted by an asterisk. The spectrum in (B) is the expansion of the boxed central region of (A).  CBDNI  Chapter 3-Structure Determination ofCBD^j  69  TOCSY experiments were recorded with the protein in D2O buffer. These assignments are reported in the appendix. Initial inspection of the iH-^N HSQC spectrum of C B D N I revealed that four of seven expected peaks from the side-chain !5NH2  r e s o n a n c e s  were either missing, very weak or  degenerate. The strong resonances from the N H 2 groups of Gln42, Gln80, and GlnlOl were 15  e2  readily assigned from the HNCACB experiment and confirmed using the N NOESY-HSQC l5  spectrum. Therefore, the unidentified side-chain amides resonances were those of Asn50, Asn81, Glnl24, and Glnl28. All four of these residues lie within the oligosaccharide-binding cleft of CBDNI  and could possibly play an important role in interacting with sugar. As a result of their  potential importance, a special effort was made to assign them. Spectra from this process are shown in figure 3.6. First, a modified version of the CBCACO(CA)HA experiment that links the side-chain carbonyls to the previously assigned C / C P and 1 3  ^72.73  a  1 3  1HP -P 2  3  of Asp/Asn and  l 3  CfV  L 3  CY  and  of Glu/Gln was recorded (Kay, 1993). Next an HNCO experiment, tuned using a total  delay of 1/(4JNH) during the first reverse INEPT sequence to favor A X 2 spin systems (Schleucher et al, 1994), was used to correlate the resonances of these side-chain carbonyls to the corresponding  1 5  NH2  groups. Finally, an HSQC spectrum, with similar delays to enhance A X 2  spin systems, was recorded overnight to help identify the yielded assignments for all seven side-chain  1 5  NH2  l 5  NH2  resonances. This strategy  groups, as well as those of the carbonyls of  almost all Asp/Asn and Glu/Gln residues (Appendix). Both side-chain amide H& protons as well as the N^ of Asn50 are completely l  2  15  2  degenerate with those of GlnlOl. Fortunately, the difference in the chemical shift of the side-chain carbonyl resonances of these two residues allowed them to be distinguished (figure 3.6, panel C). The chemical shift of the C7 carbonyl of Asn50 is unusually upfield shifted to 173.5 ppm. In 13  the absence of structural information for the bound cellotetraose, the reason for this perturbed shift this is not immediately apparent. The side-chain  1 5  NH2  groups of Asn81, Gin 124, and Gin 128,  all of which are likely to participate in hydrogen bonding to the cellotetraose, exhibit very weak  Chapter 3-Structure Determination ofCBD^]  B  H Q42eA  0i  Qi28e  iiy  y. (15  ° 13  1  H  (ppm)  £ 3  ||) 1  80t  C=179.6 ppm 1  H  (ppm)  E a a u  Figure 3.6 Spectra used to assign side-chain NH2 resonances. (A) The side-chain carbonyl  resonances are assigned. This panel shows a plane from the modified version of the CBCACO(CA)HA experiment used to connect the side-chain carbonyls to the previously assigned 13Ccc/13Cp and lHp2,p3 of Asp/Asn and 13Cp/I3Cyand lHy2,y3 of GIu/GIn. Peaks resulting from Q80 are marked by an asterisk to indicate these signals lie on a different plane, and overlap onto the one shown in this figure. (B) Using the assignments of the side-chain carbonyls the NH2 groups are assigned from an HNCO experiment, tuned using a total delay of 1/(4JNH) during the first reverse INEPT sequence to favor AX2 spin systems. (C) Nitrogen plane from the modified HNCO experiment shows the NH2 resonances of Q101 and N50 are overlapped. They are assigned on the basis of their different side-chain carbonyl chemical shifts. Peaks marked with an asterisk result from resonances that lie on another plane.  Chapter 3-Structure Determination of CBDNy  71  resonances. Only one 'H^ for the amide of Asn81 was found. It is not known if the resonances 2  from the two amide protons are degenerate or if the second is very weak. It is likely that the anomalous behavior of these side-chain amide resonances results from conformational exchange broadening, possibly due to unfavorable kinetics of cellotetraose association and dissociation from CBDNI  or due to mobility of the sugar ligand within the binding cleft. In support of this  suggestion, I note that the side-chain of Gln80, which lies on a p-strand that is part of the binding face but does not point into the binding cleft, has a strong N H 2 signal. In contrast, the sidel 5  E  chain of the adjacent Asn81, which points into the binding cleft, has a very weak N H 2 signal. 1 5  5  Aromatic Side-Chain Assignments.  Due to spectral overlap, the use of protein samples with selectively deuterated aromatic rings was an extremely useful tactic for the assignment of these side-chain spin systems. The H L  resonances from the aromatic rings were identified from homonuclear DQF-COSY, 70 msec mixing time TOCSY, and 150 msec mixing time NOESY spectra. These were recorded with samples of 1.9 mM unlabelled, 2.0 mM ([ H ]-Tyr and [H]-Trp)-labelled, and 1.9 mM ([ H ]2  2  4  2  5  5  Phe and [Fi5]-Trp)-labelled CBDNI, each in the presence of 80 mM cellotetraose. The unlabelled 2  CBDNI  sample was recorded in both  H2O  and  buffers, while the two deuterated CBDNI  D2O  samples were recorded solely in D 2 O buffer. The H and 'H of the aromatic ring spin systems 1  8  e  were then directly connected to the previously assigned CP nuclei using the (Hp)Cp(CyC5)H8 13  and (HP)Cp(CyC8C£)H£ experiments (Yamakazi et ai, 1993). Finally the assignment of the C 13  resonances of the aromatic rings were obtained from 'H- C HSQC and CT-HSQC spectra 13  acquired using the C/ N-labelled CBDNI (Santoro & King, 1992; Vuister & Bax, 1992). The 13  15  CT-HSQC was particularly useful for distinguishing the C of tryptophan residues due to their 1 3  5 1  inverted signals relative to those of other C nuclei with an even number of neighboring carbons. 13  C h a p t e r  3 - S t r u c t u r e  S t e r e o s p e c i f i c  D e t e r m i n a t i o n  A s s i g n m e n t s  a  n  o  f  72  CBDNI  d S i d e - C h a i n  T o r s i o n  A n g l e  R e s t r a i n t s .  Stereospecific assignment of 23 of the 67 residues in C B D N I with prochiral 'HP protons was obtained from a conservative analysis of the HNHB (Archer et al, 1991), 40 msec and 72 msec mixing time N TOCSY-HSQC (Marion etal, 1989c), and 50 msec C / N NOESY15  HSQC (Pascal  e t a l . ,  13  15  1994) spectra of C B D N I - These assignments and the corresponding %1  restraints were determined based on the staggered rotamer model, as outlined by Powers et al. (1993). Near complete stereospecific assignments of the diastereotopic methyls of valine and leucine residues were obtained using the elegant approach of Neri et al. (1989). As demonstrated by these authors, in high resolution HSQC spectra of 10% non-randomly fractionally C enriched 13  proteins, the Pro-R (Leu , ValY) methyls are doublets due to C - C couplings, while the Pro51  1  l3  13  S (Leu , VatY) methyls are singlets. An additional level of discrimination is provided by the use 82  2  of a constant time 'H- C HSQC experiment with a total evolution delay of 1/Jcc - 1/34 H I3  z  (Lewis Kay, pers. comm.). In the resulting CT-HSQC spectrum, the signals due to ValY, Leu , 1  51  and IleY and the signals due to ValY and Leu all appear as singlets but with opposite sign 2  2  52  (Figure 3.7). Also, the peaks due to the ThrY and He methyls, which are apparent triplets in the 2  l3  51  C HSQC due to approximately equal levels of C - C and C- C labeling (Szyperski et al, 13  13  13  12  1992), are nulled in the constant time C HSQC. Together, these two factors help simplify l 3  crowded regions of the spectrum, allowing previously overlapped resonances to be assigned. %1 restraints for 9 of the 14 valines, 6 of the 20 threonines, and all 3 isoleucines were determined on the basis of  3  JNCy  an  d  3  Jc'Cy  coupling constants (Tables 3.2 and 3.3) and  intraresidue NOE interactions, according to the staggered rotamer model. The coupling constants were determined quantitatively from C-{ N} and C-{ C'} spin echo difference CT-HSQC 13  15  13  13  spectra (Grzesiek et al, 1993b; Vuister et al, 1993). Peak volumes were used in the calculation of the coupling constants using programmes I wrote, as described by Grzesiek et al. (1993b) and Vuister et al. (1993). The %1 analysis of valine was aided by the previously determined stereospecific assignments, particularly in the case of resonances obscured by spectral overlap.  Chapter 3-Structure Determination of  1.4  73  CBDNI  1.2 HI  7T~2 HI  1.0  0.8  0.6  LO  oTs  0^6  (ppm)  (ppm)  Figure 3.7. A portion of the (A) constant time ' H - ^ C HSQC and (B) H- C HSQC spectra of 10% fractionally C enriched C B D , . Due to the biosynthetic labeling pathways, the Pro-R (Leu^ , V a F ) methyls are doublets and Pro-S (Leu** , VafY ) methyls are singlets in the regular HSQC (B). In the CT-HSQC (A), all peaks are singlets but the Pro-R (Leu^ , VafY ) methyls (filled) have the opposite sign compared to the Pro-S (Leu^ , ValT ) methyls (open). The use of constant time results in further simplification of the spectra by the elimination of Thr^ and Ile^ methyls (not shown in the spectral window) that are apparent triplets in the non-constant time HSQC. For clarity, assignments are indicated only for selected peaks. L  1 3  L3  N  1  2  1  1  2  2  1  2  2  1  Chapter 3-Structure Determination of  74 CBDNI  Table 3.2: Coupling Constants' and %l Assignments^ for Valine Residues in 2  Residue  jl 3  JNCY  (Hz)  CBDNI-  72 3  Jc'Cy  3  (Hz)  JNCY  3  (Hz) 0.7 0.6 0.6 0.3 0.7  Xl  JCCY  (Hz) 1.7 0.8 3.4 3.0 2.9  c c V17 d V34 0.4 3.4 -60° c c 180° V36 c c 180° V45 180° V47 1.5 1.1 0.7 2.7 c c -60° V48 e e e e V52 e d 0.2 1.8 1.5 0.5 V72 f 3.3 0.8 -60° V74 1.2 0.7 180° 2.1 0.7 V78 1.7 d c 0.9 0.9 c V88 0.5 3.6 180° V102 0.7 1.9 c c 180° V144 1.1 1.7 0.8 2.8 1.6 2.7 V150 d Coupling constants were determined from a C-{ N} and C-{ C'} spin echo CTHSQC spectra (Grzesiek et al, 1993b; Vuister et al, 1993). The coupling constants were corrected for a systematic underestimate in the determined coupling constant value inherent in the method by multiplication of the measured value by 1.06 ( Jc'Cy) d 1-08 ( JNCy) (Damberger et al., 1994). %1 angles were determined on the basis of Jc'Cy d JNCy values, as well as intraresidue NOEs determined with a 50 msec mixing time C/ N NOESY-HSQC and a 125 msec N NOESY-HSQC. not determined due to spectral overlap. not assigned due to conflicting evidence for the presence of a single %1 angle amongst a  13  15  13  13  a n  3  b  an  3  l3  3  3  J5  15  c  d  values and/or JNCy values and/or intraresidue NOEs. not determined due to degeneracy of the two V52 methyls in both C and 'H dimensions. i coupling constant value too small to accurately measure. 3  e  Jc'Cy  3  13  Chapter 3-Structure Determination  75  ofCBD^j  Table 3.3: Coupling Constants and %lAssignments^ for the He and Thr Residues of C B D N I • 0  Residue T8 T21 T27 T29 T58 T59 T61 T65 T67 T70  3  JNCy2  3  (Hz) 0.7 0.9  Jc'Cy2  1.0 2.0  (Hz) 2.9 2.4 3.3 2.1 0.7  d  d  e  1.7  c  e  d  d  e  d  d  e  c  1.8  e  c  Residue  Xi  60° 60°  3  T73 T87 T91 T96 T103 T105 T107 Till T115 T138  e e e  JNCy2  3  Jc'Cy2  Xi  (Hz)  (Hz)  d  d  e  0.7 0.3 c  3.2 2.4 2.4  60° 60° 60°  d  d  e  d  d  e  d  d  e  1.6 1.0  0.6 2.6  -60°  d  d  e  e  2.2 0.6 -60° 3.3 60° 1125 14 0.8 -60° 154 2.0 0.8 Coupling constants were determined from a C-{ N} and C-{ C'} spin echo CT-HSQC spectra (Grzesiek et al, 1993b; Vuister et al, 1993). The coupling constants were corrected for a systematic underestimate in the determined coupling constant value inherent in the method by multiplication of the measured value by 1.06 ( Jc'Cy) d 1 08 ( JNCy) (Damberger et al,  a  13  15  13  an  3  13  3  1994). % 1 angles were determined on the basis of Jc'Cy d J N C y > l l intraresidue NOEs using a 50 msec mixing time C/ N NOESY-HSQC and a 125 msec N NOESYHSQC. coupling constant value too small to accurately measure. not determined due to spectral overlap. not assigned due to conflicting evidence for the presence of a single % 1 angle amongst Jc'Cy values and/or J N C Y v lues and/or intraresidue NOEs. b  3  13  an  3  v a m e s  a s w e  15  a s  15  c  d  e  3  3  a  Qualitative analysis of the 3D C- C long range correlation experiment (Bax et al, 1992) 13  13  provided %2 restraints for 7 of the 12 leucine residues and 2 of the 3 isoleucine residues. As shown in Figure 3.8, a strong l H - C cross peak was observed for Ile4 and Ilel25, reflecting 5l  l3  a  a large J c 8 l C o c coupling indicative of a trans conformation ( %2=180°). i the case of Ile54, there 3  n  Chapter 3-Structure Determination of  76  CBDNy  1125  154 Residue  Figure 3.8. Strip plot of a portion of the long range C- C coupling constant experiment recorded for C J 3 D (Bax et al., 1992). Shown are 0)l( C)-(o3( H) strips for Ile4 and Ilel25, which exhibit a strong H - C cross peaks relative to the H - C autocorrelation peaks. This reflects large Jc5iCoc couplings, indicative of trans (180°) %2 dihedral angles. In contrast, no H - C cross peak is observed for Ile54 (a box indicates the position at which it is expected). Also, only a weak J H ^ - ^ C J T cross peak is observed. Together, this reflects both small Jc8lCy2 d Jc8lCa coupling constants, indicative of a gauche conformation (%2 = -60°). The C of Ile54 is marked with an asterisk indicating it is aliased in the C dimension shown in this plot. Cross peaks and auto-correlation peaks are shown without distinction of their opposite signs. Vi  Vi  13  1  Nl  1  5 l  1 3  a  J  3  L  5  L  1  3  A  2  3  +  1 3  a n  3  1 3  S l  5  L  L  3  §  1  77  Chapter 3-Structure Determination ofCBD^j  was no l H  5 L  -  1 3  C  A  peak and a weak ' H ^ - B C T peak. This indicates that both the Jc8lCa d 3  2  an  the JcSlCy2 couplings are small, suggesting the side-chain adopts a gauche conformation (%2 = 3  +  -60°). In the absence of positive evidence for this conformation, no %2 torsion restraint was included for Ile54 during the structure calculations. However, in the final ensemble of structures, this %2 angle was indeed well defined at -65° with a chi2 angular order parameter of 0.99 (S(%2); Hyberts etal., 1992).  Secondary Structure Determination.  Initial studies using CD and FT-IR spectroscopy indicated that C B D N I is composed of pstrands and devoid of helices (chapter 2). This global analysis was confirmed when the regular secondary structural elements of C B D N I were determined based on patterns of amide hydrogen exchange rates, sequential and cross strand NOEs, 13  3  JHN-HOC  coupling constants, and  C , and H chemical shifts. The information defining the p-strands in 1  a  C B D N I  a  13  0  Wishart & Sykes, 1994), negative C and 13  'HVHNi+i NOEs, cross-strand  a  ' H V H O J ,  1 3  c o i i ;  se  C, a  13  CP,  is summarized  in Figure 3.9. In general, residues in P-strand conformations are characterized by Hz, positive 'H and CP secondary chemical shifts (5 b rved - Srandom  13  3  JHN-HCX  ^8  Wishart et al., 1992;  C secondary chemical shifts, strong sequential  ' H V H N J ,  ' H N ^ H N J  NOEs, and protection from  exchange due to cross-strand hydrogen bonding (Wiithrich, 1986). Figures 3.10 summarizes the interresidue NOEs observed in the NOESY spectra of  C B D N I  (Figure 3.11), as well as the hydrogen bonds used in the structure calculations. Based on the patterns of cross-strand ^"j-'H^, H j-!H j and 'H^-'HNj N O E S , l  N  N  C B D  N  J  contains two p-  sheets, denoted A and B, each composed of five anti-parallel P-strands. Defining the exact boundaries of the regular elements of secondary structure in a protein is often difficult. For example, in the case of C B D N I , there is some ambiguity in identifying the ends of strands B3 and B4. As seen in Figure 3.10, interstrand NOEs indicate that strand B3 could be defined to start at Ile54. Correspondingly strand B4 would end at Tyrll2. The lack of cross strand 'H'Y'H"] NOEs is a result of extensive degeneracy of the H resonances in this region. However, the L  a  Chapter 3-Structure Determination of CBDN  78  ]  B2  BI.  1 10 20 30 A S P I G E G T F D D G P E G W V A Y GT D G P L D TS T G A L C V  NH Exchange 3j  A2  0 o o oooooo 9  HN-Ha  6 S 9  toooo •  ooo« 6  S  S  8  7  OO O I  S  9S  5  40 50 AVPAGSAQYGVGVVLNG  torn o o o o  8 79 9 8 8  ooo»«»©o«  S S 8 7  8  +2.0  99  •L  A Ha (ppm) 1  -2.0 I +5.0—r Ai3Ca (ppm)  J  5.0 h6.0 A13C|J(ppm) 6.0  • ll •  _1  8.0—1 A13C'  • L  (ppm) 9.0 —I  A4  A3  VAIE NH Exchange 3j  • 9  HN-Ha  60 70 80 90 100 EG T T Y T L R Y T A T A S T D VT VR AL VG QN GAP Y G T V L D T S P A L T S E P R Q V  « O0«O • • • • • • • » 0  79  S  88  999S9999  76  999  O0«««O 0 * 0 • • 0 * 0  9 7  S7S  S  —oo  9 1 0 681D10  •o*  9 7S8  989  +1.5 _ A Ha (ppm)  Li  —  1  •!  4.U—f  A C a (ppm) 13  +4.0  6.0_[ +8.0—. A C(i(ppm) 13  -4.0  J  A C (ppm) 1 3  -8.0—1  B4  A5  110 TETFTASATY NH Exchange 3j  HN-Ha  1  A C a (ppm) 13  0*0  O«0t0t  0 9 OOOO  « M * M « « O O O o o o « M « * * « o « * e  ooooo  999  8 S7 7 9 9  69  S 9 1 0 89 8  8 7 87  lilu  A Ha (ppm)  120 130 140 150 P A T PA A D D PE GQ I A F Q L G G F S A D A W T L C L D D V A L D S E V E L  9 S 8 8  m l . >_  77  8  6 7 9 9 9 9 7 6 9  77  8  ••11.LL  • _•!  +4.0—t J  •  •_-  -8.0—1 +6.0  A cp(ppm) . 13  2 0  2 llii—ll.i-i-l^.  +2.0 _ j A C (ppm) 1 3  -4  •  ""• - - i n -  •• i - h l l l . i l i l i •-• I |IU|| ~||I B  ••  79  Chapter 3-Structure Determination ofCBD^j  Figure 3.9. Summary of amide hydrogen exchange rates, J H N - H a coupling constants, and ' H , a  3  1 3  C , a  1 3  C P , and C chemical shifts used to deduce the (3-strand secondary structure of CBDNI-  The locations of the 5 p-strands of sheet A are indicated by open arrows, and those of the 5 pstrands of sheet B are indicated by solid arrows, (i) Hydrogen exchange: Filled circles indicate residues with slow hydrogen-deuterium exchange kinetics (t 1/2 > 1000 min), half-filled circles indicate those with intermediate hydrogen-deuterium exchange kinetics (10 min < ti/2 < 1000 min), and open circles indicate those with fast hydrogen-deuterium exchange kinetics (t 1/2 < 10 min) at 35 °C and pH* 5.90. (ii) J H N - H O C measured values are reported in Hz. S denotes couplings that 3  :  are too small to be determined reliably using the HMQC-J experiment, (iii) Main chain ' H , a  13  CP, and  1 3  C , a  C chemical shifts are plotted as the difference from the random coil values (8 bserved 0  ^random coil)-  change in  1 3  1 3  Residues in P-strands have a positive change in ' H and CP shift, and a negative a  C and a  1 3  13  C chemical shift (Wishart etal., 1992; Wishart & Sykes, 1994).  Chapter 3-Structure Determination of  CBDNI  F i g u r e 3.10. Alignment of the p-strands to form sheets A and B present in C B D . The NOEs used to deduce these alignments are shown by arrows. Dotted lines indicate the hydrogen bonds included in the structure calculations. Boxed amide hydrogens have slow hydrogen-deuterium exchange kinetics (filled circles in figure 3.9), while circled amide hydrogens have intermediate hydrogen-deuterium exchange kinetics (half-filled circles in figure 3.9). The positions of p-bulges are indicated by jagged lines. N1  Chapter 3-Structure Determination of CBD  0 1 2 F127HP  2  [-3  F127Hl»  G46H  a  G130H"  4  GjJgH" M25H W""" ''i 111>  G46H"  a  ~ F127H  a  Q128H"  G46H"  ^5  "^s ""• *29H~ K  . ,„ A76H l l | | B 8 s  s  L 1  a  A126H  a  V48H  a  6 , F127H  F127H  V47H  6  V48H  5  R75H  ffi  7  N  N  8  N  9 L77H  N  FT 1  T  1  1  10  1  A126 F127 R128 L129 G130  Residue Figure 3.11. G)l( H)-co3( H) strip plots of a portion of the N H S Q C - N O E S Y spectrum of C B D , recorded with a mixing time of 125 msec. Selected N O E interactions are labelled, and solid lines connect the strong NOEs of an H to the H of the previous residue, indicative of an extended p-strand conformation. This is an example of the data obtained to derive distance restraints for the structure generating process using the hybrid distance geometry/simulated annealing protocols in XPLOR 3.1. 1  l  1 5  N ]  1  ]  a  N  82  Chapter 3-Structure Determination of CBDN1  classification of these residues as having a (3-sheet conformation is not supported by the secondary chemical shifts of these residues (Figure 3.9). From the tertiary structure of C B D N I , it is found that the polypeptide backbone in these regions turn sharply, linking sheets A and B. Therefore, these sequences were not defined as part of the p-strands in C B D N I • Two strands of C B D N I  a r e  broken by (3-bulges. One bulge, classified by Promotif  (Hutchinson & Thornton, 1996) as being classical (Chan et al, 1993), begins at residue Thr87, following which Val88 and Leu89 both lie in the hydrophobic core of the protein. The second bulge starts at Leul41, after which Aspl42 and Aspl43 both have their side-chains on the exterior of the protein. The presence of both of these bulges is evident from the secondary chemical shifts shown in Figure 3.9. This emphasizes the potential wealth of structural information contained in NMR chemical shifts. From the topological arrangement of the p-strands of C B D N I , it is evident that this protein adopts a jelly-roll p-sandwich structure (Figure 3.12; Brandon & Tooze, 1991). The jelly-roll comprises strands A2 to A5 and B2 to B5, with the two short strands Al and BI appended along one side of this core motif. Strands Al / BI and A4 / B4 are not connected together by hydrogen bonding, thus defining the structure as a P-sandwich as opposed to a continuous p-barrel. This is evident from the protection patterns of the backbone amide *H protons in the hydrogen-deuterium N  exchange experiments (Figures 3.8 & 3.9). The outer edges of the p-sheets do not show protection for the backbone amide !H protons. N  Tertiary Structure.  A total of 1988 distance, hydrogen bond, and dihedral restraints were used to calculate 60 structures following the hybrid distance geometry/simulated annealing protocol (Nilges et al, 1988) with X-PLOR 3.1 (Brunger, 1992). The 25 structures with the lowest total energy and fewest NOE violations were selected for comparison. None of these had NOE violations greater than 0.4 A, and, except for the two %1 restraints involving Cys33 and Cysl40, none of the 25  Chapter 3-Structure Determination of  CBDNI  C  N Figure 3.12. Schematic diagram showing the jelly-roll P-sandwich topology of CBD |. Sheets A and B are indicated by open and solid arrows, respectively, N  and the position of the disulphide between Cys33 and Cysl40 is indicated. Strands A2-A5 and B2-B5 comprise the jelly-roll motif, with the two short strands Al and BI appended along one side. The global structure of C B D N I be c a n  envisioned by folding the figure such that sheet B lies below sheet A, and that sheet A is concave. The lengths of the strands and loops are not drawn to scale.  Chapter 3-Structure Determination ofCBD^j  84  structures had dihedral violations greater than 4.3°. Statistics for the 25 accepted structures are listed in Table 3.4. The superposition of the final ensemble of structures calculated for  CBDN 1  is shown in  Figure 3.13. The structural ensemble is clearly consistent with the jelly-roll p-sandwich topology deduced at the level of secondary structure analysis. The backbone conformation of the strands in each of the sheets is well determined, having an rmsd of 0.44 ± 0.05 A with respect to the average structure. Apart from the N- and C-termini, the regions which have the highest rms deviation from the average structure are between residues 20-30, which contains the short p-strand B1 involving residues 25-27, and in the loops between residues 37-44, 81-85, and 113-121. All these stretches contain no, or few, long distance restraints (Figure 3.14A). As shown in Figures 3.14B and 3.14C,  these stretches are also regions where the angular order parameters S(() and S\j/ are the  lowest, indicative of local disorder (Hyberts et al, 1992). The stereochemical quality of the backbone coordinates for the ensemble of 25 structures was checked using the programmes Procheck and Procheck-NMR (Morris et al, 1992; Laskowski et al, 1993). For this ensemble, 98% of the residues lie in the allowed regions of the Ramachandran plot (figure 3.15). The few residues with main chain dihedral angles that often fall outside the allowed regions, namely Glul4, Ala83, and Tyr85, are all found in the parts of  CBDNI  that exhibit high rms deviations, low angular order parameter, and the lack of regular secondary structure. As shown in Figure 3.16, the side-chains that make up the hydrophobic core of  CBDNI  are  well defined structurally. This is reflected by both low rms deviations and high angular order parameters S(%1) (Figure 3.14D, E). Of the 33 side-chains that comprise the hydrophobic core of CBDNI  , 32 have values of S(%1) greater than 0.98. The one exception is for Leu25 which has a  value of 0.62. This residue lies in the strand that is at one edge of face B, in an area that is not well defined as discussed above. In addition, the Leonard-Jones energy of each of the accepted structures is large and negative (Table 3.4) indicating that no unfavorable van der Waals contacts exist.  85  Chapter 3-Structure Determination of CBD  Table 3.4: Structural Statistics and Atomic RMS Differences^ rmsd from experimental distance restraints^ (A) (1793) rmsd from experimental dihedral restraints (deg.) (195) 0  Deviationsfrom idealized geometry Bonds(A) (2129) Angles (deg) (3841) Impropers^ (deg) (1102) o  XPLOR Energies (kcal mol" >)  <SA> 0.013 ± 0.002 0-45 ±0.13  <SA>.av 0.011 0.29  0.0019 ± 0.0002 0.54 ± 0.01 0.41 ± 0.02  0.0016 0.52 0.39  16.53 ± 3.9 2.64 ± 1.4 28.2 ± 2.3 168.2 ± 6.6 7.5 ± 1.3 13.9 ± 3.4 -569.2 ± 19.8  10.72 1.01 25.5 158.6 5.7 10.3 -612.5  6  backbone^ all heavy atoms Atomic rms differences^ (A) 0.79 ± 0.14 1.60 ± 0.20 residues 4-148 0.44 ± 0.05 1.61± 0.20 (3-sheet regions' <SA> represents the final ensemble of 25 simulated annealing structures; <SA> is the restrained minimized average structure obtained by averaging the 25 structures over residues 4148. Errors reported are ± one standard deviation. The number of restraints is given in parentheses. This includes 1705 NOE derived distance restraints, and 88 hydrogen bond restraints (44 hydrogen bonds). Torsion angle restraints include 56 (j)-angle restraints, 88 \|/-angle restraints, 42 %\-angle restraints, and 9 %2-angle restraints based on JHN-Ha> JNCy and Jc'Cy measurements, qualitative analysis of Jc8Ca values, and intraresidue and sequential NOEs from NOESY spectra. Improper torsion angle restraints maintain peptide planarity and chirality. The square well NOE ( E N O E ) using center averaging, the restrained dihedral (Edih), the improper torsion angle (Ej ), the angle (E i ), the bond (Ebnd)> d the quartic van der Waals repulsion energies (E iw) were calculated using force constants of 50 kcal mol" A , 200 kcal mok rad", 500 kcal mol" rad", 500 kcal mol' rad", 1000 kcal mol" A" , and 4 kcal mol A" . The van der Waals repulsion energy was calculated using sphere radii set to 0.75 times that supplied with XPLOR (parallhdg.pro). / EL-J is the Lennard-Jones van der Waals energy. This term was not included in any of the structure generating steps but was calculated for the final 25 structures and the restrained minimized average structure. 8 Atomic rms differences were calculated using the average structure before restrained minimization. Atoms used were N, C , and C. < This comprised residues 17-19, 25-27, 31-36, 45-50, 58-68, 72-81, 86-96, 100-108, 122129, and 137-148 inclusive. a  av  b  c  3  3  3  3  d e  c  an  mp  ang  e  0  1  vc  1  -1  h  2  1  4  a  2  1  2  1  2  -2  Chapter 3-Structure Determination ofCBD^j  Figure 3.13. Two views of the C traces of the final ensemble of 25 structures calculated for C B D N I - The co-ordinates were superposed using the backbone atoms from residues 4-148. The jelly-roll P-sandwich topology is apparent in view (A), whereas the presence of a binding cleft formed by P-sheet A is clearly evident in view (B). View A looks down on the binding cleft, with view (B) rotated by approximately 90° compared to (A). The figure was produced using the programme Molscript (Kraulis, 1991). A  Chapter 3-Structure Determination ofCBD^j  Al  10  20  BI  B2  30  AZ  40  50  B3  60  A3  70  A4  80  90  B4  100  A5  110  120  B5  130  140  150  (A) Distribution of NOE restraints per residue. Filled bars represent the number of long range NOEs (li - j I > 4), open bars represent the sum of non-trivial intraresidue, sequential and short range NOEs (1 < li - jl < 4). For every restraint the originating and destination residues were each counted once. (B, C, D) Angular order parameters S<|), Svj/, and S%1 for the cb and \|/ main chain, and side chain % 1 dihedral angles, respectively, observed in the final ensemble of 25 structures (Hyberts et al., 1992). (E, F) The rms deviation for all heavy atoms and main chain atoms, respectively, for the ensemble of 25 structures, aligned against the average structure obtained by superimposing residues 4-148. The locations of the ten p-strands are indicated on the top of the figure. Figure 3.14.  Chapter 3-Structure Determination of  -180  -135  -90  -45  CBDNI  0  45  90  135  180  Phi (degrees) Figure 3.15. Ramachandran plot of the mean minimized CBDNI structure generated by PROCHECK (Laskowski et al., 1988). The degrees of shading indicate the most favored, allowed and generously allowed regions of phi / psi space from an analysis of 118 highresolution crystal structures. Glycine residues are represented as triangles, all other nonproline residues as squares.  Chapter 3-Structure Determination of  Figure 3.16.  CBDNI  Superposition of the co-ordinates of the side-chains that make up the hydrophobic core of C B D on the C trace of the minimized average structure (thick line) of this protein. A l l heavy atoms between residues 4 and 148 were used to superimpose the 25 structures. The C B D is oriented with the binding cleft towards the page and (3-sheet B closest to the reader. This figure was produced using the programme Molscript (Kraulis, 1991). a  N 1  Chapter 3-Structure Determination of CBDN;  90  The presence of a cis-peptide bond for Ala83-Pro84 in C B D N I was initially identified based on the C7 chemical shift of the proline at 24.9 ppm. This is approximately 3 ppm upfield 13  from the value found for prolines with trans-peptide linkages (Stanczyk et al, 1989). The cispeptide bond was further confirmed by the observation of a strong H A83 to 'H p84 NOE, and 1  the lack of an 'H« 83 to A  ot  a  NOE in the 50 msec mixing time C / N NOESY-HSQC of 13  15  (Wuthrich, 1986). This NOE pattern is reversed for every other proline in C B D N I , all of  CBDNI  which are trans-linked. Pro84 lies in a turn between strands A3 and A4, classified as type Vial by Promotif (Hutchinson & Thornton, 1996). This type of turn requires a cis-linked proline at position i+2. A disulphide bond between Cys33 and Cysl40 was identified initially from the fact that the 13  CP chemical shift of a cysteine is indicative of its oxidation state. The CP shift of Cys33 is 13  47.4 ppm and that of Cysl40 is 47.5 ppm. These are within the expected range for an oxidized i  Cys, which has a random coil value of 41.8 ppm, as opposed to 28.6 ppm for the reduced form (Wishart & Sykes, 1994). This conclusion is consistent with the results of a DTNB titration of CBDNI,  demonstrating that there are no free thiols present in the protein. Cross-strand NOEs,  illustrated in Figure 3.10, also confirm the pairing of these two cysteine residues. The disulphide bond connecting Cys33 and Cysl40 is unusual in that it bridges two (3strands (Figures 3.10 & 3.17). Such covalent bridges between paired (3-strands are not observed commonly, although examples are found in azurin and chymotrypsin (Thornton, 1981). The disulphide in C B D N I is classified as a short right-handed hook by Promotif (Hutchinson & Thornton, 1996). Although the (0, \|/) angles of the cysteines and their neighboring amino acids do deviate slightly from those expected for residues in ideal anti-parallel (3-strands, the presence of the disulphide bond does not produce any pronounced distortions in strands B2 and B5. The backbone dihedral angles of the cysteines and adjacent residues still lie within the most favored regions of the Ramachandran plot and are well defined. In contrast, I observe that the %1 restraints for Cys33 and Cys 140 were often violated during the structural calculations. This may reflect conformational isomerization of the disulphide, particularly since the side-chain dihedral angles  Chapter 3'Structure Determination of CBDNI  A  B  C  Molscript ribbon diagram of the minimized average structure of CBD with residues from (3-sheet A that are implicated in oligosaccharide binding (chapter 2; Johnson et al., 1996a) shown in ball-and-stick format. Hydrophilic residues are identified with black atoms and hydrophobic residues with white atoms. The disulphide bond between p-strands B2 and B5 is also presented using gray atoms and dark gray bonds. View (A) looks directly down on the binding face, whereas view (B) is rotated by 90° to emphasize the binding cleft. These are the same molecular orientations used in figure 3.13. Strands B1-B5 are labelled in italics. F i g u r e 3.17.  NJ  Chapter 3-Structure Determination ofCBD^j  92  were determined based upon a staggered rotamer model using qualitatively estimated coupling constants.  Discussion  CBDNJ  Structure.  The structure of C B D N I in the presence of saturating concentrations of cellotetraose was determined using NMR methods. At the time this work was completed it was the third CBD structure published, the first of a Family IV CBD (Tomme et al., 1995).  CBDNI,  like all other  CBDs of known conformation, is composed of anti-parallel p-strands. However, in contrast to CBDCBHI  an  d CBDcex> C B D N I has a jelly-roll p-sandwich topology with a disulphide bond  bridging two adjacent P-strands. Furthermore, as a consequence of the strands of sheet A being shorter than those of sheet B, C B D N I contains a prominent cleft that runs across one face of the protein. As discussed below, this cleft is the binding site for cellooligosaccharides and amorphous cellulose. The individual P-strands in C B D N I are very well ordered, while the protruding loops are much less defined. In particular, the loops formed by residues 37-44, 82-84, and 113-121 are the most disordered, as evident by their high rms deviations and low angular order parameters (Figures 3.12 and 3.13). These loops form the extreme edges of the binding face, and thus a range of depths and widths are found for the dimensions of the binding cleft within the set of 25 accepted structures. An estimate of the observed variation in groove width is provided by the distances between the C atoms of Tyr43 and Alall8, which range from 24 to 31 A in the a  ensemble of structures. As will be discussed in chapter 5 using backbone N relaxation methods, 15  this is a function of true mobility and not a result of having few long-range restraints for residues in these loop regions of C B D N I •  Chapter 3-Structure Determination of CBD^i  93  Binding Mechanism.  Previously, I determined that C B D N I binds soluble cellooligosaccharides in order of increasing affinity cellotriose < cellotetraose < cellopentaose ~ cellohexaose (chapter 2; Johnson et al, 1996a; Tomme et al, 1996a). In addition to binding amorphous cellulose, C B D N I also  associates with oat- and barley-p-glucans with affinities equal to that of cellopentaose (Tomme et al, 1996a). Together, these results implied that the binding site spans approximately the length of a cellopentaose molecule. From the patterns of amide N and ' H chemical shift perturbations 15  N  resulting from the addition of cellooligosaccharides to C B D N 1, it was also demonstrated that residues in strands A1-A5 all interact with these soluble ligands (chapter 2; Johnson et al, 1996a). Therefore, P-sheet A was identified as the binding face of the N-terminal CBD of CenC. The involvement of the aromatic side-chains of Tyrl9 and Tyr85 in sugar binding was also suggested based on changes in the NMR chemical shifts and lineshapes of these residues and on perturbations of the near ultraviolet absorption spectrum of  CBDNI  due to addition of  cellooligosaccharides (Johnson etal, 1996a). Expanding upon these initial studies, I have now shown that the binding face of C B D N I is in fact a groove or cleft that extends across p-sheet A (Figure 3.17). The identification of this psheet as the ligand binding cleft of C B D N I is confirmed by the observation of intermolecular NOEs between bound cellotetraose and residues located within this region of the protein. Although the ]  H resonances of the cellotetraose are currently unassigned, NOEs from unlabelled sugar protons  to C-labelled Tyrl9 if! and W, V48 W 13  5  l  1  and 'HY , L77 U& , Y85 ' H and ' H ^ and A126 2  [  2  §  'HP were detected using a C col-edited, co3-filtered HMQC-NOESY experiment (Lee et al, 13  1994). Consistent with the reported dependence of binding affinity on the degree of oligosaccharide polymerization, the length of the C B D N I binding groove is approximately equal to that of an extended cellopentaose chain. Using the NMR-derived tertiary structure of C B D N I , it is now possible to look more closely at the residues involved in oligosaccharide binding, particularly in light of the thermodynamic studies of the interactions between C B D N I and various sugars (Tomme et al,  Chapter 3-Structure Determination of CBD^j  94  1996a). Figure 3.17 shows the exposed side-chains present on the amphipathic binding face of the averaged minimized structure of C B D N I . Two distinct features of C B D N I are evident from this figure. First, a strip of hydrophobic residues, comprised of Vall7, Tyrl9, Val48, Leul26, and Leu77, runs along the center of the binding cleft. Tyr 85, which may also be involved in binding, is located in a loop region near this hydrophobic strip. These non-polar residues may contact the pyranose rings of the oligosaccharide, providing favorable hydrophobic and van der Waals interactions. Second, there are numerous hydrophilic groups, including Asn50, Arg75, Asn81, Asp90, Thr87, Gin 124, and Glnl28, that flank the hydrophobic strip by lining the sides of the binding cleft. The polar residues likely provide hydrogen bonds to the equatorial hydroxyl groups of the sugar rings. In accordance with the relative insensitivity of sugar binding to pH and ionic strength (Tomme et al, 1996a), only two of these seven polar groups are ionizable. Although CBDNI  was investigated in the presence of saturating cellotetraose, the sugar was not included in  the structural calculations. Thus, it is not possible to state confidently that all of these polar and non-polar residues are involved directly in ligand binding. To address this issue, the specific roles played by these amino acids are being studied by site-directed mutagenesis (Jeff Kormos and Peter Tomme, Department of Microbiology, UBC) This proposed structural mechanism of oligosaccharide binding is entirely consistent with the thermodynamic parameters characterizing the association of soluble oligosaccharides to  CBDNI  (Tomme et al, 1996a). Based upon detailed calorimetric analyses, it was reported that sugar binding results in a favorable enthalpic change, compensated in part by a decrease in entropy. This implies that a predominance of polar interactions, such as hydrogen bonding, provides the primary driving force for binding. The dominant role of hydrogen bonding is observed with most other carbohydrate-binding proteins (Quiocho, 1986, 1989). Finally, and perhaps most importantly, the presence of a binding cleft in the structure of CBDNI  provides a simple explanation as to why this protein has the ability to bind soluble  oligosaccharides and amorphous cellulose, but not crystalline cellulose. The residues that mediate binding are all located within this cleft, and as a result, are unable to interact with the flat surface  Chapter 3-Structure Determination of CBDN  95  J  presented by crystalline cellulose. In contrast, soluble sugars and single strands of amorphous cellulose can bind to C B D N I by lying within this groove.  Comparison to Other Family IV CBDs. CBDNI  is the first Family IV CBD for which a structure has been determined. Figure 3.18  presents the sequences of four members of this family (Tomme et ai, 1995), aligned with the secondary structural elements identified in C B D N I - This alignment shows that the residues forming the eight strands of the jelly-roll p-sandwich motif of  CBDNI  are in general well  conserved, with deletions/insertions found in the intervening loop regions and the small p-strands Al and BI. The two cysteines that form the disulphide bridge are invariant, and most of the residues involved in ligand binding by C B D N I postulate that  CBDN2  a r e  conserved. Accordingly, it is reasonable to  from CenC, as well as the related CBDs from Thermomonospora fusca El  and Streptomyces reticuli Cell, adopt secondary and tertiary structures similar to those determined for C B D N I - Although it remains to be demonstrated if the T. fusca El and S. reticuli Cell CBDs exhibit the same specificity for soluble forms of cellulose, NMR studies of isolated  CBDN2  confirms that the second CBD from CenC does indeed bind cellooligosaccharides (E. Brun, pers. comm.). It has also been shown that the three dimensional structure of C B D N 2 is very similar to that of C B D N I (E. Brun, pers. comm.). This leads to the more tantalizing questions as to whether or not inter-domain interactions exist between  CBDNI  and  CBDN2  within the native enzyme and  why, when linked together in C B D N I N 2 , do these two domains appear to bind phosphoric-acid swollen cellulose in an independent and not co-operative manner (Tomme et ai, 1996a).  Comparison to CBDs that bind crystalline cellulose.  The dominant feature of the structure of C B D N I is the presence of a binding cleft. This distinguishes the N-terminal CenC CBD from all other characterized CBDs.  CBDCBHI  (Kraulis et  al, 1989), CBDcex (Xu et al, 1995), CBD i (Tormo et al, 1996) and CBD z (Brun et al, C p  EG  1997) all have flat binding surfaces that contain three exposed aromatic rings. Paralleling this  Chapter 3-Structure Determination of  CBDNI  A3  U P H r r B l l A E S L q P V ^ L i G T B E . H\|. K A D G ^ C V D L P G G d q ^ P l t t J A q L f VNQISJMSIFSSGIAPW 3TlpNldllN^DG|^CVDVPGGT|^Pf'^IjIpC G E IF VEiJ^kNbj^DTTTqPV; NV.. TOGUSDGKLCADVPGGTHISIITOS^IE A4  A5  B4  1  GES>JVtL,SFTAa/d IP.DMPVEVLVC L IGE'E S M A j F S F T A !£|S 11/. PV£ IRAIM B5  GGp SAE AW I LCLDDV » L DSEVE C . fimi C B D . . AG AY E FCISQV : L PTSAT C . f i m i C B D .SDE PWIFCLDDV a L LGRAE T. fusca E l HrtDElAi l l ± - T d G i V l F i V G G l . STE AW R FCVDDV: L LGGVP S. r e t i c u l i C e l l  Vi IJVLlfr . £jPA|LT|£jEjPRQVTlE|I|F'T A 3 A TYPATPAA E DPEf |V«| fl AF E 3G £ AI Ll 2 E PATRE X S F 1 S M L I F . . . PPE £- DA: |:p4iE .HALtqp|E^ETYEj^  I Y  |wdlA .yPVlM^GSY£|Ytll0|AEy  N 1  N2  Figure 3.18. Alignment of the amino acid sequences of four family IV CBDs (Tomme et al., 1995), C.fimi CenC ( C B D N I * CBDN2)> Thermomonospora fusca E l and Streptomyces reticuli Cell. The alignment was obtained using the program P H D (Rost, 1996). Boxes highlight positions where residues are conserved in three or more family members. The secondary structure of C B D N I is shown as open and filled boxes to represent the p-strands of sheet A and sheet B, respectively. The residues in sheet A that are implicated in cellooligosaccharide binding are shaded. A N C  Chapter 3-Structure Determination of CBDNI  97  structural difference, the other CBDs also stand apart from C B D N I in their affinity for crystalline cellulose and in the different thermodynamic forces that lead to carbohydrate binding by the CBDcex protein domain. Whereas the association of C B D N I with soluble oligosaccharides and phosphoric acid-swollen cellulose is enthalpically driven (Tomme et al, 1996a), the affinity of CBDcex for insoluble bacterial micro-crystalline cellulose results primarily from a favorable increase in entropy, indicative of a hydrophobic interaction (Creagh et al, 1996). Furthermore, it is known that the exposed tryptophans of CBDcex are involved in the binding event (Poole et al, 1993; Din et al, 1994; Bray et al, 1996). I postulate that CBDcex relies on hydrophobic stacking of these aromatic rings with the flat surface of crystalline cellulose as the hydroxyl groups of the glucosyl residues are involved primarily in interactions with adjacent polysaccharide chains. In contrast, C B D N I associates with a single strand of cellulose by exploiting a binding cleft in which polar residues are positioned to hydrogen bond to the exposed equatorial hydroxyl groups of the glucopyranose rings. These distinct structural and thermodynamic mechanisms highlight the complexity of cellulose as a substrate for enzymatic recognition and degradation.  Structural Similarity with 1,3-1,4-B-Glucanase.  The tertiary structure of C B D N I closely resembles those of endo-1,3-l,4-(3-glucanases from Bacillus macerans (Hahn et al, 1995a) and Bacillus licheniformis (Hahn et al, 1995c) as well as a hybrid Bacillus 1,3-1,4-p-glucanase (Keitel etal, 1993; Hahn et al, 1995b). Although the larger, l,3-l,4-(3-glucanases are composed of two (3-sheets of seven strands each, as opposed to five strands each for C B D N I , all these proteins share a common jelly-roll p-sandwich fold (Figure 3.19). Using the programme DALI (Holm & Sanders, 1995), C. fimi C B D N I and the hybrid Bacillus 1,3-1,4-P-glucanase were found to have 118 residues aligned with an rmsd of 3.7 A based on superposition of C co-ordinates. This alignment occurs despite only 8% sequence a  identity of these residues.  Chapter 3-Structure Determination of  CBDNI  Figure 3.19. Molscript ribbon diagrams of (A) the minimized average structure of Cfimi C J 3 D and (B) the crystal structure of the hybrid Bacillus 1,3-1,4-p-glucanase (Keitel et al., 1993; Hahn et al., 1995b). These two proteins share a common jelly-roll P-sandwich topology. N1  Chapter 3-Structure Determination of CBDNI  99  As pointed out by Hahn et al. (1995b), the prokaryotic 1,3-1,4-p-glucanases are topologically related to other polysaccharide degrading enzymes such as cellobiohydrolase I (Devine et al, 1994) and 1,4-p-xylanase II (Tbrronen et al., 1994), both from Trichoderma reesei. Thus, it is not surprising that the programme DALI identified these as well as Bacillus circulans xylanase (Campbell et al, 1993) and an S-lectin (Liao et al, 1994) as having tertiary  structures similar to that of  CBDNI  • The lectin also has a jelly-roll fold. It is intriguing that these  polysaccharide-binding domains resemble structurally the catalytic domains of several polysaccharide-degrading enzymes. The l,3-l,4-(3-glucanases are a distinct family of glucanohydrolases that specifically cleave l,4-(3-D-glucosidic bonds that are adjacent to the (3-1,3 linkages in mixed 1,3- and 1,4-linked [3glucans (Anderson & Stone, 1975). In addition, the hybrid l,3-l,4-(3-glucanase hydrolyses cellohexaose, but with poor efficiency (Hahn et al, 1995b). All bacterial endo-1,3-1,4-pglucanases known to date share sequence similarities with enrfo-l,3-(3-glucanases and have been classified into glycosyl hydrolase family 16 (Henrissat, 1991; Henrissat & Bairoch, 1993). The observation that C B D N I appears structurally related to these (3-glucanases is particularly interesting given that this CBD also binds the mixed 1,3- and 1,4-linked oat- and barley-(3-glucans (chapter 2; Tomme et al, 1996a). These two soluble oligosaccharides, along with lichenan from the Icelandic moss Cetraria islandica, are natural substrates for the 1,3-1,4-p-glucanases. Furthermore, the catalytic residues Glu 105 and Glu 109 of the hybrid 1,3-1,4-P-glucanases are located on opposite sides of the active site of the enzyme, whereas the similarly spaced Gin 124 and Glnl28 lie on the two sides of the C B D N I binding cleft (figure 3.20). The structural similarities between C B D N I and the 1,3-1,4-P-glucanases, combined with their binding to a common class of polysaccharides, suggests that these proteins are evolutionarily related, perhaps through divergence from a common ancestor with a jelly-roll sugar-binding motif. The unusual function of C B D N I in binding soluble or amorphous, but not crystalline glucans, may stem from an evolutionary relationship with the 1,3-1,4-P-glucanases, and not an ancestral binding-domain shared by the members of other CBD families.  Chapter 3-Structure Determination of CBDN]  Figure 3.20. Two views of the superimposition of the structures of CBD (purple) and the hybrid Bacillus 1,3-1,4-p-glucanase (green). Shown in red are the sidechains of the catalytic glutamate residues of the l,3-l,4-(3-glucanase (E105, E109) and the corresponding glutamine residues in CBD (Q124, Q128). N1  N]  101  Chapter 4 Calcium Binding By C B D N I  Abstract  The interaction of  CBDNI  with calcium was studied by NMR spectroscopy. The  association constant for calcium binding by C B D N I is (1.1 ± 0.5) x 10 M" at 35 °C pH* 6.0. 5  1  The oligosaccharide binding ability of C B D N I is not affected by the presence of this metal ion, as determined by the similarity of binding constants of calcium-loaded and calcium-free C B D N I with cellopentaose. On the basis of the observed patterns of amide chemical shift changes, and similarity with the family of Bacillus 1,3-1,4-p-glucanase structures, the putative calcium ligating atoms on CBDNi are T8 O, G30 O, D142 O and a side-chain oxygen of D142 . From analysis of lineshapes of resonances that are in intermediate exchange on the NMR chemical shift timescale, the on-rate for calcium association is determined to be (5 + 2) x 10 s~' M" . This is within a 7  1  couple of orders of magnitude of the diffusion limit, suggesting that the calcium-binding site on CBDNI  is largely preformed.  Introduction  Background  When binding studies of C B D N I with cellooligosaccharides were initially performed (chapter 2) it was noted that the resonances of approximately 13 amides near the N-terminus and adjacent to the disulphide bond were not observed in the 'H- N HSQC spectrum of the protein. 15  102  Chapter 4-Calcium Binding by CBDNI  These resonances were detected only upon addition of excess cellotetraose, but not with any other cellooligosaccharide studied. The appearance of these resonances with the addition of cellotetraose was observed for at least seven different samples of C B D N I over a period of 1.5 years, using several sugar samples with the same lot number. However, when subsequent relaxation experiments (chapter 5) were performed, it was observed that these peaks no longer appeared upon addition of cellotetraose from a new lot. The identity of the new sample of cellotetraose was checked by mass spectrometry and it was found to have the expected molecular weight. Also, the new cellotetraose eluted with the same HPLC retention time as a sample of the previously used cellotetraose, obtained from an old NMR sample. With confirmation of the identity of the new cellotetraose, I concluded that the original sample of cellotetraose, and not any of the other cellooligosaccharides, was contaminated with a substance that bound C B D N I and caused the missing amide peaks to appear. Given that the hybrid Bacillus,  the Bacillus macerans and the Bacillus licheniformis 1,3-1,4-p-glucanase structures and  CBDcip, all of which share a similar (3-jelly roll sandwich fold as C B D N I , binds calcium (Keitel et al, 1993; Hahn et al, 1995a, Hahn et al, 1995c, Tormo et al, 1996) it was strongly suspected that this metal ion could be the contaminant. More recently the structure of the family III CBD from the Thermomonospora fusca endo/exocellulase E4 was reported ( C B D E 4 ; Sakon et al, 1997). This CBD, which has a similar structure as C B D N I , also contains a bound calcium ion.  Experimental Methods  Sample preparation  The unlabelled and N-labelled protein samples were overexpressed and purified as 15  described previously (chapter 2). The sample of C B D N I nonrandomly fractionally C enriched at 13  a level of 10% is that used previously for the resonance assignment of C B D N I (chapter 3). All buffers were passed over a Chelex-100 column to remove metal ions. Plasticware, glassware and  103  Chapter 4-Calcium Binding by CBDNI  NMR tubes were soaked overnight in 4N HC1, then extensively rinsed with deionized water before use to remove any calcium present. The only exception was for the microsep concentrators (Filtron) which were soaked in IN HC1, according to manufacturers recommendations. Ethylenediamine tetraacetate (EDTA), added to the unlabelled and N-labelled C B D N I sample to l5  remove metal ions bound to the protein, was subsequently removed by extensive buffer exchange using a microsep concentrator. Some protein is lost with numerous buffer exchanges, and thus in an effort to conserve material, the 10% C-labelled C B D N I sample was not treated with EDTA. 13  As a result, it was approximately 20% calcium-loaded. The titration of this sample was only performed to obtain qualitative data on chemical shift changes with calcium binding. The buffer used for the NMR experiments presented in this chapter was 50 mM sodium chloride, 50 mM sodium acetate ( H3), 0.02% sodium azide, pH 6.0, 10% D2O / 90% H2O. Acetate only binds 2  calcium weakly, with a low K of about 100 M" (Brian Sykes, personal communication), and thus 1  a  does not compete significantly with C B D N I -  NMR spectroscopy  One dimensional NMR spectra of 0.4 mM unlabelled C B D N I samples were used to screen for metal binding. 'H- N HSQC spectra were obtained using a 1.4 mM sample of uniformly N 15  enriched  CBDNI-  15  Regular and constant-time versions of !H- C HSQC experiments were 13  recorded using a 0.5 mM sample of C B D N I nonrandomly fractionally C enriched at a level of 13  10%. Experiments were performed on a Varian Unity 500 MHz spectrometer equipped with a triple resonance probe and pulsed field gradients. All spectra were recorded at 35 °C and analyzed using a combination of NMRPipe (Delaglio et ai, 1995), FELIX v2.30 and FELIX95 (Biosym Technologies). The H- N HSQC experiments were recorded using the enhanced-sensitivity [  15  pulsed field gradient of approach of Kay et al. (1992). Selective water flip back pulses were incorporated to minimize the perturbation of the bulk water magnetization (Grzesiek & Bax, 1993; Zhang et al, 1994). The initial delays in the indirectly detected dimensions were set to l/(2*sw),  104  Chapter 4-Calcium Binding by CBDNI  resulting in a 180° first order phase shift across the transformed spectrum and inversion of aliased peaks (Bax et al., 1991).  Determination of binding constants  The binding of calcium to C B D N I at 35 °C, pH 6.0 was monitored quantitatively using 'H15  N NMR spectroscopy. A stock 0.5 M solution of calcium chloride was made by weight in  exactly the same buffer as for C B D N I - Dilute solutions were made subsequently from this stock. The pH values of the CaCl2 solutions were adjusted as necessary to pH 6.0. The initial concentration of C B D N I was 1.4 mM. Ten ifl-^N HSQC spectra with increasing concentrations of calcium were acquired as 512 x 96 complex points in the 'H and N dimensions with spectral 15  widths of 7000 and 1650 Hz respectively. Equilibrium association constants were determined by nonlinear least-squared fitting of the chemical shift data versus total calcium concentration to the Langmuir isotherm describing the binding of one ligand molecule to a single binding site. The data were fit using the programme PLOTDATA (TRIUMF, UBC, Vancouver). Only data from peaks in the >H- N HSQC spectra 15  that are in fast exchange between the free and calcium-bound forms of C B D N I on the NMR chemical shift time scale were used in the analysis. The same equations and procedures (chapter 2) used to analyse the binding of soluble cellooligosaccharides to C B D N I were applied here. To establish whether calcium binding has an effect on cellooligosaccharide binding, the association constant of fully calcium-loaded C B D N I with cellopentaose was determined. Ten H!  15  N HSQC spectra with increasing concentrations of cellopentaose were acquired using 512 x 96  complex points in the *H and N dimensions with spectral widths of 6500 and 1450 Hz , 5  respectively. The association binding constant of calcium-loaded C B D N I with cellopentaose was determined using the methods outlined in chapter 2.  105  Chapter 4-Calcium Binding by CBDNI Analysis of binding kinetics  Calcium association/dissociation kinetics were determined by lineshape analysis of residues showing intermediate uncoupled two-site exchange during the titration of C B D N I with calcium. This was done manually by comparing the experimental lineshape and frequency with that from simulated spectra. For the experimental spectra used in these comparisons, data were processed using only a linebroadening window function. Spectral simulations were obtained using values for the fraction of protein bound (fb), the total change in chemical shift of a given nucleus between the bound and free forms (A8t tal), the dissociation rate constant 0  (fc ff)  a n  0  d  the transverse relaxation  times for the YL or N nucleus in the free (T2f) and bound (T2b) protein. From the previously [  15  determined calcium-binding constant, values of fb and A8 tal were calculated. Values of T2f and to  T2b were estimated from the full linewidths at half height, and checked by comparison of the experimental and calculated lineshapes for free and fully calcium-saturated protein. Spectra were then simulated as a function of k ff using a programme written by Michael Strain (University of 0  Oregon) run as a macro in Felix version 2.3. The spectral simulations produced by this programme are based on the formalism described by Sandstrom (1982). Originally, the equations for the lineshapes were derived for one species in equilibrium with another (A ^ ^  B).  The case of C B D N I , association with calcium differs in that it is a  bimolecular process with both free protein (Pf) and calcium (Ca ) in equilibrium with bound 2+  protein (PCa ) (Pf + C a ^ ^ PCa ). As a result, the association rate is the product of the on2+  2+  2+  rate constant and the free calcium concentration. To avoid this complication, the data were fit by determining of the off-rate (fcff), which is a unimolecular process. Finally, k can be determined 0  on  as it is the product of k ff and K (k = k ff * K ). 0  a  on  0  a  Structure calculations of calcium-loaded CBDNI  Structure calculations of calcium-bound C B D N I were performed using X-PLOR v3.8. Torsion angle and distance restraints used were the same as those used earlier for the structure calculations of non-calcium-loaded C B D N I (chapter 3). A calcium atom was explicitly included in  Chapter 4-Calcium Binding by CBDNI  106  the structure generation process with distance restraints of 2.45 ± 0 . 1 5 A to its putative ligating atoms (T8 O, G30 O, D142 O, D142 O ). This restraint was determined from analysis of the 5  distances between the ligating atoms and the calcium ion in the Bacillus macerans glucanase structure (pdb code:  2AYH).  l,3-l,4-(3-  Fifty structures were determined following the simulated  annealing protocol in X-PLOR (Brunger, 1993) with the previously determined minimized average structure of C B D N I  u s e  d as the starting model.  Results  CBDNI  binds metal ions  The binding of C B D N I to a series of diamagnetic metal ions ( Ca , Zn , Cd and 2+  2+  2+  Mg ) was studied by one-dimensional NMR spectroscopy (Figure 4.1). Based on the numerous 2+  changes in the spectra of C B D N I with the addition of each ion, it is concluded that all these metals are bound by C B D N I - However, it is only with the addition of calcium that the complete appearance of the two most downfield resonances, W16 H and G7 H , previously thought to be el  N  characteristic of cellotetraose binding, are observed. As the protein concentration in each of these samples is identical, and the same ten fold molar excess of metal was added to each, it is concluded that  CBDNI  binds Ca with the highest affinity, followed by Zn , Cd and Mg , in order of 2+  2+  2+  2+  decreasing affinity. These last two bind with approximately equal affinity. Also, as each metal ion produces similar changes in the spectra of C B D N I , it is likely that all bind at the same location in the protein.  Calcium equilibrium binding association constant and binding stoichiometry  After this initial screen for metal binding, the interaction of calcium with C B D N I was further studied. The effects of calcium binding by  CBDNI  was monitored by one-dimensional 'H  spectra (figure 4.2), and 'H- N HSQC spectra (figure 4.3) recorded for ten samples with I5  Chapter 4-Calcium Binding by CBDNJ  ll .0  10.0  9.0  8.0  7.0  ppm  Figure 4.1. Screen for metal binding by C B D N I using one-dimensional I H N M R spectroscopy. Perturbation of the resonances W16 H s l and G7 H N are indicative of metal association. A 10 fold molar excess of metal ion was used for each spectrum.  Chapter 4-Calcium Binding by  108  CBDNJ  W16 Hel  11.0  JO. 5 ppm  10.0  Figure 4.2. Titration of CBDNI with calcium monitored by 15N-decoupled one-dimensional NMR spectroscopy. The molar ratio of calcium-to-protein is indicated. Resonances WI6 Hel and G7 HN exhibit intermediate two-site exchange.  Chapter 4-Calcium Binding by  CBDNI  Figure 4.3. Titration of CBDNI with calcium monitored by 1H-15N HSQC spectra.  Shown is a portion of ten overlaid spectra with the same molar ratios as indicated on figure 4.2. Arrows indicate the directions in which the amide peak shifts with added calcium. For the sake of clarity, not all the peaks are assigned. The peaks labelled * are due to impurities.  Chapter 4-Calcium Binding by CBD^J  110  increasing amounts of added calcium. The peaks most indicative of calcium binding in onedimensional 'H spectra are those of W16 H and G7 H . Based on these shifts it is estimated el  N  that, without steps being taken to remove calcium, freshly prepared samples of C B D N I are approximately 20% calcium-loaded. From analysis of figure 4.4A, it can be determined that the stoichiometry of calcium binding by C B D N I is 1:1. This arises from the fact that the titration curves have a plateau at a total calcium concentration equal to the protein concentration used in this experiment, 1.4 mM. From this data it is possible to rule out a model which describes the binding of two or more molecules of calcium, each binding with approximately equal affinity. This latter situation would have resulted in a series of curves which plateau at a total calcium concentration at least twice that of the protein, 2.8 mM. An equilibrium association constant (X ) for calcium binding was determined by a a  nonlinear least-squares fit of the H- N L  I5  HSQC  chemical shift data to a binding isotherm for a  protein with a single ligation site. For each amide, two K values were obtained, one using the data a  for the H and another for the N nucleus. Figure 4.4A shows 28 fits for the data from 14 N  15  different amides that exhibit the greatest change in chemical shift with calcium binding, and who have measurable chemical shifts at each titration point. These amides show a coincident titration indicating they all are effected by the same binding event. In total 45 fits from 28 different amides (D10, G12, G15, W16, V17, T29, G30, A31, G39, V48, L49, G51, V52, A53, R63, T65, A66, S69, D71, T73, L95, S97, R100, V102, T105, L139, L141 and A145) were used to obtain an average K value of 1.1 x 10 M" with a standard deviation of 0.5 x 10 M" at 35 °C, pH* 6.0. 5  1  5  1  a  This is a relatively high K value to be determined at the C B D N I concentrations used for a  NMR studies. To test its validity, a series of curves were calculated with varying K values. These a  simulated curves were plotted against the data obtained from the titration of R100 H , which has a N  regressed K value of 1.0 x 10 M" (Figure 4.4b). This graph emphasizes that it is hard to 5  1  a  determine accurately a K value under these conditions as large changes in K produce relatively a  a  subtle variations in the shape of the calculated curves. This graph does show however, that the K  a  Chapter 4-Calcium Binding by CBD^j  Figure 4.4. (A) Overlay of the titration data and best-fit curves of 28 15N and HN  resonances from 14 different amides (D10, G15, W16, T29, G30, A31, A53, R63, T65, A66, S69, D71, R100, L139 andL141) in the titration of CBDNI with calcium. Diamonds indicate the experimental data, the solid lines the fit to a model that describes the binding of calcium to protein in a 1:1 complex. Coincident data indicate the amides all monitor the same binding event. The arrow marks the plateau in the curve where the total sugar concentration equals the protein concentration (-1.4 mM) indicating 1:1 binding stoichiometry. (B) Simulations of the binding curves to the data for R100 HN. Squares indicate the experimental data. The lines are the simulated curves with Ka values of 50 000 M-1, 70 000 M-l, 100 000 M-l, 200 000 M-l, 300 000 M-l, 500 000 M-1 and 1 000 000 M-l.  Chapter 4-Calcium Binding by CBD^j  112  value for calcium binding by C B D N I is in the range of 1 x 10 M to 3 x 10 M . This value is 5  _1  5  -1  supported by the determination of a K value of (8.3 ± 0.2) x 10 M~' for the association of 4  a  calcium with C B D N I under identical buffer conditions, as determined by isothermal titration calorimetry (Louise Creagh, personal communication). Three amide resonances, 14, G7 and F9, exhibit a nonlinear titration during the last two or three titration points (not shown). This could be indicative of these residues sensing a second binding event. This second binding site would have a very much weaker association constant, in the order of less than 4000 M~ . Since this is weak binding and only detected for three amides, its 1  significance is unclear.  Effect of calcium on cellooligosaccharide affinity  Calcium-saturated C B D N I  w a s  titrated with cellopentaose to determine the effect that metal  binding has on the protein's affinity for cellooligosaccharides. The equilibrium affinity constant for cellopentaose binding was calculated in a similar manner to that described earlier (chapter 2) and in the previous section. As was found previously for cellopentaose-binding by -20% calciumloaded C B D N I (chapter 2), the stoichiometry of cellopentaose binding by calcium-loaded C B D N I is 1:1. This is shown by the data having a plateau at a total cellopentaose concentration equal to that of the protein, 1.4 mM (figure 4.5A). Fits using H and N chemical shift data from 14 different amides (V17, Y19, G20, V34, N  15  G44, V45, L49, R75, N81, G83, T87, GOO, G131 and L139) resulted in an average K for &  cellopentaose binding of 3.1 x 10 M with a standard deviation of 0.8 x 10 M" . The value 4  _1  4  1  obtained previously for a sample of C B D N I approximately 20% calcium-bound is (3.4 ±0.8) x 10 M" (Chapter 2; Johnson et ai, 1996b). This shows that the oligosaccharide-binding ability of 4  CBDNI  1  is not affected by calcium. Fits of the data for both the H and N chemical shift data 15  N  from 11 amide resonances are shown in figure 4.5A. Theoretical curves calculated with varying K values plotted against the data obtained for the titration of Y19 H are shown in figure 4.5B. In N  a  this case of cellopentaose binding, the K value is lower than that for calcium binding and more a  Chapter 4-Calcium Binding by  CBDNI  Total Cellopentaose Concentration (mM)  B  8.90  0.0  2.0  4.0  6.0  8.0  10.0  12.0  Total Cellopentaose Concentration (mM)  Figure 4.5. (A) Overlay of the titration data and best-fit curves of 22 15N and HN resonances from 11 different amides (VI7, Y19, G20, V34, L49, R75, N81, T87, G130, G131 and L139) in the titration of calcium-loaded CBDNI with cellopentaose. Diamonds indicate the experimental data, the solid lines the fit to model that describes the binding of cellopentaose to protein in a 1:1 complex. Coincident fits indicate that the amides all monitor the same binding event. (B) Simulations of the binding curves to the data for Y19 HN. Squares indicate the experimental data. The lines are the simulated curves with Ka values of 5 000 M-l, 10 000 M-l, 30 000 M-l, 50 000 M-l, 100 000 M-l and 1 000 000 M-l.  114  Chapter 4-Calcium Binding by CBDNI  confidence can be placed in its accuracy as changes in K values produce a greater effect on the a  calculated curves than was the case for calcium binding.  Kinetics of calcium binding  NMR spectroscopy provides a powerful method to monitor the exchange of a nucleus between different environments due to a chemical reaction, ligand binding or conformational transitions. The exchange process can be monitored by NMR even if the sites are chemically equivalent as long as they are magnetically distinct and in chemical equilibrium. To qualitatively understand the effect of exchange on an NMR spectrum, consider a nucleus that is in experiencing conformational exchange between two magnetically distinct sites. The exchange rate between the two sites is k, and the two sites differ in frequency by Av. The resonance frequency of the spin in each site can be observed for a time in the order of 1/k before the spin jumps to the other site and begins to precess with a different frequency. The finite observation time places a lower limit on the magnitude of Av required to distinguish the sites. When k « Av, distinct signals from are observed for the nucleus at each if the two frequencies. This is the slow exchange limit. At the fast exchange limit, k » Av, a single resonance appears at the population-weighted average chemical shift of the nuclei at the two sites. This was the case for the binding of cellooligosaccharides by C B D N I presented in this chapter and chapter 2. A third case arises when the exchange rate is of the order of the chemical shift separation between the two sites (k ~ Av). This situation, where the lines become very broadened and begin to coalesce, is known as intermediate exchange. In this section, detailing the binding of calcium to CBDNI,  some resonances whose chemical shifts are perturbed by calcium binding are in the fast  exchange limit, while others experience intermediate exchange. The NMR chemical-shift timescale varies by residue, as it is defined by the difference in chemical shift between the resonance frequencies of the two environments of a given nucleus.  The equations that describe the  lineshape, or bandshape, of resonances experiencing two-site exchange are complex and are summarised in Sandstrom (1982).  Chapter 4-Calcium Binding by CBDNI  115  The kinetics of calcium association/dissociation were obtained from analysis of lineshapes from residues that are in intermediate exchange on the NMR timescale. As expected from the classical model of two-site exchange, these residues appear as sharp peaks in the calcium-free form then broaden and decrease in intensity with added calcium. In the fully bound form these peaks are once again sharp. An example of this behaviour is shown for W16 H in figures 4.2 and 4.3. el  Seven different lineshapes from six different residues were analysed by the formalism of Sandstrom (1982). These were the 'H lineshapes of W16 H obtained from one-dimensional el  spectra, and the H lineshapes of G30, A31, L139, L141, and A145 and the N lineshapes of ]  l5  G30 obtained from the !H- N HSQC spectra collected during the titration. Only residues whose 15  amide broadened but did not disappear entirely were selected for analysis. Values for A;ff were G  obtained by simulation of the spectrum at as many of the titration points as possible to yield an averagefc ffvalue for each peak. From these individual points a global average of 4.5 x 10 s 2  _1  0  with a standard deviation of 0.6 x 10 s was determined. Figure 4.6 shows an example of the 2  _1  data from the calcium titration used to determine the k a rates compared with the calculated 0  lineshapes. A value for k of (5 ± 2) x 10 M s~' was determined from the experimentally 7  _1  on  determined K and k ff values. a  0  Identification of calcium-binding site  When discussing patterns of chemical shift perturbations with ligand binding, it is important to bear in mind that changes can arise either from direct interaction of the nucleus with the ligand or indirectly from conformational changes resulting from formation of the protein-ligand complex. Chemical shifts are also a very sensitive indicator of structure, with subtle conformational changes often resulting in large changes in shift. Keeping this in mind, it is possible to gain insight into the location of the calcium-binding site in C B D N I from the patterns of chemical shift perturbations upon addition of calcium. Figure 4.7 summarises the change in H and N chemical shift of each amide in C B D N I due to the binding of calcium. 15  N  Chapter 4-Calcium Binding by  1  CBDNI  1.2  I l'.0  11.2  11.0  10.8  10.6  10.4  10.6  10.4  'H(ppm)  10.8  'HCppm)  Figure 4.6. Comparison of the experimental (A) and simulated lineshapes (B) using the  data for W16 Hel. The experimental data is an expanded region of the 1 -D IH NMR spectra shown in figure 2. The simulated lineshapes were produced using a koff value of 500 s-1, T2 values of 0.02 s (free) and 0.015 s (bound), and frequency values of 5539 Hz (free) and 5256 Hz (bound). The numbers indicate the fraction of protein in the calci bound form at that point in the titration, as calculated using a Ka of 110 000 M-L.  Chapter 4-Calcium Binding by  1  CBDNI  Figure 4.7. Patterns of chemical shift perturbation due to calcium binding by CBD N ) .  The absolute value in the difference between the H and N chemical shifts of the backbone amides in the free and calcium-bound protein are plotted as negative and positive numbers, respectively. The region that shows the greatest change in chemical shift in voles the putative binding site residues T8, G30 and D142. N  1 5  118  Chapter 4-Calcium Binding by CBDNJ  C worm diagram of CBD with residues that experience the most significant change in chemical shift with calcium binding (see text), or whose amide resonances are not observed in the absence of added calcium and appear upon calcium addition (T8, V144, L146) coloured red. The top panel shows P-sheet B, which lies opposite to the binding face. The lower panel is a 90° rotation from the view shown in the top panel, the binding cleft is seen at the top of the structure in this view. Selected residues are labelled. The amino and carboxyl termini are denoted by the labels N and C, respectively. This figure was made using the programme GRASP (Nicholls et al., 1991). Figure 4.8.  a  N1  119  Chapter 4-Calcium Binding by CBD^J  The average change in the absolute value of N chemical shift due to calcium binding is 15  0.4 ppm with a standard deviation of 0.7 ppm. For the H dimension the average absolute value N  in chemical shift change is 0.05 ppm with a standard deviation of 0.08 ppm. Shown in figure 4.8 is a backbone worm diagram of C B D N I with residues that experience a change in either N or 15  H chemical shift one standard deviation greater than average in red. In addition, those residues N  not observed in the absence of added calcium (T8, V144, L146) are similarly coloured. Although this analysis is statistically questionable, as the distribution of the absolute values of the change in shifts is skewed and not Gaussian, it does provide a criterion for defining what level of perturbation is "significant". An alternative to this guage as to what changes are significant would be to indicate the location of an arbitrary number (-20%) of residues that titrate the most. Figure 4.9 shows the four overlapped R- C {  L3  [3  HSQC spectra from the titration of the 10%  C labelled C B D N I with calcium. It is clear that, when compared to the 'H- N HSQC spectra 15  (figure 4.3), the methyl resonances are less sensitive to calcium addition. Far fewer resonances change their chemical shift, and the ones that do titrate experience a much smaller perturbation. The resolved methyl resonances that shift upon calcium binding are, in order of greatest shift, T8Y,1472, L32 , L49 , L49 and A68P. With the exception of L49, all lie in regions of the 2  51  52  51  protein which contain backbone amides that also show significant chemical shift change. The methyls of L49 are in the hydrophobic core of the protein lying close to the side-chain of W16. As is shown in figures 4.2 and 4.3, W16 H also has a large change in chemical shift with added el  calcium. Due to spectral overlap, it is not possible to obtain data for the methyl groups of residues T65Y, T67Y, A3lP or T29Y, which lie close to additional residues whose amide resonances 2  2  2  titrate markedly. From figures 4.7 and 4.8 it is clear that four regions of C B D N I are affected the most by calcium binding. These include the entire N-terminus up to residue 17, residues 30-33 (P-strand B2), residues 65-68 (p-strand B3), residues 139-146 (p-strand B5) and residue 100 (p-strand B34). Aside from residues 16 and 17, no amides on the binding face of the protein show a significant change in chemical shift with calcium binding. Instead all the residues with the largest  Chapter 4-Calcium Binding by  CBDNI  o  O  O  O Q  o  CL  CL  o  ^  CM  ^  O  « s L 4 0 51  L141 82 L146 81  Q  ^  o  o  o  ) C P  J2_  1.5  J .0 HI  L49S2  0.5  L 3 2 51  0.0  ( p p m )  Figure 4.9. Titration of CBDNI with calcium monitored by 1.H-13C HSQC spectra. Shown is the region of the spectrum where the observed peaks are due to methyl groups. Four spectra with increasing amounts of added calcium are overlaid. Only residues whose methyl resonancess experience a change in chemical shift with calcium binding are assigned.  Chapter 4-Calcium Binding by CBDNI  121  change in chemical shift are on the opposite face. Each of (3-strand B2 to B5 of this face has at least one residue that shows larger than average changes in chemical shift with calcium binding, with greatest changes concentrated on fi-strands B2, B4 and B5.  Discussion  Location of the metal binding site and structure of calcium-loaded CBDNI  Chemical shift perturbations provide only qualitative data for determining the location of the calcium-binding site on C B D N I , but do not unambiguously identify the coordinating atoms. Therefore I examined the C B D N I structure for possible regions which contain residues that commonly interact with calcium. In general, these are areas high in negative charge, and thus rich in aspartic and glutamic acid residues. There are two regions that qualify as potential binding sites. The first is the loop region involving residues 6 to 11, containing the potential ligating residues E6, D10 and DU. In the minimized average structure of C B D N I (Chapter 3; Johnson et al., 1996b) the side-chain of E6 points away from D10 and DI 1. However, this is also the most disordered region of the structure and the exact location, or locations, of the side-chains may not be accurately represented with this structure. Thus the involvement of E6 cannot be ruled out based on its apparent orientation. The other possible location for calcium binding is the area around D142, D143, residues 29-32, and the N-terminal residues 7-9. In contrast to C B D N I ,  CBDN2  does not bind calcium (E. Brun, personal communication).  Thus, it is expected that the calcium-binding residues are not conserved in the sequences of these two proteins. If the residues of one of the possible binding site were conserved and another not, this would provide evidence for its location. Unfortunately, residues in both possible calciumbinding sites exhibit low sequence similarity. E6, D10 and DU in C B D N I correspond to a histidine, a serine and a glutamate in CBDN2> while D142 and D143 in C B D N I align with a serine  Chapter 4-Calcium Binding by CBD^j  122  and a glutamine in C B D N 2 - This does not differentiate the two sites, but it is consistent with the lack of binding in  CBDN2-  Strong evidence for the location of the calcium-binding site comes from a comparison of CBDNI  with the 1,3-1,4-p-glucanase structures. As discussed earlier (chapter 3) these proteins  are structurally similar, all having a (3-jelly roll fold. Both the hybrid Bacillus 1,3-1,4-p-glucanase and the Bacillus  macerans  1,3-1,4-p-glucanase bind calcium octahedrally, using P9 O, G45 O,  D207 O and D207 O (Keitel et al, 1993; Hahn et al, 1995a) (figure 4.10). Additionally, two 8  water molecules in the hybrid 1,3-1,4-p-glucanase are coordinated to the calcium ion in that crystal structure. Presumably two waters molecules also coordinate the calcium in the Bacillus  macerans  1,3-1,4-p-glucanase structure to complete the octahedral geometry. The structure of the  Bacillus  licheniformis  1,3-1,4-p-glucanase is slightly different, binding calcium in a pentahedral-  bipyramidal manner with the atoms P9 O, D207 O, D207 O and two water molecules forming 81  the pentahedral plane. Another water and the carbonyl of G45 form the apical positions (Hahn et al, 1995c). CBDNI  and these 1,3-1,4-p-glucanase structures are remarkably similar in this calcium-  binding region. All posses a bulge in the P-sheet at residue D142 for C B D N I and D207 for the 1,3-1,4-p-glucanase structures. In light of these two facts, it is therefore very likely that the calcium-binding site of C B D N I is at the same position, involving T8 O, G30 O, D142 O and a side-chain oxygen of D142 as the metal coordinating ligands. As will be discussed below, this location for the calcium-binding site is also supported by similar calcium-binding constants, and a common biological role of stabilising the folded structure for both C B D N I d the hybrid 1,3-1,4an  p-glucanase. The structure of C B D N I bound to calcium was determined using these atoms to coordinate the calcium ion. As there is no unambiguous experimental evidence these are the coordinating atoms, this structure should be considered a model. For this reason the coordinates were not deposited in the Brookhaven Protein Data Bank. One way to unambiguously experimentally determine the location of the metal binding site would be to substitute the calcium ion with Cd. 113  Chapter 4-Calcium Binding by  CBDNI  Figure 4.10. Two views of the structure of Bacillus macerans l,3-l,4-f3-glucanase showing its calcium-binding site. Drawn in red are the ligands that coordinate calcium in this structure (P9 O, G45 O, D207 O and D207 O ). Based on this structure, the corresponding atoms T8 O, G30 O, D142 O and D142 O in C B D N I are assigned as the putative calcium coordinating ligands. The calcium atom is coloured in green. 52  8  Chapter 4-Calcium Binding by CBD^j  124  A C - ' Cd HSQC could then be run to assign the resonances and identity of the backbone l3  l3  carbonyls or side chain carboxyls that ligates the metal. In almost all calcium binding proteins coordination is either octahedral or pentagonal bipyramidal (McPhalen et al., 1991). It is therefore likely that two or three water molecules also ligate calcium in the C B D N I structure. It is slightly atypical that only one carboxylate group and three backbone carbonyls ligate calcium. The presence of two or more Asp or Glu sidechains is usually the case (McPhalen et al, 1991). The involvement of three backbone carbonyl atoms ligating calcium probably accounts for the octahedral coordination. The steric constraint of having the backbone of the protein so close to the calcium probably excludes a higher coordination number. As expected for structures calculated using the same restraints, the only region of  CBDNI  that exhibits a difference between the structures calculated with calcium and the structures calculated previously (chapter 3) is in the vicinity of the calcium ion, figure 4.11. Specifically, in the calcium bound structure the backbone carbonyl of T8 twists around 180° so as to ligate the calcium ion. The backbone carbonyl, and the side-chain of D142 also twists to point toward the calcium ion (figure 4.11).  Binding affinity and kinetics  In this study it was found that C B D N I binds calcium fairly tightly, with a K of (1.1 ± 0.5) a  x 10 M at pH* 6.0, 35 °C. This value matches, within the error range, that determined by 5  _l  isothermal titration calorimetry of (8.3 ± 0.2) x 10 M (L. Creagh & C. Haynes, pers. comm.). 4  _1  Also, the K of C B D N I for calcium is similar to that determined for the hybrid Bacillus 1,3-1,4-pa  glucanase H(A16-M) which was reported as (1.0 ± 0.4) xlO M" , also at pH 6.0. (Keitel, et ai, 5  1  1994). This similarity in K values supports the proposal that the calcium-binding site of a  CBDNI  lies at the same location, and involves the same ligands, as the hybrid Bacillus 1,3-1,4-pglucanase H(A16-M).  Chapter 4-Calcium Binding by CBDjy/  G30  Figure 4.11. Model of the structure of the calcium-binding site in CBDNI. Shown in dark grey are the putative calcium-ligating ligands (T8 O, G30 O, D142 O and the sidechain of D142) of the lowest energy structure from the ensemble calculated in the presence of calcium. Shown in light grey are the same ligands in the mean minimized structure calculated without calcium (chapter 3). In black is the calcium ion, drawn using its ionic radius. In the calcium-bound model the backbone carbonyl of T8 twists 180 degrees to ligate calcium. The backbone carbonyl of D142 twists 90 degrees, and its side chain rotates to point towards the calcium ion. The position of the carbonyl of G30 is relatively unchanged. In addition to these ligands, one or more water molecules probably also coordinate calcium. This figure was made using the programme Molscript (Kraulis, 1991).  126  Chapter 4-Calcium Binding by CBDNI  When produced by Cellulomonas fimi, the CenC cellulase, from which C B D N I is derived, is secreted into the environment. It has been estimated that the concentration of calcium extracellularly is 2 mM (McPhalen et al, 1991) and 10 mM in sea water (Glusker, 1991). The calcium concentration present in the natural environment of C. fimi is not known, but it is likely to also be in the millimolar range. At these concentrations,  CBDNI  would be fully calcium saturated.  The calcium-binding association constant is also fairly high when compared to that of other extracellular proteins, though much lower than some cellular calcium binding proteins (Strynadka & James, 1989; McPhalen etal, 1991). The on-rate for calcium binding of (5 ± 2) xlO M s is within two orders of magnitude 7  _l  _1  of the diffusion-controlled limit of ~10 M"' s~' (Fersht, 1985). This implies that the calcium9  binding site of C B D N I is likely mostly preformed with no major structural rearrangement necessary to properly orient the atoms that ligate calcium. The presence of this preformed site is consistent with the relatively high binding affinity of C B D N I f° calcium. The dynamics of this r  region were studied by N relaxation methods, these results are presented in chapter 5. 15  A structural role for calcium binding by CBDNI  The apparent biological role of calcium binding by C B D N I is to stabilise its folded structure. As determined by differential scanning calorimetry (DSC), calcium binding increases the temperature of denaturation of C B D N I by 8 °C, from 49 °C to 57 °C (Louise Creagh & Charles Haynes, personal communication). The putative calcium-binding site of C B D N I ties together three segments of the protein chain that are remote in primary sequence. Having the calcium ion coordinated by ligands from various regions of the protein is common among proteins that are stabilised by calcium binding (Strynadka & James, 1989). The stabilisation of the folded structure by calcium binding also supports the choice of potential calcium-binding sites in C B D N I - The alternate site considered contains ligands solely in the N-terminal region. Binding by these residues does not seem as likely to stabilise the protein as ligands from different regions of the  Chapter 4-Calcium Binding by CBDNI  protein. It is also of note that  ill  CBDN2,  which does not bind calcium, has a higher denaturation  temperature than both the apo and calcium-loaded forms of C B D N I This stabilising role of calcium binding is supported by the finding that there is no significant difference between the K values for cellopentaose binding by C B D N I 20% calciuma  loaded or fully saturated. This indicates that the binding of oligosaccharides is not dependent upon whether or not calcium is bound. Although the effect calcium has on the binding of other oligosaccharides by C B D N I , d the effects of other metals were not checked, it is safe to assume an  that the results would be similar. This is not too surprising in light of the fact that the binding site for calcium is on the opposite side of the protein from the sugar binding site. Large structural changes with calcium binding would have to occur in order to affect the orientation of the residues on the binding face. For the class of lectins whose binding to their sugar ligand is calcium dependent, many of the structures show that the calcium ion is bound close to the sugar binding site. Often in these proteins, calcium mediates hydrogen bonding between the protein and sugar (Rini, 1995) The biological role of calcium stabilising the tertiary structure of C B D N I is shared by the Bacillus l,3-l,4-(3-glucanase enzymes. Seven native and hybrid l,3-l,4-(3-glucanase studied by  Welfle et al. (1994; 1995; 1996) all had higher denaturation temperatures and higher Gibbs free energies in the presence of calcium than EDTA. It was also found that the hybrid Bacillus 1,3-1,4(3-glucanase H(A-16-M) was more stable to guanidinium chloride denaturation calcium-loaded than calcium free (Keitel et al, 1994). Very little is known about the role calcium binding has in other CBDs. As mentioned in the introduction to this chapter, a calcium ion has been identified in the crystal structures of both CBDcip (Tormo et ai, 1996) and CBD 4 (Sakon et al, 1997). As was found for C B D N I , in both E  these structures the calcium ion is located far from the putative binding face. It is therefore likely that calcium has the role of stabilising the structure of these proteins, and does not influence cellulose binding.  128  Chapter 5 Dynamic Analysis of Ligand Binding  Abstract  Backbone amide N and side-chain methyl-containing H relaxation techniques were used 15  2  to investigate the effects of calcium binding and oligosaccharide binding on the dynamics of CBDNI.  15  N spin-lattice relaxation times (Ti), spin-spin relaxation times  (T2) and  heteronuclear  NOEs were determined for the uniformly N labelled protein. H Ti and Tip values were 15  2  determined for a sample of fractionally deuterated N C C B D N I - The N data were analysed 15  13  15  using the model-free formalism to derive the model-free parameters (S , T , R x, Sf and T ) for 2  2  E  e  S  individual backbone N-H bond vectors and values for the overall rotational correlation time (T ). M  Calcium binding dramatically reduces the values of R terms found for residues in the N-terminal ex  region of C B D N I that forms one part of the calcium-binding site. This implies that motion on the millisecond time scale is reduced upon calcium binding. Upon cellopentaose binding, R terms ex  for a set of residues that lie at the end of the (3-strands that form the binding face decrease in value. This also indicates that motion on the millisecond time scale of these is reduced upon cellopentaose binding. The deuteron relaxation techniques probe the nanosecond-picosecond time scale dynamics of methyl containing side-chain residues. It is established that methyl groups present on the binding face have a high degree of mobility in both the free and cellopentaose-bound form. It is hypothesized that this mobility is needed for C B D N I to recognize and bind different orientations of the same ligand, as well as different oligosaccharides.  Chapter 5-Dynamic Analysis of Ligand Binding  129  Introduction  Background  The determination of a three-dimensional structure of a protein, by either x-ray crystallography or NMR methods, can give the impression that its folded structure is static. In reality proteins are dynamic entities. Motion occurs on a wide variety of time scales, from rapid fluctuations of bonds and torsion angles, to conformational transitions involving the collective motion of a large number of atoms. The fast motions are characterized by frequencies on the order of gigahertz while the slower motions occur in the range of milliseconds to microseconds. The dynamics, or internal motions that occur in proteins are dramatically shown in the case of T4 lysozyme. A benzene molecule rapidly enters an engineered solvent-inaccessible hydrophobic cavity in variants of this protein that are 5-6 A from the surface (Feher et al, 1996). Benzene stabilises the protein by binding in the hydrophobic cavity created by the mutations Leu99Ala and Metl02Ala (Eriksson et al, 1992a; Eriksson et al, 1992b). This implies that relatively large fluctuations in the conformation of the protein must occur for the benzene molecule to enter the interior of this protein. The dynamics of proteins can be an important part of their functional behaviour. The dynamic behavior of proteins has been related to protein folding, ligand binding, enzymatic ability and allosteric regulation (Creighton, 1993). In the case of a binding protein , like C B D N I , a number of different possibilities can be imagined where the mobility of a region, or of specific residues, in the protein plays an important role. For instance, residues in the binding site of a protein can be disordered in the unbound state. Here the residues involved in binding are exploring many different conformations. If this region becomes more rigid upon binding, an entropic cost is entailed that must be compensated by other favorable interactions. In this chapter the dynamic behaviour of C B D N I is studied using the well developed NMR method of amide N relaxation and the recently developed technique of methyl containing 15  Chapter 5-Dynamic Analysis of Ligand Binding  130  deuteron relaxation. In particular, the changes that occur in the dynamic properties of C B D N I upon calcium and oligosaccharide binding are studied.  N  15  relaxation and dynamics  When a population of spins is disturbed by an RF pulse, the system will return to the equilibrium state. This process is the phenomenon of relaxation. Two time constant, T] and T2, are used to describe the relaxation process. Ri (1/Ti), spin-lattice or longitudinal relaxation, is the rate by which the spins return to the equilibrium population. T\ relaxation is associated with population differences and involves an exchange of energy with the surroundings. R2 (I/T2), or spin-spin relaxation, describes the rate by which the spins lose their phase coherence in the transverse plane. NMR relaxation is caused by fluctuating local magnetic fields experienced by the nuclei that arise from molecular motion. In liquids the most important relaxation mechanism is the dipoledipole interaction between neighboring spins. Chemical shift anisotropy (CSA) can also act as an efficient relaxation mechanism. CSA arises from the fact that different orientations of a nucleus with respect to the external magnetic field have different chemical shifts. Incomplete averaging of these shifts by molecular tumbling results in relaxation. The power available to cause relaxation resulting from motion at a certain frequency (00), is given by the spectral density, J(co). The Tj and T2 relaxation times and the steady-state heteronuclear NOE enhancement of an amide N nucleus are dominated by the dipolar interaction of the N nucleus with its directly 15  15  attached proton. A significant contribution due to chemical shift anisotropy is also present. The rates of these relaxation processes are given by Abragam (1961) (after Farrow et al, 1994): 1 = d [J((o -co )-t-3J(co ) + 6J(co +co )] + c J(co ) R, = — O2  ?  H  N  N  H  N  z  N  (5.1)  131  Chapter 5-Dynamic Analysis of Ligand Binding  2  2  R = ^ ~ = ^rWKO) + J(co - w ) + 3J(co ) + 6J(co ) + 6J(co + co )] + ^[3J(co ) + 4J(0)] (5.2) 19 2 6 2  H  NOE = l +  N  N  H  H  N  N  d [6J(u) + co ) + 3J(co )- J(co - oo )]T,  (5.3)  2  N  H  N  H  N  The constants d and c represent the dipolar and chemical shift anisotropy contributions, defined 2  2  as:  and C =^|YNH O(AO- ) 2  2  (5.5)  2  In these equations h is Plank's constant, yn and YN are the gyromagnetic ratios of the 'H and N l5  nuclei, respectively, CUH and CON are the *H and N Larmor frequencies, TNH is the internuclear 15  'H- N distance (1.02 A, a value that is essentially invariant), Ho is the magnetic field strength and 15  Aa is the difference between the parallel and perpendicular components of the axially symmetrical 15  N chemical shift tensor. The assumption of an axially symmetric chemical shift tensor has been  shown to be valid for peptide bonds, with A a equal to -160 ppm (Hiyama et al,  Figure  1988).  5.1 shows the dependence of T\, T2 and heteronuclear NOE for a N-H pair on the motion of a 15  N  rigid molecule. The molecular motion is described by the rotational correlation time, x . It is m  defined as the average time for the molecule to rotate by one radian. This figure clearly shows the motional dependency of these experimentally measurable parameters. The aim of this chapter is to relate Ti, T 2 and heteronuclear NOE values to the dynamics, or motions, of  CBDNI-  In the absence of any further assumptions, the three experimentally measured parameters (Tj, T2 and N-'H NOE) do not provide enough information to enable direct determination of the 15  spectral density function at the five frequencies  (0, CON, <*>H> ( ^ H - C°N)  and (COH +  ©N))  of  equations 5.1 to 5.3. To relate relaxation data to the motional properties of the protonheteronucleus bonds, specific models have been employed to provide an analytical form of the spectral density function (Peng & Wagner,  1994).  The most widely adopted, and simplest,  Chapter 5-Dynamic Analysis of Ligand Binding  11  10  132  9  -log  8  T  ?  m  Figure 5.1. Plot of the heteronuclear N O E and log Tj (i=l,2) vs. log t  m  for N relaxed by ' H - N dipolar and CSA interactions at a nitrogen frequency of 50.7 MHz. This figure is reproduced from Kay et al. (1989). 1 5  1 5  method of analysing relaxation data is the model-free formalism of Lipari and Szabo (1982a,b). This method employs a minimum number of parameters to describe the overall tumbling of the molecule, and the internal motions of the i H - ^ N amide bond vector using the following expression:  ST 2  (l+cfl^)  (1-S )T (l+a> x ) 2  • +  2  2  (5.6)  S is the order parameter squared which describes the degree of spatial restriction of the internal 2  motion of the i H - ^ N bond vector. S can vary from 0 for completely isotropic motion, to 1 for 2  133  Chapter 5-Dynamic Analysis of Ligand Binding  completely restricted motion. An effective correlation time for the more rapid local motion of the 1 5  N - ' H vector, as compared to x , is described by T where: E  m  I-_L _L  (5.7)  +  Equation 5.6 assumes that the overall tumbling of the molecule is isotropic. In the case where T x  m  E  equation 5.6 reduces to  J(co) = l + coV  (5.8)  m  The model-free analysis has been extended by Clore et al. (1990a) to approximate internal motions on two time scales differing by at least one order of magnitude; it is described by the following equation:  J(co) = Here t  s  S T„  ( S  2  l + (03X )  2  M  2  •' +1 +  - S  2  ) T  (03T)  (5.9)  2  is the effective correlation time for slow internal motions, and is included in the  relationship — = —+ — . S is expressed as the product of two order parameters (S = Sf S ) 2  2  2  2  s  describing fast and slow internal motions, Sf and S , respectively. 2  2  s  An additional term, R , can be incorporated when modeling the observed T 2 values to take ex  into account the contributions from processes other than those from dipole-dipole (DD) and chemical shift anisotropy (CSA). In many cases, these contributions arise from conformational exchange averaging occurring on the microsecond to millisecond time scale.  (5.10)  Chapter 5-Dynamic Analysis of Ligand Binding  134  Methyl group deuteron relaxation and dynamics  Recently, Kay and co-workers have developed experiments to study nanosecond-topicosecond side-chain dynamics based on the fractional incorporation of deuterium into N and 15  C-labelled proteins (Muhandiram et al, 1995). The key to the use of deuterium as a probe of  13  molecular dynamics lies in the fact that the energy of a deuteron depends critically on its local environment. This quadrupolar interaction is well understood and the interpretation of deuterium relaxation is simpler than is the case for many other nuclei. These deuterium-based relaxation experiments specifically select for  l 3  CH2D  methyl  groups present in the protein. The data are recorded as a series of two-dimensional constant-time 'H- C HSQC experiments in which the methyl peak is dependent upon the deuterium Ti(D) and 13  Tip(D) times. These values are determined by measuring relaxation of terms of the form I C D Z  Z  Z  and I C D , as well as the relaxation rate of I C , where I , C and D denote the z magnetization z  z  y  Z  Z  z  z  z  of the methyl proton, carbon, and deuteron, respectively. D is the y component of deuterium y  magnetization. The theoretical validity of these experiments, and the justification for measuring deuterium relaxation in C H 2 D versus  CHD2  groups, are discussed in Muhandiram et al. (1995)  and Yang & Kay (1996). Since the decay of the triple spin terms I C D and I C D are dominated by deuterium Z  Z  Z  z  z  y  relaxation, the following relations are excellent approximations for T](D) and Ti (D): p  1  1  T,(D)  T (D) lp  1  T,(LC D ) T,(I CJ Z  Z  T (I C D ) lp  z  z  y  (5.11)  Z  T,(LC )  (5.12)  Z  The measured Ti(D) and Ti (D) times are related to motional properties at specific sites in p  the protein through their dependence on the power spectral density function, J(co), according to the following equations (Abragam, 1961; after Kay et al, 1996):  135  Chapter 5-Dynamic Analysis of Ligand Binding  T,  _3_VqQ^ 16  VqQ^ 32 V h  2  (5.13)  [j(co ) + 4J(2co )] D  D  2  (5.14)  [9J(0) + 15J((D ) + 6J(2co )] D  D  j  e qQ .is the quadrupole coupling constant, which is 165 kHz for methyl group deuterons. n 2  Similarly as was described in the previous section for N relaxation, J(co) can be expressed as 15  (Lipari & Szabo, 1982a,b):  J«D) =  f  sx 2  m  l + (cox )  (5.15)  • +  ' ^(cox;) i >  2  m  2  J  here Sj is an order parameter for the methyl group i. S; describes the spatial restriction of motion of the C- H bond vector in the methyl group on the nanosecond-picosecond time scale, % is the l3  2  m  overall correlation time, and — —+ —^—, with T j the effective correlation time describing the e  internal motions for the C- H bond vector i. The order parameter of the bond vector about 13  2  which the methyl group rotates is denoted  S is ax  an  d is related to Sj by Sj = 0.111 S j , 2  2  ax  s  assuming tetrahedral geometry for the methyl group (Nicholson et al., 1992). A value of 1 for S axis 2  indicates complete restriction of motion of the methyl averaging axis, while an S j of 0 2  ax  s  indicates complete freedom of motion. This technique of measuring methyl deuteron relaxation rates has been previously applied to study the interaction of a Src-homology-2 (SH2) domain with a phosphorylated tyrosinecontaining peptide (Kay et al, 1996). In this study, the authors showed that motion of methyl groups in the phosphotyrosine binding pocket were restricted upon binding. Methyl groups in other regions of the binding site remained disordered in both the bound and free state. Recently, the methyl deuteron relaxation rates as a function of temperature were measured for the N-terminal SH3 domain from drk (Yang et al, 1997).  136  Chapter 5-Dynamic Analysis of Ligand Binding  Experimental Methods  Sample preparation for N relaxation experiments i5  15  N labelled C B D N I samples for the N relaxation experiments were produced as outlined 15  in Chapter 2. In many cases, the same N C B D N I sample was recycled by removing added 15  cellooligosaccharides. This was accomplished by unfolding the protein with 6 M urea directly in a microsep concentrator (Filtron) with a IK cut-off membrane. With the protein completely unfolded, all the sugar is free in solution and passes through the membrane. Extensive buffer exchange removes the urea, allowing protein refolding. The validity of this approach was confirmed by NMR spectroscopy. Spectra of the original and refolded samples are identical with no evidence of any residual soluble unfolded protein or sugar remaining after buffer exchange.  Preparation of fractionally deuterated CBD^j  A sample of N , C C B D N I fractionally deuterated to a level of approximately 40% l5  l 3  (subsequently referred to as N/ C/ H(40%) 15  13  2  CBDNI)  was prepared by expressing the plasmid  pTugNln in Escherichia coli JM101 cells grown at 30 °C using a IL fermentor. The M9 minimal media contained 2g/L of [ C ] glucose, lg/L NH S0 and lg/L C (99%) / N (99%) / H 13  15  13  6  4  l5  2  4  (50%) Celtone algal extract (Martek) in 400 mL D 0 / 600 mL H 0. 2  2  The inoculant for the 1 L culture was obtained by culturing a single colony from a plate in 5 mL TB media (100%  H2O)  for 6 hours to an OD600 of Ll. One mL of this culture was spun  down, resuspended in 10 ml of TYP media in 40% D2O/60% H2O, and grown for 2.5 hours to an OD600  °f 1-0- The entire culture was spun down and resuspended in 100 mL of M9 media  containing lg/L of unlabelled Celtone in 40% D2O/60% H2O. After growth for 4 hours to an OD600  of 0.9, all the cells were spun down and used to inoculate 1 L of the labelled media. This 1  L bacterial culture was grown for 5 hours to an ODgfjO of 1.5, at which point protein expression was induced with the addition of IPTG to a final concentration of 0.5 mM. The culture was incubated for 31 hours, the final OD600 being 5.6. All the culturing was performed at 30 °C.  Chapter 5-Dynamic Analysis of Ligand Binding  Purification of N/ C/ H(40%) l5  13  2  137  CBDNI  was as described previously (Johnson et al,  1996a; chapter 2) with the exception that the osmotic shock fractions were not subjected to Avicel binding. Instead they were combined with the H2O elution fractions from Avicel binding and loaded onto the ionic exchange column. After purification, 19 mg of N/ C/ H(40%) I5  I3  2  CBDNI  was obtained from 1 L of culture.  N  15  relaxation experiments  All spectra measuring N Ti, N T2 and 'H- N steady-state NOE values were acquired 15  15  15  using the pulse sequences described in Farrow et al (1994). These enhanced sensitivity pulse sequences employ pulsed field gradients to minimize artifacts, suppress the solvent signal and select for the coherence transfer pathway from N to H for observation. Additionally, a selective 15  1  water flip back pulse is incorporated to ensure minimum perturbation of the water magnetization. In total 11 different N relaxation series are reported here (table 5.1). All spectra were l5  recorded at 35 °C using either a Varian Unity 500-MHz spectrometer or a 600 MHz Varian Inova spectrometer (Prof. Lewis Kay, University of Toronto). Two different sample buffers were used. Initially, the buffer was 50 mM sodium chloride, 50 mM potassium phosphate (pH* 5.9), 0.02% sodium azide, 10% D2O / 90%  H2O.  When it was found that C B D N I bound calcium (chapter 4),  the buffer used in subsequent experiments was 50 mM sodium chloride, 50 mM sodium acetate (d3) (pH* 6.0), 0.02% sodium azide, 10% D 0 / 90% H 0. Samples 6 to 11 are calcium loaded 2  2  to a level of approximately 20% (chapter 4). The data for apo cellooligosaccharide-free  CBDNI CBDNI  (sample 1) were obtained by collecting spectra on in the presence of 3.8 molar equivalents of sodium  ethylenediamine tetraacetate (EDTA). Samples 2 and 4, 3 and 5, and 6 and 7 were the same samples, respectively, with solid cellopentaose added to prevent dilution of the protein sample. Samples 9, 10 and 11 were from a common pool of protein that was split in two. Using one half of the sample the unbound relaxation series was run, then solid cellopentaose added. To the other half solid cellotetraose was added.  138  Chapter 5-Dynamic Analysis of Ligand Binding  Table 5.1. N relaxation experiments performed on C B D N I 500 MHz a 600 MHz sample # sample description protein cone. buffer protein cone, buffer (mM) (mM) Apo; Calcium and sugar-free A 1.3 1 Calcium-bound A 2 1.3 3 Calcium-bound 1.9 A 4 Calcium and Cellopentaose- bound A 1.3 Calcium and Cellopentaose- bound ' 5 1.9 A Dilute, sugar-free 0.3 B 6 Dilute, Cellopentaose-bound 0.3 B 7 B Concentrated sample 5.3 8 1.6 B Sugar-free 9 1.6 B Cellopentaose-bound 10 B Cellotetraose-bound 1.6 11 Data collected using N-labelled C B D N I samples. Data collected using N / C / H (40%) CBDNI(A): 50 mM sodium chloride, 50 mM sodium acetate (d3) (pH* 6.0), 0.02% sodium azide, 10% D 0 / 90% H 0. (B): 50 mM sodium chloride, 50 mM potassium phosphate (pH* 5.9), 0.02% sodium azide, 10% D 0 / 90% H 0. Samples in buffer B are approximately 20% calcium bound. 15  b  c  a  15  b  c  15  I 3  2  c  2  2  2  2  For sample 3, the ratio of protein-to-calcium was 1:2; for sample 2 this ratio was 1:5. Using the experimentally determined association constant of C B D N I with calcium (113 000 M" ; 1  chapter 4) it is calculated that 99.5% of C B D N I is calcium-loaded in sample 3, and 99.8% calciumloaded in sample 3. In the case of the cellooligosaccharide-bound samples, sugar was added to such a level that the fraction of protein that was ligand-bound was 95%. The weight of sugar needed to achieve this level was calculated using the association constants previously determined (Tomme et al, 1996a; Johnson et al, 1996a; Chapter 2). N-labelled C B D N I samples were used for experiments collected at 500 MHz.  I5  15  N Tj  values were recorded with nine different durations of delay T= 11.1, 33.3, 55.5, 99.9, 166.5, 266.4, 410.7, 643.8 and 999.0 ms. T values were determined from spectra recorded with the 2  Chapter 5-Dynamic Analysis of Ligand Binding  139  delays T= 16.7, 33.4, 50.1, 66.8, 83.5, 100.2, 116.9, 133.6 and 150.3 ms. For each series, the spectra were acquired in mixed order. To test the reliability of the data, duplicate spectra were initially recorded for T= 55.5 ms (T[) and T= 50.1 ms (T ). 'H- N NOE values were determined 15  2  from spectra recorded in the presence and absence of a proton presaturation period of 3 s. In the case of the control no-NOE spectra, a net relaxation delay of 5 s was used, while a relaxation delay of 2 s prior to a 3 s proton presaturation period was used for the NOE spectra. Spectra were acquired using spectral widths of 6500 Hz and 1450 Hz in the 'H and N dimensions, 15  respectively. Spectra were acquired as either 96 x 1024 or 96 x 512 complex matrices. For experiments collected at 600 MHz, data were acquired using the N / C / H (40%) 13  l5  CBDNI  2  sample. Prior to data collection this deuterated sample was heated to 90 °C to unfold the  protein. This allowed any amides which might be residually deuterated from expression in 40% D 0 to fully exchange with protons in the sample buffer. Thermal unfolding of CBDNI is a 2  completely reversible process (chapter 2; Creagh et al, 1997). N Ti values were collected with I5  delays T= 11.02, 66.12, 132.25, 220.42, 308.58, 396.75, 517.98, 650.23 and 804.52 ms. T  2  values were determined from spectra recorded with the delays T= 16.66, 33.32, 49.98, 66.64, 83.30, 99.96, 116.61, 133.27 and 149.93 ms. In the case of the control no-NOE spectra, a net relaxation delay of 6.5 s was used, while a relaxation delay of 3.5 s prior to a 3 s proton presaturation period was used for the NOE spectra. Data were acquired using spectral widths of 9000 Hz and 1720 Hz in the 'H and N dimensions, respectively. Spectra were C decoupled 15  13  and acquired as 128 x 576 complex matrices.  Methyl group deuteron relaxation experiments  NMR experiments were performed using a 600 MHz Varian Inova spectrometer (Prof. Lewis Kay, University of Toronto) on a 1.9 mM sample of N/ C/ H(40%) CBDNI in 50 mM 15  13  2  sodium chloride, 50 mM sodium d-acetate, 0.02% sodium azide, 4.2 mM calcium chloride, 10% 3  D 0 / 90% H 0, pH* 6.0, 35 °C in the presence and absence of 3.2 mM cellopentaose. Pulse 2  2  sequences used for the measurement of the relaxation properties of methyl group deuterons were  140  Chapter 5-Dynamic Analysis of Ligand Binding  those described by Muhandiram et al. (1995). To obtain T](I C D ) values, nine two-dimensional Z  Z  Z  ' H - C correlation spectra were acquired with delays of 0.05, 4.1, 8.6, 13.5, 19.0, 25.1, 32.1, 13  40.3, and 50.0 ms. Values of Ti(I C ) were obtained with identical delays. Values of z  z  Ti (I C D ) were obtained using delays of 0.20, 1.3, 2.8, 4.3, 6.1, 8.0, 10.3, 12.9, and 16.0 p  Z  z  y  ms. The same delays were used when collecting data on both the free and cellopentaose bound forms of C B D N I . All spectra were obtained as 128 x 576 complex points with spectral widths of 5000 and 9000 Hz in the C and 'H dimensions, respectively. L3  Data processing  Both amide N relaxation and methyl group deuteron relaxation experiments were , 5  processed using NMRPipe (Delaglio et al, 1995). Lorentzian-to-Gaussian apodization functions were applied in both dimensions. For the constant time 'H- C HSQC experiments acquired to I3  measure methyl group deuteron relaxation, mirror image linear prediction was employed in the tl ( C) dimension (Zhu & Bax, 1992). Relative peak volumes were measured using the nlinLS 13  module built into NMRPipe. A very conservative approach was employed when peak picking the spectra. Data from peaks that lie close to each other, but not totally overlapped, were not used if their measured Ti or T2 values were similar. This results in a relatively large number of missing data points, but a high level of confidence can be placed in the final data. Using relative peak volumes and the delay values (T) as input for the programme lmquick (Dr. Neil Farrow, University of Toronto), per-residue N T-\ and T2 and methyl group H 15  2  Ti(I C ), Ti(I C D ) and T[ (I C D ) values were obtained, lmquick fits the relative peak z  z  z  z  z  p  z  z  y  volumes to a function of the form I(T) = 1(0) exp(-T/Tj), where 1/T; is the relevant relaxation rate, I(T) is the relative peak volume at time T and 1(0) is the intensity at time T=0. Errors in the measured relaxation rates are estimated using Monte Carlo procedures as described in Farrow et al. (1994).  Chapter 5-Dynamic Analysis of Ligand Binding  Steady-state ' H peaks w i t h ( I  s a t  I 5  141  N N O E values were d e t e r m i n e d f r o m the ratios o f the intensities o f the  ) a n d w i t h o u t proton saturation (I nsat)U  NOE = I  / Iunsat  s a t  T h e standard d e v i a t i o n o f the N O E v a l u e (o~ oe)  w  a  d e t e r m i n e d o n the basis o f the m e a s u r e d  s  n  b a c k g r o u n d noise levels using the f o l l o w i n g relationship  f( ^noe  NOE  2 t  _  K  t  1  J  fa  unsat  v. ^ u n s a t  ^ y  2 \ 1/2  J  A r o u t i n e i n N M R P i p e w a s used to measure the rms n o i s e o f b a c k g r o u n d r e g i o n s o f the spectra ( O t and a s a  unS  at)-  F o r the  1  5  N r e l a x a t i o n a n a l y s i s , a v a l u e for the o v e r a l l c o r r e l a t i o n t i m e for the m o l e c u l e  ( x ) w a s c a l c u l a t e d u s i n g the p r o g r a m m e tm_f77 ( D r . N e i l F a r r o w , U n i v e r s i t y o f T o r o n t o ) . T h i s m  m e t h o d uses a g r i d search i n t , o p t i m i z i n g x and S m  e  2  for each residue (equation 5.6). D a t a w i t h a  T1/T2 r a t i o d i f f e r i n g b y m o r e than one standard d e v i a t i o n f r o m average w e r e not u s e d i n the process o f d e t e r m i n i n g x  m  values. O n c e an o v e r a l l c o r r e l a t i o n t i m e was f o u n d , the data were fit to  the different spectral density functions o u t l i n e d i n the i n t r o d u c t i o n to this chapter (equations 5.6 a n d 5.10).  T h e p r o g r a m m e j _ f 7 7 ( D r . N e i l F a r r o w , U n i v e r s i t y o f T o r o n t o ) w a s u s e d for this  process. F i v e different m o d e l s o f the s p e c t r a l d e n s i t y f u n c t i o n w e r e fit to the e x p e r i m e n t a l l y determined  1  5  N b a c k b o n e T j , T2, a n d ' H -  1 5  N N O E values. T h e s e are: (1) a f u n c t i o n o f the f o r m  o f e q u a t i o n 5.8 w i t h x f i x e d at z e r o ; (2) a f u n c t i o n o f the f o r m o f e q u a t i o n 5.6 w i t h x e  e  used as a  fitting parameter; (3) a function o f the f o r m o f equation 5.6 w i t h x f i x e d at zero and i n c o r p o r a t i n g e  an R  e  x  t e r m ; (4) a f u n c t i o n o f the f o r m o f e q u a t i o n 5.6 w i t h x  e  and R  e  x  terms u s e d as fitting  parameters; a n d (5) the t w o t i m e scale v e r s i o n o f the spectral density f u n c t i o n g i v e n i n e q u a t i o n 5.9. T a b l e 5.2 summarises these m o d e l s a n d the parameters fit i n each o f them. T h e m e t h o d for m o d e l s e l e c t i o n w a s to p r o c e e d f r o m the s i m p l e s t ( m o d e l 1) to the m o r e c o m p l e x m o d e l ( m o d e l 4) as f o l l o w s ; (i) a m o d e l was d i s c a r d e d i f a l l the e x p e r i m e n t a l parameters ( T i , T2, ' H -  I 5  N N O E ) c o u l d not be fit to w i t h i n the 9 5 % c o n f i d e n c e l i m i t ; (ii) the v a l u e o f the  Chapter 5-Dynamic Analysis of Ligand Binding  142  fitting parameter must exceed its calculated error; (iii) if both (i) and (ii) are satisfied the model with the lowest x for the fits to the experimental data was chosen; (iv) finally, if none of models 1 to 4 2  (Table 5.2) satisfied criteria (i) and (ii) the two time scale model (model 5) was chosen provided that points (i) and (ii) were met. Table 5.2. Summary of the model-free parameters for N relaxation analysis model fitting parameters description 1 S Fast isotropic motion of the N-H bond. 2 S,T A fast motion (x) affects T . 3 S,R The amide experiences chemical exchange line broadening. 4 S , x , Rex Fast motions and chemical exchange affect relaxation. 5 S , Sf , x Motions occur on two time scales. l5  2  2  e  m  e  2  ex  2  e  2  2  s  For the analysis of the methyl group deuteron data the values of Tj(D) and T|p(D) were fit only to the first two models of table 5.2 according to equation 5.15. The criteria to select a model was the same as points (i) to (iii) as outlined above.  Resonance assignment of the calcium-loaded  CBDNI  The backbone N and H resonances of calcium-loaded oligosaccharide-free C B D N I were i5  N  assigned from the analysis of the HNCACB (Wittekind & Mueller, 1993) and CBCA(CO)NH (Grzesiek & Bax, 1992) experiments acquired on the C / N / H(40%) C B D N I sample. The 13  15  2  assignment process was facilitated by the use of a pulse sequence to correlate the N and H l5  N  resonances of residues that immediately follow methyl-containing amino acids in N, C and 15  13  fractionally deuterated proteins (Muhandiram et ai, 1997). Assignments of the methyl groups of calcium-loaded oligosaccharide free C B D N I were obtained from analysis of H(CCO)NH-TOCSY (Montelione et al, 1992; Grzesiek et al, 1993a) and (H)C(CO)NH-TOCSY (Logan et al, 1992; Grzesiek et al, 1993a) experiments. The data collection and processing parameters for the NMR experiments to assign calcium-loaded sugar-free  Chapter 5-Dynamic Analysis of Ligand Binding CBDNI  143  is given in table 5.3. All assignment experiments were processed and analysed using Felix  version 2.3 (Biosym). Table 5.3: Acquisition and processing parameters for NMR experiments to assign calcium-bound CBD i Experiment nucleus acq. points spectral width carrier frq. processing matrix dim. (ppm) (real) (complex) (Hz) , 13 a(3 43.0 lp, ss (90°) 38 9179 HNCACB 256 128 30 1720 119.0 lp, ss (80°) t= N t ='HN tdc, gm 256 576 9000 4.68 (15), poly , 13 ap 43.0 lp, ss (80°) 48 9179 CBCA(CO)NH 256 128 t= N 30 1720 119.0 lp, ss (80°) t ='HN tdc, gm 256 4.68 576 9000 (15), poly 4.68 lp, ss (90°) 58 9179 H(CCO)-TOCSY-NH t, = »H 256 128 1720 t= N 30 119.0 lp, ss (80°) t ='HN tdc, gm 256 9000 4.68 576 (15), poly 43.0 lp, ss (90°) 58 9179 C(CO)-TOCSY-NH 256 128 1720 119.0 lp, ss (80°) t= N 30 t='HN tdc, gm 256 4.68 9000 576 (15), poly a  N  t  =  C  ct  15  2  3  ct  t  =  C  15  ct  15  ct  15  ct  2  3  2  3  2  3  Data were collected at 600 MHz using a Varian Inova spectrometer. ct=constant time; lp=linear prediction (or mirror image linear prediction with constant-time); ss=sinebell squared (degrees shifted); sb=sinebell (degrees shifted); gm=Lorentzian-to-Gaussian multiplication with a maximum at approximately 0.1 of the acquisition time (Hz line broadening); poly=polynomial baselineflattening;tdc=time domain convolution. a  Ultracentrifugation and viscosity measurements  Ultracentrifuge sedimentation velocity runs were performed at 20 ° C on 0.35 mM samples of free and cellopentaose-bound C B D N I - A 3.4 fold molar excess of cellopentaose was used, resulting in C B D N I being 94% bound. Sedimentation equilibrium was performed on the bound  Chapter 5-Dynamic Analysis ofLigand Binding  144  form of C B D N I at a protein concentration of 0.23 mM. This complements the measurement run previously performed on an unbound sample of C B D N I (chapter 2). Viscosity measurements were performed on free and cellopentaose-bound forms of  CBDNI  using Cannon-Manning Semi-Micro type viscometers at 35 °C. Viscosity was initially measured using 1.9 mM protein samples. The cellopentaose-bound sample of C B D N I contained an 1.4 fold excess of cellopentaose, resulting in 94% binding. Three dilutions were then made, with the viscosity measured at each step. The buffer used for dilution of the ligand-bound form of  CBDNI  contained cellopentaose at the same concentration as in the undiluted sample. Thus, upon dilution, the relative concentration of cellopentaose-to-protein increased, resulting in 98% saturation after the final dilution. The buffer used for both ultracentrifuge and viscosity measurements was 50 mM sodium chloride, 50 mM potassium phosphate (pH 5.9), 0.02% sodium azide. It is estimated that under these conditions, the protein is approximately 20% calcium loaded (chapter 4). The ultracentrifuge and viscosity measurements were carried out by Les Hicks of the University of Alberta.  Results  N Ti, T and NOE values  l5  2  Tj, T2 and tH- N NOE values were determined for eleven different samples of 15  CBDNI  (Table 5.1). Shown in figure 5.2 are portions of the 2D !H- N HSQC spectra used to determine I5  the T2 values of apo C B D N I (sample 1). Examples of the decay curves to determine T\ and T2 values are shown in figure 5.3. In the analysis of the data obtained for the various samples of CBDNI,  it was not possible to determine the relaxation parameters for each backbone amide. This  is mainly due to resonance overlap, but also, in the case of sample 1, due to extreme line broadening, resulting in weak peaks for a number of resonances. Table 5.4 summarises the number of residues for which data was obtainable for each sample discussed here.  Chapter 5-Dynamic Analysis of Ligand Binding  Figure 5.2. Portion of the 1H-15N correlation spectra recorded to measure T2 values for calcium-free and oligosaccharide-free CBDNI (sample 1) recorded with delay values of (A) 16.7 ms, (B) 50.1 ms, (C) 100.2 ms and (D) 150.3 ms.  er 5-Dynamic Analysis of Ligand Binding  0  I 0  1  I  I  I  L  02  0.4  0.6  05  1  Time (s)  0 I 0  i  i  OJ02  OJ04  i 0:06  i  OJCB  i  0.1  i 0.12  i 0.14  I 0.16  Time(s)  Figure  5.3. Two parameter monoexponential curve fitting of the (A) T | and (B) T data. T j data points for residues D71 (cross), S40 (circle), N50 (diamond) and V48 (square), and T data points for residues D10 (cross), V45 (circle), E l 5 (diamond) and G44 (square) are shown. 2  2  147  Chapter 5-Dynamic Analysis of Ligand Binding  Table 5.4. Number of residues per sampleforwhich N relaxation data was obtained. sample^ 1 2 3 4 5 number of residues analysed 89 81 84 94 80 The expected number is 141. This is based on the number of residues in C B D N I less the number of proline residues and the N-terminal amino group. Refer to Table 5.1 for sample descriptions. 15  13  a  b  The value of the relaxation parameters Tj, T and !H- N NOE are shown in figure 5.4 for 15  2  calcium-loaded C B D N I free and cellopentaose-bound at both 500 MHz and 600 MHz (samples 25). The average value of Ti for sugar-free  CBDNI  is 0.53 s at 500 MHz and 0.65 s at 600 MHz.  The average T value is 0.12 s at 500 MHz and 0.11 s at 600 MHz. For cellopentaose-bound 2  CBDNI  the average Tj values are 0.53 s and 0.68 s for data at 500 MHz and 600 MHz,  respectively. The average T value is 0.12 s for the data collected at both field strengths. For 2  sample 1, the calcium- and sugar-free sample, Tj, T and !H- N NOE values are shown in figure 15  2  5.5. Also shown in this figure, for comparison purposes, are the values for the calcium-loaded form of C B D N I , sample 2. The average value of Tj for sample 1 is 0.52 s, while the average T  2  value is 0.11 s; this compares to 0.53 and 0.12 s for the Ti and T values of the calcium-loaded 2  CBDNI  (sample 2), respectively. The Ti values are very uniform throughout the protein for all the samples. Besides the  termini the only values that significantly differ from average in all samples are residues G12, G23, D71, F132 and S133. Residue L32 has a high T[ values in sample 2, but its value was not measurable in any other sample. In contrast to the relative uniformity of the Tj data, the T values show a greater variation. 2  Residues with a large T value in all samples are G12 and G23. Residue L32 has a large T value 2  2  in sample 2, but is not measurable in any other sample. Residues with T values that are 2  significantly smaller than average in all samples are T8, V17, and V34. For sample 1, residues 14 and G7 have low T values, yet an average value in sample 2. Residues E14 and A66 in sample 1 2  Chapter 5-Dynamic Analysis of Ligand Binding  Cellopentaose-free  Cellopentaose-bound  Residue  Figure 5.4. N T,, T and H - N NOE data for: (left) calcium-bound (samples 2 and 3); (right) calcium-bound cellopentaose-bound (samples 4 and 5) C B D . Open circles show data collected at 500 MHz, filled circles show data collected at 600 MHz. Bars indicate error values that are greater than the radius of the circles. The secondary structure of C B D N I is shown at the top of the diagram. Open and closed boxes indicate p-strands on sheet A (the oligosaccharide binding face) and sheet B, respectively. I 5  [  1 5  2  N )  Chapter 5-Dynamic Analysis of Ligand Binding  1  A  T 1  ]  B  1  6 2  0.50  (s)  _  (il)  41)  20  80  1  ,  120  i  i  g  0  1  „• °« 0  140  • o  -  100  I  °, 1 20  1  I  40  Ml  I  I  80  I  100  120  141)  1  •  o  i 80  _i  20  9  °  41)  fill  8  o  #  • -  i HX>  1 120  1  140  Residue  Figure 5.5. N Tj, T and H - N NOE data for calcium-free C B D (open circles; sample 1) and calcium-bound C B D (closed circles; sample 2). Bars indicate error values that are greater than the radius of the circles. The secondary structure of C B D is shown at the top of the diagram, the p-strands are labelled in this diagram. Arrows indicate the putative calcium-binding residues. 1 5  l  1 5  2  N ]  N J  N 1  Chapter 5-Dynamic Analysis of Ligand Binding  150  also have low T values but no data was obtainable for these residues in sample 2. Residues L139 2  and L141 in sample 1 have a low T , whereas the value of L141 is average in sample 2, while that 2  of L139 was not measurable in sample 2. Residue T87 has a low T value in sample 2 and 3, but 2  is not measurable in samples 4 and 5. Residue G44 has a low T value in samples 2 and 3 and an 2  average values in samples 4 and 5. For all samples, the greatest variation in steady-state 'H- N NOE values lie with the first I5  23 residues. The only noticeable variations after this point arise in the loop regions near residues 100 and 120. The C-terminal region, residues E149 to L152 have very low *H- N NOE values I5  in all samples. When comparing the 'H- N NOE values of the calcium-free and calcium-loaded I5  forms of C B D N I (samples 1 and 2), G12 has a low 'H- N NOE value in both sample 1 and 2. In 15  sample 1 residues 14 and Dil have 'H- N NOE values much lower than average, with their 15  corresponding values in sample 2 being average. Those of residues D10 and G23 are much lower than average for sample 1, but were not measurable in sample 2. In comparing the 'H- N NOE I5  results with and without cellopentaose (samples 2 to 5), residues G12 and G23 are low in all cases, residue D22 is low in sample 3 but not measurable in the other samples. It also appears that there is less scatter in the JH-^N NOE values of the first 23 residues of the samples bound to cellopentaose than in the free form of the protein.  Estimation of the overall correlation time T  m  Values for the overall correlation time (x ) of C B D N I were determined by fitting the m  relaxation data on a per-residue basis using the model-free spectral density function given in equation 5.6 (table 5.5). The values of the correlation time reported are the average of the optimal per-residue x values, using only residues with a T[/T ratio within one standard deviation of the m  2  mean Tj/T value. The error reported is the standard deviation of the x values considered in this 2  data set.  m  151  Chapter 5-Dynamic Analysis of Ligand Binding  Table 5.5. Values for the overall correlation times (x ) determined for C B D N I at 35 ° C . m  sample 1 2 3 4 5 6 7 8 9 10 11  sample description Apo; Calcium-free, sugar-free, 1.3 mM Calcium-bound, 1.3 mM, 500 MHz Calcium-bound, 1.9 mM, 600 MHz Calcium and Cellopentaose bound, 1.3 mM, 500 MHz Calcium and Cellopentaose bound, 1.9 mM, 600 MHz Dilute sample; sugar-free, 0.3 mM Dilute sample; cellopentaose-bound, 0.3 mM Concentrated sample; sugar-free 5.3 mM Sugar-free, 1.6 mM Cellopentaose-bound, 1.6 mM Cellotetraose-bound, 1.6 mM  x (ns) 7.6 ± 0.4 7.5 ± 0.5 7.8 ± 0.3 7.2 ± 0.4 7.1 ± 0.4 7.2 ± 0.4 7.3 ± 0.4 8.9 ± 0.4 7.4 ± 0.4 7.3 ± 0.3 7.2 ± 0.4 m  Equation 5.6 assumes that the tumbling of the molecule is isotropic. This assumption, in the case of C B D N I , is valid when the three-dimensional structure of this protein is considered. CBDNI  (Johnson et al, 1996b; Chapter 3), to a first approximation, is a sphere, with no one axis  being significantly longer than another. This implies that the molecule rotates isotropically in solution.  Tj/T2 analysis  The T1/T2 ratios of C B D N I samples 1 to 5 are shown in figure 5.6. Throughout much of the protein, the T1/T2 values are similar, suggesting isotropic motion of the protein in solution. T1/T2 data can be easily analysed in a qualitative way. T1/T2 values lower than average tend to indicate motions are occurring on two separate time scales, or there are fast motions present (equation 5.10; model 5, table 5.2). T1/T2 values higher than average usually indicate that the T2 values are decreased by a chemical exchange process (R ) (Clore et al., 1990b). ex  In all samples, residues with the highest T1/T2 ratios tend to occur in the N-terminal region of the protein. The highest overall T[/T ratios occur in this region in apo C B D N I (sample 1). 2  C h a p t e r  5 - D y n a m i c  A n a l y s i s  AI  A T/T  •  O  -  5  -  m  u  u  • q&  o  i  A 3  i i  A 4  i M M  i  i  A5  i mm 1  i  • ^aJb oo oo o o°oOOcP oqa* £ o ca>o Q  «0  '  i  11)0  120  1  1  1  1  1  1  i  1  1  1  1  20  40  60  Ml  lKI  120  •  B  Ml  40  u  i  2  20  T/T  ^ A2^  B i n d i n g  i  •  in  L i g a n d  M  •  -  15  o f  -  ooo5>^  -  V  1 14(1  1  •  2  5  1)  C  V  "  J  1  30  Tj/T2 io -5  o  >  1  • •  %  -  1  4  20  140  I  1  40  Ml  1 80  1  1  100  120  %,  \ 140  Residue  Figure 5.6. T1/T2 values for: (A) calcium-bound C B D ; (B) calcium-bound cellopentaose-bound C B D j . Open circles show data collected at 500 MHz, filled circles show data collected at 600 MHz. In (C) calcium-free T\/T values are shown as filled circles and calcium-bound values are shown as open circles. Arrows indicate the putative calcium-binding residues. N 1  N  2  Chapter 5-Dynamic Analysis of Ligand Binding  153  Here residues 14, G7, E14 and V17 all have noticeably higher than average T1/T2 ratios. In the calcium-bound form of C B D N I the T1/T2 ratios in this region are significantly lower, but still contain residues with higher than average values (T8 and VI7). L139 in sample 1 also has a higher than average T1/T2 ratio. For samples 2 and 3 residues T8, V17, V34, G44, N81, T87 and G130 all have higher than average T1/T2 ratios. With cellopentaose binding, the values of these ratios, with the exception of T87 for which data was not obtainable in samples 4 and 5, all decrease. Only T8, V17 and V34 remaining noticeably higher than average in the cellopentaose bound state. The Ti/T ratio of V48 is slightly higher than average in sample 4 and 5 but its value 2  was not significantly changed from what it was in samples 2 and 3.  Model-free analysis of backbone N relaxation data 15  As described in the introduction to this chapter, the relaxation parameters ( Ti, T and 'H2  15  N NOE) of each residue were fit using the spectral density functions given in equations 5.6 and  5.9. An exchange parameter, R , was also included in some models as outlined in table 5.2. The ex  optimum spectral density models were chosen as described in the experimental section. Table 5.6 summarises the number of residues that fall within each class of model for each of samples 1 to 5. Also included is the number of residues for which no model was satisfactorily fit. Table 5.6. Summary of the number of resonances occurring for each spectral density model used to fit N Tj, T2 and *H- N NOE data for the different C B D N I samples. sample model 4 5 2 3 1 4 7 1 2 9 KS ) 2 (S , x ) 15 13 8 11 10 14 4 19 7 8 3 (S2, R ) 31 11 4 (S2, x , R ) 16 15 19 5 (S2, S 2, X ) 21 38 29 47 36 1 7 3 4 6 none 15  15  13  2  2  e  ex  e  T  a  ex  S  See table 5.1 for sample descriptions.  Chapter 5-Dynamic Analysis of Ligand Binding  154  Figure 5.7 summarises the values of the fitting parameters used in the model-free analysis of free and cellopentaose-bound C B D N I at both 500 MHz and 600 MHz. Model-free data for the calcium and sugar-free form of C B D N I (sample 1) are shown in figure 5.8, with the data of sample 2 included for comparison. The average value of the square of the order parameter (S ) over all 2  residues for sample 1 is 0.79, for sample 2 this is 0.81, for sample 3 it is 0.82, sample 4, 0.80 and for sample 5 it is 0.79. The overall trends in the order parameter are similar in all forms of the protein. While there is variation in the S value for each sample, the only region of the protein 2  where the S values are consistently low is at the C-terminus, residues 148-152. 2  A x parameter was needed to fit the data throughout much of the protein in each sample. e  For all samples, longer than average T values are found at the N-terminal region up to residue 15 E  and residues G44 to V48 in samples 1 to 3 . Longer x values for the N-terminal residues are e  found in the calcium free sample than in the calcium-loaded sample. The most significant difference between  T  E  values in sugar-free and cellopentaose-bound forms of C B D N I is the  reduction in the T values for residues G44 to V48 in the sugar bound form. E  Many of the residues in C B D N I need an R term to correctly model the experimental data. ex  For all samples, R terms tend to occur at similar locations in the structure although with different ex  values. For sample 1, the R terms are highest in the N-terminal region up to residue V17, and at ex  residues L139 and C140. Obtaining data for sample 1 in the N-terminal region was hampered by fact that peaks in this region are in general very weak due to line-broadening. This line-broadening is less evident in the calcium-loaded sample. Line-broadening is indicative of the presence of motion on the chemical exchange (Rx) time scale (millisecond). For samples 2 and 3 the R e  ex  terms are largest at residues V17, V34, G44 and T87. The R term for these residues, and in ex  general for all residues, is lower in samples 4 and 5.  Chapter 5-Dynamic Analysis of Ligand Binding  Cellopentaose-bound  Cellopentaose-free I. 00  S  0.75  2  0.50 0.25 (1.00 0.08 II. 06  0.02 0.00 12  9  J  1  i  • • 0  1  1  1  • o  °  • ••••  mmmm »  f  oo  1.00  1.00  0.75  0.75  0.50 50 -  0.50  0.25  0.25  fc»  3  ^  0.00  40  60  80  Residue  0.00  40  60  ^ .  g  o .  80  Residue  5.7. Model-free parameters for: (left) calcium-bound (samples 2 and 3); (right) calcium-bound cellopentaose-bound (samples 4 and 5) C B D j . Open circles show data collected at 500 MHz, filled circles show data collected at 600 MHz. Bars indicate error values that are greater than the radius of the circles. The secondary structure of C B D is shown at the top of the diagram. Open and closed boxes indicate  Figure  N  N 1  (3-strands on sheet A (the oligosaccharide binding face) and sheet B, respectively.  Chapter 5-Dynamic Analysis of Ligand Binding  Al  BI  B2  1.00 0.75 0.50 0.25 -I 0.00  20  120  100  60  40  140  (ns)o.io 0.05  4  7  50  21  is  R ex  15  12  9  -I  I  6  3  0  20  40  60  20  40  60  80  100  120  140  100  120  140  1.00 0.75 -j 3  f  0.50 0.25 -I 0.00  Residue  Figure 5.8. Model-free parameters for calcium-free C B D (open circles; sample 1) and calcium-bound C B D (closed circles; sample 2). Bars indicate error values that are greater than the radius of the circles. The secondary structure of C B D is shown N 1  N 1  N 1  at the top of the diagram, the fi-strands are labelled in this diagram.  Chapter 5-Dynamic Analysis of Ligand Binding  157  Ultracentrifuge and viscosity measurements  Ultracentrifuge sedimentation velocity runs were performed on 0.35 mM samples of free and cellopentaose-bound forms of C B D N I using identical sample conditions. Both samples displayed single, symmetrical Schlieren peaks. Assuming an average partial specific volume of 0.73 cm g" for the bound and free forms of the protein, the calculated S20,co values are 1.60 for 3  1  the free sample and 1.70 for the cellopentaose-bound form of  CBDNI-  This slight increase in the  S20,co value is just beyond the range of expected experimental error. Based on sedimentation  equilibrium measurements, cellopentaose-bound C B D N I is monomeric protein with an apparent a  molecular mass of 15.1 kDa. This compares to a value of 14.1 kDa found previously for unbound CBDNI  in identical conditions (chapter 2). The viscosities of samples of both free and cellopentaose-bound C B D N I were measured at  1.9 mM and 1.8 mM, respectively, and also at three dilution levels. Also measured was the viscosity of the two buffers, with and without cellopentaose, used for the dilutions. This data is shown in figure 5.9. Fitting this data to a straight line results in slopes of 0.0471 cSt mM" and 1  0.0459 cSt mM , and y-intercepts of 0.762 cSt and 0.759 cSt for cellopentaose-bound and -1  unbound C B D N I , respectively.  Resonance assignment of calcium-loaded  CBDNI  Previously, resonance assignments of cellotetraose-bound and calcium-bound C B D N I were obtained (chapter 3; Johnson et al 1996b). However, at the time, it was not know that calcium was bound by the protein. By following the titration of C B D N I with the calcium contaminated cellotetraose in reverse, to the level where C B D N I was approximately 20% bound, then following a separate titration of apo-CBDNi with calcium, it was possible to assign unambiguously most of the resonances in a H- N ]  15  HSQC  spectrum of calcium-bound sugar-free C B D N I • To resolve the  ambiguous assignments, HNCACB and CBCA(CO)NH experiments were run on the calciumbound cellopentaose-free C/ N/ H(40%) C B D N I sample. The triple-resonance data obtained 13  15  2  from this 1.9 mM fractionally deuterated sample were superb, an example being shown in figure  Chapter 5-Dynamic Analysis of Ligand Binding  0.85  0.80  h  0.75„  0.50  1.0  1.5  2.0  Protein Concentration (mM)  0.0  .0  2.0  3.0  4.0  Protein Concentration (mM) Figure 5.9. (top) Values of viscosity plotted against protein concentration. Data collected in the presence of cellopentaose are shown by squares and a long dashed line. Data collected in the absence of oligosaccharide is shown by circles and a solid line, (bottom) Values of the overall rotational correlation time (x ) (diamonds; short dashed lines) of samples 6 (0.3 mM), 8 (5.3 mM) and 9 (1.6 mM), and values of viscosity plotted against protein concentration. m  Chapter 5-Dynamic Analysis of Ligand Binding  159  3.3. The assignment process was straightforward. A labelled tH-^N HSQC spectrum showing these assignments is shown in figure 5.10. The resonance assignment process was aided by the development of a new experiment that provides N-H correlations for residues that immediately follow methyl-containing amino acids 15  N  (Muhandiram et al, 1997). This triple-resonance experiment requires an N, C and fractionally 15  l3  deuterated protein sample. The correlations shown by this experiment provide a means of helping to confirm the protein sequence location while carrying out resonance assignments in much the same way that a specifically-labelled protein sample does. A spectrum generated by this experiment is shown in figure 5.11. Three of the 69 methyl containing residues precede a proline and therefore do not give rise to N-HN correlation peaks in this figure. All 66 expected peaks 15  are observed in this spectrum, although the correlations for Asn50, Ser69, and Thr96 are weak. Once the backbone N amide resonances were determined, the assignments of the methyl 15  groups of C B D N I in the calcium-bound sugar-free state were obtained. H(CCO)NH-TOCSY and (H)C(CO)NH-TOCSY experiments, together with a constant time 'H- C HSQC on a sample of 13  10% fractionally C-labelled 13  CBDNI  were used to obtain these resonance assignments (figure  5.12). An example of the data obtained from the H(CCO)NH-TOCSY and (H)C(CO)NH-TOCSY experiments is shown in figure 3.4. Following the titration of the 10% fractionally C-labelled l3  CBDNI  sample with cellopentaose (Figure 5.13) provided a means of obtaining the methyl group  assignments of the sugar-bound C B D N I sample.  Methyl deuteron relaxation and analysis  T|(I C D ), T](I C ) and Ti (I C D ) values were measured to determine deuteron Ti(D) Z  Z  Z  Z  Z  p  z  z  y  and Tip(D) values for methyl groups in free and cellopentaose-bound C B D N I , according to equations 5.11 and 5.12. Figure 5.14 shows a portion of the spectra used to determine the Tip(I C D ) rate of calcium-bound cellopentaose-free C B D N I - Examples for the decay curves for z  z  z  the T](I C D ), Ti(I C ) and Ti (I C D ) magnetization are provided in figure 5.15. Data were Z  Z  Z  z  z  p  z  z  y  obtained for 50 methyl groups in the case of unbound C B D N I , and 49 methyl groups for  Chapter 5-Dynamic Analysis of Ligand Binding  'H (ppm)  Figure 5.10. 1H-15N HSQC spectrum of calcium-loaded cellooligosaccharide-free CBDNI. Aliased peaks are marked with an asterisk and are due to arginine sidechain nuclei.  Chapter 5-Dynamic Analysis of Ligand Binding  11  10  9 111  8  7  PPM  Figure 5.11. (A) Spectrum illustrating the 15N-HN correlations of residues that immediately follow methyl containing amino acids. The peak labelled * is of opposite phase to all other peaks and arises from a CHD group of Glu 151. The peak labelled ** is due to an impurity. (B) 15N-HN HSQC spectrum of CBDNI recorded and processed in the same manner as the spectrum in (A).  Chapter 5-Dynamic Analysis of Ligand Binding  'H  (ppm)  Figure 5.12. Methyl region of a 1H-13C HSQC spectrum of 13C/15N and fractionally deuterated CBDNI calicum-loaded and cellooligosaccharide-free.  Chapter 5-Dynamic Analysis of Ligand Binding  o  O o  O CO ^  o  o  £  0  ° o  o  O  o  a a  . L4951  o 0  AI26  J^b  D o  Q  .0 HI  CM _L141S2  0  I L14151 L  7  7  S  1  L4952  C ^ ^ L 7 7 S 2  0.5  ro CJ  L3252  0.0  C ppm )  Figure 5.13. Titration of calcium-loaded CBDNI with cellopentaose monitored by 1H-13C HSQC spectra. Show is the region of the spectrum where the observed peaks are due to methyl groups. Only residues whose methyls experience a change in chemical shift are labelled.  Chapter 5-Dynamic Analysis of Ligand Binding  0.3  0.2  0.1  0.0 1  H  -O.I  (ppm)  -0.2  -0.3  0.3  0.2  0.1  0.0 1  -0.1  -0.2  -0.3  H (ppm)  Figure 5.14. Portion of the 1H-13C correlation spectra recorded to measure the decay rates of IzCzDz magnetization of methyls in calcium-bound and oligosaccharidefree C B D N I . Spectra are recorded with delay values of (A) 0.05 ms, (B) 8.6 ms, (C) 25.1 msand (D) 50.0 ms.  Chapter 5-Dynamic Analysis of Ligand Binding  (1.01  I)  0.02  0.03  (104  1105  (1.04  (1.05  Time (s)  0.65  0.01  0  0.02  (1.03  Time (s)  Figure 5.15. Examples of the decay of I C D , I C D and I C magnetization. (top) I C D magnetization decay is shown for L32^ (cross), L141 (diamond), V72Y (circle) and ¥36^ (square), (middle) I C D magnetization decay is shown for I54 (circle), V36T (diamond), Ll42 (cross) and V72T (square), (bottom) l C magnetization decay is shown for A94 (diamond), L32 (circle), V72'Y (square) and A108 (cross). z  z  y  Z  Z  Z  Z  2  z  z  Z  82  y  1  Z  81  2  51  Z  Z  1  82  z  z  2  Chapter 5-Dynamic Analysis of Ligand Binding  166  cellopentaose-bound C B D N I • Data for 47 methyl groups were obtained in both the free and bound state.  contains 99 methyl groups. Data were not obtained for many of the methyl groups  CBDNI  in the protein due to severe spectral overlap in the central part of the methyl region of the HSQC  'H-  1  3  C  spectrum, figure 5.12. Figure 5.16 provides the Ti(D) and T[ (D) values determined for both the free and p  cellopentaose-bound C B D N I - From the analysis of these data using equation 5.16, the order parameter S is 2  w  a  ax  s  determined (figure 5.17). This provides information on the nanosecond-  picosecond mobility of the methyl groups. As noted by Kay et al. (1996) S is values display a 2  ax  much greater range of values than do N S values. Excluding the methyls of the completely 15  2  disordered C-terminal residue L152, values of S j range from a low of 0.209 for L32 in 2  81  ax  s  unbound C B D N I to a high of 0.911 for L95 , also in the unbound form of 81  CBDNI-  It can be expected that the length of the side-chain could affect S j values for different 2  ax  s  residue types (Kay et al., 1996). To correct for this positional dependency, average S ; values 2  ax  s  were calculated for each methyl type (Ala CP, Thr CY, Val CY, He CY2, He C , and Leu C ; 8  CBDNI  8  contains no methionine residues), and is denoted by <S ; >. The S j values for the 2  ax s  2  ax  s  methyls of the N and C-terminal residues Al and L152 were not included in this calculation. Both the free and cellopentaose-bound values were included when calculating <S i >. These data are 2  ax s  presented in table 5.7. The difference between S j and the appropriate <S j > was then 2  2  ax  s  ax  s  determined, and is shown in figure 5.17. Table 5.7. Average S i values calculated for each methyl type found in C B D N I • methyl type <S i > 0.803 19 Ala CP Thr CY 0.767 13 ValCY 0.603 16 He CY 0.743 6 0.650 6 He C 33 0.546 Leu C 2  ax  s  2  ax s  2  8  8  a  number of values used to calculate <S j >. 2  ax s  Chapter 5-Dynamic Analysis of Ligand Binding  t — , I  1  7  •lp(D) (s)  0.030 T T 0.025  I I I I I  I  I I I 1 1 1 1 1 1 L. f N — O l —i r N CO K J O t O t O ONOMOON o-.— m H — ~ ^ — — <N>f} ON ON ON^-H —i \0>0 O l —  1 I  -y--o<o ooooV-oo^o -S <* i-po ON ON (-><  1  II I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I  0.020 0.015 0.010I h ,1 i « co o—• QLCN — CN C» 0.0050 t CN (N"^1" «£ VT) .tf ON ON u~>^ I—,— fi f i .IVJ. I iI —I ICN I  -* 4 ©  ~</->£*otofcOto  1  - r i  9  8  1  < ^j^S^-tOtO t O !£- S^S^" —  Ol<n r n r n  rT>  > J J J > >  £  >  ^ O tt oO^ o \ o r—?-xr>--"0 c o o o t O t O rf^W-i^O  — 3— •1  ~  J  ID  8S « S  ,  I 'I C OI t I O  OX I D NIC OT ON I © —11 • — ' O N ON t - v j ' » n f — I —  I— I — I JTi^o I l t ol tOOlNt—Ol CO l MD l tol l lt O_ —• NO\£j 0  —  r-O'i OO—CO i r N — 3'"^ - — N O< c o < o H —' << r o o  f N u - i O \ ON t O rrN N •-«ro—  ~ J  0  1  CN—* c y  rZ^Zn  J  Residue Figure 5.16. Tj(D) (top) and T (D) (bottom) values for methyl groups in calciumloaded CBD bound (closed circles) and not bound (open circles) to cellopentaose. lp  N1  Chapter 5-Dynamic Analysis of Ligand Binding  0.4 0.2 0 -0.2  «TS T  AT  -0.4 -0.6  —' tN —Tf — " O O O O — r-COcOTN — — (N — CN — <N -« (N  CN / N CN-tf £  >- r-rt  N h  < j - ^ C ? ^ L^j-^lU - - - - - - - - - - < N ^ O \ O N O %  r  o o  oooo  oi  <<!h  <•<•  -.v£>^<N  -  t N m m t t -a-^-q- i n  Figure 5.17. Order parameters squared, S , describing the amplitudes of the motions of the methyl averaging axis for methyl groups in (A) calcium-loaded CBD and (C) the calcium-loaded and cellopentaose-bound C B D N I complex. Differences in S a x i s from the average of the square of the order parameters for each methyl type <S > , for methyls of (B) calcium-loaded CBD and (D) the calcium-loaded and cellopentaose-bound CBD complex. (E) Differences between S values of methyls in the cellopentaose-bound and free states of calcium-loaded C B D N I • Error bars are indicated for each value. 2  A X I S  N]  2  2  N1  AXIS  2  N1  A X I S  Chapter 5-Dynamic Analysis of Ligand Binding  169  The patterns of S j - <S j > values are very similar between the bound and free forms 2  2  ax  s  ax s  of C B D N I • Aside from the N and C-terminal residues Al and L152, the methyl groups that show the lowest S i - <S i > values are L32 and L141. The side-chains of these residue occur 2  2  ax  s  ax s  across from each other in [3-sheet B, and are directly adjacent to the disulphide bond connecting C33 and C140. This is very surprising considering these residues are completely buried in the hydrophobic core of the protein. The methyl groups of V34, which also lies next to the disulphide, is also relatively disordered, but not to as great an extent as is the case for L32 and L141. Other methyl groups that have low S  2  - <S > values are V17, L25, T70, V72 and 2  axis  axis  Al 18. Residues corresponding to the most ordered methyl groups in C B D N I , V36, L49, L62, V74, V88, V89, L95 and L129, are all buried in the core of the protein.  Discussion  Comments on the model-free analysis of N relaxation 15  A disproportionately high number of residues in C B D N I needed a two time scale spectral density model to fit the experimentally determined relaxation data (table 5.6). This is apparent from the fact that few residues, aside from the C-terminal residues S148 to L152, had low T1/T2 ratios (figure 5.6) indicative of motions on two time scales (Clore et al, 1990b). This frequent occurrence of residues fit by the two time scale model correlates with the strict error analysis for choosing models and the low error in the Ti and T 2 values determined for C B D N I - The average error in T\ for samples 1 to 5 is 1.1%. For the T2 values the average error is 1.0% and for the H1  15  N NOE values it is 2.0%. This error range is much lower than reported in other studies (Farrow  et al, 1994). When the data were fit with a minimum error in T\ and T2 of 3%, and a minimum error of 5% in the iH-^N NOE values the two time scale model was need much less often to fit the experimental data (table 5.8).  170  Chapter 5-Dynamic Analysis of Ligand Binding  Table 5.8. Comparison of the spectral density models used to fit Tj, T and H- N NOE data for sample 5 with the original and introduced minimum errors. l  l5  2  model  original errors minimum errors' 7 1(S ) 27 2 (S2, T ) 15 21 3 (S2, R ) 14 15 4 (S , T , R ) 11 6 5(S2, S 2, T ) 29 7 1 2 none Minimum errors of 3% for Ti and T values and 5% in 'H- N NOE values were used. 2  e  ex  2  e  ex  t  S  a  i5  2  Although the different models used do change S values, this does not change the trends 2  observed in the data. Residues that have relatively low, or high, values of S when originally fit 2  still have relatively low, or high, values when increased error values are introduced into the data. Also, residues with R terms seem relatively unaffected by the introduction of minimum error ex  values. This dependence of model selection on the error range in the relaxation data does indicate that caution must be used in the interpretation of model-free analysis of relaxation data. In the discussion that follows this was kept in mind. The acquisition of data at two magnetic field strengths was especially useful in judging the reliability of trends in the data.  Global trends  As shown by the uniformity in Tj, T and H- N NOE data (figures 5.3 and 5.4) and the ]  15  2  derived S values (figures 5.6 and 5.7), the structure of C B D N I is well ordered throughout the 2  protein. The four residues at the C-terminus of C B D N I are disordered, but no region within the protein is similarly unstructured. In native CenC these residues form the N-terminus of  CBDN , 2  thus the observed mobility may be due to the use of a protein fragment. In general, none of the residues with T\, T or ^ - ^ N NOE values that significantly differ from average lie in regions of 2  secondary structure.  Chapter 5-Dynamic Analysis of Ligand Binding  111  The only exceptions to this are the residues in (3-strands B2 and B5 near the disulphide bond between C33 and CHO. Conformational flexibility of the region near the disulphide is also reflected by low S i - <S j > values for the methyl groups of L32 and L141 in both bound 2  2  ax  s  ax s  and free C B D N I . It is therefore apparent that this region is motionally disordered in all samples, and that this disorder is not significantly affected by calcium or cellopentaose-binding. This motion might be related to conformational restrictions resulting from the disulphide linking the two [3-strands together. Alternately, disulphide isomerization could lead to conformational flexibility, though this effect tends to occur on a slower time scale. Another possibility is that the sequence of u s e  CBDNI CBDNI,  d in all the structural and dynamic studies contains a mutation in this region. In native  and all other family IV CBDs, residue 139 is a phenylalanine, whereas in C B D N I used in  this study, it is a leucine. This mutation was introduced during the initial subcloning of pTugNln from pTZ18R-JC2. At first it was thought that this amino acid change was a result of incorrect sequencing of the original C B D N I construct. However, upon sequencing C B D N I N 2 , it  w a s  revealed that the position corresponding to L139 in C B D N I is a phenylalanine in C B D N I N 2 - It is therefore difficult to attribute the disorder observed in this region to any one effect when the native protein sequence in this region is not present. There might be subtle packing effects in this region of the core of the protein that are disrupted with the replacement of a phenylalanine by a leucine. It is useful to compare the backbone rms deviation of the ensemble of C B D N I structures with the relaxation data. The best parameter to compare the backbone rmsd to is the steady-state 'H- N NOE measurement. Smaller, or less positive, 'H- N NOE values are indicative of I5  15  internal motion on the nanosecond time scale (Kay et al, 1989), as may be expected in a disordered region of the structure. Figure 5.18 shows a plot of the backbone rms deviation and the 'H- N I5  NOE values observed in sample 1 to 5 of C B D N I - This figure reveals that, aside from the N- and C-termini, the region with the largest rmsd, the loop area around residue 120, corresponds with a protein sequence of relatively low 'H- N NOE. This indicates that the disorder observed in this 15  region, and at the termini, in structure calculation results from true mobility in the protein.  Chapter 5-Dynamic Analysis of Ligand Binding  Figure 5.18. Comparison between the rms deviation (rmsd) for the ensemble of 25 structures, aligned against the average structure (from figure 3.14), and the heteronuclear NOE values for CJ3D samples 1 to 5. High rmsd values indicate regions of the ensemble of structures with high variability. The lower the heteronuclear NOE values are, the more conformationally flexible the respective 'H- N bond vectors are on the nanosecond-picosecond time scale. N1  15  Chapter 5-Dynamic Analysis of Ligand Binding  173  The loop region near residue 83, which also exhibits a high rmsd, does not show low 'HI5  N NOE values. The high rmsd here probably arises from it lying next to the disordered loop  region near residue 120. As this hairpin loop lies at the edge of (3-sheet A, the only cross-strand NOEs to structurally define this region are to the disordered region near residue 120. The high rmsd here thus reflects its proximity to a disordered region and a lack of NOEs to ordered regions of the protein, and not motion on the nanosecond time scale. The slightly low 'H- N NOE values 15  found in the loop region around residue 100 are not reflected in a correspondingly high rmsd. This is probably due to this area being a short loop in comparison to that near residue 120. The well ordered (3-strands at both ends of the residue 100 loop most likely fix the position of this region. A general trend observed in this chapter was non-average relaxation parameters in the Nterminal region  (T1/T2  ratios, 'H- N NOE and R 15  E X  values). This reflects disorder on both the  millisecond and nanosecond time scale in this region and likely is the cause of the high rmsd values observed for residues located at the N-terminal.  Effect of calcium binding  It was previously shown (chapter 4) that C B D N 1 tightly bound calcium ions at a site lying on the face of C B D N I not involved in oligosaccharide binding. The putative atoms that ligate calcium are the backbone carbonyls T8, G30, and D142 and a side-chain oxygen of D142. The role of calcium is to stabilise the folded structure of C B D N I by tying together these three disparate regions of the protein. It is within this calcium-binding region that the most significant differences between the relaxation data of C B D N I in the calcium-free and calcium-bound form (samples 1 and 2) are found (figures 5.5, 5.6 and 5.8). Upon calcium-binding, there is a reduction in the T1/T2 ratio in the N-terminal section of CBDNI  (figure 5.6). This is reflected in a decrease in the value of the  R  E X  term in this region in the  model-free analysis (figure 5.8). This indicates that motion in this region on the millisecond time scale is reduced with calcium-binding. This effect was qualitatively recognized earlier by the  Chapter 5-Dynamic Analysis of Ligand Binding  174  presence of line-broadened amide resonances in the calcium-free form. These resonances sharpen concurrently with calcium-binding. Changes in R  E X  must be considered with caution as different processes may contribute to  this term. Besides mobility of the protein on the millisecond time scale, in principle, the R  E X  terms  could arise from chemical exchange from small amounts of calcium-bound C B D N I being present in solution according to the equilibrium below. kon  CBD + Ca ^=  CBD Ca  2+  2+  N1  k  N1  off  Following the argument of Akke et al, (1993) the rate of exchange (r ) for calcium binding is ex  given by: r  e x  =fcoff+ ^on  [Ca ] =fc ff+ Pkoff 2+  0  where (3 = [CBD i Ca] / [ C B D N I ] ; [Ca ], [CBD i*Ca] and [ C B D N I ] are the concentrations of #  2+  N  N  calcium, C B D N I bound to calcium and calcium-free C B D N I , respectively. As the relaxation experiments on the calcium-free form of C B D N I were collected in the presence of EDTA, whose affinity for calcium is much greater than that of C B D N I , it is expected that [3 is very low. This results in r effectively being equal tofc ff.From chapter 4,fc ffis estimated to be 450 s". For 1  ex  0  this rate to influence the R  E X  0  term, it must be comparable to the inverse of the time between the  refocusing pulse and the formation of the spin-echo during the CPMG sequence used in the pulse sequence to measure T . This has a value of 2000 s in this study. As r for calcium binding is _1  2  ex  significantly lower than 2000 s~' chemical exchange between calcium-free C B D N I and a small amount of calcium-bound C B D N I will not contribute to the R the R  E X  E  X  terms. This strongly implies that  terms observed for resonances from amides whose chemical shifts are perturbed by  calcium-binding in sample 1, are due to exchange between distinct conformations of the protein. Calcium-binding has a less dramatic effect on the relaxation properties of the other parts of the calcium-binding site. There is a decrease in the T[/T ratio and a decrease in R 2  E X  values of  residues L139 and L141 with calcium-binding, again indicating a reduction in motion on the millisecond time scale with binding. But the change observed in this region is not nearly as  Chapter 5-Dynamic Analysis of Ligand Binding  175  dramatic as similar changes at the N-terminus. The region around G30 also shows little change in its relaxation properties with calcium binding. The observation that the largest changes in motions occur at the N-terminal region is most likely due to this segment not being part of any secondary structure element. The two other regions that bind calcium are either a short loop that connects two (3-strands (G30), or a [3-bulge (D142). These regions, though also not part of a secondary structure element, both occur as short two-residue segments between secondary structure elements. Thus both are already relatively structurally fixed compared to the N-terminal region. The fact that the calcium-binding site is relatively disordered in the apo state seems contradictory when the previously determined on rate for calcium association is considered (chapter 4). This on-rate (k ) was found to be (5 + 2) x 10 M s , which is only two orders of 7  _1  _1  on  magnitude from being diffusion controlled. This suggests that the calcium-binding site is a structured, preformed entity with the necessary ligands correctly orientated, ready to quickly bind a calcium ion. Although the protein segment around the calcium-ligating residue T8 is mobile on the millisecond time scale, it is fairly well-ordered on the picosecond time scale. The S values, 2  representing motion on this time scale, in this region in the apo form are not significantly lower than in other regions of the protein indicating this region is not a random coil (figure 5.8). They are also not significantly altered upon calcium-binding. This suggests that protein motions on these two time scales are not linked. The rapid on-rate most likely reflects the ordered picosecond time scale motion, and is not affected by the disordered motion observed on the millisecond time scale. Additionally, the magnitude of the disorder among the native-like states necessary to produce the observed decreases in R and T1/T2 values is unknown, and may only be slight. ex  It is possible to compare these results with those from a similar N relaxation study of the 15  calcium-binding protein calbindin D9k, apo, cadmium and calcium bound (Kordel et al, 1992; Akke et al, 1993). Here, a large increase in S values at binding site II upon either calcium or 2  cadmium ligation was observed. This reflects a large decrease in mobility on the picosecond time  Chapter 5-Dynamic Analysis of Ligand Binding  176  scale of the backbone amides in this region. A similar effect was not observed for C B D N I as the binding site is already well ordered on this time scale in the apo form. Binding site I of calbindin D9k is well ordered in the apo form, consequently.there was little effect on the order parameters of amides in this region with calcium binding. Also found in for calbindin Dcik was significant conformational exchange terms (R x) in the apo and (Cd ) | state, with no significant 2+  e  conformational exchange terms found for the fully calcium-loaded protein (Ca ) . This is similar 2+  2  to what was found for  CBDNI•  Effect of cellooligosaccharide binding on the dynamics ofCBD^j  i) Backbone  relaxation data  There are limited changes in the dynamics of C B D N I with cellopentaose-binding as measured by backbone  1 5  N  relaxation. The oligosaccharide-binding site of C B D N I encompasses  one of the P-sheets of the protein. In the sugar-free form of C B D N I , the individual strands of this P-sheet are generally well ordered. They remain similarly well ordered when bound to cellopentaose. This is reflected in the uniformity of the T], T and heteronuclear N O E values 2  (figure 5.4) of these p-strand residues. The few changes in N relaxation data that do occur with l 5  cellopentaose-binding are observed at the very ends of the p-strands at the start of the loop regions of C B D N I • In particular the Ti/T ratios for V 1 7 , 2  G44, N81  and  G130 drop  significantly. These  residues are either the first or last residues of p-strands forming the binding face of C B D N I • This indicates a decrease in mobility of these residues on the millisecond time scale. The reduction in mobility is similarly shown by a decrease in R terms of these residues (figure 5.7). e x  A similar consideration of the possibility of chemical exchange and not protein motion producing R terms, as was done in the section on calcium-binding, is not possible as the off rate e x  of cellooligosaccharide association with C B D N I is unknown. It is known that sugar-binding is in the fast exchange limit of the  NMR  time scale. This is often much greater than 2000 s~, the rate 1  Chapter 5-Dynamic Analysis of Ligand Binding  177  shown in the previous section for chemical exchange, not motional process to result in the observation of R terms. ex  On the nanosecond-picosecond time scale, as reflected in heteronuclear NOE and S  2  values, there is little change in C B D N I with cellopentaose-binding. This probably reflects the fact that this protein is already well ordered on this time scale in the sugar-free form. There is a small, but consistent, drop in x values with oligosaccharide binding (table 5.5). m  This drop was observed in three separate samples of C B D N I (samples 2 and 3; samples 3 and 4; samples 9, 10 and 11). A drop in T was not observed between samples 6 and 7. This implies M  that the bound form of the protein tumbles slightly faster in solution than the unbound form. This occurs even though this complex is 829 Da, or 5%, heavier when bound to cellopentaose. The ultracentrifuge results showing a slight increase in S20,co of C B D N I when cellopentaose-bound supports this observation of faster tumbling, resulting from a slightly more compact molecule when cellopentaose-bound. Together, the observed decrease in motion on the millisecond time scale of some residues, and a general decrease in t values upon oligosaccharide binding indicates C B D N I is a less m  dynamic protein when bound. Whether this dynamic behaviour is a necessary part of oligosaccharide recognition and binding, or simply a result of steric restrictions imposed on the mobility of residues nearby the bound sugar is hard to determine. A number of previous backbone N relaxation studies have observed a similar reduction in 15  protein mobility with ligand binding. Fesik and co-workers observed a decrease in mobility of several residues on the picosecond time scale upon the binding of a tyrosine-phosphorylated peptide by a phosphotyrosine binding domain (Olejniczak et al, 1997). Hodsdon and Cistola (1997) observed decreased mobility on both the picosecond and millisecond time scales at one end of the binding site of intestinal fatty acid-binding protein upon forming a complex with palmitate. As previously mentioned the picosecond mobility of one of the calcium binding sites of calbindin D9k was dramatically reduced upon ion binding (Kordel et al, 1992; Akke et al, 1993). Conversely, an increase in motion on a picosecond time scale has also been observed upon ligand  Chapter 5-Dynamic Analysis of Ligand Binding  178  binding (Yu et al, 1996). Similar findings as presented here for C B D N I were obtained by Farrow et al (1994), where picosecond time scale disorder was not decreased upon ligand binding, but in the model-free analysis the experimental relaxation data for the complexed protein was fit with fewer exchange terms. It is clear from the different results shown by these studies that a general correlation between protein mobility and ligand-binding is not found. Instead, other protein-specific considerations such as binding affinity and ligand specificity should be taken into consideration. A protein such as C B D N I that is not highly specific to a single substrate (Tomme et al, 1996a) may need to be flexible in the unbound state in order to correctly orient the side-chains involved in binding so that productive interactions can take place with a variety of substrates. When this motion is restricted in the ligand-bound state, as seen by a drop in R values for a small subset of ex  residues on the binding face, an entropic penalty occurs. Proteins that are highly specific to a single ligand may not have to be as flexible and could be fixed in space and not entail this entropic penalty.  ii) Methyl containing deuteron relaxation  While backbone N relaxation techniques are useful in studying protein-ligand 15  interactions, a more intimate view of the binding event is obtained by studying the relaxation properties of the side-chains directly involved. The method used here for studying the dynamics of side-chain groups was methyl containing H relaxation. The work presented in this chapter is only 2  the third reported application of this technique, the second studying ligand binding. CBDNI  is well suited to this method as it contains a number of solvent-exposed methyl  containing side-chains in its binding face (figure 5.19). As a result of spectral overlap in the 'H13  C HSQC spectrum of C B D N I (figure 5.12), it was not possible to obtain data for every methyl  group in the binding face. However, a representative selection of data from methyls across the binding face was obtained.  Chapter 5-Dynamic Analysis of Ligand Binding  Figure 5.19. Structure of CBD with the methyl containing residues present in the binding face shown in green. This figure was made using the programmes Molscript (Kraulis, 1991) and Raster3D (Merrill & Murphy, 1994). N]  Chapter 5-Dynamic Analysis ofLigand Binding  180  Analysis of these data is hampered by the absence of a structure of the proteinoligosaccharide complex. Though the exact position of each methyl in relation to the substrate is unknown, the structure of C B D N I was determined from data collected on the cellotetraose-bound form of the protein. Also, intermolecular NOEs between the protein and ligand were detected for the methyl resonances of V17T, V48Y, V48Y2, L77 and A126P indicating close proximity, less 1  1  82  than 5 A, of these resonances and the bound oligosaccharide. In the unbound form, of the methyls present in the binding face, only that of V17Y is 2  significantly more disordered than methyls of similar type (figure 5.17). Both methyls of L77 have typical S is values. However, the estimation of the relative disorder of leucine methyls in 2  ax  CBDNI  are heavily biased by the large disorder found in L32 and L141. It is fair to say that the  methyls of L77 are significantly disordered compared to most leucine methyls, other than L32 and L141. The methyl groups of T87 and A83 are of typical mobility. Data is unobtainable for the methyls of residues V48 and L126. Upon cellopentaose-binding there is little change in the S j values of any of the binding2  ax  s  face methyls. Instead, the greatest change in S is occurs for T58Y and L95 which becomes 2  2  81  ax  more and less ordered, respectively, upon ligand binding. The reasons for the changes observed for these groups is not known. Surprisingly, the methyl groups V17Y, L77 and L77 remain disordered in the ligand2  82  81  bound form, even though they must be completely buried by cellopentaose. The environment in the bound state may be similar to that found in the core of the protein. The disorder of these methyls when bound to a ligand could be related to similar mobility of the bound oligosaccharide within the binding groove. "Frame-shifting" of bound ligands was suggested by the chemical shift perturbation data produced by the binding of cellooligosaccharides (chapter 2; Johnson et al, 1996a). Similar changes in chemical shifts were observed when C B D N I bound cellotriose, cellotetraose, cellopentaose and cellohexaose. This occurs even though cellotriose and cellotetraose are shorter than the length of the binding cleft. This indicates that either these ligands bind tightly in multiple orientations in the cleft, or are mobile when bound.  Chapter 5-Dynamic Analysis of Ligand Binding  181  As mentioned earlier, in the previous study using this technique (Kay et al, 1996) the methyl groups in the phosphotyrosine binding pocket of a SH2 domain were found to have their motions restricted upon binding. Methyl groups in other regions of the binding site that make contacts with the ligand remained disordered in both the bound and free state. The region where the methyls remain disordered when in contact with the peptide corresponds to the region of the peptide with relaxed binding specificity. Similarly, in the case of C B D N I where none of the methyls located in the binding cleft are highly ordered, either in the bound or free state there is also a lack of binding specificity. As shown by Tomme et al. (1996b)  CBDNI  binds a variety of oligosaccharides. These include either  [3-1,4 linked or mixed (3-1,4 and (3-1,3 glucose polymers, as well as the chemically modified forms of cellulose hydroxyethylcellulose, hydroxypropylmethylcellulose and carboxymethylcellulose. CBDNI  CBDNI  also binds chitin, chitosan, and sephadex. Also, as will be shown in chapter 6,  binds cellooligosaccharides in multiple, not a single, orientations. Motion on the millisecond time scale of residues in the binding face of C B D N I  W  A  S  previously suggested by the observation of broad and weak H$ and H signals from Y19 and Y85 e  (Johnson et al, 1996a). This was also shown by the absence of, or very weak signals of, the asparagine and glutamine groups present in the binding face, and not those in other regions of the molecule. When variants of C B D N I with either Y19 or Y85 changed to alanine a large drop in affinity was observed, as well as the presence of strong signals from all Asn and Gin groups indicating the absence of motion on the millisecond time scale (vide infra; chapter 6). Together, the results of C B D N I and those of the SH2 domain suggest that conformational flexibility of proteins coincides with areas of low protein-ligand specificity. This flexibility might be required for the recognition of a wide variety of different ligands in a common binding site by these proteins. Also, binding-site flexibility may not necessarily be reduced upon ligand-binding.  Chapter 5-Dynamic Analysis of Ligand Binding Changes in r  m  182  with viscosity  In the course of this study it was noticed that C B D N I samples experienced an increase in overall correlation time (x ) with increased protein concentration. It is expected that an increase in m  protein concentration would result in a coincident increase in the viscosity of the protein solution. It was formulated by Debye that x relates to viscosity by: m  x -  47tarL 3  — 3kT  where a is the effective radius of the solute, T is the temperature, k is the Boltzmann constant, and r) is the microviscosity of the environment of the solute. s  As is shown in figure 5.9 there is a good correlation between viscosity and protein concentration. This graph also shows that the solution containing cellopentaose, and thus the bound form of C B D N I , is slightly more viscous than that of the unbound sample. The finding that values of x generally decrease when bound to cellopentaose (table 5.5), and in solution with a m  large excess of unbound sugar, appears to conflict with this observed increase in viscosity. To achieve this drop in x the radius of the protein must decrease when cellopentaose-bound. This m  observation correlates well with the observed decrease in disorder of several loop-residues on the millisecond time scale upon ligand binding. If the relationship between viscosity and protein concentration is assumed to be linear up to 5.3 mM the viscosity of sample 8 (table 5.1; table 5.5) would be 1.002 cSt. This reflects a 30% and 20% increase in the calculated viscosity of samples 6 and 9. This compares well with the measured 24% and 20% increase in x values for sample 8 compared to samples 6 and 9, m  respectively. This increase in x values with protein concentration, plotted together with the m  increase in viscosity is shown in figure 5.9. This finding emphasizes the need to keep sample conditions as similar as possible when comparing the results of different relaxation experiments.  183  Chapter 6 Structural Analysis of Ligand Binding  Abstract  The binding of TEMPO-labelled cellotriose and cellotetraose by C B D N I was studied to provide information about the possible orientations of oligosaccharides when bound. These two ligands are bound by C B D N I slightly tighter than the unmodified sugars. The TEMPO moiety contains a free electron which increases the relaxation rates, R\ and R , of nearby nuclei. For both 2  ligands, backbone H}^ R\ and R rates were measured with the TEMPO group oxidized and l  2  reduced. From the difference in R\ and R values, the overall correlation time (T ) and distance 2  c  between the free electron and ' H nuclei were determined. It was found that these spin-labelled N  ligands are bound in multiple orientations by C B D N I - One set of orientations has the free electron near residue alanine 18, while in the other family of orientations, the free electron is close to glycine 86. These residues lie at opposite sides of the binding cleft. The binding to cellotetraose, and cellopentaose by two C B D N I variants, where tyrosine 19 and tyrosine 85 are changed to alanines, was studied. In both mutant proteins, binding affinity for both oligosaccharides is greatly reduced. A binding model is presented in which the flexibility of side-chains involved in hydrogen bonding formation to oligosaccharides is required. This flexibility allows for the binding of the same ligand in different orientations.  184  Chapter 6-Structural Analysis of Ligand Binding Introduction  Background  The structure of C B D N I was calculated (chapter 3) using data collected in the presence of a large excess of cellotetraose (-40 fold). At that time, it was not known that calcium was bound by CBDNI,  and therefore an excess of cellotetraose, with contaminating metal ion, was required to  detect all backbone amide resonances (chapters 2 and 4). Intermolecular NOEs between the protein and cellotetraose were observed (chapter 3). However, as a large excess of oligosaccharide was used, and the association of C B D N I with cellotetraose is in fast exchange on the NMR chemicalshift time scale, the chemical shifts of the sugar are heavily weighted to the unbound values. Since cellotetraose is composed entirely of (3-(l,4)-linked glucose sugars, its spectrum is highly degenerate and not easy to assign. Thus it was not possible to assign these NOEs to specific ligand-protein interactions, as required to determine the structure of the oligosaccharide-protein complex. The aim of this chapter is answer the question of how  CBDNI  interacts with its ligands.  Specifically, does the protein bind oligosaccharides in one or multiple conformations? Is the bound sugar orientated in one or two directions within the cleft? These questions are addressed in this chapter using nitroxide spin-labelled cellooligosaccharides to define the orientation of the sugar in the binding cleft, and using mutagenesis to probe the roles of two tyrosine side-chains in binding.  Theory on the use of spin labels  Structural insight into how  CBDNI  interacts with oligosaccharides was gained by studying  the binding of spin-labelled cellooligosaccharides. In this study, the spin-label used was a paramagnetic nitroxide free radical in the form of a 2,2,6,6-tetramethylpyrrolidine-l-oxyl group (1) (TEMPO).  185  Chapter 6-Structural Analysis of Ligand Binding R  7 O  (1)  The effect of the unpaired electron of the free radical is to increase the relaxation rates of the resonances of nearby nuclei in C B D N I in a distance-dependent manner. Resonances closest to the spin-label will have their relaxation rates altered the most. This effect was studied in two ways. First, qualitatively, resonances closest to the spin label will decrease in intensity and eventually disappear upon the binding of a spin-labelled molecule. This provides immediate information on the orientation(s) of ligand binding in terms of which nuclei are closest to the free electron. Second, by measuring the relaxation rates (R[, R ) of the backbone amide H nuclei, distances N  2  between the unpaired electron and the proton can be determined quantitatively. The nitroxide group contains an unpaired electron which, in general terms, interacts with protons via dipolar coupling. Due to the size of the electron magnetic dipolar moment, this interaction extends to 25 A and provides long range distance information. In contrast, NOE interactions between pairs of protons are limited to distances of less than ~5 A. The magnetic interaction of an unpaired electron and a proton is described by the modified SolomonBloembergen equations (Solomon & Bloembergen, 1956). The enhancement of the proton's spinlattice (ARi) and spin-spin (AR ) relaxation rates are given by the equations (after Kosen, 1989 2  and Gillespie & Shortle, 1997):  AR, = A  (6.1)  Chapter  6-Structural  Analysis  of Ligand  186  Binding  where K is the constant 1.23 x IO cm s~ for a nitroxide radical, r is the distance between the -32  6  2  electron and proton, x is the correlation time for the electron-proton vector and (OH is the Larmor c  frequency of the proton. A(R) and A(R ) are the difference between the spin-lattice and spin-spin 2  relaxation rates measured with the spin-label reduced (diamagnetic) and oxidised (paramagnetic), respectively. These equation are based on the assumptions that the vector between the electron and proton is free to undergo isotropic rotational diffusion, and that the distance r is constant. The correlation time x is the sum of the contributions from the relaxation of the electron c  and motions of the electron-proton vector: 1/X =1/X + 1/TR c  (6.3)  s  xs is the longitudinal relaxation time of the free electron and X R is the effective rotational correlation time of the vector, which is dependent on the motional characteristics of the protein. Typically for nitroxide radicals, X s is longer than 10~ s. Since X R is in the range of IO to IO s, x is 7  8  -9  c  essentially equal to X R . With x in the range of 10~ to 10~ s (chapter 5) and the NMR experiments carried out at 8  9  c  500 MHz (COH / 2TI) the term x (OH is greater than one. This enables x to be estimated from the c  c  ratio of AR by ARi by rearranging equations 6.1 and 6.2 to obtain: 2  1/2  VARLV 4C0  (6.4)  2  With a value of x and a measurement of either AR or ARj, the distance between the c  2  proton and electron, r, can be determined by rearranging equations 6.1 and 6.2.  r  6  =  r = K 6  3T„ 2K AR, \1 + ®WC;  4T AR, V  c  3T„  ~f~  9  1+  9  < < J  (6.5)  (6.6)  Chapter 6-Structural Analysis of Ligand Binding  187  Experimental methods  Spin labelled cellooligosaccharides  The spin-labelled compounds (2) and (3) used in this study studied were cellotriose and cellotetraose that contain the nitroxide group (1) attached at carbon 1 of the glucose located at the reducing end of the sugar chain. The nitroxide moiety (1) is approximately the same size and shape of a glucose ring. These compounds were synthesized by Lloyd McKenzie in the laboratory of Dr. Steve Withers, Department of Chemistry UBC, using the glycosynthase methodology developed in that laboratory. Full details of the synthesis and characterization of these compounds can be found in McKenzie (1997).  (3)  Preparation of CBDNI samples and titration of nitroxide-labelled cellooligosaccharides  N-labelled C B D N I used in titrations with nitroxide-labelled cellooligosaccharides was  15  obtained by recycling the protein samples from the N relaxation experiments. As mentioned in 15  chapter 5, oligosaccharide-bound C B D N I can be recovered by unfolding the protein in 6 M urea in a microsep concentrator (Filtron). Upon centrifugation, the previously bound oligosaccharide passes through the membrane. The urea was removed by extensive exchange with a 50 mM sodium chloride, 50 mM sodium acetate (d3), 0.02% sodium azide, pH 6.1 buffer. Prior to the titration of the nitroxide-labelled cellooligosaccharides, the protein samples were saturated with  Chapter 6-Structural Analysis of Ligand Binding  188  calcium by the addition of a 10 fold molar excess of calcium chloride dissolved in the same buffer as used with the protein. For the nitroxide-labelled cellotriose (2), 2.8 mg was dissolved in 20 \\L of the identical buffer as used for the protein, and added in 1, 3, 6 and 10 pJL aliquots to the 0.32 m M protein sample.  The final ligand-to-protein ratio was 22:1.  In the case of the nitroxide-labelled  cellotetraose (3), eight additions of a sugar solution were added to a 0.65 m M C B D N I sample. The estimated final ligand-to-protein ratio was 2:1.  Each point of these two titrations was  monitored by the acquisition of a ' H - N HSQC spectrum, collected with spectral widths of 6500 1 5  and 1450 Hz and with 1024 x 96 complex points in the ' H and  1 5  N dimensions, respectively. All  spectra were acquired on a Varian Unity 500 MHz spectrometer at 35 °C. Equilibrium association binding constants for these two ligands, based on their chemical shift perturbation upon ligand binding were obtained using the previously described methods (chapter 2 and chapter 4; Johnson et al, 1996a).  NMR spectroscopy of nitroxide-labelled cellooligosaccharide-CBDNj complexes Upon completion of the titration of  CBDNI  with nitroxide-labelled cellooligosaccharides, a  non-sensitivity enhanced ' H - N HSQC spectrum to estimate H 1 5  [  N  T values was obtained as 1024 2  x 96 complex points with spectral widths of 6500 and 1450 in the ' H and  l 5  N dimensions,  respectively. A series of spectra to measure t H Ti values were then obtained using a sensitivity N  enhanced ' H - N H S Q C sequence (Kay et al, 1992) as a read-out of an inversion-recovery 1 5  sequence (Carr & Purcell, 1954).  For  J  H  N  Tj measurements  of the nitroxide-labelled  cellotetraose, six spectra with delays of T= 0, 0.1, 0.2, 0.4, 0.8 and 2 s were recorded. The T=0.1 spectrum was measured twice in order to help estimate experimental error. nitroxide-labelled cellotriose, the ' H and 2 s. A l l the H L  N  N  For the  T\ series had delay values of T = 0, 0.05, 0.1, 0.2, 0.4, 0.8  T) spectra were acquired as 1024 x 96 complex points with spectral widths  of 6500 and 1450 in the ' H and  1 5  N dimensions, respectively. Selective water flip back pulses  Chapter 6-Structural Analysis of Ligand Binding  189  were incorporated into both these tH-^N HSQC sequences to ensure minimum perturbation of the water magnetization (Grzesiek & Bax, 1993; Zhang et ai, 1994). Upon completion of data collection with  CBDNI  bound to the paramagnetic  cellooligosaccharide, the nitroxide functionality was reduced to the hydroxylamine form by the addition of two molar equivalents of solid L-ascorbic acid (Sigma). This reduction changes the nitroxide functionality on the bound spin-labelled cellooligosaccharide from a paramagnetic to a diamagnetic species. The pH* of the sample was adjusted back to 6.1 and a non-sensitivity enhanced 'H- N HSQC spectrum and a *H Ti series with the same parameters as used with the l5  oxidized sample, were collected.  Data analysis and calculation ofA(Rj) and A(R2J  All NMR spectra were analysed using a combination of Felix v2.3, Felix95 (Biosym Technologies) and NMRPipe (Delaglio et al, 1995). When monitoring the binding of spinlabelled compounds with  CBDNI  by the decrease in intensity of peaks in the 'H- N HSQC, the I5  only window function applied to data was line broadening. When processing spectra to be used for peak volume measurement the data were processed with Lorentzian-to-Gaussian apodization. Relative peak volumes, absolute peak volumes and half-height linewidths were obtained using the nlinLS functionality in NMRPipe. ]  H T\ values were obtained by using relative peak volumes and the delay values (T) as N  input for the programme lmquick (Dr. Neil Farrow, University of Toronto), lmquick fits the relative peak volumes to a function of the form I(T) = 1(0) exp(-T/T|), where I(T) is the relative peak volume at time T and 1(0) is the intensity at time T=0. Errors in the measured relaxation rates are estimated using Monte Carlo procedures as described in Farrow et al (1994). Once the perresidue T) values were obtained the reciprocal of the difference between T] values collected with oxidised and reduced ligands (AR|) was determined.  190  Chapter 6-Structural Analysis of Ligand Binding  Two methods were used to measure per-residue values of AR from the non-sensitivity 2  enhanced HSQC spectra. In the first, peak volumes were used to determine AR . The ratio of a 2  peak volume in the oxidised (V ) versus the reduced (V d) form is: ox  re  V  -t/T,,  ox  _e  (6.7)  -t/T, V  red  e  where t is the total time during the HSQC pulse sequence when H magnetization is in the N  transverse plane. By measuring peak volumes, not heights, the effect of differential T relaxation 2  during the detection period can be neglected. For the pulse sequence used in this study, the value of t is 10.1 ms. This equation can be rearranged to: I  I -tAR,  (6.8)  Vred Further rearranging gives: 1 ^ = I-Hn t V V 2J f  AR = A 2  T  (6.9) V  ox  J  Values of A(l/T ) were also obtained using the proton half-height linewidths of the peaks. Since 2  (6.10)  R, = — = 7T.LW  where LW is the half-height linewidth. It is easy to show that A(l/T ) can be obtained from: <1 | = 7t(LW -LW ) (6.11) AR = A 2  A  2  0X  red  v 2y T  where LW  0X  and LW d are the half-height linewidths of the oxidised and reduced peaks, re  respectively.  Titration of mutant  CBDNI  protein samples monitored by NMR  The binding of mutant  CBDNI  protein samples to cellotetraose and cellopentaose was  studied by NMR spectroscopy. In collaboration with Jeff Kormos and Peter Tomme (Dept. Microbiology, U B C ) , two C B D N I variants were studied as part of a larger investigation of the role played by various residues in oligosaccharide-binding. The C B D N I samples studied each had one  Chapter 6-Structural Analysis of Ligand Binding  191  of the two solvent-exposed tyrosine residues (Tyrl9 and Tyr85) changed to alanine. These variants are subsequently referred to as Y19A and Y85A. N-labelled mutant C B D N I l5  CBDNI  samples were prepared by Jeff Kormos (Dept. Microbiology, UBC). Association binding constants were determined from the titration of N-labelled protein ,5  samples monitored by 'H- N HSQC spectroscopy. For each of the proteins studied, a single 15  sample was used. This sample was split in half, with cellotetraose added to one half and cellopentaose added to the other. This ensures that the protein samples studied with each of the two cellooligosaccharides are identical. The concentration of Y19A  CBDNI  was 0.49 mM. Four  aliquots of cellotetraose and five of cellopentaose was added. For the Y85A sample, a 0.58 mM sample was used, with six additions of cellotetraose and seven of cellopentaose added. For both sample a buffer of 50 mm potassium phosphate pH* 7.0 was used. This buffer is identical to that used in other binding studies performed using these mutant proteins (Jeff Kormos, pers. comm.). 'H- N HSQC spectra on 15  CBDNI  variants were collected at 30 °C using the enhanced-  sensitivity pulsed field gradient method of Kay et al. (1992) and Muhandiram and Kay (1994). Selective water flip-back pules were incorporated to minimize the perturbation of the bulk water (Grzesiek & Bax, 1993; Zhang et al, 1994). Equilibrium association binding constants were determined from the chemical shift perturbation of resonances in H- N HSQC spectra collected J  15  at each titration point. The data were analysed and K values for these protein variants were a  obtained in a similar fashion to that previously outlined (chapter 2; Johnson et al., 1996b).  Results  Binding of the nitroxide-labelled cellooligosaccharides to CBD^j  Association binding constants for the two nitroxide-labelled ligands were obtained using the methods previously described (chapter 2) with one minor modification as follows. In the case of the cellotetraose titration, very poor fits of the experimental data points to the curve calculated on  Chapter 6-Structural Analysis of Ligand Binding  192  the basis of the regressed K value were obtained. Upon the introduction of a scaling factor for the a  ligand concentration, which, along with the binding constant (K ) and bound chemical shift (8b), a  was allowed to vary, good fits of the experimental data to the calculated curve were obtained. The average value for this scaling factor was 1.8, indicating that more oligosaccharide was added than first thought. This scaling factor was not needed for the nitroxide-labelled cellotriose as the weight of this ligand used to prepare the ligand solution appeared to be accurate. The K values determined for the nitroxide-labelled cellooligosaccharides are sumarised in a  table 6.1. For the sake of comparison, also included are the values previously determined for the relevant non-nitroxide cellooligosaccharide. Table 6.1: Association constants (K ) of C B D N I binding 5 TEMPO-labelled cellooligosaccharides. ligand #a(M-') nitroxide-labelled cellotriose (2) 690 ±262 cellotriose^ 180 ±60 nitroxide-labelled cellotetraose (3) (5000) cellotetraose^ 4200 ± 720 a  u  c  Data collected at 35 °C and pH 6.1 in 50 mM sodium chloride, 50 mM sodium acetate (d ), 0.02% sodium azide and 10% D 0 / 90% H 0. The reported K values are the average of those determined from 'H and N chemical shift perturbations of G7, G15, V34, G44, T65, Q80, N81, T87, G130. The error range is one standard deviation. Data presented here are those determined in chapter 2 (Johnson et al, 1996a). Due to uncertainty in the concentration of the ligand solution only an estimate for K is given here.  a  3  2  2  a  15  b  c  a  CBDNI  binds nitroxide-labelled cellooligosaccharides in two distinct orientations  The immediately observable effect of C B D N I binding the spin-labelled oligosaccharides is the decrease in intensity of the amide resonances closest to the unpaired electron. This decrease in intensity is readily observable for a number of peaks in the the titration of  CBDNI  'H-  1 5  N HSQC  spectra collected during  with both the nitroxide-labelled cellotriose and cellotetraose. Figure 6.1  Chapter 6-Structural Analysis of Ligand Binding  H  Spin-label Reduced'  (ppm)  * ™~$>  !  H  (ppm)  Figure 6.1. 1H-15N HSQC spectra of CBDNI bound to TEMPO-label led cellotetraose.  With the spin-label oxidized the free electron attenuates the signals of nearby nuclei. When the spin label is reduced this effect on the relaxation rates is no longer present. Resonances from residues that have their signal reduced the most are labelled.  Chapter 6-Structural Analysis of Ligand Binding  shows the *H- N HSQC spectra of C B D N I 15  m  194  the presence of oxidized and reduced nitroxide-  labelled cellotetraose. It is clear that a number of resonances present in the reduced form are absent in the oxidised form. These missing resonances include V17, A18, G20, Q42 , G82, Y85 and e  T87. In this chapter resonances are said to "disappear" when their intensities fall below an arbitrary, but very low contour level. The resonances that disappear can be classified into two groups. One for residues that lie on (3-strands Al and A2 of the binding face and another for those on strands A3, A4 and the loop connecting these strands (Figure 6.2). That two regions, on opposite ends of the binding face and separated by over 22 A, are the most effected by binding the nitroxide-labelled cellooligosaccharides clearly indicates that these ligands are bound by C B D N I  i n  two distinct orientations.  Based on which resonances disappear, one orientation has the nitroxide located near A18 H , the N  other with the nitroxide flipped end-over-end near and lying G86 H . N  Figure 6.2. Schematic view of CBDNi with residues located in the binding  face labelled. The TEMPO-labelled cellooligosaccharides bind across the strands that make up this (3-sheet. Residues whose backbone H disappear with TEMPO-labelled cellotetraose-binding are underlined. N  Chapter 6-Structural Analysis of Ligand Binding  195  As shown by the ultracentrifuge results in chapter 2 and 5 there is no evidence for intermolecular association of  either free or cellopentaose-bound. This rules out the  CBDNI  possibility of a mechanism where the protein binds the spin-labelled cellooligosaccharides in a single orientation, but associates with itself end-to-end which would give rise to the relaxation rates of residues at one of the ends of the molecule being affected indirectly. During the titration of C B D N I with the nitroxide-labelled cellooligosaccharides, the resonances that disappear lose their intensities at different concentrations of added ligand. At partial saturation, when the contribution of the nitroxide to relaxation is lower, only protons closest to the nitroxide electron will disappear. As saturation increases, the effect of the spin label increases. Protons at a greater distance from the spin-label will now have their relaxation rates attenuated proportionally more. The stage at which the resonances disappear with nitroxidelabelled cellotetraose (3) binding is shown in Table 6.2. For nitroxide-labelled cellotriose (2) binding these data are shown in Table 6.3.  Table 6.2. Disappearance of resonances in H- N HSQC spectra induced by TEMPOlabelled cellotetraose (3) bindings Ligand-to-protein ratio needed to cause Resonance that disappears disappearance of resonance 0.4 : 1 A18 0.88 : 1 V17, Y19, Q42 , G86 1.15:1 G20 1.42 : 1 Y85 1.75 : 1 T87^ Data collected at 35 °C and pH 6.1 in 50 mM sodium chloride, 50 mM sodium acetate (d3), 0.02% sodium azide and 10% D 0 / 90% H 0. indicates resonance disappears upon binding the nitroxide-labelled cellotetraose but not with nitroxide-labelled cellotriose binding J  15  e  a  2  b  2  196  Chapter 6-Structural Analysis of Ligand Binding  Table 6.3. Disappearance of resonances in iJT-^N HSQC spectra induced by TEMPOlabelled cellotriose (2) binding.* Ligand-to-protein ratio needed to cause Resonance that disappears disappearance of resonance 1.09 : 1 V17, A18, Y19, G20 4.35 : 1 V48^, N50 10.89 : 1 T21* A41* Q42 , Y85, G86 21.77:1 G51^G82 undetermined Y43 Data collected at 35 °C and pH 6.1 in 50 mM sodium chloride, 50 mM sodium acetate (d3), 0.02% sodium azide and 10% D 0 / 90% H 0. indicates resonance disappears upon binding the nitroxide-labelled cellotriose but not with nitroxide-labelled cellotetraose binding unable to determine addition after which resonance disappears due to spectral overlap. 3  ft  e  fo  c  fo  a  2  2  b  c  For both TEMPO-labelled ligands, resonances on strand Al disappear at lower ligand-toprotein ratios than resonances in residues on strand A4 (figure 6.2). This may indicate that at subsaturating concentrations one set of orientations is populated first. However, without knowledge of the electron-proton distances, this could result from resonances on strand A1 being closer to the free electron than resonances on strand A4, with both orientations populated equally. The binding of the cellotriose spin-label is very weak. This is reflected in the low binding constant of 690 ± 262 M found for the association of this ligand with C B D N I _1  AS  a result, the  concentrations of spin-label used in this titration were much higher than for the cellotetraose spinlabel, and a significant amount of non-specific binding is observed. Non-specific binding is exemplified by the disappearance of residues that lie away from the binding site, or lie next to residues that appear unaffected by the spin-label. Examples of this non-specific binding include the disappearance of W16 H after 3 additions of ligand, and the disappearance of L146 H after el  the second addition. The amides of V144 and A145 do not disappear.  N  197  Chapter 6-Structural Analysis of Ligand Binding Measurement of'ARj and AR2 values  For both the nitroxide-labelled cellotriose and nitroxide-labelled cellotetraose complexes the differences in ' H R\ and R values between oxidized and reduced spin label were determined. N  2  R values were measured two ways, one using peak volumes and another using half-height ' H  N  2  line widths in non-sensitivity enhanced tH- N HSQC spectra. The peak volume method 15  monitors the effect of the nitroxide before data collection. The line width measure reflects the effect of the nitroxide during the observation period. The H R\ and R data are shown in [  N  2  figures 6.3 and 6.4 for the interaction of  CBDNI  with nitroxide-labelled cellotriose and  cellotetraose, respectively. This analysis was limited to resonances that did not disappear entirely with the protein bound to oxidised spin-label. Thus, data are not obtainable for the resonances most affected by binding of the spin-labelled'oligosaccharides. From figure 6.4 it is immediately apparent that three regions of the  CBDNI  structure have  their relaxation rates affected most by the binding of the cellotetraose spin-label. These are residues G15 to T21 on strand Al, residues Y43 to N51 on strand A2 and residue's Q80 to D90 encompassing parts of strands A3 and A4 and the loop region connecting these two strands. As an indication of which ARi, AR and ALW values are significant, average values for these parameters 2  were calculated. The average value of ARi was found to be 0.29 s with a standard deviation of _l  0.38 s", for AR the average is 8.6 s and the standard deviation is 46 s , for ALW the average 1  _l  -1  2  is 8 Hz with a standard deviation of 32 Hz. Residues that show a change in any of these values one standard deviation greater than average, or whose resonance is decreased in intensity such that AR], AR or ALW could not be measured are shown in red in figure 6.5. 2  In the case of the nitroxide-labelled cellotriose (figure 6.3), four general areas are most affected by the binding of the spin label. These areas are the same as for the cellotetraose spin label with the addition of residues 1125 to F127 on strand A5. Also, the region from residues Y43 to N51 on strand A2 seems to be more uniformly perturbed by this spin label than was the case for the cellotetraose spin label. The effects of non-specific binding are also evident, and are reflected by large values of either AR or ALW for residues V34, A35, R101 and D143 to V145. 2  Chapter 6-Structural Analysis of Ligand Binding  i  12  i  i  I  '  '  '  i  I  I  I  I  I  I  1  L.  I  i  i  i  I  I  I  I  i_  I  4  ARj -I 9  6  (S-1)  4 50  25  I  -r-1—L  200  _i_L  _l  I  I  I  I  I  L_  125  100  75 '  I  _ I  '  '  I  I  I  '  I  I  I  I  1_  i  i  150  L I  I  —1  1  1 L.  J  L  150 -  ALW (Hz) 100  -  50 0 25  50  75  100  125  150  Residue Figure 6.3. Values of ARj, AR2 and ALW resulting from the binding of  TEMPO-labelled cellotriose by CBD . Four distinct regions are affected the most: on strand Al, centered around residue A18; across the length of strand A2; on strands A3 and A4 and the loop that connects them, residues Q80 to D90; and on strand A5, centered around A126. Non-specific binding results in high values of these parameters for residues Q101 and T138 to CHO. The only explanation for all these regions being affected is the binding of the ligand in multiple orientations. N1  Chapter 6-Structural Analysis of Ligand Binding  3.0  '  i  i  i  i  i  '  i  '  i  i  i  i  t_  i  i  1 '  i  i  i  I i  i i  i L  2.5  ARj2-0 -|  (s-1)-  1 5  1.0  -  0.5  -  0.0 125  25  200  i  -r-L  i i i  i i i * i  i i  * i  i I  i  i  150  i i I—i—i—i—i—  150  AR  2  CS" ) 1  100 50 0  I"i 25  ,"i  ill  r*-i  50  If  i -i  f 75  J  75  i''I f i -i 1  100  125  • )-| 150  150  Residue Values of ARj, AR [ and ALW resulting from the binding of TEMPOlabelled cellotetraose by CBD . Three distinct regions are affected the most: on strand Al centered around residue A18; across the length of strand A2; on strands A3 and A4 and the loop that connects them, residues Q80 to D90. The only explanation for all these regions being affected is the binding of the ligand in multiple orientations.  Figure 6.4.  N1  C h a p t e r  6 - S t r u c t u r a l  A n a l y s i s  o f L i g a n d  B i n d i n g  200  C worm diagram of CBD with residues that experience the largest values of AR,, AR or ALW with TEMPO-labelled cellotetraose binding coloured red. Also coloured are residues whose amide resonances disappear with binding of this spin-labelled ligand. This view looks straight down onto the binding face of the molecule, the (3-strands that make upface are labelled. Selected residues are labelled in yellow. Figure 6.5. 2  a  N]  201  Chapter 6-Structural Analysis of Ligand Binding Calculation of T values and electron-proton distances c  As described by equations in the introduction to this chapter, once values of AR i and AR  2  are known it is facile to determine per-residue x and r values. The results of the x calculation for c  CBDNI  c  with the binding of the nitroxide-labelled cellotetraose (3) are summarised in figure 6.6.  Due to the effects of non-specific binding a similar analysis of the data for cellotriose spin-label (2) was not done. It was possible to calculate 75 individual x values for C B D N I , the values ranged c  from 1.8 x 10" s to 2.3 x 10~ s, the mean being 2.5 x 10" s with a standard deviation of 2.7 x 10  8  9  10~ s. This compares to a value of the rotational correlation time of 7.4 x 10~ s determined by 9  15  9  N relaxation methods (Chapter 5). Using this mean value of x , electron-proton distances were determined from equations 6.5 c  and 6.6. As the distances have an r dependence on x they are very insensitive to the differences 6  c  in x observed in figure 6.6. Distances were calculated using both the volume and line-width c  methods of determining AR . These three calculations of the distance are shown in figure 6.7. 2  These graphs show similar trends to those seen in figure 6.4. That is, the regions of the protein that exhibit the shortest distance to the nitroxide are found on strands Al, A3 and A4. This is not at all surprising as these distances are calculated using the AR], AR and ALW values shown in 2  figure 6.4. The distance values calculated using ARi values include the low values of 10.7 A (V17), 11.7 A (Y19), 11.4 A (G20), 11.8 A (N50), 11.5 A (G86) and 11.4 A (T87). From AR values 2  determined using peak volumes include the low values 9.2 A (V17), 9.5 A (Y19), 9.7 A (G20), 10.5 A (Y43), 10.9 A (V48), 11.6 A (G82), 11.4 A (A83), 9.1 A (G86) and 9.4 A (T87). Finally using AR values determined using half-height peak line-widths the shortest distances are 11.1 A 2  (V17), 10.6 A (G20), 11.4 A (V48), 10.0 A (N50), 11.6 A (Q80), 11.1 A (A83), 10.6 A (G86), 10.1 A (T87) and 11.0 A (D90). Based on an estimated error of 10% in the measurement of peak volumes and line-widths, the error in r calculated on AR values is likely to be at least ± 1.6 A . 2  o  The errors for r determined by ARj values are expected to be similar, on the order of ± 2 A.  Chapter 6-Structural Analysis of Ligand Binding  Figure 6.6. Values of the correlation time for the electron-proton vector (tc)  calculated from values of AR] and AR for TEMPO-labelled cellotetraosebinding. 2  202  Chapter 6-Structural Analysis of Ligand Binding  r  (  A  o  20  -  )  15  - ,  75  Residue Figure 6.7. Values of the electron-proton distance (r) calculated from values of (A) A R  l 5  (B) A R , and (C) ALW. Errors in r are estimated to be at least 2A. 2  Chapter 6-Structural Analysis of Ligand Binding  204  Cellooligosaccharide binding ofCBD^j variants  The !H- N HSQC spectra of the C B D N I variants Y19A and Y85A (Figure 6.8) are 15  similar to that of wild type C B D N I f° large majority of residues. Differences are noted for ra  resonances from nuclei near the site of mutation. Although not assigned, the overall similarity indicates that there are no gross structural changes associated with either of these mutations, such as the unfolding of part of the protein structure. Binding constants for the association of mutant  CBDNI  proteins with cellotetraose and  cellopentaose were determined from the chemical shift perturbation of the amide H and N nuclei N  15  as monitored during the titration by tH- N HSQC spectra (Figure 6.9). For each titration, 10 to 15  16 different residues, corresponding to K values from 15-31 different H and N nuclei, were N  15  a  averaged used to determine an overall K value for that ligand protein pair. As the spectra of the a  CBDNI  mutants did not exactly correspond to that of wild-type C B D N I , the amide resonances used  to determine the K values were not assigned to specific residues in the protein. a  Both these  CBDNI  variants bind cellotriose and cellotetraose with approximately equal  affinity (Table 6.4). This binding though, is much weaker than binding by wild type  CBDNI  to  these ligands. The K values for the mutant C B D N I proteins range from 0.6% to 1.4% of the R  corresponding wild-type values.  Table 6.4. Association constants (#) of C B D N I variants for two cellooligosaccharides. cellopentaose (M ) cellotetraose (M ) 84 ± 12 273 ± 30 Y19A 59 + 8 194 ± 19 Y85A 34000 ±7600 4200 ±720 wild type^  0  a  -1  _l  Data collected at 30 °C and pH 7 in 50 mM potassium phosphate and 10% D 0/ 90% H 0. The reported K values are the average of those determined from 'H and N chemical shift perturbations of 10 to 16 peaks. The error range is one standard deviation. Data presented here is that determined in chapter 2 (Johnson et al., 1996a).  a  2  15  2  b  d  Chapter 6-Structural Analysis of Ligand Binding  iTo  i o'.o HI  205  aTo  lo.o  6.0  Hi  fppm)  Wild-type  ' .,'  0  0 •  o %  °~ • •  10.0  iTo  (ppm)  II  0 * i  9.0 HI  1  „•  8.0 7.0 (ppm)  6.0  Figure 6.8 1H-15N H S Q C spectra of wild-type and two C B D N 1 variants. The  resonances corresponding the sites of mutation are indicated in the wild-type spectrum. Side-chain N H 2 groups are connected bysolid lines. All the side-chain N H 2 resonances are present in the two variants, in wild-type C B D N I the N H 2 resonances of side-chains present in the binding face are not observed.  Chapter 6-Structural Analysis of Ligand Binding  Figure 6.9. B i n d i n g o f cellopentaose by the C B D N I variant Y 8 5 A . Shown is a portion of 8 overlayed 1 H - 1 5 N H S Q C spectra in the presence o f increasing amounts o f cellopentaose. The arrows indicate which way the 1 H - 1 5 N peaks shift with added sugar. Present in these spetcra are all the N H 2 resonances. The N H 2 resonances o f residues w h i c h lie i n the binding face ( N 5 0 , N 8 1 , Q 1 2 4 and Q128) are not observed in spectra o f wild-type C B D N I . These resonances also exhibit the largest changes in chemical shift with ligand binding. The assignments given are tentative.  206  207  Chapter 6-Structural Analysis of Ligand Binding  Discussion  Binding affinity of the nitroxide-labelled cellooligosaccharides CBDNI  binds the nitroxide-labelled cellooligosaccharides slightly tighter than the  corresponding cellooligosaccharides (table 6.1). This indicates that the incorporation of the nitroxide label does not adversely effect the binding of these ligands by C B D N I , but rather slightly enhances the binding affinity. This can be explained by favorable hydrophobic and van der Waals interactions between the TEMPO group and residues on the protein surface. The TEMPO group and a glucose ring both present a relatively similar shape, both having a hydrophobic area that can interact with the protein. One way in which the nitroxide moiety (1) differs from a glucose ring is in its lack of hydroxyl groups. It is thought that the hydroxyls on the glucose rings form hydrogen bonds with hydrophilic residues on the protein, providing the primary driving force for sugar binding by CBDNI  (Tomme et  ai,  1996a). The lack of hydroxyls on (1) probably explains why the  nitroxide-labelled cellotetraose does not interact with  CBDNI  as tightly as cellopentaose. And  similarly why the nitroxide-labelled cellotriose is not bound as tightly as cellotetraose by  CBDNI-  These possible explanations for the increase in affinity of the TEMPO-labelled oligosaccharides suggest that the TEMPO group lies within the cleft when bound.  Measurement of the relaxation parameters ARj and AR2 and calculation of % and r values c  From the measured difference in the relaxation rates of H nuclei of C B D N 1 with the N  cellotetraose spin-label oxidised and reduced, per-residue i values were obtained (figure 6.6). c  The average value obtained is 2.5 ns with a standard deviation of 2.7 ns. Though lower, this value is similar to the value of 7.4 ns determined for the overall rotational correlation time as determined by N relaxation experiments (chapter 5). The reason why the value is lower than that previously l5  determined is not known, but may stem from presence of multiple binding orientations or motion of the TEMPO group when bound.  208  Chapter 6-Structural Analysis of Ligand Binding  Using the values of AR i, AR and T , distances between the free electron and H nuclei N  2  C  were determined (figure 6.7). These distances echo the trend in the data of the presence of two binding orientations, with the H resonances of A18 and G86 being the two nuclei closest to the N  free electron. Both these nuclei are approximately 9 A to 10 A from the free electron. This is probably an underestimate of the true distance. This results from there being at least two binding orientations of the spin labelled cellooligosaccharide. Upon saturation, approximately half the molecules have the spin label near these H nuclei so the full effect of the free electron is not felt N  by the nucleus. Also, one assumption of equations 6.1 and 6.2 is that the distance r is fixed. For CBDNI  binding the spin-labelled compounds in two orientations, obviously r is not fixed. With  the free and bound ligands being in fast exchange on the NMR time scale this results in the measured Tj and T values of the oxidized having significant contributions from the alternate 2  orientation, as well as staggered binding modes of the same orientation. Binding of the spin-labelled cellotriose results in attenuation of the signal from resonances in all the strands of the binding face of C B D N I (figure 6.3). This indicates that this ligand is bound in orientations with the TEMPO lying in the middle of the binding face near residues A126 and V48, in addition to orientations near A18 and G86. Figure 6.10 shows these possible orientations. One of, and possibly both, orientations Ib and Ha exist to give rise to the attenuation of the signals from A126 and V48. Orientations la and lib might exist and affect the relaxation rates of the resonances of A18 and G86, though, depending on the mobility of the ligand while bound, orientations Ib and Ha could produce the same effect. The TEMPO-labelled cellotetraose could also bind in each of these orientations, but as a result of its increased length, attenuation of signals in the middle strands of the binding face is less pronounced (figure 6.4). At the start of this study it was hoped the distances obtained would be able to be used to help determine a structure of the bound oligosaccharide. But with the discovery that C B D N I binds these spin-labelled ligands in at least two orientations an accurate structure calculation is not feasible. There is no single structure to calculate, and the restraints derived from this study are only to one atom of a long sugar chain. It is hoped that the synthesis of cellopentaose with  Chapter 6-Structural Analysis of Ligand Binding  209  Possible orientations of TEMPO-labelled cellotriose in the binding cleft of C B D N I - The binding face is drawn as a box with the relative positions of Y19 and Y85 indicated. The ligand is drawn as a narrow rectangle with the TEMPO group indicated by shading. The length of the binding face, cellotriose and TEMPO group are drawn to scale. Evidence for the orientations lb and Ha exists, while the existence of la and lib is also likely. In addition to these binding modes it is possible for each orientation to exist in orientations rotated 180° about the lengthwise axis (see figure 6.12). Figure 6.10.  Chapter 6-Structural Analysis of Ligand Binding  210  selectively incorporated C-labelled glucose rings will allow structures of the different orientations 13  of cellopentaose bound to C B D N I to be determined. However, insights into the nature of the complex can be gained from a rough model of CBDNI  with a cellopentaose molecule manually positioned in the binding cleft (figure 6.11). This  model shows the cellopentaose as a straight chain, but it is likely twisted when bound allowing the rings of Tyrl9 and Tyr85 to stack flat against the pyranose rings of the sugar. Distances between H atoms on the protein, and nuclei on the ligand where the free electron on the TEMPO might be N  are approximately 6 A . This indicates the that electron-H distances in figure 6.7 are about 3 A too N  long. Figure 6.11 clearly shows the length of cellopentaose is very similar to the length of the binding face. This corresponds with binding affinity being maximal for cellopentaose, longer chains bind with the same affinity (chapter 2; Johnson et al, 1996a; Tomme et al, 1996a).  Ligand binding by CBDNI mutants  Previous work had demonstrated the involvement of the side-chains of the tyrosine residues Y85 and Y19 of  CBDNI  in ligand binding. Difference ultraviolet absorbance  spectroscopy showed that the environments of one or more tyrosine residues are perturbed by the addition of cellooligosaccharides (Johnson et al, 1996a). Similarly, the chemical shifts of both the aromatic H and H protons of these tyrosines are perturbed by cellotetraose binding (figure 2.11; s  e  Johnson et al, 1996a), and both these residues show NOEs to cellotetraose from their aromatic protons (Chapter 3; Johnson et al, 1996b). Finally, mutation of either Y85 or Y19 to alanine drastically reduces the affinity of C B D N I to phosphoric-acid swollen Avicel (Jeff Kormos, personal communication). Carbohydrate binding sites in proteins often contain aromatic residues that closely interact with the ligand (Vyas, 1991). In particular, cellulose-CBD interactions are often mediated by the interaction of a number of aromatic residues with cellulose (Chapter 1). The binding of cellotetraose and cellopentaose to the Y19A and Y85A C B D N I mutant proteins were initially studied to see if these cellooligosaccharides were bound in a single orientation by C B D N I , and if they are immobilized while bound. In the structure of  CBDNI  er 6-Structural Analysis of Ligand Binding  Figure 6.11. Two views of a model of CBD N 1 bound to cellopentaose. These views  show only one of the possible orientations in which CBD binds cellopentaose. N]  Chapter 6-Structural Analysis of Ligand Binding  212  presented in chapter 3, cellotetraose is not long enough to interact simultaneously with both Y85 and Y19. It was thought that if cellotetraose and cellopentaose were bound in one orientation, and the ligand is immobile when bound, it would be expected that the mutation of the tyrosine that interacts with cellotetraose would result in a large decrease in affinity for that ligand. The mutation of the tyrosine that does not interact with cellotetraose would result in no change in the affinity of CBDNI  for cellotetraose. Conversely, for cellopentaose, which is expected to interact  simultaneously with both tyrosines, the removal of either tyrosine should reduce the affinity of C B D N I for  this ligand, as observed. However, recent structure calculations of C B D N  I  complexed  with cellotetraose (Dr. E. Brun, personal comm.) show that it is possible for a cellotetraose molecule to simultaneously interact with both Tyr 19 and Tyr 85. This results from the binding cleft of C B D N I becoming narrower in the structure of the cellotetraose-complex. No XPLOR energy penalty is incurred by this narrowing as very few distance restraints are found in this region of the protein to hold it in any specific conformation. The fact that the mutation of either tyrosine drastically reduces the binding for cellotetraose and cellopentaose (table 6.4) proves that both tyrosines are important for ligand binding. Cellotetraose must interact with both Y19 and Y85 while bound to maintain tight binding, in both mutant proteins binding affinity is 1-2% of wild-type affinity. This large drop in binding strength is consistent with the structure of the cellotetraose-CBDNi complex. If cellotetraose could not simultaneously interact with both tyrosines, as suggested by the sugar-free structure, and is bound in two orientations, as shown by the TEMPO-labelled cellotetraose results, its affinity for the two mutant proteins should be half that of wild-type.  Changes in the line-width of binding face residues in the  CBDNI  mutants  As mentioned in Chapter 3, a peculiar feature noticed during the resonance assignment of CBDNI  was that the side-chain NH resonances of the glutamine and asparagine residues, whose 2  side-chains point into the binding groove, are severely line-broadened. This results in a large decrease in intensity of these resonances, probably resulting from intermediate exchange on the  Chapter 6-Structural Analysis of Ligand Binding  213  NMR time scale (millisecond) between different conformations. This motion of binding face residues is not affected by cellooligosaccharide binding to any observable extent, as judged by similar line-broadening in both free and ligand-bound spectra. As these hydrophilic residues probably closely interact with oligosaccharides, a special effort was made to assign them. Most of these resonances were detected and assigned but they appear only as very weak signals. This decrease in intensity of resonances present in the binding face is not limited to NH2-containing residues but was also observed for both the H and H 8  e  resonances for Y19 and Y85 (chapter 2). In chapter 5 it was also observed that the side-chain methyls of V17 and L77 are disordered on the picosecond time scale, both when bound to cellopentaose and unbound. It was therefore very surprising to observe these resonances in the 'H- N HSQC spectra 15  of both Y85A and Y19A C B D N I (figures 6.8 and 6.9). As would be expected by their proposed close association with oligosaccharides, most likely through hydrogen-bond formation, these residue show a large change in chemical shift upon cellooligosaccharide binding (figure 6.9). Indeed, these resonances show the greatest change in chemical shift with ligand binding of any resonances in both mutant C B D N I proteins studied. The appearance of sharp, strong peaks for all the NH groups in the binding face 2  indicates a change in the motion that originally gave rise to the linebroadening. As no "extra" set of resonances was observed in the !H- N HSQC, these resonances in the two mutant proteins 15  probably have moved into the fast exchange limit on the NMR time scale. It is possible however, that the motion has shifted to the slow exchange limit, with only one conformation populated. Since the motion in wild-type protein is altered by the mutation of either Y19 or Y85, the motion must arise from the interaction of a specific residue composition of the binding face. Upon mutation of either tyrosine 85 or 19 to an alanine, the motion that gives rise to the linebroadening is eliminated. Though some of the binding face glutamines and asparagines are close to one of these tyrosines, it seems surprising that a single mutation can alter the appearance of resonances far removed from the mutation site. It is possible the motions of the binding face  Chapter 6-Structural Analysis of Ligand Binding residues are linked together by a network of hydrogen-bonds.  .  214  The removal of the tyrosyl  hydroxyl group would then disrupt any hydrogen bonds it normally forms. It therefore seems the elimination of one member of this hydrogen bonding network, Y19 or Y85, can disrupt the entire network resulting in the elimination of the line-broadening. In the absence of detailed structural information on these mutant proteins, it is not possible to suggest other reasons for the observed effect. Similar line-broadening was observed for a region of the protein Ras(l-171)GMPPNP (Ito et al, 1997) and was termed regional polysterism. This polysterism in Ras(l-171)GMPPNP was significantly reduced upon the binding of its downstream target, the Ras-binding domain of c-Raf1. As discussed in chapter 5,1 introduced the idea that the residues on the binding-face of  CBDNI  are disordered to enable this protein to recognize and bind a variety of different substrates. Ito et al. (1997) suggests a similar reason for the polysterism found in Ras(l-171)GMPPNP as this protein also recognizes a variety of target groups.  Implications for the association of CBDNI  with cellooligosaccharides  From the observed patterns of how the nitroxide-labelled cellooligosaccharides (2) and (3) attenuate the relaxation rates of the amide proton nuclei of C B D N I , it is clear that these ligands are bound in two orientations (figures 6.1, 6.3 and 6.4). The nitroxide group can lie either near [3strand A l close to Alal8, or strand A4 near residue Gly86 (figures 6.2 and 6.5).  Mobility of  bound ligand is suggested in the data from the cellooligosaccharide-binding studies presented in chapter 2. A l l the ligands studied, ranging in length from cellotriose to cellohexaose, produced similar patterns of perturbation of amide chemical shift of the protein. This indicates that these ligands are either mobile while bound, or there is a common structural change occurring with the binding of each cellooligosaccharide. Of course it is also possible for cellotetraose to bind in two orientations and also be mobile while bound. Figure 6.12 shows four possible orientations of cellotetraose.  Chapter 6-Structural Analysis of Ligand Binding  215  Figure 6.12. Four possible orientations of cellotetraose. The reducing end of the oligosaccharide is marked RI. Oxygens are coloured red. The (a) and (b) orientations are flipped 180° about the lengthwise axis. The abundance of oxygens in cellotetraose means this oligosaccharide offers similar hydrogen bonding possibilities in all of these orientations.  Chapter 6-Structural Analysis of Ligand Binding  216  It is possible that the attachment of the TEMPO moiety (1) alters the cellooligosaccharide to prevent this ligand being bound in a single orientation. This would be the case if an interaction between the protein and the anomeric hydroxyl, which is not present in (2) or (3), is especially important for C B D N I binding to it ligands. This seems unlikely as all the orientations shown in figure 6.12 have oxygens at the end of the sugar chains for C B D N I to hydrogen bond with. However in all these orientations the hydroxyls are equatorial, if the anomeric hydroxyl is needed to be in the a anomer to be bound, introduction of the TEMPO could effect binding. The anomeric configuration of the bound sugar is unknown. Also, the attachment of (1) probably does not affect its binding orientation as the nitroxidelabelled sugars studied are bound by  CBDNI  slightly tighter than the unmodified  cellooligosaccharides (table 6.1). This indicates the introduction of the nitroxide group (1) at theend of a cellooligosaccharides does not interfere with binding, probably reflecting the fact that (1) is approximately the same size and shape as a glucose ring. The methyl groups present on (1) should produce some steric hindrance to binding by C B D N I - Despite this, the ligands (2) and (3) are bound tighter than the unmodified cellooligosaccharides. As hydrogen bond formation between the oligosaccharide and the protein is the driving force for binding (Tomme et al, 1996a), the location of the oxygen atoms in four orientations of cellotetraose (figure 6.12) should be considered. It is clear that there are many hydrogen bonding possibilities in all four orientations. Also, superficially at least, there a strong resemblance between the four conformations, especially between the pairs la, 2b and lb, 2a (figure 6.12). It should be noted here that, in the absence of a detailed structure of C B D N I bound to a cellooligosaccharide, figure 6.12 is only a representation showing the orientation of a great many possible conformations. It is likely that when bound, the C5-C6 bond in the sugar can occur in any of the possible rotamers, the angle of the glycosidic bond may change, and there is no guarantee that the sugar adopts the chair conformation as shown in figure 6.12. For a mechanism where the sugar is mobile in the cleft while bound, hydrogen bonds must be continuously broken and formed. Similarly, if oligosaccharides are bound by C B D N I in two  Chapter 6-Structural Analysis of Ligand Binding  217  orientations, different hydrogen bonding combinations must be possible. The high density of hydroxyl groups on cellotetraose would make this possible. From figure 6.12 it is clear that these four orientations offer numerous, and similar, hydrogen bonding possibilities. It is likely that not all the oxygens in cellotetraose are involved in hydrogen bonding at the same time. The association of cellotetraose with C B D N I involves a AG° of -4.9 kcal mol" at 35 1  °C (Tomme et al., 1996a). There are 14 hydroxyl groups and 21 oxygen atoms in total present in cellotetraose. If the sole contribution to binding is assumed to be hydrogen bonding involving only the hydroxyls, each would contribute -0.35 kcal mol". Inclusion of hydrogen bonds to the 1  ether oxygens would reduce this further. From thermodynamic experiments involving proteins binding deoxy-sugars, it has been shown that the presence of some hydroxyls of a sugar are not necessary for binding, or even hinder binding, while others are essential. For three hydroxyls, binding of the respective mono-deoxy sugar by the lectin GS-IV is eliminated (Lemieux, 1996). This indicates that the removal of a single hydroxyl group reduces binding by as much as -6 kcal mol". For those hydroxyls whose elimination results in a measurable, but not complete, decrease 1  in affinity it is estimated that the elimination of an individual hydroxyl involves a A(AG°) of -2 kcal mol" (Quiocho, 1993) to -0.5 kcal mol" (Lemieux, 1996). Using these measurements C B D N I 1  1  likely forms 2.5 to 10 hydrogen bonds to cellotetraose. It is worth noting that these differences in free energy are not due directly to the formation of a single hydrogen bond. Instead, they are the differences between a state where water molecules hydrogen bond to the free protein and ligand, that are then replaced by hydrogen bonds between the protein and ligand. As noted earlier (chapter 3, Johnson et al., 1996b) the binding face of C B D N I contains a number of hydrophilic residues capable of hydrogen bond formation. The work of Jeff Kormos (personal communication) has shown that the substitution of any of Y19, N50, R75, N81, Y85, Q124 or Q128 to alanine dramatically reduces binding affinity. Together, these residues have a potential to form significantly more than the expected number of hydrogen bonds. For a ligand that is bound in two orientations within the binding groove of C B D N I , this excess capacity of hydrogen bond formation might be required to produce the different hydrogen  Chapter 6-Structural Analysis of Ligand Binding  218  bonding combinations necessary for productive binding. Similarly, for a ligand whose length is less than that of the cleft, cellotriose and cellotetraose, the observed mobility of the ligand may be facilitated by the presence of many hydrogen bond forming residues, with only a fraction of them being necessary to form for the ligand to be bound. It was seen earlier, the methyls of V17 and L77 are motionally disordered on the picosecond time scale (chapter 5). Disorder of binding face residues on the millisecond time scale was discussed earlier in this chapter. In both cases motional disorder of the side-chains is observed in both the free and oligosaccharide bound forms. It is possible that this motion is required for the constant reorientation of side-chains to correctly position the necessary groups to form hydrogen bonds to the ligands. This enables C B D N I to bind different ligands as well as different orientations of the same ligand. As mentioned in chapter 1, CBDcex mobile on the surface of cellulose while it is bound 1S  (Creagh et al., 1996; Jervis et al., 1997). In order to facilitate this mobility, it is likely that CBDcex l a  s o  binds cellulose in multiple orientations. As CBDcex binds cellohexaose weakly, this  hypothesis could be tested with experiments analogous to those discussed in this chapter using a spin-labelled cellohexaose molecule. The work presented in this chapter provides the first structural evidence for the manner in which a CBD binds its ligand. These results indicate that C B D N I does not bind cellulose in one specific orientation. This function is consistent with C B D N I acting as an anchor point for the catalytic domain in the native CenC cellulase. In contrast to C B D N I , cellulases bind and degrade cellulose in a single orientation. A biological role of feeding strands of cellulose into the active site of the catalytic domain would require the CBD to bind cellulose strands in one orientation to supply the catalytic domain with correctly oriented substrate. It is possible that the strand of sugar C B D N I binds is fed directly into the catalytic domain, but this would mean that one of the binding orientations of the CBD is enzymatically unproductive, resulting in a less efficient enzyme than if it bound in a single orientation.  Chapter 6-Structural Analysis of Ligand Binding  219  When discussing C B D N I it is important td keep in mind that in the intact enzyme CenC, CBDNI  is found in tandem with C B D N 2  A S  CBDNIN2-  When binding experiments with the spin-  labelled cellooligosaccharides were performed on C B D N 2 it was determined that this protein also binds ligands in multiple orientations (E. Brun, personal communication). It was also determined that C B D N I N 2 does not bind soluble glucans in a cooperative manner (Tomme et al, 1996a). However, it remains to be established whether the affect of one of the domains on the other is to influence the binding orientation, or specificity. If it is found that the two domains are independent of each other it is hard to imagine an efficient mechanism where the CenC CBDs feed the catalytic domain with substrate. Instead I believe the function of C B D N I N 2 is to act as an anchor point on cellulose for the enzyme. This would be especially true if the linker region between the CBDs and the catalytic domain is flexible, enabling it to rotate to be correctly orientated with strands of cellulose which it can then degrade.  220  Appendix Table A1. Assignment of the ' H ,  residue  1 5  N (H ) N  Al S2  1 3  C , and  C 172.89  118.27 (8.39)  P3  177.36  1 5  N N M R spectra of C B D N I i" the presence of cellotetraose (35 ° C ,  13 a c  ( H  a)  >3 P (HP)*  other  C  52.13 (4.19)  19.56 (1.56)  57.78 (5.13)  64.48 (4.66)  65.38 (4.54)  32.51 (1.98, 2.60)  y 28.12 (2.15, 2.35) 8 51.65 (4.05, 4.34)  14  109.59 (7.26)  175.34  61.39 (4.50)  37.63 (2.26)  Y 18.78 (0.95) y 26.49 2  1  (1.19, 1.34) 8 15.04 (0.91) 1  G5  112.30 (8.12)  174.84  43.80 (3.90, 4.62)  E6  125.2 0(9.43)  178.72  59.29 (4.01)  G7  112.45 (10.39)  T8  105.13 (7.07)  176.55  F9  1 19.10 (8.05)  174.83  29.48 (2.23)  Y 36.28 (2.46) 5 183.2  60.75 (4.47)  70.17 (4.47)  Y (1.18)  60.52 (4.51)  34.92 (2.86, 2.97)*  5 131.5 (7.38) e 131.74  45.68 (3.28, 4.40,) 2  (6.86) £ 128.44 (5.86) D10  118.47 (7.83)  177.12  56.24 (4.43)  40.96 (2.70, 2.58)*  Y 179.94  Dil  118.76 (8.81)  175.76  53.03 (4.81)  41.08 (2.72, 2.65)*  y 180.84  G12  110.62 (7.70)  32.35 (2.04, 2.48)  y 26.83 (2.04, 2.12)  P13  44.89 (3.86, 4.70) 175.59  64.41 (4.52)  5 49.49 (3.61, 3.89) E14  117.23 (8.75)  176.13  57.14 (3.90)  G15  106.97 (8.19)  175.05  45.52 (3.79, 4.09)  W16  125.76 (8.33)  176.03  59.53 (4.46)  27.81 (2.61)  29.12 (3.37, 2.96)*  y 36.25 (2.36,2.58) 8 183.8  5  1  129.3 (8.02) e 132.79 1  (11.04) C, 113.52 (7.22) £ 2  121.0 (6.95) e 121.04 3  (7.39) -n 122.5 (6.98) 2  VI7  116.62 (9.23)  173.61  59.97 (4.69)  36.13 (2.26)  Y 21.71 (1.04) 1  Y 19.89 (0.97) 2  A18  124.30 (8.56)  176.42  49.87 (5.65)  23.74 (1.43)  Y19  1.1.9.18 (8.86)  174.36  56.07 (5.13)  41.38 (3.09, 2.95)*  8 133.7 (6.81) e 117.3 (6.70)  G20  111.62 (9.01)  173.77  45.83 (3.82, 4.72)  3  Appendix  221  T21  108.29 (7.07)  174.70  60.73 (4.09)  69.85 (4.51)  Y 22.85 (0.81)  D22  123.07 (8.65)  176.21  53.09 (4.74)  40.04 (2.08, 2.71)*  y 179.4  G23  113.60 (7.70)  32.85 (2.15, 2.33)  Y 6.78 (1.93, 2.06) 8 49.29  P24  2  44.04 (3.89, 4.26) 177.71  62.37 (4.52)  2  (3.59, 3.64) L25  120.89 (8.22)  175.22  55.84 (4.16)  42.45 (1.47, 1.55)  y 26.31 (1.81)8' 25.55 (0.86) S 23.64 (1.57) 2  D26  124.10 (8.42)  176.58  53.19 (5.03)  44.49 (2.62)  Y 180.9  T27  117.65 (8.86)  177.65  61.34 (5.51)  68.21 (4.69)  Y 21.49 (0.82)  S28  123.57 (8.75)  176.54  62.84 (4.12)  (3.95)  T29  113.12 (8.46)  176.10  62.27 (4.46)  70.44 (4.46)  G30  112.01 (8.23)  171.27  45.30 (3.75, 4.28)  A31  124.90 (7.37)  173.39  49.93 (4.39)  21.7 (1.04)  L32  124.52 (8.25)  175.29  54.81 (3.86)  40.16 (-0.63, 0.68)*  2  Y 21.21 (1.18) 2  Y27.2 (0.73) 5 26.10 1  (0.27) 8 25.30 (0.14) 2  C33  126.39 (8.60)  174.26  55.00 (6.13)  47.35 (2.96, 2.75)*  V34  116.72 (9.02)  172.14  58.88 (4.39)  35.87 (1.71)  Y 22.36 (0.57) 1  Y A35  127.51 (8.28)  V36  129.50 (9.11)  175.47  50.60 (4.42)  19.23 (1.34)  59.54 (3.82)  32.00 (0.80)  18.81 (0.41)  2  Y 20.79 (0.80) 1  Y 21.20 (-0.30) 2  176.69  P37  62.36 (4.43)  32.42 (1.91, 2.24)  Y27.29 (1.91) 8 51.87 (3.55, 3.82)  18.77 (1.51)  A38  123.83 (8.55)  169.78  52.59 (3.91)  G39  109.67 (8.40)  175.05  46.79 (3.83)  S40  114.71 (7.43)  173.07  58.99 (4.06)  65.23 (4.12, 3.84)*  A4I  124.79 (8.65)  177.82  50.09 (4.44)  21.48 (1.34)  Q42  124.10 (8.42)  176.51  57.31 (3.25)  27.87 (1.48, 1.35)*  y 32.33 (0.59, 1.46) 8 179.5 e  Y43  122.08 (9.41)  177.82  59.97 (4.07)  36.38 (3.27)  2  112.31 (6.41, 6.80) 8 132.6 (6.97) e 118.3 (6.74)  G44  110.34 (8.56)  172.94  49.47 (3.80, 4.30)  V45  116.63 (7.11)  175.25  59.33 (4.54)  35.27 (1.58)  Y 21.84 (0.77) 1  Y 21.77 (0.95) 2  G46  117.65 (7.83)  179.66  46.21 (3.68, 4.65)  222  Appendix  V47  120.81 (8.60)  175.26  61.48 (5.13)  35.09 (1.97)  y 22.23 (1.00) 1  Y 22.73 (0.86) 2  V48  122.63 (9.59)  173.26  59.28 (5.58)  36.96 (2.01)  Y 22.70 (0.95) 1  Y 21.60 (1.01) 2  L49  128.41 (8.66)  175.02  52.95 (4.09)  41.12 (-1.51, 0.64)  Y 26.61 (0.70) 8 22.00 1  (0.22) 8 26.45 (0.51) 2  N50  124.84 (8.38)  175.88  52.92 (5.02)  40.27 (2.47, 2.75)*  Y 173.46 8  G51  111.04 (8.29)  173.44  46.65 (4.03, 4.17)  V52  125.18 (7.54)  172.93  61.30 (3.76)  33.60 (1.45)  A53  128.04 (7.89)  177.22  51.32 (4.27)  19.57 (1.36)  154  125.22 (8.51)  174.90  56.71 (4.26)  39.83 (1.50)  113.79 (7.20, 6.68)  2  Y , Y 21.09 (0.62) 1  2  • Y 25.48 (0.95, 1.45) Y 1  2  18.55 (0.70) 8 8.60 (0.15 1  E55  128.06 (8.84)  175.24  54.61 (4.60)  31.50 (1.96)  Y 35.74 (2.23) 8 183.6  E56  127.82 (8.18)  177.42  57.97 (3.40)  28.92 (1.64)  Y 34.95 (1.99) 8 182.2  G57  115.48 (9.09)  174.35  44.87 (3.49, 4.32)  T58  122.00 (8.08)  172.39  63.38 (4.16)  69.41 (4.39)  Y 21.21 (0.56)  T59  126.56 (8.47)  172.31  62.70 (4.80)  69.28 (3.78)  Y 21.66 (0.94)  Y60  128.06 (8.84)  174.96  56.58 (4.98)  43.68 (1.41, 2.47)  8 133.0 (6.61)  2  2  e 117.7 (6.83) T61  116.08 (8.84)  174.11  61.86 (5.24)  71.55 (3.86)  Y 22.23 (1.08)  L62  132.71 (9.58)  174.09  53.79 (5.38)  44.76 (1.42, 2.11)*  Y 28.41 (1.80) 8 23.12  2  1  (1.20) S 26.17 (0.80) 2  R63  127.76 (9.31)  175.05  53.94 (5.73)  34.56 (1.84)  y 28.18 (1.70) 8 43.56 (1.70) e 118.32 (8.82)  Y64  115.18 (8.18)  172.56  58.03 (4.90)  40.13 (3.65, 2.71)*  8 132.8 (6.63) e 117.32 (6.50) Y 21.86 (1.02)  T65  118.00 (8.86)  173.83  61.02 (5.37)  70.02 (3.86)  A66  128.89 (8.54)  176.91  50.46 (5.63)  24.61 (1.47)  T67  117.76 (8.73)  172.01  61.76 (4.52)  71.58 (3.78)  A68  129.41 (8.44)  177.18  50.21 (5.75)  25.65 (1.46)  S69  113.12 (8.46)  173.71  59.74 (4.27)  63.74 (3.95)  T70  118.3 (7.35)  170.06  59.44 (4.36)  70.22 (3.94)  Y 18.17 (0.53)  D71  122.74 (7.40)  175.55  54.42 (4.97)  39.86 (2.55)  Y 180.3  V72  125.73 (7.82)  171.87  60.24 (4.16)  34.68 (1.17)  Y 17.80 (-0.45)  2  Y 21.61 (0.97) 2  2  1  Y 20.44 (0.15) 2  223  Appendix  T73  125.36 (7.64)  173.50  61.46 (4.92)  70.12 (3.75)  Y 21.45 (0.91)  V74  115.64 (7.11)  173.17  57.82 (3.50)  31.48 (2.68)  Y 23.54 (0.93)  2  1  Y R75  121.90 (7.58)  174.01  54.99 (5.47)  34.31 (2.06, 1.93)*  2  18.07(0.77)  Y27.17 (1.36) 8 44.63 (3.31) e 116.12 (7.29)  A76  131.15 (8.22)  176.16  50.41 (5.15)  20.73 (0.78)  L77  125.20 (9.34)  174.48  53.47 (4.96)  45.80 (1.60, 1.97)  Y 25.05 (1.82) S 25.75 1  (0.77) S 26.93 (0.69) 2  V78  120.86 (8.17)  176.66  61.53 (5.06)  32.97 (1.62)  Y 20.83 (0.84) 1  Y 20.83 (0.89) 2  G79  116.70 (8.75)  171.66  47.40 (4.10, 4.97)  Q80  117.22 (8.02)  177.93  55.61 (3.43)  31.78 (1.72, 1.32)*  Y 35.97 (1.32, 1.57) 8 179.76 e  N81  124.48 (9.18)  174.24  51.56 (4.76)  G82  110.35 (7.01)  170.64  43.91 (3.28, 3.79)  A83  124.45 (7.91) 175.73  P84  36.26 (2.07, 2.67)*  51.21 (3.34)  16.89 (1.11)  64.08 (4.39)  32.95 (2.05, 2.46)  2  117.13 (6.81, 8.84)  y 177.84 S  2  116.60 (8.08)  Y 24.85 (1.63, 1.91) 8 49.57 (3.38)  34.94 (3.09, 3.38)*  8 133.7 (7.10) e 118.0 (6.6)  60.55 (5.28)  72.88 (4.24)  Y 22.80 (1.19)  60.46 (4.54)  33.26 (2.42)  Y 21.83 (1.01)  Y85  124.62 (9.43)  176.09  60.36 (4.02)  G86  109.38 (8.99)  172.30  45.55 (3.67, 4.09)  T87  107.21 (7.70)  175.28  V88  110.32 (8.35)  174.96  2  1  Y L89  124.53 (7.36)  174.41  55.06 (4.10)  44.02 (0.03, 1.01)  2  19.97 (0.76)  Y 27.19 (1.30) 8 23.04 1  (0.74) S 25.88 (0.62) 2  129.12 (8.62)  175.30  54.61 (4.84)  41.60 (2.55, 2.85)*  Y 181.3  T91  116.69 (8.37)  173.91  59.87 (4.88)  73.89 (4.22)  Y 22.87 (1.09)  S92  117.91 (8.64)  54.14 (5.79)  64.68 (3.70)  62.82 (4.30)  32.63 (1.73, 2.25)  D90  175.25  P93  2  Y 26.43 (1.69, 1.98) 8 49.52 (3.38, 3.45)  A94  125.73 (7.82)  177.29  51.52 (4.64)  18.83 (1.29)  L95  126.35 (9.14)  176.31  53.36 (4.75)  45.24 (1.78, 1.35)*  Y 27.00 (1.75) 8 27.69 (0.82) 8 22.80 (0.83)  T96  115.65 (8.91)  174.08  60.14 (4.82)  71.52 (4.38)  S97  115.89 (8.23)  175.09  59.46 (4.79)  63.25 (3.76, 4.05)*  Y 20.47 (1.15) 2  224  Appendix  E98  124.46 (7.78) 176.90  P99  53.13 (4.75)  30.74 (1.81, 1.96)*  Y 35.80 (2.22) 8 183.9  63.36 (4.33)  31.63 (1.72, 1.94)  y 27.50 (1.97, 2.09) 8 50.47 (3.74, 3.84)  R100  125.51 (8.06)  174.23  55.05 (4.55)  33.31 (1.81, 1.99)*  Y 27.84 (1.51) 8 43.82 (3.14, 3.34) e 115.90 (7.4  Q101  127.12 (8.50)  175.34  55.51 (4.69)  29.00 (2.08, 1.89)*  Y 34.47 (2.27) 8 1 80.9 e  V102  132.90 (9.26)  175.28  62.41 (3.92)  32.91 (0.55)  2  113.71 (6.61, 7.31) Y 21.89 (0.84) 1  Y 20.58 (0.63) 2  T103  122.41(8.12)  174.19  60.87 (5.15)  71.24 (3.93)  Y 21.62 (1.10)  E104  127.02 (9.38)  175.34  54.53 (4.97)  33.29 (1.86, 2.19)*  Y (2.42)  T105  123.29 (8.86)  174.37  61.68 (5.75)  70.85 (3.87)  Y 21.78 (1.13)  F106  125.20 (9.34)  172.19  55.82 (5.11)  41.54 (3.03, 3.25)*  8 (6.96) e 128.2 (6.72)  T107  121.41 (9.14)  175.58  61.89 (4.81)  69.38 (3.90)  y 21.48 (1.00)  AI08  132.06 (8.83)  178.47  54.38 (4.34)  20.07 (1.86)  S109  119.19 (8.14)  171.92  57.66 (4.54)  63.90 (4.30, 3.99)*  AllO  124.36 (7.18)  172.96  51.05 (4.28)  23.00 (1.06)  TI 11  1 15.08 (7.80)  175.00  60.26 (4.74)  71.32 (3.90)  Y 22.06 (1.30)  YI 12  129.18 (9.44)  50.81 (5.66)  39.50 (2.60, 3.22)*  8 131.54 (6.97)  2  2  2  2  e 116.8 (6.29) PI 13 A114  119.13 (8.41)  TI 15  111.01 (7.2)  174.93  61.72 (4.73)  33.22 (2.25, 2.37)  176.95  53.20 (3.85)  19.68 (1.41)  57.68 (4.71)  70.57 (4.06)  Y 21.26 (1.20)  63.27 (4.51)  32.49 (2.45)  Y 27.05 (2.01,)  176.63  PI 16  Y 26.78 (2.39, 2.47) 8 (4.  2  8 50.61 (3.53, 3.82) A l 17  124.93 (8.18)  177.76  50.14 (4.60)  22.23 (1.88, 1.52)  A l 18  126.55 (8.56)  177.98  54.67 (3.93)  17.74 (1.34)  D l 19  117.59 (8.41)  175.30  54.32 (4.01)  40.33 (2.77)  Y 181.3  D120  120.10 (7.59)  52.17 (5.07)  41.93 (2.81, 2.22)*  Y 180.0  62.59 (4.18)  32.38 (1.22, 1.87)  y 26.78 (1.45, 1.56)  175.09  P121  8 49.29 (3.55, 3.66) E122  121.34 (9.30)  175.46  54.35 (4.38)  G123  107.86 (7.69)  174.50  45.06 (4.06, 4.25)  Q124  120.68 (9.27)  174.24  53.57 (5.61)  29.81 (1.26, 1.93)  32.52 (1.66)  Y 35.82 (2.18, 2.37) 8 18:  y 33.20 (1.67, 2.11) 8 179.4 e  2  113.66 (8.42)  225  Appendix  1125  121.96 (8.20)  174.45  60.73 (4.64)  40.03 (1.57)  y 27.71 (1.57, 1.77) y 1  2  17.76 (0.92) 8> 15.36 (1.03) A126  127.97 (8.93)  174.62  51.06 (5.34)  22.56 (1.12)  F127  122.26 (9.13)  175.77  56.91 (5.02)  39.88 (3.12, 2.51)*  8 131.08 (6.91) e 131.09 (6.55) £ 128.4 (5.89)  Q128  123.94 (9.39)  175.77  55.85 (4.82)  29.61 (2.42, 2.12)*  Y 33.62 (2.31, 2.55) 8 179.6 e  L129  119.56 (8.07)  177.14  54.57 (4.98)  43.70 (1.10, 1.99)  2  113.44 (6.84, 7.68)  y27.19 (1.74) S 25.76 1  (0.47) 5 23.10 (0.98) 2  G130  108.65 (8.00)  174.63  44.45 (3.72,4.13)  G131  111.26 (8.12)  173.89  45.88 (3.34,3.52)  F132  120.32 (6.98)  175.45  59.66 (4.51)  8 131.8 (7.11) e 131.3  40.87 (2.34, 3.45)*  (7.35) £ 129.53 (7.25) S133  113.37 (7.80)  175.65  56.38 (4.57)  64.72 (3.45, 3.52)*  A134  134.02 (9.12)  178.72  54.59 (4.32)  18.83 (1.47)  D135  120.51 (7.99)  175.06  53.71 (4.75)  42.71 (2.46, 2.87)*  A136  122.68 (8.25)  178.05  52.07 (4.58)  19.81 (1.48)  W137  120.37 (8.19)  172.86  56.34 (4.63)  29.55 (3.23)  Y 179.9  S  1  119.9 (7.49) e  131.44  1  (10.19) C 113.13 (7.57) 2  C 130.9 (6.55) r| 123.07 3  2  (6.92) e 128.39 (5.89) 3  T138  115.16 (9.07)  173.12  61.21 (4.93)  72.40 (3.88)  Y 21.69 (1.05)  L139  129.72 (8.37)  175.61  53.25 (5.19)  44.63 (1.34, 1.59)*  Y 27.50 (1.32) 8 24.08  2  1  (0.91) S 26.30 (0.61) 2  C140  123.07 (8.72)  173.15  55.68 (5.63)  47.53 (2.72, 2.41)*  L141  127.07 (9.06)  174.36  52.74 (5.05)  47.06 (0.90, 1.03)  Y 26.05 (0.96) S 24.70 1  (-0.16) 8 23.89 (0.01) 2  D142  120.31 (9.01)  173.37  51.43 (5.18)  44.56 (2.39, 2.51)*  Y 183.9  D143  124.06 (9.03)  176.23  55.72 (4.28)  39.66 (2.60)  Y 181.4  V144  120.40 (8.53)  176.62  60.77 (5.28)  33.40 (2.06)  Y 21.46 (1.26) 1  Y 21.54 (1.03) 2  A145  129.71 (9.51)  174.35  52.60 (4.38)  23.54 (1.44)  L146  122.49 (7.61)  174.71  54.44 (5.41)  44.49 (1.25, 2.02)*  Y 27.99 (1.55) S 24.46 1  (1.06) S 26.80 (0.76) 2  D147  128.06 (8.84)  175.58  53.27 (5.19)  44.62 (2.54, 2.79)*  S148  116.39 (8.66)  174.35  56.94 (4.65)  65.30 (3.57)  Y 179.3  226  Appendix  E149  123.65 (8.20)  176.05  55.97 (4.28)  30.02 ( 1 . 7 8 , 1 . 9 6 ) *  V150  122.76 (8.03)  175.71  62.08 (4.08)  33.13 (2.01)  E151  126.98 (8.32)  L152  131.64 (7.83)  175.20  y 35.53 (2.13) 5 183.2 y  1  21.26 (0.88)  y  2  21.26 (0.88)  56.28 (4.29)  30.12 ( 1 . 9 0 , 2 . 0 4 ) *  y 35.99 (2.26) 5 180.8  56.75 (4.17)  43.55 (1.53)  y 27.33 (1.53) 8 (0.92 ) 8 (0.86)  a  asterisk denotes stereospecifically assigned (3 protons, reported as H P and tlP-^, respectively. Otherwise all other 2  methylene protons are listed by ascending chemical shift. Leucine and valine residues for which the methyls are not stereospecifically assigned are reported as 8 and y, respectively.  227  Bibliography Abragam, A. (1961) Principles of Nuclear Magnetism. Clarendon Press, Oxford. Akke, M., Skelton, N.J., Kordel, J., Palmer III, A.G. & Chazin, W.J. (1993) Effects of ion binding on the backbone dynamics of calbindin Dg^ determined by N NMR relaxation. Biochemistry 32, 9832-9844. l 5  Anderson, M.A. & Stone, B.A. (1975) A new substrate for investigating the specificity of (3glucan hydrolases. FEBS Lett. 52, 202-207. Archer, S.J., Ikura, M., Torchia, D.A. & Bax, A. (1991) An alternative 3D NMR technique for correlating backbone N with side chain HP resonances in larger proteins. J. Magn. Reson. 95, 636-641. 15  Bax, A., Clore, M. & Gronenborn, A.M. (1990) 'H- !! correlation via isotropic mixing of C magnetization, a new three-dimensional approach for assigning 'H and C spectra of C enriched proteins. J. Magn. Reson. 88, 425-431. 1  13  13  13  Bax, A., Ikura, M., Kay, L.E. & Zhu, G. (1991) Removal of Fl baseline distortion and optimization of folding in multidimensional NMR spectra. /. Magn. Reson. 91, 174-178. Bax, A., Max, D. & Zax, D. (1992) Measurement of long range C - C J couplings in a 20-kDa protein-peptide complex. J. Am. Chem. Soc. 114, 6923-6925. ,3  13  Bax, A. & Grzesiek, S. (1993) Methodological advances in protein NMR. Acc. Chem. Res. 26, 131-138. Bax, A., Vuister, G.W., Grzesiek, S., Delaglio, F., Wang, A.C, Tschudin, R. & Zhu, G. (1994) Measurement of homo- and heteronuclear J couplings from quantitative J correlation.  Methods Enzymol. 239, 79-105.  Bodenhausen, G. & Ruben, D.J. (1980) Natural abundance nitrogen-15 NMR by enhanced heteronuclear spectroscopy. Chem. Phys. Lett. 69, 185-189. Brandon, C. & Tooze, J. (1991) Introduction to protein structure. Garland Publishing Inc., New York. Bray, M.R., Johnson, P.E., Gilkes, N.R., Mcintosh, L.P., Kilburn, D.G. & Warren, R.A.J. (1996) Probing the role of tryptophan residues in a cellulose-binding domain by chemical modification. Protein Science 5, 2311-2318.  Brun, E., Moriaud, F., Gas, P., Blackledge, M.J., Barras, F. & Marion, D. (1997) Solution structure of the cellulose-binding domain of the endoglucanase Z secreted by Erwinia chrysanthemi. Biochemistry 36, 16074-16086.  Brunger, A.T. (1992) X-PLOR: A System for X-ray Crystallography and NMR. Yale University Press.  228  Bibliography  Campbell, R., Rose, D., Wakarchuk, W.W., To, R., Sung, W. & Yaguchi, M. (1993) in  Proceedings of the second TRICEL symposium on Trichoderma reesei cellulases and other hydrolases (Suominen P. & Reinikainen T. eds.) pp63-77, Foundation for Biotechnology and  Industrial Fermentation Research, Helsinki.  Carr, H.Y. & Purcell, E.M. (1954) Effects of diffusion free procession in nuclear magnetic resonance experiments. Phys. Rev. 94, 630-638. Chan, A.W.E., Hutchinson, E.G., Harris, D. & Thornton, J.M. (1993) Identification and classification of P-bulges in proteins. Protein Science 2, 1574-1590. Chervenka, CH. (1969) A manual of methods for the analytical ultracentrifuge. Spinco Division of Beckman Instruments Inc., Palo Alto, CA. Clore, G., Gronenborn, A., Nilges, M. & Ryan, C. (1987) Three-dimensional structure of potato carboxypeptidase inhibitor in solution. A study using nuclear magnetic resonance, distance geometry, and restrained molecular dynamics. Biochemistry 26, 8012-8023. Clore, G.M., Szabo, A., Bax, A., Kay, L.E., Driscoll, P.C. & Gronenborn, A.M. (1990a) Deviations from the simple two-parameter model-free approach to the interpretation of nitrogen-15 nuclear magnetic relaxation of proteins. J. Am. Chem. Soc. 112, 4989-4991. Clore, G.M., Driscoll, P.C, Wingfield, P.T. & Gronenborn, A.M. (1990b) Analysis of the backbone dynamics of interleukin-1 fj using two-dimensional inverse detected heteronuclear NH NMR spectroscopy. Biochemistry 29, 7387-7401. 15  l  Clore, G.M., Kay, L.E., Bax, A. & Gronenborn, A.M. (1991) Four-dimensional C/ C-edited nuclear Overhauser enhancement spectroscopy of a protein in solution: application to interleukin 1 beta. Biochemistry 30, 12-18. 13  l3  Coughlan, M. (1990) in Microbial Enzymes and Biotechnology. (Fogarty W.M. & Kelly CT. eds.) ppl-36, Elsevier Applied Science, London. Coutinho, J.B., Gilkes, N.R., Warren, R.A.J., Kilburn, D.G. & Miller, R.C, Jr. (1992) The binding of Cellulomonas fimi endoglucanase C (CenC) to cellulose and sephadex is mediated by the N-terminal repeats. Molec. Microbiol. 6, 1243-1252. Creagh, A.L., Koska, J., Johnson, P.E., Tomme, P., Joshi, M.D., Mcintosh, L.P., Kilburn, D.G. & Haynes, CA. (in press) Stability and oligosaccharide binding of the Nl cellulose-binding domain of Cellulomonas fimi Endoglucanase CenC. Biochemistry. Creagh, A.L., Ong, E., Jervis, E., Kilburn, D.G. & Haynes, CA. (1996) Binding of the cellulose-binding domain of exoglucanase Cex from Cellulomonas fimi to insoluble microcrystalline cellulose is entropically driven. Proc. Natl. Acad. Sci. USA 93, 12229-12234. Creighton, T.E. (1993) Proteins. Structures and molecular properties. W.H. Freeman and Company, New York. Damberger, F.F., Pelton, J.G., Harrison, CJ., Nelson, H.C.M. & Wemmer, D.E. (1994) Solution structure of the DNA-binding domain of the heat shock transcription factor determined by multidimensional nuclear magnetic resonance spectroscopy. Protein Science 3, 1806-1821.  229  Bibliography  Delaglio, F., Grzesiek, S., Vuister, G.W., Zhu, G., Pfeifer, J. & Bax, A. (1995) NMRPipe: A multidimensional spectral processing system based on UNIX pipes. J. Biomol. NMR 6, 277-293. Devine, C., Stahlberg, J., Reinikainen, T., Ruohonen, L., Pettersson, G., Knowles, J.K.C., Teeri, T.T. & Jones, T.A. (1994) The three dimensional crystal structure of the catalytic core of cellobiohydrolase I from Trichoderma reesei. Science 265, 524-528. Din, N., Forsythe, I.J., Burtnick, L.D., Gilkes, N.R., Miller, R.C.J., Warren, R.A.J. & Kilburn, D.G. (1994) The cellulose-binding domain of endoglucanase A (CenA) from Cellulomonas fimi: evidence for the involvement of tryptophan residues in binding. Molec. Microbiol. 11, 747-755. Edelhoch, H. (1967) Spectroscopic determination of tryptophan and tyrosine in proteins.  Biochemistry 6, 1948-1954.  Eriksson, A.E., Baase, W.A., Zhang, X.-J., Heinz, D.W., Blaber, M., Baldwin, E.P. & Matthews, B.W. (1992a) Response of a protein structure to cavity-creating mutations and its relation to the hydrophobic effect. Science 255, 178-183. Eriksson, A.E., Baase, W.A., Wozniak, J.A. & Matthews, B.W. (1992b) A cavity-containing mutant of T4 lysozyme is stabilized by buried benzene. Nature 355, 371-373. Farrow, N.A., Muhandiram, R., Singer, A.U., Pascal, S.M., Kay, C.M., Gish, G., Shoelson, S.E., Pawson, T., Forman-Kay, J.D. & Kay, L.E. (1994) Backbone dynamics of a free and a phosphopeptide-complexed Src homology 2 domain studied by N NMR relaxation. Biochemistry 33, 5984-6003. 1 5  Feher, V.A., Baldwin, E.P. & Dahlquist, F.W. (1996) Access of ligands to cavities within the core of a protein is rapid. Nature Struc. Biol. 3, 516-521. Fersht, A. (1985) Enzyme structure and mechanism. W.H. Freeman and Company, San Francisco. Forman-Kay, J.D., Gronenborn, A.M., Kay, L.E., Wingfield, P.T. & Clore, G.M. (1990) Studies on the solution conformation of human thioredoxin using heteronuclear N-!H nuclear magnetic resonance spectroscopy. Biochemistry 29, 1566-1572. 15  Gagne, S.M., Tsuda, S., Li, M., Chandra, M., Similie, L.B., & Sykes, B.D. (1994) Quantification of the calcium-induced secondary structural changes in the regulatory domain of troponin-C. Protein Science 3, 1961-1974.  Gardner, K.H. & Blackwell, J. (1974) The structure of native cellulose. Biopolymers 13, 19752001. Garrett, D.S., Powers, R., Gronenborn, A.M. & Clore, G.M. (1991) A common sense approach to peak picking in two-, three-, and four-dimensional spectra using automated computer analysis of contour diagrams. J. Mag. Reson. 95, 214-220. Gessler, K., Krauss, N., Steiner, T., Betzel, C , Sandmann, C. & Saenger, W. (1994) Crystal structure of p-D-cellotetraose hemihydrate with implications for the structure of cellulose II. Science 266, 1027-1029.  230  Bibliography  Gessler, K., Krauss, N., Steiner, T., Betzel, C., Sarko, A. & Saenger, W. (1995) (3-Dcellotetraose hemihydrate as a structural model for cellulose II. An x-ray diffraction study. J. Am. Chem. Soc. Ill, 11397-11406. Gilkes, N.R., Jervis, E., Henrissat, B., Tekant, B., Miller Jr., R.C., Warren, R.A.J. & Kilburn, D.G. (1992) The absorption of a bacterial cellulase and its two isolated domains to crystalline cellulose. /. Biol. Chem. 267, 6743-6749. Gill, S. & vonHippel, P. (1989) Calculation of protein extinction coefficients from amino acid sequence data. Anal. Biochem. 182, 319-326. Gillespie, J.R. & Shortle, D. (1997) Characterization of long-range structure in the denatured state of staphylococcal nuclease. 1. Paramagnetic relaxation enhancement by nitroxide spin labels. J. Mol. Biol. 268, 158-169'. Glusker, J.P. (1991) Structural aspects of metal liganding to functional groups in proteins. Adv. Prot. Chem. 42,  1-76.  Graham, R.W., Greenwood, J.M., Warren, R.A.J., Kilburn, D.G. & Trimbur, D.E. (1995) pTugA and pTugAS: vectors for high-level expression of cloned genes in E. coli. Gene 158, 5154. Griesinger, G., Otting, G., Wuthrich, K. & Ernst, R.R. (1988) Clean TOCSY for >H Spin System Identification in Macromolecules. J. Am. Chem. Soc. 110, 7870-7872. Grzesiek, S. & Bax, A. (1992) Correlating backbone amide and side chain resonances in larger proteins by multiple relayed triple resonance NMR. J. Am. Chem. Soc. 114, 6291-6293. Grzesiek, S. & Bax, A. (1993) The importance of not saturating H2O in protein NMR. Application to sensitivity enhancement and NOE measurements. J. Am. Chem. Soc. 115, 12593-12594. Grzesiek, S., Anglister, J. & Bax, A. (1993a) Correlation of backbone amide and aliphatic sidechain resonances in C/ N proteins by isotropic mixing of C magnetization. J. Magn. Reson. Series B. 101, 114-119. 13  15  13  Grzesiek, S., Vuister, G.W. & Bax, A. (1993b) A simple and sensitive experiment for measurement of Jrjc couplings between backbone carbonyl and methyl carbons in isotropically enriched proteins. /. Biomol. NMR 3, 487-493. Hahn, M., Olsen, O., Politz, O., Borriss, R. & Heinemann, U. (1995a) Crystal structure and sitedirected mutagenesis of Bacillus macerans endo-l,3-l,4-(3-glucanase. /. Biol. Chem. 270, 30813088. Hahn, K., Keitel, T. & Heinemann, U. (1995b) Crystal and molecular structure at 0.16 nm resolution of the hybrid Bacillus endo-l,3-l,4-(3-D-glucan 4 glucanohydrolase H(A16-M). Eur. J. Biochem. 232, 849-858. Hahn, M., Planas, A., Querol, E. & Heinemann, U. (1995c) Crystal structure of Bacillus licheniformis l,3-l,4-|3-D-glucan 4-glucanohydrolase at 1.8 A resolution. FEBS Letters 314, 221-224.  231  Bibliography  Henrissat, B. (1991) A classification of glycosyl hydrolases based on amino acid sequence similarities. Biochem. J. 280, 309-316. Henrissat, B. & Bairoch, A. (1993) New families in the classification of glycosyl hydrolases based on amino acid sequence similarities. Biochem. J. 293, 781-788. Hiyama, Y., Niu, C.-H., Silverton, J.V., Bavoso, A. & Torchia, D.A. (1988) Determination of N chemical shift tensor via N- H dipolar coupling in glycylglycyl[ N]glycine benzyl ester. J. Am. Chem. Soc. 110, 2378-2383. ,5  15  2  15  Hodsdon, M.E. & Cistola, D.P. (1997) Ligand binding alters the backbone mobility of intestinal fatty acid-binding protein as monitored N NMR relaxation and 'H exchange. Biochemistry 36, 2278-2290. 15  Holm, L. & Sander, C. (1995) Dali: a network tool for protein structure comparison. Trends Biochem. Sci. 20, 478-480. Hutchinson, E.G. & Thornton, J.M. (1996) PROMOTIF-A program to identify and analyze structural motifs in proteins. Protein Science 5, 212-220. Hyberts, S.G., Goldberg, M.S., Havel, TE. & Wagner, G. (1992) The Solution Structure of Eglin c Based on Measurements of Many NOEs and Coupling Constants and its Comparison with X-Ray Structures. Protein Science 1, 736-751. Ikura, M., Kay, L.E. & Bax, A. (1990) A novel approach for sequential assignment of 'H, C , and N spectra of proteins: heteronuclear triple-resonance three-dimensional NMR spectroscopy. Application to calmodulin. Biochemistry 29, 4659-4667. l3  l5  Ikura, M., Kay, L.E. & Bax, A. (1991) Improved three-dimensional 'H-^C-'H spectroscopy of a C-labeled protein using constant-time evolution. /. Biomol. NMR 1, 299-304. 13  Ito, Y., Yamasaki, K., Iwahara, J., Terada, T., Kamiya, A., Shirouzu, M., Muto, Y., Kawai, G., Yokoyama, S., Laue, E., Walchli, M., Shibata, T., Nishimura, S. & Miyazawa, T. (1997) Regional polysterism in the GTP-bound form of the human c-Ha-Ras protein. Biochemistry 36, 9109-9119. Jervis, E.J., Haynes, CA. & Kilburn, D.G. (1997) Surface diffusion of cellulases and their isolated binding domains on cellulose. J. Biol. Chem. 272, 24016-24023. Johnson, W. (1988) Secondary structure of proteins through circular dichroism spectroscopy. Ann. Rev. Biophys. Biophys. Chem. 17, 145-166. Johnson, P.E., Tomme, P., Joshi, M.D. & Mcintosh, L.P. (1996a) Interaction of soluble cellooligosaccharides with the N-terminal cellulose-binding domain of Cellulomonas fimi CenC. 2. NMR and ultraviolet absorption spectroscopy. Biochemistry 35, 13895-13906. Johnson, P.E., Joshi, M.D., Tomme, P., Kilburn, D.G. & Mcintosh, L.P. (1996b) Structure of the N-terminal cellulose-binding domain of Cellulomonas fimi CenC determined by nuclear magnetic resonance. Biochemistry 35, 14381-14394.  232  Bibliography  Kay, L.E., Torchia, D.A. & Bax, A. (1989) Backbone dynamics of proteins as studied by N inverse detected heteronuclear NMR spectroscopy: application to staphylococcal nuclease. Biochemistry 28, 8972-8979. 15  Kay, L.E., Ikura, M. & Bax, A. (1990) Proton-proton correlation via carbon-carbon couplings: A three dimensional NMR approach for the assignment of aliphatic resonances in proteins labeled with carbon-13. /. Am. Chem. Soc. 112, 888-889. Kay, L.E. & Bax, A. (1990) New Methods for the Measuring of NH-CaH Coupling Constants in N-Labelled Proteins. /. Magn. Reson. 86, 110-126. 15  Kay, L.E., Keifer, P. & Saarinen, T. (1992) Pure absorption gradient enhanced heteronuclear single quantum correlation spectroscopy with improved sensitivity. J. Am. Chem. Soc. 114, 10663-10665. Kay, L.E., Xu, G.-Y., Singer, A., Muhandiram, D. & Forman-Kay, J. (1993) A gradientenhanced HCCH-TOCSY experiment for recording side-chains 'H and C correlations in H2O samples of proteins. /. Magn. Reson. 101, 333-337. 13  Kay, L.E. (1993) Pulsed-Field Gradient-Enhanced Three-Dimensional NMR Experiment for Correlating ^C^, C , and H Chemical Shifts in Uniformly C labelled Proteins Dissolved in H 0 /. Am. Chem. Soc. 115, 2055-2057. 1 3  [  a  13  2  Kay, L.E. (1995) Field gradient techniques in NMR spectroscopy. Curr. Opin. Str. Biol. 5, 674681. Kay, L.E., Muhandiram, D.R., Farrow, N.A., Aubin, Y. & Forman-Kay, J.D. (1996) Correlation between dynamics and high affinity binding in an SH2 domain interaction. Biochemistry 35, 361-368. Kay, L.E. & Gardner, K.H. (1997) Solution NMR spectroscopy beyond 25 kDa. Curr. Opin.  Struct. Biol. 7, 722-731.  Keeler, J., Clowes, R.T., Davis, A.L. & Laue, E. (1994) Pulsed field gradients: Theory and practice. Methods Enzymol. 239, 145-207. Keitel, T., Simon, O., Borriss, R. & Heinemann, U. (1993) Molecular and active-site structure of a Bacillus 1,3-1,4-p-glucanase. Proc. Natl. Acad. Sci. U.S.A. 90, 5287-5291. Keitel, T., Meldgaard, M. & Heinemann, U. (1994) Cation binding to a Bacillus (1,3-1,4)-Pglucanase. Geometry, affinity and effect on protein stability. Eur. J. Biochem. 222, 203-214. Kordel, J., Skelton, N.J., Akke, M., Palmer III, A.G. & Chazin, W.J. (1992) Backbone dynamics of calcium-loaded calbindin Dg^ studied by two-dimensional proton-detected N NMR spectroscopy. Biochemistry 31, 4856-4866. 15  Kosen, P.A. (1989) Spin labeling of proteins. Methods Enzymol. Ill, 86-121.  Bibliography  233  Kraulis, P.J., Clore, G.M., Nilges, M., Jones, T.A., Pettersson, G., Knowles, J. & Gronenborn, A.M. (1989) Determination of the three-dimensional solution structure of the Cterminal domain of cellobiohydrolase I from Trichoderma reesei. A study using nuclear magnetic resonance and hybrid distance geometry-dynamical simulated annealing. Biochemistry 28, 72417257. Kraulis, P.J. (1991) MOLSCRIPT: A program to produce both detailed and schematic plots of protein structures. /. Appl. Cryst. 24, 946-950. Laskowski, R.A., MacArthur, M.W., Moss, D.S. & Thornton, J.M. (1993) PROCHECK: a programme to check the stereochemical quality of protein structures. J. Appl. Cryst. 26, 283-291. Lee, W., Revington, M.J., Arrowsmith, C. & Kay, L.E. (1994) A pulsed field gradient isotopefiltered 3D C HMQC-NOESY experiment for extracting intermolecular NOE contacts in molecular complexes. FEBS Lett. 350, 87-90. 13  Lemieux, R.U. (1996) How water provides the impetus for molecular recognition in aqueous solution. Acc. Chem. Res. 29, 373-380. Levy, G.C. & Lichter, R.L. (1979) Nitrogen-15 Nuclear Magnetic Resonance Spectroscopy. J. Wiley and Sons, New York. Liao, D., Kapadia, G., Ahmed, H., Vasta, G.R. & Herzberg, O. (1994) Structure of S-lectin, a developmentally regulated vertebrate (3-galactoside-binding protein. Proc. Natl. Acad. Sci. U.S.A. 91, 1428-1432. Linder, M., Mattinen, M.-L., Kontteli, M., Lindberg, G., Stahlberg, J., Drakenberg, T., Reinikainen, T., Pettersson, G. & Annila, A. (1995a) Identification of functionally important amino acids in the cellulose-binding domain of Trichoderma reesei cellobiohydrolase I. Protein Science 4, 1056-1064.  Linder, M., Lindberg, G., Reinikainen, T., Teeri, T.T. & Pettersson, G. (1995b) The difference in affinity between two fungal cellulose-binding domains is dominated by a single amino acid substitution. FEBS Lett. 372, 96-98. Lipari, G. & Szabo, A. (1982a) Model-free approach to the interpretation of nuclear magnetic relaxation in magnetism. 1. Theory and range of validity. /. Am. Chem. Soc. 104, 4546-4559. Lipari, G. & Szabo, A. (1982b) Model-free approach to the interpretation of nuclear magnetic relaxation in magnetism. 2. Analysis of experimental results. /. Am. Chem. Soc. 104, 4559-4570. Logan, T.M., Olejniczak, E.T., Xu, R.X. & Fesik, S.W. (1992) Side chain and backbone assignments in isotopically labeled proteins from two heteronuclear triple resonance experiments. FEBS Lett. 314, 413-418. Macura, S., Huang, Y., Suter, D. & Ernst, R.R. (1981) Two-dimensional chemical exchange and cross-relaxation spectroscopy of coupled nuclear spins. J. Mag. Reson. 43, 259-281. Marion, D., Ikura, M., Tschudin, R. & Bax, A. (1989a) Rapid recording of 2D NMR spectra without phase cycling. Application to the study of hydrogen exchange in proteins. /. Mag. Reson. 85, 393-399.  234  Bibliography  Marion, D., Kay, L.E., Sparks, S.W., Torchia, D.A. & Bax, A. (1989b) Three dimensional heteronuclear NMR of N-labeled proteins. J. Am. Chem. Soc. Ill, 1515-1517. 15  Marion, D., Driscoll, P., Kay, L., Wingfield, P., Bax, A., Gronenborn, A. & Clore, G. (1989c) Overcoming the overlap problem in the assignment of H NMR spectra of larger proteins by use of three-dimensional heteronuclear iH-^N Hartmann-Hahn-multiple quantum coherence and nuclear Overhauser-multiple quantum coherence spectroscopy: Application to interleukin 1(3. Biochemistry 28, 6150-6156. ]  Mattinen, M.L., Kontteli, M., Kerovuo, J., Linder, M., Annila, A., Lindeberg, G., Reinikainen, T. & Drakenberg, T. (1997a) Three-dimensional structures of three engineered cellulose-binding domains of cellobiohydrolase I from Trichoderma reesei. Protein Science 6, 294-303. Mattinen, M., Linder, M., Teleman, A. & Annila, A. (1997b) Interaction between cellohexaose and cellulose binding domains from Trichoderma reesei cellulases. FEBS Letters 407, 291-296. Mcintosh, L. & Dahlquist, F. (1990) Biosynthetic incorporation of N and C for assignment and interpretation of nuclear magnetic resonance spectra of proteins. Q. Rev. Biophys. 23, 1-38. 15  13  Mcintosh, L.P., Wand, A.J., Lowry, D.F., Redfield, A.G. & Dahlquist, F.W. (1990) Assignment of the backbone 'H and N NMR resonances of bacteriophage T4 lysozyme. Biochemistry 29, 6341-62. 15  McKenzie, L. (1997) Ph.D. Thesis (in preparation). University of British Columbia. McPhalen, C.A., Strynadka, N.C.J. & James, M.N.G. (1991) Calcium-binding sites in proteins: a structural perspective. Adv. Prot. Chem. 42, 77-144. Merritt, E.A. & Murphy, M.E.P. (1994) Raster3D Version 2.0, a program for photorealistic molecular graphics. Acta. Cryst. D. 50, 869-873. Miller, J. (1972) Experiments in Molecular Genetics. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, N.Y. Montelione, G.T., Lyons, B.A., Emerson, S.D. & Tashiro, M. (1992) An efficient triple resonance experiment using carbon-13 isotropic mixing for determining sequence-specific resonance assignments of isotopically-enriched proteins. J. Am. Chem. Soc. 114, 10974-10975. Morris, A.L., MacArthur, M.W., Hutchinson, E.G. & Thornton, J.M. (1992) Stereochemical quality of protein structure coordinates. Proteins 12, 345-364. Muhandiram, D. & Kay, L. (1994) Gradient-enhanced triple-resonance three-dimensional NMR experiments with improved sensitivity. J. Mag. Res. Series B. 103, 203-216. Muhandiram, D.R., Yamazaki, T., Sykes, B.D. & Kay, L.E. (1995) Measurement of H T| and Tip relaxation times in uniformly C-labeled and fractionally H-labeled proteins in solution. J. Am. Chem. Soc. 117, 11536-11544. 2  13  2  Muhandiram, D.R., Johnson, P.E., Yang, D., Zhang, O., Mcintosh, L.P. & Kay, L.E. (1997) Specific N , NH correlations for residues in N, C fractionally deuterated proteins that immediately follow methyl-containing amino acids. J. Biomol. NMR. 10, 283-288. l5  15  l 3  235  Bibliography  Neri, D., Szyperski, T., Otting, G., Senn, H. & Wuthrich, K. (1989) Stereospecific nuclear magnetic resonance assignments of the methyl groups of valine and leucine in the DNA-binding domain of the 434 repressor by biosynthetically directed fractional C labeling. Biochemistry 28, 7510-7516. 13  Nicholls, A. Sharp, K.A. & Honig, B. (1991) Protein folding and association: Insights from the interfacial and thermodynamic properties of hydrocarbons. Proteins 11, 281-292. Nicholson, L.K., Kay, L.E., Baldisseri, D.M., Arango, J., Young, P.E., Bax, A. & Torchia, D.A. (1992) Dynamics of methyl groups in proteins as studied by proton-detected C NMR Spectroscopy. Application to the leucine residues of staphylococcal nuclease. Biochemistry 31, 5253-5263. 13  Nilges, M., Clore, G.M. & Gronenborn, A.M. (1988) Determination of three-dimensional structures of proteins from interproton distance data by hybrid distance geometry-dynamical simulated annealing calculations. FEBS Lett. 229, 317-24. Olejniczak, E.T. & Fesik, S.W. (1994) Two-dimensional nuclear magnetic resonance method for identifying the H / C signals of amino acid residues preceding proline. /. Am. Chem. Soc. 116, 2215-2216. a  a  Olejniczak, E.T., Zhou, M. & Fesik, S.W. (1997) Changes in the NMR-derived motional parameters of the insulin receptor substrate 1 phosphotyrosine binding domain upon binding to an interleukin 4 receptor phosphopeptide. Biochemistry 36, 4118-4124. Ong, E., Greenwood, J.M., Gilkes, N.R., Kilburn, D.G., Miller Jr, R.C. & Warren, R.A.J. (1989) The cellulose-binding domains of cellulases: tools for biotechnology. Trends Biochem. Sci. 7, 239-243. Ong, E., Gilkes, N.R., Miller, R.C, Jr., Warren, R.A.J. & Kilburn, D.G. (1993) The cellulosebinding (CBDcex) domain of an exoglucanase from Cellulomonas fimi: production in Escherichia coli and characterization of the polypeptide. Biotechnol. Bioeng. 42, 401-409. Otting, G. (1993) Experimental NMR techniques for studies of protein-ligand interactions. Curr. Opin. Str. Bio. 3, 760-768. Pace, C.N., Vajdos, F., Fee, L., Grimsley, G. & Gray, T. (1995) How to measure and predict the molar absorption coefficient of a protein. Protein Science. 4, 2411-2423. Pascal, S.M., Muhandiram, D.R., Yamazaki, T., Forman-Kay, J.D.F. & Kay, L.E. (1994) Simultaneous acquisition of N and C-edited NOE spectra of proteins dissolved in H2O. J. Magn. Reson. 103, 197-201. 15  l3  Peng, J.W. & Wagner, G. (1994) Investigation of protein motions via relaxation measurements.  Methods Enzymol. 239, 563-596.  Poole, D.B., Hazlewood, G.P., Huskisson, N.S., Virden, R. & Gilbert, H.J. (1993) The role of conserved tryptophan residues in the interaction of a bacterial cellulose binding domain with its ligand. FEMS Microbiol. Lett. 106, 77-84.  Bibliography  236  Powers, R., Garrett, D., March, C., Frieden, E . , Gronenborn, A. & Clore, G . (1993) The highresolution three-dimensional solution structure of human interleukin-4 determined by multidimensional heteronuclear magnetic resonance spectroscopy. Biochemistry 32, 6744-6762. Quiocho, F . A . (1986) Carbohydrate-binding proteins: tertiary structures and protein-sugar interactions. Ann. Rev. Biochem. 55, 287-315. Quiocho, F.A. (1989) Protein-carbohydrate interactions: basic molecular features. Pure & Appl. Chem. 61, 1293-1306. Quiocho, F . A . (1993) Probing the atomic interactions between proteins and carbohydrates. Biochem. Soc. Trans. 21, 442-448. Reinikainen, T., Ruohonen, L . , Nevanen, T., Laaksonen, L . , Kraulis, P., Jones, T . A . , Knowles, J.K.C. & Teeri, T.T. (1992) Investigation of the function of mutated cellulose-binding domains of Trichoderma reesei cellobiohydrolase I. Proteins 14, 475-482. Reinikainen, T., Teleman, O. & Teeri, T.T. (1995) Effects of pH and ionic strength on the adsorption and activity of native and mutated cellobiohydrolase I from Trichoderma reesei. Proteins 22, 392-403. Rini, J.M. (1995) Lectin structure. Annu. Rev. Biophys. Biomol. Struct. 24, 551-577. Rost, B. (1996) PHD: predicting one-dimensional protein structure by profile-based neural networks. Methods Enzymol. 266, 525-539. Sakon, J., Irwin, D., Wilson, D.B. & Karplus, P.A. (1997) Structure and mechanism of endo/exocellulase E4 from Thermomonospora fusca. Nature Struc. Biol. 4, 810-818. Sambrook, J., Fritsch, E.F. & Maniatis, T. (1989) Molecular cloning: a laboratory manual. Cold Spring Harbor Laboratory Press, Cold Spring Harbor NY. Sandstrom, J. (1982) Dynamic NMR Spectroscopy. Academic Press, London. Santoro, J. & King, G.C. (1992) A constant-time 2D Overbodenhausen experiment for inverse Ccorrelation of isotopically enriched species. J. Magn. Reson. 97, 202-207. Sattler, M . & Fesik, S.W. (1996) Use of deuterium labeling in NMR: overcoming a sizable problem. Structure 4, 1245-1249. Schleucher, J., Schwendinger, M . , Sattler, M . , Schmidt, P., Schedletzky, O., Glaser, S.J., S0rensen, O.W. & Griesinger, C. (1994) A general enhancement scheme in heteronuclear multidimensional N M R employing pulsed field gradients. J. Biomol. NMR 4, 301-306. Solomon, L . & Bloembergen, N. (1956) Nuclear magnetic interactions in the H F molecule. J. Chem. Phys. 25, 261-266. Stanczyk, S.M., Bolton, P.H., Dell'Acqua, M . & Gerlt, J.A. (1989) Observation and sequence assignment of a cis prolyl peptide bond in unliganded staphylococcal nuclease. / . Am. Chem. Soc. I l l , 8317-8318.  237  Bibliography  Strynadka, N.CJ. & James, M.N.G. (1989) Crystal structures of the helix-loop-helix calciumbinding proteins. Anna. Rev. Biochem. 58, 951-998. Surewicz, W., Mantsch, H. & Chapman, D. (1993) Determination of protein secondary structure by fourier transform infrared spectroscopy: A critical assessment. Biochemistry 32, 389-394. Szyperski, T., Neri, D., Leiting, B., Otting, G. & Wuthrich, K. (1992) Support of 'H NMR assignments in proteins by biosynthetically directed fractional C-Labeling. /. Biomol. NMR 2, 323-334. l3  Thornton, J.M. (1981) Disulphide bridges in globular proteins. J. Mol. Biol. 151, 261-287. Tomme, P., Warren, R.A.J., Miller, R.C., Jr., Kilburn, D.G. & Gilkes, N.R.(1995) Cellulosebinding domains: classification and properties, in Enzymatic Degradation of Insoluble Polysaccharides (Saddler J.N. & Penner M., eds) ppl42-161, American Chemical Society, Washington, DC. Tomme, P., Creagh, A.L., Kilburn, D.G. & Haynes, CA. (1996a) interaction of polysaccharides with the N-terminal cellulose-binding domain of Cellulomonas fimi CenC. 1. Binding specificity and calorimetric analysis. Biochemistry 35, 13885-13894. Tomme, P., Gilkes, N.R., Guarna, M.M., Haynes, C.A., Hasenwinkle, D., Jervis, E., Johnson, P., Mcintosh, L.P., Warren, R.A.J. & Kilburn, D.G. (1996b) Cellulose-binding domains: versatile affinity tags for large scale purification, concentration and immobilization of fusion proteins. Ann. N. Y. Acad. Sci. 799, 418-424. Tomme, P., Kwan, E., Gilkes, N.R., Kilburn, D.G., Warren, R.A. (1996c) Characterization of CenC, an enzyme from Cellulomonas fimi with both endo- and exoglucanase activities. J.  Bacteriol. 178, 4216-42323.  Tormo, J., Lamed, R., Chirino, A.J., Morag, E., Bayer, E.A., Shoham, Y. & Steitz, T.A. (1996) Crystal structure of a bacterial family III cellulose-binding domain: a general mechanism for attachment to cellulose. EMBO J. 15, 5739-5751. Torronen, A., Harkki, A. & Rouvinen, J. (1994) Three dimensional structure of endo-l,4-pxylanase II from Trichoderma reesei: two conformational states in the active site. EMBO J. 13, 2493-2501. Vuister, G.W. & Bax, A. (1992) Resolution enhancement and spectral editing of uniformly C enriched proteins by homonuclear broadband C decoupling. J. Magn. Reson. 98, 428-435. l3  13  Vuister, G.W., Wang, A.C. & Bax, A. (1993) Measurement of three-bond nitrogen-carbon J couplings in proteins uniformly enriched in ^N and C J. Am. Chem. Soc. 115, 5334-5335. 1 3  Vyas, N.K. (1991) Atomic features of protein-carbohydrate interactions. Curr. Opin. Struct. Biol. 1, 732-740. Welfle, K., Misselwitz, R., Welfle, H., Simon, O., Politz, O. & Borriss, R. (1994) Microcalorimetric determination of the thermostability of three hybrid (l,3-l,4)-(3-glucanases. J. Biomol. Struct. Dyn. 11, 1417-1424.  Bibliography  238  Welfle, K., Misselwitz, R., Welfle, H . , Politz, 0. & Borriss, R. (1995) Influence of Ca + on conformation and stability of three bacterial hybrid glucanases. Eur. J. Biochem. 229, 726-735. 2  Welfle, K., Misselwitz, R., Politz, O., Borriss, R. & Welfle, H. (1996) Individual amino acids in the N-terminal loop region determine the thermostability and unfolding characteristics of bacterial glucanases. Protein Science 5, 2255-2265. Wishart, D., Richards, F. & Sykes, B. (1992) The chemical shift index: a fast and simple method for the assignment of protein secondary structure through N M R spectroscopy. Biochemistry 31, 1647-1651. Wishart, D. & Sykes, B. (1994) The C chemical-shift index: a simple method for the identification of protein secondary structure using C chemical-shift data. J. Biomol. NMR 4, 171-180. l 3  1 3  Wishart, D.S., Bigam, C.G., Yao, J., Abildgaard, F., Dyson, H.J., Oldfield, E . , Markley, J.L. & Sykes, B.D. (1995) ' H , C and N chemical shift referencing in biomolecular NMR. J. Biomol. NMR 6, 135-140. 1 3  1 5  Wittekind, M . & Mueller, L . (1993) H N C A C B , a high-sensitivity 3D N M R experiment to correlate amide-proton and nitrogen resonances with the alpha- and beta-carbon resonances. J. Magn. Reson. Series B. 101, 201-205. Wuthrich, K. (1986) NMR of Proteins and Amino Acids. J. Wiley and Sons, New York. Xu, G.-Y., Ong, E . , Gilkes, N.R., Kilburn, D . G . , Muhandiram, D.R., Harris-Brandts, M . , Carver, J.P., Kay, L . E . & Harvey, T.S. (1995) Solution structure of a cellulose-binding domain from Cellulomonas fimi by nuclear magnetic resonance spectroscopy. Biochemistry 34, 69937009. Yamazaki, T., Foreman-Kay, J.D. & Kay, L . (1993) Two-dimensional N M R experiments for correlating C P and l H ^ chemical shifts of aromatic residues in C-labelled proteins via scalar couplings. J. Am. Chem. Soc. 115, 11054-11055. 13  /e  l3  Yang, D. & Kay, L . E . (1996) The effects of cross correlation and cross relaxation on the measurement of deuterium T i and Tip relaxation times in C H 2 D spin systems. / . Mag. Res. Ser. B. 110, 213-218. 13  Yang, D., Mok, Y . , Forman-Kay, J.D., Farrow, N.A. & Kay, L . E . (1997) Contributions to protein entropy and heat capacity from bond vector motions measured by NMR spin relaxation. / . Mol. Biol. 272, 790-804. Yanisch-Perron, C , Vieira, J. & Messing, J. (1985) Improved M l 3 phage cloning vectors and host strains: nucleotide sequences of the M13mp and pUC vectors. Gene 33, 103-119. Yu, L . , Tse-Dinh, Y . & Fesik, S.W. (1996) Backbone dynamics if the C-terminal domain of Escherichia coli topoisomerase I in the absence and presence of single-stranded D N A . Biochemistry 35, 9661-9666.  Bibliography  239  Zhang, O., Kay, L . E . , Olivier, J.P. & Foreman-Kay, J.D. (1994) Backbone ' H and N resonance assignments of the N-terminal SH3 domains of drk in folded and unfolded states using enhanced-sensitivity pulsed field gradient NMR techniques. J. Biomol. NMR 4, 845-858. 1 5  Zhu, G. & Bax, A. (1992) Improved linear prediction of damped N M R signals using modified "forward-backward" linear prediction. J. Mag. Reson. 100, 202-207. Zuiderweg, E.R. & Fesik, S.W. (1989) Heteronuclear three-dimensional N M R spectroscopy of the inflammatory protein C5a. Biochemistry 28, 2387-91.  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0061647/manifest

Comment

Related Items