Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Assessment of protein surface hydrophobicity by spectroscopic methods and its relation to emulsifying… Arteaga Mac Kinney, Guillermo Eleazar 1994

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

Item Metadata


831-ubc_1994-953063.pdf [ 10.38MB ]
JSON: 831-1.0088026.json
JSON-LD: 831-1.0088026-ld.json
RDF/XML (Pretty): 831-1.0088026-rdf.xml
RDF/JSON: 831-1.0088026-rdf.json
Turtle: 831-1.0088026-turtle.txt
N-Triples: 831-1.0088026-rdf-ntriples.txt
Original Record: 831-1.0088026-source.json
Full Text

Full Text

ASSESSMENT OF PROTEIN SURFACE HYDROPHOBICITY BY SPECTROSCOPICMETHODS AND ITS RELATION TO EMULSIFYING PROPERTIES OF PROTEINSbyGUILLERMO ELEAZAR ARTEAGA MAC KINNEYB. Eng., Instituto Tecnologico y de Estudios Superiores de Monterrey (Mexico), 1984M.Sc., The University of British Columbia, 1988A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF PHILOSOPHYinTHE FACULTY OF GRADUATE STUDIES(Department of Food Science)THE UNIVERSITY OF BRITISH COLUMBIAJune 1994® Guillermo Eleazar Arteaga Mac KinneyIn presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.Department of FdThe University of British ColumbiaVancouver, Canada(Signature)Date J4, \qq4DE-6 (2/88)ABSTRACTLife as we know it originated in water and the interactions between water and organiccompounds has been one of the forces that directs life forms on earth. Thirty five years ago,Kauzmann (1959) introduced the concept of the “hydrophobic effect” and suggested it to be the mainstabilizing force of the native conformation of proteins. Geometric constraints in most proteinsprevent the shielding from solvent of all hydrophobic residues, thus in the native conformation ofmost proteins, some hydrophobic groups are exposed or solvent accessible. These exposed residueshave a potential influence on several protein functions. In order to fully understand these functions,quantification of exposed or effective protein hydrophobicity is essential. Fluorescence probemethods are widely used in food protein studies to estimate surface hydrophobicity, however,controversy still exits about the mechanisms and type of interactions involved in these methods.The purpose ofthis thesis was to critically compare different sectroscopic methds for theirpotential to the study surface hydrophobicity of a group of 10 purified food proteins. The proteingroup consisted of the main proteins from cow’s milk (afl-, I- and ic-casein, a-lactalbumin, f3-lactoglobulin and bovine serum albumin) and some egg white proteins (ovalbumin, lysozyme,ovotransferrin and ovomucoid). The proteins were analyzed in their native state or in four differenturea concentrations (2, 4, 6, and 8 M). Five different methods were evaluated: 4th derivative UVabsorption and two-dimensional fluorescence spectroscopy, which are two methods not commonlyused for food proteins; nuclear magnetic resonance (NMR) and Raman spectroscopy, two techniquesnot previously used for the study ofprotein hydrophobicity; and a fluorescence probe method usinganilino-naphthalene-sulphonate (ANS) and cis-parinaric acid (CPA). Fourth derivative UVabsorption as well as fluorescence spectroscopy give information related only to aromatic aminoacids, mainly Trp and Tyr residues, while both NMR and Raman spectroscopies have the potentialof detecting aliphatic as well as aromatic groups.A second objective of this work was to relate the information obtained with the differentspectroscopic methods to the emulsify-ingproperties (i.e., emulsion activity and emulsion coalescenceUstability) of the proteins. Within this objective, the technique of artificial neural networks (ANN)was evaluated as a new methodology for the development of structure-functionality relationshipsin food proteins.The well established spectral changes due to unfolding were observed in both UV-absorptionand fluorescence spectroscopy. Urea caused a blue shift in the UV-absorbance spectra of mostproteins. Deconvolution of the spectra by computing its fourth derivative was very useful indetecting the relatively small shifts and in assigning spectral regions to specific side chain type,yielding parameters which were significantly correlated with (log) ANS and (log) CPAhydrophobicities. Urea caused marked changes in the fluorescence spectra ofmost proteins, causinga red shift in the emission peak position and decreased in fluorescence intensity, suggestingexposure of Trp and Tyr residues to solvent and collision quenching of fluorescence. For threeproteins (a-lactalbuinin, f3-lactoglobulin and ovotransferrin) fluorescence intensity increased as ureaconcentration increased. This phenomenon has been previously reported to occur for thedenaturation of the two whey proteins but not for ovotransferrin unfolding. The width of theemission peak as well as the shape parameter of peak asymmetry were also affected by urea. Asignificant correlation was found between the position of the emission peak at an excitationwavelength of 290 nm and (log)ANS (r0.485, n=40, p=O.002) and (Iog)CPA (r=O.453, n=40,p=O.004).Urea denaturation simplified the NMR spectra of the proteins, the degree of changedepended on the protein and urea concentration. The NMR spectra of the proteins in 8 M urea weresimilar to those calculated assuming random coil conformation. Using basic principles of NMR,linewidths were used as a crude estimate of mobility/exposure. Proteins with less orderedconformation (e.g., caseins) had NMR resonances with smaller linewidths than more compact andglobular proteins (e.g., a-lactalbumin, lysozyme). Significant correlations were found betweenlinewidths of the methyl proton signal of the native proteins and their ANS and CPAhydrophobicities. The emulsifying properties were also significantly correlated to the methyl signalUIlinewidths. Urea denaturation also caused a downfield shift of the main aliphatic peaks. This wasin accordance with several reports, which have indicated that buried residues tend to presentupfield shifts as compared to the residues in a standard state. Using a set of assumptions, acomputer program was written to estimate the number of exposed aliphatic residues (lie, Leu andVal) employing information of the experimental NMR methyl peak. For native proteins, thiscalculated “aliphatic exposure” was significantly correlated to ANS and CPA hydrophobicities. ANMR cross saturation experiment was also used to estimate exposure of hydrophobic residues ofproteins. For native proteins, a highly significant regression model (R2=0.992, F=240, n=8) wasfound relating the change in NMR integrated areas due to cross saturation and CPAhydrophobicity.Contrary to the large changes observed by NMR spectroscopy, urea caused only smallchanges in the Rainan CH stretching area of proteins. The broadness ( 100 cm’) of tlils proteinRaman region, caused difficulties in detecting spectral alterations. Maximum likelihooddeconvolution was used to expand and quantify these changes. Four deconvoluted peaks were foundto form the broad CH stretching area. Caseins tended to have peaks at higher Raman shifts andwith larger linewidths than whey and egg proteins. Signfficant regression equations were derivedrelating some parameters from these peaks to ANS and CPA hydrophobicities and to the emulsionproperties of the proteins.In order to relate the large number of spectroscopic variables to the two emulsifyingproperties, principal components regression (PCR) and artificial neural networks (ANN) were usedand compared. The prediction ability of ANN was found to be superior to that of PCR, especiallywith cross validation.Overall, these results support the use of fluorescence probes for the estimation of surfacehydrophobicity of food protein. Of the several spectroscopic methods evaluated, NMR spectroscopymay have the greatest potential for analysis of surface residues of proteins. Results also indicatedthe great potential of ANN in elucidating the structure-functionality relationships of proteins.ivTABLE OF CONTENTSABSTRACT iiTABLE OF CONTENTS vLIST OF TABLES xLIST OF FIGURES xiACKNOWLEDGEMENTS xviiiGENERAL INTRODUCTION 1LITERATURE REVIEWA. THE CONCEPT OF H’YDROPHOBICITY 41. Introduction 42. Intermolecular noncovalent interactions 53. Hydrophobicity, surface tension and surface free energy 84. Hydrophobicity and thermodynamics 115. Protein solvent accessible surface area and hydrophobicity 126. Determination of protein hydrophobicity 14a. Hydrophobicity scales 15b. Fluorescence probe methods 20C. Aliphatic and aromatic hydrophobicity of proteins 27B. SPECTROSCOPIC METHODS FOR THE STUDY OF PROTEINHYDROPHOBICITY 301. Derivative ultraviolet spectrophotometry 302. Intrinsic fluorescence 323. Proton nuclear magnetic resonance in protein chemistry 34a. Introduction 34b. Basic principles 35c. Nuclear magnetic resonance of proteins 35i. Introduction 35ii. Aromatic and aliphatic resonances 37iii. Assignments of NMR resonances 37iv. NMR of denatured and native proteins 40v(a). Ribonuclease.(b). Caseins .(c). Lysozymev. NMR linewidths and protein conformationvi. Spin diffusion and protein mobility41414243464. Raman spectroscopy in protein chemistrya. Introductionb. Basic principlesc. Raman studies of proteinsi. Important Raman bands in proteinsii. Raman studies of selected proteins(a). Lysozyme(b). Ribonuclease(c). Milk proteinsiii. Raman and protein hydrophobicity(a). Introduction(b). Effect of chemical environment on Ramanbands(c). The Raman C-H vibrations(d). Rainan vibrations of water1. Chemistry of urea2. Urea as a protein denaturing agent66673. Application of urea denaturation in protein chemistry 71a. Introduction 71b. Biochemistry related studies 72c. Food protein related studies 74i. Effect of urea on some egg white proteins 77ii. Effect of urea on whey proteins 79iii. Effect of urea on milk caseins 82D. EMULSIFICATION AS A FUNCTIONAL PROPERTY OF PROTEINS 851. Protein functionality 852. Prediction of protein functionalitya. Artificial neural networks3. Emulsification as a functional property of proteins..a. Introduction and definitionsb. Emulsion formationc. Emulsion stabilityi. Creamingii. Flocculation4747485051515153535454555763C. EFFECT OF UREA ON PROTEINS 66..8586888889959698viiii. Coalescence.100iv. Estimation of coalescence stability. 1044. Physicochemical properties of egg white and milk proteins 109a. Introduction 109b. Amino acid composition of proteins 109c. Milk proteins 110i. Caseins 110ii. Whey proteins 118d. Egg white proteins 122i. Introduction 122ii. The four main egg white proteins 1225. Emulsifying properties of milk and egg white proteins 127a. Milk proteins 127i. Introduction 127ii. Caseins 127iii. Whey proteins 131b. Egg white proteins 1416. Effect of urea on emulsifying properties of proteins 149a. Introduction 149b. Effect of urea 149c. Summary of the effect of urea on protein stabilizedemulsions 152MATERIALS AND METHODSA. PROTEIN SAMPLES AND CHEMICALS 1531. Preparation of a- and ic-casein 1532. Buffer and urea solution preparation 1553. Protein solution preparation 155B. SPECTROPHOTOMETRIC MEASUREMENTS (FOURTH DERIVATiVESPECTROSCOPY) 156C. TWO-DIMENSIONAL FLUORESCENCE SPECTROSCOPY 156D. RAMAN SPECTROSCOPY 157E. NUCLEAR MAGNETIC RESONANCE (NMR) SPECTROSCOPY 1581. One dimensional NMR of native and denature4 proteins 1582. Spin diffusion experiments of native proteins 159WiF. FLUORESCENCE PROBE HYDROPHOBICITY 1591. General methodology and rational 1592. ANS probe hydrophobicity 1613. CPA probe hydrophobicity 163G. EMULSIFYING PROPERTIES 1631. Emulsifying activity index (EM) 1632. Emulsion stability 164H.. STATISTICAL ANALYSIS 165RESULTS AND DISCUSSIONA. A CLASSIFICATION OF SPECTROSCOPIC METHODS FOR THE STUDYOF PROTEIN CONFORMATION 167B. FLUORESCENCE PROBE HYDROPHOBICITY OF PROTEINS 169C. EMULSIFYING PROPERTIES OF PROTEINS 1741. Emulsion activity index and emulsion stability of protein 1742. Relationship between probe hydrophobicity and emulsifyingproperties 178D. FOURTH DERIVATWE UV SPECTROSCOPY 182E. FLUORESCENCE SPECTROSCOPY OF PROTEINS 1961. Band-shape analysis of intrinsic fluorescence spectra ofproteins 1962. Two-dimensional fluorescence spectroscopy of proteins 1963. Effect of urea on fluorescence peak position 1984. Effect of urea on fluorescence intensity of proteins 2045. Effect of urea on fluorescence peak width 2066. Effect of urea on peak asymmetry 211vu’7. Summary and comments.214F. NUCLEAR MAGNETIC RESONANCE (NMR) SPECTROSCOPY OFPROTEINS 2161. Effect of denaturation on the NMR spectra of proteins 2162. NMR spectra parameters, probe hydrophobicity values andemulsifying properties of proteins 2383. Estimation of aliphatic group exposure using NMR simplexoptimization 2584. NMR cross saturation and hydrophobicity 2645. Summary and comments 278G. RAMAN SPECTROSCOPY OF PROTEINS 2801. Raman spectra of amino acids: CH stretching area 2802. General features of the Raman CH stretching area of proteins . . ... 2873. Effect of urea denaturation on the Raman CU stretching area ofproteins 2914. Raman spectra parameters, probe hydrophobicity and emulsifyingproperties 3125. Summary and conclusions 319H. PREDICTION OF PROTEIN FUNCTIONALITY USING ARTIFICIALNEURAL NETWORKS 3211. Artificial neural networks (ANN) 3212. Development of structure-functionality relationshipsusing ANN and principal component regression (PCR) 3213. Prediction ability of ANN and PCR models 324a. Prediction of emulsion activity index 324b. Prediction of emulsion stabffity 329c. Summary and conclusions 329GENERAL CONCLUSIONS 331REFERENCES CITED 335ixLIST OF TABLESTable 1. Selected hydrophobicity scales.21Table 2. Proton magnetic resonance position of amino acids residuesfor computing the spectra of random coil proteins 38Table 3. Twelve main physical factors affecting the stability andrheology of emulsions 97Table 4. Amino acid composition of the model proteins 114Table 5. Effect of urea on the fourth derivative spectra parametersof the proteins analyzed 184Table 6. Spectroscopic parameters used in the development of artificialneural networks and principal component regression models forpredicting emulsifying properties of proteins 322Table 7. Comparison of artificial neural networks and principal componentregression for prediction of emulsifying activity index andemulsion stability using 26 or 28 spectroscopic variables 327xLIST OF FIGURESFig. 1. Radar chart of the amino acid composition of caseins 111Fig. 2. Radar chart of the amino acid composition of whey proteins 112Fig. 3. Radar chart of the amino acid composition of egg proteins 131Fig. 4. Pulse profile used for the NMR cross-saturation experiments 160Fig. 5. Fluorescence probe hydrophobicities (HO) of the 10 proteins analyzed 170Fig. 6. Fluorescence probe hydrophobicities (HO) (log scale) of the 10 proteinsanalyzed 171Fig. 7. Effect of urea on ANS hydrophobicity (HO) of the analyzed proteins.Within each protein, HO is expressed taking the value for the nativeprotein as 100 172Fig. 8. Effect of urea on CPA hydrophobicity (HO,A) of the analyzed proteins.Within each protein, HOcrA is expressed taking the value for the nativeprotein as 100 173Fig. 9. Effect of urea on emulsifring activity index (EM) of the analyzed proteinsEM expressed as absorbance at 500 inn 175Fig. 10. Effect of urea on emulsion stability (ES=% change in EM) of theanalyzed proteins . 176Fig. 11. Relationship between log CPA and log ANS hydrophobicities (HO) andemulsifying activity index (EM) of native and urea-denatured protein(2, 4, 6 and 8 M urea). EM expressed as absorbance at 500 nm 179Fig. 12. Relationship between log CPA and log ANS hydrophobicities (HO) andemulsion stability (ES=% change in EM) of native and urea-denaturedprotein (2, 4, 6 and 8 M urea) 180Fig. 13. Fourth derivative spectra of native, urea-denatured andSDS-heated -lactoglobu1in 183Fig. 14. Effect of urea concentration on the position of the Trp minimum of thefourth derivative spectra of the 10 proteins analyzed 188Fig. 15. Effect of urea concentration on the Tyr geometric factor of thefourth derivative spectra of the 10 proteins analyzed 189Fig. 16. Effect of urea concentration on the position of the Tyr minimum of thefourth derivative spectra of the 10 proteins analyzed 190xiFig. 17. Relationship between Trp fourth derivative spectra parameters withthe logarithm of ANS or CPA fluorescence probe hydrophobicities 192Fig. 18. Relationship between Tyr fourth derivative spectra parameters withthe logarithm of ANS or CPA fluorescence probe hydrophobicities 193Fig. 19. Relationship between Phe fourth derivative spectra peak position withthe logarithm of ANS or CPA fluorescence probe hydrophobicities 194Fig. 20. Two dimensional fluorescence spectra of native andurea-denatured 51-casein 197Fig. 21. Effect of urea concentration and excitation wavelengthon the emission peak position of caseins 199Fig. 22. Effect of urea concentration and excitation wavelength on theemission peak position of whey proteins 200Fig. 23. Effect of urea concentration and excitation wavelength on theemission peak position of egg proteins 201Fig. 24. Relationship between emission peak position (excitation wavelength 290 nm)with the logarithm of ANS or CPA fluorescence probe hydrophobocities 203Fig. 25. Effect of urea concentration on the relative fluorescence intensity(integrated fluorescence area) of the 10 proteins analyzed 205Fig. 26. Calculated fluorescence intensity per aromatic residue of urea-denaturedand native proteins 207Fig. 27. Effect of urea on the emission peak width of caseins 208Fig. 28. Effect of urea on the emission peak width of whey proteins 209Fig. 29. Effect of urea on the emission peak width of egg proteins 210Fig. 30. Effect of urea concentration on the emission peak asymmetry of the10 proteins analyzed (excitation wavelength of 270 nm) 212Fig. 31. Effect of urea concentration on the emission peak asymmetry of the10 proteins analyzed (xcitation wavelength of 297 nm) 213Fig. 32. ‘H-NMR spectrum of native;1-casein in 0.1 M phosphate d-buffer(pD,=7.5). Temperature 22°C. Protein concentration 1.5% 217Fig. 33. ‘H-NMR spectrum of urea-denaturedcç1-casein (8 M urea-d4 in 0.1 Mphosphate d-buffer 0.1 M, pDa=7.5). Protein concentration 1.5% 218Fig. 34. Aliphatic region of the1H-NMR spectrum of ovotransferrin 219Fig. 35. Aromatic region of the ‘H-NMR spectrum of ovotransferrin 220xuFig. 36. Aliphatic region of the ‘H-NMR spectrum of lysozyme 221Fig. 37. Aromatic region of the ‘H-NMR spectrum of lysozyme 222Fig. 38. Aliphatic region of the ‘H-NMR spectrum of bovine serum albumin 223Fig. 39. Aromatic region of the ‘H-NMR spectrum of bovine serum albumin 224Fig. 40. Aliphatic regiolA of the ‘H-NMR spectrum of 3-lactoglobulin 225Fig. 41. Aromatic region of the ‘H-NMR spectrum of -lactoglobulin 226Fig. 42. Computer simulated1H-NMR spectrum ofcE1-casein in arandom-coil conformation 228Fig. 43. Computer simulated ‘H-NMR spectrum of -casein in arandom-coil conformation 229Fig. 4-4. Computer simulated ‘H-NMR spectrum of ic-casein in arandom-coil conformation 230Fig. 45. Computer simulated ‘H-NMR spectrum of a-lactalbumin in arandom-coil conformation 231Fig. 46. Computer simulated ‘H-NMR spectrum of JI-lactoglobulin in arandom-coil conformation 232Fig. 47. Computer simulated ‘H-NMR spectrum of bovine-serum albumin in arandom-coil conformation 233Fig. 48. Computer simulated1H-NMR spectrum of ovalbumin in arandom-coil conformation 234Fig. 49. Computer simulated1H-NMR spectrum of lysozyme in arandom-coil conformation 235Fig. 50. Computer simulated ‘H-NMR spectrum of ovomucoid in arandom-coil conformation 236Fig. 51. Computer simulated ‘H-NMR spectrum of ovotransferrin in arandom-coil conformation 237Fig. 52. Relationship between calculated total hydrophobic NMR areaper residue and Bigelow’s total hydrophobicity 239Fig. 53. Relationship between experimental total hydrophobic NMR areaper residue and Bigelow’s total hydrophobicity 240Fig. 54. Relationship between some parameters of the methyl ‘H-NMR peakand log fluorescence probe hydrophobicities 242xliiFig. 55. Relationship between some parameters of the methyl ‘H-NMR peakand emulsif5ring properties of proteins. Native proteins only 243Fig. 56. Prediction of emulsifring properties of proteins using parametersof the methyl ‘H-NMR peak. Native protein only 244Fig. 57. Effect of urea on the linewidth at 10% peak height of the methyl‘H-NMR peak of proteins 245Fig. 58. Effect of urea on the linewidth at 50% peak height of the methyl1H-NMR peak of proteins 246Fig. 59. Relationship between some NMR peak parameters andlog fluorescence probe hydrophobicities (HO) 248Fig. 60. Relationship between some NMR peak parameters andemulsifring properties of proteins 249Fig. 61. Effect of side chain exposure on secondary shifts of aliphatic protonsof lysozyme. Methyl protons of fle, Leu and Val were only considered 251Fig. 62. Average resolution of the aromatic ‘H-NMR region of nativeand denatured proteins 254Fig. 63. Effect of urea on aromatic ‘H-NMR resolution (ARES).ARES was expressed as a percentage of the resolution ofthe simulated NMR spectra 256Fig. 64. Relationship between aromatic ‘H-NMR resolution (ARES)and log fluorescence probe hydrophobicities (HO).ARES was expressed as a percentage of the resolution ofthe simulated NMR spectra 257Fig. 65. Relationship between aromatic ‘H-NMR resolution (ARES)and emulsitying properties of proteins 259Fig. 66. 1H-NMR methyl peak of f3-casein, ovotransferrin and -lactogIobulin 260Fig. 67. Simplex optimization fitting results for native lysozyme 261Fig. 68. Relationship between % content of aliphatic amino acids andlog fluorescence probe hydrophobicities 263Fig. 69. Effect of cross saturation on the ‘H-N)4R spectrum of nativea11-casein 266Fig. 70. Effect of cross saturation on the1H-NMR spectrum of native -casein 267Fig. 71. Effect of cross saturation on the ‘H-N1IR spectrum of native ic-casein 268Fig. 72. Effect of cross saturation on the ‘H-NMR spectrum of native ct-lactalbumin ... 269xivFig. 73. Effect of cross saturation on the ‘H-NMR spectrum of native -lactoglobu1in... 270Fig. 74. Effect of cross saturation on the ‘H-NMR spectrum of nativebovine serum albumin 271Fig. 75. Normal and cross-saturated1H-NMR spectra of native andreduced ovalbumin 272Fig. 76. Effect of cross saturation on the ‘H-NMR spectra of native lysozyme 273Fig. 77. Effect of cross saturation on the ‘H-NMR spectra of native ovomucoid 274Fig. 78. Effect of cross saturation on the ‘H-NMR spectra of native ovotransferrin .... 275Fig. 79. Relationship between the change in aliphatic (CHAL) and aromatic (CHAR)1H-NMR area due to cross saturation and log fluorescence probehydrophobicities. Native proteins only 276Fig. 80. Relationship between the change in aliphatic (CHAL) and aromatic(CHAR)1H-NMR area due to cross saturation and emulsifying propertiesof the protein 277Fig. 81. Raman spectra of alanine and isoleucine in the CH stretching region.Amino acids were dissolved in 1.0 M DC1 in D20 to a final concentrationof 1.0 M for alanine and 0.2 M for isoleucine 281Fig. 82. Reman spectra of valine and leucine in the CH stretching region.Amino acids were dissolved in 1.0 M DC1 in 1)20 to a finalconcentration 0.2 M 282Fig. 83. Raman spectra of lysine and aspartic acid in the CH stretching region.Amino acids were dissolved in 1.0 M DC1 in 1)20 to a finalconcentration of 1.0 M 283Fig. 84. Reman spectra of proline and threonine in the CH stretching region.Amino acids were dissolved in 1.0 M DC1 in 1)20 to a finalconcentration 1.0 M 284Fig. 85. Reman spectra of phenylalanine (0.2 M in 1 M DCI in D20) in thewavenumber shift regions of 2800-3200 cm1 and 200-1800 ciii’ 285Fig. 86. Reman spectra of tyrosine and tayptophan in the CH stretching region.Amino acids were dissolved in 1.0 M DCI in 1)20 to a finalconcentration 0.2 M 286Fig. 87. Reman spectra of solid alanine and solid isoleucine in theCH stretching region 288Fig. 88. Reman spectra of -lactoglobulin and ovotransferrin in theCH stretching region. Native and urea denatured samples.Protein concentration was 5% 289xvFig. 89. Raman spectra of -casein and ovalbumin in theCH stretching region. Native and urea denatured samples 290Fig. 90. Raman spectra of neat dioxane and of 10% (v/v) dioxane in D20in the in the CH stretching region 293Fig. 91. Raman spectra and resulting maximum likelihood deconvolutedspectra of native and urea denatured f3-casein 295Fig. 92. Effect of baseline correction, smoothing and maximum likelihooddeconvolution on the Raman spectrum of native bovine serum albumin 296Fig. 93. Effect of baseline correction, smoothing and maximum likelihooddeconvolution on the Raman spectrum of urea denatured bovine serumalbumin 297Fig. 94. Effect of baseline correction, smoothing and maximum likelihooddeconvolution on the Raman spectrum of native ic-casein 298Fig. 95. Effect of baseline correction, smoothing and maximum likelihooddeconvolution on the Raman spectrum of urea denatured ic-casein 299Fig. 96. Experimental and fitted Raman spectra of a saturated solutionof KNO3 in the wavenumber shift region 1000-1100 cm1.A Lorentzian peak shape was assumed for fitting the spectrum 300Fig. 97. Maximum likelihood deconvoluted Raman spectra of urea denaturedand nativea11-casein 302Fig. 98. Maximum likelihood deconvoluted Raman spectra of urea denaturedand native (3-casein 303Fig. 99. Maximum likelihood deconvoluted Raman spectra of urea denaturedand native ic-casein 304Fig. 100. Maximum likelihood deconvoluted Raman spectra of urea denaturedand native a-lactalbumin 305Fig. 101. Maximum likelihood deconvoluted Raman spectra of urea denaturedand native -lactoglobulin 306Fig. 102. Maximum likelihood deconvoluted Raman spectra of urea denaturedand native bovine serum albumin 307Fig. 103. Maximum likelihood deconvoluted Raman spectra of urea denaturedand native ovalbumin 308Fig. 104. Maximum likelihood deconvoluted Raman spectra of urea denaturedand native lysozyme 309xviFig 105. Maximum likelihood deconvoluted Raman spectra of urea denaturedand native ovomucoid 310Fig. 106. Maximum likelihood deconvoluted Raman spectra of urea denaturedand native ovotransferrin 311Fig. 107. Relationship between some parameters of the CE1 deconvoluted Rainanpeak and log CPA hydrophobicity (HO,A) 313Fig. 108. Relationship between some parameters of the CE2 and CE3 deconvolutedRaman peaks and log CPA hydrophobicity (HO) 314Fig. 109. Relationship between some parameters of the deconvoluted Raman peaksand log CPA hydrophobicity (HOA) 315Fig. 110. Relationship between some parameters of the deconvoluted Raman peaksand emulsifying properties of the proteins. Native proteins only 317Fig. 111. Relationship between some parameters of the deconvoluted Raman peaksand emulsifying properties of the proteins 318Fig. 112. Prediction ability of the trained artificial neural network for emulsionstability and emulsifying activity index 325Fig. 113. Non-cross validated prediction ability of the developed principal componentregression model for emulsion stability and emulsion activity index 326xviiACKNOWLEDGMENTSI wish to thank my supervisor, Dr. S. Nakai for his invaluable guidance and support throughoutmy graduate studies. I especially thank him for trusting me and for giving me the freedom topursue my ideas. His understanding and patience with my “organized” mess is also appreciated.I also wish to thank Dr. E. Li-Chan for her advice and moral support during some critical timesof my graduate program, but especially, for treating me as a scientist when I was only astudent. I also extend my thanks to the other members ofmy research committee, Dr. F. Aubkeand Dr. T. D. Durance for their helpful advice and comments on my research program andthesis. To my friend Angela, my eternal gratitude for the many things she helped me with. Fordoing the monumental work of proofreading this work, for her babysitting, for her advice andsupport and for never asking ‘When are you going to finish with your thesis?”. Your friendshipAngela will be one of my best memories. I would not have been able to complete my Ph.D.without the help, love, understanding and sacrifice of Cecilia, my best friend and wife. Herconfidence in me was the force behind this work. Two people who did everything in their powerto prolong the completion of my degree were my son, Emilio and my daughter, Paulina. To themI dedicate this work. Through them, I rediscovered the child within me. Their love andimagination helped me to keep my sanity. I would like to thank my parents, Doris and Juan,for their support, love and understanding. I also thank my “second set” of parents, my parents-in law for their support and love. Thanks go to all of them for their “sacrifice” in coming manytimes to Vancouver to visit and help us. I also want to dedicate this work to my sister, Monicaand my brother Sergio. Without knowing, they have influenced me the most. Through them, Ilearned to be humble and to try to do the very best with my life. Life has giving me so much,not because I deserved it, but because I am one of the lucky ones that has my health and forthat I thank you God.XVII’GENERAL INTRODUCTION“Proteins” is a word derived from the Greek “proteios” meaning “primary”. The word“protein” is also related to the Greek God “Proteus”, which could assume many different forms.Since proteins perform a wide variety of bielogical functions, proteins are doubly well named(Robson and Gamier, 1988). Proteins are possibly the most reactive among the major foodcomponents. They can react with other proteins, reducing sugars, fats and many other foodcomponents. In many foods, sensory qualities depend upon physicochernical properties andinteractions of protein components with other elements in the system. Such properties, whichaffect the influence of proteins in food systems, are functional properties of proteins (Kinsella,1982).Proteins can be viewed as linear macromolecules with nonrepetitivespecific covalent tructuresbeing able to adopt relatively fixed three-dimensional structures, or conformations. The covalentstructure is determined by the structure ofthe 20 different amino acids and by the order in whichthey are linked together, based on the genetic information, into a polypeptide chain (Creighton,1992). Recent advances in molecular biology have shown that structure is crucial for a properunderstanding of protein function. Since chemistry can be defined as the discovery of the orderthat exists in chemical systems and establishing quantitative rules which describe this order, theprediction ofthe three dimensional structure ofproteins from the amino acid sequence is basicallya chemical problem ofgreat complexity. The study ofthe relationship between chemical propertiesand structure is an important part of modern chemistry. When structural parameters of a seriesof compounds are related to some biological property of the same compounds, the term QSAR iscommonly used. The study of quantitative structure-activity relationships (QSAR) is essentiallythat offinding quantitative relationships between the biological activities of compounds and theirchemical structures. In statistical terms, the biological activity data is taken as dependentvariable(s), while the chemical data is taken as independent variables.1Chemical data used to describe the molecular structures of biologically active compounds are themost important elements of a QSAR study. Most descriptors which have been used are based onlinear free energy-related variables related to hydrophobic, electronic and steric properties.In Food Science, QSAR studies have stimulated a better understanding of the factors that affectprotein functicnality, and in some cases it has been possible to predict the functionality of selectedproteins by using empirical models (Nakai and Li-Chan, 1988). The manner in which foodproteins behave in a given food system or food application is a manifestation of the fundamentalphysicochemical properties of the proteins under the given conditions. One of the most importantphysicochemical property of a protein is its hydrophobicity. Although hydrophobicity and/orhydrophobic interactions are generally regarded to be very important to both protein structureand protein function, a quantitative assessment of these phenomena has been shown difficult aswell as controversial. A variety of methods have been proposed, knit there is no consensus onwhich method gives the “true” hydrophobicity for use in QSAR.The objectives of this thesis were: (1) to evaluate different spectroscopic methods for theestimation of surface hydrophobicity of proteins derived from milk and egg, (2) to study the effectof urea denaturation (i.e., unfolding) on emulsifying properties of proteins derived from milk andegg and (3) to study the relationship between surface hydrophobicity and emulsifying propertiesof native and urea-denatured proteins derived from milk and egg.In previous work from our research group and by other groups, changes in protein conformation(i.e., denaturation) were brought about by heating, pH changes, and chemical or enzymaticmodffication. In this work, urea was used as the main protein denaturant. Contrary to otherdenaturants, such as heating, urea tends to unfold most proteins, thus increasing the solventexposure of most protein groups. This well-established unfolding is important for this work, sincesurface hydrophobicity is strongly related to the solvent exposure of hydrophobic groups. Ureatends to produce little interference with most spectroscopic methods, espcially 1H NMR and2Raman, and most proteins remain soluble at high urea concentrations, avoiding reductions insignal to noise ratios due to turbidity/coagulation effects.Most modem scientists recognize that the complexity of external reality is so great that tocomprehend it we must describe it in terms of simplified models. The relationship of models toreality has concerned thinkers since Plato (428-348 BC). Idealists, like Plato and Hegel(1770-1831), hold that concepts such as space, order and causation are attributes of the humanmind which organize our perception of the world. This school of thought believes that modelspreexist in ou minds. A second school is that of the Empiricists such as David Hume (1711-1776).They believe that our sense organs record and define the order that exist in the world about us.Space, time, forces and so forth are attributes of this world that exists independently of us(Eisenberg and Crothers, 1979). The concept of protein hydrophobicity was developed to explaincertain observations and is at best a simplified representation of reality. Is hydrophobicity realor just a thought in our minds? This work by no means endeavored to answer this question, butit evaluated the value of several methods for the study of protein hydrophobicity in some foodproteins.3LITERATURE REVIEWA. THE CONCEPT OF HYDROPHOBICITY1. IntroductionBefore any physical/chemical property of a system can be quantified a proper andaccepted definition of the property is needed. Everyone agrees on the meaning of terms such asboiling point or refractive index. However, no such consensus exists regarding a definition of theterm hydrophobicity.Hydrophobicity has been usually explained and defined in phenomenological or empirical ways,which do not lend themselves to thermodynamic analysis (Duncan-Hewitt, 1990).Hydrophobic effects have been said to arise because “like prefers like”, suggesting tliat similarmolecules will interact in a favourable way only if their physicochemical characteristics aresimilar. Another school of thought explains hydrophobicity in terms of special characteristics ofthe solvent (i.e., water). Hydrophobicity has also been related to classical intermolecularinteractions (e.g., van der Waals interactions).In simple terms a hydrophobic substance is one that “dislikes” water. Two basic approaches havebeen applied for predicting/estimating the degree that a substance “likes” water or itscompatibility with water:(1) Physicochemical parameters such as interfacial tension, refractive index, dielectricconstant and dipole moments, are commonly used to predictlestimate water likeness ofsmall compounds. The basic premise in these cases is that if a molecule is similar towater (e.g., high dipole moment), it will interact easily with water. It is important topoint out that all these properties are a reflection of the internal structure andintermolecular interactions occurring in the pure solutes. The solvation ability ofsolvents, including water, has also been related to its physicochemical properties (seeReichardt (1988) for a review in this area). As molecules become more complex, theusefulness ofsimple physicochemical properties as predictors ofwater likeness decreases;in addition, their determination and/or estimation becomes more difficult. For largemolecules other parameters, in many cases quite empirical, are used in estimating waterlikeness.4(2) Aqueous solubility/insolubility is the most common phenomenon to use for estimatingwater likeness of small and large molecules. The centuries old empirical rule of “likedissolves like” is a macroscopic manifestation of intermolecular interactions betweensolvent and solute molecules which determines mutual solubility (Reichardt, 1988).Tanford (1980) defines hydrophobic substances as substances that are readily soluble inmany nonpolar solvents but only sparingly soluble in water. Substances that presentvery high intermolecular cohesion with low solubility in all solvents are not hydrophobiccompounds.Essentially, the first approach uses information on the structure and intermolecular forces ofthe solute, while the solvent (i.e., water) has a minor role. In the second approach, the point ofview is completely opposite; the behaviour of the solute in the solvent determines if a compoundis hydrophobic or not.2. Intermolecular noncovalent interactionsIntermolecular noncovalent interactions (INI) and aqueous solubility are strongly related,thus a basic understanding of INI is needed. Intermolecular noncovalent interactions were firstproposed by van der Waals in 1873 to explain the deviation in volumetric behaviour of manygases from the ideal gas equation. He proposed an equation containing correction factors, whichfor n moles of gases is:(P+ )(V-nb)=nRTThe quantities a and b are characteristics of each gas and are determined experimentally. Thetermn2a/tP arises from the forces of intermolecular attractions, while nb accounts for the finitevolume of the individual molecules; both effects were neglected in the deviation of the ideal gasequation. The value of a is proportional to the forces between a pair of molecules, but the totaleffect for a given volume is the summation over all the pairs of molecules. The constant brepresents the volume that, for each mole of gas, is effectively unavailable, and for n moles, the5volume nb is subtracted from the overall measured volume to obtain the ideal free volume (Brey,1978).Reichardt (1988) c1assifled attractive van der Waals interactions in two main categories: the firstcategory comprises the dipole-dipole, dipole-induced dipole and dispersion forces. These forcesare non-specific and cannot be completely saturated. The second group is formed by hydrogenbonding forces and the forces of charge or electron transfer.Dipole-dipole forces are electrostatic interactions between molecules which have a permanentdipole moment (i.e., polar molecules), while dipole-induced dipole forces are associations thatinvolve a molecule possessing a permanent dipole moment and a molecule with an induceddipole moment.Dispersion forces, also called London dispersion forces, are universal for all molecules and theyare primarily responsible for the interaction of molecules which possess neither free charge norpermanent dipole moment (i.e., non-polar molecules). London forces are attractive in nature, butthey are called dispersion forces, because the molecular property of polarizability (cx) involvedin these forces, also determines the extent to which a substance can disperse a beam of lighttransmitted into its various coloured components.For two individual molecules (i.e., in the gas phase) the magnitude of the London dispersionforces depends upon a and the inverse sixth power of the intermolecular separation. This strongdistance dependence makes the dispersion forces between two individual molecules effective atonly short range. Polarizability (a) is an electric property of the molecule related to the ease withwhich the charges (electron and protons) respond to the presence of an electric cloud. Electricpolarizabiity Ce) may be estimated by measuring the refractive index (RI) and using theMaxwell relation (R12=e). Large values of RI suggest the presence of strong dispersion forces.Due to this fact, RI is sometimes used as a parameter for quantitative structure activityrelationships (QSAR) of small molecules.6For an assembly of molecules (i.e., condensed phase), dispersion forces are, to a firstapproximation, additive and the interaction energy between two colloidal particles (e.g., proteinmolecules) can be computed by summing the attraction between all interparticle molecule pairs.According to Hamaker (1937) the result of these summations predicts that the Londoninteraction energy or force between colloidal particles decays much less rapidly than thatbetween individual molecules. The magnitude of this interaction energy or force depends onthree factors: (1) the distance separating the particles, (2) the diameters of the particles, and (3)the intrinsic properties of the particles forming the system, represented by the Hamakerconstant (A) of the system. If for a given system A is known, the interaction energy or force dueto dispersion forces at any distance is easily calculated. For two particles of the same sizeseparated by very short distances (in comparison to their diameter) the interaction force (invacuum) depends on the squared separation distance:F=-A(lJ24)(x2)Although Hamaker (1937) considered Ate be independent ofthe separation, more recent theorieshave shown that it is not usually the case. According to the Dzyaloshinskii, Lifshitz andPitaevskii (DLP) theory (Nir, 1976; van Oss, 1988), the Hamaker constant must be replaced bya Hamaker function which, in general, is a function of separation and temperature. The DLPtheory predicts a monotonic decreas in A as separation distance increases and at largeseparations, A becomes a constant.If the particles are immersed in a liquid, the presence of the dispersion medium lowers theinteraction energy. In this case the A has to be replaced by an effective A. For the interactionbetween two particles, 1 and 2, in a dispersion medium 3 the effective Hamaker constant (A132)as:A1=A12+A.A3-The interaction between particles of phases made of different substances is assumed to be equa1to the geometric mean of the attractions of each particle or phase to itself:7A12=(A11* and S— (*)9Using this equation, the effective A becomes:A _(A ½A ½(A ½A ½£132_ ‘33 .‘ ‘22 33If the two particles are of the same material, this expression becomes:A _(A ½A ½2‘13 “11 33 IThree important conclusions are derived from these equations:(1) The interaction between particles of the same material is always attractive (A131 isalways positive).(2) Repulsive dispersion forces can occur when A has a value intermediate between A11 andA.(3) When the particles and the dispersion medium are similar(A11A ), the value ofA1is very small. Since in this case particle-particle interactions are as favourable asparticle-solvenL interactions, the particles will be soluble or form a stable dispersion.The London dispersion energy or force of a system made of a particle (1) and water (3) (i.e., A13)can be taken as a measurement of the hydrophobicity of the substance; however, since thedetermination of A for colloid systems is quite difficult and prone to errors (see Nir, 1976 for areview), alternate routes in quantilijing the London dispersion forces in colloidal systems havebeen taken.3. Hydrophobicity, surface tension and surface free energyA macromolecular manifestation of the degree and type of interactions present in asubstance is its surface tension. Surface tension can be defined as the amount of work requiredto expand the surface by unit area (Shaw, 1992). The forces responsible for surface/interfacialtension include dispersion, hydrogen bonding, dipole-dipole, etc., and for some compounds suchas mercury, metal bonding. These forces can be considered additive. For example, the surfacetension of water can be taken as a sum of a dispersion force contribution () and a hydrogenbonding contribution (f). In the case ofhydrocarbon oils the surface tension is taken as due onlyto the dispersion forces:8Ywater ==When water and a hydrocarbon oil are mixed, an interface is formed. The resulting surfacetension at the interface (i.e., interfacial tension) is due to the combined effect of the forcesforming the surface tension of water and the hydrocarbon oil. Fowkes (1964) using an approachsimilar to the one used for the calculation of the effective Hamaker constant, proposed that thehydrocarbon oil-water interfacial tension (y) is the geometric mean of the hydrocarbon oilsurface tension y0 (due only to dispersion effects) and the water surface tension (due todispersion and hydrogen bonding):_Tow— ro+:w+r w’-”ro yw’Since surface tension is numerically equal to the surface free energy per unit area, by using theabove reasoning, the surface free energy due to dispersion, hydrogen bonding, etc., can beestimated. Free energy is a thermodynamic parameter that encompasses both entropic andenthalpic effects in a reaction or interaction. It gives a quantitative estimation ofthe most likelydirection of change (e.g., attraction or repulsion).van Oss’s group (1985, 1988, 1990 and 1992) has shown that two groups offorces are responsiblefor the surface/interfacial tension of a substance or system. The first type of forces is calledLifshitz-van der Waals forces and clusters together the dispersion, dipole-dipole and dipole-induced forces. The second group, which includes hydrogen bonding forces, are the Lewis acid-base CAB) forces (electron acceptor/electron donor interaction).Using similar combination rules as for the estimation of effective Hamaker constant, van Oss’sgroup stated that the surface free energy (or interaction energy) between two materials, i andj, immersed in a liquid w is given by= /,W +y - ..1’.BA negative value of G indicates that the interaction at a minimum equilibrium distance(assumed to be =4.6 A) between material i and j in water will be an attraction.9For proteins, surface tension components are computed from contact angle measurementsbetween a dry or hydrated layer of protein and solvents, for which the contribution of LW andAB forces to their surface tension is known.As a quantitative measurement of the hydrophobicity of a protein, van Oss (1992) proposed touse the free energy of interaction (in vacuo) between the protein (p) and water (w):= -L2G-y +y,,The smaller this attraction (i.e., AGE,, more positive) the more “hydrophobic” the protein. Wateris tne least “hydrophobic” substance; the free energy of interaction of water to itself is AG=-145.5 mJIm2. This value can be taken as the lower limit of hydrophobicity. Zein, a veryhydrophobic protein from corn, has a eG=-103.5 mJ/m2 while the AG for fibrinogen, ahydrophilic protein, is -134.9 mJ/m2.The hydrophobicity approach of van Oss’s group brings together the concepts of intermolecularnoncovalent interactions and surface thermodynamics. This approach has sound theoretical basisand has been shown to predict, for a limited number of proteins, some protein-solventphenomena (e.g., solubility and chromatographic behaviour). Some of the most importantlimitations of this approach when applied to proteins are:1. For the determination of the contact angles a dry or hydrated layer of protein is used,and it is assumed that in this form, the protein is in its native conformation. In additionthe determination of contact angle is not straightforward, requiring considerableexpertise.2. It is well known that some proteins present surface denaturation at the air-liquidinterface; this conformational change of the protein causes a decrease in the interfacialtension, a fact which helps in the formation of emulsions and foams. This surfacedenaturation of proteins is avoided in the van Oss approach by using a thick layer ofhydrated protein, thus the protein parameters obtained by this methodology may havelimited practical value when in the process under study, protein surface denaturationis likely to occur, which is the case in many food processing operations, such asemulsification, whipping, etc.More research is needed in this area to determine the practical value of this methodology.104. Hydrophobicity and thermodynamicsIn general, two thermodynamic parameters determine the most likely direction of a chemicalor physical process: the change in enthalpy and the change in entropy. The change in enthalpy(H) is related to the energy of the system, while the change in entropy (AS) is associated withdegree of randomness of the system. These two parameters contribute to the change in freeenergy (oG) of the process:AG = oH - ToSKauzmann in 1959 proposed that since the interaction ofnonpolar amino acid chains with wateris entropically unfavourable, these groups tend to be located in the interior of native globularproteins. This phenomenon is called the hydrophobic bond or hydrophobic effect. It is importantto point out that in the concept of the hydrophobic effect, the attractions between two nonpolargroups in water are not due to dispersion forces but to their thermodynamically disadvantageouscontacts with water.In studying the thermodynamics of solution both enthalpic and entropic effects must be takeninto account. As an example, consider the mixing of two pure liquids A and B, where the onlyinteractions present in both A and B are dispersion (London) forces. Mixing of A and B is drivenby an entropic effect (i.e. AS is positive), because more disorder is present in the mixture ABthan in the individual components. On the other hand, mixing is opposed by the enthalpies ofinteraction (oH is positive), since the forces leading to AB attractions are smaller than thoseleading to AA or BB attractions.Kauzmann (1959) noted that nonpolar groups tend to organize water molecules about them inhydrogen-bond cages. This ordering causes an unfavourable decrease in entropy of the water,thus the conformation of a native protein tends to minimize the contact of nonpolar groups tothe solvent, in order to minimize its free energy. Recent studies have shown that while at lowtemperature the insolubility of nonpolar groups in water is entropy driven, at high temperature11(130 -160°C) insolubility is due only to enthalpy effects (Diii, 1990; Creighton, 1991; Murphy andFreire, 1992).When protein is unfolded (e.g., with urea) the polypeptide backbone is unwound and thenonpolar side chains become more exposed to the solvent. The transfer of small nonpolarmolecules such as methane and benzene to water from nonpolar solvents have been used tosimulate this aspect of protein denaturation.Quantification of the hydrophobic effect in a protein has been traditionally made by calculatingthe free energy of transfer for all the amino acid side chains forming the protein from a(arbritarily selected) nonpolar solvent to water. These transfer free energies can be easiiycalculated using solubility data of each amino acid in both water and the nonpolar solvent.Transfer free energy values are sometimes refered to as hydrophobicities. More discussion of thistype of hydrophobicity is given in the section on Hydrophobicity Scales.5. Protein solvent accessible surface area and hydrophobicityThe term “accessible surface area” (ASA) was introduced by Lee and Richards in 1971 toquantitatively describe the extent to which atoms on the protein surface can form contact withwater. For a particular protein atom it is defined as the area over which the centre of a watermolecule can be placed while retaining van der Waals contacts with that atom and notpenetrating any other atom.This term was later shown to be related to hydrophobic (i.e., transfer) free energies (Chothia,1974). The ASA of a protein is not its area of molecular surface. It is measured as one proberadius (1.4 A) away from the molecular surface and encloses half a layer of solvent in additionto the molecular volume.In order to calculate ASA of a protein its tertiary structure is needed. As mentioned by Lee andRichards (1971) this ASA is a static ASA since the protein is assumed to be fixed in space andthe methodology does not consider potential flexibility or movement of groups in the structure.12The ASA concept has mainly been used to quantify to what extent the hydrophobic effect isinvolved in the formation and stability of the native protein structure (Eisenberg andMcLachlan, 1986); however, the same concept can be used to estimate the “static” surfacehydrophobicity of a native protein by calculating its hydrophobic ASA.Miller et al. (1987) found that for 46 monomeric proteins (molecular weight range 4000 to35,000) the ASA was related to the protein molecular weight (MW) by the following power law:ASA = 6.3*MW073This ASA was found to be, on an atomic level, 57% nonpolar, 24% polar and 19% charged, whilethe molecular surface area buried inside the protein was 58% non-polar, 39% polar and 4%charged. On an amino acid level, it was also observed that on average the protein interior isenriched in large aliphatic and aromatic residues; Val, Leu, Tie and Phe ionstitute 44% of theinterior of the proteins, but only 14% of the surface. Interestingly, both Trp and Tyr wire foundto be located preferentially on the protein surface. The protein surface was found to be rich incharged residues. On average Asp, Glu, Lys and Arg constitute 27% of the protein surface andonly 4% of the interior. Very often the nonpolar residues are organized in hydrophobic clusters.These clusters are functionally very significant since they enable proteins to bind viahydrophobic interactions to other proteins, to lipids, to carbohydrates and in the case of someenzymes, to substrates and effectors. Thus, both the amount and the arrangement ofhydrophobic groups are important.The ASA of a native protein when compared to the ASA for the corresponding unfolded proteinhas also been related to the hydrophobic free energy (Zielenkiewicz and Saenger, 1992).The ASA of an unfolded polypeptide can be easily calculated by assuming an extended “standardstate” for each type of amino acid (Rose et al., 1985a) and adding up the surface areas of theindividual residues. In one approach (Rose et al., 1985a), the extended standard state for aresidue X is taken to be the surface area of that residue in the extended tripeptide Giy-X-Gly.13More recently the tripeptide Ala-X-Ala has been suggested to be a better model (Zielenlciewiczand Saenger, 1992).The ASA of an unfolded polypeptide (ASA) is linearly related to its molecular weight (Milleret al., 1987):ASA= 1.48*MW+21Thornton et aL (1986) developed an alternative concept to ASA to define those portions of aprotein structure which are accessible and “protrude” into solvent. To evaluate the protrusionof a residue, an equimomental ellipsoid to fit the a-carbon backbone coordinates is ralculated,the absolute size of the ellipsoid is arbitrarily chosen to include a specified percentage of theatoms. A 90% effipsoid will include 90% of the atoms, with 10% lying outside or protruding fromthe globular shape. A protrusion index (H) can be assigned to each residue, specifying the %ellipsoid at which that residue first become external. For example, aliresidues which are outsidethe 90% effipsoid are assigned a P1 of 9. In the case of myoglobin and lysozyme, a highcorrelation was found between high P1 values and the location of antigenic peptides.6. Determination of protein hydrophobicityBased on the above discussion it is evident that the term hydrophobicity has differentmeanings for different researchers. Due to this situation, a large number of methods have beenproposed to measure the hydrophobicity of substances. The same is also true for proteins.The quantitation of protein hydrophobicity can be an essential step for accurate prediction ofprotein functionality (Nakai, 1983). Given the present lack of methodology derived fromtheoretical basis for determining the absolute hydrophobicity of proteins, empirical methodswhich may have practical relevance are currently used. A detailed description of these can befound in the recently published book of Nakal and Li-Chan (1988). Only two approaches:hydrophobicity scales and fluoresence probes, which are those relevant to this thesis, will bediscussed in detail.14a. Hydrophobicity scalesAccording to Rekker and Mannhold (1992) lipophilicity and its analogue hydrophobicitycan be defined as the physical property of the molecule which governs its partitioning into thenonaqueous partner of an immiscible or partially immiscible solvent pair. The partitioncoefficient (P) can siirply be described as (Rekker and Mannhold, 1992):P = cJc or log P = log c0 - log cwhere c0 and ç represent the molar concentration of the partitioned compound in organic andaqueous phase, respectively.Historically, hydrophobicity has been related to P. Proteins can be considered condensationpolymers of amino acid residues, thus if the P of amino acids is known or can be measured, andassuming additivity, the P of a protein can be calculated. Although the P for proteins can beestimated (Nakai and Li-Chan, 1988), P or some related property is usually calculated.Many scales for the estimation of P or some related property of proteins have been publishedin the literature. These scales can be classified in two groups: (1) those derived from solubilitydata, and (2) those calculated empirically from molecular structure, especially X-ray patterns(Nakal and Li-Chan, 1988).Solution scales are based upon comparisons of solubilities in aqueous and nonaqueous solvents;from this data, partition coefficients are calculated and then thermodynamic parameters (e.g.,free energy of transfer) are estimated. Empirical scales measure the partition of amino acidsusing the extent to which each residue type is found buried within the protein interior. Aspointed out by Rose et al. (1985a) the concept of “buriedness” is a property of the three-dimensional structure and, unlike solubiity, may be only nominally defined. A more detaileddiscussion in the relationship between solvent accessible surface area and hydrophobicity isgiven in the section “Protein solvent accessible surface area and hydrophobicity”.In the biochemistry area, the concept of hydrophobicity is usually related to one of the forcesresponsible for the conformation and stability of native proteins (Dill, 1990). Kauzmann (1959)15defined the term “hydrophobic bonding” as follows: “Since the non-polar side chains have a lowaffinity for water, those polypeptide chain configurations in proteins which bring large numberof these groups into contact with each other, and hence tend to remove them from the aqueousphase, will be more stable than other configurations, other things being equal. One can considerthat the side chains of the above-mentioned amino-acids will form intramolecular “micelles”analogous to the micelles known to occur in aqueous soaps and detergents. This tendency of thenon-polar groups ofproteins to adhere to one another in aqueous environments has been referredto as hydrophobic bon’ing [italics by Kauzmannl. The hydrophobic bond is probably one of themost important factors involved in stabilizing the folded conformation ofproteins”. As mentionedbefore (see section Hydrophobicity and thermodynamics) the hydrophobic bond or hydrophobicinteraction, as it is currently called, was explained by Kauzmann (1959) in thermodynamicterms (i.e, enthalpy, entropy and free energy changes when a non-polar molecule is transferredfrom a non-aqueous to an aqueous environment).Later, Tanford’s group (Tanford, 1962; Whitney and Tanford, 1962; Nozaki and Tanford, 1971;Tanford, 1980) quantified the effect of hydrophobic interactions on protein stability. Thisquantification was done in terms of free energy of transfer (G), estimated using solubility dataof a hydrocarbon molecule (or amino acid side chain) from a pure hydrocarbon phase (or anarbritarily selected nonpolar phase) to an aqueous phase. The larger the AG, the moreunfavorable the nonpolar to water phase transfer is and the more hydrophobic the amino acidis. For example, the nonpolar residues Trp and Leu have AG of 3.0 and 2.4 kcal/res, respectively,while the polar residues Lys and Thr have a value of 1.5 and 0.45 kcal/res, respectively. Thesevalues are valid only when the amino acid side chains are transferred from ethanol to water(Bigelow, 1967). If it is assumed, as Tanford (1962) did, that ethanol is a good model for theorganic interior of proteins, the burying of a Trp side chain (i.e., indole) inside a protein willprovide about 3 kcal of stabilization free energy. The choice of using alcohols (e.g., methanol oroctanol) as a model for the protein interior has been criticized. Since alcohols have the potential16ofhydrogen bonding, especially with Trp side chains, this will affect solute solubility by specificsolute-solvent interactions. Wolfenden and Radsika (1986) indicated that Trp related compoundsappear to be extremely hydrophobic if the organic reference phase consists of molecules that canform hydrogen bonds to indole derivatives.Tanford (1962) provided experimental values for the transfer free energies (hydrophobicityindices) for 15 amino acids (Gly, Ala, Val, Leu, lie, Phe, Pro, Met, Tyr, Thr, Ser, Asn, Gin, Aspand Glu, the last two in the neutral form). He computed values for three residues (Trp, Arg andLys). There were no values for two residues (Cys and His) and for charged Asp and Giu. Tanford(1962) calculated the free energy of unfolding due to hydrophobic interaction by multiplying theG of a given type of side chain by the number of residues having that side chain. The total freeenergy of unfolding due to hydrophobic interactions is then calculated by adding the productsfor all protein side chains (Z&G). Tanford (1962) reported the Z8G for f-lactoglobulin, myoglobinand ribonuclease of 192,173 and 100, respectively, which indicates that hydrophobic interactionsare more important for the stability of f-lactoglobuiin than for ribonuclease. Tanford (1962)discussed these values in terms of stability of the native conformation and not as a property ofthe protein, that is, Tanford (1962) did not indicate that native f3-lactoglobulin is morehydrophobic than native ribonuclease, or that native -lactogIobulin has the potential of moreintermolecular interactions than ribonuclease.Bigelow (1967) was the first one to consider the free energy of unfolding due to hydrophobicinteractions as a property ofthe native protein and named ithydrophobicity. In order to compareproteins of different sizes, Bigelow (1967) computed the average hydrophobicity (110) asTanford’s G divided by the number of residues. For 150 proteins, over 50% of them had avalue between 1000 and 1200 kcallres. Bigelow (1967) related HO to solubility in a qualitativefashion, and for a limited set of proteins he suggested that HO is inversely related to aqueoussolubiity.17Zimmerman (1968) published hydrophobicity indices for all the 20 amino acids based onTanford’s (1962) experimental data. After a few years, Nozaki and Tanford (1971) provided arefined set of indices for 11 amino acids. In 1975, Jones published a full hydrophobicity scalewhich was obtained by adjusting the Zimmerman scale according to the data of Nozaki andTanford (1971). This scale has served as one of the best tools for the study of various propertiesof protein molecules, and has stimulated the proposal of over 40 other scales (Ponnuswamy,1993).Wolfenden et al. (1981) determined the free energy of completely stripping water away from theresidue side chains, AGh, by measuring the partition of side chain analogs between aqueoussolution and the dilute vapor phase. These authors have proclaimed that AGh is a bettermeasure of the change that occurs in side-chain environments upon folding, because in thisprocess interior groups are displaced from a hydrated state to the one that is solvent shielded.Interestingly, there is no correlation between the Wolfenden et al. (1981) scale and the scale ofNozaki and Tandford (1971).In 1987, Cornette and co-workers compared 38 hydrophobicity scales for their ability to identifrthe characteristic period of cz-helix. They reported that experimentally determined scales assignPro as hydrophobic, while statistical scales consider it hydrophilic, reflecting its tendency to beexposed in a protein molecule, despite its hydrophobic character. Val is generally morehydrophobic in the statistical scales than in the experimental scales. High variation in theassignment of His and Thr is observed among the scales. For example, the experimental scaleof Jones (1975) considers both His and Thr hydrophilic, His being 0.14 as hydrophilic as Asn,and Thr being equally hydrophilic as Asn. The statistical scale of Rose et aL (1985) considers Hisas hydrophobic (2.6 times less hydrophobic than fle) and Thr 0.2 times as hydrophilic as Asn.The statistical scale of Eisenberg and MacLachan (1986) considers both amino acids to behydrophobic, being approximately 3 times less hydrophobic than lie.18A comparison of several scales was also made by Hopp (1987) for their ability to detect antigenicdeterminants bound by antibodies.Nakai et al. (1988) used cluster analysis to study the relationship between 222 published indicesrepresenting various physicochemical and biochemical properties of amino acid residues,including hydrophobicity. They report that several subclasses of hydrophobicity scales could beidentified: preference of inside and outside, accesible surface area, surrounding hydrophobicityand other mostly experimental scales including transfer free energy, partition coefficients, HPLCparameters and polarity.As noted by Nakai and Li-Chan (1988) there is no standard rule for the selection of amino acidscales. In biochemistry, the three most popular scales are the Jones (1975), Fauchere and Pliska(1983) and Kyte and Doolittle (1982) scales. The first is one of the earliest experimental scales,the second and third are the most popularly used experimeEntal and empiriaI scale(Ponnuswamy, 1993). Other scales commonly used are the Eisenberg consensus scale (Eisenberget al., 1987), the surrounding hydrophobicity scale, recently updated by Ponnuswamy (1993), theMiller’s (1987) statistical scale and two statistical scales of Rose et al. (1985a): average areaburied upon unfolding and mean fractional area loss. In the area offood science, the hydrophobicconcept of average hydrophobicity discussed before, proposed by Bigelow (1967), is mostcommonly used to estimate the hydrophobicity of food proteins.Recently three new scales have appeared in the literature. Urry et al. (1992) reported ahydrophobicity scale based on inverse temperature transitions of amino acid residues as a guestin a repeating peptide sequence which occurs in elastin. This scale is a functional hydrophobicityscale that may be more related with protein folding than other scales. Holibrook et al. (1990)used artificial neural networks to predict the surface exposure of amino acids from proteinsequences and reported a hydrophobic scale which is highly correlated to other scales.Using a multivariate analysis (i.e., principal component), Hellberg et aL (1987) obtained threedescriptors scales, encompassing 29 variables of amino acids. These scales are related to the19three principal properties of amino acids: hydrophobicity, bulk and electronic properties. Thesescales are reported in Table 1.Tanford (1962) was very clear in indicating the limitations of using simplified models (i.e.,hydrophobic scales) to study phenomena related to hydrophobicity. He wrote: It is not likelythat all non-polar parts of the molecule are shielded from the solvent in an actual nativestructure, nor has it been established that the unfolded form of a globular protein is socompletely unfolded as to permit free contact of all parts of the molecule with the soivent.’. Thisstatement points out the main limitation of using hydrophobic scales: the tertiary structure ofthe protein is not taken into account in the computation.b. Fluorescence probe methodsFluorescent probes or fluorescent dye binding is a procedure which, although i does notyield quantitative information on the distribution of individual side chains, provides anindication on the availability ofhydrophobic regions in protein molecules (Kronman and Robbins,1970). In addition, extrinsic fluorescence probes allow selective examination of a particularcomponent or environment of a complex biomolecular assembly (Johnson, 1992).Fluorescence probes can be defined as small molecules which undergo changes in one or moreof their fluorescence properties as a result of noncovalent interaction with a protein or othermacromolecules (Brand and Gohlke, 1972). Molecules with greater dipole moments in the excitedstate than in the ground state will show solvent-dependent emission spectra reflecting themobility of the molecule in the solvent and polar character of the solvent. Such dyes arefrequently referred to as “polarity” probes (Brand and Gohlke, 1972).Fluorescent compounds such as 1-anilino-8-naphthalene sulfonate (ANS) and cis-parinaric acid(CPA) have extremely low quantum yields in water. In the presence of many proteins, thequantum yield of the probe shows dramatic increase with concomitant shift in fluorescenceemission maxima. The dependence of ANS and CPA fluorescence on solvent composition20Table1.Selectedhydrophobicityscales.t.3(1)Bigelow(1967).Ethanoltowatertransfer freeenergy(kcalfres).(2)Jones (1975)asreportedinCornetteetal.(1987).Organicsolvent(ethanolordioxane)towatertransferfreeenergy(kcailres).(3)FauchereandPliska(1983).Octanoltowaterfreeenergyoftransfer(kcal/mol).(4)KyteandDoolittle(1982).Hydropathyindex(kcallmol).(5)Ponnuswamy(1993).Averagesurroundinghydrophobicity(kcaVmol).(6)Roseetal. (1985a).Meanfractional arealoss. (7)Roseetal. (1985a).Averageareaburieduponfolding(A2).(8)Milleretal. (1987).Interiortosurfaeefreeenergyoftransfer (kcallmol).(9)Urryetal. (1992).Inversetemperaturetransition(°C).(10)Hollbrooketal. (1990).Artificialneuralnetworksweights.(11)Eisenberg(1987).Freeenergyoftransfer (100xkcallmol). (12)Hellbergetal.(1987).Hydrophobicprincipal property.Amino(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)acidlie2.953.151.804.53.140.88158.00.74105873-4.44Leu2.402.171.703.81.990.85164.10.6554553-4.19Val1.701.871.224.22.530.86141.00.61244254-2.69Ala0.750.870.311.80.850.7486.60.20456250.07Gly0.000.100.00- that interactions of the probe with hydrophobic regions of the proteins are responsiblefor the wavelength shift and increase in quantum yields (Kronman and Robbins, 1970).Weber and Laurence (1954) first described that while only green fluorescence is emitted whenANS and related molecules are dissolved in aqueous solutions, intense blue fluorescence occurredwhen the chromophores adsorbed to bovine serum albumin. These authors also reported amarked increase in quantum yield of naphthalene dyes in the presence of some heat denaturedproteins and serum albumin. Intense fluoresence was also observed in organic solvents.Subsequently Stryer (1965), who demonstratad that these dyes adsorbed specifically to the hemebinding site(s) on apomyoglobin and apohemoglobin, initiated many investigations related totheir interactions with proteins, membranes, subcellular particles, and cells. Since thefluorescence of these type of compounds dissolved in nonpolar solvents, is similar to thatobserved when they are adsorbed to proteins and membranes, they became known as‘hydrophobic probes”. The chemical structure of ANS and CPA is shown below:T H%_HNH SO H CH CH - 03 2Hcc2)7_c_01ANS CPA22There are three major types of hydrophobic probes (Haugland, 1992):1) Anionic probes which include fluorescent analogs of fatty acids, fatty acyl fluoresceinamides and naphthalenesulfonates (i.e., aromatic sulfonic acids). Naphthalenesulfonatesconsist of a compact nonpolar fluorophore with a heavily hydrated (in aqueous solution)anionic substituent that positions the probe at the lipid or non polar-water interface.These types of anionic probes have the least structural resemblance to intrinsicmembrane components.2) Cationic probes such as carbocyanine and dialkylaminostyryl dyes. These dyes arecommonly used as neuronal tracers.3) Neutral probes, including both nonpolar and zwitterionic (electrostatically neutral)probes that have, or may have specific affinity to membranes. With the exception ofsterol analogs, these probes only vaguely resemble intrinsic membranes.Diphenythexatriene (DPH), Nile red and N-phenyl-1-naphthylamine are the mostpopular neutral probes.Among all these hydrophopic probes, ANS is the most frequently used. Slavik (1982), reviewedits properties and applications to the study of membranes. As mentioned before, its fluorescenceemission depends critically on the polarity of the environment. The emission maximum alsopresents a red shift (up to 50 nm) in water. As mentioned by Nakai and Li-Chan (1988) simpleinterpretation of ANS results in terms of hydrophobicity is not recommended. Penzer (1972)reported that the fluorescence emission of ANS was enhanced and blue shifted in concentratedaqueous solutions of some salts (e.g., MgCl, CaC12NaOH) as when ANS binds to proteins. Hesuggested that molecular rigidity rather than solvent polarity is the dominant factor influencingthe energy and quantum yield of ANS. Haugland (1992) indicates that bis-ANS may be superiorto 1,8-ANS as a probe for nonpolar sites in proteins, since it often binds with a much higheraffinity.Chen (1990) also points out that the interpretation of fluorescence spectra of ANS-proteincomplexes in terms of site polarity, is difficult. This point is illustrated by the work of Weber etal. (1979), who have obtained the 2.8 A resolution X-ray structure of the ANS-chymotrypsincomplex. The dye exhibits a marked fluorescence enhancement and a blue-shifted emissionspectrum. The dye is exposed to solvent on pne side and to polar amino acid side chains on theother, yet apparently reports a “hydrophobic” binding site. According to Chen (1990) the23“polarity” of binding sites is an outmoded abstraction which derived from trying to extrapolatesolution conditions to protein environments. He writes: “The use offluorescence probes has neverreliably shown that the bind site is hydrophobic, in the sense that only apolar amino acids arein the vicinity. Probes such as ANS, when they bind to macromolecules and become fluorescent,always exhibit a blue shift, probably because they are immobilized and protected from solventreorientation. If these ligands were true probes of site polarity, they would sometimes be foundto have a yellow, green or orange fluorescence”. According to Turner and Brand (1968), theemission maximum of adsorbed ANS gives the most reliable estimate of binding site polarity.These authors correlated the emission maximum ofANS to Z, an empirical solvent polarity scale.The use of CPA as a hydrophobic probe was suggested by Sklar et al. (1977). Since the structureof CPA isomers are similar to naturally occurring fatty acids, the perturbing effects of thesemolecules on biological membranes should be minimal and can be bfosynthetically incorporatedinto cellular membranes (Nakai and Li-Chan, 1988). In systems with coexisting gel (solid) andliquid crystalline (fluid) phase lipids, trans-parinaric acid partitions preferentially into the solidphase, whereas the cis acid (CPA) exhibits an almost equal distribution (Haugland, 1992).Parinaric acid’s extensive unsaturation makes it susceptible to oxidation and photodecomposition(Haugland, 1992).The study of the binding of small molecules to proteins has a long tradition in the biochemistryarea and there are several approaches for obtaining and analysing binding data (Weber, 1992).Typically, the parameters obtained from binding experiments are: the number of binding sites(n) and the affinity of each site for the probe. In chemical terms, these two parameters arereferred to as the stoichiometry of the binding and the association or dissociation bindingconstant (K), respectively.From a biochemical point ofview, in order to have a clear understanding of the binding reaction,both n and K are necessary. For example, suppose the binding of a small hydrophobic moleculeto two proteins is to be compared. Protein A has 10 binding sites and protein B has only five.24For both proteins all binding sites are independent of each other (an assumption usually madein simpler binding models). The dissociation constants (K.) are i05 and 10.6 mol/liter forprotein A and B, respectively. The question “which protein is more hydrophobic ?“ could beanswered in two different ways. If only n is considered, the answer will be A since this proteinhas more hydrophobic binding sites (10 versus 5); on the other hand, the answer would beprotein B if only K. is used since the binding affinity of protein B is an order of magnitudelarger than for protein A, indicating that protein B binds the hydrophobic probe more tightly.Comparison of fluorescence intensity values at equal molar concentration could also be used toestimate the “polarity” of the binding sites. Fluorescence intensity values were used by Wickeret al. (1986 and 1989) to study the thermal transitions in tilapia myosin. These authorsindicated that prior to the onset of gelation, a change in the hydrophobic character of myosin(i.e., increase in fluorescence intensity in myosin-ANS systems) occtrred.For the determination of n and K for a given system, several methods have been reported. Influorescence binding studies, where a significant change in a fluorescence property of the probe(i.e., intensity, polarization, etc.) occurs when it binds to the protein, the most common approachis to titrate a fixed quantity of protein with different amounts of the probe. Alternatively,titration of a fixed quantity of probe with protein can also be performed (Weber, 1992).The resulting binding isotherm can be analyzed using one of the various linearized forms of thebinding equation derived for a single noninteracting binding site. The most common linearizedform is the Scatchard equation or plot (Scatchard, 1949). Skiar et al. (1977) reported a Scatchardplot for the binding of CPA to serum albumin. Recently, Laligant et al. (1991) reported that theScatchard equation could not be used, due to turbidity affects at high protein contents, to studythe binding of CPA to -lactoglobulin. The application of the Scatchard equation to bindingphenomena has recently been criticized by Klotz (1989).Weber and Young (1964) showed a convenient way to determine n and K. In this method x=F/Fo,where F is the observed fluorescence intensity and Fo is the intensity observed when all the25probe at a concentration Do, is bound to the protein. Then by measuring the dependence of x asa function of protein concentration (P), at a constant probe concentration (Do), and plotting xagainst P (P expressed in mol/liter), the initial slope in this plot is nJ(Do+K) for x< 0.1. Thus,the slope is directly related to n, the number of binding sites and to the dissociation constant,K.Kato and Nakai (1980) proposed a method for the estimation of effective hydrophobicity ofproteins. This method is essentially a titration of a constant amount of CPA with increasingconcentration of protein. The initial slope (S0) (i.e., slope at 0 concentration of protein) of the plotof fluorescence intensity against protein concentration (expressed in %) was used as an indexof protein hydrophobicity. No explanation is given on the rationale behind the use of the slopeas a measure of hydrophobicity. In this paper, the authors indicated that the “fluorescencemethod of Sklar et al. (1977) was applied to determine the effective hydrophobicity of proteins”(Kate and Nakai, 1980). However, the methodology reported by Skiar et al. (1977) does notexactly correspond to the one used by Kate and Nakai (1980). Kate and Nakal (1980) reportedgood correlations between S0 and interfacial tension and emulsifying activity of model proteins.Hayakawa and Nakai (1985) applied the same titration method for estimating ANShydrophobicity. It is important to point out that the Kate and Nakai (1980) method forestimation of effective hydrophobicity is commonly used in the study offood proteins (Nakai andLi-Chan, 1988). A discussion on these studies is given in the section of’Protein Hydrophobicityand Functionality”.In the biological area, the primary application of DPH has been to estimate “membrane fluidity”based on fluorescence depolarization measurements (Haugland, 1992). Tsutsui et al. (1986)reported a method to estimate triglyceride binding capacity of proteins. In this method DPH isfirst dissolved in a corn oil-heptane solution; after removal of the heptane by evaporation, theresultant DPH in corn oil is then allowed to interact with aqueous protein solution. Followingincubation, the fluorescence intensity is measured and is take as an indication of fat-binding by26the protein molecule and used as an index of relative hydrophobicity (Tsutsui et aL, 1986).Several steps are involved in this method, limiting its wide application.In theory, DPH can be used at different pHs since it is a neutral molecule. This is not the casefor the anionic probes ANS and CPA. At acidic pH, ANS and CPA cannot be used for accuratedetermination ofhydrophobicity since the quantum yields ofundissciated and dissociated formsof the probes are different and charge effects may be difficult to account for. For CPA, poorsolubility of the probe (undissociated form) at pH<5, limits its application at acidic pH.Hayakawa and Nakai (1985) indicated that the binding sites of CPA on protein molecules maydiffer from those of ANS, therefore these authors classi1r the hydrophobicity of proteinsmeasured with CPA into aliphatic hydrophobicity due to aliphatic amino acid residues and thatone measured by ANS as aromatic hydrophobicity due to aromatic amino acid residues. Thisprobe selectivity has not been confirmed experimentally; however, experimental results haveshown that for some proteins, ANS and CPA hydrophobicities do not follow parallel trends(Nakai and Li-Chan, 1988), suggesting that it is highly probable that protein binding sitesand/or binding interactions of CPA and ANS are different.Although fluorescence probe methods have certain limitations, they have become widely adopted,due not only to their simplicity but also to their great detection sensitivity to nonpolarenvironment (Nakai, 1983; Nakai and Li—Chan, 1988). As pointed out by Kato and Nakai (1980)analysis of protein hydrophobicity using CPA (fluorescence slope method) requires 10 mmcompared to a minimum of 2 h for hydrophobic partition and 5 h for hydrophobicchromatography.c. Aliphatic and aromatic hydrophobicity of proteinsIt has been suggested that differences may exist between hydrophobic interactionscaused by aliphatic and aromatic side chains (Hayakawa and Nakai, 1985; Burley and Petsko,271988). Aromatic residues (Tyr, Trp and Phe) are more bulkier than aliphatic ones (Val, Leu, lie).Also they are generally far less common in proteins than aliphatic ones (Rose et al., 1985a).The question, “are aromatic residues more hydrophobic than aliphatic ones?’ is difficult toanswer. &G values for the transfer from alcohol to water suggest that aromatic side chains aremore hydrophobic (Jones, 1975). Oi the other hand, cyclohexane to water AG (Radzicka andWolfenden, 1988) and vacuum to water AG values (Wolfenden et al., 1981) indicate higherhydrophobicity for aliphatic residues. Miller et al. (1987) reports that the amino acidcomposition of the interior (residues with less than 5% accesibility) of 37 monomeric proteins ismade up of over 30% of aliphatic and 14% aromatic residues, while the surface (residues withmore than 5% accesibility) is 12% aliphatic and 9% aromatic. Using these partition of residuesbetween the protein surface and interior these authors calculated transfer AG for amino acidresidues. These AG values suggest that the hydrophobicity of Tie, Lei, VaT and Phe (0.74, 0.65,0.61, 0.67 kcallmol, respectively) are higher than for Trp (0.45 kca]Imol) and ‘i’yr (-0.22 kcal/mol).This latter amino acid was classified as hydrophilic in this scale.According to the distribution of accessibility to solvent of amino acids residues reported by Roseet al. (1985a), aliphatic residues tend to be more buried than aromatic. Interestingly, amino acidresidues branched at C3 (Val, Tie) and the aromatic residues favour sheet conformation(Hamaguchi, 1992).Statistical analysis performed by Ikai (1980) showed that the relative volume of a proteinoccupied by aliphatic side chains (Ala, Tie, Tie and VaT) of proteins of thermophilic bacteria issignificantly higher than that of ordinary proteins. This author reports a good correlationbetween the “aliphatic index” and thermostability of proteins. The aliphatic index (Al) of aprotein was calculated according to the following formula:AT = X + aX + b(Xfle + X)where X, X, Xe and X are mole percent (100 x mol fraction) of Ala, VaT, Ile, and Leu. Ascoefficients a and b were taken as 2.9 and 3.9, the relative volumes of aliphatic side chains to28that of alanine side chains were taken. High values of Al were reported for a-lactalbumin and-lactoglobulin. Ikai (1980) speculated that since the effect of temperature on solubility ofaliphatic amino acids is somewhat different from that on aromatic residues, suggesting thataliphatic hydrophobicity increases more rapidly with increasing temperature, the correlationbetween thermostability of globular proteins and A! is reasonable. Probably, aromatic residuesare found more exposed than aliphatic ones due to their bulkiness which hinders effective burialin the protein interior. These surface-exposed aromatic residues may be important in terms ofintermolecular hydrophobic interactions, thereby affecting solubility. On the other hand, theburied hydrophobic residues (i.e., mainly aliphatic) may stabilize protein structure throughintramolecular interactions and thus affect functionality which depends on flexibility or rigidityof the protein molecule as well as the exposure of hydrophobic residues upon processing such asheating or whipping. Schein (1989) indicates that the aqueous affinity ofTrp imidazole ring andthe hydroxyl group of Tyr were understimated by early hydrophobicity measurements.It is important to point out that while the state of aromatic residues in proteins can be studiedby several spectroscopic methods (i.e., UV absorption, and fluorescence, Raman spectroscopy,nuclear magnetic resonance, etc.), fewer techniques are available for the study of aliphaticresidues. The CH stretching region in the vibrational spectra of organic compounds is oftencomplicated by superposition ofdifferent types ofbands. Besides the CH stretching fundamentalbands, overtone and combination bands appear and strong Fermi resonance may occur betweenthem (Matrai et al., 1990)29B. SPECTROSCOPIC METHODS FOR THE STUDY OF PROTEIN HYDROPHOBICITY1. Derivative ultraviolet spectrophotometrvVisible and near UV spectra originate from relatively low-energy electronic transitions.Three main groups of chemical species are involved: compounds containing metals (especiallytransition metals), such as prosthetic groups in metalloproteins, large aromatic ring structuresand conjugated double-bond systems. Electronic transitions are usually very broad because theyinciude a large number of closely spaced bands each corresponding to a vibrational energychange (Franks, 1988). For wavelengths above 230 nm, the absorbing components in proteinsare essentially the aromatic amino acids Tyr, Trp and Phe. Trp has the largest molar extinctioncoefficient (5500 l/cm/mol at 278 nm), roughly four times larger than Tyr (1340 at 275 nm) and27 times larger than that of Phe (190 J/cm/mol at 257 nm). Since the molar absorbance ofphenylalanine is relatively low as compared to Tyr and Trp, and most proteins contain asignificantly larger number of Tyr and Trp residues relative to Phe, UV spectroscopy cannotusually give adequate information about Phe residues in proteins (Hamaguchi, 1992). It isimportant to point out that the spectrophotometric analysis of the environment of both Trp andTyr residues in proteins containing both amino acids is frequently limited to qualitative aspects,because of the overlapping ofTrp and Tyr bands. According to Dufiach et al. (1983) Trp residuesare less frequently used as internal probes than Tyr ones, since the Trp bands have a lowerresponse to changes in environment.Incorporation of the aromatic amino acids into a protein generally produces a red shift (i.e., longwavelength shift) in relation to the free amino acids. Thus, it has been customary to concludethat blue shifts in the UV spectra of proteins arise from the exposure to the polar solvent ofchromophores that had been previously buried in nonpolar regions of the molecule. Although inmany instances this conclusion is valid, a blue shift is not conclusive evidence for enhancedexposure since this type of spectra shift can have other origins (e.g., changes in the30polarizability, hydrogen bonding state, etc.) (Kronman and Robbins, 1970). The UV absorptionspectrum of a protein is changed by additives such as organic solvents, salts and by variationin pH and temperature. However, since these changes are generally very small, traditionally twodifferent techniques have been used to detect them: difference spectra and perturbation spectra.More detail in these techniques can be found in Kronman and Robbins (1970).The more recent technique of derivative spectroscopy is now recognized as a tool for analysis ofconformational changes and interactions in proteins (Padros et al., 1984).Recently, fourth derivative spectroscopy has been used to study the characteristics of aromaticamino acids in proteins (Yesilada et al., 1992; Mozo-Villaria et al., 1991; Padros et al., 1984;Padros et al., 1982). Fourth derivative spectroscopy provides better resolution than the first orsecond derivatives and as for all even-numbered derivatives, the maxima of the derivatives areat the same wavelength as the original absorption spectrum (Padros et al., 1984). Changes inthe environment of the aromatic residues (Tyr, Trp and Phe) produce changes in the fourthderivative spectra, reflecting the alteration in electronic energy levels. For proteins where theratio of Tyr to Trp is equal or smaller than 4, the bands due to Tyr and Trp are well resolved(Dufiach et al., 1983). For Tyr residues, change in the strength and type of hydrogen bonds, inaddition to “polarity” effects, also produce shifts in the position of the bands. In addition to peakpositions, some geometric parameters of the peaks have also been suggested for quantifrng inmore detail changes in the spectra (Padros et aL, 1982; Dufiach et al., 1983). Model studiesshowed that contrary to Tyr and Trp bands, Phe bands presented very little solvent dependencyon its fourth derivative spectra (Padros et aL, 1982). In addition, the small absorbance coefficientof Phe makes it difficult to observe its spectra. An alternative procedure to fourth-derivativespectroscopy for the study of aromatic residues in protein proposed by Metzer et al. (1972) isbased on the fitting of log-normal distribution curves to the experimental absorption curve.312. Intrinsic fluorescenceIn 1957, Teale and Weber reported the first thorough investigation of the fluorescenceexcitation and emission spectra of the aromatic amino acids and since then fluoresencespectroscopy has been widely used to study conformational properties ofproteins. The usefulnessof this method is due to the rich variety of molecular details that it can reveal about proteins,including the solvent exposure of amino acid side chains, the existence of protein conformers,the rate of rotational diffusion of a protein, and the distance between sites on a protein (Eftink,1991). Fluorescence spectra properties such as excitation and emission spectra contours andmaximum positions, quantum yield, fluorescence decay time (lifetime) and polarizationproperties are very responsive to changes in the environment of a fluorophore. Thisresponsiveness is much greater than that observed in related types of spectral measurements,such as UV-absorbtion spectroscopy (Eftink, 1991). The key to this enrironmental responsivenessis that fluorescence spectroscopy is in essence a kinetic measurement. The characteristics of afluoresence signal depends on the competition between two processes: (1) the rate of theradiative transition of the excited state back to the ground state, and (2) the rate of a varietyof other excited-state reactions that may lead to quenching (i.e., nonradiative return to groundstate), entry into the triplet manifold, resonance transfer of the excitation energy to a secondchromophore, rotation of the fluorophore and relaxation of solvent molecules or other polargroups around the excited dipole (Eftink, 1991).As for UV absorption, the fluorescence spectra (i.e., emission spectra) of proteins is the sum ofthe contribution from the three aromatic amino acids Trp, Tyr and Phe. In most cases, Trp isthe most valuable intrinsic probe of proteins since the fluorescence spectrum of ‘l’yr is veryinsensitive to solvent effects and the one of Phe is very weak. When all three are present in aprotein (i.e., class B protein), pure emission from Trp can be obtained by excitation atwavelengths above 295 nm. Below 290 and 270 nm there is an onset of absorption into Tyr andPhe, respectively, but resonance energy transfer can occur between these residues and Trp, so32that emission from the latter usually is dominant for class B proteins. For proteins that haveonly Phe and Tyr, for example insulin, zein and ribonuclease (class A proteins), emission fromTyr is usually observed at around 304 nm, corresponding to the value observed with the freeamino acid (Hamaguchi, 1992; Kronman and Robbins, 1970).In order to obtain more information on protein conformation with fluorescence spectroscopy,recording of intrinsic emission spectra at different excitation wavelengths is useful (Myers,1988). This technique can be called two-dimensional fluorescence spectroscopy (Pau et al., 1988).Assuming no significant Phe contribution, only Tyr excites at 270 nm, and if its emission isefficiently relayed to Trp, then only the Trp emission is apparent. The width (or the half bandwidth) of the peak is a measure of the efficiency of the energy transfer; the smaller the widththe higher the energy transfer. At 297 nm excitation, only Trp residues are excited. If the Trpis in a nonpolar environment its emission wavelength will be as low as 325 nm. However, if theTrp is exposed to the solvent its emission will be at about 350 nm. The position of the emissionband in the native protein, when compared to other proteins, may give an indication of theexposure of Trp+Tyr residues to the solvent. However, this position cannot be taken as a directmeasurement of the hydrophobicity ofthe chromophore(s) environment, since other mechanisms(i.e., thermal motions) can also cause changes (Eftink, 1991).If a conformational change causes Tyr residues to move away from Trp, then the energy relaymay be broken, and at 280 or 290 nm excitation, both emission bands may be evident (Myers,1988). The fluorescence intensity can also be used to detect conformational changes; exposureofa fluorophore to the solvent usually causes a decrease in fluorescence intensity due to solvent-quenching (Kronman and Robbins, 1970). Two-dimensional fluorescence spectroscopy has beenapplied with success for the identification of bacteria (Pau et al., 1988).333. Proton nuclear magnetic resonance in protein chemistrya. IntroductionNuclear magnetic resonance (NMR) spectroscopy is one of the few techniques availablewith the capacity to obtain detailed information about biomolecular phenomena. With NMR, anindividual nucleus in a molecule can be “observed” by monitoring that nucleus line (James,1975). Until recently the direct examination of the resonances of the H atoms (protons) ofproteins was severely limited by insufficient sensitivity and resolution, but developments inNMR technology have alleviated, though certainly not eliminated, these problems (Jardetzkyand Roberts, 1981; Yamane, 1971).The increased availability ofinexpensive minicomputers coupled with the advent ofcommerciallyavailable high field super-conducting magnets has added an impotant series of experimentalNMR techniques that can yield precise information about the structural and dynamicscharacteristics of macromolecules. Techniques such as two-dimensional NI’IR spectroseopy,correlated spectroscopy (COSY) and especially nuclear Overhauser effect (NOE) have made itpossible to obtain detailed information (i.e., three dimensional structure) of oligonucleotides andsmall proteins (Cheatman, 1989; Jardetzky and Roberts, 1981). To give some idea of thewidespread use of NMR in chemistry and biochemistry, a computer search for the period1985-1987 turned up over 500 books and review articles and over 7000 citations, not includingpapers in which NMR was used for routine identification of organic compounds (Rabenstein andGuo, 1988). This discussion will focus on the use of one dimensional proton NMR (‘H NMR) forthe study of protein conformation with a special focus in its use to estimate solvent exposure ofamino acid residues in proteins.34b. Basic principlesNMR spectroscopy, like many other forms of spectroscopy, depends upon the absorption ofenergy by the molecule (i.e., its nuclei) being studied. NMR is based on the absorption of radiofrequency electromagnetic radiation (1& Hz) by certain types of atomic nuclei placed in astrong external magnetic field (Cheatam, 1989). It has been known since the 1920’s that manyatomic nuclei have an angular momentum arising from their inherent property of rotation orspin. Since nuclei are electrically charged, the spin corresponds to a current flowing about thespin axis, which in turn generates a small magnetic field (Pykett, 182). In general the magneticdipoles of the nuclei with spin will be pointing in random directions; however, in the presenceof an external magnetic field, nuclei with nonzero spin orient themselves with the magnetic fieldlines. Spinning nuclei behave rather like tiny tops or gyroscopes. If the axis of a spinninggyroscope is tipped away from the vertical, the gyroscope will rotate about its former axis in amotion describing the wall of a cone; this motion is called precession (Pykett, 1982). Similarlyif the bulk magnetization vector (M), which is defined as the vector that represents the net effectof all the magnetic moments of the nuclei of a given nuclear species, is tipped away from thevertical, M will precess about the vertical (by convention the z axis). Such a tipping can beachieved by applying electromagnetic radiation in the radio-frequency range. The frequency ofthe applied electromagnetic radiation must match the natural precessional frequency of theparticular nuclei (Pykett, 1982).c. Nuclear magnetic resonance of proteinsi. IntroductionThe first NMR study of a protein was the publication in 1957 of a low-field (40 MHz ‘H NMR)of bovine pancreatic ribonuclease A, which revealed four distinct regions of overlapping peaks.These features were interpreted in terms of the spectra of the constituent amino acids of the35protein molecule (Markley and Ulrich, 1984). The NMR spectrum of a protein inherentlycontains much structural information. Even a small protein will have on the order of a thousandprotons, with individual resonance lines characteristic of the chemical environment of each typeof proton (Roberts and Jardetzky, 1970).The detailed interpretation of a protein 1H NMR spectrum is based on the analysis of theintensity, position and width of resonance absorbance lines and is based on the followinggenerally verified premises (Roberts and Jardetzky, 1970):1. The intensity of a resonaxce absorption line is directly proportional to the number ofnuclei in a given chemical environment and is independent of any other variables.2. The position of a resonance absorption line (chemical shift) is determined by: (a) thedensity and geometric configuration of the electrons absorbing nucleus and (b)permanent local magnetic fields such as those originating from other nuclei andunpaired electrons in close proximity (1-10 A) to the absorbing nucleus. The position ofNMR peaks is known as the chemical shift and is very sensitive to the envirpnment ofthe protons; the chemical shift can be used to assign chemical groups (i.e., methylene,aromatic, etc.) to specific peaks.3. The width of resonance absorption lines is determined by the rate of atomic motions.Absorption lines in macromolecules have a half width (Av½) of 10 to 100 Hz. In addition,the band shape is related to the types of interactions involved and flexibility, diffusionrates and indirectly to exposure of the residue.Thus, it is possible to assign a given absorption line to a chemical group (e.g., a specific sidechain or functional group of an amino acid) on the basis of line intensities and position, to drawinferences about its mobility in space (i.e., exposure) from line widths, and to study its electronicconfiguration using line position and multiplicity (Roberts and Jardetzky, 1970). The informationcontent of the NMR spectrum of a small enzyme, for example Staphylococcal nuclease, may beset at 3 x 855=2565, i.e., 3 parameters (line intensity, position and width) for each of the 855proton resonance lines, whereas the information content of its ultraviolet spectrum is at best 3X 11= 33 (the same 3 parameters x 11 aromatic residues). In general a difference of two or moreorders ofmagnitude is found between the information content ofhigh resolution NMR and otherforms of spectroscopy (Roberts and Jardetzky, 1970).36ii. Aromatic and aliphatic resonancesInformation on the NMR spectra of amino acids is essential for the interpretation ofprotein NMR spectra. A compilation of the chemical shifts, intensities and ov½ of the mostcommon amino acids is given in Table 2. Analysis of this table shows that the chemical shiftsof aliphatic protons are in the upfield area of 0.8 to 3 ppm. This area is useful to analysealiphatic hydrophobic amino acids (Ala, lie, Leu, Met and Val). In contrast the chemical shiftsof the aromatic protons (i.e., those from the aromatic hydrophobic amino acids Tyr, Trp and Phe)are ia the downfield area of 6.5 to 8 ppm. Based on the above observation the NMR spectrumof a protein is usually divided into an aliphatic (0.5-3 ppm) and an aromatic region (6-9 ppm).In simple molecules the estimation of lipophilicity or hydrophobicity is rather straightforwardas shown by Ben-et and Tatarskyt (1972). These authors estimated the hydrophilic-lipophilicbalance (HLB) of several nonionic surfactants (i.e., polyethoxylated emulsifiers). The HLB valueis defined as one fifth of the ratio between the weight of the hydrophilic part to the lipophilicpart of the molecule. According to the general chemical structure of this compounds, Ben-et andTatarsky (1972) considered the peaks around 4 ppm to be from hydrophilic and the other peaksto be hydrophobic. The integrated NMR area of these two groups of peaks was used to estimateHLB values. Due to the complexity and tertiary structure of proteins this approach cannot beapplied to estimate effective or surface hydrophobicity of proteins.iii. Assignments of NMR resonancesAssignments of resonances occurs on two levels. First the resonance type (e.g., histidineC2 proton) must be determined. At the second level it is desirable to identil& a resonance witha particular residue in protein (e.g., histidine 119).37Table 2. Proton magnetic resonance positions of amino acid residues for computing the spectraof random-coil proteinsA.Equivalent Resonance position WProton type Protons/Residue (ppm)B (Hz)Leucine (JH3 3 0.899 7.5CH3 3 0.943 7.5CH4JH 3 1.649 20.0Isoleucine 6CH3 3 0.885 10.0JH3 3 0.943 10.0JH2 1 1.190 30.0‘yCH2 1 1.478 30.0fCH 1 1.894 25.0Valine ‘yCH3 3 0.969 8.53 0.942 8.5CH 1 2.130 25.0Ala.nine 1CH3 3 1.395 18.0Threonine CH3 3 1.232 16.0Lysine 4JH2 2 1.47 1 30.08CH2 2 1.708 15.0CH2 2 1.808 15.0sCH2 2 ‘ 3.023 22.0Arginine CH2 2 1.719 28.0f3CH2 2 1.840 24.06CH2 2 3.020 22.0Proline ‘CH2 2 2.030 21.0POll2 2 2.110 25.06CH2 2 3.653 30.0Glutamic acid fCH2 2 1.980 20.02 2.300 20.0Gluta.mine CII2 2 2.070 20.0‘yCH2 2 2.379 20.0Aspartic acid CH2 2 2.837 55.0Asparagine CH2 1 2.755 30.0130112 1 2.831 30.038Table 2. (Continuation).Equivalent Resonance position WProton type ProtonsfResidue (ppm)B (Hz)Methionine eCH3 3 2.128 10.03CH2 2 2.060 22.0CH2 2 2.633 16.0Cysteine fCH2 1 3.278 8.01 2.958 8.0Histidine fCH2 2 3.180 18.0Imidazole CH-4 1 7.140 10.0Imidazole CH-2 1 8.120 10.0Tyrosine f3CH2 2 2.980 30.0C3,5 H 2 6.857. 17.0C2,6 H 2 7.149 17.0Phenylsbnine 13CH2 1 2.991 30.013CH2 1 3.223 30.0Ring protons 5 7.339 30.0Tryptophan f3CH2 2 3.258 27.0Indole C2 H 1 7.244 10.0Indole C5,6 H 1,1 7.167,7.244 15,15Indole C4,7 H 1,1 7.649,7.504 18,18ASource: McDonald and Philips (1969) and Bundi and Wüthrich (1979)Bat 220 MHz from internal 2,2,dimethyl-2-silapentane-5-sulfonate (DSS)c1 width at half height of the resonance39iv. NMR of denatured and native proteinsThe spectrum of a denatured peptide or protein can usually be approximated to a highdegree by the sum of the spectra of the constituent amino acids (Jardetzky and Roberts, 1981).However, in native proteins the positions of the resonance lines are partially determined bytertiary conformation since they are shifted considerably when protein configuration is changedfrom extended random coil where the residue side chains are in an aqueous environment to thenative conformation having many residues in an internal protein environment (McDonald andPhilips, 1969). McDonald and Philips (1969) developed a widely used procedure for computingsome of the characteristics of the NMR spectrum at 220 MHz of any protein in an extendedrandom-coil configuration in neutral aqueous solution. An essential assumption of this methodis the fact that the NMR spectrum of a random-coil protein is not affected by its amino acidsequence (i.e., primary structure), but only by the amino acid compoition of the molectle. Whilethe NMR spectrum of a random-coil protein provides no direct information on the nature of thefolded native protein, it does provide a necessary reference for measurements of spectral changesthat occur upon denaturation and renaturation.In native proteins, the proton resonance that can be most easily resolved in the NMR spectrumin D20 is the histidine C2 resonance, since its line position (7.91 ppm) is relatively far awayfrom other resonances. For this reason histidine resonances have received a great attention(James, 1975).A survey of proteins studied by NMR in 1984 indicated that detailed NMR studies had beenreported for more than 50 different proteins (Markley and Ulrich, 1984); by now this numbershould be over 100.The following discussion is organized according to studies on some selected proteins, in order todemonstrate the type of information that can be obtained from NMR experiments.40(a). RibonucleaseBovine pancreatic ribonuclease (m.w. 13,700) is one of the most thoroughly characterizedproteins. As mentioned before the NMR spectrum of this enzyme was the first one to bepublished. Ribonuclease thermal and acid denaturation have been investigated by proton NMRat 220 MHz (McDonald and Phillips, 1972). The ribonulease NMR spectra at 72.5°C presentssharp extreme high field resonances, which can be attributed to the highly shielded methylprotons of Val, Leu and fle. This protein contains four histidine, three phenylalanine and sixtyrosine residues; tryptophan is absent. Resonance lines appearing in the 6.8-7.8 ppm regionsof the spectrum of the denatured form of the protein can be assigned to the aromatic protons ofthese component amino acids. In the native form of the protein, side chain mobility seems to begreatly constrained. Upon heating the NMR spectrum showed improvement in resolution, a factthat suggests an increase in the mobility (i.e., exposure) of certain irotein residues. In the caseof this enzyme, it is known from X-ray diffraction studies that two of the four histidine residuesare located in the active site cleft. All four resonances are resolved at 100 MHz and subsequentstudies have provided exact assignments for those resonances (James, 1975). The technique ofNMR difference spectroscopy has also been used for ribonuclease. The difference spectrum isgenerated by subtracting the unperturbed spectrum from the perturbed spectrum. In principle,only those resonances which have been either broadened, narrowed, or shifted will show up inthe difference.(b). CaseinsThe caseins are examples ofhighly aggregated systems with sufficient internal flexibilityto enable high resolution NMR signal to be obtained. A detailed NMR study of a5, B and iccaseins was made by Leslie et al. (1969). The NMR spectrum of B-casein, suggested a low degreeof structural organization in the molecule, even after aggregation. The NMR spectrum of iccasein was found to be different from the one of B-casein in that signals from aromatic and41terminal methyl protons were only well resolved after heating or treatment with urea. Thespectrum was consistent with a structure of ic-casein in which the aromatic and most of the nonpolar residues are restricted in motion, but the hydrophilic groups had considerable molecularmotion (Leslie et al., 1969). ‘H-NMR spectra of bovine ic-casein and of bovine casein micelleswere recently studied by P.ollema et al. (1988). The observed spectra consisted of a superpositionof spectra with relatively small linewidths and strongly broadened spectra. The former spectraindicated the existence of protein regions with substantial conformational mobility. For both thealiphatic and aromatic regions good resolution was obt&ned.In order to study the contribution of nonspecific electrostatic interactions between ic- and ccasein, Kason et al. (1974) used NMR, together with other techniques, to study the interactionof the synthetic polymer polyethylenemine witha1-casein. When the NMR pattern ofa,1-caseinwas calculated by the method of McDonald and Phillips (1969) a general agreement in the shapeof the calculated and the experimental pattern was found. Decreased resolution and spectrumamplitude in the NMR spectrum of;1-casein occurred by mixing the protein with the polymer,suesting complex formation. Also a downfield shift in the methylene protons of the polymerwas found on binding with the casein; from 3.1 ppm in the polymer to 3.4 ppm in the polymer-protein mixture. The authors suggested that electrostatic interaction between the protein andthe polymer could be the cause of the observed low-field shift.(c). LysozymeBoth hen egg white (HEW) and human lysozyme (m.w. 14,600) have been investigatedby NMR. HEW lysozyme undergoes reversible thermal denaturation from 68 to 74°C and at 79°Cis believed to be in an extended random-coil configuration except for the four disulfide bridgeswhich remain intact. The simulated spectrum of lysozyme based on its amino acid compositionof the enzyme was found to be very similar to the experimental spectrum obtained at 79°C(McDonald and Philips, 1969). When a protein is unfolded, the broad bands of the native42spectrum give rise to sharper and better resolved peaks. Bradbury and Norton (1974) reportedthat even at 80°C in D20 (pD 2.8) lysozyme retained some noncovalent interactions, since furthertreatment with GdnHCI or mercaptoethanol caused sharpening of the aromatic region of thespectrum. Reversible thermal, acid, and chemical (GdnHC1) denaturation of HEW was studiedusing NMR by McDonald et al. (1967 and 1969). Typical changes in the NMR spectrum upondenaturation were found, such as increased resolution and high similarity to the calculatedspectrum. In a similar type of experiment Bradbury et al. (1972) reported that, according toNMR data, lysozyme in 3 M urea was completely unfolded below pH 3 and virtually native atpH>3.5. The complete assignment of all the aromatic proton resonances in the NMR spectrumof lysozyme has been made (Bundi and Wüthrich, 1979). Recently, Van Dael et al. (1993) usedNMR to study the unfolding of equine milk lysozyme.v. NMR linewidths and protein conformationThe 1H NMR spectra of native proteins are usually heavily overlapped. A protein with200 amino acid residues (molecular weight of20,000 dalton) may contain 4000 protons whichall resonate in the narrow region of about 10 ppm (300 Hz at 300 MHz). The higher themolecular weight of the protein, the larger the number of nonequivalent protons in the spectrumand the broader the linewidth of each of their resonance lines. In the NMR spectra of manyproteins sharp peaks are often encountered together with a broad envelope. Basic principles ofNMR indicate that the faster the molecular motion of a group the sharper its resonance line.Thus, linewidth provides a good measure of molecular motions (Akasaka and Kainosho, 1985).For the Streptomyces subtiisin enzyme inhibitor (SSI) the sharp resonances observed in the 1-2ppm chemical shift range correspond to side chain methyl protons in the exposed segments ofthe protein (Akasaka and Kainosho, 1985).Due to conformational rigidity, proteins with a high proportion of ordered secondary structuretend to produce broader spectra than less organized proteins. Since the molecular freedom of the43amino acid side chains increases upon unfolding, unfolding causes a significant increase in theoverall resolution of the NMR spectrum. The unfolded polypeptide chain gives a NMR spectrumvery similar to the one calculated based on the amino acid composition of the protein.The rate of tumbling of a protein molecule is roughly inversely proportional to the molecularweight of the protein. Thus, if no rapid internal motions such as group rotations and segmentalmotions, are present and if the rotation of the whole molecule may be considered effectivelyisotropic, the minimum linewidth increases proportionally with the molecular weight (Akasakaand Kainosho, 1985). For proteins with a molecular weight of io and iC)5 the expectedminimum linewidths for a methylene group are 10 and 100 Hz, respectively. These linewidthsare equivalent to 0.03 and 0.33 ppm, respectively, in a 300 MHz instrument. To a firstapproximation the line sharpening could be related to exposure of amino acid side chains.However, as in native proteins, resonance positions are usually shifted from their “standard”position (i.e, chemical shift in a model tetrapeptide) due to the formation of secondary or tertiarystructure; this chemical shift dispersion may itselfcause an increase in linewidths. For example,both protons of the p-carbon (CH2)of Ser in the “standard state” resonate at 3.885 ppm (Bundiand Wüthrich, 1979), but in the case of Ser-36, Ser-72, Ser-85, Ser-86 and Ser-100 of lysozymeat 35 °C and pH 3.8, these protons resonate at 3.59 and 4.51, 3.75 and 4.24, 3.89 and 4.17, 3.89and 4.17, and 4.11 and 4.20 ppm, respectively (Redfield and Dobson, 1988). These assignmentswere obtained using detailed analysis of the two dimensional (2-D) NMR spectra of the protein.Interestingly, the n-carbons of Ser-85 and Ser-86 which are significantly more exposed to thesolvent than the n-carbons of the other Ser (Shrake and Rupley, 1973) are also the ones thatpresent the smallest secondary shifts (defined as the difference between the resonance positionsof a protein proton to the resonance position of the same protons in the “standard state”) intheir n-carbons. This fact suggests that side chain secondary shifts may also be related to theexposure of the amino acid side chain.44Several factors of the system under study must be considered in order to correctly interpretNMR linewidths. An example of erroneous interpretation of NMR linewidths is the paper byFossel et al. (1986). These researchers reported the rapid determination of malignant tumorsby ‘H-NMR spectroseopy ofplasma (water-suppressed NMR spectra). These authors showed thatNMR methyl and methylene linewidths from plasma of patients with malignancy appear to belower than linewidths from controls. The water-suppressed proton NMR spectrum of plasma isessentially that of the lipid of all plasma lipoproteins and the few low-molecular-weightmolecules present in appreciable concentrations. The proton signals of proteins are highlyheterogenous and abundant, resulting in an extremely broad, poorly defined contribution to thespectrum. Superimposed on this background are other, better defined resonances in plasma,which arise mainly from aliphatic lipid protons of lipoproteins (Fossel et al., 1986). Theseauthors report that the average lipoprotein-lipid linewidth of methyl and methylene beaks for44 normal subjects was 39.5±1.6 Hz, while the mean linewidth for 81 patients with untreatedcancer was 29.9±2.5 Hz. They suggested that the lack of supramolecular ordering wasresponsible for the narrowing of methyl and methylene resonances in the plasma of patientswith malignant tumors. Problems arose when the test was subjected to large generalpopulations. The NMR linewidths were found veiy sensitive to the distribution of plasmalipoproteins and tryglycerides. Changes in these levels may result for a variety of reasons: diet,trauma, and disease other than cancer. Many clinical assessments of this tests have beenreported, and most of them have shown that this test is not useful for screening anasymptomatic population for cancer (Smith and Chmurny, 1990). For example, Verdery et al.(1989) reported that methyl and methylene linewidths were strongly positively correlated withhigh density lipoprotein levels and inversely correlated with triacyiglycerol-rich lipoproteinslevels. They also showed that after a meal containing fat, the methyl, methylene, and averagelinewidths decreased progressively, reaching a nadir simultaneously with the occurrence of thepeak of triacyiglycerol level, and then increased to baseline.45Recently, multivariate analysis was applied to discriminate the serum from rabbits implantedwith a carcinoma from that of a control group (Kruse et al., 1991). Digitally defined parts (88data points) of the spectral profiles of the lipoprotein methylene and methyl signal were used:0.78-0.89 ppm and 1.15-1.28 ppm, respectively. At a high probability level samples fromimplanted rabbits were classified as malignant as soon as the tumors developed.More sophisticated NMR techniques have also been suggested to estimate the mobility and/orexposure of amino acid side chains in proteins. Determination of the hydrogen-deuteriumexchange rates of amide NHs, number and intensities of the nuclear Overhauser enhancements(NOE), and secondary shifts, coupling constants and intensities of the main-chain resonancesare related to mobility and/or exposure (Wüthrich, 1986; Saudek et aL, 1989).In addition to the high degree of technical sophistication and expertise involved in thesemethodologies, their present applicability is limited to small proteins (<150 amino acids).vi. Spin diffusion and protein mobilityFrom a quantum mechanic point of view a NMR signal is the result of the absorptionof energy to change the state of an assembly of nuclei from a lower energy to a higher energystate. This transition is effected only when the energy of the quanta carried by the radio-frequency field equals the difference in magnetic energy between the two energy states (Pykett,1982). After the excitation pulse ends, the nuclei will tend to return to the equilibrium, i.e., lowenergy state). The velocity or rate that a nucleus returns to equilibrium depends essentially onthe mobility and electronic environment ofthe proton. The return to equilibrium is characterizedby two principal “relaxation” mechanisms characterized by a time constant T: the spin-latticerelaxation (T1) also called longitudinal relaxation, and the spin-spin relaxation (T2) also calledtransverse relaxation. The width of spectral lines is proportional to l/T2. In deuterated solventsin the absence of paramagnetic species, the spin-lattice relaxation is governed by the mutualdipole-dipole interactions of the protons and should in principle provide valuable information on46interproton distances and intramolecular motions. However, in large molecules at high frequencythe spin-lattice of an individual proton is no longer the result of the combined spin-latticecontributions from the fluctuating dipoles coupling to its neighbour, but is largely influenced bythe exchange of spin magnetization (i.e., energy) between many protons. This effect is calledcross-relaxation or spin diffusion (Kalk and Berendsen, 1976; Akasaka, 1983).The extent of spin diffusion depends on interproton distances and molecular motions of themacromolecule. Protons which are relatively inmobilized will present a higher degree of spindiffusion (Akasaka, 1983). Thus by measuring the extent of spin diffusion it may be possible toseparate mobile from rigid protons.Akasaka (1983) used spin diffusion to study the dynamics of Streptomyces subtili$in inhibitor.For this purpose, a cross-saturation pulse is applied at a given frequency before the actual NMRpulse, the rigid protons become saturated and their NMR signals re decreased significantly.Since the signals of the more mobile protons are hardly affected, the signals of exposed and/ormobile segments become more resolved. A similar method was also applied by Akasaka (1979)to study some conformational aspects of myosin.4. Raman spectroscopy in protein chemistiya. IntroductionRaman spectroscopy is named after the Indian scientist C.V. Raman, who, with KS.Krishnan, observed that some of the light scattered by a liquid is changed in wavelength. TheRaman spectrum is a vibrational spectrum of a molecule (Carey, 1982). In the past decadeRaman and resonance Raman spectroscopies have been applied to such an extent to biochemicalproblems that a multidisciplinary field of research called “bio-Raman” has been developed(Carey, 1988). Developments in optics, lasers and electronics continue to promote advances in47experimental Raman spectroscopy and these advances have important ramifications forbiochemical applications (Carey, 1982; Carey, 1988).Raman spectroscopy provides an excellent tool for elucidating the structural features ofproteins.Relationships between spectra and structure (environment as well) have been accumulated andare used in the spectral analysis. Raman spectra of colorless proteins with visible laserexcitation consist of an overlap of bands arising from vibrations of the main chain and side-chains of amino acid residues. Some of the Raman bands of the main chain and side chains,including amide, disulfide, methionine, aromatic amino acid, histidine, and ca”boxyl groups, havesufficient intensities to be generally useful (Harada and Takeuchi, 1986).Raman spectroscopy is widely used for quantifying the secondary structure of proteins and tostudy the strength of hydrogen oonds to tyrosine -OH, exposure to hydrophobic/hydrophilicenvironments of tryptophan, and conformation of disulfide bonds in proteins.Contrary to infrared spectroscopy (IR), vibrations of nonpolar bonds, such as C-H give rise tointense or moderately intense Raman bands. Aromatic groups also produce high Ramanintensities. These facts suggest that Raman spectroscopy may be useful in studying aromatic andaliphatic hydrophobicity of proteins.In general C-H stretching vibrations of aliphatic and aromatic groups occur in the 2800-3100cm1 range, while C-H bending vibrations are observed in the 1420-1500 cm’. Aromatic groupstend to produce very strong Raman signals in the 800-1300 cm’ range (e.g., 1006 cm’ in Phe)(Carey, 1988).b. Basic principlesWhen a beam of monochromatic radiation strikes a substance, several things may happen to it.The radiation may be absorbed, reflected or it can be scattered, or it can induce fluorescence(MacLeod, 1973). The term scatter means that the radiation is randomly directed from thescattering medium in all directions. Three forms of scatter can be produced: Tyndall, Rayleigh48and Raman scattering. Tyndall scattering is due to the presence of colloidal particles in the lightpath (e.g. fine dust and debris). Rayleigh scattering is very similar to Tyndall scattering, themain difference is in the size of the particles causing the scattering; in the Rayleigh phenomenonthe particles are much smaller. An important aspect of these two types of scattering is that thescattered radiation is at the same wavelength as the exciting radiation. In Raman scattering thescattered radiation is at a different wavelength from the primary exciting radiation (MacLeod,1973).The physical origin of Raman scattering lies in inelastic collision between the molecules and thephotons of the light beam. An inelastic collision means that there is an exchange of energybetween the photons and the molecule with a consequent change in energy, and hencewavelength of the photon. Moreover, since total energy is conserved during the scatteringprocess, the energy gained or lost by the photon must equal an energy change within themolecule, thus, by measuring the energy gained or lost by the photon we can probe changes inmolecular energy (Carey, 1988).Two approaches can be used to explain the origins of the Raman effect: classical electromagnetictheory and quantum mechanics. The classical electromagnetic theory considers that the effectof the interaction between the molecule and an electromagnetic field is an induction of a dipolemoment or polarization, P. in the molecule:P=aEwhere P is the polarizability and E is the electromagnetic field. Since most molecules areanisotropic, that is, they are not spherically symmetrical, the a may be different in the x, y andz directions. Thus a is a 3 X 3 tensor, but as a simplffication it can be treated as a scalar. Thepolarizability is not a time independent constant since the molecule is vibrating. Thepolarizability theory predicts the three basic phenomena observed in a Raman experiment. Itpredicts an unshifted frequency component (i.e., Rayleigh scattering), a component at a lowerfrequency (i.e., Stokes shift) and a component at a higher frequency (i.e, antiStokes) of the49incident light (Tobin, 1972; Rousseau and Ondrias, 1984). The quantum mechanical theory issometimes called the Placzek polarizability concept. This theory was proposed in the early 1930sto explain the experimentally observed intensities of the Raman Stokes and anti-stokesscattering (Baranska et al., 1987). A molecular vibration is Raman active to the extent that itchanges the molecular polarizability. For centrosymmetric molecules, symmetric vibrationschange the polarizability but not the dipole moment while the situation is reversed forantisymmetric vibrations. Vibrations which preserve Local symmetry centers, e.g., the stretchingof homonuclear bonds, are weak in the infrared spectrum but strong in the Raman spectrum.On the other hand, vibrations ofhighly polar bonds produce large changes in the dipole momentbut small changes in the polarizability. In particular, water, which is a good infrared absorber,is a poor Raman scatterer, and produces relatively little interferences in Raman studies ofaqueous solutions. In general, Raman and infrared spectra are cmplementary, in terms ofselections rules and relative intensities, and it is useful in structural studies to have bothavailable (Carey, 1982; Spiro, 1974).c. Raman studies of proteinsRaman studies of proteins and related compounds (such as amino acids and peptides)began before most modern techniques of biochemical analysis. In the 1930s Edsall and hiscolleagues at Harvard obtained Raman spectra of amino acids, and the same research groupobtained the first Raman spectrum of lysozyme in 1958 (Carey, 1983: Carey, 1982). Pioneerssuch as Edsall routinely took two to three days to record a spectrum (Carey, 1983).Raman spectroscopy of proteins can be used to obtain the following type of information (Carey,1983):1. Quantitation of polypeptide conformation, a-helix, 13-sheet, B-turn, etc.2. Characterization of cysteine-SH side chains.3. Conformation of disuhide -S-S-linkage.4. Strength of hydrogen bonds to tyrosine -OH.5. Exposure to hydrophobic/hydrophilic environments of aromatic amino acid side chains,especially tiyptophan.50By using conformation sensitive peaks, it is possible to study the effect of different factors (e.g.,temperature, pH, chemical agents) on the protein conformation. Many proteins have beenstudied by Raman spectroscopy, including hormones, toxins, and enzymes. Information obtainedwith these proteins can be useful in the study of food proteins.i. Important bands in the Raman spectrum of proteinsThe observed Raman spectrum of a protein recorded under nonresonance conditionsconsists of contributions from various amino acid side chain modes, together with modesoriginating from the peptide backbone (Carey, 1982). Some bands or lines are commonly usedin the analysis of the Raman spectrum of proteins. These bands can be classified in two maingroups: conformation sensitive bands and side chain vibration bands.Conformational sensitive bands (i.e., amide I and amide III bands) are related to the peptidebackbone and are widely used to characterize the secondary structure of protein. Bandekar(1992) recently reviewed the role of amide modes on protein conformation. Side-chain vibrationare assignable to specific chemical groupings.ii. Raman studies of selected proteinsIn order to illustrate the use of Raman spectroscopy for conformational analysis ofproteins, Raman studies of three food-related proteins will be discussed in the following section.(a). LysozymeLord and Yu (1970) presented a detailed study of the Raman spectrum ofhen’s egg whiteHEW) lysozyme in water. The Raman spectrum of native HEW lysozyme exhibited severalcharacteristically intense lines whose origins in terms of amino acids residues were identifiedby comparison with the superimposed spectrum of the constituent amino acids. For the nativeenzyme the amino acids with aromatic side groups gave rise to very intense Raman lines. The51S-S stretching band at 509 cm’, arising from four cystine residues per molecule was alsointense (Lord and Yu, 1970). Although this study demonstrated the potential of Ramanspectroscopy as a method to study protein conformation, technical limitations at that time madeunsuccessful the detailed investigation of lysozyme thermal and pH denaturation using Ramanspectroscopy. With the use ofmore powerful lasers and better spectrometers two research groups(Brunner and Sussner, 1972; Chen et al., 1973) reported Raman studies of the thermaldenaturation of aqueous lysozyme. Conflicting results were obtained between the two groups,but experimental conditions were different and may account for the discrepancies. Brunner andSussner (1972) reported that significant changes occurred in several Raman lines whenlysozyme was thermally denatured and that the Rarnan line at 504 cm4, due to S-S stretchingwas temperature dependent; it was essentially zero (i.e., all clisulfide linkages broken) abovethe denaturation temperature (78°C). Contrary to these results, Cheh et al. (1973) repärted thatlittle quantitative changes in the Ranian spectra of lysozyme occurred upon heating in thetemperature range of 32-76°C, and that at the most only one of the four disuffide bondsdisappeared upon denaturation. A briefreport (Fox and Tu, 1979) indicated that the amide I andwide IU bands of lysozyme change upon heating, suggesting conformational change in theenzyme from predominantly a-helix to random coil as the temperature is increased. Thetryptophan residues of lysozyme were investigated using Raman spectroscopy by Miura et al.(1988). They elegantly demonstrated that two of the six tryptophan residues in the protein werein hydrophobic environments. The effects of high-temperatures and pressures on the Rainanspectra of lysozyme were reported recently by Remmele et aL (1990). In another recent report(Miura et al., 1991), the state of H-bonding and hydrophobic interactions of the six Trp sidechains of lysozyme bound to substrate-analogue inhibitors was investigated using Ramanspectroscopy. Li-Chan and Nakai (1991) applied Raman spectroscopy to investigate structuralchanges in lysozyme in gels formed by heating in the absence or presence of the reducing agentdithiothreitol.52(b). RibonucleaseThe Raman spectrum of bovine pancreatic ribonuclease in solid state and aqueous solutionswere compared by Koenig and Fushour (1972). The spectrum of the aqueous solution showedfour bands in the skeletal region at 1109, 1078 and 1059 cm’, while only two bands at 1101 and1030 cm’ were detected in the solid spectrum. In going from the solid state to aqueous solutionno change in the intensity ratio of the C-S and S-S stretching modes at 654 and 516 ciii’ wasfound, suggesting no change in disulfide angles upon dissolution (Koening and Fushour, 1972).Laser Raman spectroscopy has also been used to study the thermal unfolding of ribonuclease(Chen and Lord, 1976; Verma and Wallach, 197Th). In general the results obtained by Ramanspectroscopy agreed well with those obtained by other physical methods. Alteration in theintensity ratio of the tyrosine doublet at 854 and 830 cm4 was found upon heating (Chen andLord, 1976). The low ratio iíi in native ribonuclease near room temperature is indicative ofa substantial amount of strong hydrogen bonding or buried tyrosine residues. As the aqueoussolution of the enzyme was heated, the strong hydrogen bonds were broken, and the intensityratio I/I went above unity (1.5 at 70°C), which indicated that nearly all the tyrosinesresidues became weakly or moderately bonded or exposed. The melting temperature of theenzyme as followed by the tyrosine intensity ratio was about 63°C, which agrees well withvalues determined by other physical methods (Lord, 1977).(c). Milk proteinsThe objective of most of the Raman studies of milk proteins has been the determinationof its secondary structure (Frushour and Koening, 1974; Byler et al., 1988; Byler and Susi,1988). According to Byler et al. (1988) lyophilized a, and 6-casein have around 10% a-helicalstructure, around 20% B-sheet structure and from 20 to 35% turns. Secondary structure werealso estimated for a-lactalburnin and B-lactoglobuiin from their Raman spectra.53The alkaline denaturation of B-lactoglobulin was studied using Raman spectroscopy by Fushourand Koening (1975). Their results indicated that the average tyrosine in the denatured state isin a more hydrophobic environment than in the native state. Also, tryptophan vibrationsappearing at 833 cm’ in the spectrum of the native protein became weaker as the pH increasedto 11.0. When the pH was increased from 6.0 to 11.0, the amide Til line shifted from 1242 to1246 cm’, broadened and decreased in intensity. This was consistent with the conversion of 13-sheet regions on 13-lactoglobulin to disordered conformation. Miura et al. (1988) reported thestate of tryptophan side chain in a-lactalbumin. Using Raman spectroscopy they found that twoof the four tryptophan residues per protein molecule were in a hydrophobic environment. In arecent paper, Nonaka et al. (1993) used Raman spectroscopy to study the thermally inducedgelation of whey proteins.iii. Raman spectroscopy and protein hydrophobicity(a). IntroductionRaman spectroscopy has been used to study the state of Tyr and Trp of proteins. Signalsfrom the aromatic part ofPhe residues have been shown to be rather insensitive to environment.Tu (1986) presented a detailed discussion for the Raman spectra of Tyr, Trp and Phe. In thisthesis more emphasis will be placed in the analysis of aliphatic groups of proteins, an area thathas received very little attention in the literature.In amino acids, peptides and proteins, there are many C-H stretching vibrational bands in the2800-3000 cni’ region. Since they are clustered together, C-H stretching vibrational bands canbe distinguished from other bands quite easily. Although these bands have structuralimplications they have not been extensively studied Cru, 1986).The fact that vibrational modes of nonpolar groups (e.g., C-H) give rise to detectable Ramansignals, suggests its potential use to estimate protein hydrophobicity. When measuring the54hydrophobicity of a protein, the quantification can be made at two levels. In the first level theprotein conformation (i.e., three dimensional structure) is not considered, and the total proteinhydrophobicity is calculated using the amino acid composition of the protein and a scale ofhydrophobicity for each amino acid. This approach is basically a classification or scaling scheme;residues classified as hydrophobic re given a high or low value in the scale, separating themfrom the hydrophilic residues. A similar methodology can be implemented using Ramanspectroscopy. Since enough information is available to assign the major band of a protein Ramanspectrum, it is possible to classilr the Raman bands as hydophobic or hydrophilic. In theory, thesignals from aromatic and aliphatic groups could be integrated, and since these signals areproportional to the concentration of these groups in the protein, a total hydrophobicity valuecould be derived. Clarke et al. (1991) used this group frequency approach to estimate thearomatic composition of gasolines and aviation fuels. Raman bands indicative of aromaticcomponents of a fuel mixture were easily detectable on the basis of their small line width andhigh intensity. Regardless ofthe possible experimental difficulties encountered when comparingdifferent proteins, the above approach could not measure the hydrophobicity at a second level,that is the effective or surface hydrophobicity.(b). Effect of chemical environment on Raman bandsSince surface hydrophobicity is basically the estimation of the exposure of hydrophobicgroups, signals arising from exposed versus buried hydrophobic groups must be diferent enoughto be quantified. Thus, the first question to answer is: to what extent are Raman bands fromnonpolar groups sensitive to changes in environment?In spectroscopy, the term solvatochromism is used to describe the change in position, intensityand/or band width of a band, accompanying a change in the polarity of the medium (Reichardt,1988). These changes are a result of physical intermolecular solute-solvent interactions such asweak non-specific interactions (e.g., dipole-dipole, dipole-induced dipole, etc.) and strong specific55association of solute with solvent molecules, usually of the hydrogen bond type. Theories ofsolvent effect on spectra assume principally that the chemical state (e.g., protonation) of theisolated and solvated chromophore containing molecules are the same and treat this effect onlyas a physical perturbation of the chromophore (Reichardt, 1988).For ir’frared (IR) and Raman spectra, the vibrational spectrum of a A-B bond depends not onlyon the strength of the bond between A and B, but may also be affected by environmental factors.The position, intensity and band width of the bands may be affected (Reichardt, 1988). For JRspectroscopy, solvent induced chang’s have been extensively studied (see Reichardt, 1988), anddifferent theoretical models have been proposed to explain solvent effects. For example, Ikramand coworkers (1993) studied the effect of solvent-solute interactions on the JR spectra ofdimethyldichiorosilane in a number of solvents. The spectral vibrational shift for the Si-Clstretching band was correlated with the refractive index and dielecti4c constants of the’solvents.In comparison to JR spectroscopy, solvent effects have received less attention with respect toRaman spectroscopy. Possibly, the more dramatic change in an environment of a moleculeconsists in going from the gaseous to liquid phase. For this type of transition, most simpleRaman stretching bands are displaced to lower Raman shifts. For example the band ofacetone in gas phase is at 1740 cm’ while in liquid phase is at 1715 cm’. This observationsuggests that formation of intermolecular interactions decreases bond strength, lowering theRaman shift. A greater decrease in Raman shift may be observed for valence vibration ofstrongly polar groups (e.g., 0-H, C=0, etc.). That is why some researchers recommend measuringthe Raman spectra in non-polar solvents (Baranska et al., 1987). Hydrogen bonding alsoproduces a decrease in the Raman shift. For example in the spectrum of 4-chiorophenol thestretching of the hydrogen bonded -0-H occurs at around 3250 cm’, while for a free (i.e, in Cd4)-0-H is at 3619 cm4 (Baranska et al., 1987). Transition from the liquid to the solid state mayaffect the spectrum in two ways. Often a simpler spectrum is observed; some bands disappearbecause of the smaller number of conformations in which the molecules can exist. On the other56hand, local electric fields in the crystal lattice can cause splitting of some bands and can alsoaffect band intensity (Baranska et al., 1987). Rea (1960) reported that the intensity of a Ramanband of a solute is dependent on the refractive index of the solution. When instrumental effectsare factored out, the intensity shows a general increase with a refractive index of the solvent.Jones et al. (1965) reported that the relative intensity of the 459 cm4 of CCI4 ranges from 0.55in methanol to 1.93 in carbon disulfide with 1.0 being the intensity of pure CC14.Kamogawa and Kitagawa (1989) reviewed the use of Raman difference spectroscopy (RDS) tostudy solvent effects on the vibrations of nonpolar groups. They point out that conventionalRaman spectrometers are not adequate for RDS. In their experiments, these authors used anoptical multichannel detector. For the case of mixing two liquids, A and B, they proposed thatthe frequency shift difference (u), measured from the frequency of pure liquid A, arises fromA-A and A-B interactions, being a function of the composition o( the mixture. A theoreticalconsideration suggested that the frequency shift becomes larger as the polarizability of the Bspecies increases and the intermolecular distance is reduced. Kamogawa and Kitagawa (1989)also reported changes in i as a function of ethanol concentration in solutions of ethanol plusorganic solvents. While mixing ethanol with acetonitrile or water causes a positive shift in thestretching bands of ethanol, mixing ethanol with benzene or CCI4 cause a negative u.(c). The Raman C-H vibrationsEnvironmental effects on the Raman bands due to vibrations of C-H bonds have beenstudied in more detail in lipid containing systems than in proteins. Raman spectroscopy is alsoquite useful for studying the structural characteristics of phospholipid molecules in model andbiological systems. The fluidity of the membrane assemblies can be examined since some of thefatty acid hydrocarbon chain vibrations are sensitive to conformation. The choline and phosphategroups of phospholipids can also be studied. The application of Raman spectroscopy to studymembrane systems have been reviewed by Carey (1982) and Parker (1983).57The C-H stretching vibrations appear in the range 2700-3000 cm’. These bands are generallyintense in the Raman spectra and the vibrations are more or less localized within the methyl(-CH3)and methylene (-CH2)groups and give rise to correlatable frequencies. As for many othervibration bands, CH stretching bands can be either symmetric or antysimmetric. Symmetricbands are usually stronger and appear at a lower Raman shift than antisymmetric (Baranskaet al., 1987). In n-alkanes, methyl antisymmetric C-H stretching bands usually appear in thefrequency range 2965-2969 cm1, while methyl symmetric C-H stretching bands fa1l near 2884cm1. Methylene antisymmetric and symmetric stretching bands for n-alkanes appear in theranges 2912-2929 and 2849-2861 cm, respectively (Lin-Vien et al., 1991). The characteristicregions usually contain additional bands arising from various forms of the C-H oscillatorcouplings and from Fermi resonance (Baranska et aL, 1987). The C-H stretching spectra ofaromatic hydrocarbons appear in the range 3100-3000 cm4, which is characteristic ofunsaturated groups (Baranska et al., 1987).The C-H stretching vibration region in the Raman spectra of lipids can be used to identifydifferent states of order. Although both C-H bending and stretching bands in fatty acids areaffected by chain architecture, the stretching region has been reported to be more sensitive(Verma and Wallach, 1977a).The intensity ratio of the 2850 and 2885 cm’ lines (II) increases as the “looseness” of thelateral packing of disordered hydrocarbon chains increases (Larsson and Rand, 1973). When thehydrocarbon chains are in a liquid state (i.e., melted or dissolved in Cd4) the symmetricstretching vibrations of the Cl2 groups at about 2850 cm’ dominate. If on the other hand, thelipid is in a crystalline state, the intensity of the symmetric vibration peak of the CH3 groupsat 2885 cm’ dominates. The same researchers, using mixtures of 1-propanol and water as amodel, also studied the effect of changes from apolar towards a more polar environment of CU2and H3 groups. They reported that as the water content increases and the environment of thepropanol becomes more polar the 2940 cm4 increases compared to the other C-H stretching58vibrations. A crystalline urea inclusion compound of stearic acid also presented an increase inthe relative intensity of the 2930 cm4 peak assigned to asymmetric CH2 stretching, as comparedto the fatty acid crystal. Since in the inclusion complex the hydrocarbon chains are surroundedby urea while in the crystal form the hydrocarbon regions are in contact only with otherhydrocarbons, they suggested that changes in this area are due to the effects of differentenvironments. Comparison of a micellar solution versus a solution of sodium dodecyl sulfate alsoshowed that a higher exposure of the hydrocarbon chain (i.e., in a detergent solution below thecritical micelle concentration) gives an increased relative intensity of the 2930 cm4 peak. Theyalso examined the Raman scattering in the 2900 cm4 region of certain proteins (albumins, flactoglobulin, insulin, collagen and keratin) as well as polypeptides (poly-L-lysine and poly-Lleucine), and the peak corresponding to the asymmetric vibrations of the CH2 groups, at about2930 cmt, was always found to be the dominating peak No comparison between th proteinswas made. Larsson and Rand (1973) concluded that the relative intensity of the 2930 cm1 peakto the other C-H stretching vibration peaks (e.g., 2850 cm1) is increased with increased degreeof polar environment of the hydrocarbon chains. In a later publication, Verma et al. (1975)reported the position of the —2940 cm’ band for chymotrypsinogen, bovine serum albumin,ribonuclease, poly-l-lysine and poly-L-glutamate to be 2926, 2940, 2951, 2936, and 2923 cm’,respectively.Verma and Wallach (1976) reported that the Raman intensity ratio I293O/Iq for erythrocyteghosts or membranes (EM) increases with temperature. They reported that at 22°C, the CHstretching region from EM is characterized by three strong bands that lie at —2930 cm’, —2880cm’ and —2850 ciii’, representing contributions from both the protein and lipid component of themembranes. According to Wallach and Verma (1975) the 2850 cm’ band, assigned to symmetricCH2 stretching, arises predominantly from membrane lipids, while the 2940 cni’ band, assignedto CH2 antisymmetric stretching, contains a greater contribution from membrane proteins.Verma and Wallach (1976) assigned bands near 2930 and 2960 ciii’ to symmetric and59antisymmetric CH3 stretching from amino acids, suggesting that in proteins the C-H stretchingbands are likely to be broad because of the diverse amino acids responsible for these bands. The2880 cm’ band was assigned to symmetric CH3and CH2 stretching mainly from membrane lipidacyl chain segments of length greater than four CH2 (Verma et al., 1975), but mixed with the2875 cm4 band from proteins, resulting in a broad scattering between 2875 and 2890 em1. Asthe temperature is raised from 22°C to above 42°C at pH 7.4, the —2930 cnf’ bands becomeprominent relative to the temperature-insensitive bands featured at 2850 cm4 (Verma andWallach, 1976). At a fixed temperature, increasing the pH from 5 to 9 produces a decrease inthe I2/I intensity ratio. Heating at 60-70 °C from 15 mm causes an increase of this intensityratio from 1.1 to 1.5 (Wallach and Verma, 1975). In contrast to membrane systems, the 2940cm1band ofbovine serum albumin does not change with temperature relative to other scatteringpeaks (Wallach and Verma, 1975). Since other spectroscopic techiiiques show that membraneproteins in these systems undergo a conformational transitions around 40°C and NMR resultsshow that methyl side chains from membrane proteins become more mobile as temperature israised, Wallach and Verma (1975) suggested that the Raman changes observed are related totransitions in both membrane proteins and lipid.Verma et al. (1975) reported that the 2890 cm1:2850 cm1 intensity ratio in lipid containingsystems could be related to acyl chain mobility. They suggested that increasing values of thisratio are related to restricted chain mobility. Lis et al. (1976) used Raman spectroscopy to studylipid-protein systems made of dimyristoyl and dipalmitoyl phosphatidylcholines and smallamounts (< 0.2 mM) of cytochrome c, cytochrome c oxidase, albumin and fibrinogen. Proteinsaffected the lipid hydrocarbon C-H stretch Raman modes.In an interesting paper, Verma and Wallach (1977b) studied the effects of thermal denaturationat low pH on the C-H stretching region of ribonuclease-egg lecithin mixtures. These authorsreported that denaturation resulted in an increase in the intensity of the CH3 stretching bandin relation to the 2850 cm1 CH2 stretching band of lecithin. As in the other papers from this60research group previously discussed, they did not evaluate the protein alone but in a protein-lipid system. In this case, ribonuclease was mixed with a 4% dispersion of egg lecithin. The finalprotein concentration was 4% in a 1:1 mixture ofH20/D.The reported Raman C-H stretchingregion of pure ribonuclease is a broad envelope with a major band at 2930 cm1.Weaker peaksoccur at 2870 ciii’, 2890 ciii’ and 2975 ciii’. All the bands, are broad, probably because ofslightly dissimilar contributions of various amino acid methyl and methylene residues withdifferent environments and chemical linkages (Verma and Wallach, 197Th). For the protein-lecithin system, the intensity ratio IfI,changed from 1.3 at 20°C to 1.7 at 45°C. The positionof the band at 2930 ciii’ did not change upon denaturation. Verma and Wailach (197Th) alsoreported the spectra of methanol, dimethyl sulfoxide and t-butanol for the pure solvents and forvarious mixtures of H20 and D20. In all these solvents, the C-H Raman stretching bands shiftto higher Raman wavenumbers when the solvents are mixed with water as compared to thebands in the pure solvents. For example, pure t-butanol has two strong bands at 2921 cm’ and2977 ciii’, and a weaker band at 2945 cm’. In t-butanollH2Othe strong bands appear at 2922ciii’ and 2980 ciii’, while 2945 ciii’ band appears at 2950 cu1.The ratio of the intensity of the2945-2950 ciii’ weak band to the 2977-2980 cu1 strong band increases in a nonlinear fashionwith the proportion of water in the system, being 0.55 in the pure solvent and 0.69 for 10% tbutanol (Verma and Wallach, 197Th).Based on the effects of water on the C-H stretching bands of this rather limited set of solvents,Verma and Wallach (1977b) concluded that the increase in the band at 2930 cm1 ofribonucleaseis due to the exposure of previously buried aliphatic residues (lie, Leu and Val) to water. It isimportant to point out that they did not analyse the thermal denaturation of pure ribonucleasebut a a mixture of ribonuclease-phospholipid.Li-Chan and Nakai (1991b) reported that an increase in the intensity of the C-H stretchingvibration band at 2938-2942 ciii’ relative to the broad water line at 3250 cu’ occurred uponlysozyme gelation. Although these authors suggested that this increase reflects an increased61solvent exposure of aliphatic side chains, this conclusion should be taken with caution sincematrix effects may be responsible for the observed changes. In another publication, the sameauthors demonstrated that lysozyme-corn oil emulsion (10% protein, 0=0.25) made by eithervortex or sonication gave different Raman C-H stretching spectra (Li-Chan and Nakai, 1991a).Difference spectra, obtained by subtracting the protein and oil spectra from the spectra of theemulsion showed that some peaks shifted to higher wavenumber.For the study of complex molecules by spectroscopic methods, it is usually convenient to firstanalyze smaller molecules in model systems. In a series of more than 10 papers publishedbetween 1938 to 1965, J. T. Edsall and coworkers at Harvard University used Ramanspectroscopy to study the structure of amino acids and related compounds in solution. Thesestudies clearly demonstrated the dipolar nature of amino acids at their isoelectric point (forexample see Edsall, 1937). This research group (Takeda et aL, 1940) reported that ionizationeffects altered the CH stretching vibrations of glycine. While a normal CH2 group attached tosaturated carbon atoms in a hydrocarbon shows two stretching frequencies, a symmetrical onenear 2850 cm’ and an asymmetrical one near 2925 cul’, in glycine, in the cation and dipolar ionforms, these vibrations were at higher frequencies: 2970 and 3015 cm’ for the dipolar ion and2974 and 3017 cm’ for the cation. The line around 2970 cul’ was the more intense of the twoand since it was polarized, it was assigned to a symmetrical vibration. Formation of the anionspecie (i.e., removal of the positive charge on the ammonium group) produced a large alterationin these vibrations. The anion form only gave one broad line near 2940 cm’. These researchersalso reported that alanine, which contains no methylene groups, gave three strong lines near2890, 2950 and 3000 cm’. For this amino acid the position of these lines was not affected byionization.Gaber and Peticolas (1976) gave a detailed analysis on the origin and physical significance ofthe structure-sensitive features in the Raman spectra of phospholipids. In phospholipids, theprincipal bands are the CH2 symmetric (2850 cm’) and antisymmetric (2890 cm’) stretchings.62The Cf!3 stretching bands are shoulders, with the Cl!3 symmetric stretch at 2930 cm’ and Cl!3antisimmetric stretch at 2960 cm’. Upon dissolution of the phospholipid in chloroform theantysimmetric Cl!2band broadens, loses intensity and shifts to higher Raman shifts by 10 cm’,while less change is observed in the symmetric band. The intensity ratioI2JIdrops from 2in the solid phospholipid to 0.77 in solution. These authors suggested that the increase in the2890 cm’ intensity observed when a solid phospholipid is compared to one in solution is due toa Fermi resonance effect between Cl!2 bending modes and stretching modes. In the solidphospholipid, interchain (lateral) packing and the presence of a single conformation promoteFermi resonance, causing an increase in intensity. Another important observation reported bythese researchers is that when a solid hydrocarbon is melted new lines apppear in its Ramanspectrum, while previously sharp features become broader. This broadening is probably due tothe several conformations present in the liquid, while in the solid only one conformation exists.Lippert et al. (1976) proposed a method for determining the secondary structure of proteins bylaser Raman spectroscopy. In this method a scale factor (C) is used. This C factor is the relativeintensity of the methylene deformation observed in the protein at 1448 cm’ in comparison tothat in poly-L-lysine. Since Raman intensity is proportional to concentration, the C factor shouldbe proportional to the number of CH and CH2groups in a protein as long as the vibrations haveconstant intensity across proteins. These authors reported that this was not the case. Theywrote: “It is interesting to notice that for a protein, insulin the parameter C is larger in thecrystal than in the solution. We have other, unpublished, Raman spectra of additional proteins,lyophilized, crystalline, and in solutions which also show a tendency for C to increase withdecreasing amount of solvent. Although extremely speculative at this time, it may be possiblethat further investigation of the parameter in proteins may yield information about theinteraction between the hydrophobic C-H moiety and its aqueous environment”. Areas et aL(1989) reported that the C parameter for a lyophilized phospholipase is 0.71 while for the sameprotein in solution is 0.75. Treatment of the phospholipase with 1 % SDS increases the C63parameter to 0.85. Although these authors suggested that the increase in C value correspondsto the exposure ofhydrophobic groups to solvent, the highly speculative nature of this parametersuggests that until more experimental evidence is available, the interpretation of the Cparameter must be taken with caution.(d). Raman vibrations of waterIt is well documented that the Raman spectra ofliquid water and ice show significant differences(Brooker et al., 1989; Hibben, 1937). The high Raman shift region of the water spectrum at3000-3700 cm’) is assigned to intramolecular stretching of the 0-H covalent bond. Its shape isdue to the the perturbation induced by the noncovalent OH••• intermolecular hydrogen bonds onthe covalent bond. The low-intensity peak at about 1640 cm’ is attributed to the bending of thewater molecule, and the band at <1000 cm’ is related to vibrational motions (Auzanneau et al.,1991). The very low Raman shift bands (60 and 190 cm’) have been assigned to eitherintermolecular hydrogen bonds or to dipole-induced-dipole contributions (Auzanneau et al.,1991). Early research (Hibben, 1937) showed that an increase in temperature caused adiminution in the intensity of the 3220 cni’ bands and an increase in the 3630 cm1 band. Theopposite effect has been shown in going from water to ice. The effect of solute is generallysomewhat similar to that of a temperature increase (Auzanneau et al., 1991). Thus, a highfrequency shift is related to a weakening of hydrogen bonds.Cavatorta et al. (1976) used Raman spectroscopy to study lysozyme-water interactions. Theyreport the effect of lysozyme on the Raman spectrum of water mainly in the 0-H stretchingregion. A decrease in the intensity of the OH Raman peak relative to the protein C-H stretchingpeak was observed. Thus, a plot of I,JJ<, presented upward concavity. The first moment of the0-H band shifts to higher Raman shifts as the protein concentration increases. Intensity firstderivative plots showed. that the greatest shape changes take place in higher Raman shiftregions, while the lower Raman shift area presented mainly intensity changes. Samanta and64Wairafen (1978) reported that instead of a upward concavity, the relationship between proteinconcentration and I/I< presented a small downward concavity, indicating that 13H rises abovewhat it should be if there were no interactions. At a concentration of 0.6 g lysozyme/g water themolar scattering power of water in the 0-H region is 30% greater than it would be if nointeractions occurred between lysozyme and water. In a subsequent article Allota et al. (1981)reported more detailed experiments in this area and concluded that the interaction betweenlysozyme and water does not produce any significant change to the 0-H Raman spectrum. Thisconclusion was also supported by depolarization ratio measurements. They explain theinsensitivity of Raman spectroscopy to detect changes in terms of a relatively small amount ofwater implicated, suggesting that although only 5% of the total water may be involved ininteractions, the interaction itselfmay be strong. These studies suggest that Raman spectroscopymay be somewhat insensitive for the study of protein-water interactions using the 0-Hstretching bands of water. Zhukovskii et al. (1990) reported that the state of water in proteinsolutions can be studied by following the changes in the infrared absorption band at 1985 nm.The coefficient of band asymmetry defined as the ratio of the intensities of its longwave andshortwave components, serves as a quantitative measure of the degree of stability of the water.The less hydrophobic protein ribonuclease presented a greater change in y with increasingtemperature than the changes observed for cytochrome-c and chymotrypsinogen. These resultsgo in hand with the stabilization of water structure due to the hydrophobic effect.Portmann et al. (1992) applied a method to deconvolute the Raman stretching band of water inorder to study the effects of sugars and temperature on different water structural components(i.e., more or less rigidly hydrogen bonded species). Heating (25°C to 45°C) resulted in a decreasein the the quasi-crystalline component in favour of the less organized components. While sugarsat 10% concentration did not affect the area of the decomposed bands, the bands shifted tohigher wavenumber. This phenomenon may be caused by a structure promoter effect of sugars(Portmann et al., 1992).65C. EFFECT OF UREA ON PROTEINS1. Chemistry of ureaUrea or carbamide(H2N-CO-NH)was known as long ago as 1773 as a compound whichis always present in urine. Urea is an important product of the metabolism of mammals. Inmammals, food proteins are metabolized to amino acids, carbon dioxide and ammonia. In theliver these compounds are converted into urea which is transported into the blood and excreted,by the kidneys, into the urine (Orten and Neuhaus, 1982). According to Longinans and de Bussy(1968), pure urea was isolated from urine by Prout in 1821 and synthesized by Wöhler in 1828.This synthesis was of great scientffic importance because previously it had been generallyaccepted that carbon compounds could only be formed by some mysterious life force (Longmansand de Bussy, 1968). -From a chemical point of view, urea is an amide functional derivative of carbonic acid (HO-COOH), the latter compound being a derivative of acetic acid. The presence of the carbonyl group(-C=O) makes the acid derivatives polar compounds. The amide derivatives such as urea, arecapable of strong intermolecular hydrogen bonds, which are responsible for their high boilingpoints. Pure urea is a colourless crystalline solid having a melting point of 132.7°C (Morrisonand Boyd, 1983). Contrary to the guanidine group (NH=C(NH2),urea is weakly basic, formingsalts with strong acids (Morrison and Boyd, 1983).Large quantities of urea are used as fertilizer. It is also used in the resin and plastic industries,as a feed additive, and in the synthesis of different compounds such as barbituric acids. As adiamide, urea is capable for forming polymers, some of them important in the production ofmolded plastics (Morrison and Boyd, 1983). Urea forms crystalline inclusion complexes oradducts with several inorganic salts and with straight chain aliphatic organic compounds butnot with branched chain hydrocarbons. This phenomenon is used in the petroleum industry for66the separation of straight chain aliphatic hydrocarbon from a hydrocarbon mixture (Longmansand de Bussy, 1968).The partition coefficient (log P) of urea in octanol-water system is -1.09, while in diethylether-water, chloroform-water, and oil-water the log P value is -3.33,-3.85, and -3.82, respectively(Rekker, 1977). These values suggest that urea is a hydrophilic compound. It is important topoint out that calculation of the log P values of urea based on the hydrophobic fragmentalconstant (see Rekker, 1977) gives a value of -4.53.Urea is very soluble in water. At 25 °C the maximum solubility in water is 75 g/100 g of water,which is equivalent to 8.019 molar or 12.612 molal. Over the complete solubility range, urea-water solutions do not follows Raoult’s law (P1 = x1P0), which suggests either a molecularattraction or complex formation between urea and water (Chang, 1981).2. Urea as a protein denaturing agentUrea and guanidine hydrochloride (GdnHCl) are the most common protein denaturants.Aqueous solutions ofurea yield cyanate (N=C-OH) on standing, which may carbaniylate variousamino acid side chains of proteins. GdnHCI has no such disadvantage and is a strongerdenaturant than urea (Hamaguchi, 1992). Urea is a nonionic compound, while GdnHC1 affectsthe ionic strength of a solution. Urea does not react directly with disulphide bonds, however theresulting protein unfolding may promote disuphide interchange reactions (Hamaguchi, 1992).The observation that concentrated aqueous solutions of urea denature proteins, was most likelymade over a hundred years ago. Lewis (1931) referred to work done in 1900 and 1902 regardingthe increased solubility of proteins in saturated aqueous urea solutions. In his review, Lewis(1931) pointed out that the mechanism of the effect of urea on proteins is a controversy, whichstill remains unresolved today. Two mechanisms may be responsible for the effect of urea onproteins. The first one is called the “environmental effect” (Lewis, 1931). This effect refers to themodification ofthe properties of the solvent water upon dissolution ofurea. This mechanism has67also been called the indirect effect of urea, where urea alters the structure of water in a waythat facilitates the solvation of a nonpolar compound by water (Kuharski and Rossky, 1984).Urea is known to modif’ many properties of water such as dielectric constant, viscosity, surfacetension, density, refractive index, etc. (Sober, 1970).Urea may also react and/or combine with a protein or nonpolar compound in a purely chemicalsense. This is called the direct mechanism, where urea directly interacts with or solvatesnonpolar groups (Kuharski and Rossky, 1984). The indirect mechanism has received moreattention, and urea is usually referred to as a water ‘structure breaker” (Kinsella et al., 1989).However, in recent years, interest has shifted to the role of the direct effect (Kuharski andRossky, 1984).The aqueous solubility of almost all of the constituent parts of a protein increases withincreasing urea or GdnHCl concentration (Whitney and Tanford, 1962; Tanford, 1970). Basedon data of free energy of transfer of amino acids from water to 8 M urea, Whitney and Tanford(1962) indicated that aqueous urea solutions appear to accommodate nonpolar side chains ofamino acids better than water does and presented a calculation showing that the free energy oftransfer of the hydrophobic groups of 3-lactoglobulin (4Tyr, 4 Phe, 21 Leu, 9 lIe and 9 Val), fromthe interior of a protein to a solvent (i.e., native protein unfolded protein) was 20 kcal morenegative (i.e., more favorable) when the solvent was 8 M urea than when it was water. Whitneyand Tanford (1962) also reported that contrary to some alcohols, urea solubilized the chargedamino acid asparagine, which suggests that in urea solutions, the presence of two kinds ofpolarmolecules (urea and water) produces a more favorable solvent structure for the solubilization ofboth polar and nonpolar groups. This may explain the ability of urea and GdnHCI to unfoldproteins even though it is not clear how they exert their solubilizing effect (Pace, 1986).It was pointed out in the section “Chemistry of Urea”, that urea forms hydrogen bonds easily,and early research suggested that the denaturation effect of urea on globular proteins wasmainly due to the breaking of interpeptide hydrogen bonds which water cannot break In one68of the more influential papers in protein chemistry, Mirsky and Pauling (1936) attributed thedenaturation action of urea or alcohols to their ability to form hydrogen bonds, thus disruptingthe internal hydrogen bonds of the protein. Kamoun (1988) pointed out that this erroneousexplanation is sustained even in current biochemistry textbooks.Creighton (1983) wrote the following on the effect of urea on proteins:“Urea decreases the hydrophobic interaction and this is likely to be the basis of its denaturatingaction. Initial rationalizations focused upon their [urea and GdnHC1J obvious potential forhydrogen bonding, and they were considered to break protein hydrogen bonds. However, furtherreflections, plus some experimental data, indicated that they were no more potent in this respectthan water. Model compound studies demonstrated that both increase the solubilities ofnonpolarmolecules, including those of amino acid side chains, in proportion to their accessible surfacearea, diminishing the magnitude of the hydrophobic effect by up to one third...and that they[urea and GdnHCll directly interact with both the folded and unfolded state to produce a widerange of effects, depending upon the local geometry of the interacting groups in the protein.’Creighton (1983) does not indicate why nonpolar compounds are more soluble in urea than inwater and how urea interacts with the protein. Roseman and Jencks (1975) were very criticalto the “structure breaking” effect (indirect effect) as an argument for explaining the increase insolubility of nonpolar solutes by urea. They almost sarcastically wrote:The kind of confusion between correlation and cause has done nothing to further ourunderstanding and, in fact, only serves to conceal our ignorance of the nature of the drivingforces for [urea-nonpolar solutes] interactions in aqueous solutions. One reason for the popularityof “solvent structure” is that the addition of almost anything to water causes changes in itsstructure so that an “explanation” for any experimental result is immediately at hand. Thisadvantage must, however, be tempered by the dictum that a theory that is not refutable by anyconceivable event is nonscientific”.Based on the results ofa computer simulation study ofa system containing 200 water molecules,1 urea molecule and a model apolar spherical solute (roughly equivalent to neopentane),Kuharski and Roosky (1984) suggested that urea tends to displace water molecules from theapolar solvation shell, causing changes in the apolar hydration shell. The displacement of watermolecules from a configuration of restricted freedom (hydration shell) to one of higher mobility(bulk water) will have both positive enthalpy (endotherinic) and entropy components. If theentropic contribution is greater, this displacement is thermodynamically favorable. Results from69these simulations suggest that in the solubility enhancement of nonpolar molecules by urea theindirect mechanism is not as important as the specific interactions of urea with the solute.In a recent review article, Creighton (1991) summarized the present opinion regarding the effectof solutes on protein stability. Salutes affect proteins by altering the physical properties ofwaterand by interacting directly with the proteins. The two effects can be distinguished by dividingthe free energy of a molecule in aqueous solution between the energy cost of making a cavity inthe solution in order to accommodate the molecule, and establishing the interactions betweenthe molecule and the water upon filling the cavity (Creighton, 1991). Changes in the firstcomponent should be general to all molecules, and should be reflected in the surface tension ofwater. Increasing the surface tension should make it more costly to make a cavity and will tendto decrease the solubilities of other molecules. Urea and GdnHC1 slightly increase the surfacetension of water, however, they interact more favourably than water with the nonpolar surfacesof proteins. Since the unfolded protein has a higher nonpolar exposed area than a native protein,urea and GdHCI stabilize unfolded conformations (Creighton, 1991). Interestingly, more than50 years ago, Anson and Mirsky (1933) concluded that the denaturation ofbovine serum albuminby reagents of the amide class is described as the result of a specific interaction of thedenaturating agent with the protein which favors the denatured form.In an important paper, Alonso and Dill (1991) presented a molecular theory based on statisticalthermodynamics, to predict the effects of nonspecific small-molecule solute action on proteinstability. In this theory, folding is assumed to be driven by solvophobic interactions. The solventdependence of the solvophobic interactions is taken from transfer experiments on amino acidsinto aqueous solutions ofurea and GdnHC1. An important aspect ofthis theory is that it predictsthat at neutral pH, thermal and solvent (i.e., urea and GdnHC1) denaturation will tend toproduce similar conformation in proteins. It also predicts that at any given denaturantconcentration the free energy of folding, the radius of the unfolded molecule and the exposureof nonpolar residues to solvent or to other polar groups depend on the length of the protein70chain, the composition (i.e., the fraction of residues that are nonpolar) and the free energy oftransfer of an average nonpolar residue.3. Application of urea denaturation in protein chemistrya. IntroductionThe urea denaturation has been reported in the literature for many proteins. A computersearch in Chemical Abstract covering the period 1969-1992, using tha search terms “urea andprotein and denaturation” gave a total of 82 references (English language only).In most of the studies, the urea denaturation per se is not the main objective of the research.Urea is commonly used as an agent to obtain an unfolded conformation of the protein(s) ofinterest and then the effect of folding on some protein property, having as a baseline theunfolded state, can then be studied.In recent years and due to the strong interest in genetic manipulation of proteins (i.e., proteinengineering), urea and GdnHCl denaturation have regained importance in protein chemistry.To be biologically active, all proteins must adopt specific folded three-dimensional structures.Yet the genetic information for the protein specifies only the primary structure, the linearsequence of amino acids in the polypeptide backbone. Since most purified proteins canspontaneously refold in vitro after being completely unfolded, the three-dimensional structuremust be determined by the primary structure. How this occcurs has come to be known as “theprotein folding problem” (Creighton, 1990). By studying urea and/or GdnHCl denaturation, theresearcher can get a better understanding of both the forces and mechanisms responsible forprotein folding. Relevant thermodynamic and kinetic parameters can also be obtained.The observation that some proteins acquire a unique structure, the “molten globule” state, witha partially folded conformation that can be distinguished from either the native or the fullydenatured forms under mild denaturation conditions (Hirose, 1993) has also caused a renewed71interest in protein denaturation by urea. Several food proteins including f-lactoglobulin, alactalbumin, ovalbumin, ovotransferrin, lysozyme and serum albumin have been found toacquire a molten globule state under specific conditions such as acidic pH, disulfide bondsreduction, or low concentration of denaturant (Hirose, 1993).b. Biochemistry related studiesIn the biochemistry area, two types of urea denaturation studies are usually made; dynamicand equilibrium studies. In dynamic studies, also known as kinetic studies, the time-course ofdenaturation at a given urea concentration is followed. Denaturation is estimated by measuringsome property of the protein, e.g., enzymatic activity, absorbance or fluorescence maxima. Datais usually analysed using the first order integrated rate equation and a reaction order, whichis a measure of the speed of the denaturation process at a defiuied urea concentiation andtemperature. By performing experiments at different urea concentrations or temperatures andusing the Arrhenius equation the activation energy for the denaturation reaction can be obtained(Brown and Yada, 1991).Equilibrium denaturation studies involve the incubation of the protein with the denaturant fora defined time or equilibration period, ranging from minutes to hours, and then estimating thefraction of the protein folded using any observable protein property such as fluorescenceintensity. Denaturation curves are usually obtained by measuring the protein parameter atdifferent (6 to 10) denaturant concentrations. In most cases a two-state mechanism is presumed.This mechanism assumes that the protein can only have two conformations, the native anddenatured state, and that the denaturation is reversible. At a given denaturant concentrationand after correction for solvent effects, the equilibrium denaturation constant (Ku) and freeenergy of unfolding (eG) can be calculated. The free energy of unfolding extrapolated to zerodenaturant concentration is a measure of the stability of the native protein. For most globularproteins, free energy of unfolding in water calculated using urea or GdnHCl denaturation curves72falls in the range of 4-10 kcallmol (Hamaguchi, 1992). As expected, the AGD becomes morenegative, indicating that unfolding becomes more favorable, as the denaturant concentration isincreased. Equilibrium denaturation curves can be used to estimate the differences inconformation stability among proteins differing slightly in chemical structure because ofdifferences in amino acid sequences or alterations resulting from chemical or geneticmodification. Urea denaturation curves can also give some information about the foldingmechanism of the protein. For example it can be shown that unfolding does not follow a twostate-mechanism. Distinct transitions at different urea concentrations may indicate the existenceofdistinct structural elements while transitions at lower urea concentrations imply less stability.The determination and analysis of urea and GdnHCl denaturation curves has been reviewed byPace (1986).A protein has many chromophores and each of them can be used as a probe of localconformation, thus, in many studies involving analysis of denaturation curves, severaltechniques are used in order to get a multi-faceted picture of local changes due to a structureperturbing effect.Lindsay and Pain (1990) used urea denaturation curves to study the folding and conformationof penicillin G acylase (PA). These researchers use several methods: near and far-ultravioletcircular dichroism, steady-state and time-resolved fluorescence spectroscopy and differentialsedimentation velocity to study the PA conformation. The 86 KDa enzyme was found to bespherical and stable, unfolding over a narrow range of urea concentration in an apparentlycooperative fashion, with a mid denaturation point of 4.5 M urea. For a given protein, the middenaturation point, defined as the concentration of denaturant at which the protein is 50%unfolded, has been found to depend on the technique of measuring the unfolding (e.g.,fluorescence versus CD), on the protein concentration and other environmental variables (e.g.,pH, salt concentration, etc). Mo et aL (1992) reports a urea mid-transition point for jf3-73tropomyosin measured with far-UV CD (222 nm) of 2.3 and 2.8 M for protein concentration of0.05 and 5 mg/mL, respectively.Fluorescence and CD spectroscopies were used by Banik et al. (1992) to study the ureadenaturation of the , repressor, a dimeric protein involved in RNA polymerisation. Three phasesof denaturation are present in its denaturation. A mid-transition point of 3.5 M urea, based onthe CD signal at 222 nm, was obtained. In a recent article, Mrnjana et al. (1993) reported theequilibrium urea and GdnHCL denaturation of bovine liver protein disulfide isomerase (PDI)using fluorescence and CD spectra. With urea denaturation, PDI exhibits a single reversibleunfolding transition and the concentration of urea required to half-denature the enzyme is 4.8M with a AG0 of 5.8 kcallmol.c. Food proteins related studiesUrea has also been used in the study of several food related proteins. The ureadenaturation offood related proteins such as ovalbumin and bovine serum albumin (Gordon andJencks, 1962) and hemoglobin (Simko and Kauzmann, 1962) were among the first ones to bestudied in detail.In systems where extensive protein-protein interactions are present, the role of hydrophobicinteractions can be qualitatively determined by using urea. A typical example of this is givenby the recent work of Prudencio-Ferreira and Areas (1993). These authors studied the protein-protein interactions occurring during the extrusion of soya protein isolate. By measuring thesolubility of the control and extruded sample in different buffer systems: no additives, 8 M ureaand/or 0.1 M mercaptoethanol, they concluded that disulphide linkages, hydrophobic andelectrostatic interactions are the important forces responsible for the structure of extrudates.Urea has also been used to study the effect of unfolding on protein functionality. It is wellknown that for most proteins urea denaturation is reversible. Mo et al. (1992) reports thattotally unfolded tropomyosin in 6 M urea refolds in less than 5 s upon tenfold dilution into74buffer, while for bovine a-lactalbuxnin, f-lactoglobulin, cytochrome c and parvalbumin refoldingis completely or essentially restored within 0.01 s (Ptitsyn and Semisotnov, 1991). Although therefolding was estimated by far-UV CD, which detects major changes in secondary structure andnot tertiary structure, experimental evidence indicates that for most proteins refolding upondecreasing the urea concentration by dilution, should be considered to be very fast.In the studies reported by Hayakawa and Nakai (1985) and Kinsella (1981) refolding effects aredisregarded. Hayakawa and Nakai (1985) studying the solubility of milk and soy proteins, usedseveral methods to partially denature the proteins, i.e., temperature, pH andurea-i j-mercaptoethanol (ME). Their methodology for the urea+ME treatment involved theincubation of the protein in a pH 8 buffer containing 8M urea and 100 mM ME; after standingfor two hours at room temperature, the urea was removed by gel filtration. Most likely, removalof the urea will cause refolding of the protein, thus any observatioi made in the prtein (e.g.,hydrophobicity) after dilution in a benign buffer, cannot be directly related to the unfoldedconformation ofthe protein. Kinsella et al. (1989) in a review article of the relationships betweenstructure and function of milk proteins citing the work of Waniska et al. (1981), indicated thatcomplete disruption of the secondary and tertiary structure of bovine serum albumin (BSA) byurea, destroyed emulsifying activity (EA). After analysis ofthe methodology reported by Waniskaet al. (1981), it can be concluded that the above conclusion is somewhat misleading. The ureadenaturation methodology of Waniska et al. (1981) involved incubation of the protein with 8 Murea for 3 hours at room temperature, then dialysis was performed to remove the urea and theprotein freeze dried. Thus, it cannot be concluded that breaking of secondary and tertiarystructure of BSA by urea destroyed EA, since when EA was measured the protein had secondaryand tertiary structures, a product of the refolding process.Based on the above discussion, it is evident that if the effect of unfolding by urea on someprotein property is to be studied, measurements need to be made in the presence of urea as is75customarily done in the biochemistry area. However, the presence of relatively high ureaconcentration (4-8 M) may cause some interference, which needs be taken into account.In a rather intriguing conclusion, Ismond et al. (1988) reported that apparent surfacehydrophobicity measured by a fluorescence probe method, of the fababean storage protein,vicilin, decreases as urea or GdnHCI unfolding increases. Since unfolding causes the exposureof previously buried amino acids, it is more likely for surface hydrophobicity to increase than todecrease upon unfolding. Interference, due to the presence of either denaturant during theestimation of apparent surface hydrophobicity, is the most likely explanation of their results. Itis important to point out that these authors do not report in detail the methodology used for theestimation of fluorescence probe hydrophobicity. However, based on their other experimentalprocedures, it seems that they measured fluorescence probe hydrophobicity in the presence ofthe denaturant. -Since urea and GdnHCI interfere with “hydrophobic interactions” and these interactions areprimarly responsible for the binding of the hydrophobic fluorescence probes to protein, it is verylikely that the result reported by Ismond et aL (1988) was due to interference of the denaturantwith the binding of the probe with the protein. Ismond et al. (1988) comment regarding theirresults as follows: “The decrease in both aromatic and aliphatic [hydrophobicityl So wasunexpected; deterioration ofconfonnation to a random coil should parallel by an increase in theexposure ofhydrophobic residues. The decreased values observed here may reflect intermolecularaggregation as denaturation progressed.” Regarding the possibility of interferences due to thedenaturants, they write: “The decrease in fhydrophobicityl S was not attributed to fluorescencequenching; preliminary studies showed that the binding [of the probes] was not impacted by thepresence of urea or GdnHC1.” They do not report what type of preliminary experiments theyused.It has also been suggested that some unfolded proteins cannot strongly interact withfluorescence probes as they have no stable hydrophobic clusters (Ptitsyn and Semisotnov, 1991).76However, several investigations indicate that thermal denaturation usually causes an increasein probe hydrophobicity (Nakai and Li-Chan, 1988).Niwa et al. (1989) reported that white-flesh fish actomyosins exhibited an increase in probehydrophobicity (measured as the relative fluoresence intensity of a conjugate of protein withANS) with increase ofurea concentration up to 2 M. Red-flesh fish and elasmobranh actomyosinexhibited less increase. Urea hardly affected the fluorescence intensity of actomyosins frommammalian sources such as beef, pork and whale, and for chicken actomyosin a small decreasein intensity occurred with increasing urea concentrations. These results suggest that theintramolecular hydrophobic interaction in actomyosin is the weakest in white-flesh fish. Theeffect of urea and temperature on the inactivation of actomyosin and myosin C&2 -ATPaseactivity from several fish species was investigated by Arai et al. (196).Klotz and Shikama (1968) reported that the binding of methyl orange, an anionic probe, tobovine serum albumin is suppressed by urea. Even 1 M urea affects the binding and totalinterference is obtained at a urea concentration of 6 M. However, the inhibition is largelyreversible if urea is removed either by dialisis or dilution. The binding of this anion to asynthetic polymer, polyvinylpyrrolidone, which is unordered and swollen in water, is alsosuppressed by urea (Klotz and Shikama, 1968). Urea may also cause the same type ofinterference for the binding between proteins and the anionic probes ANS and CPA, which arewidely used in protein chemistry as polarity probes.i. Effect of urea on some egg white proteinsOvalbumin denaturation by urea was studied in detail by Mckenzie et al. (1963a and1963b). The effect of urea on the levorotation of ovalbumin was strongly dependent on the ureaconcentration and pH. At pH values near 3 in 7 M urea there is an immmediate increase inrotation with no further change with time. However under neutral or alkaline conditions thereaction is very slow; at pH 6.1 and 7 M urea the rotation reaches a plateau value after 2 h.77Sedimentation, diffusion and viscosity results suggest that ovalbumin in 7 M urea behaves asa random coil polymer. The unfolded protein shows a tendency to aggregate, most likely due todisulfide interchange reactions, especially at neutral and alkaline pHs.Meyer and Kauzman (1962) reported that 6 M urea (pH 7.0- 7.5) completely destroys the a-helixof ovalbumin as measured by optical rotatory disoersion (i.e., b0 term of the Moffit equation).They also make the observation that when an increasing amount of SDS is present along withurea, a considerable amount of a-helix is restored.Nishimura et al. (1989) reported that treatment of ovalbumin with a solution containing 4 Murea and 2 M GdnHCI produced a hydrophobic ovalbumin (i.e., 40% higher fluorescence probehydrophobicity). After removing urea-GdnHCI by dialysis, the secondary structure of thismodffied ovalbumin was very similar tc. that of the native protein. However, the increased probehydrophobicity of the modified ovalbumin (i.e., urea-GndHC1 treated) suggests that under theseconditions, renaturation is not complete.Azari and Feeney (1961) reported that iron-free ovotransferrin (OT) was much more sensitiveto urea denaturation and other physical and chemical treatments than iron ovotransferrin (FeOT), as measured by susceptibility ofdisulfide bond cleavage by sulfite and chromogenic capacitywhich is related to iron binding capacity. These authors reported that both OT and Fe-OT werequite stable or any induced changes are rapidly reversible at pH values of 5.5, 7.5 and 8.5 inconcentration of urea from 1 to 8 M. Extensive changes in the chromogenic capacity occurredwhen OT is incubated at pH 9.5 in 4, 6, and 8 M urea.Yeh et al. (1979) indicated that a-helix and p-sheet content decreases with a concomitantincrease in random coil structure, according to the CD data, when OP was in 0.1 M Tris-HCIbuffer, pH 7.8 containing 7.2 M urea. Renaturation by 10-fold dilution of the protein-ureamixture after incubation for 2 h at room temperature caused complete recovery of the native CDspectrum.78Data from laser light scattering and transient electric birefringence suggests thaturea-denatured OT had more globular and larger molecular dimension than native OT. NativeOT was best described as a prolate ellipsoid with a major axis a=68 A and a minor axis b=21A (axial ratio p=O.3l), while for the denatured OT, a=84 A and b=42 A (p=O.50) (Yeh et al.,1979).Ovomucoid has been shown to be highly resistant to urea denaturation, especially under acidicconditions (Li-Chan and Nakai, 1989). However, Waheed et aL (1977) reported that ovomucoidat neutral pH (0.06 M phosphate buffer, pH 7) was denatured by urea as measured by viscosity,UV spectroscopy and fluorescence measurements. According to these authors, the ureadenaturation of ovomucoid was very slow and occurred in two steps involving at least threeconformational states: native, intermediate and fully denatured state. The intermediate statewas obtained at 6 M urea, while the denatured state was obtainedat around 9 M uiea.Lysozyme is known to be resistant to the action of urea to a concentration of nearly 8 M(Tanford, 1970). Hamaguchi and Rokkaku (1960) reported that the intrinsic viscosity and opticalrotatory dispersion constant of native lysozyme at 33 °C and pH 6.4 (0.03 M phosphate buffer)was not affected significantly by 8 M urea. The facts that both chicken ovomucoid and lysozymeare small proteins (28 kD and 14.3 kD) with a high number of disulfide bonds (9 and 4,respectively) (Li-Chan and Nakai, 1989) may be partially responsible for their stability.ii. Effect of urea on whey proteinsO’Neill and Kinsella (1987) reported that little change in the wavelength of maximumintrinsic fluorescence occurred when -lactoglobulin was in urea solution of up to 3 M. At higherurea concentrations the fluorescence emission maximum of this protein revealed a red shift. Theeffect of urea on the relative fluorescence intensity was somewhat surprising; instead of thedecrease in fluorescence intensity expected at high urea concentrations, an increase wasobserved. For most proteins, as aromatic residues become more exposed to the solvent, a79decrease in fluorescence intensity is observed due to solvent quenching. The binding of smallaliphatic molecules such as alkanones with -lactoglobulin was also affected by urea (O’Neill andKinsella, 1987). Urea concentration as low as 1 M decreased the binding of alkanones and at 5M urea, a five fold decrease in the binding constant for 2-nonanone as compared to the nativeprotein is observed (O’Neill and Kinsella, 1987). In contrast to alkanones, the binding of retinolto 3-lactoglobulin was apparently unaffected by the presence of 8 M urea (Fugate and Song,1980).According to Cupo and Pace (1983) the urea midpoint denaturation point of f3-lactoglobulin was4.97 M (25°C, 0.15 M KCI, pH 2.75) with a free energy change of denaturation of 5.84 kcal/mol.The same authors reported that chemical modification of the protein using the only free SHgroup as the target group affects the urea stability. While the mercaptoethylamine-lactoglobulin (f3-lactoglobulin-S-S-CH2-CHN)derivative had a midpoint denaturation of 1.68M, the mercaptocarboxyethyl, mercaptohydroxyethyl and mercaptopropyl derivatives hadmidpoint denaturation of 4.47, 4.19 and 4.23 M, respectively.The main characteristics of serum albumin are its acidity, high solubility and stability againstdenaturation (Peters, 1985). Serum albumin can withstand 8 M urea or 6 M GdnHCI even at44 °C resulting in a temporary loss of helix but without obvious harm (Tanford, 1968). Whenthe disulfide bonds of albumin were deliberately broken by reduction, further unfolding of themolecule can be detected beyond that seen in acid, alkali, 8 M urea or 6 M GdnHC1 (Tanford,1968). This complete unfolding increased the intrinsic viscosity from 23 to 52 cm3 g’. At thisstage, the mo1ecule can be considered a long peptide chain in random configuration, withoutdetectable ordering of its structure (Peters, 1985). McKenzie et al. (1963a and 1963b) reportedthat serum albumin was in a random coil conformation when it was dissolved in 7 M urea. Theseauthors observed gelation of serum albumin and ovalbumin at high urea concentrations.Moriyama et al. (1993) reported a mid point transition at 5 M urea for serum albumin based onchanges in helix content measured using CD. These authors also indicate that upon addition of80small amounts of SDS (<0.3mM) to the solution ofprotein in urea, the helicity of the protein wasmostly recovered.Differential scanning calorimetric (DSC) studies reported by Yamasaki et al. (1991) showed thatserum albumin (2%) in 2 M urea had a denaturation temperature of 45 °C, which is 20 °C lowerthan for the protein in 2 M NaC1. At higher than 2 M urea, the DSC curves did not have anydenaturation peak.The unfolding/refolding of a-lactalbumin has recently become an area of active research (Hirose,1993). Under mild denaturation conditions, such as low ionic strengths and low pH, or in thepresence of intermediate concentrations of denaturants, several proteins exhibit a foldingintermediate known as a “molten globule”. The molten globule state of apo-a-lactalbumin is themost extensively characterized (Hirose, 1993). Moderate concentrations of GdnHCl (1-1.5 M) atpH 8.0 induce the formation of the molten globule state (Xie et al 1991).At pH 2.0 in the presence of 9 M urea guinea pig a-lactalbumin is completely unfolded asdetected by ‘H-NMR (Dobson et aL, 1991). While two of the five Trp are exposed to solvent inits native conformation, urea denatured a-lactalbumin (8 M urea, pH 2-3, 25 °C, 0.15 M KCI)has four Trp exposed (Shukla, 1973). Native a-lactalbumin (pH 6, 25°C, 0.15 M KCI) had afluorescence emission maximum at 342 nm, compared to 353 nm for urea treated a-lactalbuminat pH 2.7-6.5 (Shukla, 1973). The a-helix content of a-lactalbumin was less than 5% in 8 M urea(pH 7.0). At this pH, the native protein has an apparent a-helix content of 35-40% (Shukla,1973). Although lysozyme has a primary sequence highly homologous with a-lactalbumin, thelatter is more sensitive to denaturation. Some studies (Siegel et al., 1972; Awad et al., 1972)have shown that a high arginine:lysine ratio stabilizes proteins in urea and GdnHCI solutions.This ratio is 0.083 for a-lactalbumin, and for lysozyme is 1.79. Bovine serum albumin and flactoglobulin have an arginine:lysine ratio of 0.415 and 0.2, respectively.At high concentration (10%) whey proteins form a gel in the presence of 8 M urea or 6 MGdnHC1 without any heat treatment (Katsuta and Kinsella, 1990). A similar phenomenon has81also been reported for ovalbumin (Mackenzie et al., 1963). Egelandsdal (1984) reported that 6%ovalbumin solutions in the presence of 6 M GdnHCI and 0.1 M mercaptoethanol failed to forma gel, suggesting that disulfide interchain reactions may be involved in this type of gelation.iii. Effect of urea on milk caseinsCaseins are usually isolated from milk by methods involving urea or some similar denaturantat high concentrations. For example in the classical Zittle and Custer (1963) method, purificationofcç1-casein is done by a urea-sodium chloride procedure with a final precipitation of impuritiesby ethanol. w-Casein is prepared by extraction of whole casein with urea-sulfuric acidsupplemented with ethanol precipitation. The native conformation of these proline-rich proteinsmay therefore not fully be recovered within a reasonable time interval after removal of urea(Holt, 1992). Graham et al. (1984) were unable to demonstrate any appreciable differencebetween the circular dichroism spectra of f3-casein isolated with and without exposure to urea.However the possibility of more subtle conformation differences remains, perhaps associatedwith special conformational states adopted during micelle assembly, or prior to phosphorylation,or on binding to the colloidal phosphate (Holt, 1992).Contrary to the egg and whey proteins described before, the caseins can be consider non-globularproteins with a large proportion of their secondary structure consisting of random coils. Thecaseins possess high contents of apolar amino acids, a large number of proline residues (7, 14.2,and 10.2 mol% for a,, [ and ic-casein, respectively) and have a strong tendency to selfassociation into micelles (Kinsella et al., 1989). The high proline content, which is uniformlydistributed throughout the molecules, minimizes the formation of such secondary structure asa-helix. Aithough these facts suggest that re1atively little change in conformation occurs whenthe proteins are denatured by urea, the presence of urea interferes with self-association and thisfact may bring about important changes in protein structure, especially in solutions of highprotein concentration.82Monomeric disperions ofa,1-casein are obtained by employing strongly dissociating media suchas 6 M urea at pH 7.3 and 3 M GdnHCI (Whitney, 1988). CD studies of Creamer et al. (1981)show a low proportion of secondary structure in native;1-casein, but more than in 4.5 MGdnHCI or 6 M urea. The hydrodynamic properties of f-casein suggest that it behaves like anexpanded coil of low axial ratio, rather than a compact globular protein (Holt, 1992). Indissociating solvents such as 6.6 M urea and 3 M GdnHC1, f3-casein exists as a monomerindependent of temperature, pH, or ionic strength (Whitney, 1988). Although1H-NMR and31P-NMR indicated that a high degree of side chain mobility was present in -casein, changesin some NMR parameters were obtained when the protein was dissolved in 8 M urea ascompared to 2JJo (Andrews et al., 1979). Creamer et al. (1981) found that some secondarystructure remains in fI-casein even in the presence of high concentrations of urea or GdnHCI.The heterogeneity of w-casein in milk due to posttranslational glycosylation and disulfideinterchange makes the interpretation of information regarding their size, shape, andconformation difficult. Evidence obtained in dissociating solvents such as 5 M GdnHCI at pH 5.0,7 M urea at pH 8.5, and 33% acetic acid in 0.15 M NaCl indicates heterogeneous mixtures ofpolymers linked together in ic-casein by intermolecular disulfide bonds with mean molecularweights of 88 kD to 118 kD (Whitney, 1988). Cleavage of these disulfides at pH 12 produces adecrease in the observed molecular weight to —23 kD. Similar values are obtained for ic-caseinin dissociating solvents after disulfide reduction with mercaptoethanol (Swaisgood, 1973).ic-Casein displays a strong tendency to interact with a- and f3-caseins except in solutions withdissociating agents such as urea (Swaisgood, 1973).Rollema et al. (1988) reported that the NMR spectra of ic-casein micelles (30 mglmL ic-casein in50 mMD20-phosphate buffer, pH 7.0) and of ic-casein monomers (20 mg/mL ic-casein in 50 mMD20-phosphate buffer, 6 M urea-d4 pH 7.0) showed a typical random-coil pattern.Incorporation of urea into gels for zone electrophoresis of a- and -casein results in completedissociation to monomers, allowing identification of these polymorphic proteins. Native ic-casein,83on the other hand, yields a broad, smeared zone near the origin due to the heterogeneousmixture of covalent polymers. Incorporation ofmercaptoethanol in the electrophoretic procedureallows the separation of the casein monomers. Under these conditions the gel electrophoresis ofreduced ic-casein from an individual cow producing a single variant reveals one intense stainingband and roughly five bands staining with less intensity, all of which disappear upon rennintreatment. These bands are most likely ic-caseins with varying amounts of carbohydrate(Swaisgood, 1973).Clarke and Nakai (1972) reported that urea produces changes in the fluorescence properties ofa conjugate between ic-casein and ANS. Urea caused reduction in relative fluorescence intensityand energy transfer, red spectral shift and decreased polarisation of the conjugate. Althoughthese authors explained these changes in terms of alterations in the conformation of ic-casein,it is also possible to have these modifications in fluorescence propeities if urea is corsidered tointerfere with the binding of ANS with ic-casein.84D. EMULSIFICATION AS A FUNCTIONAL PROPERTY OF PROTEINS1. Protein functionalityIn Food Science, functional properties of proteins are those physical and chemicalproperties which affect the behavior of proteins in food systems during processing, storage,preparation and consumption. It is these characteristics which influence the quality andorgwwleptic attributes of foods (Kinsella, 1982). According to Cheftel et al. (1985) functionalproperties of foods can be classffied into three main groups: (a) hydration properties (dependenton protein-water interaction), (b) properties related to protein-protein interaction, and (c) surfaceproperties. The first group encompasses such properties as water absorption and retention,wettability, swelling, adhesion, dispersibility, solubility, and viscosity. The second group ofproperties is operative during such phenomena as precipitation, gelation, and the fo!’mation ofother structures such as protein dough and fibers. The third group ofproperties relates primarilyto surface tension, emulsification, and foaming characteristics of the protein.2. Prediction of protein functionalityThe use of quantitative structure activity relationships (QSAR), which is based on theformation of empirical models that use linear free energy-related parameters as the independentvariables and activity (e.g., emulsffication, foaming, gelation, etc.) as the dependent variable, hasproduced a better understanding on the factors that affect protein functionality, and in somecases it has been possible to predict functionality of selected proteins by using empirical models(Nakai, 1983; Nakal and Li-Chan, 1988; Nakai and Li-Chan, 1993).The QSAR approach is not a new idea. According to Dunn (1989) the first quantitativerelationship study was published by the Japanese scientist Furukawa in 1918, in which thebiological activity of perfumes was investigated. The first QSAR paper was published by Hanschin 1962 (Hansch et al., 1962) in which the relationship between chemical structure of85phenoxyacetic acids and their biological activity as plant growth regulators was investigated.Hansch is considered to be the father of the QSAR field.The basic assumption for the QSAR approach is that variation in activity or functionality canbe correlated with hydrophobic, electric and steric parameters. In the case of small molecules(mw<1000 daltons) the above energy-related parameters are relatively easy to calculate ormeasure. However, for large and complex molecules, such as proteins, it is often difficult, if notimpossible, to determine them. Instead well-defined physicochemical properties such assolubility, viscosity and surface tension are commonly used as independent variables.Solubility, a functional property in itself has been thought to be the most important factor thatdetermines protein functionality (Kinsella, 1976). However the assumption that proteins withhigh solubility have good functional properties is not always correct, especially with emulsifringproperties (Nakai, 1983; Cheftel et aL, 1985). Various investigations have shown that whenhydrophobic parameters were used in conjunction with solubility, better correlations to variousfunctional properties were obtained than when only sohibility was used (e.g., Townsend andNakai, 1983; Voutsinas et al., 1983; Li-Chan et al., 1984, 1987). A recent review in theapplication of QSAR in structure and function of food proteins has been published by Nakai andLi-Chan (1993).a. Artificial neural networks: a new tool in structure-functionality studiesStatistical techniques play an essential role in the quantitative analysis ofthe structure-activity relationships in food proteins (Kinsella, 1976; Nakai and Li-Chan, 1988). So far,regression and multivariate analysis techniques, such as principal component and discriminantanalysis have been used (Nakai and Li-Chan, 1988). However, these techniques have somelimitations, which affect their applicability. In addition, they are not very efficient in analyzingnon-linear systems (Pineda, 1989; White, 1989; Aishima and Nakai, 1991). Functionality86analysis by artificial neural networks, an alternative technique to parametric statistics, ispresented in this thesis.What are artificial neural networks? Artfficial neural networks or artificial neural nets are anew technique for the analysis of multivariate data. Artificial neural networks (ANN) are basedon simplified mathematical descriptions of what is known about the physical structure andmechanism ofbiological cognition and learning (Wythoff, 1993). In simple terms artfficial neuralnets (ANN) are a computer simulation of biological parallel signal processing (Lawrence, 1991;Smith and Walter, 1991; Katz et al., 1992). Artificial neural nets are usually compossed ofcomputer simulated layers of processing elements or artificial neurons. The artificial neuronreceives some number of input signals, and calculates its own output in a two step process; firstthe weighted sum of its inputs is computed and then the resulting sum or activation level ispassed through an output or transfer function (e.g., a sigmoid funtion) to obtain the neuronoutput. The processing units are usually arranged in layers. The first layer consists of neuronswhich take on the input values of the variables and is called the input layer. The input layerdoes not process the inputs, but serves merely to distribute them to the next layer. Followingthe input layer are one or more hidden layers; they receive no input from, and produce no outputto the outside world (Wythoff, 1993). Hidden layers extract information or features from inputs.Finally, the output layer produces the network output. For a given problem, the number ofinputneurons is fixed by the number of input variables, and the number of output neurons is fixedby the number of values which are desired. Input and output neurons are equivalent toindependent and dependent variables of the problem, respectively. Selection of the number ofhidden neurons and hidden layers is commonly done empirically, since no absolute rule existsand it seems largely dependent upon the complexity ofthe problem (Lawrence, 1991). In general,to improve the ANN prediction accuracy using a “training set”, the number of neurons in thehidden layer needs to be increased. Alternatively, to improve the generalization capabilities, thusimproving the performance with new cases, the size of the hidden layer needs to be reduced. If87too many hidden neurons are used, the ANN will tend to perform very well with the trainingset, but when presented with new cases, its prediction ability may be very low (i.e., overfitting).In this case the ANN does not learn to generalized, but only “memorizes” the individual trainingset pattern. On the other hand, if too few hidden neurons are used, the prediction ability of theANN may be poor, making it useless (i.e., underfitting) (Wythoff, 1993).The weights on the input lines represent the strength of the connection to a neuron, andlearning rules (such as the back-propagation algorithm) alter these weights in order to createthe desired inputJouput response from an ANN. The “knowledge” or functionality of ANN isencoded in the values of the weights (Katz et al., 1992). Despite the simple nature of theindividual artificial neurons, the network as a whole has impressive capabilities. A thoroughexplanation of ANN theory is given elsewhere (Rumeihart et al., 1986; Eberhart and Dobbins,1990; Wythoff, 1993). Artificial neural nets are not an expert system since they do not use arule-based architecture. Artificial neural nets learn by association, just like humans. A set oftraining facts (i.e., inputs and outputs) are first presented, and using an error minimizationalgorithm, the correct weights between the neurons are optimized in order to correctly predictthe training outputs using the training inputs. Once the ANN has “learned” the associationbetween these inputs and outputs, a new set of inputs can be presented and the ANN willpredict the corresponding value of the outputs (Eberhart and Dobbins, 1990; Jansson, 1991;Lawrence, 1991; Smith and Walter, 1991).3. Emulsffication as a functional property of proteinsa. Introduction and definitionsEmulsification is one of the most important functional properties of food proteins. Someexamples offood emulsions are milk, butter, cheese, mayonnaise and ice-cream. Since emulsionsare a type of colloidal system, a discussion on these systems, will follow.88The simplest type of colloidal system consists of a single dispersed phase of particles in a secondcontinuous phase called the dispersion medium (Dickinson, 1992). He (1992) defines phase asa region of the system either spatially connected or discontinuous, in which there is constancyin measurable intensity variables like density, pressure, and dielectric constant. A system ofsolid particles smaller than 1 im suspended in a liquid is called a colloidal dispersion (or sol).If the particles are larger than 1 pm and the particles are prone to settle under gravity the termsuspension is used. A dispersion or suspension of liquid droplets in a liquid continuous phaseis an emulsion (Dickinson, 1992). In many food emulsions, the dispersed and continuous phasesmay not be pure liquids, but may be partially crystallized (oil phase) or gel (aqueous phase) asin many dairy desserts (Dickinson, 1992).Most colloidal systems are polydisperse, that is the particles are not all in the same size or havethe same properties. Closeness to pure monodispersity is a desirable attribute in colloids forfundamental studies in the laboratory because many theories relating to cofloids apply to theideal monodisperse situation. A powder containing nonporous particles of colloidal dimensions,has a high surface area. Consider the change in the surface area of a piece of matter as it issuccessively subdivided. A 1-cm cube, initially with an area of 6 cm2, when divided into 1012cubes of 1 pm edge length has an area of 6 m2, whereas when divided into 1018 cubes of 10 nmedge length, it has an area of 600 m2. For colloidal particles the proportion of molecules at thesurface may be as high as 30%; hence, surface properties become increasingly important thesmaller the particle size (Parfltt, 1989).The study ofemulsions has been concentrated in three main areas: emulsion formation, emulsionstability and emulsiflers.b. Emulsion formationThe manufacturing of emulsions is a highly energetic process. Fresh interface isgenerated using high shear forces, and the nascent droplets must be rapidly protected against89premature coalescence by adsorption of emulsifier at the interface (Dickinson and Stainsby,1987). Rapid coalescence is often very important with proteins, because the development of astabilizing surface film around new droplets is very slow (Hailing, 1981). From a thermodynamicpoint of view, the dispersion of a nonpolar substance in water is not a spontaneous process andin many cases a large amount of mechanical energy must be supplied t the system. The maincomponent of this thermodynamic barrier is the formation of new oil-water interface (Shaw,1992). Except for the relatively small amount needed for the surface free energy, all thismechanical energy is dissipated as heat. The free energy (aG) required to increase the area ofoil-water interface by an amount AA is:AG =‘yAAwhere ‘y is the interfacial tension (Dickinson, 1992). Interfacial tension usually lies between theindividual surface tension of the two liquids in question. Interfacial tensions of many alkaneswith water are in the range of 21.8±0.7 inN m1. It is well known that a reduction in yleads toa reduction in the intensity of energy necessary to obtain a certain droplet size. Roughlyspeaking, to halve the mean droplet size requires four times the energy intensity.Many different systems are available for emulsion formation. The method most widely used inthe food industry to make fine emulsions is a high-pressure homogenization using a MantonGaulin valve (Dickinson, 1992). A homogenizer is a device in which dispersion of the liquid isachieved by forcing the mixture (oil-i.water+emuisifier) through a small orifice under very highpressures. This results in very fine particles of about 1 jim or smaller (Gopal, 1968). Dropletsare disrupted by a combination of turbulence and intense laminar shear flow (Dickinson, 1992).The high power consumption leads to the warming of liquids by several degrees after a singlepassage through the homogenizer. Colloidal mills are also popular devices for emulsionformation. In this case the premix is fed through a narrow gap made by a rotor, rotating at highspeed, and a static surface. The gap may be as small as 100 jim. Very strong shear flows are setup and the liquid interfaces are violently torn apart (Gopal, 1968). Colloid mills are very90versatile. The input materials may be liquids or pastes. Particle diameters of the order of 2 pmare easily obtained (Gopal, 1968). For laboratory scale, various types ofblenders and mixers areavailable, most of them producing average droplet sizes in the range of 5-10 im with variousdegrees of polydispersibility (Dickinson, 1992). Emulsification can also be produced by high-energy ultrasonic waves. Droplets smaller than 1 can be formed, but there may be a wideor bimodal spread of sizes (Dickinson, 1992). The advantage of ultrasonic homogenizers is thatvery small sample sizes can be used.Tornberg and Hermansson (1977) compared emulsions made with four different types ofemulsifying equipment, namely, an Ultra-Turrax, a Sorvall omnimixer, a valve homogenizer andan ultrasonic apparatus. An emulsion stability test measuring creaming stability (see sectionEmulsion stability) was the parameter used to compare the emulsions made with 2.5% protein(soy protein isolate, sodium caseinate or whey protein concentrate)and 40% oil by weight. Theauthors found that the type of emulsifying apparatus, the emulsifying time and intensity all hada marked influence on the properties of the protein stabilized emulsions. They showed that forall types of apparatus and proteins, increasing the emulsifying intensity and time, led to theformation of more stable emulsion up to a certain limit, whereupon no further improvement increaming stability was obtained by additional increases in these factors.Most of the processes that are important during emulsion formation occur on a time-scale ofmilliseconds or less: droplet formation, emulsifier adsorption, emulsifier spreading, and dropletcollision (Walstra, 1983). To make a stable emulsion, the time lag between creation and collisionof small droplets must be long compared with the time for emulsifier adsorption and formationofa stabilizing layer at the oil-water interface. Diffusion-controlled adsorption ofmacromoleculesat the oil-water interface is slower than for small-molecule surfactants (Walstra, 1983). Howeverchanges in interfacial tension associated with variations in the chemical nature and compositionof the emulsifying materials offer only a relatively small degree of control over droplet size, ascompared with variations in energy input which may extend over several orders of91magnitude (Dickinson, 1992). Under turbulent flow conditions ofhigh-pressure homogenization,protein particles such as casein micelles are adsorbed at the oil water interface more quicklythan individual protein molecules (Walstra, 1983), suggesting that relative rates of adsorptionbased on diffusion under quiescent conditions, as in most laboratory experiments with planarinterfaces, may give misleading information regarding commercial emulsification (HaIling, 1981).In comparing the thermodynamics of adsorption between different compounds such as proteinsand surfactants the concepts of efficiency of adsorption and surface activity need to bedistinguished. Efficiency of adsorption relates to the molar concentration in the bulk phaseneeded for the saturation of the interface, while surface activity is related to the strength ofadsorption of the compound to the interface (Dickinson, 1992). In general, proteins are moreefficient adsorbers than small-molecule surfactants, but since proteins are completely displacedfrom the interface by surfactants, they are less surface active. ‘flue more effective interfacepacking of the surfactant lowers the surface energy to a greater extent than polymer molecules(Dickinson, 1992). When proteins are compared, a protein is said to be more surface active whenit gives a lower surface tension or higher surface pressure at similar bulk concentrations(Dickinson, 1992). Various proteins have been shown to give emulsions of different droplet sizes.Pearce and Kinsella (1978) reported that droplet sizes ranged by 2.5 fold between differentsoluble proteins.Mita et al. (1973) founded that for the emulsffication of benzene with a blender, the largestdroplets tended to be formed at pH values near the p1 of the protein. With BSA, coalescence onstorage was slower under conditions where the largest droplets were obtained. Surprisingly, theeffect of surface tension was not very clear. It is possible that these effects are due to the surfacerheology of the protein film opposing the deformation necessary for droplet breakup, so themaximum rigidity near p1 leads to the largest droplets.92In the food area two parameters have been used to measure the ability of a substance formaking an emulsion: emulsifying capacity and emulsifying activity index. These parameters donot consider the stability of the final emulsion.Emulsion capacity (EC) is usually defined as the maximum amount of oil that a given amountof substance can emulsify under standard conditions. EC ha a long tradition of use in the meatprocessing area. Originally, it was assumed that the higher the EC for a food ingredient thebetter its performance in a final food product. However, recent research has criticised the useof EC as a quality predictive variable. HaIling (1981) indicates that EC is possibly a measureof the emulsion inversion point. The surface properties of proteins seem to be the major sourceof influence on the EC test since a substantial proportion of the protein becomes associated withthe oil droplets during the test (HaIling, 1981).Pearce and Kinsella (1978) developed a turbidimetric method to estimate the interfacial area ofan emulsion. According to the Mie theory of light scattering, the turbidity of a dilute suspensionof spherical particles, which are larger than the wavelength of the light, is related to itsinterfacial area (Cameron et al., 1991). A simple formula based on the turbidity, volume fractionof dispersed phase and weight of protein was used to calculate a proposed emulsifying activityindex (EM) of the protein. EM was defined as:EM = 2T/ØCwhere C is the weight of protein per unit volume of aqueous phase before the emulsion isformed, P is the turbity and 0 is the volume fraction of dispersed phase. According to Pearce andKinsella (1978) the EM has units of area of interface stabilized per unit weigtht of protein (e.g.,m2g”). However, recently Cameron et al. (1991) pointed out that the Pearce and KinselIa (1978)formula of EM was in error, and proposed a new formula with the following form:EAJ0= 2T/C(14)Due to its simplicity, EM is possibly the most popular parameter for comparing the effectivenessof different proteins as emulsifiers (Dickinson and Stainsby, 1988). However, because of the93assumptions used in deriving the formula for EM and the experimental methodology generallyused, this technique has several limitations. As pointed out by Dickinson and Stainsby (1988)one problem with the EM concept is that the relationship between turbidity and mean dropletsize is not so simple as Pearce and Kinsella (1978) suggested. For droplets that are randomlypositioned, the specific turbidity (turbidity per unit weight of droplets in a given volume)depends on three variables: the droplet diameter (assumed to be monodisperse), the wavelengthof light and the ratio m of the refractive index of oil and aqueous phases (for food emulsion mwill be close to 1.10). At this value of m for a monodispersed system, the specific turbidity firstrises with increasing particle diameter, and then falls (Dickinson and Stainsby, 1988). Althoughwith coarse emulsions (diameter >4 tim) the turbidity increases with increasing weight fractionsof droplets, the test is not sensitive to fine droplets which may contribute most of the interfacialarea (Dickinson and Stainsby, 1988). Because of the large angle of acceptance of somespectrophotometers, some apparatus do not measure true turbidity. This may incorporate anadditional error to EM (Cameron et al., 1991).Dickinson and Stainsby (1988) pointed out that another problem with the original EMmethodology is the sample dilution performed before turbidity determination. In the methodologyof Pearce and Kinsella (1978), after the emulsion is prepared the emulsion is diluted 1:1000 to1:5000 in 0.1% SDS an anionic detergent in order to obtain absorbance of less than 0.4. At thisconcentration SDS is well above its critical micelle concentration of 8.27 mM. This dilutionmedium stabilizes the emulsion droplets and significantly improves the reproducibility of themethod. It is very likely that SDS displaces the proteins from the interfaces, which may causechanges in the droplet size, thus, the measured turbidity may not be related to the originaldroplet size of the emulsion. Cameron et al. (1991) used as a diluting medium a solutioncontaining 1% (w/v) dextran and 0.1% (v/v) of NP-40, a polyglycol ether nonionic detergent.Dextran was included in the diluent to increase viscosity and reduce the rate of flotation of oildroplets, and the detergent was used to prevent flocculation and coalescence of oil droplets.94Dickinson and Stainsby (1988) concluded that the EM test is useful only for comparing similaremulsions in one laboratory. It can assess total interfacial area only for emulsions having anarrow droplet size distribution, and its use can be quite misleading when comparing emulsionsof different polydispersity.Cameron et al. (1991) presented a comparison of EM values (original and corrected formulas)calculated by turbidity to the ones computed using diameter and 0 for three sizes of latex beads.Turbidity derived EM were highly inaccurate. For latex beads with nominal diameter of 2.02,10.2 and 19.6 tim, the nominal and EM0values were 0.304 and 0.217, 0.0616 and 0.0276, and0.0345 and 0.0167. In spite of its lack of accuracy, turbidity derived EM values permit validcomparison of proteins within one laboratory (Cameron et aL, 1991). As pointed out by Pearceand Kinsella (1978), EM values are not a property of the protein, but a property of the systemas a whole. Emulsion volume, the type of emulsifring apparatus, tile speed and duration of itsoperation, the protein concentration and the type and amount of oil all effect EM.c. Emulsion stabilityAny emulsion is potentially unstable and with time undergoes destabilization, separationand breakdown. Stability implies no tendency towards structural change. A stable food emulsion,therefore, is one in which the number and arrangement of the droplets changes imperceptiblyslowly within the time-scale of interest (Dickinson, 1992).With liquids which are highly immiscible, such as tryglycerides and water, the primaryprocesses ofinstability are creaming, flocculation and coalescence. Creaming and flocculation willalmost always occur to some limited extent, but can usually be tolerated up to a certain levelsince they are generally reversible. On the other hand, droplet coalescence, with the associatedrelease of free fat from oil-in-water emulsions, is normally unacceptable at any level (Dickinson,1992).95Emulsion stability is both a thermodynamic and kinetic phenomena. Based on thermodynamicconsiderations any system will try to reach a global minimum in its internal free energy. Foran emulsion, this minimum is when the dispersed phase becomes united in a singlehomogeneous lump. In practical terms, what is more important is to know the rate of changeover the time-scale of interest, which could be hours to years.The study of colloid stability has a long tradition in physical chemistry. Some theoreticalconcepts derived from these studies, such as hydrophobicity, long range interactions and thecolloid interaction theories, have found to be useful in other areas. The basis of the stability ofcolloids, including emulsions has been described theoretically from the point of view of energiesof interactions. Several interactions are operative in an emulsion system, and the net effectdetermines overall stability. A detailed review of these interactions can be found in the recentreview article of Das and Kinsella (1990a).Dickinson (1992) summarized the relative importance of twelve key factors in relation to theseparate processes of creaming, flocculation and coalescence. This information is given in Table3. For purpose of this thesis, only the process of coalescence will be discussed in detail.i. CreamingCreaming is the movement of oil droplets, under gravity or in a centrifuge, to form aconcentrated layer at the top of an oil-in-water emulsion, with no accompanying change in thedroplet-size distribution. In the early stages there is merely a vertical concentration gradient ofdroplets, but later a distinct boundary may appear between an upper cream layer and a lowerdepleted layer (Dickinson, 1992). Creaming is reversible; the original uniform distribution ofdroplets can usually be re-established by gentle mixing (Dickinson and Stainsby, 1987). Sutheim(1946) pointed out that creaming is a special case of sedimentation, and both of thesephenomena obey a rather simple law known as Stokes’ law of sedimentation.96Table 3. Twelve main physical factors affecting the stability and rheology of emulsions (O=notimportant, 1= sometimes important, 2= often important, 3= generally important).Factor Creaming Flocculation Coalescence RheologyDroplet size 3 2 1 1Droplet size distribution 3 2 0 2Droplet volume fraction 3 3 3 3Density differences 3 0 0 0between phasesRheology of continuous 3 3 2 3phaseRheology of dispersed 0 0 0 1phaseRheology of adsorbed 0 0 3 2layer -Thickness of adsorbed 1 2 3 - 2layerElectrostatic interactions 1 3 2 1Steric (polymer) 0 3 2 2interactionsFat crystallization 0 0 3 3Liquid crystalline phases 1 1 2 2Source: Dickinson (1992)97Creaming results from a density difference between the two phases, since the densities of mostedible oils and melted fats are lower than water, and is not necessarily accompanied by dropletcoagulation or flocculation, although it facilitates its process (Shaw, 1992). The theoretical rateof creaming (v) depends on the droplet size (d), density difference (tsp) between oil (pr,) and theaqueous phase(p) and the Newtonian viscosity (11) of the medium as expressed by Stokes’equation:v = Apgd2/18’rThe sign of Ap or (PD-PC) determines the sign of V. In the majority of oil in water emulsions,upward sedimentation (negative v in cm/sec) is observed. Increasing the gravity constant (g),accelerates the creaniing rate. Centrifugation is employed in the milk industry for this purpose.The speed of creaming is inversely proportional to the viscosity (ri) of the continuous phase. Amedium of high viscosity acts like a brake and slows down the motin of any dispersed particle.In addition, the velocity of creaming is directly proportional to the square of the diameter, whichindicates that a relatively small change in particle size bring a large change in the speed ofcreaming. Using the above equation, it can be calculated that a particle of 10 pm, as a fatglobule in fresh milk (i0.1 poise, Ap = PD-PC = 0.93-1.03 = -0.1), will cream at the rate of 4.7 cmper day, whereas a particle of less than 1 pm, as in homogenized milk will cream less than 1120cm per day or 0.33 cm per week. Thus, for practical purposes creaming is practically eliminatedby homogenizing.ii. FlocculationAggregation is the process whereby two or more colloidal particles become stuck togetherby forces of unspecified origin and magnitude. The terms coagulation and flocculation are usedto distinguish between the formation ofrigid aggregates and loose aggregates (Dickinson, 1992).Forces between coagulated particles are strong, whereas those in flocculated aggregates (floes)98are weak. In contrast to coagulation, the process of flocculation is normally regarded asreversible (Dickinson, 1992).Flocculation occurs because of net attractive long-range interactions between different droplets.These net interparticle attractive forces arise from the combined effects of a very large numberof separate forces between pairs of atoms or molecules on different particles. Except for highlypolar materials, it is the sum of the London dispersion forces that contributes nearly all the vander Waals attraction between colloidal particles (Dickinson, 1992; Das and Kinsella, 1990a).The total energy of interaction between the particles in a hydrophobic colloid is obtained bysummation of the electric double layer and van der Waals energies. If this interaction iscalculated as a function of distance between particles, the so-called potential energy curves areobtained. At short particle separation, van der Waals attraction will predominate (Shaw, 1992).For some systems, a second minimum (i.e., attraction) is present at ielatively large interparticleseparations. If the minimum is moderately deep compared to the average thermal energy of theparticles, given by kT, where k is the Boltzman constant (1.380658 x 10 J•K’) and T istemperature in °K, it gives rise to a loose, easily reversible flocculation (Shaw, 1992).Flocculation is usually observed for relatively large droplets (>1 inn) and not for relatively small(<0.2 pm) droplets (Das and Kinsella, 1990a). Flocculation kinetics is dependent on emulsiondroplet size. In practical food emulsions, larger droplets (>2pm) flocculate quickly under shearflow (orthokinetics flocculation). Creaming promotes flocculation. Polydispersed systems usuallyflocculate faster than the monodisperse emulsion Was and Kinsella, 1990a).Flocculation of emulsions depend on the nature of the protein stabilizing the emulsion, and onthe solution conditions such as pH and ionic strength Was and KInsella, 1990).Tornberg and Ediriweera (1987) studied the flocculation behavior ofprotein stabilized emulsions.Caseinate, whey protein concentrate, bovine plasma and soy protein presented different degreesof flocculation and in all cases flocculation increased with increase in ionic strength.99Being the precursor to creaming and coalescence in many situation, it can be seen thatflocculation is an important phenomenon in food colloids. Sometimes the onset of flocculationitself is perceived by the consumer as an indication of poor product quality. An example of thisis the “feathering’ of homogenized coffee cream (Dickinson, 1992). Flocculation is usuallymeasured using droplet-size or rheological techniques. Tornberg and Edirweera (1987) used anarbritary flocculation scale based on optical microscopy at 400X.iii. CoalescenceCoalesence is the irreversible coming together of two or more emulsion droplets to forma larger single droplet. In relation to emulsion shelf-life, coalescence, with the associated releaseof free fat from oil-in-water emulsions, is a much more severe form of instability than creaming;when perceived by the consumer in a food product, it is almost invariably considered quiteunacceptable (Dickinson, 1992). When the dispersed phase is liquid, extensive coalescenceusually produces oiling off, i.e., free oil on the surface. Thermodynamic stability is reached whenall the droplets in an emulsion have coalesced into a single homogeneous region of oil or aqueousphase. At this point the emulsion is said to have been broken, thus resistance to coalescence isessential for long-term stability (Dickinson and Stainsby, 1987).Essentially, coalescence is produced by droplet collisions due to Brownian motions, shear flowor creaming, and to the breaking of the adsorbed films at the oil-water interface (Das andKinsella, 1990a). Various factors, such as solubility, type and concentration ofthe emulsifier, pH,salts, phase-volume ratio, temperature, and properties of the film, affect coalescence stability ofemulsions (Das and Kinsella, 1990).In a system containing no emulsifier, when two droplets come close together, their coalescencedepends on the stability of the thin film of continuous phase separating them (Dickinson, 1992).For these type of systems, lamella rupture is the cause of coalescence. Hailing (1981) defIneslamella as the thin layers of continuous phase liquid separating two adjacent droplets.100According to Fisher and Parker (1985), the first step in emulsion breakdown is drainage ofcontinuous phase or bulk liquid from between the droplets. These authors indicated that mostof the common food emulsion stabilizers slow down the drainage process, either by creating arelatively thick liquid crystalline film or, as for proteins, by having parts of the moleculeprotruding into the continuous phase. As drainage proceeds, the droplets are increasingly pulledtogether by van der Waals forces of attraction, and these forces must be balanced by some forceof repulsion (e.g., electrostatic forces, steric repulsion, hydration repulsion, etc) if the dropletsare not to contact and coalesce (Fisher and Parker, 1985).For protein-stabilized emulsion the adsorbed protein film at the oil-water interface is the mainfactor influencing coalescence stability. Dickinson and Stainsby (1987) indicated that “the longterni stability of emulsions and foams depends in large part on the thickness and strength ofadsorbed films at the oil-water and air-water interface.” Dickinsn (1989) pointed out thatpresent theories of droplet coalescence have limited applicabifity in protein-stabilized systems.Many researchers (Powrie and Tung, 1976; Das and Kinsella, 1990a) clearly point out that inemulsions, the nature of the interfacial film, that is the oriented emulsifier molecules aroundeach droplet, have a definitive role in coalescence stability.Experimental evidence in model systems suggests that coalescence is a two-stage processinvolving film thinning followed by film rupture. A pair of droplets will coalesce when their filmis thin enough to be ruptured by random fluctuation. For droplets with no adsorbed films, theorypredicts that coalescence is promoted by large droplet diameters, weak repulsion between thedroplets and a low interfacial tension. For protein-stabilized emulsion, the high viscoelasticityof protein films, presenting both liquid and solid like characteristics, inhibits film rupture(Dickinson, 1992). Biswas and Haydon (1962) suggested that the main part of themacromolecular film is on the “continuous phase” side of the interface, thus explaining whywater drops are not stabilized by highly viscoelastic BSA or pepsin ifims.101The oldest and simplest explanation for stabilization of emulsion droplets by proteins is that theadsorbed macromolecular film forms a sort of “protective skin” around the droplets which actsas a mechanical barrier to coalescence. According to this widely accepted interpretation,coalescence stability is favoured by a protein film which has a high surface viscosity to slowdown the rate of drainage and a high surface elasticity to reduce susceptibility to spontaneousrupture (Dickinson, 1989a). Recently, a model for estimating the coalescence stability ofemulsion in the cream layer was proposed (Adachi et aL, 1992). Since coalescence in the creamlayer is not a diffusion process, this model considers the forces working on a particle based onthe DLVO theory and Stokes’ law.Mita et al. (1973) investigated the rate of coalescence of large polydispersed protein-coatedbenzene droplets (25-57 pm) in water using an optical microscopic technique. In this techniquemore than 1000 globules per sample were examined. Five different proteins were studied: bovineserum albumin, casein, gluten, lysozyme and ovalbuinin. A protein concentration of 1% and anoil volume fraction of 0.5 were used. There was a tendency for the mean droplet diameter toincrease, and for the standard deviation of the fitted log-normal distribution function todecrease, near the isolectric point (p1) of the proteins. Both the interfacial tension and the rateof coalescence between the globules were minimized as the p1 of the protein was approached.Salt concentrations above 0.05 M increased coalescence stability. Other groups, using lowerprotein concentrations (0.003-0.01%) have reported minimum coalescence stability at p1 ofseveral proteins (Halling, 1981). These results suggest that there may be considerable differencesin the properties and strength of protein films formed at different concentrations.According to Dickinson and Stainsby (1987), at high surface coverage, dispersions are stericallystabilized. Steric stabilization is the generic term for stabilization of colloidal dispersion bymacromolecules. At its simplest, an adsorbed macromolecule layer prevents the approach ofparticles to separations at which their mutual van der Waals attraction would be sufficient forcoagulation to occur (Dickinson and Stainsby, 1987).102The higher the surface protein load and the greater the compactness of the protein in the film,the higher the surface viscosity and elasticity and the lower the rate of coalesence (Boyd et at,1972). Based on the study of the coalescence stability of emulsion stabilized by lysozyme, bovineserum albumin and -casein, Phillips (1977 and 1981) concludes that coalescence in protein.stabilized emulsion will be reduced when the adsorbed protein layers are thick, highly hydratedand charged. He also suggested that the rheological properties of the proteins layers coveringthe droplets seem to be of secondary importance. More recent research (Tornberg andEdiriweera, 1988; Das and Kinsella,, 1990a) points out that high surface load and hence greaterfilm thickness and surface viscosity are not as important in coalescence stability. Das andKinsella (1990b) reported that for -lactoglobulin-stabilized emulsions, the more stable emulsionwas obtained at pH 9.7 where the surface load was minimum. The least coalescence-stableemulsion was obtained around the p1 where in fact a thicker adsoibed layer was observed.Unfolding of proteins at the interface may cause an increase in the number of attachments tothe interface, making a stronger film, thus maximizing coalescence stabffity (Tomberg andEdiriweera, 1988). Contrary to small surfactants, proteins adsorb to an interface with manypoints of attachment and thus there is a high energy barrier for desorption (Das and Kinsella,1990a).According to Boyd et at (1972) both surface viscosity and surface elasticity are involved incontrolling coalescence. Coalescence is caused by the force pressing adjacent globules together,giving rise to a compressive stress. With time the stress increases and ultimately exceeds acritical value, causing breakdown of the film that results in coalescence (Boyd et at, 1972).Das and Kinsefla (1990a) indicated that the interfacial tension of the oil/water system is notrelated to the coalescence stability of emulsion, while Dickinson (1992) suggested the reverse.Due to their amphiphilic nature, proteins are forcefully adsorbed to the interface between oil andwater, causing a pronounced reduction of the interfacial tension that readily facilitatesemulsification. According to Kato and Nakai (1980) more hydrophobic proteins orient themselves103more readily than less hydrophobic ones at the interface, with their polar groups directedtowards the aqueous phase and their hydrophobic groups towards the nonaqueous phase,lowering surface free energy. Hydrophobic proteins tended to decrease to a greater extent theinterfacial tension and also had higher values of EM (Kate and Nakai, 1980). These authors didnot study the effect of interfacial tension or hydrophobicity on coalesecense stability, but theysuggested that effective hydrophobicity, estimated with a fluorescence probe, plays a major rolein stabilizing the protein lipid interaction.Das and Kinsella (1990b) studied the effect of heat-induced cnformational instability of Jlactoglobulin on its adsorption behavior at the oil/water interface and coalescence stability ofemulsion droplets. A strong correlation between surface hydrophobicity and coalescence rateconstant was observed when the average volume-surface diameter (d,) of the emulsions werekept constant. These authors indicated that for -lactoglobulin, conformational instability wasmore important than film thickness in controlling droplet coalescence.The type of oil making the emulsion can also have an effect on coalescence. Das and Kinsella(1993) reported that for whey proteins stabilized milkfat-peanut oil emulsions, at a constant 0,droplet diameter and polydispersibility of droplet size tended to decrease with increasedproportion of peanut oil. Emulsions stored at 50°C showed increasing coalescence stability withincreasing peanut oil content of the dispersed phase.iv. Estimation of coalescence stabilityThe rate of coalescence is perhaps mostly affected by the emulsion (mean) droplet size.Emulsions with large droplets coalesce faster, even when they possess a thick film of adsorbedproteins (Das and Kinsella, 1990a). When the droplet sizes are large (e.g., d,> 5-7 iim), thecoalescence is virtually independent of the nature and thickness of the adsorbed film. Collisionof two bigger droplets produce great change in momentum causing an increase in probability of104film rupture. In a polydispersed emulsion, the bigger droplets coalesce faster so that the relativenumber of small droplets in the emulsion increases (Das and Kinsella, 1990a).Since coalescence causes the average size of the droplet to increase, changes in droplet size overtime or after a specific treatment are usually related to coalescence stability. Volumemeasurements of separated phases (i.e., oiling off) and changes in viscosity have also been used.Droplet size is the most important fundamental property for the characterization of emulsions.Most research work with food emulsions requires a knowledge of the droplet size distributionof the emulsions, because it influences properties such as stability, viscosity, texture andmouthfeel. Droplets ranging from 0.1 to 100 pm in diameter may occur in food emulsions (Dasand Kinsella, 1990a). Since most food emulsions are polydispersed, it is usually necessary tomeasure the particle size distribution of the emulsion.A large number of methods are available for measuring droplet size. These methods can beclassified as direct and indirect methods. The oldest method is the direct observation of thedroplets with the aid of a microscope. With the optical microscope quantitative measurementsof droplet diameters of less than 0.5 pm is not practicable Was and Kinsefla, 1990a). Electronmicroscopy has also been used, but because of difficulties in sample preparation, reproducibility,and possible artifacts it is difficult to use as a routine analysis (Das and Kinsella, 1990a).Indirect techniques measure some property of the particles that is related to their size. TheCoulter counter and light scattering techniques are the two main ways to determine the dropletsize distribution for fine 0/W emulsions. Other techniques which are sometimes used areturbidimetry and photosedimentation. The spectroturbidimetric method developed by Waistra(1965, 1968) is a popular method. In this method, light transmission through a diluted sampleof emulsion having absorbance <0.1 is measured in the range 400-800 nm using aspectrophotometer with a small angle of acceptance. From the experimental spectrum a reducedturbidity spectrum is calculated and compared to theoretical turbidity curves generatedassuming log-normal or other suitable distributions. Theoretical and experimental curves are105compared for the best match. The shift in the abcissa or x-axis gives the as well as theparameter c, which is the relatively standard deviation ofthe surface-weighted size distributionof the emulsion. Although the principles behind the method are complex, from a practical pointof view this method is simple and has been used by several researchers (e.g., Haque andKinsella, 1988; Haque and Kinsella, 1989; Singh et al, 1993). A more detailed discussion onthese techniques can be found in the following references: Dickinson (1992), Das and Kinsella(1990a) and Dickinson and Stainsby (1982).For the estimation of coalescence stability several methods have been reported in the literature.Since typical food emulsions are relatively resistant to coalescence “accelerated” tests arenormally used, in order to avoid storing the food emulsions for long periods. However, as pointedout by Das and Kinsella (1990a), rapid methods for the determination of stability are oftenunreliable. A different mechanism of instability may be involved inthese accelerated methods.This is particularly true for centrifugal and ultracentrifugal methods as well as methodsinvolving heating, freezing, and thawing (Das and Kinsella, 1990a). Waistra (1989) clearlyindicated that quick tests for coalescence stability are mostly unreliable.One of the most popular rapid methods for estimating coalesence stability is the one firstproposed by Pearse and Kinsella (1978). In this paper the exact procedure for estimating the so-called emulsion stability index (ESI) was not clearly described. In the Methods section of theirarticle they wrote: “The emulsion under test was held at constant temperature while beinggently stirred. Periodically aliquots of the emulsion were taken for emulsion and turbiditymeasurement as described above [as per EAIJ.” While in their result section, they introduced anew procedure. They show a figure (Fig. 8 in the original paper) reporting “the heat stability ofan emulsion stabilized by 0.1% SDS detergent, pH 8.0, 0.1M NaCl, 0=0.25. Heated in a boilingbath, diluted 1,2500 (ESI 3.3 h).” Assuniing that the decay in absorbance or turbidity (7) overtime (t) is first order:dT/dt=kT106They defined ESI as:ESI = TAt/ATwhere T is turbidity, AT, is the change in turbidity occurring during the time interval At. Theterm At/AT is the inverse of the rate of change of turbidity over time and since for first orderreactions this rate is rLot constant, it is multiplied by a T value to obtain the time that takes fora emulsion having a value of T, to change its turbidity by AT. If the experimental data fitsperfectly to a first ordey decay curve, ESI does not depend on the time interval chosen, howeverthis is an unlikely case for real systems. It is important to point out that the original ESI valueof Pearce and Kinsella (1976) is related to coalescence stability (Leman and Kinsella, 1989).Leman and Kinsella (1989) clarified the ESI methodology. They wrote: ‘The emulsion was heldat constant temperature (approximately 25°C) while being gently stirred. The emulsion was thendiluted serially with water and sodium dodecyl sulfate (SDS) solution to give final dilutionbetween 1000 and 5000 and an SDS concentration of 0.1%. The SDS stabilized emulsion wasthen heated to 98°C and the absorbance of the emulsion determined at 500 nm”. A question thatarises with this methodology, is why was the SDS stabilized emulsion heated?. Apparently theemulsion was heated to accelerate coalescence. As mentioned before, there is the posibility thatSDS displaces the protein from the interface, thus changing its original properties. Possibly,because the technique is ill defined, the original methodology of ESI is seldomly used.Kato et al. (1981) and Voutsinas (1982) modified the original ESI methodology as follows: theemulsions are held at room temperature and at different time intervals, aliquots are takendirectly from the bottom of the container and diluted with SDS and absorbance at 500 nm isrecorded. The half-life (mm) of the absorbance decay with time, determined graphically, is usedas ESI. Essentially, this methodology has been used by several investigators from both ourresearch group (e.g., Arteaga et al., 1993) and from other groups (Kate et al., 1985a; Paulson andTung, 1988; Mine et al., 1991).107Contrary to the original ESI concept, the process of instability being measured with the abovemodification is creaming and not coalescence. From a theoretical basis creaming is not asimportant as coalescence (Dickinson, 1992; Das and Kinsella, 1990a). In addition creaming isusually reversible. Since aliquots are taken from the bottom of the container, changes inabsorbance or turbidity are related to changes in the vertical profile or concentration gradientof the dispersed phase (i.e., creaming). This creaming phenomenon causes the 0 in the aliquotanalysed to be different than that found in the original emulsion, causing errors in theestimation of the intefacial area if the the initial 0 is used in the calculations. Although thismodified methodology is widely used, few researchers are aware of these facts, and a few of themhave erroneously suggested that coalescence stability is being measured as in the originalmethodology.1084. Some physicochemical properties of egg-white and milk proteinsa. IntroductionThe primary importance of proteins in the diet is as a source of amino acids. Of thetwenty-two amino acids most commonly occurring in proteins, eight are known to be essentialto humans; that is they must be supplied in the diet to maintain growth and health. Proteinsfrom some sources, especially non animal proteins, are deficient in certain amino acids (e.g.,cereal proteins are deficient in lysine) and must be combined with other proteins to provide anadequate balance. Of the many proteins available in the human diet, the nutritional value ofmost animal derived proteins is usually the highest (Pomeranz, 1991). Apart from nutritionalconsideration, proteins play a large part in the organoleptic properties ofmany foods (tee sectionProtein Functionality). In this regard, animal proteins usually have better functionality thanvegetable proteins. However, the production ofanimal proteins is less energy efficient than plantproteins and in some regions of the world, especially developing countries, more difficult thanplant proteins. Pimentel (1984) indicated that anywhere from 7 to 88 kcal of fossil energy arerequired to produce 1 kcal of animal protein, while the production of protein in soybeans andcorn grain required only 0.7 to 3 kcal of fossil energy per 1 kcal of plant protein.In this thesis animal proteins from two sources were studied: egg white and milk proteins. Inthe following sections a description and comparison of some of the physicochemical andfunctional properties of these proteins is given.b. Radar charts of the amino acid composition of milk and egg white proteinsAnalysis of the amino acid composition of proteins can give some insight on thephysicocheniical properties and functional attributes of proteins. The amino acid composition ofa protein can be viewed as a point in an 18 dimensional space of 18 axes representing the109content of amino acids (Nishikawa and Ooi, 1982). Saito et al. (1990) proposed a simple andclear graphical approach to depict the amino acid composition of a protein in relation to an“average protein”. The amino acid composition of the “average protein” was taken as the averageamino acid composition calculated using a total of 356 proteins as reported by Nishikawa andOoi (1982). In this methodology the amino acid composition of a protein expressed in molarpercent is normalized by the mean value and the standard deviation (SD) (i.e., Z normalization)obtained from the data of 356 proteins. A radar plot is constructed where the scale of the axisrepresenting the contents of individual amino acids is graduated from -3 SD (center of the circle)to +3 SD (outer circle). The inner circle indicates the mean value representing the “averageprotein”. This methodology was used in this thesis and the resulting radar plots are shown inFig. 1-3. The amino acid composition of the proteins used in this work and of the “averageprotein” are reported in Table 4.-c. Milk proteinsi. The caseinsAs for many other foods, cow’s milk is made of water, fat, proteins, sugars and minerals.Some components of milk, including milk fat, lactose and most proteins are specificallycharacteristic of synthesis in the mammary gland. Other constituents, such as vitamins, salts,are found in many other biological sources. The composition of milk reflects the fact that it isthe sole source of food for very young mammals (Swaisgood, 1985). Milk is roughly 88% waterand contains 30-35 g of protein per liter. Milk proteins have been traditionally classffied in threecategories: caseins, whey proteins, and proteins associated with the lipid phase (Pomeranz,1991).110Caseins and whey proteins exhibit genetic polymorphism since they are products of co-dominant,allelic and autosomal genes. In addition, posttranslational modification is also common in thecaseins (Swaisgood, 1985; Eigel et al., 1984).The caseins are defined as those proteins that precipitate from raw skim milk by acidificationto pH 4.6 at 20°C (Eigel et al., 1984). 4lthough this definition is operational, the differentproteins forming this fraction have similar properties. Caseins were originally classified intoseveral families on the basis of electrophoretic mobility. The three main caseins families are: afl-,a,2- f- and ic-casein. Classffication based on their primary sequence is nw recommended (Eigelet al., 1984).Caseins comprise the largest fraction ofbovine milk; approximately 80% of the total milk protein(Swaisgood, 1985). Caseins in milk are not present in a solution state, but in a colloidal stateforming so-called casein inicelles. The casein micelle is a highly hydrated spherical cmplex ofproteins with calcium phosphate (Swaisgood, 1985). The primary function of this complex is totransport calcium and phosphate to the neonate in a soluble form. It also forms a coagulum inthe stomach ofthe young, allowing the slow release ofnutrients down the digestive tract (Farrelland Thompson, 1988; Holt, 1992). Holt and Sawyer (1988) indicated that casein micelles preventthe potential catastrophic precipitation of calcium phosphate within the golgi vesicles of themammary secretory cell. It has also been suggested that caseins are organized in micelles inorder to have milk with a reduced viscosity, allowing milk to be withdrawn, rather quickly, fromthe mammary gland by sucking (Holt and Sawyer, 1988). Holt (1992) wrote a recent review onthe structure and stability of bovine casein micelles.The a, caseins are all phosphoproteins that are precipitated by low concentrations ofCa2exceptin the presence of ic-casein.a1-Casein is by far the major protein in these fractions. Human milkdoes not contain a,-caseins. At present, the five genetic variants are designated A, B, C, D, andE. The B variant is predominant in Bos taurus (Eigel et al., 1984). The most common variantconsists of 199 amino acids with 8 phosphorylated Ser, with a calculated mw of 23,614111KT:LHYFW H FWX81—casejn f3—caseinKTDLHYFWic—caseinFig. 1. Radar chart of the amino acid composition of caseins. The amino acid composition of theproteins were normalized by the mean value and the standard deviation (SD) of 356 proteinsreported by Nishilcawa and Ooi (1982). The scale of the axis, representing the contents ofindividual amino acids, was graduated from -3 SD (center) to +3 SD (outer circle). Inner circleindicates the mean value. Amino acid standard one letter codes were used, except for E=Glu+Glnand D=Asp+Asn.112K TR A KTHYFWc—1acta1bumin fr-lactoglobulinKTIIYFWSerum albuminFig. 2. Radar chart of the amino acid composition of whey proteins. The amino acid compositionof the proteins were normalized by the mean value and the standard deviation (SD) of 356proteins reported by Nishikawa and Ooi (1982). The scale of the axis, representing the contentsof individual amino acids, was graduated from -3 SD (center) to +3 SD (outer circle). Inner circleindicates the mean value. Amino acid standard one letter codes were used, except for E=Glu-i-Glnand D=Asp+Asn.113K TYFWOvalbumin LysozymeKT KTM,;HYFWOvomucoid OvotransferrinFig. 3. Radar chart of the amino acid composition of egg proteins. The amino acid composition ofthe proteins were normalized by the mean value and the standard deviation (SD) of 356 proteinsreported by Nishikawa and Ooi (1982). The scale of the axis, representing the contents ofindividual amino acids, was graduated from -3 SD (center) to +3 SD (outer circle). Inner circleindicates the mean value. Amino acid standard one letter codes were used, except for E=GIu+Glnand D=Asp+Asn.114Table4.AminoAcidcompositionof themodelproteins(numberof residuespermolecuie).Amino(1Y(2)A(3)A(4)A(5)A(6)A(7)5(8)c(9)C(10)°PRO5SD5acidlIe111013810142562735.02.0Leu1722813226132850137.02.8Val1119116103630649158.82.2Ala9571531446341253118.83.5Gly952631519125158.12.8Met5621441321121.91.1Cys00285356830161.41.1Phe89444272032653.81.6Trp211422361101.10.8Tyr10494419103263.31.5Pro17352028281522874.71.9Ser1616137728371047126.82.7Thr5914783415737145.92.2Asp744911413221783010.53.0Glu24181381659525671410.63.5Lys1411912155919658136.53.1Asn857125130000-‘GIn1521145920000-‘-‘i’Arg645132315112964.42.1His5533217811242.11.2Proteincodes:(1)a81-Casein,(2)f3-Casein,(3)ic-Casein,(4)a-Lactalbumin,(5)fI-Lactoglobulin, (6)Bovineserumalbumin,(7)Ovalbumin, (8)Lysozyme,(9)9votransferrin, (10)Ovomucoid.References:AEigel etal. (1984).5Nisbet etal. (1981).CLi..ChanandNakai(1989).DJeltschandChambon(1982).EpRoHAverageltproteinofMishikawaandOoi(1982).Aminoacidcompositionandstandaddeviation(SD)inunitsofmole%.FForthishypotheticalproteinGlu=Glu+GlnandAsp=Asp÷Asn.C)I(Swaisgood, 1973; Eigel et al., 1984). Bigelow’s average hydrophobicity for this casein (1,170cal/res) is in the high range of that observed for most globular proteins. A total of 32% of theresidues are hydrophobic. Based on the classification of the Chemical Abstract Service (1990)lie, Leu, Met, Val, Phe, Trp and Tyr were considered hydrophobic. Similarly to -casein, a,casein does not contain any Cys residues. (Swaisgood, 1973; Eigel et al., 1984).Analysis of the primary sequence suggests three hydrophobic areas: 1-44, 90-113, and 132-199.Correspondingly, charged and polar groups are concentrated in two areas: 45-89 and 114-131.The polar segment 45-89 is of special importance since all of the phosphoseryl residues arelocated in this area. The segment 43-79 is highly acidic, containing twelve side-chain carboxylgroups from Glu and Asp, two c-amino groups, and seven phosphoseryl residues; the net chargeat pH 6.6 is -22.7, while the rest of the molecule has a net charge of only 1 or 2.;1-Caseincontains 17 Pro residues spaced rather evenly in the two largest hdrophobic areas. 1’he largenumber of Pro is responsible in part of the lack of secondary structure in this protein. Themolecule is most likely amphiphilic, with as many as possible of the hydrophobic residues buriedin a globule, while the acidic segment, being highly solvated, would occupy an unusually largehydrodynamic volume (Swaisgood, 1975).Seven genetic variants of -Casein, also a Ca2 sensitive phosphorylated protein, are known. Thepredominant polymorph in species ofBos is A2 with 5 phosphorylated serines (Eigel et al., 1984).This fi-casein consists of a single polypeptide chain containing 209 residues, with a calculatedmw of 23,983 (Eigel et aL, 1984). A total of 34% residues are classified as hydrophobic. Thiscasein has the highest value of calculated Bigelow’s average hydrophobicity value (1,335 cal/res)ofthe caseins. Roughly, two thirds ofthe C-terminal portion is extremely hydrophobic, especiallythe regions 80-90, 110-120, 133-140 and 200-209. A large concentration of negative charges islocated in the N-terminal region (1-21). The net charge of this portion, which contains all thephosphate groups, is -12 at pH 6.6, while the remainder of the chain has a net charge of116essentially zero. A total of 35 Pro residues are present in this casein. Pro occurs five times inpairs and other times in sequence of alternating residues, and in very hydrophobic regions(Swaisgood, 1975; Eigel et al., 1984).3-Casein presents a strong temperature-dependent association; only monomers are present atlow temperature (4°C) but at room temperature large polymers are formed. Spectroscopicmeasurements suggests a transfer of residues from polar to a less polar environment uponpolymer formation. An intricate balance between hydrophobic and electrostatic interactions ismost likely responsible for this association (Swaisgood, 1975).The major component of human casein micelles is f-casein containing from 0 to up to 5phosphoserine residues per molecule. For both bovine and human -caseins, hydrophobicinteractions appear to be dependent upon the conformational state of the molecule (Slattery etaL, 1989). -ic-Caseins are soluble at all concentration of Ca2, except at extreme levels where gelation andsalting out occur (Swaisgood, 1975). This casein family consists of a major carbohydrate-freecomponent and a minimum of six minor glycoproteins (Eigel et al., 1984). Two genetic variants,designated A and B are known. The A variant with one phosphorylated Ser tends to bepredominant in most breeds. ic-Casein with two phosphate groups has also been reported (Eigelet aL, 1984). The primary structure of the major carbohydrate-free component consists of 169amino acids with a calculated molecular weight of 19,007. A total of 34% residues are classifiedas hydrophobic. Its calculated Bigelow’s average hydrophobicity (1,223 cal/res) is between thoseof cx- and -casein (Kinsella and Whitehead, 1989). This casein contains 20 Pro and one Cysresidue. The bond sensitive to chymosin (rennin) occurs between Phe15 and Met. Thehydrolytic products are para-ic-casein (residues 1-105) and macropeptide (residues 106-169)(Swasigood, 1985). The macropeptide contains all of the carbohydrate, linked through hydroxyloxygen of threonyl residues and the only phosphoseryl residue (Swaisgood, 1985). The netnegative charge of the macropeptide is twice that of ic-casein. This acidic charge is not due to117organic phosphate groups since only one or two are present. The macropeptide contains anunusually high frequency of Ser and Thr residues (Eigel et aL, 1984).ic-Casein displays a strong tendency to interact with a11- and -casein except in solutions withdissociating agents such as urea. When isolated, this casein undergoes self-association topolymer size in the order of 650,000; this phenomenon is insensitive to temperature, ionicstrength or concentration effects (Swaisgood, 1975). This association has been suggested to besimilar to the formation of micelles by small detergents (Swaisgood, 1984). Swaisgood (1975)suggests that ic-casein appears to have a fairly rigid globular structure for about 2/3 of themolecule, to which a flexible, highly solvated tail is attached.Recently, molecular modeling, sequence-based secondary structure prediction techniques andFourier transform and Raman spectroscopy data were used to build unrefined three dimensionalstructures in vacuo of the milk caseins (Kumosinski et al., 1991; Farrrell et al., 1993).ii. Whey proteinsWhey proteins can be defined as those proteins remaining soluble at pH 4.6 and 20°Cafter casein removal from skim milk or whole milk (Bottomley et al., 1990). Whey proteinsaccount for 20% (5-7 g/l) of the total protein in cows milk, whereas in human milk, whey proteins(i.e., noncasein proteins) quantitatively exceed the caseins (McKenzie and White, 1991). Themajor whey proteins in cow’s milk, in order of decreasing concentration are f3-lactoglobulin (2-4gil), a-lactalbumin (1-1.5 gIl), the proteose-peptone fraction (0.6-1.8 g/l), immunoglobulins (0.6-1.0g/l) and serum albumin (0.1-0.4 g/l) (Swaisgood, 1985). Discussion will be limited to alactalbumin, f3-lactoglobulin and serum albumin, which were the proteins used in this work.Contrary to the caseins, whey proteins are more compact and globular, having disulfide bondswhich confer a certain degree of structural constraint which imparts stability. Furthermore,when compared with the caseins, whey proteins are more heat-sensitive, less calcium-sensitive,118and can engage in thiol-disulfide interchange to form oligomeric structures (Kinsella andWhitehead, 1989).a-Lactalbumin accounts for approximately 25% of the whey proteins. Three genetic variants A,B and C are known. Only the B variant has been observed in milk from western cattle. Theprimary sequence of a-lactalbumin B consists of 123 amino acids with a calculated mw of 14,175.The proportion ofhydrophobic amino acids forming this protein is 26%. Compared to the caseins,this protein has a lower value of calculated Bigelow’s hydrophobicity (1,118 cal/res) (Kinsella etal., 1989). The protein has 8 Cys residues all of them involved in disulfide bonds (Eigel et aL,1984). Results from different spectroscopic techniques suggest that two of the four tryptophanre3idues (possibly Trp 104 and Trp 118) are exposed to solvent at neutral pH and 25°C(Whitney, 1988). The protein has a marked propensity for self-association at low pH (pH 3.5 to4), and to a lesser but significant degree at neutral pH (Kromman, 1989).With the onset of lactation, a-lactalbumin is formed in the mammary gland and alters thesubstrate specificity of galactosyltransferase from N-acetylglucosamine to glucose, enablinglactose synthesis to be effected. Thus, a-lactalbumin is a “specifier” protein (McKenzie andWhite, 1991).a-Lactalbumin has a very stable conformation between pH 5.4 and pH 9.0. It is the most stableof the principal whey proteins (Bottomley et al., 1990). It has been found to possess a strong Ca2binding site, and much of the recent research on the protein has been concerned with itscalcium-binding properties and the large conformational changes associated with metal binding(McKenzie and White, 1991). The metal-ion binding properties of this protein in relation to itsconformational properties was reviewed by Kronman (1989). Binding of small lipophiliccompounds (e.g., ANS) has been shown to afford protection against pepsin cleavage. This effectmay have physiological significance in nutritional transport (Hirai et al., 1992). cz-Lactalbuniinpresents a clear homology in sequence and conformation with hen egg-white lysozyme and a119great number of studies have been carried out in this area. For a recent review see McKenzieand White (1991). Experimental observations suggest that the structure of a-lactalbumin is lessstable than that of lysozyrne.-Lactoglobulin, the major whey protein, is the most extensively characterized and best describedof all food proteins (Kinsella and Whitehead, 1989). Seven genetic variants of f3-lactoglobulin areknown. The A variant is the most common variant in bovine milk (Eigel et al., 1984). Thisprotein consists of 162 amino acids with a calculated mw of 18,277. Of the total residues 33%are hydrophobic. -Lactoglobulin is especially rich in Leu, containing 20 residues. Its calculatedBigelow’s hydrophobicity (1,202 callres) is the highest of the whey proteins (Kinsella et al., 1989).According to X-ray crystallography results, the molecule consists ofantiparallel f3-sheets, formedby nine strands wrapped around to form a flattened cone. The core of the protein fold is aneight-stranded, cross-hatched 3-barrel composed of apolar amino acid side chains lining theinternal cavity. This hydrophobic core binds apolar ligands (e.g., retinol) and there is a Trpresidue associated with binding (Kinsella and Whitehead, 1989; Bottomley et al., 1990). Basedon this binding of apolar substances by f3-lactoglobulin, it has been suggested that a possiblebiological function of this protein is vitamin A transfer from maternal milk to the neonate(Kinsella and Whitehead, 1989).Five Cys residues are present in this protein, four of which are involved in disulfide bonds. Theone free thiol group appears to be distributed equally between positions 119 and 121, its locationaffecting the position of one -SS- bridge (Eigel et al., 1984). The presence of this free Cysprovides a potential for inter- and intramolecular disulphide link interchange duringconformational changes associated with pH alterations or heat treatment. Due to electrostaticinteractions, the protein tends to exists as a dimer in solution (Kinsella and Whitehead, 1989).120Serum albumin prepared from milk is physically and immunologically identical to blood serumalbumin, the major protein ofplasma. The protein prepared from blood plasma has been studiedextensively, and is one of the “classical” proteins in protein chemistry since it is readily availableand relatively easy to purifr (see the review article by Peters (1985) for more detailedinformation). Serum albumin consists of 582 residues and has a calculated mw of 66,267. Theproportion of hydrophobic amino acids is 28%. This protein has the lowest calculated Bigelow’shydrophobicity (989 cal/res) (Kinsella et al., 1989) and is the most acidic protein (isoelectric point4.2-4.5) of the milk proteins. Overall, this protein has low content of Trp and Met and a highcontent of Cys and charged amino acids (Asp, Glu, Lys and Arg). Gly and lie are lower than inthe average protein.The protein has 17 intramolecular disulfide bonds and only one free-SH. The disufide bondsappear to be well shielded from the solvent. None is accessible toreducing agents rn the pHrange 5-7, but they become progressively available as the pH is raised or lowered. Of the twoTrp residues in bovine albumin, one is apparently more exposed (Peters, 1985). The molecule isvisualized as having three major domains, each having an ellipsoid shape consisting of two largedouble ioops and a small double loop. The domains are dissimilar in hydrophobicity, net chargeand ligand binding sites (Eigel et aL, 1984). This loop structure gives the molecule considerableflexibility in spite of its high content of disulphide bonds (Peters, 1985). The high total charge,about 185 ions per molecule at pH 7, aids in its solubility, and the many disulfide bonds, afeature of most extracellular proteins, contributes to its stability. The overall net charge at pH7 is in the range -12 to -18 (Peters, 1985). Serum albumin has two biological functions: atransport function for fatty acids and many other substances and, in plasma, it helps toregulated blood osmotic properties (Bottomley et al., 1990; Kinsella and Whitehead., 1989). Themaximum number of fatty acids which can be bound to albumin in vitro is about six, however,normally in vivo one to two are carried. Limited proteolysis experiments suggests that theprimary fatty acid binding site is the fragment TA (377-582) (Peters, 1985).121d. Eg white proteinsi. IntroductionThe avian egg is the perfect microenvironment for the development of an embryo. It isan axenic (i.e., sterile) environment which requires limited interactions with the extrinsicenvironment (Board and Tranter, 1990). In the egg, the embryo is protected by the egg shell andalso by the egg white or albumen. The protection given by this latter system is especiallyimportant before and during th first few days of incubation when it provides both a physical(i.e., high viscosity and maintaining the yolk in a central location of the egg) and a chemical (i.e.,high pH and presence of antimicrobial proteins) defense against microbial infection (Board andTranter, 1990).The albumen is also a food storage system for the developing embryo. Rej*esentingapproximately 60% of the total egg weight. It occurs in four layers in most chicken eggs (Boardand Tranter, 1990). Water is the major constituent of albumen (=88%). The protein, lipid,carbohydrate (both free and forming glycoprotein) and ash content of albumen are as folliows:9.7-10.6%, 0.03%, 0.4-0.9% and 0.5-0.6%, respectively (Powrie and Nakai, 1985).Albumen may be regarded as a protein system consisting of ovomucin fibers in an aqueoussolution of numerous globular proteins (Powrie and Nakai, 1985). Albumen contains more than15 proteins (Li-Chan and Nakai, 1989). In the following section, only information regarding thefour egg-white proteins used in this thesis will be given.ii. The four main egg white proteinsOvalbumin is the most abundant egg white protein, making up more than half (= 54%) of theegg white proteins by weight. Three proteins, differing only on their phosphorus content, arefound in purified ovalbumin; fraction A1 having two phosphates per molecule, A2 one and A3without phosphates. The proportions of these fractions are about 85%, 12%, 3% for A1, A2 and122A3, respectively (Froning, 1988). Ovalbumin is a monomeric phosphoglycoprotein made of 385residues with a molecular weight of 44,500, with 35% of its residues being hydrophobic. Of themajor egg proteins, ovalbumin is the one that has the highest value of Bigelow’s hydrophobicity(1,110 kcal/res) (Powrie and Nakai, 1986). Ovalbumin contains four sulfhydryl groups and onedisulfide bond. Three weakly reactive SH groups are present in native ovalbumin (Froning,1988). The single carbohydrate moiety, made of a core of two N-acetylglucosamine and fourmannose units, is attached to an asparagine residue (Asn-292) (Conchie and Strachan, 1978).Based on the primary sequence of ovalbumin reported by Nisbet et al. (1981) the followingobservations can be made: the N-terminus of the protein is blocked by an acetyl group. The C-terminal portion is a Cys-Val-Ser-Pro-COOH. The fragment of the C-terminal (365-385) isrelatively hydrophobic containing 2 Cys residues. A fragment including the N-terminal region(1-20) is also hydrophobic and contains 1 Cys residue. The heptape5tide, Glu-A1a-GlyVal-Asp-Ala-Ala, which is liberated from ovalbumin by subtilisin, under specific conditions, with theconcomitant formation of plakalbumin, is made of residues 346-352. Since this peptide isobtained from partial proteolysis of native ovalbumin, it is probably highly exposed.Phosphoserine groups of ovalbumin have a high degree of mobility and appear to be on thesurface of the protein, an observation that has been found true for other phosphorylated proteins(Holt and Sawyer, 1988). The crystal structure of ovalbumin was reported recently (Stein et al.,1991). Doi et al. (1987) using CD observed 49% helix, 13% f’-sheet, and 24% unordered structurein native ovalbumin.Hunt and Dayhoff (1980) did a computer-aided comparison of ovalbumin with all proteinsequences available at that time. This analysis revealed an unexpected and definite relationshipbetween the sequence of ovalbumin and that of two protease inhibitors: human antithronibin-IHand human plasma1-proteinase inhibitor. They proposed that all three proteins be grouped ina protein superfamily termed ovalbumin-antithrombin superfamily. It is important to point outthat although ovalbumin and the two protein inhibitors possess some of the same structural123modifications (e.g., the three are glycoproteins) their positions do not usually correspond whenthe sequences are aligned.Ovotransferrin, also called conalbumin, belongs to the transferrin protein family, which is agroup of homologous iron-binding glycoproteins (mw ‘80,000) which are widely distributed invarious biological fluids (e.g., lactoferrin in milk) (Williams et al., 1982). Due to it,s strong ironbinding properties (binding constant near 1020), this protein is one of the anti-bacterial ageuiin egg white (Li-Chan and Nakai, 1989). On a dry basis it makes up, on average, 12% of thealbumen (Powrie and Nakai, 1985). It was named conalbumin because it precipitated withovalbuinin during purification of this protein (Feeney and Osuga, 1988). The ovotransferrinmRNA codes for a 705 amino acid preprotein with a 19 amino acid signal protein, thus theprimary sequence of the protein consists of 686 amino acids (JeThsch and Chambó’n, 1982).Williams and coworkers (1982) reported a calculated mw of 77,770, a figure which includes 1900for the glycan.Transferrin molecules are divided into two lobes of similar size which show marked sequencehomology. A total of 15 disulfide bonds are found in ovotransferrin and are present in a similarposition in human sero- and lactotransferrin, also some residues which may be involved in iron-binding (e.g., tyrosine and histidine) are located in conservative positions (Wiliams et al., 1982).Penner et al. (1983) using a periodate oxidation treatment reported that 4 to 5 Tyr are essentialfor iron binding activity and that Met residues are not involved in this binding. They alsoreported that periodate oxidation in the presence of 8 M urea oxidized the 11 Met, while in theabsence of urea only 5 Met were modified. The calculated Bigelow’s hydrophobicity forovotransferrin is 1080 cal/residue (Powrie and Nakai, 1985). Athough this proteins contains arelatively large number of aromatic residues, only 28% of its amino acids are hydrophobic.As compared with the uncomplexed protein, the iron complex of ovotransferrin is much morestable to thermal denaturation, as well as to physical and chemical treatments. This fact has124been exploited to protect the heat-sensitive egg white proteins during pasteurization (Froning,1988). Ovotransferrin is also more heat sensitive but less susceptible to surface denaturationthan ovalbumin. Ovotransferrin is more heat stable at pH 9 than at neutral pH (Powrie andNakai, 1986).Ovomucoids are heat-resistant glycoproteins containing 20-25% carbohydrate, and exhibitingantitrypsin and or antichymotrypsin activity. Chicken ovomucoid (c-ovomucoid) makes up 12%of the dry matter of albumen. The primary sequence of c-ovomucoid consists of 185 amino acidswith a mw of 28,000 (Froning, 1988). Ovomucoids are present in eggs of many different avianspecies, and are usually devoid of Trp and are high in disulfide bonds (9 in the the case ofc-ovomucoid). Chicken ovomucoid has the lowest value for Bigelow’s hydrophobicity (920cal/residue) of all the egg white proteins (Powrie and Nakai, 1985). This value is due o the lowproportion (23%) of hydrophobic residues forming the protein. Ovomucoids are formed of threedomains, capable of forming complexes with trypsin-like or chymotrypsin-like enzymes. In thecase of c-ovomucoid, the only strong binding site is for trypsin (Feeney and Osuga, 1988). Dueto its small size, readily measurable biological activity (both in vivo and in vitro), facilepurification and presence in many different species extensive research has been carried out inthese proteins. In a review article by Laskowski et al. (1987) a little more than 100 species werediscussed. Feeney and Osuga (1988) also reviewed some recent research on ovomucoids.Lysozyme is a ubiquitous enzyme, which has the ability to lyse certain bacteria by hydrolyzingthe -1,4-linkage between muramic acid and N-acetylglucosamine ofmucopolysaccharides in thebacterial cell wall (Li-Chan and Nakai, 1989). Lysozyme occurs in domestic hen egg-white to theextent of —3.4% dry basis (Powrie and Nakai, 1985). It is the most extensively studied lysozymeand is representative of a class of lysozyme, designated c-type lysozyme (c for chicken)125(McKenzie and White, 1991). Interestingly, all reported mammalian lysozymes have proved tobe of the c type (McKenzie and White, 1991).Hen egg white lysozyme, from now on termed lysozyme, is a relatively small and stable enzymemolecule. It consists of 129 amino acids with a mw of 14,300. It is a very basic protein, with anisolectric point of 10.7 (Li-Chan and Nakai, 1989). Lysozyme has four disulfide bridges and isrich in Arg (11 residues). For lysozyme the sum Asp+Glu (7+2) is substantially less thanLys-i-Arg (6+11), leading to a high isoelectric point (McKenzie and White, 1991).Its reported primary sequence shows 26% of its residues being hydrophobic. Many of thesehydrophobic residues are exposed to solvent, for example Val-2, Phe-3, Leu-17, Phe-34, Leu-75,Trp-123, Pro-70, and Pro-79, with Trp-62, Trp-63, fle-98, Trp-108, and Val-109 being on thesurface of the cleft which divides the molecule in two lobes (McKenzie and White, 1991). As forovomucoid, Bigelow’s hydrophobicity value is low (970 cal/residue) (Vowrie and Nakai, 1985). Forthis protein, hydrophobic and hydrophilic side chains tend to occur in large clusters rather thanbeing randomly distributed throughout the polypeptide. A good proportion ofthe first 39 residuesof the N-terminal region are hydrophobic, the next 40 residues are mainly hydrophilic, and thethird terminal of the sequence is once again hydrophobic. The determination of the three-dimensional structure of lysozyme was the first elucidation of the X-ray crystal structure of anenzyme (Blake et al., 1965).Its basic nature has been applied in the development of procedures for isolation; it has also beenimplicated with electrostatic interactions with negatively charged residues. These interactionsmay have considerable, practical importance on properties ofegg in storage (Li-Chan and Nakai,1989).1265. Emulsifying properties of milk and egg white proteinsa. Milk Proteinsi. IntroductionMilk proteins are used in manufactured foods either for their functionality or for theirnutritional properties (i.e, for fortification) (Mulvihill, 1991; Evans, 1982). Milk proteins are oneof the most common protein additives in processed foods (Kinsella and Whitehead, 1989).Currently, a large selection of high-protein dairy powders, e.g., caseinates, coprecipitates, andwhey powders are commercialy available. These display a highly variable range of functionalproperties because of differences in composition and processing treatments.In general, both commercially available caseins and purified caseins exhibit ‘‘ery goodemulsifying properties. oii in water emulsion stabilization is a property which the soluble Na-and K-caseinates possess by virtue of their flexible and amphiphilic nature. Casein stabilizedemulsions are heat resistant and also tolerant to high salt concentrations and high pH (Evans,1982).ii. The caseinsAlthough the main casein fractions (a-, f- and ic-casein) present some structural similarities(see section on physicochemical properties of milk proteins), they present different emulsifyingproperties. Compared to whey proteins, fewer studies have been published comparing theemulsifying properties of purified caseins.According to Mulvihill (1991), sodium caseinate is a more effective interfacial tension depressorthan whey proteins, blood plasma, gelatin or soy protein. It diffuses more quickly to an interfaceand on reaching the interface adsorbs more quickly than the other proteins, probably becauseof direct and rapid anchoring of freely available hydrophobic segments.[27Using three commercial protein samples, Tornberg and Hermansson (1977) reported that inemulsions made with 2.5% protein and 40% oil, the creaming stability values for the proteinsfell in the following order: soy protein isolate > whey protein concentrate> casein.Mitchell et al. (1970) compared the spread and adsorption behaviour of several milk proteins.They observed that f3-casein presented a more rapid adsorption to the air-buffer interface thana- and ic-casein, this latter protein presented the slowest adsorption rate. Of the three caseinsat a constant concentration, 3-casein produced the luwast interfacial tension, followed by ;- andic-casein. Disulfide bonds of ic-casein appear to constrain the complete unfolding of this proteinsince upon reduction of these bonds ic-casein had a interfacial behavior similar to f3-casein. Thegreater efficiency of adsorption and surface activity of-casein compared to the other caseins notonly reflects the rate of diffusion of the native protein to the interface, but also its molecularflexibility, charge, orientation, unfolding and facility for molecular packing in the film (Lemanand Kinsella, 1989).The order of surface activity reported for the individual milk proteins is f-casein > serumalbumin> a-lactalbumin > a,-casein ic-casein> f3-lactoglobulin (Mulvihill, 1991).For f3-casein, monolayer coverage forming a thick film of tightly packed folded molecules (>7.7A2 per residue) is complete at a concentration of 0.1 mg/100 ml. Further adsorption is retardedbecause of charge repulsion and steric factors (i.e., folding and looping of the f-casein becomespronounced) until substrate concentration reaches 10 mgIlOO ml when multilayer formationoccurs. Since ic-casein is more resistant to complete unfolding in the interface, ic-casein filmscontain more protein and have a higher resistance to shear deformation and compressibilitycompared to 3-casein films (Leman and Kinsella, 1989). Dickinson (1989b) summarized somerecent studies regarding the surface and emulsifring properties of caseins.The study of peptides obtained from partial hydrolysis of caseins has been recently an activearea of research. Shimizu et al. (1984) reported that removing a hydrophobic peptidecorresponding to the 23 N-terminal region of cç1-casein did not affect, at high protein128concentrations, the EM of the protein. At neutral pH, this peptide was found to have an EMsimilar to thea1-casein (emulsion made with 2% protein and 20% oil concentration). Removalof this hydrophobic peptide fromcç1-casein decreased the amount of protein adsorbed onto theoil globule surface by almost half (9.76% versus 5.53%). This peptide was found to adsorb ingreat amount to the interface (72.5% of bulk concentration adsorbed). In a subsequent paper,Shimizu and coworkers (1986) used HPLC to remove small amounts of other peptides from theiroriginal 23 N-terminai a, peptide (AC-23N). When these contaminants were removed, the EM,at neutral pH, of the AC-23N purified peptide decreased to almost 0. Since at neutral pH thenet charge of the peptide is low, electrostatic effects may be involved. This study suggests thateven for simple peptides, as for proteins, emulsi1,ing properties are very complex, being affectedby coexisting substances, emulsification conditions and other factors. In another similar study,Shimizu (1983) reported that the adsorption ofa1-casein to the oil globule surface is ery tight;even in the presence of 5 M urea only 50% of the bound protein could be extracted. Theseauthors also reported that removal of the 23-N terminal peptide decreased the strength of theadsorption and the creaming stability of emulsions.Lee et al. (1987) compared the emulsiIying properties ofa hydrophobic and a hydrophilic peptideobtained from the enzymatic hydrolysis of 3-casein. Both peptides had very low EM values atpH 7, while at acidic or alkaline conditions high values were observed. No adequate explanationwas given by Lee and coworkers regarding this pH effect. Thick adsorbing films observed byelectron microscopy suggest that a high degree of peptide association occurred on the oil dropletsurface.Reductive alkylation of whole or ofa tryptic hydrolyzate of f3-casein was reported to increase theEM compared to a nonalkylated sample (Touati et al., 1990). Chobert et al. (1987) studied theeffect on solubility and emulsifying properties of the covalent attachment of polymethionyl andpolyvalyl chains (2 to 6 residues), through isopeptide bonds to lysyl residues on whole casein.The emulsifying activity of these chemically modified caseins could not exclusively be explained129in terms of solubility and hydrophobicity. This suggested that the balance between hydrophilicand lipophilic forces played an important role in emulsifying properties of these caseins.The solubility and emulsifying properties of i-casein and its caseinomacropeptide (CMP) werestudied by Chobert et al. (1989b). The hydrophilic peptide showed similar emulsifying activitybut less emulsion stability as compared to ic-casein. Similar results were also obtained when f3-casein was treated with trypsin; at neutral pH all emulsions made with hydrolyzates (0.1%protein concentration, 0=0.25) were found to be less stable than those of native 3-casein(Chobert et al., 1989a).Murphy and Fox (1991) reported some functional properties of a f-casein enriched (BCE) andan (x-/1c-casein enriched (AKE) fraction obtained from ultrafiltration of sodium caseinate. Interms of emulsifying properties, the BCE fraction was more surface active than Na-caseinate.The AKE fraction was the least surface active. Though forming larger fat globules at comparableemulsification power inputs, both the AKE fraction and sodium caseinate had higher emulsionstability ratings than the BCE fraction. These authors suggested that the nature of protein-protein interactions around the fat globules affect the emulsion stability (creaming). Theincreased ic-casein content of the most stable emulsions could lead to the formation of a morecohesive, stable film around the fat globule, through intermolecular clisuiphide bonds orincreased hydrophobic association with other casein constituents. An important observation fromthis study is that, for these samples, emulsion stability did not correlate with the interfacialbehavior of the proteins. Mulvihill and Fox (1989) reported similar results for the purifiedcaseins. Lorenzen and Reimerdes (1992) reported that1-casein stabilized emulsions presenta slower rate of creaming than f3-casein stabilized emulsions. For both proteins, enzymaticdephosphorylation increased the creaming stability 6- and 10-fold, respectively.Dickinson and Stainsby (1987) reported that for n-hexadecane/water emulsion (50% oil, 0.8 %protein, 2 M NaC1, pH 7.5, 25°C) the flocculation stability of the caseins was in the order f3- =ic- >a91-casein. When the continuous phase was 15 mM CaC12 only ic-casein gave adequate130stability. This observation is consistent with the behaviour of the proteins in solution. It is notclear why a,1-casein emulsions are the least stable. The observation that oil droplets andpolystyrene latex particles have a higher electrophoretic mobility when coated witha81-caseinthan with either f3- or ic-casein (Dickinson and Stainsby, 1987) suggests that factors other thanelectrostatic effects are involved.Robson and Dalgleish (1987) found that different caseins are capable of displacing one anotherfrom interfaces and that a displacement order appears to exist. Thus, the more hydrophobic 3-casein displaces the less hydrophobic cz1-casein from the oil/water interface during aging of acaseinate emulsion. A similar effect is seen when an emulsion prepared using cx61-casein istreated with 13-casein (Dickinson and Stainsby, 1988).The topography of cc1- and 3-casein at the oil/water interface has recently been reported. Foracasein immunochemical methods were used (Ametani et aL, 1989) and for 3-casein, kinetics oftrypsin-catalysed hydrolysis was employed (Leaver and Dalgleish, 1990). Fora,1-casein, it wasfound that all antigenic determinants are exposed to the aqueous phase when the protein isadsorbed to the oil surface. For f3-caseins, it was found that with the exception of largehydrophilic segments (N-terminal region residues 1-28) most of the protein lies flat on theoil/water interface.iii. Whey proteinsAccording to Mancini (1992) whey protein concentrate (WPC) is the “star” of the dairyingredient segment, showing 11 per cent growth in 1991. The same author indicated that threeproperties of WPC are responsible for their popularity in formulated foods: high solubility (evenat acidic pH), high nutritive value and good emulsifring properties. Harper et al. (1980)described six different model food systems (bread dough, cake, salad dressing, coffee whitener,starch pudding and whipped topping) to evaluate the most important functional properties ofwhey proteins.131Competitive price of the whey proteins is another important factor influencing its use. Horton(1992) reported that the cost of WPC (34% protein) is about US$2.75 per kg on a protein basis,while non-fat milk solids are US$6.75 per kg same basis.A great number of studies have been published in the area of emulsi,ing properties of wheyproteins, especially on the properties of J3-lactoglobulin, the main whey protein, as well as onserum albumin. Since a comprehesive review in this area is out of the scope of this thesis, thissection will zfly include information relevent to this thesis and some recent developments in thearea.Three reviews on the functional properties of whey proteins have recently been published by deWit (1989), Kinsella and Whitehead (1989) and Morr and Ha (1993). As pointed out by de Wit(1989), a wide variety of model systems and varying conditions have been used to determine theemulsification properties of whey proteins andlor the effect of compositional or environmentalfactors on these properties. However, comparison of results is difficult because the methods ofanalysis have not been standardized.It is commonly accepted that whey proteins have lower emulsification properties than caseins(de Wit, 1989; Blenford, 1990). This is especially true when the native proteins are compared.Compared to caseins, whey proteins are more heat-sensitive, less calcium-sensitive and canengage in thiol-disulfide interchanges to form oligomeric structures (Kinsella and Whitehead,1989). The effect of treatments before emulsffication and environmental factors, especially PH,tend to have a greater effect on whey proteins.Whey proteins are less surface active than caseins (Jackson and Pallansch, 1961; Mitchell et al.,1970), which is a reflection of their more compact and rigid structure. For the three main wheyproteins, a-lactalbumin, f-lactoglobulin and serum albumin, a heat treatment (0.03% proteinconcentration, 85°C for 20 mm) increased the surface activity of the proteins by 10, 30 and100%, respectively (Mitchell et al., 1970). Jackson and Pallansch (1961) reported that 3-lactoglobulin decreased the interfacial tension at the butter oil-water interface with increasing132temperatures to the point of denaturation, whereupon its interfacial activity was depressed. Ina system made of butter oil-water, whey albumins (i.e., x-lactalbumin and serum albumin) weremore surface active than f-lactoglobulin. The reverse was true when the aqueous medium wasprotein-free skim milk (i.e., protein-free plasma) (Jackson and Pallansch, 1961). Roedhi andJelen (1988) reported that the order of surface activity (dynes/cm) for individual native wheyproteins in distilled water at concentrations approximating normal contents in whey is (surfacetension and protein concentration): -lactoglobulin (45.0, 0.29%) = a-lactalbumin (45.9, 0.13%)> serum albumin (48.9, 0.25%). Heating (80°C, 50 mm) slightly decreased (< 10%), the surfaceactivity of serum albumin and f3-lactoglobulin, but improved that of x-lactalbumin.The adsorption of whey protein onto the surface of a fat globule is selective and influenced bypH, presence of salts, protein concentration and temperature (Yamauchi et al., 1980). Theseauthors reported that in an emulsion made with whole whey protein at pH 7, lactoferrin,3-lactoglobulin and immunoglobulins were well adsorbed on the coconut oil globule surface, whilethe relative amount of a-lactalburnin in the adsorbed protein layer was less and no adsorptionof serum albumin was observed. The selectivity in the adsorption was dependent on pH. Serumalbumin did not adsorb at pH 3, 7, and 9, but showed marked adsorbtion at pH 5. On the otherhand, relative adsorption of3-lactoglobulin increased at higher pH, while that of a-lactalbuminwas higher at lower pH (Yamauchi et al., 1980).Britten et al. (1991) compared the emulsifying properties of a whey protein isolate, a sodiumcaseinate and their mixtures. Both were commercial samples. Interfacial pressure and proteinload determination both indicated preferential adsorption of casein over whey proteins.Emulsions (1% w/v protein concentration, 30% oil) containing casein showed more creaming (i.e.,phase separation) upon storage, but less stirring-induced (SI) coalescence than the whey protein.Heat treatment (80°C, 30 mm) of the casein solution before emulsion formation reducedcreaming and SI coalescence but induced slight sedimentation. On the other hand, whey proteinemulsions showed higher SI coalescence but less creaming under storage than casein emulsions.133Creaming, but not SI coalescence, was inhibited by heating whey protein solutions. Heating ofwhey proteins increased the viscosity of the emulsion and nearly formed a gel.Patel and Kilara (1990) related physicochemical properties of six spray-dried whey proteinconcentrates to their emulsifying and foaming properties. Free fat and denaturation enthalpywere positively related to foaming and emulsifying properties.In an extensive study, Toro-Vazquez and Regenstein (1989) compared physicochemicalparameters and emulsifying properties of several protein additives, including a whey proteirconcentrate and a sodium caseinate sample. Elizalde et al. (1988 and 1991) studied the emulsionstabilizing properties of a variety of food protein products in relation to the water and oil uptakeof proteins and the viscosity of the external phase. Emulsion instability values (related more tocreaming than to coalescence) of an emulsion made with whey protein concentrate (1% w/vprotein concentration and 0=0.23) were higher than one made withsodium caseinateor bovineserum albumin. In these studies the whey protein concentrate was found to be the less desirableprotein additive in terms of emulsion stability. These authors suggested that high water andoil absorption, together with high viscosity, have a positive effect on emulsion stability. Theyconclude that the balance between the hydrophilic and lipophilic characteristics of a protein aswell as the magnitude of these characteristics may be the factors which determine emulsionproperties. Knowledge of water and oil adsorption capacity of the protein and composition of theemulsion can be used to predict emulsion instability (Elizalde et al., 1991).Recently, Singh et al. (1993) reported the emulsifying properties of protein fractions preparedfrom heated milk. Emulsions were made with 2.5% w/v protein and 0=0.4. Their results showedthat aggregated protein systems, i.e., casein micelles, a whey protein/ic-casein rich fraction andic-casein-depleted micelles produced lower values of EM but higher creaming stabilities thanhighly dispersed protein systems, e.g., sodium caseinate or whey protein isolate.Enzymatic hydrolysis of a commercial whey protein concentrate and the emulsion capacity (EC)of the resulting peptide fractions were reported by Turgeon et aL (1992a). These authors134reported that the peptide fractions have low EC. In another paper from the same research group(Turgeon et al., 1992b), the interfacial properties ofpeptides resulting from the tryptic hydrolysisof f3-lactoglobulin were reported. These authors proposed the term “amphipolarity” to describeclustering of hydrophobic and hydrophilic residues in distinct regions. It is highly possible thatthe spatial distribution, more than total numbers, of hydrophobic and hydrophilic residues inproteins is the feature determining interfacial behavior in proteins.Emulsification and gelation properties of partially insolubilized whey protein concentrates werereported by Beuschel et al. (1992). These researches reported that for these products solubilityand emulsi1ing properties were not strongly related.Recently, in a clever experiment, McClements et al. (1993) demonstrated the occurence of oilexchange between oil-in-water emulsion droplets stabilized by a whey protein isolate and asodium caseinate sample (protein concentration 0.5% w/w, 0=0.2). bii exchange was measuredindirectly using differential scanning calorimetry (DSC). Here, the time dependence of thecrystallization behavior of emulsions containing a mixture of n-hexadecane and octadecanedroplets was examined. The rate of exchange was faster for emulsions containing whey proteinisolate than the one containing casein. Very little coalescence was observed in emulsions evenafter 29 days of storage. These authors proposed that the mechanism for oil exchange is asfollows: the binding of the oil to the protein at the interface, the displacement of the protein-oilcomplex from the interface, the diffusion of the complex through the aqueous phase, itsincorporation at the interface of another droplet, and the release of the oil.For 4% w/w oil-in-water emulsion, stabilized with 0.16% protein (fat/protein ratio of 25), de Witet al. (1988) reported that a-lactalbumin had the best emulsifying properties (emulsifyingactivity and stability) when compared to f3-lactoglobulin, serum albumin and immunoglobulins.Emulsion stabilized with a-lactalbumin, f3-lactoglobulin, serum albumin and immunoglobulinhad volume to surface diameters (d) offat globules of0.34, 0.54, 1.20 and 1.26 pm, respectively.After one week of storage dvs values were 0.34, 0.66, 1.20 and 1.17 pm, respectively. In other135systems, -lactoglobu1in has been shown to have better emulsifying properties than alactalbumin, especially with respect to emulsion stability (Slack et at, 1986).According to data reported by Kato et al. (1983), taking emulsifying activity index (EM) andemulsion stability (ES) of ic-casein as 1, -lactoglobulin and serum albumin had relative EM andES of 0.9 and 1.1, respectively. Emulsions were made with 0.2% protein in 0.1 M phosphatebuffer, pH 7.4 with 0 = 0.25. In this study, heat denaturation was found to decrease EM andES values for -lactoglobu1in and serum albumin. Surface hydrophobicity (So), estimated usingthe Kato and Nakaa (1980) method, was also found to decrease upon heat denaturatien. Forthese proteins, So values of-lactoglobulin and serum albumin were found to be almost an orderof magnitude higher than for ic-casein. Reported So values for ic-casein, -lactoglobulin andserum albumin were 430, 2700 and 3200, respectively (Kato et al., 1983). It is well known thatboth f-lactoglobu1in and serum albumin have a strong tendency fo binding nonpolar ligands.This decrease in surface hydrophobicity values, upon heat treatment of these two proteins,probably does not reflect changes in So, but may be related to conformational changes affectingspecific ligand binding sites of the proteins.In another study by the same research group (Kato et al., 1985a), the relationship betweensurface properties and flexibility of proteins was investigated. In this study emulsions weremade with 0.1% protein concentration in 0.06 M phosphate buffer, pH 7.4 with 0=0.25. Underthese conditions, EM values for the main whey proteins were in the following order: serumalbumin — ic-casein > a-lactalbumin > J3-lactoglobulin. Good positive correlation was observedbetween emulsifying activity and the digestion velocity of proteins determined by achymotrypsin. Since digestion velocity is an indirect measure ofprotein flexibility, these authorssuggested that flexibility ofprotein structure is an important factor affecting protein emulsifyingproperties. For the whey proteins and ic-casein the digestion velocity reported was in thefollowing order (actual number in parenthesis, %/min): ic-casein(4.8) > f3-lactoglobulin(4.3) >136serum albumin(4.0) > a-lactalbumin (2.9). This order is different than the one obtained whenthe EM of the proteins are used for the ranking.Intramolecular crosslinked serum albumin was very resistant to protease digestion but itsemulsifying properties were only slightly decreased (Kato et aL, 1986a).In an additional study from this research group (Kato et aL, 1985b) on emulsions made with0.1% w/v protein in 0.02 M phosphate buffer pH 7.4 and 0=0.25, the whey proteins analyzed(a-lactalbumin, -lactoglobulin and serum albumin) were all found to produce similar EMvalues, which were higher than that of ic-casein. a-Lactallximin and -lactoglobulin were foundto produce similar ES values, which were higher than those of serum albumin and ic-casein.Dickinson and Matsumura (1991) demonstrated that at the oil-water interface of emulsions atime-dependent polymerization of f3-lactoglobulin through disulphide bonds occurs. While onlymonomers are detected in the adsorbed protein immediately after Jrnulsion formation with -lactoglobulin, on storing the emulsion the amount of polymerized protein and size of oligomersincreases with time. After 72 h of storage the relative amounts of monomer, dimer, trimer andtetramer or higher are 38, 26.1, 8.2 and 27.5%, respectively (Dickinson and Matsumura, 1991).This work suggests an important role of free protein sulphydryl groups in the development ofthe high surface viscoelasticity of adsorbed globular protein at the oil-water interface. Murphyand Howell (1990) reported that thiolation of serum albumin, by attachment of homocysteineresidues, to either 16 or 21 amino groups of the protein impaired the whipping and gellingproperties of the protein but there was little change in emulsifying properties in terms of dropletsize and creaming stability.The ability of whey proteins to stabilize o/w emulsions seems to be seriously affected by the pHand the ionic strength of the aqueous phase. Shimizu et al. (1985) investigated the emulsifyingproperties of f3-lactoglobulin at pH 3—9 and also the effect of pH on some conformationalparameters. Emulsions were made with 1% w/v protein and 0=0.2. Increasing the pH produceda curvilinear increase in EM. 3-Lactoglobu1in at pH 3 showed an EM 30% lower than the value137found at pH 9. Two methods were used to estimate surface hydrophobicity (Ho): hydrophobicchromatography and fluorescence probe (ANS). For both methods a decrease in Ho was observedwith increasing pH; the fluorecence probe method Ho at pH 9 was 10% of the value at pH 3.Thus, in this experiment, positive correlation between Ho and EM was not found. By usingsurface pressure measurements and urea and GdnHC1 denaturation curves by intrinsicfluorescence, these researchers suggested that the low EM of this protein at pH 3 was due tolow denaturability (flexibility) of the molecule at this pH. To further prove this point, theycleaved the intramolecular disuffide bonds of the protein with 0.1% 3-mercaptoethanol, toincrease its flexibility, and measured EM and urea denaturability at pH 3 for the reducedprotein. Reduced f-lactoglobulin became more susceptible to urea denaturation than the nativeprotein. Reduction also caused about a 50% and 300% increase in EM and ANS Ho, respectively.As mentioned before estimation of Ho using anionic probes such as ANS at different pH valuesis not recommended, since electrostatic effects affecting the probe are difficult to account for.Thus, the effects of pH on the probe Ho reported by Shimizu et al. (1985), need to be taken withcaution. Unfortunately, this paper has been cited by other researchers (e.g., Kinsella et al., 1989)as an argument to minimize the importance of surface hydrophobicity.Recently, Chen and Dickinson (1993) reported that at pH 3 the surface coverage off3-lactoglobulin in oil-water emulsions is 1.05 rng/m2,while at pH 5 and pH 7 it is 1.6 mg/rn2.While at pH 7 a large amount of Tween 20 (surfactantlprotein molar ratio (R) 25) is neededto displace the protein from the interface, at pH 3, R=15 produces complete displacement. Theseobservations confirm the data reported by Shimizu et al. (1985). In the case of emulsionsstabilized by -casein at pH 7, a value of R20 produced complete displacement (Chen andDickinson, 1993). Relkin et al. (1993) reported that the thermal stability of 3-lactoglobu1in,evaluated from the denaturation temperature, is maximum at the isoelectric pH in acetate bufferand at pH 3.5 in distilled water. They also confirmed the observations of Shinilzu et aL (1985),and suggested that the protein is more compact at acidic pHs, leading to higher denaturation138temperature, and more flexible at basic pHs since high reactivity offree SH groups was observedat pH>7.0.Acylation and alkylation of 3-lactoglobu1in produced a 3 to 6 fold increase in EM values but notin coalescence stability (Creuzenet et al., 1992).In one of the more comprehensive and well-thought studies, Das and Kinsella (1990b) reportedthe effect ofheat denaturation on the adsorption of 3-lactoglobulin at the oil/water interface andon coalescence stability of emulsions. Emulsions were made with a protein concentration of 0.3w/v % in 0.1 M phosphate buffer, pH 6.5 and 0 = 0.2. Kinetics of coalescence upon storage atroom temperature for up to 5 days was measured using a spectrophotometric method. Heatingof f3-lactoglobulin increased its Ho measured by ANS probe, and different time-temperaturetreatments were used to obtain protein samples with different values of Ho. It was observed thatincreasing Ho provided greater resistance to coalescence. Since the average volume surfacediameter of the emulsions were kept nearly constant, the changes in coalescence reflect changesin film stability. The monolayer adsorption density (mg/ni2)of f-lactoglobulin decreased withincreasing surface hydrophobicity and the coalescence stability increased. The protein was moretightly bound to the interface as bulk surface hydrophobicity increased. The least stable of theemulsions was given by the most compact conformation of the protein in the bulk solution priorto emulsification. This corresponds to the lowest area per molecule, indicating maximumretention of native conformation at the interface. Only the first layer of adsothed proteinappears to impart stability against coalescence. Coalescence stability of -lactoglobulin alsocorrelated in a positive way with the strength of the adsoied protein toward detergentdisplacement. This correlation indicates that the energy barrier to coalescence is related to theprobability ofdesorption ofprotein from the interface. Overall, the results from this investigationsuggested that the conformation ofthe protein may be more important than the thickness of theadsorbed layer in providing resistance against coalescence.139The effects of pH, ionic strength, and thiol reagent induced structural changes on theemulsifying properties of serum albumin were studied by Haque and Kinsella (1988b). Animportant characteristic of this study is that the energy input during emulsification was keptconstant by using a computerized controlled single-piston recirculating valve homogenizer.Increasing the energy input increased the surface area of the emulsion. Addition of the reducingagent dithiothreitol (DT) at 1 mM tended to increase the EA of the protein, but at higher DTI’concentrations (8 mM) the EA decreased dramatically, especially at higher energy inputs. At pH8, the EA was greater than at lower pH valu.s, but the energy required to make the emulsionwas higher. ES was maximum at pH 4, where the protein is believed to have an uncoiledconformation.A comparison between emulsions prepared with serum albumin or egg yolk low densitylipoprotein (LDL) was made by Mizutani and Nakamura (1985). Results clearly showed thatLDL had far superior emulsifying properties than serum albumin, both in terms of EM andcoalescence stability. Upon storage for 15 days of a serum albumin emulsion (proteinconcentration of 0.3%, 0=0.75) the globular size changed from 3 am to 4.2 pm.Haque and Kinsella (1989) compared the emulsifying activity of serum albumin and delipidatedcasein. Three different methods were used to estimate mean globule diameter, namely electronmicroscopy, computerized optical microscopy and turbidity method of Walstra. EM was alsomeasured. Emulsions were made with a molar concentration of 0.135 mM for both proteins and0=0.4, at different energy inputs. In terms of EA (i.e., m2 of interfacial area per g protein) thedelipidated casein had on average, a higher (=10%) EA than serum albumin. But when EA wasexpressed per mol of protein (MEA), serum albumin was far superior. The difference in mwbetween serum albumin (67,000 daltons) and casein is very large and a given mass of casein willcontain 3.30 more molecules, and hence more molecular surface, than serum albumin. Theseauthors suggested that MEA may be a better measure of the EA of proteins. Serum albumintended to produce markedly (2 to 5 times) smaller droplets than casein. Formation of serum140albumin emulsions at high power density caused a decrease in EA. This change may be areflection ofthe intermolecular state ofthe protein. Increased intermolecular association appearsto strengthen the interfacial film whereas dissociation weakens the film (Haque and Kinsella,1988b).b. Egg white proteinsContrary to the milk proteins, relatively few papers have been published about the emulsifyingproperties of egg white proteins. This is understandable since egg white as a food ingredient has,in general, poor emulsifying properties and is not used as an emulsffier. It has been suggestedthat egg white may have some emulsifying properties in some food systems (e.g., snack and layercakes) (Shen and Jackson, 1991).As previously pointed out (see section on some physicochernical properties of egg white proteins),these proteins are relatively easy to purify in large quantities. Although not as wellcharacterized as some of the milk proteins, they provide good model proteins for studyingstructure-functionality relationships. Baldwin (1990) and Froning (1988) as well as MacDonnellet al. (1955) indicate that the two main functional properties of egg white in foods are foamingand gelation/coagulation. From a nutritional standpoint, egg white proteins are considered oneof the best sources of high quality protein and are routinely used as a standard for proteinquality determination (Froning, 1988). Eggs (white+yolk) as a food ingredient aremultifunctional. Few ingredients can match the superior foaming, gelation, binding andemulsification attributes provided by eggs. Extensive research has been done to develop otheringredients that could fully replace eggs in various formulated products. Although partialreplacement has been possible with such products as soy protein isolate, none of the substitutesoffer the superior multifunctional ability of eggs.As pointed out by Shen and Jackson (1991) soy proteins do not have all the properties of egg (ormilk) proteins and may provide only the properties needed for a given application. While the141same egg (or milk) protein can be used in a number of applications, different soy proteinpreparations may be needed for each.Nakai and Li-Chan (1985) used a quantitative structure-activity approach to the structure andfunctionality modification of whey proteins in an attempt to obtain whipping and gellingproperties similar to those of egg white. Although whipping of whey protein concentrates wasimproved by pepsin hydrolysis and gelling by a polyphosphate treatment, the results of angelfood cake making were still unsatisfactory (Nakai and Li-Chan, 1985).Contrary to egg white, egg yolk has excellent emulsifying properties, being an importantemulsifying ingredient in the manufacture of mayonnaise, salad dressing and cakes. Theemulsifying components in yolk are phospholipids, lipoproteins, and proteins (Powrie and Nakai,1985). Phosvitin, a highly phosphorylated glycoprotein present in egg yolk, was found to havea twofold and tenfold EM value as compared with serum albumin aiid ovalbumin, respectively,for emulsions made with 0.1% protein and 0=0.25. The emulsion stability (creaming) ofphosvitin was also far superior than the one for the two other proteins (Kate et al., 198Th).Kate et al. (1985c) reported that the egg white protein ovomucin, which is a sialoglycoproteinresponsible for the gel structure of thick egg white, has good emulsifying and foaming properties.At a protein concentration of 0.25% and 0=0.2, the relative EM values for serum albumin,soluble, sonicated and reduced ovomucin were 1.0,0.8, 0.9 and 1.0, respectively. Under the sameconditions the relative emulsion stability (creaming) values were 1, 1, 2 and 3.8. The dissociationof this protein proceeded in the order of soluble, sonicated and reduced (i.e., treatment with -mercaptoethanol). Surface hydrophobicity (CPA method) also increased in the order of soluble,sonicated and reduced ovomucin. This protein closely resembles phosvitin as a polyanion-typeprotein.The good functional properties of ovomucin may be due to the large number of carbohydrate sidechains in addition to the large molecular size (Kate et at, 1985c). This protein was not used inour research.142Pearce and Kinsella (1978) reported that at a protein concentration of 0.5% in 0.1 M phosphatebuffer, pH 6.5, ovalbumin and lysozyme have EM values of 49 and 50 m2g4, respectively. Underthe same conditions 3-lactoglobulin, sodium caseinate and a whey protein powder have EMvalues of 153, 166 and 142 in2g1. Ofmore than 10 proteins analysed by these researchers, thesetwo egg white proteins produced the lowest EM values. The sane authors indicated that forovalbumin, increasing the number of passages through a homogenizer produced a decrease inthe number of globules formed (N). This behaviour was unusual since most protein presenteda signfficant increase in N with increased number of passages. The fact that ovalbumin is easilydenatured at surfaces and that a marked depletion of protein from solution was noted duringemulsion formation (Pearce and Kinsella, 1978) suggests that for this and other proteins theremay be an optimum degree of homogenization for maximum dispersion.Kato and Nakai (1980) reported that for emulsions made with b.2% protein and 0=0.25,ovalbumin and lysozyme produced very low EM values (— 55 m2 g’), ovotransferrin gave anintermediate value (100 m2 g’’) and serum albumin gave the highest EM value (155 in2 g’).These authors also indicated that heat denaturation of ovalbumin increased both the EAT andSo (CPA method with 0.002% SDS). Native ovalbumin had very low So (60) and EAT (—55 m2g’), but after heating 1,2,3,4, and 5 mm at 85°C the So (%1) values were 62,400,700,800 and900. Corresponding EM (m2 g’) values were 50,60,70,85 and 100, respectively. A similar trendwas also found for lysozyme. The So and EM values for native and heat denatured lysozyme(heated 6 mm at 85°C) were: 100 and 55 in2g, and 1250 and 150 m2 g’. A surfactant treatmentwith either SDS or linoleate also produced significant increases in both surface hydrophobicityand EM.Keshavarz and Nakai (1979) reported that the interfacial tension (dynes/cm) at the cornoil/water interface was 21.6, 20.5 and 9.3 and 6.0 for 0.2% solution of ovotransferrin,f34actoglobulin, lysozyme and serum albumin, respectively. Tsutsui et al. (1986) reported that143ovomucoid gave a larger value of EM than ovalbumin, although for both proteins the EM wasvery small (<5 m2 1)The effects of partial denaturation on surface properties of ovalbumin and lysozyme wasinvestigated by Kato et al. (1981). The denaturation treatment consisted of heating 20 mL of0.2% protein in 0.01 M phosphate buffer, pH 7.4 for 1-6 mm, followed by immediate cooling ofthe solution to room temperature. The So of ovalbuniin and lysozyme increased markedly as heatdenaturation proceeded. Emulsion stability (creaming stability) also increased with denaturation,and correlated linearly with So (CPA method). The ES value for native lysozyme and ovalbuniinwas 2 mm, for the proteins with the highest level of denaturation was 20 and 26 mm,respectively. The structure of ovalbumin, as estimated by the the ellipticity at 222 nm, wasfound to be more sensitive to heat than that of lysozyme. Despite a relative small conformationalchange in lysozyme upon denaturation, its So greatly increased (Kate et al., 1981). Theseauthors reported an 18% decrease in helix content in lysozyme and a 28% decrease in ovalbuminon heat denaturation at neutral pH, while for both proteins So increased by a factor of 20. It wasalso reported that the increase in So of heat denatured proteins did not change even 1 h afterdenaturation, suggesting slow renaturation. SDS-treated ovalbumin also showed improved ES.Changes in ellipticity at 222 nm of SDS-bound ovalbumin were found to be smaller than thoseof the heat denatured one, although the So of the SDS-bound protein increased to almost thesame values as those of heat-denatured ovalbuinin. An important conclusion of this research isthat denaturation is not necessarily detrimental to protein functionality. For these proteins, thesurface properties were improved considerably by heat denaturation if no coagulation occured.Kato et al. (1983) reported that ovalbumin heated to 20°C (control), 70°C, 75°C, 78°C and 80°Cat a heating rate of 1°C/mm had relative EM and ES (creaming) values of 1 and 1, 1.7 and 1.0,2.2 and 1.4, 3.0 and 2.3, and 3.8 and 2.8. The So (CPA method) for ovalbumin as well as the Sofor 7S globulin and ic-casein increased with denaturation, while that of J3-lactoglobulin andserum albumin decreased with heat denaturation. The EM and ES values of native ovalbumin144were reported to be 6 and 8 times less than those of serum albumin. Voutsinas and Nakai (1983)also reported that thermal denaturation improved the emulsifying properties of ovalbuniin.Heating a 1% protein solution (pH 1.0) for 15 mm at 100°C caused a four-fold increase in EATin spite of decreasing the solubility of the protein by almost 60%. ES values for the heat treatedprotein was 22.5 times larger than that for the native protein. The So values (CPA method) forthe native and denatured protein were 6 and 296, respectively. These papers clearlydemonstrated that the emulsitring properties of proteins are influenced by both hydrophobicityand solubility. Although solubility is very important, it cannot alone fully explain theemulsifring properties especially when the proteins are heat denatured. Heating does not haveuniform effects on the emulsifying properties of different proteins: for some proteins (e.g.,ovalbumin and gelatin) heating improved these properties, for others (e.g., -lactog1obu1in andcasein) it adversely affected both EM and ES. -The digestion velocity for native ovalbumin and lysozyme by chymotrypsin were found to be 0.5and 0.3 %hnin; these low values suggest rigid or folded molecules (Kato et aL 1985a). Thedigestion velocity for ovotransferrin was found to be 1.7%/mm. Although not as high as for otherproteins (e.g., 2.9 for a-lactalbumin, 4.3 for f3-lactoglobulin) it was higher than that found for theother egg white proteins. The EM value (expressed as absorbance at 500 nm) for ovotransferrin(0.5) was also higher than for lysozyme (0.13) and ovalbumin (0.2). Acetylation of these twolatter proteins increased the digestion velocity by a factor of 4 and 10, and the EM by a factorof 5 and 7, respectively. On the other hand, acetylation of ovotransferrin only increased thedigestion velocity and EAJ values by a factor of 1.6 and 2, respectively (Kato et aT. 1985a).Kato et al. (1985b) also reported that ovotransferrin has greater EM and ES values thanovalbumin and lysozyme. In this study, EM value for ovotransferrin was found to be similar tothe one given by the soy protein uS. These researchers also indicated that the ES values forovaibumin were independent of protein concentration; these proteins gave very low ES valueseven at 1% protein concentration.145In a subsequent paper from this research group (Kato et aL, 1986b), the relationship betweenprotein flexibility and emulsifying and foaming properties of heat-denatured ovalbumin andlysozyme were studied. The heat transition points for EM of ovalbumin and lysozyme were 74°Cand 82°C, being very close to those of So. Correlation analysis showed that So was more stronglyrelated to EM than protease-digestion velocity. Both parameters were significantly correlatedto EM. The digestion velocity values suggested that ovalbumin was more flexible than lysozyme.An increase in exposure of hydrophobic residues on the molecular surface of total egg whiteprotein, as measured by So (CPA method), and formation of disulfide bonds in total egg whiteprotein occurred during heating, causing egg white proteins to polymerize by intermolecularsulfhydryl-disulfide exchange (Mine et al., 1990). According to CD data, the secondary structureof native ovalbumin consists of 40.6% helix, 15.8% 13-sheet, 15.5% 13-turn and 28.2% unordered,while for a protein heated to 100°C in distilled water, pH 9.5 and cooled to room temperature,the secondary structure was made of 15.8% helix, 46.8% 13-sheet, 5.5% 13-turn and 3 1.9%unordered (Mine et al., 1990). Koseki et al. (1989) also concluded, from the results of chemicalmodifications, that part of some ovalbuxnin hydrophobic regions containing SH groups came tobe exposed on the surface of the molecule during heat denaturation.Ikura et al. (1992) used monoclonal antibodies to identify the structural regions of the ovalbuminmolecule that change during heat denaturation. Four new epitope regions at the surface ofovalbuinin after heat denaturation were detected: Asn159-Lys81,Phe1-M t211,VaL-Lys andfle-Lys9.Mine et al. (1991) studied the emulsifying and structural properties of ovalbumin. Theemulsifying properties of this protein were dependent on pH, concentration ofprotein dispersion,0 and presence of salts. The pH was the most important variable in the EM. The EM ofovalbumin was high at pH 3. For an emulsion made with 1% w/v protein and 25% w/w oil, theEM values (expressed as absorbance at 500 nm) were 0.796, 0.526, 0.483 and 0.439, at pHs 3,5, 7 and 9, respectively. The EM and ES values were very low when the protein concentration146was below 0.5%. At a protein concentration of 0.1% and 50% oil the EM value was 0.131 andthe ES (creaming stability) was less than 15 s (Mine et aL, 1991). Fat globule size was alsominimal at pH 3 and increased with increasing pH. The So (CPA method) of the protein at acidicpHs (pH 3 and 5) was found to be greater than at neutral pH (Mine et al., 1991). Limitationsof using fluorescence probe methods at different plls have already been pointed out, thus theseobservations must be taken with caution. The line widths of 31P NMR spectra of ovalbumin atpH 3 or 8 suggested that the conformation of ovalbumin was more flexible at acidic pHs thanat neutral. These researchers conclude that the relatively high EM of ovalbumin at acidic pHis related to greater flexibffity of the protein and greater surface hydrophobicity.In a subsequent publication, Mine et al. (1992) reported the effects of limited proteolysis on theemulsifying properties of ovalbumin. EkE was decreased when a peptide (residues 1-22) from theN-terminal region was cleaved with pepsin under acidic conditions.This N-terminal peptide ishydrophobic and is the core of the protein. On the other hand, digestion with subtilisin cleaveda large part of the C-terminal region (346-385), which most likely was on the surface of theprotein, and formed a plakalbumin-like protein, yielding a 3-fold EM value than the nativeprotein. The authors do not give a clear explanation for these observations. It may be possiblethat these limited proteolysis caused some changes in the conformation of the protein andaffecting their interfacial behaviour. 31P NMR spectra of native and modified ovalbumin andovalbumin emulsions suggested that the phosphate residue at phospho-Ser 344 has a restrictedmobility due to interactions between part of the molecule around this residue and the oil phase(Mine et al., 1992).Deamidation, that is the conversion of Gin and Asn residues to Glu and Asp with the use ofproteases at alkaline pH, have been shown to improve, to a limited extent, the emulsifying andfoaming properties of lysozyme and ovalbumin (Kato et al., 1987a). This process increases thenegative charge of the surface of the molecule.147Heating egg white proteins in the dry state have been reported to increase emulsifying andfoaming properties (Kato et al., 1990a) and also gelation properties (Kate et aL, 1990b).Enthalpy of denaturation (JI) was found to be linearly related to EAT and ES (creaming) (Kateet al., 1990a). Heating of egg white in the dry state causes a substantial increase in molecularflexibility and surface hydrophobicity (Kate et al., 1989; Kate et aL, 1990b). Calorimetricanalysis of egg white proteins as a function of heating time showed that the enthalpy ofdenaturation (H) dramatically decreased with an increase of heating time, while denaturationtemperature (Td) only decreased slightly (Kate et al., 1990b). The reduction of All of egg whiteproteins by dry heating may enhance the rate of diffusion of the molecules to the interface,facilitating protein-protein interactions during the formation of the interfacial film. Proteinshaving high entropy due to the reduction of AH and more flexible polypeptide chains shouldfacilitate a greater degree of association by hydrophobic and electrotatic interactions (Kate etal., 1990a).The emulsifying properties of ovalbumin were also improved by coupling with dextran. Solubleprotein-dextran conjugate prepared by coupling ovalbumin to cyanogen bromide activateddextran had almost tenfold higher values for both EM and ES (creaming) as compared to nativeovalbuinin(Kato et al., 1988a). Noncovalent mixing failed to generate the high values ofEAI andES. The authors indicated that the majority of the conjugates had a mw of 250,000 daltens andthat the composition of the subunit was 1 mol of ovalbumin and 3 mol of dextran, which wasextended as a prolonged polymer. The emulsifying properties of an ovalbumin-pectin conjugatewere not improved. The authors suggested that structural differences between thesecarbohydrates may be responsible for the different properties of their conjugates. Dextran hasa highly branched network structure, while pectin has a single-chain structure (Kate et al.,1988a). In a subsequent paper, Kate et al. (1990c) reported the formation of a functionaldextran-ovalbumin conjugate by storing for 3 weeks at 60°C and 65% relative humidity a148mixture of protein and dextran (weight ratio 1 to 5). This conjugate had also excellentemulsifying properties.6. Effect of urea on emulsifying properties of proteinsa. IntroductionSince in this work proteins were denatured with urea and protein stabilized emulsionswere made in the presence of different concentrations of urea, it is pertinent to review the effectof urea on emulsions. However, it is important to point out that limited information was foundin this area. The main questions which need to be answered are: (1) Do high urea concentrationsaffect the behavior of emulsffiers? and (2) If the answer to (1) is yes, in which way and to whatextent? -b. Effect of ureaHigh urea concentrations most likely affect the behavior of emulsifiers, either byaffecting the structure of water (see section Urea as a protein denaturant), or by specificinteractions with emulsifiers. The dipolar properties of urea, although small, suggests that itmay have some emulsification properties. Although reports were not found in this area, theobservation that addition of a 20% urea-0.2% imidazole solution to acetic acid-solubilized milk,makes a uniform emulsion (Le and Nakai, 1970), suggests that urea may have some emulsifyingability.The effect of urea on micelle formation in an aqueous solution of surfactant and on thedenaturation of proteins is generally interpreted in terms of the breakdown of water structureand subsequent weakening of “hydrophobic bonds”. Both of the phenomena of micelle formationand protein denaturation involve the equilibrium distribution of hydrocarbon chains betweenhydrophilic and hydrophobic environments and the effect of urea is to displace the position of149the equilibrium in favour of an increased concentration of the hydrocarbon chain in thehydrophilic media (Shaw, 1992; Jones, 1973).Urea is well known for increasing the critical micelle concentration (CMC) of some surfactants.Shaw (1992) writes: “Organic molecules may influence CMC’s at higher additive concentrationby virtue of thdr influence on water structuring. Sugars are structure-makers and as such causea lowering of CMC, whereas urea and formamide are structure breakers and their additioncauses an increase in CMC.” Bruning and Holtzer (1961) reported that the values for CMC (inmoles per 1) for the cationic detergent n-dodccylmethylammonium bromide in 0.0, 0.5, 2.0 and6.0 M urea were 0.0142,0.0156,0.0204 and 0.0454, respectively. These authors pointed out thatthe effect of urea is small when compared to the effect of acetone. Aqueous solutions of thesurfactant in 6.6 M acetone showed no evidence of micelles even at detergent concentrations ashigh as 0.12 M (Bruning and Holtzer, 1961). Interestingly, based on these experiments, Bruningand Holtzer (1961) concluded that urea does not break hydrophobic bonds.The interfacial tension (‘, dynes cm1) of an hydrocarbon-aqueous urea system is smaller thanfor a hydrocarbon-water system. At 25°C, the y for water-n-decane is 51.7, while for the 8 Murea-n-decane system is 46.5 (Jones, 1973). This observation also indicated that urea had asmall surface activity. Commercial emulsiflers and proteins tend to depress thewater-hydrocarbon interfacial tension to a greater extent (=20-30 dynes cm’).According to Jones (1973), urea did not affect the surface activity of the surfactant ndodecyltrimethylammonium bromide (DTAB). Although urea raised the CMC of DTAB,experimental results indicated that the surface coverage of DTAB in water-, 3 M and 6 M urean-decane interface are all 3 x 10b0 mole/cm2.This author concluded that the invariance ofsurface coverage with urea indicates that urea does not significantly affect the adsorption ofDTAB at the aqueous n-decane interface over the range of urea concentration up to 6 M. Thisimplies that there are not specific interactions between the urea molecule and DTAB micelles.This author indicated that similar results were obtained for SDS.150In a more recent study, Joos and Serrien (1989) have reported the effects of urea and fructoseon the dynamic ‘y of lower alkanols at the air-water interface. For solutions containing urea thedynamic ‘y are higher and for solution with sucrose the dynamic y is lower than those for purealkanol solutions without these additives. These authors discussed their results in terms of thewater “structure breaker” effect of urea. Urea presumably decreased the structuralization ofwater molecules around hydrocarbon chains so that the surfactant is less surface active and theadsorption rate constant decreased. For pentanol the absorption rate constant to the air-waterinterface was reported to be 5.5 x i0 cm 51; in the presence of 6 M urea this parameter was 3.8x iO-3 cm s’. The effect of urea is much less than that obtained by decreasing the number ofcarbons in the alkanol by one, the adsorption rate constant for butanol in water was s4.1 x i(Y3cm s1). Fructose, a “structure former” substance increased the adsorption rate constant. Theadsorption rate constant is independent of temperature, and hence the adsorption rate s merelyentropic. The desorption rate constant, which does not depend in the hydrophobic effect, was notaffected by either urea or fructose (Joos and Serrien, 1989). These two studies suggested thaturea affects, to a rather limited extent, the behavior of low mw emulsifiers at planar interfaces.A recent paper (Breslow and Guo, 1990) reported the somewhat surprising observation thataqueous solutions of urea and GdnHCI have a greater y than water. The y of a 5 M aqueoussolution of GdnHC1 was reported to be 74.9 dynes cm1, compared to 72.75 cm1 for pure water.Urea and GdnHCI are usually referred to as chaotropic agents, because of the disordering theycause. These compounds are also termed salting-in-agents, because they increase solubility ofnonpolar substances (e.g., benzene) in water. From a thermodynamic point of view, chaotropicagents can affect two different solubility energy terms. The first one is the energy required tomake a “hole” or cavity in the solvent, into which the solute can go, and the other is the energyof solute-medium solvation energy (Breslow and Guo, 1990). In many cases it has been assumedthat chaotropic agents disrupt the structure of water, decreasing the energy required to makea new surface in the solvent (i.e., cavitation energy). If this is the case, chaotropic agents will151tend to decrease the yof water. These authors concluded that the main effect of these chaotropicagents is in increasing the solvation of the nonpolar molecules by acting as a favorable bridgebetween solutes and the water. Although controversy still exists regarding whether meat batteris a true emulsion (Regenstein, 1988), a recent study (Gordon and Barbut, 1990) reported thatthis multicomponent system showed some emulsion-like characteristics. In the Gordon andBarbut (1991, 1992) studies, several chemical agents, including urea, were added to the meatbatter and their effect evaluated. Urea was added to a level of 4.5% of the meat block + urea.Most chemicals tested did not produce any detectable effect. Addition of an even higher ureaconcentration (5% of the aqueous phase) improved water retention and did not affect fatexudation or gel strength in a model meat batter (Whiting, 1987).c. Summary of the effect of urea on irotein stabilized emulsionsBased on these reports, it can be suggested that the effect of urea on a protein stabilizedemulsion can be represented by:E(u) = f{ S(u), 1(u), D(u) }were E(u) is the emulsion property (e.g., EAI or emulsion stability) at an urea concentration u,S(u) is the urea solvent effect, 1(u) is the effect of urea on protein-lipid, protein-protein, andlipid-lipid interactions, and D(u) is the urea protein denaturation effect.Although high concentrations of urea affect several properties of water (e.g., density, viscosity,surface tension and dielectric constant) its effect is usually small. The term 1(u), includes theeffect of urea on several interactions. If the increase of CMC of surfactants in the presence ofurea is taken as a typical measurement, this effect seems to be also small. The next effect, theprotein denaturation effect, is the one we are concerned about and since denaturation has beenshown to greatly affect the surface properties of proteins, it is expected to be the most importanteffect in protein stabilized emulsions made in the presence of high urea concentrations.152MATERIALS AND METHODSA. PROTEIN SAMPLES AND CHEMICALSThe following proteins from cow’s milk and chicken egg white were used: a-IactalbuminNo. L-5385 (approximately 85% pure by electrophoresis); f3-lactogibulin No. L-0130 (containsf3-Iactoglobulin A and B); bovine serum albumin No. A-7511 (essential fatty acid free); f3-caseinNo. C-6905 (minimum 90% pure by electrophoresis); ovalbumin No. A-5503 (99% pure byelectrophoresis); lysozyme No. L-6876 (3x crystallized); ovotransferrin No. C-0755 (conalbuniin,contains less than 0.002% Fe); ovomucoid No. T-2011 (chicken egg white trypsin inhibitor,purified ovomucoid, free of ovoinhibitor); all were purchased from Sigma Chemical Co., St. Louis,MO. Sodium dodecyl sulphate-polyacrylamide gel eleetrophoresis (SDS-PAGE) using the Phastsystem (Pharmacia) with Coomassie staining indicated that the purity of the proteins wasgreater than 90%. Deuterium oxide (No. D-4501, 99.9 atom % D), urea (Molecular BiologyReagent No. U-5378), deuterated urea (No. U-4877, Urea-d498+ atom % D), sodium deuteroxide(No. S-7630, 99+ atom % D), dextran sulfate (No. D-6001, sodium salt, average mw of 500,000),and 8-anilino-1-naphthalenesulfonic acid (ANS) (No. A-5144) were also from Sigma ChemicalCo., St Louis, MO. Cis-parinaric acid (CPA) (P-1901) was from Molecular Probes, Inc., JunctionCity, OR. All other chemicals used were analytical reagent grade or better.1. Preparation of a..1- and ic-caseina- and ic-Casein were prepared according to the methodology of Zittle and Custer(1963). One liter of raw milk from the UBC dairy farm was centrifuged at 4°C for 30 mm at4,000 x g in order to separate the fat. The resulting skim milk was diluted with one volume ofdistilled deionized water, and the pH of this mixture was adjusted to 4.6 by the dropwiseaddition of 2 M HCI. This procedure was carried out at 37°C. The casein precipitate was filteredthrough cheese cloth, washed with water, and dissolved in 11 of 6.6 M urea. Two hundred ml153of 3.5 M H2S04 and 2 1 of water were added to lower the pH to 1.5. The mixture was leftstanding at room temperature for 2 h. Then, the precipitate formed was removed by filtrationthrough a Whatman No. 4 paper filter. This precipitate (P1), which contains a- and 3-caseins,was used for thea1-casein isolation as described below. ic-Casein was precipitated from thesupernatant by addition of 132 g (NH4)2S0.The precipitate was collected by centrifugation at1000 x g for 15 mm at 4°C, washed with cold water (4°C) and recentrifuged. The precipitatewas suspended in water and 1 M NaOH was added to give a pH of 7.5 and the mixture waslyophilized. The resulting crude ic-casein was purified as follows: a 1% solution of crude ic-caseinat pH 7.0 was mixed with two volumes of ethanol. Then 1 M ammonium acetate in 75% ethanolwas added until a precipitate formed. The supernatant was decanted and the precipitatedissolved in water by addition of 1 M NaOH to bring the pH to 7.5. The solution was dialysedfor 3 days at 4°C against distilled deionized water and then freeze dried.The preparation ofa,1-casein was as follows (Zittle et al., 1959): the P1 precipitate was dissolvedin 150 ml of 6.6 M urea containing 3.2 g of NaC1 per 150 ml: Then, 215 ml of distilled deionizedwater was added to lower the urea concentration to 4.7 M and precipitate theci1-casein. The a,-.casein precipitate was dissolved again in 6.6 M urea-NaCI solution and precipitated again bythe addition of water. The final precipitate (P2) was washed first with 4.7 M urea and then withcold water and finally dissolved in water after adjusting the pH to 7.5. The solution was dialysedfor 3 days at 4°C against distilled deionized water and then freeze dried.To further puriira1-casein the following procedure was used (Zittle and Custer, 1963): to 2%aqueous solution of the P2 precipitate at pH 7.2, one volume of ethanol (95%) was added, andthen 1 M ammonium acetate in 75% ethanol was added until precipitation was observed. Theprecipitate was discarded and the cç1-casein, which was present in the supernatant, wasprecipitated at pH 5.0 by addition of 3 M HCL The precipitate was collected by centrifugationat 4000 x g for 15 mm at 4°C, washed with cold water and recentrifuged. The precipitate wassuspended in water and 1 M NaOH was added to give a pH of 7.5 and the solution was dialysed154for 3 days at 4°C against distilled deionized water and then freeze dried. For both a- and iccasein, analysis by SDS-PAGE showed >85% purity.2. Preparation of buffer and urea solutionsA 0.1 M sodium phosphate, pH 7.5±0.1 buffer was used for all proteins. This wasprepared according to the method of Dennison (1988). For Raman spectroscopy and NMRexperiments, the buffer was made in D20 and pH adjustment was carried out with concentrated(30%) NaOD solution. This buffer in D20 will be referred to as d-buffer. Urea solutions forRaman and NMR were made with deuterated urea in d-buffer (d-urea). No pH correction dueto isotope effects were made. A two point pH calibration, using pH 6 and 7 std. buffer solutions,was used to calibrate the pH meter. To denature the proteins to different levels, urea or d-ureaat four concentration of 2,4, 6 and 8 M were used. The concentration of the urea solutions wasroutinely checked by measuring refractive index and using the values reported by Sober (1970).To prevent decomposition of the urea solutions, with the resulting formation of cyanate, ureasolutions were kept frozen at -30°C until used. The storage times for the frozen urea solutionswere less than 3 weeks.3. Protein solution preparationA stock protein solution was routinely made by weighing =55 nig of protein in a 1.5 mlmicrocentrifuge tube and then adding 1 ml of the buffer or urea solution. After vortex mixingfor 30 sec and standing for a minimum of 60 nun, the solution was centrifuged for 10 miii at7,000 x g. From this stock solution, appropriate dilutions were made for analysis. Since someproteins tended to give solutions of high viscosity at this protein concentration (=5%), pipettingfrom the stock solution was carried out with a positive displacement automatic pipette(Microman, Gilson Medical Electronics, France). Protein concentration was estimated bymeasuring the absorbance at 280 nm and using the following extinction coefficients (absorptivity155of a 1% solution measured in a 1 cm light path at 280 nm): for cx- , 13- and ic-casein, 10.0, 4.5and 10.5, respectively (Eigel et al., 1984), for a-lactalbumin, 13-lactoglobulin and bovine serumalbumin, 20.1, 10.0 and 6.5, respectively, for ovalbumin, lysozyme, ovotransferrin and ovomucoid,7.5, 26, 11.1 and 4.55 (Sober, 1970) or by a microassay method based on the Biuret reactionusing serum albumin as standard (Sigma Diagnostic Procedure No. 541).B. SPECTROPHOTOMETRIC MEASUREMENTS (FOURTH DERIVATiVE SPECTROSCOPY)Ultraviolet (UV) absorbance in the 330-250 nm range was measured with a ShmadzuUV-160 double beam spectrophotometer (Shimadzu Co., Kyoto, Japan) with speed setting of slow(50 nm/mm). Protein concentration used was from 0.4 to 0.7 mg/ml. All spectra were recordedagainst a reference cell containing buffer or the appropriate urea solution. After the UVabsorbance spectra were measured, the fourth derivative spectra were calculated using the built-in program of the spectrophotometer. As recommended by Padros et al. (1982) a derivativewavelength difference W) value of 1.8 nm was used. This value depends on the measuringwavelength range and the setting of the parameter N. For a range 100 nm, a setting of N=3was used. The resulting data were then transferred to a personal microcomputer for analysis.Measurements were done in duplicate and mean values are reported. As expected for this typeof instrumental measurements, they were found highly reproducible.C. TWO-DIMENSIONAL FLUORESCENCE SPECTROSCOPYThe steady-state emission spectra from 300-450 nm at different excitation wavelengths(270,275,280,290 and 297 nm) were recorded on a Shimadzu RF-540 spectrofluorophotometer(Shimadzu Co., Kyoto, Japan) at room temperature. The protein concentration used was in therange 0.025-0.05 mg/mi. These low values were used to prevent quenching effects. Excitation andemission slits were each set at 5 nm, scan speed at “slow” (15 nm/mm) and a samplingwavelength interval of 0.5 nm. Quartz cuvettes of 10 mm pathlength (Helima Ltd., Germany)156were used. Spectra were corrected for fluorescence of buffer blanks. The blank corrected spectrawere converted to microcomputer files by digitizing using the system SigmaScanTMversion 3.90(Jandel Scientific, Corte Madera, CA). The spectra at each excitation wavelength were fitted toa log normal band shape in order to estimate peak maximum and peak width using thespectroscopic/chromatographic software system LabCalc (Galactic Industries Corp., Salem,NH). Three dimensional plots were generated using SYGRAPH version 5.01 (Wilkinson, 1990a).Measurements were done in duplicate and mean values are reportedD. RAMAN SPECTROSCOPYNon-resonance Raman spectra were recorded on a JASCO Model NR-1100 laser Ramanspectrophotometer (Japan Spectronic Co., Tokyo, Japan). An argon-ion laser (Spectra-PhysicsModel 168B) tuned at 488 nm was used as the excitation source. The Raman scattering ofsamples placed in glass haematocrit capillaries (Nichiden-Rika Glass Co., Ltd) was collectedusing standard 90° geometry. Raman spectra were measured at ambient temperature withouttemperature control under the following conditions: 200 mW laser power setting at power supply,2 mm slit height, sampling speed 120 cm miii’ with data collected every cm’, with a spectraresolution of 5.0 cm’ at 19,000 ciii’. Protein concentration was 50 mg/mI in d-buffer or in durea solution. In order to increase the signal to noise ratio, a minimum of 8 scans per samplewere measured and co-added. Before measuring the protein samples, wavelength calibration wasroutinely performed by measuring the Raman spectrum of a saturated solution of KNO3 inwater, and if necessary correcting the wavelength scale of the instrument to have the KNO3Rarnan peak at 1050±2 cm’. All computation on the recorded spectra were performed by usingthe software LabCalc’ (Galactic Industries Corp., Salem, NH) with Squares Tools (SpectrumSquare Associate, Ithaca, NY). Measurements were done in duplicate and mean values arereported.157E. NUCLEAR MAGNETIC RESONANCE (NMR) SPECTROSCOPY1. One dimensional NMR of native and denatured proteinsProton magnetic resonance spectra were obtained on a Varian Fourier Transform-HighResolution spectrometer operating at a frequency of 300 MHz using sodium 2,2-dimethyl-2-silylpentane-5-sulphonate as an external frequency reference. In order to improve signal to noiseratio a minimum of 128 free-induction decays (FID) were signal averaged, and Fouriertransformation was used to convert the resulting FID to a spectrum. Total acquisition time was20-30 minutes. As recommended by Radford et al. (1991), dioxane, at a final concentration of 1mM was used as an internal frequency and intensity reference. Peak positions are indicated inparts per million (ppm) in relation to the methyl protons of the external reference aftercorrecting the spectrum, if necessary, to obtain a resonance positiän of the dioxane peaks of3.743 ppm (Radford et al., 1991). Spectra were measured at ambient temperature (20-23°C) orat 80°C after 5 mm of equilibration in the spectrometer. Protein concentration was 15 mg/mIin d-buffer or d-urea. NMR. tubes of 5 mm o.d. (Wilman Glass, Co., Vineland, NJ) and a samplevolume of 0.6 ml were used. For each protein, the preparation of the reduced-urea denaturedsample was as follows: 0.1 ml of the stock protein solution (50 mg/mi) in 8 Md-urea solution wascombined with 10 of f-mercaptoethanol (ME) and 0.6 ml of 8 M d-urea solution. This mixturewas heated in boiling water ( 95°C) for 10 mm and cooled with tap water for 10 mm. Then, 25p1 of 10% v/v of solution of dioxane in 1)20 was added and the mixture was vortex mixed for 15sec. From this solution, 0.6 ml were transferred to the NMR tubes. The same procedure was usedfor the preparation of the non-reduced samples, with neither addition of ME nor heat treatment.The instruments’ computer software was used for Fourier transformation, baseline correctionand estimation of chemical shifts. From an expanded plot of the aliphatic and aromatic areas,peak widths at different peak heights, were measured manually.1582. Spin diffusion experiments of native proteinsFor the spin diffusion experiment, the cross saturation method reported by Akasaka(1983) was used. Only samples in d-buffer (i.e., native proteins) were analyzed. The samplepreparation procedure described above was used with no addition of 3-ME and no heattreatment. After some preliminary experiments and with advice of staff in the NMR Laboratoryof the Chemistry Department, the pulse sequence reported in Fig. 4 was used. The instrument’scomputer software was used for data manipulation (i.e., Fourier transformation and integrationof resonances). Within each protein, the normal and cross saturated spectrum was measured i