@prefix vivo: . @prefix edm: . @prefix ns0: . @prefix dcterms: . @prefix skos: . vivo:departmentOrSchool "Applied Science, Faculty of"@en, "Chemical and Biological Engineering, Department of"@en ; edm:dataProvider "DSpace"@en ; ns0:degreeCampus "UBCV"@en ; dcterms:creator "Singhal, Anupam"@en ; dcterms:issued "2013-04-30T00:00:00"@en, "2012"@en ; vivo:relatedDegree "Doctor of Philosophy - PhD"@en ; ns0:degreeGrantor "University of British Columbia"@en ; dcterms:description """This thesis describes the development of novel microfluidic technologies for rapid, high-­‐throughput screening and selection of monoclonal antibodies (mAbs) from single cells. Microfluidic devices were used to compartmentalize single antibody-­‐ secreting cells (ASCs) in small-­‐volume chambers (i.e. hundreds of picoliters to nanoliters) in order to concentrate secreted mAbs for measurement of antigen binding kinetics and affinities using a novel microfluidic fluorescence bead assay. Microfluidic single-­‐cell antibody screening was performed on ASCs harvested from antigen-­‐ immunized mice and purified by fluorescence-­‐activated cell sorting (FACS). Following microfluidic selection of ASCs producing antigen-­‐specific mAbs, ASCs were sequentially recovered from the microfluidic device and subjected to single-­‐cell RT-­‐PCR to amplify the antibody-­‐encoding heavy and light chain genes. Antibody genes for selected high-­‐ affinity mAbs are sequenced and cloned into expression vectors for recombinant production in mammalian cell lines. Nearly 200 high-­‐affinity mouse mAbs to the model antigen hen egg lysozyme (HEL) were selected as a validation of this technology, representing a ten-­‐fold increase in the number of high affinity anti-­‐HEL mAbs previously selected using single-­‐cell micro-­‐technologies and the traditional hybridoma approach. Microfluidic single-­‐cell mAb screening also yielded important insights into affinity maturation, immuno-­‐dominance, and antibody stereotypy in the adaptive immune system. By circumventing time-­‐consuming limiting dilution and clonal expansion in the hybridoma approach, microfluidic single-­‐cell screening will enable selection of mAbs from other animal species (e.g. rabbits, humans) for both therapeutic and research applications."""@en ; edm:aggregatedCHO "https://circle.library.ubc.ca/rest/handle/2429/43575?expand=metadata"@en ; skos:note """ MICROFLUIDIC  TECHNOLOGIES  FOR  RAPID,  HIGH-­‐ THROUGHPUT  SCREENING  AND  SELECTION     OF  MONOCLONAL  ANTIBODIES  FROM  SINGLE  CELLS     by  Anupam  Singhal    B.A.Sc.,  The  University  of  Toronto,  2004    A  THESIS  SUBMITTED  IN  PARTIAL  FULFILLMENT  OF  THE  REQUIREMENTS  FOR  THE  DEGREE  OF    DOCTOR  OF  PHILOSOPHY  in  THE  FACULTY  OF  GRADUATE  STUDIES  (Chemical  and  Biological  Engineering)     THE  UNIVERSITY  OF  BRITISH  COLUMBIA  (Vancouver)    November  2012       ©  Anupam  Singhal,  2012   ii Abstract  This   thesis   describes   the   development   of   novel   microfluidic   technologies   for  rapid,  high-­‐throughput  screening  and  selection  of  monoclonal  antibodies  (mAbs)  from  single   cells.   Microfluidic   devices   were   used   to   compartmentalize   single   antibody-­‐secreting   cells   (ASCs)   in   small-­‐volume   chambers   (i.e.   hundreds   of   picoliters   to  nanoliters)  in  order  to  concentrate  secreted  mAbs  for  measurement  of  antigen  binding  kinetics  and  affinities  using  a  novel  microfluidic   fluorescence  bead  assay.  Microfluidic  single-­‐cell   antibody   screening   was   performed   on   ASCs   harvested   from   antigen-­‐immunized  mice  and  purified  by   fluorescence-­‐activated  cell   sorting  (FACS).  Following  microfluidic  selection  of  ASCs  producing  antigen-­‐specific  mAbs,  ASCs  were  sequentially  recovered  from  the  microfluidic  device  and  subjected  to  single-­‐cell  RT-­‐PCR  to  amplify  the  antibody-­‐encoding  heavy  and   light   chain  genes.  Antibody  genes   for   selected  high-­‐affinity   mAbs   are   sequenced   and   cloned   into   expression   vectors   for   recombinant  production  in  mammalian  cell  lines.  Nearly  200  high-­‐affinity  mouse  mAbs  to  the  model  antigen   hen   egg   lysozyme   (HEL)   were   selected   as   a   validation   of   this   technology,  representing   a   ten-­‐fold   increase   in   the   number   of   high   affinity   anti-­‐HEL   mAbs  previously  selected  using  single-­‐cell  micro-­‐technologies  and  the  traditional  hybridoma  approach.  Microfluidic   single-­‐cell  mAb   screening   also   yielded   important   insights   into  affinity   maturation,   immuno-­‐dominance,   and   antibody   stereotypy   in   the   adaptive  immune   system.   By   circumventing   time-­‐consuming   limiting   dilution   and   clonal  expansion   in   the   hybridoma   approach,   microfluidic   single-­‐cell   screening   will   enable  selection  of  mAbs  from  other  animal  species  (e.g.  rabbits,  humans)  for  both  therapeutic  and  research  applications.   iii Preface  I   conducted   the   vast  majority   of   the  work   described   in   this   thesis,  which  was  jointly  designed  by  Dr.  Carl  Hansen,  Dr.  Charles  Haynes,  Dr.  John  Schrader  and  myself.    All  animal  work  described  in  this  thesis  was  conducted  in  collaboration  with  the  laboratory   of   Dr.   John   Schrader   at   the   Biomedical   Research   Centre   located   at   the  University   of   British   Columbia   (UBC).   Animal   immunizations,   harvesting   and  purification   of   antibody-­‐secreting   cells,   and   ELISPOT   assays   were   performed   by   Dr.  Welson  Wang,   a   postdoctoral   fellow   in   Dr.   Schrader’s   laboratory.   Andy   Johnson   and  Justin  Wong  performed  all  FACS  sorting  at  the  UBC  Biomedical  Research  Centre  and  Life  Sciences  Institute.  Dr.  Michael  Williams  at  the  Biomedical  Research  Centre  also  assisted  in   the   preparation   of   fluorescent   protein   conjugates   and   provided   training   in   cell  culture  methods.  All   LabView   software   (Appendix   B)   was   developed   in   conjunction   with   an  undergraduate   engineering   physics   student,   Daniel   Da   Costa,   during   an   8-­‐month  internship  in  the  Hansen  lab.  When  Dan  started  his  co-­‐op  internship,  he  was  provided  with  sample  LabView  code   that   I  had  previously  developed  and  used   to  automate   the  microscope  hardware.  Dan  re-­‐wrote  the  majority  of  this  software  to  produce  a  robust  platform   for   screening   many   chambers   in   a   single   device   using   custom   autofocus  algorithms   that   we   developed   together.   Dan   also   produced   completely   new   LabView  software   (Appendix   B.2)   to   perform   semi-­‐automated   analysis   of   fluorescence   images  and  binding  kinetics.  This  replaced  manual  methods  of  image  analysis  that  I  previously  used.   Finally,   Daniel   assisted   in   fabrication   of  microfluidic   devices   used   in   this  work.  Dan   also   deserves   credit   for   both   the   bead   trap/filter   and   novel  multiplexer   designs   iv that   were   incorporated   into   the   final   microfluidic   device   architecture.   Dan’s  contributions   to   this   project   are   described   in   his   co-­‐op   report,   entitled   “Microfluidic  Technology  for  Screening  and  Selection  of  Monoclonal  Antibodies  from  Single  Cells”.1  Most  of  Chapter  2  was  previously  published  as  a  manuscript  in  the  ACS  journal  Analytical   Chemistry,   entitled   “Microfluidic  measurement  of   antibody-­‐antigen  binding  kinetics  from  low-­‐abundance  samples  and  single  cells”  by  Singhal,  A.,  Haynes,  C.A.  and  Hansen,  C.L.  82(20):8671-­‐9  (2010).  The  work  described  in  this  publication  was  jointly  designed  by  all  three  co-­‐authors,  and  I  performed  all  of  the  experiments  and  wrote  the  manuscript.  Chapters  3  and  4  serve  the  basis  for  manuscripts  currently  in  preparation.     All  work  conducted  and  described  in  this  thesis  was  approved  by  the  Research  Board  of  Ethics  at   the  University  of  British  Columbia   (Ethics  Certificate  Number  A08-­‐0493).       v Table  of  Contents   Abstract  .............................................................................................................................................  ii   Preface  .............................................................................................................................................  iii   Table  of  Contents  ............................................................................................................................  v   List  of  Tables  ....................................................................................................................................  x   List  of  Figures  ...............................................................................................................................  xii   Acknowledgements  ..............................................................................................................  xxviii   Dedication  ...................................................................................................................................  xxix   Chapter    1:  Introduction  ..............................................................................................................  1  1.1   Antibodies  and  the  Vertebrate  Adaptive  Immune  System  ..............................................  2  1.2   Methods  for  Antibody  Screening  and  Selection  ................................................................  10  1.2.1   Single-­‐Cell  Methods  for  Antibody  Selection  ...............................................................  14  1.3   Microfluidics   –   An   Enabling   Technology   for   Screening   and   Selection   of  Antibodies  from  Single  Cells  .................................................................................................................  17  1.4   Aims  of  this  Thesis  .........................................................................................................................  24   Chapter    2:  Microfluidic  Measurement  of  Antibody-­‐Antigen  Binding  Kinetics  from   Low  Abundance  Samples  ..........................................................................................................  25  2.1   Antibody-­‐Antigen  Binding  Properties:  Binding  Affinity,  Selectivity  and  Kinetics  ................................................................................................................................................................  25  2.1.1   Mathematical  Model  for  Antibody-­‐Antigen  Binding  ..............................................  26  2.2   Methods  and  Parameters  for  Antibody  Screening  and  Selection  ..............................  30  2.3   Materials  and  Methods  ................................................................................................................  32  2.3.1   Microfluidic  Device  Fabrication  and  Control  ............................................................  32   vi 2.3.2   Reagent  Preparation  ............................................................................................................  33  2.3.3   Fluorescence  Microscopy  ...................................................................................................  34  2.3.4   Cell  Culture  ...............................................................................................................................  35  2.3.5   Assay  Operation  .....................................................................................................................  35  2.3.6   Data  Analysis  ...........................................................................................................................  40  2.4   Results  .................................................................................................................................................  41  2.4.1   Microfluidic   Fluorescence   Bead  Measurements   Reflect   Intrinsic   Antibody-­‐Antigen  Binding  Kinetics.  ..................................................................................................................  47  2.4.2   Microfluidic   Fluorescence   Bead   Measurements   Exhibit   Low   Detection  Limits  and  Minimal  Sample  Consumption.  ................................................................................  55  2.4.3   Measurement   of   Binding   Kinetics   of   Antigen   and   Antibody   Secreted   from  Single  Cells.  ..............................................................................................................................................  58  2.4.4   Extensions  of  the  Microfluidic  Fluorescence  Bead  Assay  ....................................  61  2.4.4.1   Direct   Measurements   of   Antibody-­‐Antigen   Equilibrium   Binding  Affinities  ..............................................................................................................................................  61  2.4.4.2   Measurement   of   Antibody-­‐Antigen   Binding   Kinetics   and   Selectivity  Using  Optical  and  Spatial  Multiplexing.  .................................................................................  64  2.5   Conclusion  .........................................................................................................................................  66   Chapter    3:  Microfluidic  Single-­‐cell  Sorting,  Recovery,  and  Robust  Amplification  of   Antibody  Heavy  and  Light  Chain  Genes  from  Single  Cells  .............................................  67  3.1   Structure  of  Antibody  Heavy  and  Light  Chain  Genes  .....................................................  69  3.2   Primer  Design  for  Reverse-­‐Transcription  Polymerase  Chain  Reaction  (RT-­‐PCR)  of  Antibody  Heavy  and  Light  Chain  Genes  .....................................................................................  71   vii 3.3   Materials  and  Methods  ................................................................................................................  74  3.3.1   Cell  Culture  ...............................................................................................................................  74  3.3.2   Cell  Lysis  and  mRNA  Purification  ...................................................................................  75  3.3.3   RT-­‐PCR  Reaction  Mix  and  Cycling  Conditions  ..........................................................  77  3.3.4   Analysis,  Purification,  and  Sequencing  of  RT-­‐PCR  Products  ..............................  79  3.3.5   Microfluidic  Single-­‐Cell  Sorting,  Lysis,  and  Recovery  ............................................  80  3.4   Results  .................................................................................................................................................  82  3.4.1   RT-­‐PCR   Optimization   for   Single-­‐Cell   Amplification   of   Mouse   Heavy   and  Light  Chain  Antibody  Genes  .............................................................................................................  82  3.4.2   Microfluidic  Single-­‐Cell  Sorting,  Lysis,  and  Recovery  ............................................  91  3.5   Conclusions  .......................................................................................................................................  95   Chapter     4:   Rapid,   High-­‐Throughput   Screening   and   Selection   of   High   Affinity   Monoclonal  Antibodies  from  Single  Antibody-­‐Secreting  Cells  ....................................  97  4.1   Experimental  Methods  ................................................................................................................  99  4.1.1   Mouse  Immunization,  Harvesting  and  Purification  of  ASCs  ...............................  99  4.1.2   Reagent  Preparation  ..........................................................................................................  100  4.1.3   Microfluidic  Device  Design  and  Operation  ...............................................................  101  4.1.4   Sequencing   of   Antibody   Heavy   and   Light   Chain   Genes   and   Recombinant  Expression  of  Selected  mAbs  .........................................................................................................  113  4.2   Results  ...............................................................................................................................................  115  4.2.1   Kinetic  Screening  and  RT-­‐PCR  Amplification  of  Antibody  Genes  from  Single  Hybridoma  Cells  ..................................................................................................................................  115   viii 4.2.2   Microfluidic  Screening  and  Selection  of  mAbs  from  ASCs  Purified  from  Mice  Immunized  with  HEL  ........................................................................................................................  116  4.2.3   Antibody-­‐Antigen  Binding  Kinetics  and  Affinities  of  Novel  Anti-­‐HEL  Mouse  mAbs    .....................................................................................................................................................  120  4.2.4   Analysis  of  Heavy  and  Light  Chain  Genes  from  Novel  Anti-­‐HEL  Mouse  mAbs  .....................................................................................................................................................  127  4.2.5   Cloning  and  Expression  of  Novel  Anti-­‐HEL  Mouse  mAbs  ..................................  141  4.3   Conclusions  .....................................................................................................................................  142   Chapter    5:  Conclusions  and  Future  Work  ........................................................................  148  5.1   Selection  of  mAbs  for  Multiple  Functional  Binding  Properties  ...............................  149  5.2   Increasing  Capacity  to  Screen  Larger  Numbers  of  ASCs  .............................................  151  5.3   Selection  of  mAbs   from  Other  Animal   Species   (e.g.  Humans,  Rabbits,   etc.)   and  Cell  Types  ....................................................................................................................................................  156  5.4   Other  Insights  into  the  Adaptive  Immune  System  .........................................................  157   References  ...................................................................................................................................  159   Appendices  ..................................................................................................................................  171  Appendix  A  -­‐  Primer  Designs  for  Amplifying  Mouse  Antibody  Genes  .............................  171  A.1   Highly  Degenerate  Primer  Set   for  Amplifying  Mouse  Heavy  and  Light  Chain  Antibody  Genes.  ..................................................................................................................................  171  A.2   Low   Degeneracy   Nested   PCR   Primer   Set   for   Amplifying   Mouse   Heavy   and  Light  Chain  Antibody  Genes.    (continued  on  next  page)  ...................................................  172  A.3   (continued  from  previous  page)  Low  Degeneracy  Nested  PCR  Primer  Set  for  Amplifying  Mouse  Heavy  and  Light  Chain  Antibody  Genes.  ............................................  173   ix Appendix  B  -­‐  Labview  Software  for  Hardware  Automation  and  Image  Analysis  ........  174  B.1   Custom  LabView  Software  to  Automate  CCD  Camera,  Brightfield  Illumination,  Microscope  Stage  Control,  and  Microfluidic  Valve  Operation.  .......................................  174  B.2   Custom  LabView  Software  for  Automated  Analysis  of  Images.  ..........................  175         x List  of  Tables    Table  1.1        Diversity  of  human  antibodies  generated  by  combinatorial  (imprecise)  gene  recombination  and  heavy/light  chain  pairing.  Antibody  genes  are  further  diversified  by  somatic  hypermutation.  .................................................................................................................................  5  Table  2.1     Analytical   solutions   to   first-­‐order   differential   equations   describing  antibody-­‐antigen  binding  under  the  condition  that  [Ab]  ≈[Ab]t=0  >>  [Ag]0.  .........................  27  Table  2.2     Antibody-­‐antigen   binding   kinetics   measured   using   the   microfluidic  fluorescence  bead  assay.  .............................................................................................................................  43  Table  3.1     Number  of  V  region  gene  segments  that  encode  human  and  mouse  antibody  heavy  and   light   chains.  A   range  of  values  provided   for   some  gene   segments   to   reflect  differences   in   the   published   literature.   Data   selected   from   Janeway,2   Arnaout   et   al.,18  and  Tiller  et  al.19,58  .........................................................................................................................................  71  Table  3.2     One-­‐step  RT-­‐PCR  cycling  protocol.  ................................................................................  78  Table  4.1     Antibody-­‐antigen   binding   kinetics   and   affinities   from   single   D1.3   and  HyHEL-­‐5  hybridoma  cells  measured  by  a  microfluidic   fluorescence  bead  assay  using  a  HEL-­‐Dylight488   fluorescent   conjugate.     Results   represent   the   average   and   standard  deviation  of  replicate  measurements  performed  on  multiple  distinct  D1.3  (n  =  30)  and  HyHEL-­‐5  (n  =  5)  cells.  .................................................................................................................................  116  Table  4.2     Measured   antibody-­‐antigen   binding   kinetics   and   affinities   from   over   70  anti-­‐HEL  mAbs  selected  using  the  microfluidic  single-­‐cell  screening  approach.  .............  124  Table  4.3     Range  of  kinetic  and  equilibrium  rate  constants  for  anti-­‐HEL  mAbs  selected  using   the   microfluidic   single-­‐cell   screening   approach   from   three   different   HEL-­‐immunized  BALB/c  mice.  .........................................................................................................................  124   xi Table  4.4   Binding  kinetics,  affinities,  VDJ  gene  usage  and  number  of  amino  acid  (AA)  substitutions   in   kappa   and   heavy   chain   gene   sequences   for   select   subset   of   selected  anti-­‐HEL  mAbs.    (n/a  =  not  amplified,   i.e.  the  corresponding  kappa  or  light  chain  gene  did  not  amplify  by  RT-­‐PCR).  [continued  on  next  page]  ...............................................................  131  Table  4.5     Binding  kinetics,  affinities,  VDJ  gene  usage  and  CDR  sequences  for  anti-­‐HEL  mAbs   encoded   by   the   Vκ5-­‐43   and   VH3-­‐8   genes.     (n/a   =   not   amplified,   n/r   =   not  reported).  mAbs   are   listed   in   order   of   binding   affinity   to  HEL   (highest   affinity   at   the  top).  mAbs  marked  with  an  asterisk  are  encoded  by  both  Vκ5-­‐43  kappa  and  VH3-­‐8  heavy  chains.   The   heavy   chain   diversity   (D)   region   of   some   mAbs   was   not   identified.  [continued  on  next  page]  ..........................................................................................................................  138  Table  4.6   Nucleotide  sequences  of  the  heavy  chain  junction  region  for  anti-­‐HEL  mAbs  encoded  by  the  VH3-­‐8  gene.    mAbs  marked  with  an  asterisk  also  utilize  the  same  kappa  chain  gene  (Vκ5-­‐43).  ...................................................................................................................................  140  Table  4.7   Comparison   of   binding   kinetics   of   M1_R06C01   anti-­‐HEL   mouse   mAb  selected  from  single  ASC  and  produced  by  recombinant  expression  in  mammalian  cells.  Reported   error   represents   the   calculated   standard   deviation   of   multiple   replicate  measurements.  Values  measured  only  once  are  reported  without  error  bars.  ................  142     xii List  of  Figures  Figure  1.1           Microfluidic  pipeline  for  single-­‐cell  antibody  screening  and  selection.  ......  2  Figure   1.2       A   schematic   drawing   (A)   and   crystal   structure   (B)   of   the   antibody   IgG  molecule.  Figures  reproduced  from  the  following  websites:  .........................................................  3  Figure  1.3     Antibody   heavy   and   light   chains   can   be   divided   into   3   hypervariable  regions   (complementary-­‐determining   regions,   CDRs)   and   4   framework   regions.  Comparison   of   antibody   heavy   and   light   chains   reveals   that   sequence   differences   are  largely  concentrated  to  the  CDR  regions.  The  heavy  and  light  CDR3  regions,  which  are  the   sites   of   V(D)J   recombination,   exhibit   the   greatest   sequence   diversity.   Figure  reproduced   from   Janeway’s   Immunobiology   with   permission   from   Garland   Science   /  Taylor  and  Francis  LLC,  2011.2  ...................................................................................................................  6  Figure  1.4     The  hypervariable  (CDR)  regions  of  both  antibody  heavy  and   light  chains  lie   in   the  antigen-­‐binding  domain  of   the   folded  antibody  molecule.  Figure  reproduced  from   Janeway’s   Immunobiology   with   permission   from   Garland   Science   /   Taylor   and  Francis  LLC,  2011.2  ...........................................................................................................................................  7  Figure  1.5     Generation  of  antigen-­‐specific  antibodies  by  the  adaptive  immune  system.    Antigen   binds   surface-­‐displayed   antibody   on   a   subset   of   naïve   B   cells,   which   are  subsequently   stimulated   to   proliferate   (clonal   selection).   Somatic   mutations   in  proliferating   cells   alter   the   expressed   antibodies,   and   the   clonal   selection   process   is  iterated   to   generate   antibodies   that   bind   antigen   with   high   affinity   and   specificity  (affinity   maturation).   This   process   creates   two   cell-­‐types:   plasma   cells   that   secrete  soluble  antibodies  into  the  blood  and  other  tissues  and  memory  B  cells  that  accelerate  the  immune  response  to  host  re-­‐infection  with  the  same  antigen.  .............................................  9   xiii Figure  1.6     Hybridoma   method   for   producing   antibodies   of   a   defined   specificity.    Antibody-­‐secreting   cells   (ASCs)   from  animals   (e.g.  mice)   immunized  with   antigen   are  fused  to  cancer  (myeloma)  cells   in  order  to  generate   immortalized  ASCs  (hybridoma).  The   hybridoma   cells   are   screened   by   limiting   dilution   to   identify   stable   clones   that  secrete  antigen-­‐specific  mAbs.  Figure  adapted  with  permission  from  Joyce  et  al  (Nature  Publishing  Group,  2010).30  ........................................................................................................................  11  Figure  1.7     Screening  and  selection  of  synthetic  antibody  libraries.  (A)  Antibody  genes  are  synthesized  or  amplified   from  B  cells  and  diversified  by   in  vitro  mutagenesis   (e.g.  error-­‐prone  PCR  or  DNA  shuffling).  The  antibody  genes  are  expressed  on  the  surface  of  a   vector   (B)   and   panned   with   antigen   in   order   to   select   antigen-­‐specific   mAbs.   The  process   is   iterated  several   times   in  order   to   increase  affinity  and/or  specificity  of   the  selected   antibodies.   Figure   reproduced   from   Hoogenboom   with   permission   from  Nature  Publishing  Group,  2005.46  ...........................................................................................................  14  Figure  1.8     Selected   Lymphocyte   Antibody   Method   (SLAM)   for   identifying   antigen-­‐specific  mAbs  from  single  antibody-­‐secreting  cells  (ASCs).  ASCs  are  mixed  with  antigen-­‐coated   sheep   red   blood   cells   (SRBCs)   and   blood   complement   on   a   glass   slide   and  incubated  at  37°C  for  an  hour.  Binding  of  secreted  antibodies  to  antigen  triggers  lysis  of  red  blood  cells  in  the  area  around  each  ASC,  thus  forming  visible  “plaques”  on  a  sealed  glass  slide.  ASCs  are  manually  recovered  and  subjected  to  single-­‐cell  RT-­‐PCR  followed  by  cloning  and  expression  of  antibody  genes.  Figure  reproduced  with  permission  from  Babcook  et  al.  (PNAS,  1996).50  .................................................................................................................  17  Figure  1.9     Fabrication   of   single   (A)   and   multilayer   (B   &   c)   polydimethylsiloxane  (PDMS)  microfluidic  devices.  (A)  Soft   lithography.  Replica  molding  of   lithographically-­‐ xiv patterned  master  molds  using  PDMS  liquid  polymer.    After  the  PDMS  polymer  is  cured  into   a   solid   substrate,   it   is   removed   from   the   master   mold,   input/output   ports   are  manually  punched  through  the  device  and  the  microfluidic  channels  are  sealed  against  a  glass  slide.    (B)  Multilayer  soft  lithography  (MSL).    Replica  molding  is  performed  using  multiple  master  molds,   and   the   resulting  PDMS   layers   are   aligned   and  bonded   into   a  monolithic  structure.  (C)  Pressure  applied  to  a  fluid-­‐filled  channel  on  the  control  layer  deflects   the   elastomeric  membrane   separating   it   from   the   channel   on   the   flow   layer,  thus  closing   the  reversible  valve  structure.  Figures   (A)  adapted  with  permission   from  McDonald   et   al.   (Electrophoresis,   2000),85   (B)   reproduced   from  Unger   et   al.   (Science,  2000),86    and  (C)  courtesy  of  C.  Hansen.  ..............................................................................................  19  Figure  1.10     Integration   of   multiple   microfluidic   valves   into   higher-­‐order   fluidic  structures   (pumps,   fluidic   mixers,   and   fluidic   multiplexing   structures)   in   single  microfluidic  devices  fabricated  by  multilayer  soft  lithography.73,86,87  Pumps  are  used  to  meter  precise  volumes  of  fluidic  reagents,  ranging  from  100  pL  to  1  nL.  Viscous  forces  dominate   inertial   forces   for   fluid   flow   in  microfluidic  channels  (i.e.   laminar   flow),  and  thus   fluidic   mixers   are   required   to   accelerate   the   mixing   of   chemical   reagents.    Multiplexing   structures   facilitate   the   selection   of   one   or  more   fluidic   reagents  with   a  reduced   number   of   valves   (2logN)   compared   to   the   number   of   reagent   inputs   (N).  Figure  courtesy  of  C.  Hansen  with  permission.  .................................................................................  20  Figure  1.11     Concentration  enhancement  in  small-­‐volume  chambers  (<1  nL)  enables  detection   of   antibodies   secreted   by   single   antibody-­‐secreting   cells   (ASCs).   ASCs  harvested   from   immunized   animals   typically   survive   for  ~1-­‐2   days   in   culture.     Thus,   xv mAbs   from   single   ASCs   cannot   be   detected   in   cell-­‐culture   plates   using   standard  laboratory  tests  (>1  nM  detection  limit).  ............................................................................................  21  Figure  1.12     Methods   for   screening   antibodies   secreted   by   single   cells   using  micro-­‐fabricated  wells  (A  and  B)  and  droplet  encapsulation  (C).  (A)  Microengraving  method.    Single   cells   in   PDMS   micro-­‐wells   secrete   antibodies   that   are   captured   on   a   “printed  microarray”   that   is   imaged   after   incubation   with   fluorescently-­‐labeled   antigen.   (B)  ISAAC   method   (see   text   for   details).     (C)   Schematic   (above)   and   microscope   images  (below)   of   microfluidic   devices   to   encapsulate   single   cells   in   water-­‐in-­‐oil   emulsion  droplets,   incubate   and   detect   secreted   antibodies.   Scale   bars   are   100   μm.   Figures  reproduced  with  permission  from  Love  et  al.  (Nature  Publishing  Group,  2006)  (A)14,  Jin  et   al.   (Nature   Publishing   Group,   2009)   (B)51,   and   Köster   et   al.   (Lab   on   a   Chip,   2008)  (C)91……………..    .................................................................................................................................................  23  Figure   2.1         Graphical   depiction   of   first-­‐order   antibody-­‐antigen   binding   kinetics.  Concentration  of  antibody-­‐antigen  complex   is  on  the  y-­‐axis,  whereas   time   is  on  the  x-­‐axis.   Antibody-­‐antigen   complex   follows   bimolecular   exponential   association   kinetics  during   the   association   phase,   and   first-­‐order   exponential   kinetics   during   the  dissociation  phase.  Equations  describing  the  rates  of  growth  and  decay  in  concentration  of  antibody-­‐antigen  complex  are  presented  in  Table  2.1.  ............................................................  28  Figure  2.2        Microfluidic  fluorescence  bead  measurements  of  antibody-­‐antigen  binding  kinetics.     (A)   Device   schematic   showing   control   channels   (orange)   for   selecting   six  reagent   inlets  (blue)  and  actuating  sieve  valves  on  the  reagent  outlet  channel  (green).    (B)  Microscope   image  of   device  with   food   coloring   to   visualize  distinct   reagent   inlets  (yellow   and   green)   and   control   channels   (red).     (Insets)   Brightfield   (top)   and   xvi fluorescence   (bottom)   images   of   beads   trapped   using   sieve   valves   at   20X   and   100X  magnification,  respectively.  [continued  on  next  page]  ..................................................................  36  Figure  2.3        Antibody-­‐antigen  association  kinetics  measured  from  multiple  beads   in  a  single  field-­‐of-­‐view  (FOV).    In  this  experiment,  fluorescently  labeled  hen  egg  lysozyme  is  binding  bead-­‐immobilized  anti-­‐HEL  D1.3  mouse  mAb.    Reported  error   represents   the  calculated   standard   deviation   from   multiple   replicate   measurements.   Dissociation  kinetics  measured  on  multiple  beads  in  a  single  FOV  were  also  consistent  to  within  20%    (data  not  shown).  ...........................................................................................................................................  41  Figure  2.4        Microfluidic  fluorescence  bead  measurements  of  antibody-­‐antigen  binding  kinetics.     Direct   fluorescent  measurements   of   association   and   dissociation   kinetics   of  (A)  D1.3  mAb   and  HEL-­‐Dylight488   conjugate,   (B)  HyHEL-­‐5  mAb   and  HEL-­‐Dylight488  conjugate,  (C)  LGB-­‐1  mAb  and  enhanced  green  fluorescent  protein  (EGFP).  (D)  Indirect  measurement   of   dissociation   kinetics   of   D1.3   mAb   and   HEL   using   HEL-­‐Dylight488  conjugate.   Solid   lines   represent   experimental   fits   using   mass-­‐action   equations  (equations   2.7a-­‐c).   Reported   error   represents   the   calculated   standard   deviation   of  multiple   replicate   measurements.     Adapted   with   permission   from   Singhal   et   al.  (American  Chemical  Society,  2010).112  .................................................................................................  44  Figure   2.5         Effect   of   fluorophore   stability   on   measured   antibody-­‐antigen   binding  kinetics.     (A)  Photobleaching  rates  of   fluorescent  dye  molecules  under  100W  Hg  lamp  illumination   using   100X   oil-­‐immersion   objective   (NA   1.30).   (B)   Effect   of   fluorescent  exposure   times   on   measured   association   kinetics   of   D1.3   mAb   and   HEL-­‐Dylight488.  Reported   error   represents   the   calculated   standard   deviation   of   multiple   replicate   xvii measurements.   Values  measured  only   once   are   reported  without   error  bars.  Adapted  with  permission  from  Singhal  et  al.  (American  Chemical  Society,  2010).112  .......................  49  Figure  2.6             Effect   of   different   bead   composition   and   capture   antibodies   on  measured   antibody-­‐antigen   binding   kinetics.  Measured   binding   kinetics   and   affinities  from   both   conditions   were   consistent   within   experimental   error   (see   Table   2.1).  Adapted  with  permission  from  Singhal  et  al.112  (American  Chemical  Society,  2010).112  51  Figure  2.7             Measured   dissociation   kinetics   of   mouse   mAb   from   antibody   capture  beads.   No   dissociation   of   D1.3   mAb-­‐Dylight488   conjugate   from   Rabbit   anti-­‐Ms   pAb  coated   beads   was   observed   over   3   days.   Reported   error   represents   the   calculated  standard  deviation  of  multiple  replicate  measurements.  Adapted  with  permission  from  Singhal  et  al.112  (American  Chemical  Society,  2010).112  ................................................................  52  Figure  2.8           Effect  of  antigen  re-­‐binding  on  measured  antibody-­‐antigen  dissociation  kinetics.    Dissociation  kinetics  of  D1.3  mAb  and  HEL-­‐Dylight488  conjugate  were  similar  both   in   the   presence   and   absence   of   a   large   concentration   of   competitive   antigen   (2  mg/mL  HEL).  Measured  binding  kinetics   from  both  conditions  were  consistent  within  experimental   error   (see   Table   2.1).   Adapted   with   permission   from   Singhal   et   al.112  (American  Chemical  Society,  2010).112  .................................................................................................  53  Figure  2.9           Effect   of   mass   transport   on   measured   antibody-­‐antigen   binding  kinetics.   Association   and   dissociation   kinetics   of   D1.3   mAb   and   HEL-­‐Dylight488  conjugate   were   similar   over   a   range   of   flow   rates   (~3-­‐14   µL/hr).   Fixed   error   bars  represent   the   calculated   ratio   of   the   standard   deviation   to   mean   value   of   measured  D1.3/HEL  kinetic   rate   constants   reported   in  Table  2.1   (25%  and  10%   for  kon   and  koff,   xviii respectively).   Adapted   with   permission   from   Singhal   et   al.112   (American   Chemical  Society,  2010).112  ............................................................................................................................................  54  Figure  2.10           Sensitivity   and   detection   limit   of   antibody-­‐antigen   binding   kinetics  measurements.  (A)  Measured  association  kinetics  of  D1.3  mAb-­‐Dylight488  conjugate  on  rabbit   anti-­‐mouse   pAb   coated   beads.   (Inset)   Schematic   of   bead   assay   for   measuring  binding  kinetics  of  fluorescently  labeled  mouse  mAb  and  rabbit  anti-­‐mouse  pAb  coated  beads.    Solid   lines  represent  experimental   fits  using  mass-­‐action  equations  (equations  2.7a-­‐c).   (B)   Association   kinetics   of   HEL-­‐Dylight488   conjugate   on   beads   with   varying  amounts  of  immobilized  D1.3  mAb  (shown  in  %  bead  coverage).  Bead  fluorescence  data  is   plotted   after   subtraction  of   bead   autofluorescence   at   time   zero.  No   change   in  bead  fluorescence   was   observed   when   beads   were   not   covered   with   D1.3  mAb   (0%   bead  coverage).   Reported   error   represents   the   calculated   standard   deviation   of   multiple  replicate  measurements.  [continued  on  next  page]  ........................................................................  56  Figure  2.11           Antibody-­‐antigen  binding  kinetics  measured  using   antibodies   secreted  from  a  single  cell.    (A)  Microscope  image  of  D1.3  hybridoma  cell  loaded  into  microfluidic  device  adjacent  to  rabbit  anti-­‐mouse  pAb  coated  beads  trapped  using  a  sieve  valve.    (B)    “Single-­‐cycle”  binding  kinetics  from  a  single  bead  containing  D1.3  mAbs  secreted  from  a  single  cell  and  subject  to  increasing  concentrations  of  HEL-­‐Dylight488  conjugate.  Solid  lines   represent   three   experimental   fits   using  mass-­‐action   equations   corresponding   to  each   concentration   of   fluorescently   labeled   HEL.   Reported   error   represents   the  calculated   standard   deviation   of   multiple   replicate   measurements.   Adapted   with  permission  from  Singhal  et  al.112  (American  Chemical  Society,  2010).112  ............................  60   xix Figure  2.12           Direct   measurement   of   equilibrium   dissociation   constants   by  measuring   equilibrium   bead   fluorescence   using   immobilized   D1.3   mAb   and   varying  concentrations   of   HEL-­‐Dylight488.   Solid   line   represents   experimental   fits   using   a  Langmuir   isotherm  equation.  Value  of  Kd   estimated  by   the  concentration  at  which   the  equilibrium   bead   fluorescence   was   equal   to   the   half-­‐maximal   value.   Adapted   with  permission  from  Singhal  et  al.112  (American  Chemical  Society,  2010).112  ............................  61  Figure  2.13           Simultaneous  screening  of  binding  kinetics  and  selectivity  of  multiple  antibody-­‐antigen   interactions  using  optical  and  spatial  multiplexing.     (A)  Schematic  of  assay   to   screen   binding   of  m  antibodies   on   distinct   beads   to  n   distinct   antigens   each  labeled  with  a  spectrally  unique  fluorophore.  (B)  False-­‐colored,  overlay  of  images  taken  with   distinct   fluorescence   filter   cubes   to   identify   anti-­‐lysozyme  mAbs   (red)   and   anti-­‐EGFP  mAbs   (green).   (C)   Measured   association   and   dissociation   kinetics   of   3   distinct  mAbs   (HyHEL-­‐5,   D1.3,   and   LGB-­‐1)   interacting   with   2   different   antigens   (HEL-­‐Dylight633   conjugate   and   EGFP).   Solid   lines   represent   experimental   fits   using  mass-­‐action   equations.   Adapted  with   permission   from  Singhal   et   al.112   (American  Chemical  Society,  2010).112    ...........................................................................................................................................  65  Figure  3.1           Antibody   heavy   and   light   chain   genes   are   constructed   from   variable  region   segments   (V,D,J)   that   are   joined   by   somatic   recombination.   Leader   (L)   and  constant   (C)   regions   are   joined  by  mRNA  splicing.   Figure   reproduced   from   Janeway’s  Immunobiology  with  permission  from  Garland  Science  /  Taylor  and  Francis  LLC,  2011.2      .................................................................................................................................................  70  Figure  3.2           Design   of   primers   for   PCR   amplification   of   antibody   heavy   (IgH)   and  light   (IgL)   chain   genes.     3’   primers   are   designed   to   the   antibody   constant   (C)   region.   xx Degenerate   primers   to   the   5’   region   can   be   designed   either   to   the   leader   (L)   or   1st  framework  (FWR1)  region  of  VH  and  VL  genes.  ................................................................................  72  Figure  3.3           Degenerate   primers   are   mixtures   of   oligonucleotides   with   similar  sequences   designed   to   amplify   genes   with   highly   related   sequences.     The   level   of  degeneracy  depends  on  the  number  of  base  positions  and  the  variation  at  each  position.    In   this   example,   the   5’   primer   has   4-­‐fold   degeneracy   (2   positions   X   2   bases   at   each  variable  position)  whereas  the  3’  primer  has  6-­‐fold  degeneracy  (3  bases  at  1st  variable  position  X  2  bases  at  2nd  variable  position).  .......................................................................................  73  Figure  3.4           Nested  PCR.  Multiple  rounds  of  PCR  are  performed,  in  which  a  unique  set  of  primers  internal  to  the  template  DNA  are  used  in  each  successive  PCR  round.  In  semi-­‐nested   PCR,   one   primer   is   re-­‐used   and   one   internal   primer   is   designed   for   each  successive  round  of  PCR.  Nested  PCR  is  used  to  increase  amplification  specificity  for  the  target  gene.    .................................................................................................................................................  74  Figure  3.5           RT-­‐PCR  experiment  for  amplifying  genes  from  antibody-­‐secreting  cells.    Cells   are   enumerated   using   a   haemocytometer.   The   protocol   is   repeated   with   serial  dilutions  of  cell  lysate  in  order  to  determine  the  detection  limit  of  RT-­‐PCR  reactions  for  mouse  β-­‐actin  and  antibody  heavy  and  light  chain  genes.  ...........................................................  76  Figure  3.6           Microfluidic  device  for  sorting,  lysis,  and  mRNA  bead  capture  from  single  cells.  (A)  Schematic  of  microfluidic  device  containing  9  reagent  inlets  (left),  8  chambers  (one   cell   per   chamber)   and   one   fluidic   outlet   (right).   (B)   (expanded   view   of   boxed  region  in  A)  Each  chamber  contains  a  partially  closing  sieve  valve  used  to  trap  cells  and  beads.     Cells   are   lysed   in   the   chamber   to   release   cellular   mRNA   that   is   captured   on  oligo(dT)  beads.  Beads  are  sequentially  eluted  from  each  chamber  and  recovered  from   xxi the  output  port  for  single-­‐cell  RT-­‐PCR  amplification.  (C)  Brightfield  microscope  image  of  single   chamber   containing   a   stack   of   oligo(dT)   beads,   an   antibody-­‐secreting   cell,   and  antibody-­‐capture   beads.  Microscope   image   is   rotated   90°   counter-­‐clockwise   from   the  schematic  drawings  in  (A)  and  (B).  ........................................................................................................  81  Figure  3.7           RT-­‐PCR  of  mouse  β-­‐actin  and  antibody  heavy  and  light  chain  genes  using  purified  mRNA  from  different  concentrations  of  D1.3  hybridoma  cell  lysate.    (A)  RT-­‐PCR  products  visualized  on  a  1%  DNA  agarose  gel  with  a  100bp   ladder.    The  β-­‐actin  gene  product  appears  as  a  single  band  with  ~500  bp  in  size.    Multiplex  PCR  of  both  heavy  and  light  chain  reactions  also  appear  as  a  single  ~400  bp  band.  Both  heavy  and  light  chain  gene  products  were  amplified  as  confirmed  by  excising,  purifying,  and  sequencing  the  DNA  products.  DNA  melting  curve  analysis  for  both  mouse  β-­‐actin  (B)  and  multiplexed  heavy   and   light   chain   RT-­‐PCR   reactions   (C).   Plotted   is   the   change   in   fluorescence  intensity   (dI/dT)  at   each   temperature.  The   large   fluorescence   signal   change  at  ~52°C  coincides  with  the  primer  melting  temperature.  .............................................................................  83  Figure  3.8           Multiplex  RT-­‐PCR  of  mouse  heavy  and  light  chain  genes  of  mRNA  purified  from   ~106   D1.3   hybridoma   cells   using   highly   degenerate   primers   at   two   different  concentrations  (160  nM  and  600  nM).  Lower  primer  concentrations  resulted  in  reduced  amplification  of  both  specific  amplicons  and  non-­‐specific  primer  dimers.  Shown  is  a  1%  DNA  agarose  gel  with  100  bp  ladder.  ....................................................................................................  84  Figure  3.9           Multiplex   RT-­‐PCR   of   mouse   antibody   genes   at   different   annealing  temperatures.     4   different   touchdown   PCR   protocols   were   tested   with   annealing  temperatures   varying   from   (A)  65°C-­‐55°C   and  60°C-­‐50°C   to   (B)   55°C-­‐45°C   and  50°C-­‐40°C.   Amplification  was   successful   using   template   concentrations   greater   than   ~100   xxii cell   equivalents,   with   significant   non-­‐specific   amplification   observed   in   all   reactions.  Shown  is  a  1%  DNA  agarose  gel  with  100bp  ladder.  ......................................................................  86  Figure  3.10           Multiplex   RT-­‐PCR   of   mouse   antibody   genes   on   serial   dilutions   of  oligo(dT)   bead-­‐purified   RNA   from   HyHEL-­‐5   (A),   D1.3   (B),   and   CD1d   (C)   mouse  hybridoma   cells   using   a   highly   degenerate   primer   set134   (left)   and   1st   round   primers  from  a  low  degeneracy  primer  set58  (right)  at    600nM  concentration.  Single-­‐cell  RT-­‐PCR  sensitivity  using  low  degeneracy  primers  obtained  for  all  three  hybridoma  cells.    Shown  is  a  1%  DNA  agarose  gel  with  100bp  ladder.  .....................................................................................  88  Figure  3.11           Single-­‐plex   RT-­‐PCR   of   mouse   antibody   genes   from   mouse   hybridoma  cells  using  1st  round  primers  from  a  low  degeneracy  primer  set58.  mRNA  from  3.5×105  D1.3   cells   and   6×105   CD1d   cells  was   purified   using   oligo(dT)   beads.   The   beads  were  then  split  into  two  equal  parts  and  mixed  with  RT-­‐PCR  reaction  mix  containing  primers  at   a   concentration   of   600nM   for   heavy   and   light   chain   amplification,   respectively.    Shown  is  a  1%  DNA  agarose  gel  with  100bp  ladder.  ......................................................................  89  Figure  3.12           Single-­‐plex  RT-­‐PCR  of  mouse  antibody  genes  from  primary  antibody-­‐secreting   cells   (ASCs)  harvested   from  mice   immunized  with  hen  egg   lysozyme   (HEL).    Cells  were  sorted  by  fluorescence-­‐activated  cell  sorting  (FACS).  ASCs  were  lysed  and  the  mRNA  from  serial  dilutions  of  cell  lysate  was  purified  using  oligo(dT)  beads.  The  beads  were   then   split   into   two  equal  parts   and  mixed  with  RT-­‐PCR   reaction  mix   containing  low   degeneracy   primers58   at   600nM   for   heavy   and   light   chain   amplification,  respectively.   Amplification   in   the   heavy   chain   NTC   reaction   was   due   to   reagent  contamination,   which   was   removed   when   using   fresh   primer   solutions   and   RT-­‐PCR  reagents  (data  not  shown).    Shown  is  a  1%  DNA  agarose  gel  with  100bp  ladder.  ............  91   xxiii Figure  3.13           RT-­‐PCR   of   mouse   antibody   genes   from   mRNA   purified   on   oligo(dT)  beads  from  single  HyHEL-­‐5  hybridoma  cells  sorted  in  a  microfluidic  device.  Beads  from  chambers  with  a  single  cell  (“1  cell”)  and  without  cells  (“NTC”)  were  alternately  eluted  and   recovered   from   the   output   port   in   a   stainless   steel   pin   and   Tygon   tubing.   The  output  port  was  washed  with  1X  PBS  and  10%  bleach  in  between  each  sample  elution.    Significant   cross-­‐contamination   occurred   between   samples.   Shown   is   a   1%   DNA  agarose   gel   with   100   bp   ladder.   Pixel   intensities   are   inverted   to   highlight   amplified  products.    ................................................................................................................................................  93  Figure  3.14           RT-­‐PCR  of  mouse  β-­‐actin  (A)  and  antibody  heavy  and  light  chain  genes  (B)   from   mRNA   purified   on   oligo(dT)   beads   from   single   HyHEL-­‐5   hybridoma   cells  sorted   and   recovered   from   a  microfluidic   device.   Beads   from   chambers  with   a   single  cell   (“HyHEL-­‐5   cell”   or   “D1.3”)   and  without   cells   (“NTC”)  were   alternately   eluted   and  recovered   by   manual   pipetting   with   a   new   gel-­‐loading   tip   for   each   sample.   Cross-­‐contamination  between  samples  occurred  if  the  output  port  was  insufficiently  washed  with  1X  PBS  in  between  each  sample  elution.  Shown  is  a  1%  DNA  agarose  gel  with  100  bp  ladder.    .............................................................................................................................................  93  Figure  3.15           RT-­‐PCR  of  antibody  heavy  and  light  chain  genes  from  D1.3  hybridoma  cells   sorted   and   recovered   from   a   microfluidic   device   without   on-­‐chip   cell   lysis.  Chambers   with   and   without   cells   were   alternately   eluted   and   recovered   by   manual  pipetting  with  a  new  gel-­‐loading  tip  for  each  sample.  Samples  were  directly  transferred  to  RT-­‐PCR  reaction  mix  without  dedicated  RT  primers.  Antibody  heavy  and  light  chain  genes   were   successfully   amplified   from   eluted   samples   from   chambers   containing  single   D1.3   cells   as   well   as   two   chambers   loaded   with   multiple   cells   (2   and   6   cells,   xxiv respectively).   No   cross-­‐contamination   between   samples   was   observed.   RT-­‐PCR  performed   using   low   degeneracy   primers58   at   600nM   concentration.   Shown   is   a   1%  DNA  agarose  gel  with  100bp  ladder.  .....................................................................................................  95  Figure  4.1           Microfluidic  screening  and  selection  of  mAbs  from  single  cells.  .................  98  Figure  4.2           Representative  results  from  ELISPOT  assay  to  determine  frequency  of  antigen-­‐specific   (i.e.   anti-­‐HEL)   and   IgG-­‐secreting   ASCs   from   FACS-­‐enriched   mouse  splenocytes  (Image  prepared  by  Dr.  Welson  Wang,  Biomedical  Research  Centre,  UBC)  .....      ...........................................................................................................................................  100  Figure  4.3           Microfluidic  device  for  screening  single  antibody-­‐secreting  cells  (ASCs).  (A)   Device   schematic   depicting   9   fluidic   inlets,   1   fluidic   outlet,   and   112   chambers   (8  rows   ×   14   columns)   addressed   using   a   row   multiplexer   and   column   valves.   (B)  Schematic  of  single  microfluidic  chamber  (volume  ~1  nL)  containing  a  bead  filter/trap  and   sieve   valve   to   modulate   flow   rate   through   chamber.   (C   and   D)   Bright-­‐field  microscope  images  of  sub-­‐nanoliter  microfluidic  chambers  containing  single  ASCs  and  antibody-­‐capture  beads  at  20X  (C)  and  40X  magnification  (D).  ..............................................  104  Figure  4.4           Schematic   of   microfluidic   chamber   while   performing   single-­‐cell  antibody  screening  and  selection.  See  text  for  details.  (Page  1  of  6).  ....................................  105  Figure  4.5           Heavy  chain  genes  from  four  single-­‐cell  selected  anti-­‐HEL  mouse  mAbs  amplified   in   triplicate  by  RT-­‐PCR.  All   amplicons  were   extracted   and  purified   for  DNA  sequencing.  Comparison  of  DNA  sequences  of  amplicons  was  performed  to  verify  that  assigned  somatic  mutations  were  not  generated  by  polymerase  errors  during  RT-­‐PCR.  RxCy   nomenclature   designates   the   row   and   column   address   for   the   microfluidic   xxv chamber  from  which  the  cells  were  recovered.  Shown  is  a  1%  DNA  agarose  gel  with  100  bp  ladder……….    .............................................................................................................................................  114  Figure  4.6           Multiplex  RT-­‐PCR  amplification  of  antibody  heavy  and  light  chain  genes  from  eluted  chambers  containing  no  cells   (no-­‐template  control,  NTC),  a  D1.3  cell,   and  HyHEL-­‐5  cell.  No  amplification  was  observed  from  eluted  chambers  containing  no  cells.    Amplified   gene   products  were   extracted   and   purified   from   the   gel   and   sequenced   to  confirm  that  they  correspond  to  the  corresponding  hybridoma  cell-­‐line.  Shown  is  a  1%  DNA  agarose  gel  with  100  bp  ladder.  ..................................................................................................  116  Figure  4.7           Identification   of  microfluidic   chambers   containing   single   cells   secreting  anti-­‐HEL  mAbs.    After  flushing  chambers  with  fluorescently  labeled  antigen  (i.e.  14.3nM  HEL-­‐Dylight488   conjugate),   high-­‐resolution   fluorescence   imaging   of   all   chambers   is  performed.   The   maximum   fluorescence   bead   intensity   in   each   chamber   is   measured  and   a   threshold   is   set   equal   to   2   standard   deviations   larger   than   the   average  fluorescence  of  no-­‐cell  control  chambers  (95%  confidence  interval).  .................................  118  Figure  4.8     Measured  binding  kinetics  and  affinities  from  ~70  anti-­‐HEL  mAbs  selected  by   microfluidic   single-­‐cell   screening.   Equilibrium   dissociation   constants   (A),   on-­‐rate  constants   (B),   and   off-­‐rate   constants   (C)   plotted   in   rank   order   of   affinity,   as   well   as  histograms  of  these  binding  constants  (D-­‐F).  ..................................................................................  125  Figure  4.9     No   correlation   observed   between   equilibrium   and   kinetic   binding   rate  constants   for  over  70  anti-­‐HEL  mAbs  selected  by  microfluidic  single-­‐cell  screening.  R2  values  correspond  to  linear  regression  of  the  plotted  data.  ......................................................  126  Figure  4.10     Single-­‐cell  RT-­‐PCR  amplification  of  antibody  heavy  and  light  chain  genes  from   ASCs   secreting   anti-­‐HEL   mAbs.   Cells   are   sequentially   recovered   from   the   xxvi microfluidic  device   (left-­‐to-­‐right).  RxCy  nomenclature  designates   the   row  and  column  address  for  the  microfluidic  chamber  from  which  the  cells  were  recovered.  Shown  is  a  1%  DNA  agarose  gel  with  100  bp  ladder.  ..........................................................................................  128  Figure  4.11     Light   (A)   and   heavy   (B)   chain   gene   usage   for   anti-­‐HEL   mouse   mAbs  selected  by  microfluidic  single-­‐cell  screening.  ................................................................................  129  Figure  4.12     Amino  acid  sequences  of  anti-­‐HEL  mAbs  encoded  by  IgHV3-­‐8  heavy  (A)  and   IgK5-­‐43   kappa   (B)   chain   genes.     HyHEL-­‐10,   HyHEL-­‐26,   HyHEL-­‐8,   HyHEL-­‐63,  F10.6.6,   and  D44.1  are  hybridoma-­‐generated  anti-­‐HEL  mAbs,  whereas  X25   is   an  anti-­‐DNP  mAb.  Boxed  residues  are   those   that  contact  HEL   in   the  HyHEL-­‐10/HEL  complex.  Sequences  are  aligned  and  clustered  using  ClustalX.  Truncated  heavy  chain  sequences  (i.e.   M2_R01C08   and   M3_R01C03)   were   not   used   for   clustering.   CDR   regions   are  highlighted  by  shaded  boxes.  ..................................................................................................................  137  Figure  4.13     Sample   association   and   dissociation   curves   of   recombinant   M1_R6C01  mAb   binding   to   HEL-­‐Dylight488   fluorescent   conjugate   (14.3   nM   concentration)  measured   using   the   microfluidic   fluorescence   bead   assay.   Solid   line   represents  experimental  fit  using  mass-­‐action  equations  (Chapter  2,  equations  2.7a-­‐c)  ...................  141  Figure  4.14   The  HEL   protein   can   be   sub-­‐divided   into   three   non-­‐overlapping   regions  that   bind   to   distinct   (“complementation”)   groups   of   mAbs.146   D1.3,   HyHEL-­‐5,   and  HyHEL-­‐10  are  representative  members  of  the  three  different  complementation  groups.  Image  reproduced  from  Batista  et  al  (Cell  Press,  1998).115  .......................................................  146  Figure  5.1           Bi-­‐functionalized   beads   for   the   simultaneous   capture   of   mAbs   and  antibody-­‐encoding  mRNA  from  single  cells.    (Top)  Scheme  for  chemical  conjugation  of  secondary   mAbs   and   oigo(dT)     to   beads   using   carbodiimide   chemistry.   (Bottom)   xxvii Microscope  images  of  bi-­‐functionalized  beads  trapped  by  a  microfluidic  sieve  valve  (A).  Captured   on   the   bead   surface   are   fluorescently   labeled   synthetic   DNA   (B)   and  fluorescently   labeled  mouse  mAbs   (C).   (Bottom)  Measurement   of   binding   kinetics   of  antigen  and   single   cell-­‐secreted  antibodies  on  bi-­‐functionalized  beads  as  described   in  Chapter  2.  Figure  adapted  from  US  Patent  Application  2012/0015347  A1.185  [continued  on  next  page]    ................................................................................................................................................  154           xxviii Acknowledgements  I  would   like   to   thank  my  supervisors  Dr.  Carl  Hansen  and  Dr.  Charles  Haynes.     I   owe  both  of  these  mean  a  great  deal  for  for  their  mentorship,  support,  and  encouragement  both  personally  and  professionally  over  the  last  7  years.    I  would  like  to  also  thank  the  many  members   of   the  Hansen   Lab   that   have   helped  me   and  made  my   experience   so  enjoyable  and  rewarding.    Of  particular  note,   I  would   like   to   thank  Dan  Da  Costa,  Tim  Leaver,  Veronika  Sasse,  Jeffrey  Ng,  Dr.  Kevin  Heyries,  Jens  Huft,  Michael  Van  Insberghe,  and  Hans  Zahn.    I  would  also  like  to  thank  members  of  Schrader  Lab  and  the  Biomedical  Research   Centre   for   their   training   and   countless   valuable   discussions:   Dr.   Michael  Williams,  Dr.  Welson  Wang,  Dr.  Yanni  Wang  and  Dr.  Christy  Thompson.    This  work  would  also  not  have  been  possible  without  the  love  and  support  of  so  many  friends  and   family.    Neha  Bangar:  you  made  so  much  of  my   life  possible  over   the   last  several   years.    I   will   cherish   every  moment   of   this   experience.    My   best   friends   Brad  Atcheson  and  Adam  White:  your  friendship  is  the  greatest  reward  that  I  have  received  in  Vancouver  and  I  hope  our  paths  “convergently  evolve”  in  the  future.    To  my  new  little  brothers:  Harshanvit  Singh  and  Shreyas  Rangan.    And  to  everyone  who  supported  me  physically  and  mentally  through  the  most  physically  challenging  episode  of  my  life:  Jay  Legouillux,  Dr.  Rob  Lloyd-­‐Smith,  Dr.  Michael  Gilbart,  and  Dr.  Rozeela  Nand  (!).    To  Dr.  Ken  Bryant  (Ustad,  Guruji,  friend,  mentor….):    Thank  you  for  helping  me  find  my  “sur”!    To  my   family:  Mummy,  Daddy,  Didi,  Tamara  Didi,  Tauji  and  Tia...   I   cannot  describe   in  words  how  much  I  love  you.    This  work  is  truly  your's  as  much  as  it  is  mine.   xxix   Dedication              To  all  the  fathers  who  made  me  who  I  am.    To  my  grandfathers,  Ram  Kishore  Aggarwal,  Padam  Prakash  Aggarwal,  and  Kunwar  Sen  Goyal.    I  can  only  hope  that  I  will  one  day  measure  up  to  some  fraction  of  you  all.     1 Chapter    1: Introduction  Antibodies  are  proteins  produced  by  the  vertebrate  adaptive  immune  system  to  defend  against  infectious  bacteria,  viruses,  and  other  foreign  agents.2  In  addition  to   their   natural   role   in   immunity,   antibodies   that   bind   target   antigens   with   high  affinity   and   selectivity   are   routinely   used   for   protein   purification,   cell   sorting,  histology,  and  other  research  and  diagnostic  applications.3–8  Antibodies  are  also  the  most   rapidly   growing   class   of   therapeutics   and   the   second   largest   group   of   drugs  after  vaccines,9  with  25  products  approved   for  clinical  use  and  over  425  others   in  development   for   the   treatment   of   cancer,   as  well   as   cardiovascular,   autoimmune,  and   infectious  diseases.10–12  Although   several   techniques   for   producing   antibodies  have  been  developed  over   the   last   three  decades,   the  discovery  of  new  antibodies  remains  an  expensive  and  time-­‐consuming  process.13–15     This   thesis   describes   the   development   of   novel   microfluidic   technologies   for   rapid,  high-­‐throughput  screening  and  selection  of  antibodies  for  both  therapeutic  and   research  applications.  Described  is  a  microfluidic  pipeline  (Figure  1.1)  for  screening  and  selection  of  monoclonal  antibodies  from  single  primary,  antibody-­‐secreting  cells  (ASCs).   In   this   pipeline,   animals   are   first   immunized   with   a   target   antigen.   Next,  ASCs   are   harvested   from   the   immunized   animals   and  purified  using   fluorescence-­‐activated  cell  sorting  (FACS).  Following  purification,  ASCs  are  arrayed  as  single  cells  into  sub-­‐nanoliter  chambers  in  a  microfluidic  device  and  screened  by  a  fluorescence  bead   assay   for   production   of   high   affinity,   antigen-­‐specific  mAbs.   ASCs   producing  antigen-­‐specific  mAbs  are  sequentially  recovered  from  the  device  and  subjected  to  single-­‐cell   RT-­‐PCR   to   amplify   the   antibody-­‐encoding   heavy   and   light   chain   genes.   2 Finally,   antibody   genes   for   high-­‐affinity   mAbs   are   sequenced   and   cloned   into  expression  vectors  for  recombinant  production  in  mammalian  cell  lines.  This  thesis  describes   the   development   and   validation   of   this  microfluidic   single-­‐cell   antibody  selection  platform.        Figure  1.1           Microfluidic  pipeline  for  single-­‐cell  antibody  screening  and  selection.     1.1 Antibodies  and  the  Vertebrate  Adaptive  Immune  System  Antibodies   have   two   primary   functions:   one   is   to   bind   target   molecules,  collectively   referred   to  as  antigens;   the  other   is   to   recruit   immune  cells  and  other  defense   agents   to   degrade   or   clear   the   bound   antigens   from   the   host   organism.2  Distinct   regions   of   the   antibody   molecule   perform   these   two   separate   functions.  Antibodies   are   symmetric   “Y-­‐shaped”   proteins   consisting   of   4   polypeptide   chains,  two   identical   “heavy”  chains  and   two   identical   “light  chains”   (Figure  1.2).  The   two  arms  of  the  Y-­‐shaped  antibody  molecule  contain  identical  antigen-­‐binding  sites  and  are   called   the   antibody   variable   (Fv)   region,   so   called   because   the   gene   sequence  encoding  this  region  varies  between  antibodies.  Sequence  diversity  in  the  antibody  variable   region   generates   extensive   conformational   and   chemical   diversity.  Antibodies  bind  target  antigens  through  a  combination  of  different  molecular  forces,  including   electrostatic   and   van   der   Waals,   as   well   as   hydrophobic   and   hydrogen   Amplify, sequence, and express antibody genes Antigen-speci!c mAbs Screen and select single ASCs secreting antigen-speci!c mAbs Micro"uidic device Immunize animal (mice, rabbits, humans) with target antigen Harvest and purify ASCs 3 bonding  interactions.16  The  base  of  the  Y-­‐shaped  antibody  molecule  is  referred  to  as  the  antibody  constant  (Fc)  region,  which  can  take  one  of  four  or  five  different  forms,  each  of  which  recruits  different  immune  effectors.  For  example,  antibodies  with  a  γ-­‐type  heavy  chain  within  the  Fc  region,  known  as  immunoglobulin  G  (IgG)  antibodies,  can  bind  phagocytic   cells,   such   as  macrophages   and  neutrophils,  which   engulf   the  bound   antigen.2   Conversely,   IgE   antibodies,   which   utilize   an   e   type   heavy   chain  within   the  Fc   region,   trigger   inflammatory   responses  by  binding   to  mast   cells   and  basophils.2  Unlike  the  antibody  variable  region,  the  gene  sequence  for  the  constant  region  is  conserved  across  all  antibodies  of  the  same  sub-­‐type  produced  in  the  same  species.         Figure  1.2      A  schematic  drawing  (A)  and  crystal  structure  (B)  of   the  antibody  IgG  molecule.   Figures  reproduced  from  the  following  websites:    http://www.biology.arizona.edu/immunology/tutorials/antibody/structure.html   (A)   and   http://www.doctortipster.com/8245-­‐immune-­‐defense-­‐against-­‐viruses-­‐not-­‐based-­‐on-­‐ antibody-­‐production-­‐study-­‐says.html  (B).         A   B   4 Antibodies  are  naturally  produced  by  bone  marrow-­‐derived  cells  (B  cells)  in  jawed   vertebrates,   including   mice,   rabbits,   and   humans.   B   cells   undergo   gene  recombination   in   order   to   produce   two   antibody-­‐encoding   genes,   encoding   the  antibody  heavy  and  light  chain  protein  chains,  respectively.2  A  human  heavy  chain  gene   is   generated   by   combining   three   different   gene   segments:   1   out   of   ~60  different   variable   (V)   regions,   1   out   of   ~25   diversity   (D)   regions,   and   1   out   of   6  junction   (J)   regions   (Table   1.1).2   This   combinatorial   process,   called   VDJ  recombination,   can   in   theory   produce   over   7000   unique   heavy   chain   genes.  Similarly,  each  human  B  cell  produces  only  one  of  roughly  500  light  chain  genes  by  recombining  1  out  of  40  Vκ  (kappa)  genes  with  1  of  5  Jκ  -­‐genes,  or  by  recombining  1  of  30  Vλ  (lambda)  genes  with  1  of  4  Jλ  genes  (Table  1.1).2  Combinatorial  pairing  of  heavy   and   light   chains   can   therefore   produce   well   over   a   million   different  antibodies,   each   characterized   by   a   distinct   antigen-­‐binding   site.   The   antibody  repertoire   is   further  diversified  by  nucleotide   trimming  and  enzymatic  addition  of  non-­‐templated   bases   at   the   junctions   of   recombined   gene   segments.   Importantly,  recombination  of  a  single  heavy  and  light  chain  gene  in  each  B  cell  excludes  all  other  possible  recombination  events  (allelic  exclusion);  thus,  all  antibodies  produced  by  a  single   B   cell   share   an   identical   sequence,   structure,   and   function.   Accounting   for  both  combinatorial  and  junctional  diversity,  human  B  cells  are  thought  to  be  capable  of   producing   roughly   one   hundred   trillion   (~1014)   unique   antibody   molecules  (Table   1.1).   However,   since   a   human   produces   between   1010   -­‐   1011   B   cells,   any  particular   individual   produces   only   a   small   percentage   (~0.01%)   of   the   total  antibody  diversity.17     5 Table   1.1         Diversity   of   human   antibodies   generated   by   combinatorial   (imprecise)   gene   recombination   and   heavy/light   chain   pairing.   Antibody   genes   are   further   diversified   by   somatic  hypermutation.  Data  for  table  taken  from  Janeway,  Arnaout  et  al.,  and  Tiller  et  al.2,18,19       Human  Antibodies     Antibody  Chain   Heavy   Light  (κ  or  λ)   Variable  (V)  Segments   56-­‐65   40(κ)  30(λ)   Diversity  (D)  Segments   23-­‐27   -­‐   Joining  (J)  Segments   6   5(κ)  4(λ)   V(D)J  combinations   ~7500-­‐10,000   ~300   #  of  Heavy  +  Light  Pairs     ~2-­‐3  ×  106   Junctional  Diversity   ~3  ×  107   Total  Antibody  Diversity   ~  1014    The  immense  diversity  of  antigen-­‐binding  sites  makes  it   likely  that  a  subset  of  B  cells  will  express  an  antibody  that  binds  any  antigen.  Naïve  B  cells,  those  cells  that  have  yet  to  encounter  antigen,  express  their  antibody  as  cell-­‐surface  receptors  (the  B  cell  receptor,  or  BCR).  The  binding  of  antigen  to  these  receptors  triggers  the  B  cell  to  rapidly  proliferate:  dividing  two  to  four  times  daily  for  3  to  5  days,  resulting  in   a   clone   of   approximately   1000   antibody-­‐producing   daughter   cells.2   B   cells  undergoing   clonal   expansion   also   produce   high   levels   of   the   enzyme   activation-­‐induced   deaminase   (AID),   which   acts   on   antibody   genes   to   induce   somatic  hypermutation.   These   mutations   are   often   concentrated   in   three   hypervariable  regions,   known   as   complementary-­‐determining   regions   (CDRs),   in   both   the  antibody  heavy  and  light  chains  (Figure  1.3).  Collectively,  the  heavy  and  light  chain   6 CDR   regions  define  most   of   the   antigen-­‐binding   surface   of   the   antibody  molecule;  thus,  mutations   to   these  regions  often  result   in  amino  acid  substitutions   that  alter  antibody-­‐antigen  binding  specificity  and  affinity   (Figure  1.4).  While  many  of   these  substitutions   will   abolish   antibody-­‐antigen   binding,   others   result   in   increased  binding  specificity  and  affinity,  conferring  the  B  cell  with  a  selective  advantage  over  the   original   parent   B   cell.   Through   this   iterative   “affinity   maturation”   process,  antibodies  that  bind  their   target  antigen  with  very  high  affinity  and  specificity  can  evolve  over  several  weeks  or  months  (Figure  1.5).20         Figure  1.3     Antibody  heavy  and   light  chains  can  be  divided   into  3  hypervariable  regions   (complementary-­‐determining   regions,   CDRs)   and   4   framework   regions.   Comparison   of   antibody  heavy  and  light  chains  reveals  that  sequence  differences  are  largely  concentrated  to   the   CDR   regions.   The   heavy   and   light   CDR3   regions,   which   are   the   sites   of   V(D)J   recombination,   exhibit   the   greatest   sequence   diversity.   Figure   reproduced   from   Janeway’s   Immunobiology  with  permission  from  Garland  Science  /  Taylor  and  Francis  LLC,  2011.2     7   Figure  1.4     The  hypervariable  (CDR)  regions  of  both  antibody  heavy  and  light  chains  lie  in   the   antigen-­‐binding   domain   of   the   folded   antibody   molecule.   Figure   reproduced   from   Janeway’s   Immunobiology  with   permission   from  Garland   Science   /   Taylor   and   Francis   LLC,   2011.2     8 During   clonal   selection   and   affinity   maturation,   B-­‐cells   give   rise   to   two  classes   of   cellular   progeny:   plasma   cells   and   “memory”   B   cells.   Plasma   cells   are  terminally-­‐differentiated  B  cells  that  no  longer  respond  to  antigen  stimulation,  and  whose   primary   function   is   to   serve   as   antibody-­‐producing   “factories”,   with  antibodies  accounting  for  10-­‐20%  of  all  protein  synthesis  in  these  cells.2  While  some  of   these   antibodies  may   still   be   displayed   on   the   plasma   cell  membrane,  most   of  these  antibodies  are  secreted  into  the  blood  and  other  tissues  at  extraordinary  rates,  reaching  several  thousand  molecules  per  second.21  Plasma  cells  are  enriched  in  the  spleen,   lymph  nodes,  and  bone  marrow,  and  have  been  reported  to  migrate   to   the  blood   1   week   after   antigen   re-­‐stimulation   (“boost”).22,23   Memory   B   cells   are  quiescent  cells   that  circulate  throughout  the  body   long  after   the  primary  antigenic  challenge,  and  are  named  for  their  role  in  accelerating  the  immune  response  to  re-­‐infection  with  the  same  antigen.  Unlike  plasma  cells,  memory  B  cells  do  not  secrete  antibodies;   rather,   they   display   antigen-­‐specific   antibodies   on   their   cell   surface.  Binding   of   antigen   to   memory   B   cells   triggers   a   vigorous   response,   consisting   of  rapid  cellular  proliferation  and  differentiation  into  plasma  cells.24    This  thesis  focuses  on  methods  for  screening  and  selection  of  antigen-­‐specific  antibodies   from   plasma   cells,  which   are   uniquely   tractable   for   single-­‐cell   analysis  because   of   their   high   rates   of   production   and   secretion   of   antibody   mRNA   and  protein,  respectively.       9   Figure  1.5     Generation   of   antigen-­‐specific   antibodies   by   the   adaptive   immune   system.     Antigen  binds  surface-­‐displayed  antibody  on  a  subset  of  naïve  B  cells,  which  are  subsequently   stimulated  to  proliferate  (clonal  selection).  Somatic  mutations  in  proliferating  cells  alter  the   expressed  antibodies,  and  the  clonal  selection  process  is  iterated  to  generate  antibodies  that   bind  antigen  with  high  affinity  and  specificity  (affinity  maturation).  This  process  creates  two   cell-­‐types:  plasma  cells   that   secrete   soluble  antibodies   into   the  blood  and  other   tissues  and   memory   B   cells   that   accelerate   the   immune   response   to   host   re-­‐infection   with   the   same   antigen.         Naive B-cell repertoire A!nity maturation (somatic hypermutation) Clonal selection (proliferation) Antibody-secreting Plasma cells “Memory” B cells No clonal expansion Non-functional antibody High a!nity antibody Antigen Cell di"erentiation 10 1.2 Methods  for  Antibody  Screening  and  Selection  At   the   end   of   the   19th   century,   Paul   Ehrlich,   Emil   von   Behring   and  Shibasaburo   Kitasato   conducting   pioneering   studies   that   demonstrated   the  presence   of   therapeutic   antibodies   in   the   blood   sera   of   antigen-­‐immunized  animals.25,26  This  work  laid  the  foundations  for  vaccination  and  passive  serotherapy  strategies   against   diphtheria,   tetanus,   and   other   pathogens.   However,   antibody  serum   derived   from   other   animal   species   induces   cross-­‐reactive   antibodies   in  humans   and   often   results   in   undesirable   off-­‐target   effects,   in   part   because   the  antibodies   are   derived   from   many   cells   (i.e.   polyclonal)   and   are   therefore   multi-­‐reactive.27   Widespread   adoption   of   antibodies   in   both   research   and   therapeutic  applications  was   facilitated  by  Kohler   and  Milstein’s   seminal   invention   in  1975  of  the  “hybridoma”  method  for  producing  antibodies  of  a  defined  specificity.28   In  this  method,   spleen   cells   (i.e.   splenocytes)   from   immunized   rodents   (e.g.   mice,   rats,  hamsters)  are  harvested  and  fused  with  a  cancer  (i.e.  myeloma)  cell-­‐line  to  generate  hybrid  cells   (hybridoma)   that  both  secrete  antibodies  and  can  be  expanded   in  cell  culture  using  selective  media  (Figure  1.6).  These  cells  are  grown  for  over  a  week  in  order   to   obtain   sufficiently   high   antibody   concentrations   such   that   the   culture  supernatant   can   be   screened   by,   for   example,   an   enzyme-­‐linked   immunoassay  (ELISA),   to   identify   cell   subpopulations   secreting   antigen-­‐specific   antibodies.   This  screening  process   is   typically   iterated  several   times  using   limiting  cell  dilutions   to  select   clones   that   are   producing   antigen-­‐specific   antibodies   of   a   single   specificity,  termed  monoclonal  antibodies  (mAbs).29           11   Figure  1.6     Hybridoma   method   for   producing   antibodies   of   a   defined   specificity.     Antibody-­‐secreting  cells  (ASCs)  from  animals  (e.g.  mice)  immunized  with  antigen  are  fused  to   cancer  (myeloma)  cells  in  order  to  generate  immortalized  ASCs  (hybridoma).  The  hybridoma   cells   are   screened  by   limiting  dilution   to   identify   stable   clones   that   secrete   antigen-­‐specific   mAbs.  Figure  adapted  with  permission  from  Joyce  et  al  (Nature  Publishing  Group,  2010).30      The  success  of   the  conventional  hybridoma  technique   for  producing  rodent  mAbs   has   spurred   great   interest   in   adapting   this   technique   for   the   production   of  antibodies   from   larger   animals.   For   example,   Knight   and   coworkers   developed   a  rabbit   plasmacytoma   cell   line   capable   of   forming   rabbit   hybridoma,   thus   enabling  the   selection   of   antibodies   to   antigens   not   immunogenic   in   rodents.31   In   order   to  circumvent   the   immunogenicity   of   mouse   or   rabbit   mAbs   when   used   as   human  therapeutics,  mAbs   generated   using   the   hybridoma   approach   can   be   “humanized”  by   substituting   human   sequences   into   regions   of   the  mouse  mAbs   that   are   not   in  direct   contact   with   the   antigen.32–34   Although   several   therapeutic   mAbs   (e.g.  Herceptin)  have  been  produced  in  this  way,  “humanization”  of  mAbs  is  a  laborious,  iterative  process  since  the  process  of  humanization  often  produces  mAbs  that  either  no  longer  bind  the  target  antigen  or  remain  sufficiently  immunogenic  in  humans  to  preclude   their  use   as   therapeutics.35  Efforts  have   therefore  been  made   to  develop  human   mAbs   by   fusion   of   human   ASCs   with   human   cancer   cell-­‐lines,   though  technical   limitations   have   limited   widespread   adoption   of   this   approach.36   Isolate ASCs H L Myeloma cells Hybridomas Immunize with antigen Screen with native or recombinant antigen Antigen-speci!c mAbsCell culture 12 Alternatively,  hybridoma  cell   lines  have  been  generated  by  fusing  mouse  myeloma  cell-­‐lines   with   ASCs   from   transgenic   “humanized”   mice   in   which   the   endogenous  mouse   Ab   genes   have   been   inactivated   and   replaced   with   functional   human   Ab  genes.37–40   Most   clinically   approved   “fully”   human   therapeutic   mAbs   (6   out   of   7)  have  been  developed  using  humanized  mice.41  Common   to   all   hybridoma  methods   is   the   problem   that  many   of   the   fused  cells  either  stop  secreting  mAbs  or  fail  to  expand  in  culture  due  to  genetic  instability.  Whereas   the   spleen   of   an   immunized   mouse   may   contain   tens   to   hundreds   of  thousands   of   antigen-­‐specific   ASCs,   a   typical   fusion   will   generate   fewer   than   50  hybridoma   clones   secreting   antigen-­‐specific  mAbs,  many   of   which   are   genetically  unstable  or  secrete  low-­‐affinity  mAbs.28,42,43  The  pool  of  hybridoma  clones  therefore  grossly   underrepresents   the   true   antibody   diversity   in   the   immunized   animal.  Generation   of   high-­‐quality   commercial   mAbs   (e.g.   high   affinity   and/or   antigen  specificity)  may   require   screening  hundreds  or   thousands  of   clones   from  multiple  animal   immunizations   and   hybridoma   fusions,   representing   a   considerable  investment   in   time   and   cost.35  Alternative  methods   for   immortalizing  ASCs   in   cell  culture,  such  as  viral  transformation,  have  similarly  low  transformation  efficiencies  and,  hence,  also  underrepresent  the  native  antibody  repertoire.44,45    Antibody   screening   techniques   based   on   generation   of   synthetic   antibody  libraries  can  circumvent  both  animal  immunization  and  hybridoma  generation.46  In  these  methods,  diverse   libraries  of  antibody  genes  are  generated   in  vitro  by  error-­‐prone   polymerase   chain   reaction   (PCR),   DNA   shuffling   or   other   recombinant  methods,  and  these  antibody  genes  are  expressed  on  the  surface  of  phage,  yeast,  or   13 other   recombinant   vectors   (Figure   1.7).46   The   library   of   surface-­‐displayed  antibodies   is   screened   by   panning   over   an   antigen-­‐covered   surface   or   by  fluorescence-­‐activated   cell   sorting   (FACS)   in   order   to   select   for   antigen-­‐specific  antibodies.  The  process   is   iterated  using   the   selected  antibodies   in  order   to   select  for   high   binding   affinity   and/or   specificity.   These   newer   approaches   can   be  expensive   and   time-­‐consuming  due   to   difficulties   in   generating   and  maintaining   a  diverse  antibody   library,  as  well  as   the  challenge   to  establishing  an  effective  post-­‐screening  validation  step  to  discard  unstable,  insoluble,  or  non-­‐specific  binders.  As  a  result,   most   research-­‐grade   antibodies   continue   to   be   produced   using   the  hybridoma  method.   The   application   of   these  methods   to   discovery   of   therapeutic  mAbs   is   likewise   challenging.   In   order   to   reduce   possible   immunogenicity   for  therapeutic   applications,   synthetic   libraries   are   often   generated   entirely   from  human   antibody   gene   sequences.   However,   the   use   of   human   genes   as   starting  materials  for  library  generation  is  not  sufficient  to  guarantee  the  creation  of  a  “truly  human”  antibody  since  in  vitro  screening  cannot  re-­‐capitulate  the  negative  selection  processes  that  reject  self-­‐reactive  antibodies  in  humans.47  Indeed,  almost  a  third  of  rheumatoid   arthritis   patients   treated   with   Humira,   a   clinically   approved   human  therapeutic   mAb   developed   using   phage   display   technologies,   were   found   to  develop  anti-­‐Humira  antibodies  and  were,  in  turn,  less  likely  to  have  clinical  benefit  or  remission.48     14   Figure  1.7     Screening  and  selection  of  synthetic  antibody  libraries.  (A)  Antibody  genes  are   synthesized  or  amplified  from  B  cells  and  diversified  by  in  vitro  mutagenesis  (e.g.  error-­‐prone   PCR  or  DNA  shuffling).  The  antibody  genes  are  expressed  on   the  surface  of  a  vector   (B)  and   panned  with  antigen  in  order  to  select  antigen-­‐specific  mAbs.  The  process  is  iterated  several   times   in   order   to   increase   affinity   and/or   specificity   of   the   selected   antibodies.   Figure   reproduced  from  Hoogenboom  with  permission  from  Nature  Publishing  Group,  2005.46     1.2.1 Single-­‐Cell  Methods  for  Antibody  Selection  In   order   to   reduce   the   time   and   expense   of   both   hybridoma   and   synthetic  library  screening,  a  number  of  approaches  have  recently  been  developed  to  isolate  antigen-­‐specific  mAbs  directly  from  single  antibody-­‐secreting  cells  (ASCs).23,49–57  In  these  methods,  tissue  samples  (e.g.  blood,  cerebrospinal  fluid,  spleen,  bone  marrow)  are   harvested   from   immunized   animals   or   naturally   infected   humans.   ASCs   are  enriched   from   these   tissue   samples   by   incubating   cells  with   fluorescently-­‐labeled  antibodies  to  known  plasma  cell-­‐surface  markers  (e.g.  CD138+  in  mice  or  CD19+  in   NATURE BIOTECHNOLOGY VOLUME 23 NUMBER 9 SEPTEMBER 2005 1107 heterologous expression, secretion and fold- ing, with proteolysis and antigen-antibody accessibility. Therefore, many of these display and screening systems, although elegant in nature31,32, are not widely used today for anti- bodies. However, a recently described approach bypasses most of these problems: it is based on anchoring the antibody fragment on the periplasmic face of the inner membrane of E. coli followed by disruption of the outer mem- brane, incubation with fluorescently labeled antigen and sorting of the protoplasts. This very promising and versatile display method is directly compatible with (filamentous) phage display, combi es the ease of E. coli-based library constructio s with the power of ce l sorting, and therefore, is likely to become widely used. Other selection platforms. Directed evolution platforms recently devel- oped for antibody fragments include retroviral display34, display based on protein-DNA linkage35,36, microbead display by in vitro compart- mentalization37, in vivo-based growth selection based on the protein fragment complementation assay (PCA)38 or other systems39 and even single-molecule sorting40. Although each of these methods will have specific theoretical advantages, to date, their validation with antibody fragment libraries has been limited, and their advantages over more established systems (e.g. regarding the truly monovalent nature of the method, eukaryotic expression advantages, increase in library size or selection efficiency) remain to be demonstrated. (For a more in-depth discussion of library-display technologies, including PCA and two- hybrid systems, that are available but have not yet been used in combi- nation with antibody fragments, see ref. 41.) To establish a platform to select recombinant antibody libraries in the IgG format, the preferred format for many applications, researchers recently displayed small libraries of IgGs on the surface of mammalian cells. After homologous integration of a single-gene copy in each cell, the population was sorted by flow cytometry to obtain a clone with sevenfold affinity improvement (W.D. Shen, Amgen, personal commu- nication). In the future, bigger combinatorial IgG format–based libraries may be built using vaccinia virus–based vectors42, or diversity may be introduced in vivo by using B-cell lines that hypermutate a carrier anti- body gene constitutively43 or upon i ducti n44 or that harbor i duc- ible hypermutable enzymes involved in this process in nature45. Some of these newer selection and diversification methods may open novel applications for the directed evolution of antibodies and other proteins (see also accompanying review on p. 1126–1136). Strat gies to select and s reen antibody libraries Individual clones of a recombinant si gle-chain Fv (scFv) or Fab library theoretically can be directly scr ened for antig n binding, f r example, using binding assays based on ELISA or filter-based screening. Screening is limited by the number of clones that can be examined, hence in many applications the frequency of antigen-reactive clones is too low, and the libraries too large (with tens of millions to billions of clones) to do this efficiently. The connection between genotype and phenotype in phage- or ribosome-display libraries provides a means to select for clones bindi g to a desirabl antigen, thereby increasing the frequency of antigen-reactive clones, enriching the clones with best binding affinity, or the clones with certain predefined binding characteristics. Typically many more clones can therefore be sampled compared with screen- ing procedures. Many different selection methods and experimental approaches have been developed that separate clones that bind from those that do not (Fig. 3). Selection procedures. For phage-display libraries, selection involves exposure to antigen to allow antigen-specific phage antibodies to bind their targets during biopanning. This is followed by recovery of antigen- bound phage and subsequent infection in bacteria. Although ideally, only one round of selection would be required, nonspecific binding limits the enrichment that can be achieved per selection round and therefore, in most cases, recursive rounds of selection and amplification are needed to select the best binders from the library (Fig. 2a). Phage display–based selections are now a relatively standard procedure in many molecular biology laboratories (a more detailed description of these proced es is provided elsewhere10,46 and references therein). For more complex selections such as those using cells or tissues, it can be instructive to use enrichment studies with control phage antibodies to optimize the efficiency of the selection method and to compare different selection approaches, and then tune the selection strategy accordingly to Phage display Protein-mRNA link via: Protein-DNA display Growth selection via: Display on: Microbead via in vitro compartmentalization Coupling of geno to phenotype Selective pressure on phenotype Screening Amplification + Antibody gene pool Displayed library Selected antibody lead Synthetic DNA Cloning of genetic diversity B-cells Selection cycle Mutagenesis and selection cycle -ribosome display -mRNA display -Yeast -Bacteria -Mammalian cells -Retroviruses -..... -Yeast 2-hybrid -Protein fragment complementation a bSteps in antibody selection Selection platformsFigure 2 Creating and selecting recombinant antibody libraries. (a) First, antibody diversity is generated from synthetic V genes or cloned from B cells. Next, antibody phenotype (boxes in green, blue and orange) is coupled to its genotype (wavy line) via a phenotype-genotype link (green) packaged in a host (purple) (shown here schematically for phage display). As a result, each host particle expresses (or displays) a unique antibody on its surface. The repertoire of antibodies displayed on these host particles is subjected to The process is repeated and eventually antibodies binding to antigen are confirmed by screening. (b) Different selection platforms for conventional antibodies. Color code as for a (see text for details and citations). Ka tie R is R E V I E W © 20 05 N at ur e Pu bl is hi ng G ro up h ttp :// ww w. na tu re .c om /n at ur eb io te ch no lo gy 15 humans)  followed  by  fluorescence-­‐activated  cell  sorting  (FACS).19,58  Antibody  heavy  and  light  chain  genes  are  amplified  by  single-­‐cell  reverse  transcription  polymerase  chain   reaction   (RT-­‐PCR),   cloned   into   expression   vectors   and   expressed  recombinantly  in  bacterial,  yeast,  or  mammalian  cells.  The  expressed  antibodies  are  subsequently  screened  for  binding  affinity  and  specificity  to  the  target  antigen.  This  general  workflow  has  yielded  antibodies  from  highly  enriched  antigen-­‐specific  ASC  populations  produced  by  humans  exposed  to  tetanus  toxoid,  anthrax,  dengue  virus,  rotavirus,   and   influenza   virus.23,59–62  Remarkably,   up   to  80%  of   human  ASCs   from  blood   taken   7   days   after   a   booster   shot   with   influenza   vaccine   were   found   to  produce  influenza-­‐specific  mAbs.23  It  remains  to  be  seen  whether  this  approach  can  be   generally   applied   for   the   selection   of   antibodies   to  more   poorly   immunogenic  antigens  that  may  fail  to  generate  highly  enriched  ASC  populations,  such  as  mutated  proteins   expressed   in   human   cancers   and   endogenous   proteins   overexpressed   in  autoimmune  disorders.  Under   these   circumstances,   laborious  and   time-­‐consuming  cloning  and  expression  of  thousands  of  antibodies  may  be  required  for  downstream  screening.    An   assay   for   screening   antibodies   secreted   by   single   ASCs   could   facilitate  rapid   and   inexpensive   single-­‐cell   antibody   selection   without   the   need   to   amplify,  clone  and  express  mAbs  from  all  (i.e.  antigen-­‐specific  and  antigen-­‐nonspecific)  ASCs.  Among   the  most   successful   examples   of   this   strategy   is   the   Selected   Lymphocyte  Antibody   Method   (SLAM),   in   which   a   hemolytic   plaque   assay   is   used   to   identify  single   cells   secreting   antigen-­‐specific   mAbs   (Figure   1.8).50   In   SLAM,   antigen   is  chemically   conjugated   to   the   surface  of   sheep   red  blood   cells   (SRBCs).  The   SRBCs   16 are   then   mixed   with   ASCs   and   blood   complement,   such   that   the   SRBCs   vastly  outnumber   the  ASCs.  The  mixture   is   then  placed   in  between   two  wax-­‐sealed  glass  slides   and   incubated   at   37°C   for   less   than   an   hour.   During   this   time,   antibodies  secreted  by  ASCs  bind  to  the  antigen-­‐covered  SRBCs.  Blood  complement  then  binds  to   these  antibodies,  which  triggers   the   lysis  of   the  SRBCs  (i.e.  hemolysis).  Antigen-­‐specific  ASCs  are   located   in  regions  on   the  glass  slide  devoid  of  SRBCs  (“plaques”)  (Figure  1.8).  Each  ASC  is  then  manually  recovered  using  a  micropipette  and  can  be  subjected  to  single-­‐cell  reverse  transcription  polymerase  chain  reaction  (RT-­‐PCR)  to  amplify   the   antibody   heavy   and   light   chain   genes.   The   amplified   genes   are  sequenced,   cloned   and   expressed   in  mammalian   cell-­‐lines   for   production.   Despite  success  of  the  SLAM  method,  the  hemolytic  plaque  assay  is  not  suitable  for  selecting  mAbs  based  on  antigen-­‐binding  affinity  and  selectivity  and  is  further  not  amenable  to  high-­‐throughput  automation.  The  development  of  new  technologies  for  sensitive,  high-­‐throughput   screening   of   mAbs   from   single   cells   could   facilitate   rapid  generation  of  mAbs  for  both  research  and  therapeutic  applications.       17   Figure  1.8     Selected  Lymphocyte  Antibody  Method  (SLAM)  for  identifying  antigen-­‐specific   mAbs  from  single  antibody-­‐secreting  cells  (ASCs).  ASCs  are  mixed  with  antigen-­‐coated  sheep   red  blood  cells  (SRBCs)  and  blood  complement  on  a  glass  slide  and  incubated  at  37°C  for  an   hour.   Binding   of   secreted   antibodies   to   antigen   triggers   lysis   of   red   blood   cells   in   the   area   around   each  ASC,   thus   forming   visible   “plaques”   on   a   sealed   glass   slide.   ASCs   are  manually   recovered  and  subjected  to  single-­‐cell  RT-­‐PCR  followed  by  cloning  and  expression  of  antibody   genes.  Figure  reproduced  with  permission  from  Babcook  et  al.  (PNAS,  1996).50       1.3 Microfluidics   –   An   Enabling   Technology   for   Screening   and   Selection   of   Antibodies  from  Single  Cells  Microfluidics   can  be  broadly  defined   as   technologies   that  manipulate   small  volumes   of   fluids   (femtoliters   to   nanoliters,   or   10-­‐15   –   10-­‐12   L)   in   channels   with  dimensions   of   tens   to   hundreds   of   microns.63   Microfluidics   offer   fundamental  advantages   in   chemical   analysis,   namely   precise   control   of   reagents,   reduced  analysis  times,  and  lower  cost  compared  with  traditional  analytical  methods.  In  the  last  two  decades,  microfluidic  technologies  have  progressed  from  an  initial  proof-­‐of-­‐concept   demonstration64   to   commercial   applications   in   chromatography,65  molecular   detection,66   and   high-­‐throughput   DNA   sequencing67.   Microfluidic   Proc. Natl. Acad. Sci. USA Vol. 93, pp. 7843–7848, July 1996 Immunology A novel strategy for generating monoclonal antibodies from single, isolated lymphocytes producing antibodies of defined specificities (PCR!antibody-forming cells!VH and VL genes!immunoglobulin!plaque assays) JOHN S. BABCOOK, KEVIN B. LESLIE, OLE A. OLSEN, RUTH A. SALMON, AND JOHN W. SCHRADER* The Biomedical Research Centre, 2222 Health Sciences Mall, The University of British Columbia, Vancouver, British Columbia, V6T 1Z3 Canada Communicated by George A. Palade, University of California at San Diego, La Jolla, CA, April 11, 1996 (received for review July 5, 1995) ABSTRACT We report a novel approach to the generation of monoclonal antibodies based on the molecular cloning and expression of immunoglobulin variable region cDNAs gener- ated from single rabbit or murine lymphocytes that were selected for the production of specific antibodies. Single cells secreting antibodies for a specific peptide either from gp116 of the human cytomegalovirus or from gp120 of HIV-1 or for sheep red blood cells were selected using antigen-specific hemolytic plaque assays. Sheep red blood cells were coated with specific peptides in a procedure applicable to any antigen that can be biotinylated. Heavy- and light-chain variable region cDNAs were rescued from single cells by reverse transcription–PCR and expressed in the context of human immunoglobulin constant regions. These chimeric murine and rabbit monoclonal antibodies replicated the target spec- ificities of the original antibody-forming cells. The selected lymphocyte antibody method exploits the in vivo mechanisms that generate high-affinity antibodies. This method can use lymphocytes from peripheral blood, can exploit a variety of procedures that identify individual lymphocytes producing a particular antibody, and is applicable to the generation of monoclonal antibodies from many species, including humans. The enormous diversity of immunoglobulin antigen-binding regions is generated by a series of unique genetic and cellular mechanisms that operate during lymphocyte development and immune responses. It permits the isolation of antibodies that bind an unlimited range of molecular conformations (1). The hybridoma technique (2) enabled the reproduction of specific monoclonal antibodies (mAbs) and revolutionized the exploi- tation of antibodies in research, industry, and medicine. How- ever, in general, the hybridoma technology is restricted to the generation of rodent mAbs. Moreover, it results in the im- mortalization of only a small fraction of the specific antibody- forming cells available in an immunized animal. Newer tech- niques based on the screening of libraries of randomly recom- bined immunoglobulin heavy and light chain cDNAs (3, 4) are restricted by practical limits to the size of libraries and the requirement for the antibody to be properly folded and expressed in bacteria. Furthermore, the generation of these libraries disrupts the pairing of light and heavy chains that were somatically mutated and coselected in single cells in vivo during immune responses; thus, combinatorial techniques fail to fully exploit the immense power of the immune system to generate high-affinity antibodies. We report here a conceptually distinct method for gener- ating mAbs that overcomes these limitations. It involves first identifying within a large population of lymphoid cells a single lymphocyte that is producing an antibody with a desired specificity or function, and then rescuing from that lymphocyte the genetic information that encodes the specificity of the antibody (Fig. 1). The selected lymphocyte antibody method (SLAM) permits the reproduction of the high-affinity anti- bodies generated during in vivo immune responses in multiple species. Over 30 years ago, Nossal and Lederberg (5, 6) pioneered the use of micromanipulation techniques to analyze the spec- ificity of antibodies secreted by single cells. Techniques that permitted screening of large populations of cells to directly identify single cells that produced antibody of a particular specificity followed, first identifying cells producing antibodies specific for a bacterial antigen by their adherence to the relevant bacteria (7) and, subsequently, identifying cells pro- ducing antibodies specific for heterologous erythrocytes by formation of hemolytic plaques (8). The hemolytic plaque assay has since been modified to detect cells producing anti- bodies specific for a wide range of antigens that can be attached to erythrocyte surfaces. Although these methods allowed identification of single antibody-forming cells (AFCs) and analysis of the specificity of the ntibodies they produced, the AFCs died rapidly, precluding the generation of clones that would produce mAbs (9). Our method provides a means to clone cDNAs that encode the specificity of the antibody produced by such a single cell. The SLAM strategy thus har esses the pow r of techniques like plaque assays for screening large numbers of cells for AFCs producing specific antibodies, allowing the generation of mAbs with desired characteristics. The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked ‘‘advertisement’’ in accordance with 18 U.S.C. §1734 solely to indicate this fact. Abbreviations: SLAM, selected lymphocyte antibody method; AFC, antibody-forming cell; hCMV, human cytomegalovirus; SRBC, sheep red blood cells; IL-3, interleukin 3. *To whom reprint requests should be addressed. e-mail: john@brc. ubc.ca. FIG. 1. Strategy for cloning immunoglobulin VH and VL cDNAs from single cells producing specific antibodies. 7843 18 technology   has   also   enabled   high-­‐throughput   biochemistry,68,69   drug   discovery,70  chemical  synthesis,71  structural  biology,72–74  molecular  diagnostics,75–77    and  single-­‐cell  analysis.78–83  Rapid  prototyping  of  complex  microfluidic  devices  has  been  enabled  by  two  inventions:   firstly,   the   invention   of   a   replica   molding   technique,   called   soft  lithography,   for   fabricating   microfluidic   channels   from   lithographically-­‐patterned  master   molds   (Figure   1.9A);84,85   and,   secondly,   a   method   to   fabricate   valves   by  multilayer   soft   lithography   (MSL)  whereby   channels   are   fabricated   and   aligned   in  multiple  layers  of  a  single  device86  (Figure  1.9B).  Both  of  these  fabrication  methods  utilize   a   silicone   rubber   known   as   polydimethylsiloxane   (PDMS),   which   is  manufactured   as   a   two-­‐part   viscous   fluid   that,   when   mixed,   solidifies   by   room-­‐temperature   vulcanization   (RTV).     PDMS   exhibits   a   range   of   useful   material  properties   for   microfluidic   devices,   the   most   important   of   which   are   gas  permeability  for  maintaining  viability  of  biological  specimens  and  for  enabling  dead-­‐end  loading  of  liquids  into  channels,  transparency  to  allow  imaging  of  devices  using  standard   optical   and   fluorescence   microscopes,   and   a   relatively   low   Young’s  modulus  (<  1  MPa).  This  latter  property  enables  the  simple  fabrication  of  pneumatic  valves  whereby  application  of  pressure  to  a  fluid-­‐filled  channel  on  one  layer  deflects  a   thin   elastomeric   membrane   to   seal   off   a   channel   on   an   adjacent   layer   of   the  microfluidic   device     (Figure  1.9C).    MSL   facilitates   the   integration   of   thousands   of  micro-­‐valves   in   a   single   PDMS   device,   which   can   be   used   to   build   higher-­‐level  components   including   fluidic   mixers,   peristaltic   pumps,   and   fluidic   multiplexing  structures  (Figure  1.10).       19   Figure  1.9     Fabrication  of  single  (A)  and  multilayer  (B  &  c)  polydimethylsiloxane  (PDMS)   microfluidic   devices.   (A)   Soft   lithography.   Replica   molding   of   lithographically-­‐patterned   master   molds   using   PDMS   liquid   polymer.     After   the   PDMS   polymer   is   cured   into   a   solid   substrate,   it   is   removed   from   the   master   mold,   input/output   ports   are   manually   punched   through   the   device   and   the   microfluidic   channels   are   sealed   against   a   glass   slide.     (B)   Multilayer  soft  lithography  (MSL).    Replica  molding  is  performed  using  multiple  master  molds,   and   the   resulting   PDMS   layers   are   aligned   and   bonded   into   a   monolithic   structure.   (C)   Pressure   applied   to   a   fluid-­‐filled   channel   on   the   control   layer   deflects   the   elastomeric   membrane  separating  it  from  the  channel  on  the  flow  layer,  thus  closing  the  reversible  valve   structure.   Figures   (A)   adapted   with   permission   from   McDonald   et   al.   (Electrophoresis,   2000),85  (B)  reproduced  from  Unger  et  al.  (Science,  2000),86    and  (C)  courtesy  of  C.  Hansen.     CAD File High-resolution photomask 1. Laser printing 2. Photolithography UV light Master mold (Photoresist on Silicon wafer) 3. Replica molding Cast PDMS liquid polymer on mold and cure 4. Remove PDMS replica, punch input/output ports, seal against glass slide Micro!uidic device mold !at substrate A B C 100 µm 100 µm Membrane P re ss ur e In le t Flow Channel Control Channel Glass Slide Glass Slide P re ss ur e In le t Membrane Control Channel Pneumatic valve Open state Pneumatic valve Closed state Glass slide Control channel Flow c annelP re es ur e in le t Membrane Glass slide Control channel Pr ee su re in le t Membrane Bulk PDMS Bulk PDMS 20   Figure  1.10     Integration  of  multiple  microfluidic  valves  into  higher-­‐order  fluidic  structures   (pumps,   fluidic   mixers,   and   fluidic   multiplexing   structures)   in   single   microfluidic   devices   fabricated  by  multilayer  soft   lithography.73,86,87  Pumps  are  used  to  meter  precise  volumes  of   fluidic  reagents,  ranging  from  100  pL  to  1  nL.  Viscous  forces  dominate  inertial  forces  for  fluid   flow   in   microfluidic   channels   (i.e.   laminar   flow),   and   thus   fluidic   mixers   are   required   to   accelerate  the  mixing  of  chemical  reagents.    Multiplexing  structures  facilitate  the  selection  of   one   or   more   fluidic   reagents   with   a   reduced   number   of   valves   (2logN)   compared   to   the   number  of  reagent  inputs  (N).  Figure  courtesy  of  C.  Hansen  with  permission.    Miniaturization   offers   a   particular   advantage   for   the   detection   of   mAbs  secreted   by   single   antibody-­‐secreting   cells   (ASCs).   Despite   secreting   thousands   of  antibody  molecules  per  second21,  single  ASCs  in  conventional  96-­‐  or  384-­‐well  plates  (>10   μL   volumes)   would   require   approximately   10   weeks   to   secrete   antibody  concentrations   detectable   by   standard   enzyme-­‐linked   immunoassays   (ELISA)  (Figure  1.11).  Confining  single  ASCs  to  small  (e.g.  sub-­‐nanoliter)  volumes  can  enable  efficient   detection   of   secreted   antibodies   from   single   ASCs   within   hours   of  harvesting   and   laboratory   culture   (Figure   1.11).     Indeed,   Nossal   and   Lederberg   21 confined   single   ASCs   in  microdroplets   and   demonstrated   that   single   ASCs   secrete  antibodies  of   a   single   specificity,   in  a   seminal   study   from  1958   that   first  provided  experimental  evidence  for  the  “one  cell-­‐one  antibody”  theory.88,89     Figure  1.11     Concentration   enhancement   in   small-­‐volume   chambers   (<1   nL)   enables   detection   of   antibodies   secreted   by   single   antibody-­‐secreting   cells   (ASCs).   ASCs   harvested   from  immunized  animals   typically  survive   for  ~1-­‐2  days   in  culture.    Thus,  mAbs   from  single   ASCs   cannot   be   detected   in   cell-­‐culture   plates   using   standard   laboratory   tests   (>1   nM   detection  limit).        A  number  of  groups  have  recently  used  small-­‐volume  compartmentalization  to   screen   mAbs   from   single   ASCs   using   either   micro-­‐fabricated   wells14,51,90   or  emulsion  (i.e.  water-­‐in-­‐oil)  droplets91,92.  Love  and  coworkers  used  a  microengraving  method   to   fabricate   small  micro-­‐wells   in  a  PDMS  substrate   (Figure  1.12A).14  ASCs  were   manually   pipetted   on   to   the   surface   of   the   substrate   and   single   ASCs   were  allowed   to   settle   into   the   wells   by   gravity.   The   PDMS   substrate   was   then   sealed  against  a  functionalized  glass  slide  that  captured  secreted  antibody  from  each  well   Antibody-secreting cell (ASC) 96-well plate volume > 1µL Ab  1cell 1000 Abs / s 10 weeks 10 +L 1nM Ab  1cell 1000 Abs / s 1hr 1nL  6 nM 22 to   produce   a   printed   “microarray”   that   was   then   incubated   with   fluorescently  labeled  antigen  and  secondary  antibodies.  Fluorescence   imaging  of   the  microarray  was  performed  to  determine  which  wells  contained  single  ASCs  secreting  antigen-­‐specific   mAbs.   Repeating   this   method   with   multiple   slides   and   incubating   each  printed   microarray   with   a   different   concentration   of   fluorescent   antigen,   the  apparent  affinity  of  each  secreted  mAb  was  estimated.90  In  a  similar  manner,  Kishi  and  coworkers  demonstrated  that  mAbs  from  single  ASCs  in  micro-­‐fabricated  wells  can  be  detected  using  a   fluorescence   surface  assay   (immunospot  array  assay  on  a  chip,   ISAAC)   and   that   selected   ASCs   can   be   retrieved   from   the   array   by  micromanipulation  in  order  to  amplify  heavy  and  light  chain  genes  for  subsequent  cloning   and   expression   of   mAbs   (Figure   1.12B).51   Finally,   Merten   and   coworkers  developed   microfluidic   devices   to   encapsulate   single   hybridoma   cells   in   aqueous  droplets  (Figure  1.12C)  and  demonstrated  the  detection,  sorting  and  enrichment  of  droplets   containing   ASCs   secreting   mAbs   that   inhibit   enzymatic   activity   (e.g.  angiotensin  converting  enzyme  1,  ACE-­‐1)  based  on  a   fluorescence  assay.92  Despite  screening  tens  to  hundreds  of  thousands  of  ASCs,  these  micro-­‐technologies  typically  identify   very   few   ASCs   secreting   antigen-­‐specific   mAbs   (~20),   of   which   the   vast  majority   of   mAbs   bind   target   antigen   with   low   affinities   (Kd   <   100nM);51,90   thus,  these   methods   produce   comparable   yield   and   quality   of   mAbs   to   conventional  hybridoma  methods  (see  Chapter  1,  Section  1.2  above).  Of  these  micro-­‐technologies,  only  the  ISAAC  method  has  previously  been  used  to  recover  ASCs  for  amplification,  cloning,  and  expression  of  antibody  genes.  Thus,   the  work  described   in   this   thesis  focused  on  the  development  of  a  novel  micro-­‐technology  for  rapid,  high-­‐throughput   23 selection  of  high  affinity  antigen-­‐specific  mAbs  from  single  ASCs.  Specifically,  a  fully  integrated   microfluidic   device   was   designed   and   fabricated   by   multilayer   soft  lithography  for  single-­‐cell  handling,  screening  of  high  affinity  antigen-­‐specific  mAbs,  and   selective   recovery   of   ASCs   for   downstream   amplification,   cloning,   and  expression  of  antigen-­‐specific  mAbs.           Figure  1.12     Methods   for   screening   antibodies   secreted   by   single   cells   using   micro-­‐ fabricated  wells  (A  and  B)  and  droplet  encapsulation  (C).  (A)  Microengraving  method.    Single   cells  in  PDMS  micro-­‐wells  secrete  antibodies  that  are  captured  on  a  “printed  microarray”  that   is  imaged  after  incubation  with  fluorescently-­‐labeled  antigen.  (B)  ISAAC  method  (see  text  for   details).     (C)   Schematic   (above)   and   microscope   images   (below)   of   microfluidic   devices   to   encapsulate   single   cells   in   water-­‐in-­‐oil   emulsion   droplets,   incubate   and   detect   secreted   antibodies.   Scale   bars   are   100   μm.   Figures   reproduced   with   permission   from   Love   et   al.   (Nature  Publishing  Group,  2006)   (A)14,   Jin   et   al.   (Nature  Publishing  Group,  2009)   (B)51,   and   Köster  et  al.  (Lab  on  a  Chip,  2008)  (C)91.       Glass Glass 50-100 +m 50-100 +m Microarray of secreted products PDMS PDMS 1 23 4 A C Mice or volunteers immunized with antigen (for example HEL, HBs or InV) Microarray chip (230,000 wells) Detection of ASCs Retrieval of ASC Single-cell RT-PCR for amplification of Ab cDNA Construction and cloning of IgH and IgL IgH IgL Ab-cDNA cloning and transfection ELISA for specific binding 5–6 d7–8 h B 24 1.4 Aims  of  this  Thesis  This   thesis   describes   the   design   and   fabrication   of   a   novel   fully-­‐integrated  microfluidic   system   that   enables   sensitive   screening   and   selection   of   mAbs   from  single  antibody-­‐secreting  cells  (ASCs)  through  direct  and  accurate  measurement  of  their   binding   affinities   and   selectivity   to   a   target   antigen,   as   well   as   automated  recovery   of   single   cells   for   RT-­‐PCR   amplification   of   antibody   genes   in   order   to  sequence,   clone,   and   express   antigen-­‐specific  mAbs   produced   by   these   cells.   This  work  involved:  1. The  development  of   an  ultrasensitive  microfluidic   fluorescence  bead  assay  for  measuring  antibody-­‐antigen  binding  kinetics  and  selectivity  from  small  amounts  of  antibody  sample  (Chapter  2);  2. The   development   of   methods   to   sort   and   recover   single   cells   from  microfluidic  devices  and  RT-­‐PCR  amplify  heavy  and  light  chain  genes  encoding  antigen-­‐specific  mAbs  (Chapter  3);  3. The   integration   of   antibody   screening   from   hundreds   of   single   cells  with   recovery   and   amplification   of   antibody   genes   from   cells  producing  novel,  antigen-­‐specific  mAbs  (Chapter  4).  By  screening  monoclonal  antibodies  from  single  cells,  the  proposed  technology  will  enable   rapid   and   high-­‐throughput   selection   of   monoclonal   antibodies   for  therapeutic  and  biomedical  research  applications.       25 Chapter    2: Microfluidic  Measurement  of  Antibody-­‐Antigen  Binding   Kinetics  from  Low  Abundance  Samples  An   ultrasensitive   microfluidic   fluorescence   bead   assay   for   measuring  antibody-­‐antigen  binding  kinetics  from  low  abundance  samples  is  described.         2.1  Antibody-­‐Antigen   Binding   Properties:   Binding   Affinity,   Selectivity   and   Kinetics  The  binding  of  antibodies  to  target  antigens  is  typically  characterized  by  two  properties:  affinity  and  selectivity.  Selectivity  refers  to  the  ability  of  an  antibody  to  bind  variants  of  a   target  antigen;   that   is,   a   cross-­‐reactive  antibody  will  bind  many  different   structural   isoforms   (glycoforms,   post-­‐translational   modifications,   or  species  homologues)  of  an  antigen,  whereas  a  selective  antibody  will  bind  a  specific  structural   isoform  of  an  antigen.  On   the  other  hand,  antibody  affinity  refers   to   the  “strength”  with  which  it  binds  to  a  target  antigen,  and  is  governed  by  a  combination  of  net  favorable  electrostatic  and  van  der  Waals  forces,  as  well  as  hydrophobic  and  hydrogen   bonding   interactions.   Based   on   differences   in   the   number,   types   and  geometries  of  molecular  contacts,  antibody-­‐antigen  interactions  can  exhibit  a  broad  range   of   binding   affinities,   with   equilibrium   dissociation   constants   ranging   from  micromolar  to  sub-­‐nanomolar  (10-­‐5  –  10-­‐10  M).93         26 2.1.1 Mathematical  Model  for  Antibody-­‐Antigen  Binding  Antibody-­‐antigen   interactions   are   non-­‐covalent   and   usually   reversible.94   In  such   cases,   the   kinetics   of   the   binding   reaction   can   be   described   by   the   following  first-­‐order  differential  equation:   !!" 𝐴𝑏𝐴𝑔 = 𝑘![𝐴𝑏][𝐴𝑔] −  𝑘![𝐴𝑏𝐴𝑔]   (2.1),  in  which  [Ab],  [Ag],  and  [AbAg],  represent  molar  concentrations  of  antibody,  antigen  and  antibody-­‐antigen  bound  complex,  respectively.  This  equation,  often  referred  to  as  the  law  of  mass  action,  can  be  physically  interpreted  in  the  following  manner:  1)  antibody   and   antigen   molecules   collide   in   solution   due   to   random   diffusion   and,  thus,  the  probability  of  collision  and  binding  is  directly  proportional  to  the  antibody  and   antigen   solution   concentrations;   2)   the   dissociation   of   antibody-­‐antigen  complex  is  a  random  event  and  is  therefore  only  proportional  to  the  concentration  of  antibody-­‐antigen  complex.    The  proportionality  constants,  kf  and  kr,  are  called  the  forward   and   reverse   kinetic   rate   constants,   respectively.   This   equation   can   be  solved  analytically  when  one  of  the  two  interacting  molecules  is  present  in  excess  or  is  immobilized  on  a  substrate,  such  that  [Ab]  =  [Ab]t=0  +  [AbAg]  ≈  [Ab]t=0  .    Under  these  circumstances,  the  analytical  solution  to  equation  2.1  is  described  by  the  equations  in  Table  2.1  and  depicted  in  Figure  2.1.       27 Table  2.1     Analytical  solutions  to   first-­‐order  differential  equations  describing  antibody-­‐ antigen  binding  under  the  condition  that  [Ab]  ≈[Ab]t=0  >>  [Ag]0.       Association  phase:    𝐴𝑏𝐴𝑔 = [𝐴𝑏]! [!"]![!"]!!!! 1 − 𝑒!! !        (2.2a)  𝜏 = !!! !" !!!!                                                                                          (2.2b)   Chemical  Equilibrium:  𝐴𝑏𝐴𝑔 = [!"]![!"]!!!!                        (2.3a)  𝐾! = !!!!           (2.3b)   Dissociation  phase:      𝐴𝑏𝐴𝑔 = [𝐴𝑏]! [!"]![!"]!!!! 𝑒!! !                    (2.4a)  𝜏 = !!!                     (2.4b)         28   Figure   2.1         Graphical   depiction   of   first-­‐order   antibody-­‐antigen   binding   kinetics.   Concentration   of   antibody-­‐antigen   complex   is   on   the   y-­‐axis,   whereas   time   is   on   the   x-­‐axis.   Antibody-­‐antigen   complex   follows   bimolecular   exponential   association   kinetics   during   the   association   phase,   and   first-­‐order   exponential   kinetics   during   the   dissociation   phase.   Equations   describing   the   rates   of   growth   and   decay   in   concentration   of   antibody-­‐antigen   complex  are  presented  in  Table  2.1.        According  to  this  simple  but  proven  model,  antibody-­‐antigen  binding  is  predicted  to  occur   at   an   exponential   rate,  with   a   time   constant  dependent  on   the   total   antigen  concentration  as  well  as  both  the  forward  and  reverse  rate  constants  (equation  2.2).    The   concentration   of   antibody-­‐antigen   complex   reaches   a   plateau   at   a   maximum  value,   corresponding   to   the   condition   of   chemical   equilibrium,   where   the   rate   of  antibody-­‐antigen  association  equals  the  rate  of  dissociation  of  the  antibody-­‐antigen  complex.  The  amount  of  antibody-­‐antigen  complex  formed  at  chemical  equilibrium  can  be  conveniently  quantified  with  respect  to  the  equilibrium  dissociation  constant  (Kd,   in   units   of   molarity),   where   a   lower   value   of   Kd   represents   a   higher   binding  affinity  (equation  2.3).  For  a  system  containing  antibody-­‐antigen  complex  initially  at  equilibrium,  this  model  also  predicts  that  the  amount  of  antibody-­‐antigen  complex  decreases  exponentially,  governed  solely  by  the  reverse  rate  constant,  when  the  free   29 (unbound)   state   of   one   of   the   interacting   molecules   is   completely   removed   from  solution  (equation  2.4).      By  considering  the  antibody-­‐antigen  interaction  as  two  independent  physical  processes   of   diffusive   transport   and   intrinsic   reaction,   the   forward   and   reverse  kinetic  rate  constants  can  be  expressed  as:  𝑘! = !!!!"!!!!!"     (2.5a)  𝑘! = !!!!""!!!!!"       (2.5b),  where   kon   and   koff   are   the   intrinsic   on-­‐rate   and   off-­‐rate   constants,   and   k+   is   the  characteristic   diffusive   rate   constant.95   The   diffusive   rate   constant   k+   is   highly  dependent   upon   the   geometric   and   solution   conditions   under   which   antibody-­‐antigen   interactions   occur.     If   antibody-­‐antigen   binding   occurs   on   a   spherical  surface,   as   when   antibodies   are   expressed   on   a   cell   surface,   the   diffusive   rate  constant  can  be  expressed  as:  𝑘! = 4𝜋𝐷𝑎 𝑁!       (2.6),  where  D  represents  the  effective  diffusion  coefficient,  a  represents  the  radius  of  the  spherical   binding   surface,   and  NA   represents   Avogadro’s   number.   If   one   assumes  that  the  antigen  is  a  small  protein  (D  <  10-­‐10  m2/s)  and  the  antibody  is  expressed  on  a  cell  with  diameter  of  1  to  10  μm,  the  diffusion  rate  constant  is  then  predicted  to  be  on  the  order  of  1012  M-­‐1s-­‐1.    Stringent  orientation  requirements  limit  kinetic  on-­‐rate  constants  for  most  antibody-­‐antigen  complexes  to  less  than  105-­‐107  M-­‐1s-­‐1.96,97  Thus,  antibody-­‐antigen  binding  on  the  cell  surface   is  reaction-­‐limited  (k+  >>  kon),  making  the   forward   and   reverse   rate   constants   equal   to   the   intrinsic   on-­‐rate   and   off-­‐rate   30 constants,   respectively   (equations   2.5).   The   half-­‐lives   of   antibody-­‐antigen  interactions  typically  range  from  several  minutes  to  several  hours,  corresponding  to   koff   values   greater   than   10-­‐3   -­‐   10-­‐4   s-­‐1.   Taken   together,   the   majority   of   naturally  produced  antibodies   therefore  bind   their  antigen  with  an  equilibrium  dissociation  constant   ranging   from   100   pM   to   10   µM.59,98   Antibodies   with   Kd   values   close   to  100pM   are   estimated   to   be   near   the   affinity   limits   that   can   be   selected   by   the  immune  system,  as  antibodies  that  bind  antigen  with  longer  half-­‐lives  than  cellular  endocytosis   rates   should   theoretically   provide   no   selective   advantage   during   the  adaptive  immune  response.93  It  is  possible,  however,  that  antibodies  with  Kd  values  less  than  100  pM  are  produced  by  the  immune  system  simply  by  chance.     2.2 Methods  and  Parameters  for  Antibody  Screening  and  Selection  Antigen  binding  affinity  and  selectivity  are  the  two  parameters  that  typically  determine   the   suitability   of   an   antibody   for   particular   research   and   therapeutic  applications.  For  instance,  therapeutic  antibodies  for  long-­‐term  protection  to  human  influenza   virus   would   ideally   cross-­‐react   with   a   variety   of   hemagglutinin   (HA)  surface  proteins  present  on  different  viral  strains.49  Conversely,  antibodies  used  for  treatment  of  cancerous  tumors  must  often  specifically  bind  a  particular  genetically  mutated   or   glycosylated   state   of   a   protein.99   The  most   useful   antibodies   for   both  research   and   therapeutic   applications   are   typically   those   that   bind   their   target  antigen   with   moderate   to   high   binding   affinities   (i.e.   equilibrium   dissociation  constants  less  than  or  equal  to  10  nM).12     31 Antibody   binding   affinity   and   selectivity   are   typically   assessed   using   an  enzyme-­‐linked   immunosorbent   assay   (ELISA)   or   related   assay,   in   which   varying  amounts  of  antibody  are  titrated  on  a  surface  with  bound  antigen  and  the  amount  of  bound   antibody-­‐antigen   complex   is   measured   using   a   secondary   antibody   with   a  fluorescent   reporter.   ELISA  measurements   do   not   provide   any   information   about  antibody-­‐antigen   binding   kinetics.     However,   there   are   numerous   applications   in  which   knowledge   of   antibody-­‐antigen   binding   kinetics   may   be   useful   for   the  selection  of  research-­‐grade  or  therapeutic  mAbs.    For  instance,  antibodies  with  high  on-­‐rate   constants   (>106  M-­‐1s-­‐1)  may  be  particularly  useful   for  diagnostics   and  bio-­‐sensing,  as  well  as  for  viral  neutralization.100,101  Conversely,  therapeutic  antibodies  that  bind  their  target  antigens  with  off-­‐rate  constants  less  than  10-­‐4  s-­‐1  (i.e.  half-­‐lives  of  hours  to  days)  could,  in  principle,  be  administered  in  lower  dosages,  reducing  the  cost  and  side-­‐effects  of  these  therapies.11,102      A  number  of  detection  techniques  exist  to  measure  antibody-­‐antigen  binding  kinetics,   including   surface   plasmon   resonance   (SPR)   spectroscopy,   fluorescence  polarization,   ellipsometry,   quartz   crystal   microbalance   (QCM)   sensing,   and  interferometry.103–108  SPR  spectroscopy,   the  most  widely  used  of   these   techniques,  facilitates   real-­‐time,   label-­‐free  detection  of  antigen  binding   to  surface-­‐immobilized  antibodies  by  detecting  refractive  index  changes  at  the  binding  surface.  SPR  arrays  have   previously   been   used   to   screen   antibodies   produced   by   phage   display   and  rabbit  hybridoma.109,110  However,  since  refractive  index  changes  are  proportional  to  the  mass  bound  to  the  sensor  surface,  SPR  spectroscopic  measurements  are  poorly  suited  for  the  detection  of  low  molecular  weight  molecules  (<200  Da),  as  well  as  low   32 abundance  samples  (<200  pg)  such  as  antibodies  secreted  by  single  cells.111  Back-­‐scattering   interferometry   (BSI)   is   an   alternative,   label-­‐free   technique   capable   of  measuring   binding   of   low   molecular   weight   molecules   and   has   lower   detection  limits  than  SPR  spectroscopy.106  However,  as  a  solution-­‐phase  method,  BSI  does  not  enable   direct   measurement   of   dissociation   kinetics,   cannot   be   easily   extended   to  make  multiplexed  kinetic  measurements  of  multiple  antibody-­‐antigen   interactions,  and   is   limited   in   its   ability   to   measure   binding   kinetics   in   complex   mixtures.    Moreover,  in  both  BSI  and  SPR  spectroscopy,  measurement  sensitivity  is  affected  by  temperature   fluctuations   and   bulk   refractive   index   shifts   during   buffer   exchange.  More   importantly,   neither   these   nor   other   currently   available   methods   allow   for  measurement   of   antibody-­‐antigen   binding   kinetics   from   very   low   abundance  samples.     To   address   this   need,   a   microfluidic   fluorescence   bead   assay   was  developed   to   enable   screening   of   antibody-­‐antigen   association   and   dissociation  kinetics  in  order  to  characterize  and  select  antibodies  secreted  by  single  cells.     2.3 Materials  and  Methods   2.3.1 Microfluidic  Device  Fabrication  and  Control  All  microfluidic  devices  were  fabricated  using  multilayer  soft  lithography.86,87  Devices  were  composed  of  two  layers  of  poly(dimethylsiloxane)  (PDMS)  elastomer  (GE  RTV  615)  bonded   to  No  1.5  glass  coverslips   (Ted  Pella,   Inc.).  The  microfluidic  device  was  fabricated  with  a  push-­‐down  geometry,  in  which  the  flow  channels  were  bonded   by   oxygen   plasma   directly   to   the   cover-­‐glass,   while   the   control   channels  were   situated   above   the   flow   channels,   separated  by   a   thin   (~10  µm),  deflectable   33 PDMS  membrane.   Thus,   reagent   samples  were   brought   in   direct   contact  with   the  cover-­‐glass,  allowing  for  imaging  of  the  device  with  a  100X  high  numerical  aperture  (N.A.  1.30)  oil-­‐immersion  objective  (working  distance  ~  200  µm).  The  devices  were  designed   in  AutoCAD  software   (Autodesk)  and  printed  on  high-­‐resolution   (20,000  dpi)   transparency   masks   (CAD/Art   Services).   Master   molds   were   fabricated   in  photoresist   on   silicon  wafers   (Silicon  Quest)   by   standard   optical   lithography.   The  control  master  molds  were  fabricated  out  of  20-­‐25  µm  high  SU-­‐8  2025  photoresist  (Microchem).  The  flow  master  molds  were  fabricated  with  12  µm  rounded  SPR220-­‐7.0  photoresist  channels  (Rohm  and  Haas)  and  6µm  SU-­‐8  5  photoresist  (Microchem)  channels  with  rectangular  cross-­‐section.  Microfluidic  valves  were  actuated  at  30  psi  pressure   that   was   controlled   using   off-­‐chip   solenoid   valves   (Fluidigm   Corp)  controlled   using   LabView   7.1   software   and   a   NI-­‐6533   DAQ   card   (National  Instruments).  Compressed  air  (3-­‐4  psi)  was  used  to  push  reagent  solutions  into  and  through  the  device.     2.3.2 Reagent  Preparation  Protein   A-­‐coated   5.5   µm   diameter   polystyrene   beads   (Bangs   labs)   were  incubated   with   a   1   mg/mL   solution   of   Rabbit   anti-­‐mouse   polyclonal   antibodies  (pAbs)   purchased   from   Jackson   Immunoresearch   and   used   without   further  purification.  All  antibody  and  antigen  solutions  were  prepared  in  PBS/BSA/Tween  solution  consisting  of  1X  PBS,  pH  7.4  (Gibco)  with  10  mg/mL  BSA  (Sigma)  and  0.5%  Polyoxyethylene   (20)   sorbitan   monolaurate   (similar   to   Tween-­‐20,   EMD  Biosciences).  Lysozyme   from  chicken  egg  white   (HEL)  was  purchased   from  Sigma,   34 and  the  D1.3  and  HyHEL-­‐5  mouse  monoclonal  antibodies  (mAbs)  to  lysozyme  were  generously  provided  by  Dr.  Richard  Willson  (University  of  Houston).  The  anti-­‐GFP  mouse  mAb   (LGB-­‐1)   was   purchased   from   Abcam.   Fluorescent   protein   conjugates  were   prepared   using   Dylight488   and   Dylight633   NHS   esters   (Pierce)   and   were  purified   using   Slide-­‐A-­‐Lyzer™   dialysis   cassettes   (Pierce).   The   concentration   of  fluorescent   conjugates  was  measured  by   spectrophotometry   (Nanodrop).   In  order  to  minimize  protein  denaturation,  fluorescent  HEL  conjugates  were  labeled  at  a  dye-­‐to-­‐protein   ratio   (D/P)   of   less   than   1,  whereas   the  D1.3-­‐Dylight488   conjugate  was  prepared  at  a  D/P  of  ~5.     2.3.3 Fluorescence  Microscopy  The   microfluidic   devices   were   imaged   on   a   Nikon   TE200   Eclipse   inverted  epifluorescence   microscope   equipped   with   green   (470/40   nm   excitation,   535/30  nm   emission)   and   red   (600/60   nm   excitation,   655   nm   long-­‐pass   emission)  fluorescence   filter   cubes   (Chroma   Technology).   Fluorescence   images   were   taken  using  a  16-­‐bit,  cooled  CCD  camera  (Apogee  Alta  U2000)  and  a  100X  oil   immersion  objective  (N.A.  1.30,  Nikon  Plan  Fluor).  The  fluorescence  sensitivity  was  adjusted  by  binning   pixels   on   the   CCD   detection   camera   and   by   modulating   the   fluorescence  exposure  times  (20  ms  -­‐  1  s)  with  a  computer-­‐controlled  mechanical  shutter  (Ludl  Electronic  Products).  During  antibody-­‐antigen  binding  experiments,  the  image  focal  position   was   held   constant   on   the   center   of   the   beads   by   minimizing   the   bead  diffraction  pattern  observed  under  bright-­‐field  illumination.     35 2.3.4 Cell  Culture    Mouse  D1.3  hybridoma  cells  were  grown  in  6  mL  petri  dishes  (Nunc)  using  RPMI   1640   medium   (Gibco)   with   10%   fetal   calf   serum   (FCS)   in   a   cell   culture  incubator  (37°C,  5%  CO2).  Cells  were  passaged  approximately  once  a  week  by  serial  dilutions   (5-­‐fold)   in   fresh   medium.   Prior   to   loading   into   the   device,   cells   were  washed  by  centrifugation  at  1500  rpm  and  re-­‐suspended  in  1X  PBS,  pH  7.4  (Gibco)  in   order   to   remove   free   antibodies   in   the   cell   medium.   Cell   concentration   was  quantified  using  a  haemocytometer  and  brightfield  microscope.     2.3.5 Assay  Operation  The  microfluidic   device   consisted   of   six   flow   input   channels,   each   used   for  loading  a  distinct  reagent  and  controlled  with  an  independent  control  valve,  which  join   into  a  common   flow  output  channel   (Figure  2.2  A  and  B).  The  output  channel  was  partitioned   into  discrete  ~200  pL  chambers  by  actuating  a  set  of  microfluidic  “sieve”  valves  which,  when  actuated,  acted  as  filters  to  immobilize  large  particles  (>  1µm)  while  still  allowing  fluid  exchange.83  The  microfluidic  device  consisted  of  low  fluidic   dead   volume   upstream   of   the   bead   capture   area   (<4   nL),   such   that   this  volume  was   displaced   in   approximately   1   second   based   on   the   typical   flow   rates  used  in  this  study  (~10  µL/hr).       36   Figure   2.2         Microfluidic   fluorescence   bead   measurements   of   antibody-­‐antigen   binding   kinetics.     (A)   Device   schematic   showing   control   channels   (orange)   for   selecting   six   reagent   inlets  (blue)  and  actuating  sieve  valves  on  the  reagent  outlet  channel  (green).    (B)  Microscope   image  of  device  with  food  coloring  to  visualize  distinct  reagent  inlets  (yellow  and  green)  and   control  channels  (red).    (Insets)  Brightfield  (top)  and  fluorescence  (bottom)  images  of  beads   trapped  using   sieve   valves   at   20X   and  100X  magnification,   respectively.   [continued  on  next   page]   Control layer Flow layer Glass slide Open State Closed state Valve Control Channels Flow Input Channels Flow Output Channel A Control layer Flow layer Glass slide Sieve valve Open State Closed state B 100µm 10µm 20µm 37   Figure   2.2   [continued   from   previous   page]   (C-­‐E)   Schematics   of   bead   assay   for   direct   measurement  of  association  and  dissociation  kinetics  of  immobilized  mAbs  and  fluorescently   labeled  antigen  (C  and  D,  respectively),  and  indirect  measurement  of  dissociation  kinetics  of   immobilized   mAbs   and   unlabeled   antigen   molecules   (E).   Adapted   with   permission   from   Singhal  et  al.112  (American  Chemical  Society,  2010).     At   the   start   of   the   experiment,   the   flow  output   channel  was   flushed  with   a  PBS/BSA/Tween  solution  from  the  top  and  bottom  flow  inlets  in  order  to  pre-­‐coat  the   hydrophobic   channel   walls   to   reduce   nonspecific   binding.   Next,   a   solution  containing  Protein  A  beads  coated  with  Rabbit  anti-­‐mouse  pAb  (d  =  5.5  µm,  ~106-­‐107  beads/mL)  was  loaded  through  the  device  to  the  fluidic  outlet.    The  microfluidic  sieve  valves  were  then  actuated  to  immobilize  the  beads  against  the  traps,  and  the  fluidic  outlet  channel  was  again  washed  for  1  min  with  PBS/BSA/Tween  solution  to  remove  any   free  rabbit  pAb   in  solution.  The  beads  were  then   incubated   for  ~1-­‐10  min  with  a  1-­‐100  µg/mL  solution  containing  the  mouse  mAb  of  interest.  Again,  free  mouse   antibody   was   washed   out   of   the   fluidic   output   channel   using  PBS/BSA/Tween   solution   for   1   min.   To   measure   the   rate   of   antibody-­‐antigen  association,  the  beads  were  flushed  with  a  solution  of  fluorescently  labeled  antigen  and   fluorescently   imaged   at   defined   time   intervals   (Figure   1C).   When   chemical  equilibrium  between  the  antibody  and  antigen  was  reached,  as  detected  by  a  plateau  in   bead   fluorescence,   the   beads   were   flushed   with   PBS   buffer   and   imaged   to  measure   the  rate  of  antibody-­‐antigen  dissociation.  The  process  was  repeated  with  multiple  solutions  of  varying  concentrations  of  fluorescently  labeled  antigen  (10  pM  –  500  nM),  each  loaded  onto  the  microfluidic  device  from  a  separate  fluidic  inlet.  At  the  concentrations  used  in  this  study,  we  did  not  detect  any  increase  in  fluorescence  intensity   relative   to   the   bead   autofluorescence   when   fluorescent   antigen   was  flushed  over  control  beads  without  antigen-­‐specific  mouse  mAbs.      A   second  version  of  microfluidic  bead  assay  was   implemented   to   indirectly  measure  dissociation  kinetics  between  antibodies  and  unlabeled  antigen  molecules   39 by   displacement  with   fluorescently   labeled   antigen   (Figure   1D).   In   this   assay,   the  antibody   of   interest   was   captured   on   Rabbit   anti-­‐mouse   pAb-­‐coated   Protein   A  beads,   and   the   beads   were   subsequently   washed   with   unlabeled   antigen   at   high  concentration   (>1   µM)   to   saturate   all   antibody   binding   sites.   Beads   were   then  washed  with   a   solution  of   fluorescently   labeled   antigen   (10  nM)  while   imaging   at  defined   time   intervals.  Dissociation  of   the  unlabeled   antigen  was   then   inferred  by  accumulated  fluorescence  on  the  beads.  In  order  to  measure  the  antigen  binding  kinetics  from  antibodies  secreted  by  single  cells,  a  solution  of  RPMI-­‐1640  medium  containing  freshly  washed  hybridoma  cells   (~105   cells/mL)   was   loaded   into   the   device.   The   control   valve   was  momentarily  opened  to  allow  for  a  single  hybridoma  cell  to  be  trapped  by  the  first  sieve   valve   in   the   fluidic   outlet   channel.   Subsequently,   a   solution   of   RPMI-­‐1640  medium  containing  Protein  A  beads  coated  with  Rabbit  anti-­‐mouse  pAb  (d  =  5.5  µm,  ~106-­‐107  beads/mL)  was  loaded  into  the  device,  such  that  1-­‐2  beads  were  brought  into  close  proximity  with  the  hybridoma  cell.  The  hybridoma  cell  was  then  allowed  to   incubate  next  to  the  beads  for  1  hour,  and  subsequently  washed  for  1  min  with  PBS/BSA/Tween  buffer  to  wash  out  any  free  antibody  in  solution  and  halt  antibody  secretion   from   the   cell.   Kinetic   measurements   of   antibody-­‐antigen   binding   were  then   performed   while   flushing   the   beads   with   cycles   of   increasing   fluorescent  antigen  solution  (~5-­‐50  nM)  and  PBS/BSA/Tween  buffer.         40 2.3.6 Data  Analysis      Fluorescent   images   were   analyzed   using   MaximDL   4   imaging   software.  Fluorescent  intensities  were  measured  by  selecting  line  profiles  through  the  beads  and  recording  the  maximum  intensity  at  the  bead  surface  (Appendix  A).  Analysis  of  multiple   beads   in   a   single   field-­‐of-­‐view   during   antibody-­‐antigen   binding  experiments   confirmed   that   measured   binding   kinetics   were   insensitive   to  systematic  variations  caused  by  non-­‐uniform  binding  of  antigen  to  the  bead  surface,  differences   in   bead-­‐to-­‐bead   binding   capacity,   variation   in   position   in   the   flow  channel  and  non-­‐uniform  illumination  over  the  field  of  view  (Figure  2.3).  Error  in  all  measurements  was  estimated  to  be  less  than  10%  of  the  fluorescence  intensity.    The  measured   fluorescence   bead   intensities   were   assumed   to   be   proportional   to   the  concentration   of   antibody-­‐antigen   complex   ([AbAg])   and  were   fit   to   the   following  standard   first-­‐order   mass-­‐action   equations   (Section   2.1.1)   using   nonlinear   least  squares  minimization:  𝐹 𝑡 = 𝐹!"# − 𝐹! [!"]![!!]!!!! 1 − 𝑒! !!"[!"]!!!!"" ! + 𝐹!   (2.7a)  𝐹 𝑡 = 𝐹!"# − 𝐹! [!"]![!"]!!!! 𝑒!!!""! + 𝐹!         (2.7b)  𝐹 𝑡 = 𝐹!"# − 𝐹! [!"]![!"]!!!! + 𝐹!           (2.7c),  in   which   F(t)   represents   the   measured   bead   fluorescence   at   time   t,     F0   and   Fmax    represent   the   background   and   maximum   bead   fluorescence,   respectively,   [Ag]0  represents   the   solution   concentration   of   antigen   (in   M),   Kd   is   the   equilibrium  dissociation  constant  (in  M),  and  kon  and  koff  represent  the  intrinsic  association  and   41 dissociation  rate  constants,  in  units  of  M-­‐1s-­‐1  and  s-­‐1,  respectively.  In  addition  to  the  binding   rate   constants   (Kd,   kon   and   koff),   F0   and   Fmax   were   constants   fitted   by   the  model.   In   agreement  with   this  model,   all   measured   antibody-­‐antigen   interactions  obeyed   simple   bimolecular   association   and   first-­‐order   dissociation   kinetics.     All  reported  errors  represent  the  calculated  standard  deviation  from  multiple  replicate  measurements.       Figure  2.3        Antibody-­‐antigen  association  kinetics  measured  from  multiple  beads  in  a  single   field-­‐of-­‐view   (FOV).     In   this   experiment,   fluorescently   labeled   hen   egg   lysozyme   is   binding   bead-­‐immobilized   anti-­‐HEL   D1.3   mouse   mAb.     Reported   error   represents   the   calculated   standard  deviation  from  multiple  replicate  measurements.  Dissociation  kinetics  measured  on   multiple  beads  in  a  single  FOV  were  also  consistent  to  within  20%    (data  not  shown).     2.4 Results  Measurements   of   antibody-­‐antigen   binding   kinetics   using   the   microfluidic  fluorescence  bead  assay  were  validated  using  the  model  antigen  hen  egg  lysozyme   kon = (1.8 ± 0.2) × 106 M-1s-1 0 5000 10000 15000 20000 25000 30000 35000 0 1 2 3 4 5 M ax F lu or es ce nc e In te ns ity (a .u .) Time (min) 5 4 3 2 1 1 5 4 32 42 (HEL).     HEL   is   a   14.7   kDa   protein   of   known   structure   that   can   hydrolyze  polysaccharides,   such   as   those  present   in  bacterial   cell  walls.113  HEL   is   frequently  used  as  a  model  antigen   in   immunological  research  because   it   is   inexpensive,  very  soluble  in  water  (~20  mg/mL),  and  highly  immunogenic  in  mice.    The  latter  fact  has  enabled   the   production   of   dozens   of   hybridoma   cell-­‐lines   secreting   monoclonal  antibodies  that  bind  HEL  with  a  wide  range  of  affinities  (10  µM  >  Kd  >  10  pM).114  Of  these,  two  particular  mAbs,  D1.3  and  HyHEL-­‐5,  have  significantly  different  binding  kinetics   and   were,   thus,   selected   for   testing   and   validation   of   the   microfluidic  fluorescence  bead  assay    (Figure  2.4A  and  2.4B  and  Table  2.2).    Association   and   dissociation   rate   constants   measured   in   device   for   the  D1.3/HEL   interaction   were   1.87   ±   0.48   ×   106   M-­‐1s-­‐1   and   2.10   ±   0.25   ×   10-­‐3   s-­‐1,  respectively,  and  were  consistent  with  values  of  1.0  -­‐  2.0  ×  106  M-­‐1s-­‐1  and  1.15  -­‐  3.04  ×  10-­‐3  s-­‐1  previously  measured  using  surface  plasmon  resonance  (SPR)  spectroscopy,  stopped-­‐flow   fluorescence   quenching,   and   competitive   ELISA.115,116   A   ten-­‐fold  smaller  association  rate  constant  previously  reported  for  the  D1.3/HEL  interaction  (1.67  ×  105  M-­‐1s-­‐1)  may  be  attributed  to  differences  between  the  full  D1.3  mAb  used  in   the   microfluidic   bead-­‐based   measurements   and   the   recombinant   single-­‐chain  antibody  fragment  used  by  Bedouelle  and  coworkers.117    A  variation  of  the  microfluidic  bead  assay  using  fluorescently-­‐labeled  HEL  as  a  competitive  antigen  was  used  to  indirectly  measure  the  dissociation  rate  constant  between  D1.3  mAb  and  unlabeled  HEL   (Figure  2.2E   and  2.4D).   In   this   assay,  D1.3  mAbs   immobilized   on   beads   were   first   saturated   with   unlabeled   HEL   and  subsequently   washed   with   fluorescently-­‐labeled   HEL.   Measurements   of   the   43 accumulated   bead   fluorescence   faithfully   reflected   the   D1.3/HEL   dissociation  kinetics  provided  the  labeled  HEL  was  at  a  sufficiently  high  concentration  to  ensure  that   dissociation  was   rate-­‐limiting   (i.e.   kon[Ag]   >   koff   ,   or,   equivalently,   [Ag]   >  Kd).  Using   this  method,   the   dissociation   rate   constant   of   D1.3   and   unlabeled  HEL  was  measured   to   be   1.45  ±   0.30   ×   10-­‐3   s-­‐1,   in   close   agreement  with   direct   dissociation  measurements  between  D1.3  and  fluorescently-­‐labeled  HEL  (Table  2.2).       Table  2.2     Antibody-­‐antigen   binding   kinetics   measured   using   the   microfluidic   fluorescence  bead  assay.    Reported   error   represents   the   calculated   standard   deviation   of   multiple   replicate   measurements.  Data  taken  from  Singhal  et  al.112       Antibody/Antigen  pair   kon  (M-­‐1s-­‐1)   koff  (s-­‐1)   Kd  =  koff  /  kon  D1.3  mAb/HEL-­‐Dylight488   (1.87  ±  0.48)×106   (2.10  ±  0.25)×10-­‐3   1.20  ±  0.42nM  D1.3  mAb/HEL-­‐Dylight633   (1.27  ±  0.22)×106   (2.15  ±  0.23)×10-­‐3   1.75  ±  0.46nM  HyHEL-­‐5  mAb/HEL-­‐Dylight488   (5.75  ±  0.71)×106   (1.69  ±  0.30)×10-­‐4   30.0  ±  7.4pM  LGB-­‐1  mAb/EGFP   (5.00  ±  0.72)×104   (5.15  ±  0.89)×10-­‐3   106  ±  28nM     Figure   2.4         Microfluidic   fluorescence   bead   measurements   of   antibody-­‐antigen   binding   kinetics.     Direct   fluorescent   measurements   of   association  and  dissociation  kinetics  of   (A)  D1.3  mAb  and  HEL-­‐Dylight488  conjugate,   (B)  HyHEL-­‐5  mAb  and  HEL-­‐Dylight488  conjugate,   (C)   LGB-­‐1  mAb  and  enhanced  green  fluorescent  protein  (EGFP).  (D)  Indirect  measurement  of  dissociation  kinetics  of  D1.3  mAb  and  HEL  using   HEL-­‐Dylight488  conjugate.  Solid  lines  represent  experimental  fits  using  mass-­‐action  equations  (equations  2.7a-­‐c).  Reported  error  represents   the   calculated   standard   deviation   of  multiple   replicate  measurements.     Adapted  with   permission   from   Singhal   et   al.   (American   Chemical   Society,  2010).112     0 0.2 0 .4 0 .6 0 .8 1 1.2 0 5 10 15 20 25 30 35 40 N or m al iz ed F lu or es ce nc e T im e (m in ) 125µg/m L 80µg/m L 40µg/m L 33µg/m L 20µg/m L 0 0.2 0 .4 0 .6 0 .8 1 1 .2 0 0 .25 0 .5 0 .75 1 0 0.2 0 .4 0 .6 0 .8 1 1.2 0 10 20 30 40 50 60 70 N or m al iz ed F lu or es ce nc e T im e (m in ) 428ng/m L 214ng/m L 107ng/m L 54ng/m L 42.8ng/m L 21.4ng/m L 0 5000 10000 15000 20000 25000 30000 35000 40000 45000 50000 0 10 20 30 40 50 60 70 B ea d F lu or es ce nc e T im e (m in ) koff = 1 .45 ± 0.30 X 10 -3 s -1 0 0 .2 0 .4 0 .6 0 .8 1 1.2 0 50 100 150 200 250 300 350 400 N or m al iz ed F lu or es ce nc e T im e (m in ) 214ng/m L 107ng/m L 54ng/m L 21.4ng/m L 10.7ng/m L 0 0.2 0 .4 0 .6 0 .8 1 1 .2 0 1 2 3 4 A B DC In  comparison  to  the  D1.3  mAb,  HyHEL-­‐5  binds  HEL  with  a  nearly  four-­‐fold  larger   association   rate   constant   (5.75   ±   0.71   ×   106   M-­‐1s-­‐1)   and   ten-­‐fold   smaller  dissociation   rate   constant   (1.69  ±   0.30   ×   10-­‐4   s-­‐1)   (Figure   2.4B).   Thus,  HyHEL-­‐5   is  found   to  bind  HEL  with  a  ~40-­‐fold  smaller  equilibrium  dissociation  constant   than  D1.3  (30  pM  vs.  1.2  nM)  (Table  2.2).  Previous  measurements  of   the  HyHEL-­‐5/HEL  interaction   using   particle-­‐counting   fluorescence   immunoassay   (PCFIA)   and  stopped-­‐flow  fluorescence  polarization  produced  a  similar  equilibrium  dissociation  constant  (25  pM)  and  dissociation  rate  constant  (2.2  ×  10-­‐4  s-­‐1),  but  a  three-­‐  to  five-­‐fold   larger   association   rate   constant   (1.5–3.3×107  M-­‐1s-­‐1).105   Immobilization  of   the  mAb  in  the  microfluidic  bead  assay  may  result  in  slower  association  kinetics  when  compared  with   solution-­‐phase   fluorescence  polarization  measurements.    HyHEL-­‐5  mAb  and  HEL  are   known   to  bind  with  near  diffusion-­‐limited  kinetics,   a   regime   in  which   the   association   rate   constant   scales   linearly   with   the   effective   diffusion  coefficient   (equation   2.6,   where   D   ≅   DmAb   +   DHEL).95,105   Since   the   translational  diffusion   coefficient   of  HEL   is  much   larger   than   that   of   the  HyHEL-­‐5  mAb   (DHEL   ≥  3×DmAb),   immobilization   of   the   mAb   would   reduce   the   apparent   association   rate  constant   by   less   than   25%.118,119   Similarly,  mAb   immobilization  would   reduce   the  effective   rotational   diffusion   by   less   than   10%,   based   on   the   rotational   diffusion  coefficients  of  HEL  and  the  three-­‐  to  five-­‐fold   larger  HyHEL-­‐5  mAb  molecule  (DR  ∝  1/R3  where  R   =   radius   of   the  molecule).118,120   Although  mAb   immobilization  does  not  have  an  appreciable  effect  on  effective  diffusion  of  the  HyHEL-­‐5/HEL  pair,   it   is  possible   that   binding   of   HEL   to   bead-­‐immobilized   HyHEL-­‐5  mAb   results   in   steric  hindrance   of   adjacent   mAb   molecules,   resulting   in   slower   association   kinetics   46 measured   using   the   bead   assay   when   compared   to   solution-­‐phase   fluorescence  polarization  measurements.  To   demonstrate   that   the   microfluidic   bead   assay   can   be   used   to   measure  binding   kinetics   of   a   previously   uncharacterized   antibody,   binding   kinetics   were  measured  for  a  commercially  available  mouse  monoclonal  antibody  (LGB-­‐1,  Abcam)  to  enhanced  green  fluorescent  protein  (eGFP)  (Figure  2.4C).  In  this  instance,  native  eGFP  fluorescence  was  measured,  eliminating  the  need  for  an  exogenous  fluorescent  label.    The  measured  association  and  dissociation  rate  constants  for  the  LGB-­‐1/eGFP  interaction   were   5.00   ±   0.72   ×   104   M-­‐1s-­‐1   and   5.15   ±   0.89   ×   10-­‐3   s-­‐1,   respectively  (Table  2.2).  Collectively,   the   measured   binding   kinetics   of   the   anti-­‐lysozyme   and   anti-­‐eGFP   mAbs   span   nearly   four   orders   of   magnitude   in   equilibrium   dissociation  constants   (30  pM   to  0.1  µM),  with  association   rate   constants  varying   from  5×104-­‐106  M-­‐1s-­‐1  and  dissociation  rate  constants  ranging   from  10-­‐3-­‐10-­‐4   s-­‐1   (Table  2.2).   In  principle,  the  microfluidic  bead  assay  can  be  used  to  characterize  stronger  antibody-­‐antigen  interactions  than  the  HyHEL-­‐5/HEL  interaction;  however,  measurements  of  binding   interactions   with   dissociation   rate   constants   lower   than   10-­‐4s-­‐1   must   be  taken  over  several  days  or  weeks.  On  the  other  hand,  the  bead-­‐based  assay  can  be  readily   used   to   measure   binding   interactions   weaker   than   the   LGB-­‐1/eGFP  interaction.   By   optimizing   the   fluidics   for   rapid   solution   exchange   (<1   s),   the  practical   upper   limit   in  measurable   rate   constants   is   approximately   10-­‐1   s-­‐1.   Thus,  the   microfluidic   bead-­‐based   assay   should   enable   characterization   of   antibody-­‐ 47 antigen  interactions  that  span  greater  than  five  orders  of  magnitude  in  kinetic  rate  constants  and  at  least  seven  orders  of  magnitude  in  equilibrium  binding  affinities.         2.4.1 Microfluidic   Fluorescence   Bead   Measurements   Reflect   Intrinsic   Antibody-­‐Antigen  Binding  Kinetics.      A  series  of  experiments  were  performed   in  order   to  verify   that  bead-­‐based  fluorescence   measurements   reflected   intrinsic   antibody-­‐antigen   binding   kinetics,  and   were   unaffected   by   artifacts   arising   from   fluorescent   labeling   of   the   antigen,  antibody  immobilization,  or  mass  transport  effects.      Fluorescent   labeling   of   HEL   did   not   alter   the   intrinsic   D1.3/HEL   binding  kinetics,   as   indicated   by   the   agreement   between  microfluidic   bead  measurements  using   fluorescently   labeled  HEL  and  previously  reported  measurements  using  SPR  spectroscopy  with   unlabeled  HEL.121   In   addition,   no   differences  were   observed   in  bead-­‐based  kinetic  measurements  of  the  D1.3  mAb  binding  to  HEL  labeled  with  two  different   fluorophores,   Dylight488   and   Dylight633   (Table   2.2).   Photobleaching   of  fluorophores   did   not   affect   the   measured   binding   kinetics,   as   was   confirmed   by  measuring   the   photobleaching   rates   of   the   fluorescent   dyes   used   in   this   study  (Dylight488,  Dylight633,  and  eGFP)  and  selecting  fluorescence  exposure  times  (e.g.  100ms)  that  resulted  in  less  than  5%  reduction  in  bead  fluorescence  (Figure  2.5A).      Indeed,   measured   binding   kinetics   were   consistent   over   a   large   range   of  fluorescence  exposure  times  (20  –  500  ms),  whereas  exposure  times  of  greater  than  1   s   resulted   in   substantial   photobleaching   and   an   artificial   increase   in   measured   48 association   and   dissociation   binding   kinetics  when   compared   to   intrinsic   kinetics  (Figure  2.5B).  Association   and   dissociation   rate   measurements   for   the   D1.3/HEL  interaction   were   also   unaffected   by   the   antibody   bead   immobilization   chemistry  employed,   as   demonstrated   by   measurements   using   both   silica   and   polystyrene  beads  coated  with  either  rabbit  or  goat  anti-­‐mouse  polyclonal  antibody  (Figure  2.6).  We   further   verified   that  multivalent   binding   between   the   rabbit   anti-­‐mouse   pAbs  and  fluorescently-­‐labeled  D1.3  mAb  resulted  in  no  detectable  dissociation  over  the  course   of   3   days,   which   would   otherwise   artificially   accelerate   the   measured  antibody-­‐antigen  binding  kinetics  (Figure  2.7).  Indeed,  the  nearly  irreversible  bond  between   rabbit   pAb   and   the   mouse   mAbs   was   critical   for   this   bead   assay,   as  attempts   to   measure   D1.3/HEL   binding   kinetics   using   Protein   A   beads   without  Rabbit   anti-­‐mouse   pAbs   were   unsuccessful   due   to   rapid   dissociation   (and   low  affinity)  of  protein  A  /  mouse  mAb  complexes  (data  not  shown).         49     Figure  2.5        Effect  of  fluorophore  stability  on  measured  antibody-­‐antigen  binding  kinetics.    (A)   Photobleaching   rates   of   fluorescent   dye  molecules   under  100W  Hg   lamp   illumination  using   100X  oil-­‐immersion  objective  (NA  1.30).  (B)  Effect  of  fluorescent  exposure  times  on  measured   association   kinetics   of   D1.3   mAb   and   HEL-­‐Dylight488.   Reported   error   represents   the   calculated   standard   deviation   of   multiple   replicate   measurements.   Values   measured   only   once  are  reported  without  error  bars.  Adapted  with  permission  from  Singhal  et  al.  (American   Chemical  Society,  2010).112         0 0 .2 0 .4 0 .6 0 .8 1 1 .2 0 5 10 15 20 25 30 35 40 45 N or m al iz ed F lu or es ce nc e T im e (s ) Dyligh t488 Dyligh t633 EGFP 1.00E-02 1 .00E-01 1 .00E+00 10 100 1000 As so ci at io n R at e (s -1 ) E xposure tim e (m s) A B k bleach = 9 .19 X 10 -2 s -1 k bleach = 9 .40 X 10 -2 s -1k bleach = 3 .97 X 10 -1 s -1 50   Several  experiments  were  conducted  to  verify  that  mass  transport,  including  diffusion   limitation,   did   not   affect   microfluidic   bead   measurements   of   antibody-­‐antigen  binding  kinetics.    In  the  diffusion-­‐limited  regime,  antibodies  adjacent  on  the  bead   surface   compete   for   fluorescent   antigen,   thus   reducing   the   apparent  association   rate   constant.   Similarly,   antigen   that   dissociates   and   then   rebinds   to  adjacent   antibodies   would   reduce   the   apparent   dissociation   rate.25,21   However,  nearly   identical   association  and  dissociation  kinetics  were  measured   for   the  D1.3-­‐HEL   interaction   by   varying   the   amount   of   bead-­‐immobilized   D1.3   mAb   over   two  orders   of   magnitude   (Figure   2.10B).   Moreover,   dissociation   kinetics   of   the   D1.3  antibody   and   fluorescently   labeled   HEL   were   similar   both   in   the   presence   and  absence  of  a  high  concentration  (~2mg/mL)  of  competitive  unlabeled  HEL  antigen  (Figure  2.8).  Thus,  no  diffusion  limitation  was  observed  in  the  form  of  competition  between   antibodies   adjacent   to   one   another   on   the   beads.   Association   and  dissociation   rate   constants   of   the   D1.3-­‐HEL   interaction   remained   constant   over   a  range  of  flow  rates  from  3-­‐14  µL/hr,  suggesting  no  other  effects  of  mass  transport  on  the  measured  kinetics  (Figure  2.9).     Figure  2.6             Effect   of   different   bead   composition   and   capture   antibodies   on   measured   antibody-­‐antigen   binding   kinetics.   Measured   binding  kinetics   and   affinities   from  both   conditions  were   consistent  within   experimental   error   (see  Table   2.1).   Adapted  with  permission   from  Singhal  et  al.112  (American  Chemical  Society,  2010).112     0 0 .2 0 .4 0 .6 0 .8 1 1 .2 0 5 10 15 20 25 30 35 40 45 N or m al iz ed F lu or es ce nc e T im e (m in) G oat an tiM s C oated S ilica B eads R abb it antiM s C oa ted P rote in A Po lystyrene B ead Bead / Capture pAb kon (M-1s-1) koff (s-1) Kd (nM) Goat anti-Mouse pAb-coated silica beads 1.7 106 2.4 10-3 1.4 Rabbit anti-Mouse pAb-coated polystyrene beads 1.1 106 2.1 10-3 1.9 52   Figure  2.7             Measured   dissociation   kinetics   of   mouse   mAb   from   antibody   capture   beads.   No   dissociation   of   D1.3   mAb-­‐Dylight488   conjugate  from  Rabbit  anti-­‐Ms  pAb  coated  beads  was  observed  over  3  days.  Reported  error  represents  the  calculated  standard  deviation  of   multiple  replicate  measurements.  Adapted  with  permission  from  Singhal  et  al.112  (American  Chemical  Society,  2010).112           5000 7500 10000 12500 15000 0 1000 2000 3000 4000 5000 Fl uo re sc en ce In te ns ity T im e (m in) 53   Figure  2.8           Effect  of  antigen  re-­‐binding  on  measured  antibody-­‐antigen  dissociation  kinetics.    Dissociation  kinetics  of  D1.3  mAb  and  HEL-­‐ Dylight488   conjugate   were   similar   both   in   the   presence   and   absence   of   a   large   concentration   of   competitive   antigen   (2   mg/mL   HEL).   Measured  binding  kinetics  from  both  conditions  were  consistent  within  experimental  error  (see  Table  2.1).  Adapted  with  permission  from   Singhal  et  al.112  (American  Chemical  Society,  2010).112       0 0.2 0 .4 0 .6 0 .8 1 1.2 0 5 10 15 20 25 30 35 40 45 N o rm al iz ed F lu o re sc en ce T im e (m in) w ithout H E L w ith 2m g/m L H E L ko! (s-1) With 2mg/mL HEL 2.23×10-3 Without HEL 2.00×10-3 54   Figure  2.9           Effect  of  mass  transport  on  measured  antibody-­‐antigen  binding  kinetics.  Association  and  dissociation  kinetics  of  D1.3  mAb   and  HEL-­‐Dylight488  conjugate  were  similar  over  a  range  of  flow  rates  (~3-­‐14  µL/hr).  Fixed  error  bars  represent  the  calculated  ratio  of  the   standard   deviation   to   mean   value   of   measured   D1.3/HEL   kinetic   rate   constants   reported   in   Table   2.1   (25%   and   10%   for   kon   and   koff,   respectively).  Adapted  with  permission  from  Singhal  et  al.112  (American  Chemical  Society,  2010).112     0.00E+00 1.00E+06 2.00E+06 3.00E+06 4.00E+06 5.00E+06 0.00E+00 5.00E-04 1.00E-03 1.50E-03 2.00E-03 2.50E-03 3.00E-03 0 5 10 15 k o n ( M -1 s-1 ) k o (s -1 ) Flow rate (μL / hour) ko kon 2.4.2 Microfluidic   Fluorescence   Bead  Measurements   Exhibit   Low  Detection   Limits  and  Minimal  Sample  Consumption.      In   order   to   quantify   the   limit   of   detection   (LOD)   and   minimal   sample  consumption  required   for   the  microfluidic  bead  assay,  measurements  of  antibody-­‐antigen   binding   kinetics   were   made   using   varying   amounts   of   bead-­‐immobilized  mAb.   Using   the   measured   kinetic   on-­‐rate   constant   for   fluorescently-­‐labeled   D1.3  mAb  binding  to  Rabbit  anti-­‐mouse  pAb  coated  Protein  A  beads  (kon  =  1.10  ±  0.11  ×  106  M-­‐1s-­‐1),  the  amount  of  bead-­‐immobilized  D1.3  mAb  was  varied  over  two  orders  of  magnitude  by  modulating  the  loading  time  of  mAb  on  the  bead  (Figure  2.10).  The  limit   of   detection   (LOD)   was   determined   to   be   2%   of   the   bead   antibody-­‐binding  capacity   by   setting   a   threshold   three   standard   deviations   larger   than   the   mean  equilibrium   bead   fluorescence   from   replicate   measurements   of   control   beads  without  D1.3  mAb  (i.e.  0%  immobilization).  From   both   the   manufacturer’s   specifications   and   steric   considerations,   a  single   5.5μm   diameter   bead   can   bind   at   most   4×106   antibody   molecules   (~6.6  amol);   thus,   the   measured   LOD   of   2%   of   the   bead   surface   corresponds   to  approximately  ~8×104  antibodies  or  ~132  zeptomoles   (Figure  2.10C).   In  contrast,  SPR   spectroscopy   requires   at   least   200   pg   (~109   molecules)   of   immobilized  antibody   in   order   to   generate   a   detectable   refractive   index   change.111   D1.3/HEL  binding  kinetics  were  measured  using  as  little  as  2  million  D1.3  mAb  molecules  (~3  attomoles)  loaded  into  the  microfluidic  device.         56     Figure  2.10           Sensitivity   and   detection   limit   of   antibody-­‐antigen   binding   kinetics   measurements.   (A)   Measured   association   kinetics   of   D1.3   mAb-­‐Dylight488   conjugate   on   rabbit  anti-­‐mouse  pAb  coated  beads.   (Inset)  Schematic  of  bead  assay   for  measuring  binding   kinetics  of   fluorescently   labeled  mouse  mAb  and  rabbit  anti-­‐mouse  pAb  coated  beads.    Solid   lines   represent   experimental   fits   using   mass-­‐action   equations   (equations   2.7a-­‐c).   (B)   Association   kinetics   of   HEL-­‐Dylight488   conjugate   on   beads   with   varying   amounts   of   immobilized  D1.3  mAb  (shown   in  %  bead  coverage).  Bead   fluorescence  data   is  plotted  after   subtraction   of   bead   autofluorescence   at   time   zero.   No   change   in   bead   fluorescence   was   observed  when  beads  were  not  covered  with  D1.3  mAb  (0%  bead  coverage).  Reported  error   represents  the  calculated  standard  deviation  of  multiple  replicate  measurements.  [continued   on  next  page]       57   Figure   2.10         [continued   from   previous   page]   (C)   Equilibrium   bead   fluorescence   of   HEL-­‐ Dylight488   conjugate   varies   linearly   with   the   amount   of   immobilized   D1.3   mAb.   (Inset)   Expanded   view   of   graph   to   highlight   detection   limit   of   bead   assay.   Solid   lines   represent   experimental  fits  using  mass-­‐action  equations  (A  and  B,  equations  2.7a-­‐c)  and  a  line-­‐of-­‐best-­‐fit   (C).  Adapted  with  permission  from  Singhal  et  al.112  (American  Chemical  Society,  2010).112        In  theory,  the  minimum  sample  consumption  of  the  microfluidic  bead  assay  could  be  reduced   even   further   by   reducing   losses   associated   with   channel   dead   volumes,  optimizing   the   capture   efficiency   of   antibodies   on   beads,   as   well   as   using  microfluidic  pumps  to  achieve  flow  rates  less  than  1  µL/hr.86  In  its  present  form,  the  microfluidic   bead   assay   reported   here   enables   antigen-­‐antibody   binding   kinetics  measurements  with   a   four-­‐order  of  magnitude   reduction   in  both  LOD  and   sample  consumption   when   compared   with   SPR   spectroscopy   and   a   previously   reported  microfluidic   fluorescence   assay   for   measuring   protein-­‐protein   binding  kinetics.111,122       0 5 10 15 20 25 0% 20% 40% 60% 80% 100% E qu ilb iru m B ea d Fl uo re sc en ce (X 1 03 ) Amount of bead-immobilized D1.3 (% capacity) 6 7 8 9 0.0% 1.0% 2.0% 3.0% C 58 2.4.3 Measurement   of   Binding   Kinetics   of   Antigen   and   Antibody   Secreted   from  Single  Cells.      The   low   detection   limit   of   the   fluorescence   bead   assay   enabled  measurements  of  antigen  binding  kinetics   from  antibodies  secreted  by  single  cells.    To   this   end,   rabbit   anti-­‐mouse   pAb   coated   Protein   A   beads   and   single   D1.3  hybridoma  cells  were  loaded  adjacent  to  one  another  in  the  microfluidic  device  and  were   co-­‐incubated   for   1   hour   at   room   temperature   (Figure   2.11).   Subsequently,  antibody-­‐antigen  binding  kinetics  were  measured  by  recording  the  fluorescence  of  a  single   bead   washed   with   buffer   and   successively   higher   concentrations   of  fluorescent   antigen,   in   a   manner   analogous   to   the   single-­‐cycle   kinetics   technique  used  with   SPR   spectroscopy.121,123   Association   and   dissociation   rate   constants   for  the  D1.3/HEL  interaction  were  successfully  measured  using  antibodies  secreted  by  a  single  D1.3  hybridoma  cell,  and  the  single-­‐cell  measurements  were  consistent  with  measurements   on   purified   antibodies   (Figure   2.11   and   Table   2.2).   By   loading   the  beads   into   the   microfluidic   device   after   the   cell,   free   antibodies   from   the   cell  medium  were  washed  out  of  the  device,  thereby  ensuring  that  only  mAbs  secreted  by   a   single   cell   bound   to   the   adjacent   beads.   No   antibody-­‐antigen   binding   was  detected  in  control  experiments  in  which  beads  were  loaded  into  the  fluidic  output  channel   previously   loaded   with   cell-­‐free   medium   containing   anti-­‐HEL   mAbs,  confirming  minimal  mixing  of  reagents  during  sequential  loading  into  the  device.  Antibody-­‐secreting   cells   are   known   to   secrete   thousands   of   antibodies   per  second   at   37°C,   and  would   therefore   secrete   enough   antibodies   in   approximately  one  hour   to   saturate   the   surface  of  a   single  5.5  µm  diameter  bead  with  maximum   59 binding   capacity   of   ~4×106   antibody   molecules.21,124   However,   hybridoma   cells  likely  secrete  antibodies  at  a  reduced  rate  when  incubated  in  the  microfluidic  device  at   room   temperature   (i.e.   non-­‐physiological   conditions).   Although   direct  measurements  of  actual  antibody  secretion  rates  were  confounded  by  time-­‐varying  concentrations  of  secreted  antibodies  and  an  unknown  bead  capture  efficiency,  the  cell/bead   incubation   time  was   varied   in   order   to   estimate   the  minimum  antibody  secretion  rate.    Based  on  the  minimum  incubation  time  (~5  min)  and  detection  limit  of   the   assay   (~8×104   antibodies),   single   hybridoma   cells   secreted   at   least   200  antibodies/second  when  incubated  at  room  temperature  in  the  microfluidic  device.         60   Figure  2.11           Antibody-­‐antigen  binding  kinetics  measured  using  antibodies  secreted  from  a   single   cell.     (A)   Microscope   image   of   D1.3   hybridoma   cell   loaded   into   microfluidic   device   adjacent   to   rabbit   anti-­‐mouse   pAb   coated   beads   trapped   using   a   sieve   valve.     (B)     “Single-­‐ cycle”  binding  kinetics   from  a   single  bead   containing  D1.3  mAbs   secreted   from  a   single   cell   and   subject   to   increasing   concentrations  of  HEL-­‐Dylight488   conjugate.   Solid   lines   represent   three  experimental   fits  using  mass-­‐action  equations  corresponding   to  each  concentration  of   fluorescently   labeled   HEL.   Reported   error   represents   the   calculated   standard   deviation   of   multiple  replicate  measurements.  Adapted  with  permission   from  Singhal  et  al.112   (American   Chemical  Society,  2010).112         Hybridoma cell Antibody capture beads A 10µm 0 0.2 0 .4 0 .6 0 .8 1 1.2 0 20 40 60 80 100 120 N or m al iz ed F lu or es ce nc e T im e (m in ) 107ng/m L H EL488 214ng/m L H EL488 428ng/m L H EL488 kon (M-1s-1) koff (s-1) Kd (1.2 ± 0.1) X 10 6 (1.9 ± 0.2) X 10 -3 1.6 ± 0.2 (nM) B 2.4.4 Extensions  of  the  Microfluidic  Fluorescence  Bead  Assay   2.4.4.1 Direct   Measurements   of   Antibody-­‐Antigen   Equilibrium   Binding   Affinities  The  microfluidic   fluorescence  bead  assay  was  used   to  directly  measure   the  equilibrium   binding   affinities   of   antibody-­‐antigen   interactions.     In   particular,   the  equilibrium   bead   fluorescence   was   measured   at   different   concentrations   of  fluorescently   labeled  HEL  using  beads   loaded  with  a   fixed  amount  of   immobilized  D1.3  mAb  (Figure  2.12).  By  fitting  the  equilibrium  bead  fluorescence  measurements  with  a  Langmuir  isotherm  (equation  2.7c),  the  D1.3  mAb/HEL  interaction  was  found  to   have   an   equilibrium   dissociation   constant   equal   to   1.67   nM,   consistent   with  kinetic  measurements  from  the  fluorescence  bead  assay  (Table  2.2).     Figure  2.12           Direct   measurement   of   equilibrium   dissociation   constants   by   measuring   equilibrium   bead   fluorescence   using   immobilized   D1.3  mAb   and   varying   concentrations   of   HEL-­‐Dylight488.  Solid  line  represents  experimental  fits  using  a  Langmuir  isotherm  equation.   Value  of  Kd   estimated  by   the   concentration  at  which   the  equilibrium  bead   fluorescence  was   equal   to   the   half-­‐maximal   value.   Adapted  with   permission   from   Singhal   et   al.112   (American   Chemical  Society,  2010).112           0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 20 40 60 80 100 120 140 160 Eq ui lib riu m B ea d Fl uo re sc en ce ( X 10 4 ) [H E L-D ylig ht488] (nM ) Kd = 1 .7 nM 62     Alternatively,   the   equilibrium   binding   affinity   of   an   antigen-­‐antibody  complex  can  be  measured  relative  to  a  known  affinity  for  the  same  antibody  bound  to   a   second   capture   reagent   (such   as   anti-­‐immunoglobulin,   denoted   antiAb).    Assuming  the  two  binding  reactions  are  independent,  both  reactions  take  the  form  of  the  classic  single-­‐sorbate  Langmuir  isotherm:  [𝐴𝑏𝐴𝑔] = [𝐴𝑔]! [!"]![!"]!!!!,!"!#         (2.8a),  [𝐴𝑏𝐴𝑛𝑡𝑖𝐴𝑏] = [𝐴𝑛𝑡𝑖𝐴𝑏]! [!"]![!"]!!!!,!"!#$%!"     (2.8b),  Taking  the  ratio  of  these  two  equations:  [!"!#$%!"][!"!#] = [!"!#$!"#]![!"!#]! [!"]!!!!,!"!#[!"]!!!!,!"!#$%!"   (2.9a),  When   the   free   antibody   concentration   is   less   than   the   equilibrium   dissociation  constants  of  both  reactions  ([Ab]0  <500-­‐fold)  as  well  as  many  primers  with  low  degeneracy  (<10-­‐fold).58,134         Figure  3.2           Design  of  primers  for  PCR  amplification  of  antibody  heavy  (IgH)  and  light  (IgL)   chain  genes.    3’  primers  are  designed  to  the  antibody  constant  (C)  region.  Degenerate  primers   to  the  5’  region  can  be  designed  either  to  the  leader  (L)  or  1st  framework  (FWR1)  region  of  VH   and  VL  genes.       IgH IgL (Igκ or Igλ) FWR1L C FWR1L C L1 L2 VH CHJD5’ 3’ L1 L2 VL CLJ5’ 3’ 73     Figure  3.3           Degenerate  primers  are  mixtures  of  oligonucleotides  with  similar  sequences   designed  to  amplify  genes  with  highly  related  sequences.    The  level  of  degeneracy  depends  on   the  number  of  base  positions  and  the  variation  at  each  position.    In  this  example,  the  5’  primer   has  4-­‐fold  degeneracy  (2  positions  X  2  bases  at  each  variable  position)  whereas  the  3’  primer   has  6-­‐fold  degeneracy  (3  bases  at  1st  variable  position  X  2  bases  at  2nd  variable  position).     One  potential  drawback  of  using  degenerate  primers  for  RT-­‐PCR  of  antibody  genes   is   the   amplification   of   non-­‐specific   products   as   a   result   of   the   reduced  specificity  of  degenerate  primers.  One  approach   for  compensating   for   the   reduced  specificity   of   degenerate   primers   is   to   perform   nested   PCR,   in   which   multiple  rounds  of   PCR   are  performed  using   a   unique   set   of   primers   internal   to   the   target  gene   at   each   successive   round   (Figure   1.4).   In   order   to   optimize   RT-­‐PCR  amplification  of  mouse  antibody  heavy  and   light   chain  genes   from   low  abundance  template   (e.g.   single   cells),   a   variety   of   conditions  were   tested,   including  different  primer   concentrations,   annealing   temperatures,   primers   with   varying   levels   of  degeneracy,  as  well  as  nested  PCR  primers.       74   Figure  3.4           Nested  PCR.  Multiple   rounds  of  PCR  are  performed,   in  which  a  unique   set  of   primers  internal  to  the  template  DNA  are  used  in  each  successive  PCR  round.  In  semi-­‐nested   PCR,  one  primer  is  re-­‐used  and  one  internal  primer  is  designed  for  each  successive  round  of   PCR.  Nested  PCR  is  used  to  increase  amplification  specificity  for  the  target  gene.     3.3 Materials  and  Methods   3.3.1 Cell  Culture  Mouse   hybridoma   cells   (D1.3,   HyHEL-­‐5,   CD1d)   were   grown   in   6   mL   petri  dishes  (Nunc)  using  RPMI  1640  medium  (Gibco)  with  10%  fetal  calf  serum  (FCS)  in  a  cell  culture  incubator  (37°C,  5%  CO2).  Cells  were  passaged  approximately  once  a  week   by   serial   dilutions   (5-­‐fold)   in   fresh   medium.   Prior   to   RT-­‐PCR   experiments,  cells  were  washed  by  centrifugation  at  1500  rpm  and  re-­‐suspended   in  1X  PBS,  pH  7.4  (Gibco)  in  order  to  remove  cell  medium.  Cell  concentration  was  quantified  using  a  haemocytometer  and  brightfield  microscope.   5’ 3’ 5’ 3’ 1st round 2nd round 75 3.3.2  Cell  Lysis  and  mRNA  Purification  105–106   antibody-­‐secreting   cells   (e.g.   mouse   hybridoma   cells)   were  concentrated   into  a  pellet  by  centrifugation  at  1500  rpm  and  the  supernatant  was  decanted.  Cells  were  re-­‐suspended  in  alkaline  or  non-­‐ionic  lysis  buffer.  Alkaline  lysis  buffer  consisted  of  100  mM  Tris-­‐HCl,  pH  7.5,  500  mM  LiCl,  10  mM  EDTA,  pH  8,  1%  Lithium   dodecyl   sulfate   (LiDS),   and   5  mM   dithiothreitol   (DTT)   (Invitrogen).   Non-­‐ionic   lysis   solution   consisted   of   0.5%   NP-­‐40   or   Triton   X-­‐100   in   1X   PBS,   pH   7.4  (Gibco).   10X   serial   dilutions   of   cell   lysate  were  prepared  by  diluting  10  μL   of   cell  lysate  into  90  μL  of  lysis  solution.  mRNA   from   cell   lysate   solutions   was   purified   using   oligo(dT)25   beads  (Dynabeads   mRNA   Direct   kit,   Invitrogen).   For   each   reaction,   50   μL   of   2.8μm  diameter   oligo(dT)25   Dynabeads   (107   beads/mL)   was   transferred   into   a   1   mL  DNAse-­‐   and   RNase-­‐free  microcentrifuge   tube   (Axygen)   and   placed   on   a  magnetic  tube   rack   (LifeSep1.5S,   Bangs   Labs)   for   1   minute   (Figure   3.4).   The  superparamagnetic  Dynabeads  formed  a  tight  pellet  that  adhered  to  the  sidewall  of  the   microcentrifuge   tube   adjacent   to   the   magnet,   allowing   the   supernatant   to   be  removed  by  pipet.  Beads  were  re-­‐suspended  in  100  μL  of  alkaline  lysis  solution.  The  magnetic  wash   step  was   repeated   and   the  beads  were   re-­‐suspended   in   cell   lysate  solution.   Cell   lysate   was   incubated   with   oligo(dT)25   Dynabeads   for   5   minutes.  Following  incubation,  beads  were  concentrated  by  magnet  and  washed  in  100  μL  of  solution  consisting  of  10  mM  Tris-­‐HCl,  pH  7.5,  0.15  M  LiCl,  1  mM  EDTA,  and  0.1%  LiDS   (Wash   buffer   A).   This   wash   step   was   repeated   to   remove   all   LiDS   using   a  solution  of  10  mM  Tris-­‐HCl,  pH  7.5,  0.15  M  LiCl,  1  mM  EDTA  (Wash  buffer  B).     76   Figure  3.5           RT-­‐PCR  experiment  for  amplifying  genes  from  antibody-­‐secreting  cells.    Cells  are  enumerated  using  a  haemocytometer.  The   protocol  is  repeated  with  serial  dilutions  of  cell  lysate  in  order  to  determine  the  detection  limit  of  RT-­‐PCR  reactions  for  mouse  β-­‐actin  and   antibody  heavy  and  light  chain  genes.       Antibody-secreting cells Chemical Lysis Mix serial dilutions of lysate with beads Magnetic oligo(dT) bead solution Cell lysate Remove supernatant Magnetic oligo(dT) beads mRNA bead capture bead-bound mRNA Remove lysate Puri!ed mRNA on oligo(dT) beads Add RT-PCR mix (primers, dNTPs, enzyme) One-step RT-PCR C C 77 As  LiDS  is  a  strong  protein  denaturant,  it  must  be  removed  prior  to  subsequent  RT-­‐PCR  reactions.   Finally,   the   bead   solution   was   placed   on   the   magnetic   tube   rack,   the  supernatant  was  decanted,  and  beads  were  re-­‐suspended  in  RT-­‐PCR  reaction  mix.       3.3.3 RT-­‐PCR  Reaction  Mix  and  Cycling  Conditions  RT-­‐PCR  primers  were  ordered  in  dried  lyophilized  form  from  IDT  Technologies.    A  100μM  stock  solution  of  each   individual  primer  was  prepared   in  RT-­‐PCR-­‐grade  (i.e.  DNase-­‐free   and   RNase-­‐free)   water   (Qiagen).   Two   distinct   primer   sets   for   mouse  antibody   genes  were   tested:   a   highly   degenerate   primer   set   (Appendix   A.1)134   and   a  low-­‐degeneracy  nested  PCR  primer  set  (Appendix  A.2).  Working  solutions  of  5’  and  3’  primers  were  prepared  by  mixing  individual  primers  in  equal  amounts  such  that  each  primer  was  at  8μM  concentration.  As  a  control,  commercial  PCR  primers  were  ordered  for   amplifying   the  mouse   β-­‐actin   gene   (514bp),   similar   in   size   to   the   antibody   heavy  and  light  chain  genes  (Stratagene).  The  sense  and  anti-­‐sense  primer  sequences  directed  against   the   mouse   β-­‐actin   gene   were   TGTGATGGTGGGAATGGGTCAG   and  TTTGATGTCACGCACGATTTCC,  respectively.    RT-­‐PCR   reactions   were   performed   in   0.2mL   PCR   reaction   tubes   on   a   96-­‐well  PTC-­‐100  benchtop  thermal  cycler  (MJ  research).    Each  RT-­‐PCR  reaction  was  performed  in  a  50  μL  reaction  volume  prepared  using  the  Qiagen  One-­‐step  RT-­‐PCR  kit.    For  single-­‐plex  PCR  reactions  of  heavy  or  light  (kappa)  chain  genes,  the  reaction  mix  consisted  of  21   μL  RT-­‐PCR   grade  water,   10   μL   5X   reaction   buffer,   7.5   μL   5’   primer  mix,   7.5   μL   3’  primer  mix,  2  μL  enzyme  mix  containing  both  RT  and  DNA  polymerase  enzyme,  and  2  μL  dNTP.  For  multiplex  RT-­‐PCR  reactions,  the  50  μL  reaction  volume  was  maintained  by   78 using  only  6μL  of  water  with  7.5  μL  of   both   forward   and  both   reverse  primer  mixes.    The  final  concentration  of  each  primer  in  the  PCR  reaction  was  600  nM.  One-­‐step   RT-­‐PCR   amplification   was   performed   in   order   to   minimize   sample  losses   and   potential   contamination   during   transfer   steps   from   RT   to   PCR   reactions  (Table  3.2).  As  the  RT  enzyme  may  be  active  at  room  temperature,  all  RT-­‐PCR  samples  were  prepared  on  ice  and  transferred  directly  to  the  thermal  cycler  at  50°C  for  30  min  for   reverse   transcription.   The   subsequent   15   min   hold   at   95°C   simultaneously  denatures   the   RT   enzyme   while   activating   the   hot-­‐start   DNA   polymerase.   Each   PCR  cycle   consisted   of   three   temperature   steps:   a   denaturation   step   at   94°C,   a   primer  annealing   step   (45-­‐65°C),   and   a   DNA   extension   step   at   72°C.     A   variety   of   primer  annealing   temperatures   were   tested,   ranging   from   45   to   65°C.     For   touchdown   PCR  experiments,   the  annealing   temperature  during   the   first  PCR  cycle  was  set  5°C  higher  than   the   primer   melting   temperature   (e.g.   55°C   for   a   Tm   =   50°C)   and   progressively  reduced  by  1°C  each  cycle  for  the  first  10  cycles.        Table  3.2     One-­‐step  RT-­‐PCR  cycling  protocol.   Step   Temperature  (°C)   Hold  Time   #  of  cycles  Reverse  transcription  (RT)   50   30  min   1  Hot-­‐start  Activation   95   15  min   1  Denaturation   94   30  sec   40  Anneal   55     30  sec   40  Extension   72   1  min   40  Final  extension   72   10  min   1   79 The  annealing  temperature  was  then  held  constant  for  the  remaining  40  PCR  cycles.  A  final  extension  step  at  72°C  for  10  min  was  used  to  complete  the  PCR  reaction  for  any  partially  amplified  products.  For  nested  PCR  reactions,  3.5  μL  reaction  product  from  the  1st   round   of   PCR   was   mixed   with   50   μL   reaction   mix   containing   the   same   reaction  mixture  as  the  1st  round  but  with  the  nested  primers.    The  cycling  protocol  for  2nd  round  of  nested  PCR   reactions  was   identical   to   the  1st   round,   except   that   the   initial  RT   step  was  omitted.     3.3.4 Analysis,  Purification,  and  Sequencing  of  RT-­‐PCR  Products    RT-­‐PCR   amplification  products  were   analyzed  by  DNA   gel   electrophoresis.   1%  agarose  gels  were  prepared  by  dissolving  0.5  g  of  agarose  (Sigma)  in  50  mL  of  1X  TAE  buffer  containing  0.5  μL  of  10,000X  SYBRSafe  concentrated  solution  (Invitrogen).  The  size   of   the   PCR   products   was   compared   against   a   1   kb   Plus   ladder   (Invitrogen)   and  fluorescence   images   were   taken   with   the   Alpha   Imaging   gel   imaging   system   (Alpha  Innotech).    Melting  curve  analysis  was  also  performed  on  RT-­‐PCR  products  by  adding  SYBRSafe   DNA   dye   at   10X   concentration   and   measuring   the   sample   fluorescence   at  every   0.5°C   over   a   range   of   approximately   60-­‐100°C   using   the   Chromo4   Opticon  thermal  cycler  (Bio-­‐Rad).      For   sequencing,   DNA   samples   were   extracted   from   the   gels   by   a   scalpel   and  purified  using  spin  columns  as  per  the  Qiagen  MinElute  kit.    Samples  were  eluted  from  the   spin   columns   in   low   TE   buffer   and   stored   in   a   -­‐20°C   freezer.   DNA   samples   and  associated   primers   were   submitted   to   the   Nucleic   Acid   Protein   Service   Unit   (NAPS,  http://www.msl.ubc.ca/services/naps)  for  standard  Sanger  sequencing.   80 3.3.5 Microfluidic  Single-­‐Cell  Sorting,  Lysis,  and  Recovery  Prototype  microfluidic  devices  were  designed  and   fabricated  with  8   chambers,  approximately  ~1  nL  in  volume,  each  with  an  independent  valve  to  control  reagent  flow  through  the  chamber  (Figure  3.6).  In  order  to  trap  large  particles  (e.g.  cells,  beads),  all  chambers   contains   a   partially   closing   sieve   valve   that   are   controlled   in   parallel.   The  device   consists   of   9   inlets   controlled   by   a  multiplexing   valve   structure   to   enable   the  injection  of  different  solutions  through  the  device  and  into  a  single  output  port.  In   a   typical   experiment,   a   solution   containing  mouse  hybridoma   cells   (105-­‐106  cells/mL)  was  injected  into  the  device  and  into  the  channel  upstream  of  the  microfluidic  chambers  with  the  sieve  valves  closed.  Single  cells  were  sorted  into  chambers  by  briefly  opening   each   chamber   valve   when   a   cell   approached   the   chamber   inlet   as   observed  under  a  brightfield  microscope.  Cells  were   typically   trapped  at   the   chamber   interface  where   the   height   of   the   chamber   reduced   from  ~12-­‐14   μm   to   ~3   μm,   several   times  smaller   than   the   ~10-­‐15   μm   diameter   hybridoma   cells   (Figure   3.6C).   The   remaining  cells   in   the   channel   were   then   flushed   into   the   outlet   using   1X   PBS   solution.  Subsequently,   a   solution   of   2.8   μm   diameter   oligo(dT)   beads   was   injected   into   the  device   and   several   hundred   beads   were   stacked   against   the   sieve   valve   in   each  chamber.  An  alkaline   lysis   solution  (1%  LiDS)  was   injected   into  each  chamber   to   lyse  cells   and   capture   poly(A)   tailed-­‐mRNA   on   the   downstream   stack   of   oligo(dT)   beads.  After   flushing   lysis   solution   out   of   the   device   using   1X   PBS   solution,   the   sieve   valves  were  opened  and  beads  from  each  chamber  were  sequentially  eluted  to  the  output  port.    Samples   were   recovered   either   by:   1)   removal   of   the   stainless   steel   pin   and   Tygon  tubing  connected  to  the  output  port;  or,  2)  manual  pipetting  of  solution  out  of  the  outlet   81 port  using  a  gel-­‐loading  pipet  tip.  Recovered  samples  were  pipetted  directly  into  50μL  of  RT-­‐PCR  reaction  mix  and  transferred  to  a  benchtop  thermal  cycler  for  amplification  according  to  one  of  the  protocols  described  above    (see  section  3.4.3).     Figure  3.6           Microfluidic  device  for  sorting,  lysis,  and  mRNA  bead  capture  from  single  cells.  (A)   Schematic   of   microfluidic   device   containing   9   reagent   inlets   (left),   8   chambers   (one   cell   per   chamber)  and  one  fluidic  outlet  (right).  (B)  (expanded  view  of  boxed  region  in  A)  Each  chamber   contains  a  partially  closing  sieve  valve  used  to  trap  cells  and  beads.    Cells  are  lysed  in  the  chamber   to  release  cellular  mRNA  that  is  captured  on  oligo(dT)  beads.  Beads  are  sequentially  eluted  from   each   chamber   and   recovered   from   the   output   port   for   single-­‐cell   RT-­‐PCR   amplification.   (C)   Brightfield   microscope   image   of   single   chamber   containing   a   stack   of   oligo(dT)   beads,   an   antibody-­‐secreting   cell,   and   antibody-­‐capture   beads.   Microscope   image   is   rotated   90°   counter-­‐ clockwise  from  the  schematic  drawings  in  (A)  and  (B).   A B C Column valve Column valve Sieve valve 82 3.4 Results   3.4.1 RT-­‐PCR   Optimization   for   Single-­‐Cell   Amplification   of   Mouse   Heavy   and   Light  Chain  Antibody  Genes    The   sensitivity   of   a   highly   degenerate   primer   set   (Appendix   A.1)   for   RT-­‐PCR  amplification  of  antibody  heavy  and   light  chain  genes  was   tested  using  serial   ten-­‐fold  dilutions  of   a  mouse  hybridoma  cell   lysate   (Figure  3.7).    Heavy  and   light   chain   genes  were  successfully  amplified   from   initial   template   concentrations   ranging   from   tens   to  thousands  of  cell  equivalents;  however,  no  PCR  products  were  detected  from  template  concentrations   less   than   or   equal   to   a   single   cell   (Figure   3.7).   Successful   RT-­‐PCR  amplification  of  mouse  β-­‐actin  genes  from  0.1  to  1  cell  equivalents  of  mRNA  suggested  that   loss   of   template   during   purification   was   not   responsible   for   unsuccessful  amplification  of  antibody  gene  products  at  low  template  concentrations.    The  poor  amplification  of  antibody  genes  at  low  template  concentrations  may  be  the  result  of  poor  primer  binding  to  template  and/or  amplification  of  competing,  non-­‐specific   products.  DNA  melting   curve   analysis  was  performed   in  order   to  detect  non-­‐specific   products   in   RT-­‐PCR   samples.136   Based   on   the   known   gene   sequence   for   the  mouse   β-­‐actin   as  well   as   the   antibody   heavy   and   light   chain   genes   from   the   selected  mouse  hybridoma  cells  (e.g.  D1.3  and  HyHEL-­‐5),   the  theoretical  melting  temperatures  of  all  amplicons  were  estimated  to  be  within  the  range  of  82-­‐85°C.137,138  Compared  to  the   β-­‐actin   RT-­‐PCR   reactions,   amplification   of   antibody   heavy   and   light   chain   genes  generated   significantly   greater   levels   of   non-­‐specific   products   with   melting  temperatures   less   than  80°C   (Figure  3.7B  and  C).  At   template   concentrations  of  0.1-­‐1   83 cell  eq.,  these  non-­‐specific  products  were  present  even  though  antibody  genes  were  not  successfully  amplified.     Figure  3.7           RT-­‐PCR  of  mouse  β-­‐actin  and  antibody  heavy  and  light  chain  genes  using  purified   mRNA   from   different   concentrations   of   D1.3   hybridoma   cell   lysate.     (A)   RT-­‐PCR   products   visualized  on  a  1%  DNA  agarose  gel  with  a  100bp  ladder.    The  β-­‐actin  gene  product  appears  as  a   single   band   with   ~500   bp   in   size.     Multiplex   PCR   of   both   heavy   and   light   chain   reactions   also   appear   as   a   single   ~400   bp   band.   Both   heavy   and   light   chain   gene   products  were   amplified   as   confirmed  by  excising,  purifying,  and  sequencing   the  DNA  products.  DNA  melting  curve  analysis   for  both  mouse  β-­‐actin  (B)  and  multiplexed  heavy  and  light  chain  RT-­‐PCR  reactions  (C).  Plotted  is   the  change   in   fluorescence   intensity   (dI/dT)  at  each   temperature.  The   large   fluorescence  signal   change  at  ~52°C  coincides  with  the  primer  melting  temperature.     Beta-actin Template conc (cell equivalents) Template conc (cell equivalents) Antibody H+L NT C 14 00 14 0 14 1. 4 0. 14 NT C 14 00 14 0 14 1. 4 0. 14 Beta-actin Antibody H+L A B C 10X dil. 1400 NTC 10X dil. 1400 NTC 84 To   increase   specificity   of   antibody   heavy   and   light   chain   RT-­‐PCR   reactions,   a  range   of   primer   concentrations  was   tested.  Whereas   low   primer   concentrations  may  produce   insufficient   template   amplification,   high   primer   concentrations   can   result   in  mis-­‐priming   and   non-­‐specific   amplification.139   For   typical   RT-­‐PCR   reactions,   gene-­‐specific   primer   concentrations   range   from   50   to   200   nM.139   However,   the   optimal  primer   concentration   for   RT-­‐PCR   reactions   using   degenerate   primers   is   reaction-­‐specific   because,   in   each   reaction,  many   primers  will   not   bind   to   the   template,  while  primers  that  bind  may  have  variable  affinities  due  to  possible  base  mismatches.  RT-­‐PCR  amplification  of  antibody  heavy  and  light  chain  genes  was  successful  over  a  broad  range  of  primer  concentrations  ranging  from  160  nM  to  1.2  μM  (Figure  3.8).  RT-­‐PCR  reactions  with   lower   primer   concentrations   (i.e.   160   nM)   resulted   in   both   a   reduction   in   the  amount   of   both   specific   amplicons   and  primer   dimers   compared   to   reactions   using   a  higher  primer  concentration  (600  nM).  Thus,  reduction  in  primer  concentration  did  not  appear  to  be  a  robust  strategy  for  improving  RT-­‐PCR  sensitivity  of  antibody  genes  using  this  highly  degenerate  primer  set.       Figure  3.8           Multiplex   RT-­‐PCR   of  mouse   heavy   and   light   chain   genes   of  mRNA   purified   from   ~106  D1.3  hybridoma  cells  using  highly  degenerate  primers  at  two  different  concentrations  (160   nM  and  600  nM).  Lower  primer  concentrations  resulted  in  reduced  amplification  of  both  specific   amplicons  and  non-­‐specific  primer  dimers.  Shown  is  a  1%  DNA  agarose  gel  with  100  bp  ladder.     # of beads # of beads} } NTC 107 106 105 104 103 NTC 107 106 105 104 103 Loss of some sample during gel loading Primers @ 600nM Primers @ 160nM 85 Annealing   temperatures   also   affect   reaction   specificity,   as   increases   in  temperature  reduce  the  likelihood  of  base  mismatches  during  primer-­‐template  binding;  however,  increases  in  annealing  temperature  may  also  come  at  the  expense  of  reduced  amplification  efficiency.  A  range  of  annealing  temperatures  from  45°C  to  65°C  flanking  the   primer   melting   temperature   (Tm   ~   50°C)   was   tested   using   “touchdown”   PCR.  Touchdown   PCR   can   improve   amplification   specificity   by   using   high   annealing  temperatures   during   early   PCR   cycles   to   increase   primer   binding   stringency,   thus  selecting   for   target   template   amplification   over   the   formation   of   non-­‐specific  products.140,141  Once  template  concentration  has   increased  after  the   initial  PCR  cycles,  target  template  can  outcompete  non-­‐specific  template  for  primer  binding,  thus  enabling  gradual   reduction   of   annealing   temperature   in   later   PCR   cycles   in   order   to   improve  amplification  efficiency.  The  annealing   temperature   for   the   first  PCR  cycle  was  varied  between   50°C   to   65°C,   followed   by   progressively   lower   temperatures   (1°C   lower   per  cycle)   for   the  next   10  PCR   cycles,   and   then   a   constant   annealing   temperature   for   the  final   40   PCR   cycles.   However,   changes   to   annealing   temperatures   did   not   improve  amplification  efficiency  at   low  target  template  concentrations,  nor  reduce  non-­‐specific  amplification  (Figure  3.9).  The  lack  of  improvement  in  RT-­‐PCR  sensitivity  when  altering  both   primer   concentrations   and   annealing   temperatures   suggested   that   these   highly  degenerate  primers  may  not  be  suitable  for  amplification  of  antibody  genes  from  single  cells.   86   Figure  3.9           Multiplex  RT-­‐PCR  of  mouse  antibody  genes  at  different  annealing  temperatures.    4   different   touchdown   PCR   protocols   were   tested  with   annealing   temperatures   varying   from   (A)   65°C-­‐55°C   and   60°C-­‐50°C   to   (B)   55°C-­‐45°C   and   50°C-­‐40°C.   Amplification   was   successful   using   template   concentrations   greater   than   ~100   cell   equivalents,   with   significant   non-­‐specific   amplification  observed  in  all  reactions.  Shown  is  a  1%  DNA  agarose  gel  with  100bp  ladder.   } }65-55C (1C per cycle)55C for 40 cycles 60-50C (1C per cycle)50C for 40 cycles NT C 38 0 38 3. 8 0. 38 Template conc (cell equivalents) NT C 38 0 38 3. 8 0. 38 Template conc (cell equivalents) } }55-45C (1C per cycle)45C for 31 cycles 50-40C (1C per cycle)40C for 40 cycles NT C 38 0 38 3. 8 0. 38 Template conc (cell equivalents) NT C 38 0 38 3. 8 0. 38 Template conc (cell equivalents) A B 87 The  RT-­‐PCR   sensitivity   of   a   low  degeneracy  nested  primer   set58   (Appendix  A.2)   was   directly   compared   to   that   of   the   above-­‐described   highly   degenerate  primers.   RT-­‐PCR   was   performed   on   dilutions   of   purified   mRNA   from   multiple  different   mouse   hybridoma   cell-­‐lines   (D1.3,   HyHEL-­‐5,   and   CD1d)   in   order   to  estimate  amplification  performance  across  different  cells.  Notably,  mouse  antibody  genes   were   successfully   amplified   from   all   cells   at   all   tested   template  concentrations,  including  concentrations  as  low  as  0.1  cell  equivalents,  (Figure  3.10)  using   only   the   1st   round   primers   from   the   low   degeneracy   set.   These   results  demonstrate   that   primer  design   is   critical   for   successful   amplification  of   antibody  genes   from   single   cells,   and   that   primers   with   reduced   degeneracy   may   be   less  prone  to  nonspecific  amplification  than  high  degeneracy  primers.  Despite   successful   RT-­‐PCR   amplification   of   all   three   hybridoma   cell-­‐lines  using  the  1st  round  low  degeneracy  primers,  only  a  single  PCR  product  was  visible  on  the  DNA  gel  from  multiplex  PCR  reactions  of  the  CD1d  hybridoma  (Figure  3.10).  Extraction   of   this   product   from   the   gel   and   DNA   sequencing   confirmed   that   this  product  corresponded  to  an  antibody  light  (kappa)  chain,  indicating  a  failure  of  the  heavy  chain  RT-­‐PCR  reaction  (Figure  3.10C).  In  an  attempt  to  improve  heavy-­‐chain  amplification   efficiencies,   single-­‐plex   RT-­‐PCR   reactions   were   carried   out   by  purifying  mRNA   on   oligo(dT)   beads   from   both  mouse   D1.3   and   CD1d   hybridoma  cells,  and  splitting  the  beads  into  two  equal  parts  for  amplification  using  heavy  and  light  chain  primers,  respectively  (Figure  3.11).       88   Figure  3.10           Multiplex   RT-­‐PCR   of   mouse   antibody   genes   on   serial   dilutions   of   oligo(dT)   bead-­‐purified  RNA  from  HyHEL-­‐5  (A),  D1.3  (B),  and  CD1d  (C)  mouse  hybridoma  cells  using  a   highly  degenerate  primer   set134   (left)   and  1st   round  primers   from  a   low  degeneracy  primer   set58   (right)   at     600nM   concentration.   Single-­‐cell   RT-­‐PCR   sensitivity   using   low   degeneracy   primers  obtained  for  all   three  hybridoma  cells.    Shown  is  a  1%  DNA  agarose  gel  with  100bp   ladder.     NT C 15 00 15 0 15 1. 5 0. 15 Template conc (cell equivalents) NT C 15 00 15 0 15 1. 5 0. 15 Template conc (cell equivalents) Highly degenerate primers Nested 1st round Primers NT C 73 50 73 5 73 7 0. 7 Template conc (cell equivalents) Template conc (cell equivalents) NT C 73 50 73 5 73 7 0. 7 NT C 73 50 73 5 73 7 0. 7 Template conc (cell equivalents) Template conc (cell equivalents) NT C 73 50 73 5 73 7 0. 7 NT C 21 50 21 5 21 2 0. 2 Template conc (cell equivalents) NT C 21 50 21 5 21 2 0. 2 Template conc (cell equivalents) A B C HyHEL-5 D1.3 CD1d HyHEL-5 D1.3 CD1d 89 Successful  amplification  of  both  antibody  genes   from  both  cell-­‐lines   indicated   that  heavy   and   light   chain   amplification   reactions   can   compete   in   multiplex   reactions  (Figure  3.10)  and,  thus,  should  be  performed  independently.           Figure  3.11           Single-­‐plex   RT-­‐PCR   of   mouse   antibody   genes   from   mouse   hybridoma   cells   using  1st   round  primers   from  a   low  degeneracy  primer   set58.  mRNA   from  3.5×105  D1.3   cells   and  6×105  CD1d  cells  was  purified  using  oligo(dT)  beads.  The  beads  were  then  split  into  two   equal   parts   and  mixed  with   RT-­‐PCR   reaction  mix   containing   primers   at   a   concentration   of   600nM  for  heavy  and  light  chain  amplification,  respectively.    Shown  is  a  1%  DNA  agarose  gel   with  100bp  ladder.     Measurements   of   template   concentrations   based   on   cell   equivalents   are  imprecise;  that  is,  it  is  unclear  whether  mouse  hybridoma  cells  contain  comparable  amounts   of   antibody-­‐encoding   mRNA   as   primary   antibody-­‐secreting   cells   (ASCs)  harvested  from  immunized  animals.  Thus,  in  the  absence  of  an  absolute  measure  of  template   concentration,   single-­‐plex   RT-­‐PCR   reactions   were   performed   on   serial  dilutions  of  purified  mRNA  from  primary  ASCs  (i.e.  CD138+  splenocytes)  harvested  from  mice  immunized  with  the  model  antigen  hen  egg  lysozyme  (see  Chapter  4  for   Nested 1st round Primers Kappa chainHeavy chain NTC D1.3 CD1d NTC D1.3 CD1d }} 90 methods).  ASCs  were  sorted  into  RT-­‐PCR  tubes  by  fluorescent-­‐activated  cell  sorting  (FACS)   and   chemically   lysed.   Serial   ten-­‐fold  dilutions  of   cell   lysate  were  prepared  and  mRNA  was  purified  from  each  dilution  using  oligo(dT)  beads.  The  beads  were  split   into   two   equal   parts   for   amplification   with   low   degeneracy   heavy   and   light  chain   primers   (Appendix   A.2),   respectively.   Importantly,   RT-­‐PCR   amplification   of  both   heavy   and   light   chain   genes   was   successful   down   to   single-­‐cell   template  concentrations   (Figure   3.12).  Whereas   robust   amplification   of   kappa   chain   genes  occurred   at   template   concentrations   of   0.1   cell   equivalents,   heavy   chain   reactions  were   successful   at   a  minimum   template   concentration   of   1   cell   equivalent.   Heavy  chain   reactions   produced   significant   amounts   of   both   large   (~600   bp)   and   small  (~250  bp)  non-­‐specific  products,  though  the  larger  products  were  outcompeted  by  specific   amplicons   when   using   high   initial   template   concentrations   (i.e.   100   cell  equivalents,   Figure   3.12).   Although   the   amount   of   specific   amplification   of   heavy  chain   genes   was   modest   compared   to   light   chain   genes   at   single-­‐cell   template  concentrations,   the   established   RT-­‐PCR   protocols   should   be   suitable   for  amplification   of   both   heavy   and   light   chain   genes   from   single   ASCs   selected   by  microfluidic  screening.  The  following  section  describes  methods  to  sort  single  ASCs  in  microfluidic  devices  in  a  manner  compatible  with  detection  of  secreted  antibodies  (Chapter   4)   and   subsequently   lyse   and   recover   selected   cells   for   RT-­‐PCR  amplification  of  antibody  genes.     91   Figure  3.12           Single-­‐plex  RT-­‐PCR  of  mouse  antibody  genes  from  primary  antibody-­‐secreting   cells  (ASCs)  harvested  from  mice  immunized  with  hen  egg  lysozyme  (HEL).    Cells  were  sorted   by   fluorescence-­‐activated   cell   sorting   (FACS).   ASCs   were   lysed   and   the   mRNA   from   serial   dilutions  of  cell  lysate  was  purified  using  oligo(dT)  beads.  The  beads  were  then  split  into  two   equal   parts   and   mixed   with   RT-­‐PCR   reaction   mix   containing   low   degeneracy   primers58   at   600nM  for  heavy  and  light  chain  amplification,  respectively.  Amplification  in  the  heavy  chain   NTC  reaction  was  due  to  reagent  contamination,  which  was  removed  when  using  fresh  primer   solutions  and  RT-­‐PCR  reagents  (data  not  shown).    Shown  is  a  1%  DNA  agarose  gel  with  100bp   ladder.     3.4.2 Microfluidic  Single-­‐Cell  Sorting,  Lysis,  and  Recovery  Single   hybridoma   cells   were   sorted   into   ~1   nL   microfluidic   chambers  containing  oligo(dT)  beads  trapped  with  a  sieve  valve  (Figure  3.6  and  Section  3.4.5).  Cells  were   lysed  using  a  harsh  alkaline   lysis   solution   (1%  LiDS)   in  order   to  purify  cellular  mRNA     on   the   oligo(dT)   beads.  With   several   hundred  beads   per   chamber  and  an  estimated  binding  capacity  of  106  mRNA  molecules  per  bead,  the  total  bead  binding   capacity   greatly   exceeded   the   number   of   mRNA   molecules   in   single  mammalian  cells  (105-­‐106).142–144  Thus,  all  cellular  mRNA  should  be  captured  on  the  bead  stack,  with  the  remaining  cellular  components  (DNA,  proteins,  carbohydrates,  and  lipids)  washed  out  of  the  device.   Heavy chain Nested 1st round Primers Kappa chain Nested 1st round Primers NTC 100 cells 10 cells 1 cell 0.1 cells NTC 100 cells 10 cells 1 cell 0.1 cells 92 Beads   loaded   with   cellular   mRNA   were   sequentially   eluted   from   each  chamber  and  recovered  by  removal  and  replacement  of   the  stainless  steel  pin  and  Tygon  tubing  connected  to   the  output  port.  This  solution  was  mixed  with  reaction  mix  (e.g.  primers,  dNTPs,  and  enzyme)  and  placed  on  a   thermal  cycler   for  RT-­‐PCR  amplification.   As   samples   from   all   microfluidic   chambers   were   eluted   through   a  common   outlet   port,   a   test   for   cross-­‐contamination   between   samples   was  performed   by   alternating   elution   of   chambers   with   and  without   hybridoma   cells.  Amplification   of   antibody   genes   from   all   samples   confirmed   significant  contamination   between   samples,   which   was   not   removed   by   washing   the   output  port  both  with  1X  PBS  and  10%  bleach  solution  following  elution  of  each  chamber  (Figure  3.13).  This  cross-­‐contamination  was  caused  by  a   large  dead-­‐volume   in   the  outlet  port  (diameter  ~  500  μm)  that  is  not  adequately  flushed  by  the  much  smaller  feed  channel  (100  μm  wide  ×  100  μm  high).      An  alternative  method  of  recovering  samples  from  the  device  was  therefore  developed.  Devices  were  fabricated  with  a   large  elution  port  (diameter  ~  1.2  mm)  such   that   solution   could   be   extracted   by   manual   pipetting   out   of   the   outlet   port  using  a  gel-­‐loading  pipet  tip.  With  this  new  device  design,  very  little  contamination  was  observed  between  eluted  samples  when  the  output  port  was  thoroughly  flushed  by  repeated  pipetting  of  a  1X  PBS  solution  and  a  new  pipet   tip  was  used  between  each  eluted  sample  (Figure  3.14).       93   Figure  3.13           RT-­‐PCR  of  mouse  antibody  genes  from  mRNA  purified  on  oligo(dT)  beads  from   single  HyHEL-­‐5  hybridoma  cells  sorted  in  a  microfluidic  device.  Beads  from  chambers  with  a   single  cell  (“1  cell”)  and  without  cells  (“NTC”)  were  alternately  eluted  and  recovered  from  the   output  port  in  a  stainless  steel  pin  and  Tygon  tubing.  The  output  port  was  washed  with  1X  PBS   and  10%  bleach   in   between   each   sample   elution.     Significant   cross-­‐contamination  occurred   between   samples.   Shown   is   a   1%  DNA  agarose   gel  with  100  bp   ladder.   Pixel   intensities   are   inverted  to  highlight  amplified  products.       Figure  3.14           RT-­‐PCR   of   mouse   β-­‐actin   (A)   and   antibody   heavy   and   light   chain   genes   (B)   from   mRNA   purified   on   oligo(dT)   beads   from   single   HyHEL-­‐5   hybridoma   cells   sorted   and   recovered  from  a  microfluidic  device.  Beads  from  chambers  with  a  single  cell  (“HyHEL-­‐5  cell”   or   “D1.3”)   and   without   cells   (“NTC”)   were   alternately   eluted   and   recovered   by   manual   pipetting  with  a  new  gel-­‐loading   tip   for  each  sample.  Cross-­‐contamination  between  samples   occurred   if   the   output   port  was   insufficiently  washed  with   1X  PBS   in   between   each   sample   elution.  Shown  is  a  1%  DNA  agarose  gel  with  100  bp  ladder.   Chambers eluted in order from left to right NTC 1 cell NTC 1 cell NTC 1 cell Chambers eluted in order from left to right Beta-actin Antibody Heavy and Light chain Nested 1st round Primers Microfluidic elution (in order from left-to-right) Off-chip control NTC D1.3 }} NTC D1.3 NTC D1.3 A B NTC NTC HyHEL-5 cell HyHEL-5 cell HyHEL-5 cellNTC 94   The  above-­‐described  method  for  microfluidic  single-­‐cell  sorting,  lysis,  mRNA  purification  and  recovery  is  a  robust  method  for  RT-­‐PCR  amplification  of  expressed  genes   from   single   cells.     However,   the   chemical   lysis   step   in   this   method   may  preclude  its   integrated  use  with  methods  to  screen  secreted  antibodies  from  these  same  cells  as  described  in  Chapter  2.  Alkaline   lysis  solutions  containing  sodium  or  lithium   dodecyl   sulfate   (SDS   or   LiDS)   are   powerful   protein   denaturants;145   thus,  while   facilitating  purification  of   cellular  mRNA,   these   solutions  may  also  denature  secreted   antibodies   present   on   antibody-­‐capture   beads   adjacent   to   the   cell   in   the  microfluidic   chamber   (Figure   3.6C).   Denaturation   of   secreted   antibodies   will,  therefore,  restrict  screening  of  secreted  antibody  function  to  be  performed  prior  to  cell   lysis   and   recovery.   In   attempt   to   circumvent   this   limitation,   single   cells  were  eluted   from  microfluidic  chambers  without   lysis  or  mRNA  bead  purification.    Cells  were  recovered  from  the  device  output  port  by  manual  pipetting  with  a  gel-­‐loading  tip   and   transferred   directly   to   RT-­‐PCR   reaction   mix.     In   this   manner,   successful  amplification   of   both   heavy   and   light   chain   genes   was   performed   (Figure   3.15).    Interestingly,   no   cell   lysis   step  was   required,   suggesting   that   cells   lyse   in   RT-­‐PCR  reaction   mix   solution.   Moreover,   PCR   primers   for   heavy   and   light   chain   genes  successfully   primed   the   reverse   transcription   reaction   without   dedicated   RT  primers,  such  as  oligo(dT)  or  random  hexamers.  A  further  benefit  of  this  approach  is  the  fact  that  mRNA  is  retained  within  the  cell  during  elution  and  recovery,  reducing  the   chances   of   cross-­‐contamination   if   a   small   number   of   oligo(dT)   beads   is   left  behind   in   the   output   port   during   pipet   recovery   (Figure   3.14).   Indeed,   no   cross-­‐contamination   between   samples   was   observed   after   alternating   elution   of   95 microfluidic  chambers  containing  cells  or  no  cells   (Figure  3.15).  Most   importantly,  microfluidic   single-­‐cell   sorting   and   recovery   followed   by   optimized   RT-­‐PCR  amplification  (e.g.  using  low  degeneracy  primers)  resulted  in  >90%  success  rates  for  amplifying  antibody  genes  from  single  cells  from  three  different  mouse  hybridoma  cell-­‐lines  (D1.3,  HyHEL-­‐5,  and  CD1d).           Figure  3.15           RT-­‐PCR   of   antibody   heavy   and   light   chain   genes   from  D1.3   hybridoma   cells   sorted  and  recovered  from  a  microfluidic  device  without  on-­‐chip  cell  lysis.  Chambers  with  and   without   cells   were   alternately   eluted   and   recovered   by   manual   pipetting   with   a   new   gel-­‐ loading   tip   for   each   sample.   Samples   were   directly   transferred   to   RT-­‐PCR   reaction   mix   without   dedicated   RT   primers.   Antibody   heavy   and   light   chain   genes   were   successfully   amplified   from   eluted   samples   from   chambers   containing   single   D1.3   cells   as   well   as   two   chambers   loaded   with   multiple   cells   (2   and   6   cells,   respectively).   No   cross-­‐contamination   between  samples  was  observed.  RT-­‐PCR  performed  using  low  degeneracy  primers58  at  600nM   concentration.  Shown  is  a  1%  DNA  agarose  gel  with  100bp  ladder.     3.5 Conclusions  This   chapter   described   optimized   methods   for   RT-­‐PCR   amplification   of  antibody  heavy  and  light  chain  genes  from  low  template  concentrations.  Single-­‐cell  RT-­‐PCR  sensitivity  was   confirmed  using   low  degeneracy  primers   to  amplify  genes  from  mouse  hybridoma  cells  as  well  as  primary  antibody-­‐secreting  cells  harvested   Nested 1st round Primers Chambers eluted in order from left to right Ch 4 D1.3 cell Ch 5 NTC Ch 6 D1.3 cell Ch 7 NTC Ch 8 Six D1.3 cells Ch 2 Two D1.3 cells 96 from   immunized  mice.   Future   experiments   harnessing  microfluidic   single-­‐cell   RT-­‐PCR   and  high-­‐density   digital   PCR  may   enable   absolute   quantification   of   antibody-­‐encoding  mRNA  in  single  cells.75,81  Optimized   RT-­‐PCR   protocols   were   coupled   with   microfluidic   devices   for  sorting,  lysis,  and  recovery  of  single  cells.  Successful  amplification  of  antibody  genes  was  performed   from  single  mouse  hybridoma  cells  both  with  and  without  on-­‐chip  chemical   lysis   and   mRNA   bead   purification.   Moreover,   single-­‐cell   RT-­‐PCR  amplification  of   antibody  genes  was   successfully  performed  without  dedicated  RT  primers  in  the  reaction  mix,  indicating  that  gene-­‐specific  PCR  primers  can  prime  the  reverse  transcription  reaction.  Notably,  microfluidic  single-­‐cell  sorting  and  recovery  followed   by   optimized   RT-­‐PCR   amplification   resulted   in   >90%   success   rates   for  amplifying  antibody  genes  from  single  cells  from  three  different  mouse  hybridoma  cell-­‐lines  (D1.3,  HyHEL-­‐5,  and  CD1d).      The  microfluidic   device   architecture   for   sorting   and   recovering   single   cells  can  be  readily  adapted  to  screen  secreted  antibodies  from  single  antibody-­‐secreting  cells   co-­‐incubated   with   antibody-­‐capture   beads   (Figure   3.13).     The   following  chapter   will   describe   the   integration   of   a   microfluidic   fluorescence   bead   assay  (developed   in   Chapter   2)   and   the   optimized   RT-­‐PCR   methods   described   in   this  chapter  in  order  to  perform  screening  and  selection  of  novel  monoclonal  antibodies  from  single  primary  antibody-­‐secreting  cells  (ASCs)  from  immunized  animals.  This  approach  can  be  extended  to  the  amplification  of  single  ASCs  from  other  vertebrate  species,  such  as  rabbits  and  humans,  using  RT-­‐PCR  primers  complementary  to   the  unique  antibody  gene  sequences  present  in  each  species.19,50,58       97 Chapter    4: Rapid,   High-­‐Throughput   Screening   and   Selection   of   High   Affinity   Monoclonal   Antibodies   from   Single   Antibody-­‐ Secreting  Cells  This  chapter  describes   the  development  of  a  novel  pipeline  (Figure  4.1)   for  screening   and   selection   of   high   affinity   antigen-­‐specific   monoclonal   antibodies  (mAbs)   from   single   primary   antibody-­‐secreting   cells   (ASCs).   Animals   are   first  immunized   with   a   target   antigen.   ASCs   are   then   harvested   from   the   immunized  animals   and   purified   by   labeling   known   plasma   cell   surface   markers   to   perform  fluorescence-­‐activated   cell   sorting   (FACS).   Following  purification,  ASCs   are   loaded  into   a  microfluidic   device   and   loaded   as   single   cells   into   an   array   of  microfluidic  chambers   (volume  ~1   nL).   By   concentration   enhancement   (see   Chapter   1,   Figure  1.11),   single   ASCs   secrete   sufficient   numbers   of   mAbs   within   1   hour   for  measurements   of   antibody-­‐antigen   binding   kinetics   and   affinities   using   the  microfluidic   fluorescence   bead   assay   described   in   Chapter   2.   ASCs   producing  antigen-­‐specific  mAbs  are  sequentially  recovered  from  the  device  and  subjected  to  single-­‐cell   RT-­‐PCR   to   amplify   their   antibody   heavy   and   light   chain   genes   using  methods  described   in  Chapter  3.  Antibody  genes   for  high-­‐affinity  mAbs  were   then  cloned  into  expression  vectors  for  recombinant  production  in  mammalian  cell  lines.       98     Figure  4.1           Microfluidic  screening  and  selection  of  mAbs  from  single  cells.     This   technology   was   validated   through   the   selection   of   nearly   200   high-­‐affinity  mouse  mAbs   to   the  model   antigen   hen   egg   lysozyme   (HEL)   by   screening  fewer   than   1000   ASCs   harvested   from   HEL-­‐immunized   mice.   Over   80%   of   these  mAbs  bound  HEL  with  equilibrium  dissociation  constants  (Kd)  less  than  or  equal  to  1   nM   and   on-­‐rate   constants   (kon)   greater   than   106   M-­‐1s-­‐1.   Microfluidic   single-­‐cell  screening   yielded   a   10X   greater   number   of   high-­‐affinity   antigen-­‐specific   mAbs  compared   to   recently   described   single-­‐cell   micro-­‐well   screening   approaches  (Chapter  1,  Figure  1.12)  and  conventional  hybridoma  technology  (Chapter  1,  Figure  1.6),   without   requiring   time-­‐consuming   clonal   expansion   and   sub-­‐cloning   steps  required   in   the   latter   process.51,90114114,146   As   described   below,   this   microfluidics-­‐based   pipeline   for   screening   mAbs   from   single   cells   offers   the   potential   both   to  select   antigen-­‐specific   mAbs   with   desired   binding   kinetics   and   affinities,   and   to  improve   our   understanding   of   affinity  maturation   and   immune-­‐dominance   in   the  vertebrate  adaptive  immune  system.       Amplify, sequence, and express antibody genes Antigen-speci!c mAbs Screen and select single ASCs secreting antigen-speci!c mAbs Micro"uidic device Immunize animal (mice, rabbits, humans) with target antigen Harvest and purify ASCs 99   4.1 Experimental  Methods   4.1.1 Mouse  Immunization,  Harvesting  and  Purification  of  ASCs  BALB/c   and   C57JBL/6   strain   mice   (6-­‐wk   to   1-­‐year   old)   were   given   intra-­‐peritoneal   (IP)   injections   of   HEL   emulsified   in   Freund’s   incomplete   adjuvant   and  were  boosted  every  2  weeks  (2-­‐10  injections).  Mice  were  sacrificed  1  week  after  a  final  subcutaneous  booster   injection  with  HEL  in  Alum  and  single-­‐cell  suspensions  from   the  dissected   spleen  were  prepared  using   a   cell   strainer.  Mouse   splenocytes  were   then   stained  with   fluorescently-­‐labeled  antibodies   and   sorted  by   fluorescent  activated  cell  sorting  (FACS)  to  enrich  for  CD138+  cells,  a  known  cell  surface  marker  for  mouse  plasma  cells.58  An  ELISPOT  assay  (Figure  4.2)  was  used  to  determine  the  frequency  of  IgG-­‐  and  anti-­‐HEL  secreting  ASCs  in  the  FACS-­‐enriched  cell  population.  BALB/c  mice   yielded   50,000-­‐150,000   CD138+   sorted   into   500   μL   of   cell  medium  (Gibco  RPMI  1640  medium  with  10%  fetal  calf  serum  and  0.5%  mercaptoethanol).  The  cells  were  spun  down  into  a  pellet  by  centrifugation  at  1500rpm  for  5min  and  450  μL  of  cell  medium  was  removed.  The  cells  were  re-­‐suspended  in  the  remaining  ~50  μL  to  obtain  a  concentrated  cell  solution  (>106  cells/mL),  suitable   for   loading  into  microfluidic  devices.       100   Figure  4.2           Representative   results   from   ELISPOT   assay   to   determine   frequency   of   antigen-­‐ specific   (i.e.   anti-­‐HEL)   and   IgG-­‐secreting   ASCs   from   FACS-­‐enriched   mouse   splenocytes   (Image   prepared  by  Dr.  Welson  Wang,  Biomedical  Research  Centre,  UBC).     4.1.2 Reagent  Preparation  Protein   A-­‐coated   5.5   µm   diameter   polystyrene   beads   (Bangs   labs)   were  incubated  with  a  1mg/mL  solution  of  Rabbit  anti-­‐mouse  polyclonal  antibodies   (pAbs)  purchased   from   Jackson   Immunoresearch   and   used   without   further   purification.   All  antibody  and  antigen  solutions  were  prepared   in  PBS/BSA/Tween  solution  consisting  of  1X  PBS,  pH  7.4  (Gibco)  with  10mg/mL  BSA  (Sigma)  and  0.5%  Polyoxyethylene  (20)  sorbitan  monolaurate  (similar  to  Tween-­‐20,  EMD  Biosciences).  Lysozyme  from  chicken  egg  white  (HEL)  was  purchased  from  Sigma.  D1.3  and  HyHEL-­‐5  mouse  hybridoma  cell-­‐lines   were   generously   provided   by   Dr.   Richard   Willson   (University   of   Houston).  Fluorescent   protein   conjugates  were   prepared   using  Dylight488   and  Dylight633  NHS  esters  (Pierce)  and  were  purified  using  Slide-­‐A-­‐Lyzer™  dialysis  cassettes  (Pierce).  The  concentration   of   fluorescent   conjugates   was   measured   by   spectrophotometry  (Nanodrop).  In  order  to  minimize  protein  denaturation,  fluorescent  protein  conjugates  were  labeled  at  a  dye-­‐to-­‐protein  ratio  (D/P)  of  less  ~5.   101 4.1.3 Microfluidic  Device  Design  and  Operation  A  microfluidic  device  for  screening  antibody-­‐secreting  cells  (ASCs)  was  designed  and   fabricated   (Figure   4.3)   using   multilayer   soft   lithography.86,87   The   valves   were  designed   in  a  push-­‐down   format   such   that   flow  channels   could  be  directly  bonded   to  thin   (No.  1)   coverglass   for  high-­‐resolution  optical   imaging   (see  Chapter  2,   section  2.3  for  a  complete  description  of   fabrication  methods).  The  device  consists  of  an  array  of  112  discrete  chambers  (8  rows  ×  14  columns),  9  fluidic  inlets  for  different  reagents,  and  one   common   fluidic   outlet.   The   outlet  was  used  both   for   flushing   reagents   out   of   the  device  and  for  sample  recovery.  Reagent  flow  was  directed  into  particular  chambers  by  using  a  combination  of  a  multiplexing  valve  structure87  to  select  a  particular  row  in  the  chamber   array   and   a   separate   valve   for   each   column   of   chambers.   In   addition   to  enabling   reagent   flow   through   a   particular   chamber,   this   valve   architecture   enabled  fluid  flow  to  be  directed  into  an  arbitrary  set  of  chambers  in  a  single  row,  all  chambers  in   a   particular   column,   or   all   chambers   in   the   array.   Dedicated   valves   upstream   and  downstream   of   the   chamber   array  were   also   used   to  wash   each   row   of   the   chamber  array  before  delivery  of  a  new  reagent  into  the  chambers.  The   chambers   were   designed   for   reversible   trapping   of   ASCs   and   antibody-­‐capture   beads   (Figure   4.3B).   A   filter,   consisting   of   an   array   of   pillars   spaced   2-­‐3   μm  apart,  was  used  to  trap  cells  and  beads  >5μm  in  diameter.  The  entry  and  exit  regions  of  the   chambers   have   rounded   cross-­‐sectional   profiles  with   a   height   of   10-­‐15  μm  and   a  width  of  150  μm;  the  rounded  channel  profiles  enabled  integration  of  valves  to  partition  and  isolate  each  chamber  from  adjoining  chambers.  The  middle  chamber  regions  have  rectangular  cross-­‐section  with  a  height  of  2   -­‐  4  μm  aligned  to  a  partially-­‐closing  sieve   102 valve83   (see   Chapter   2,   Figure   2.2A   for   valve   design)   used   to  modulate   the   flow   rate  through   each   chamber.   The   total   volume   each   of   microfluidic   chamber   was  approximately  1  nL  (10-­‐9  L).  At   the   start   of   the   experiment,   the   sieve   valves  were   closed   in   order   to   create  high   flow   impedance  and   low   flow   rate   through   the   chambers,   thereby  enabling   cells  and  beads  to  be  trapped  upstream  of  the  bead  filter.  By  opening  the  sieve  valve,  the  flow  rate  through  the  chamber  was  increased,  causing  the  cells  (but  not  the  rigid  beads)  to  deform   and   squeeze   through   the   bead   filter   and   into   the   fluidic   outlet   for   manual  recovery  using  a  pipette.  Recovered  cells  were  then  transferred  to  RT-­‐PCR  reaction  mix  for   off-­‐chip   amplification   of   antibody   genes.   Antibody-­‐capture   beads   retained   in   the  microfluidic   chambers   were   used   for   subsequent   measurement   of   antibody-­‐antigen  binding  kinetics.  Importantly,  the  device  and  chamber  architecture  enabled  recovery  of  cells   without   subjecting   the   chambers   to   cell   lysis   solutions,   thus   avoiding   chemical  denaturation  of  secreted  antibodies  and  antigen  on  the  capture  beads.  Figure  4.4  shows  a  schematic  of  a  single  chamber  on  the  device  during  a  typical  mAb   screening   experiment.   At   the   start   of   the   experiment,   all   chambers   were   first  washed   with   1X   PBS   buffer   (Figure   4.4A).   Next,   ASCs   in   cell   culture   medium   (~106  cells/mL)   were   loaded   into   each   row   of   the   chamber   array   at   a   low   flow   rate   (~10  µL/hr).  Single  ASCs  were   loaded   into  each  chamber  by  using  a  microscope  to  observe  cells   approaching   a   chamber;   the   corresponding   column  valve  was   then  momentarily  opened  to  direct  reagent  flow  into  the  chamber  (Figure  4.4B).  In  this  manner,  over  100  single  cells  were  loaded  into  distinct  chambers  in  15-­‐20  minutes.  After  cell  loading  into  chambers,  all   remaining  cells   in  channels  upstream  and  downstream  of   the  chambers   103 were  washed   out   of   the   device.   A   solution   of   antibody-­‐capture   beads   in   cell  medium  was  then   loaded   into   the  chamber  array  and  beads  were  sequentially   trapped  against  the  bead  filter   in  all  112  chambers  (Figure  4.4C  and  D).  Antibodies  present   in  the  cell  medium  were   flushed  out  of   the  microfluidic  chamber  by   the   incoming  bead  solution,  ensuring   that   the   beads   captured   only   antibodies   secreted   by   adjacent   single   cells.  There  was  no  detectable  cross-­‐contamination  of  secreted  mAbs  between  chambers,  as  evidenced  by  the  lack  of  detectable  antigen  binding  to  beads  in  control  chambers  with  no  cells.   In  contrast,  antigen  bound  to  beads   in  all  chambers  (including   those  without  cells)   when   beads   were   loaded   into   the   device   chambers   either   before   or  simultaneously  with  cells.    Cells  and  adjacent  beads  were  co-­‐incubated   in  each  microfluidic  chamber  for  1  to   2   hours   to   allow   for   mAbs   to   be   secreted   and   captured   on   beads   at   detectable  concentrations  (Figure  4.4E).  After  incubation,  all  chambers  were  washed  with  1X  PBS  to  remove  mAbs  free  in  solution  (Figure  4.4F).  All  chambers  were  then  flushed  with  a  solution  of  fluorescently  labeled  antigen  (e.g.  HEL-­‐Dylight488  or  HA-­‐Dylight488)  for  5  to  10min  (Figure  4.4G  and  H).  The  chambers  were  again  washed  for  5  minutes  with  1X  PBS  prior  to  automated,  high-­‐resolution  fluorescence  imaging  (Figure  4.4I).       104     Figure  4.3           Microfluidic  device  for  screening  single  antibody-­‐secreting  cells  (ASCs).  (A)  Device   schematic   depicting   9   fluidic   inlets,   1   fluidic   outlet,   and   112   chambers   (8   rows   ×   14   columns)   addressed   using   a   row   multiplexer   and   column   valves.   (B)   Schematic   of   single   microfluidic   chamber   (volume   ~1   nL)   containing   a   bead   filter/trap   and   sieve   valve   to   modulate   flow   rate   through   chamber.   (C   and   D)   Bright-­‐field   microscope   images   of   sub-­‐nanoliter   microfluidic   chambers   containing   single  ASCs   and  antibody-­‐capture  beads   at  20X   (C)   and  40X  magnification   (D).     9 reagent inlets 8-row multiplexer 14 columns Output/ elution port 112 chambers Inlet channel Outlet channel Column valves Sieve valve Bead trap/ !lter A B DC 105   Figure  4.4           Schematic   of   microfluidic   chamber   while   performing   single-­‐cell   antibody   screening  and  selection.  See  text  for  details.  (Page  1  of  6).       A B 106   [Figure   4.4   con’t   -­‐   Schematic   of   microfluidic   chamber   while   performing   single-­‐cell   antibody   screening  and  selection.  See  text  for  details.  (Page  2  of  6).]   C D 107   [Figure   4.4   con’t   -­‐   Schematic   of   microfluidic   chamber   while   performing   single-­‐cell   antibody   screening  and  selection.  See  text  for  details.  (Page  3  of  6).]   E F Y Y Y Y Y Y Y Y Y Y 108   [Figure   4.4   con’t   -­‐   Schematic   of   microfluidic   chamber   while   performing   single-­‐cell   antibody   screening  and  selection.  See  text  for  details.  (Page  4  of  6).]       G H 109   [Figure   4.4   con’t   -­‐   Schematic   of   microfluidic   chamber   while   performing   single-­‐cell   antibody   screening  and  selection.  See  text  for  details.  (Page  5  of  6).]       I J Cell recovery for RT-PCR Image of positive chamber Image of negative chamber 10μm10μm 110   [Figure   4.4   con’t   -­‐   Schematic   of   microfluidic   chamber   while   performing   single-­‐cell   antibody   screening  and  selection.  See  text  for  details.  (Page  6  of  6).]   K L Time Fl uo re sc en ce Time Fl uo re sc en ce 111 Bright-­‐field  and   fluorescence   images  of  all   chambers  were   taken  using  a  Nikon  TE200   Eclipse   inverted   fluorescence   microscope   equipped   with   an   automated   xyz-­‐microscope   stage   (Prior   Scientific),   a   computer-­‐controlled   mechanical   shutter   (Ludl  Electronic   Products),   a   custom   LED   circuit   for   bright-­‐field   illumination,   a   100X   oil  immersion  objective  (N.A.  1.30,  Nikon  Plan  Fluor)  and  an  EMCCD  camera  (Hamamatsu).  Fluorescence   images   were   taken   using   green   (470/40   nm   excitation,   535/30   nm  emission)   and   red   (600/60   nm   excitation,   655   nm   long-­‐pass   emission)   fluorescence  filter  cubes  (Chroma  Technology).  Custom  LabView  software  was  developed  in  order  to  automate  hardware  control  and  image  acquisition  (Appendix  B.1).  A  custom  autofocus  algorithm   was   developed   that   consisted   of   taking   a   series   of   bright-­‐field   images   at  different  focal  (z-­‐axis)  positions  and  computing  the  variance  in  pixel  intensity  for  each  image.  The  image  with  maximum  variance  corresponded  to  the  focal  plane  at  the  edge  of   the  bead  surface  and,  along  with   the  known  bead  diameter,  was  used   to   locate   the  optimal   focal  position  at   the  center  of   the  beads,  as  confirmed  by  the  distinctive  bead  diffraction  pattern.    The   maximum   fluorescence   intensity   on   the   antigen-­‐bound   beads   in   each  chamber  was  measured  using  custom  LabView  software  (Appendix  B.2).  A  fluorescence  threshold  was  selected  by  computing  the  average  plus   two  standard  deviations  of   the  maximal   sorbate   fluorescence   intensity   from   multiple   control   chambers   deliberately  not  loaded  with  a  cell.  Chambers  with  fluorescence  intensity  higher  than  this  threshold  were  deemed  positive  for  antigen-­‐specific  mAbs  (95%  confidence  interval).  Cells  from  each   positive   chamber   were   recovered   from   the   device   (i.e.   ~1-­‐1.5   hours   after   112 incubation   in  microfluidic  devices)   in  order  to  minimize  the  chances  of  cell  death  and  mRNA  degradation  prior  to  RT-­‐PCR  amplification.  Selected   cells   from   chambers   deemed   positive   for   antigen-­‐specific  mAbs  were  sequentially   eluted   in   1X   PBS   into   the   output   port   and   manually   recovered   using   a  pipette  with  a  gel-­‐loading  tip  (Figure  4.4J).  After  each  sample  was  recovered,  the  elution  port   was   thoroughly   washed   by   repeated   pipetting   of   1X   PBS   solution.   Cells   were  directly  transferred  to  85  μL  of  Qiagen  One-­‐Step  RT-­‐PCR  reaction  mix  containing  20  μL  of  RT-­‐PCR  buffer,  42  μL  of  RNase-­‐free  water,  4  μL  of  dNTPs,  and  4  μL  of  RT  and  PCR  enzymes.  Each  sample  was  then  split   into  two  equal  parts,  to  which  7.5  μL  of  forward  and  reverse  primer  solutions  (8  μM)  were  added  for  single-­‐plex  amplification  of  heavy  and   light   chain   genes,   respectively.   RT-­‐PCR   amplification   was   performed   using  previously-­‐published   low   degeneracy   primers   that   annealed   to  mouse   V   gene   leader  and   constant   region   sequences58   (see  Appendix  A.2   and  Chapter   3   for   list   of   primers  and   discussion   of   primer   design).   As   roughly   90%   of   mouse  mAbs   utilize   the   kappa  (Igκ)  light  chain,  RT-­‐PCR  amplification  was  only  performed  using  primers  for  the  kappa  light   chain.   A   mixture   of   primers   was   used   to   amplify   heavy   chain   genes   encoding  different  antibody  isotypes  (e.g.  IgA,  IgM,  IgG1,  IgG2a,  etc.).  RT-­‐PCR  reaction  mixes  were  then  transferred  to  a  bench-­‐top  thermal  cycler  for  RT-­‐PCR  amplification  (see  Chapter  3,  Table  3.2  for  cycling  protocols).      Following   cell   recovery,   antibody-­‐antigen  binding  kinetics   and   selectivity  were  measured  on  beads  retained  in  the  device  using  a  microfluidic  fluorescence  bead  assay  (see   Chapter   2).   For   kinetic   measurements,   chambers   deemed   positive   for   antigen-­‐specific  mAbs  were  first  washed  with  1X  PBS  and  dissociation  kinetics  were  measured   113 by   time-­‐lapse   imaging   (Figure  4.4K).   Subsequently,   each   chamber  was   flushed  with   a  solution   of   fluorescently   labeled   antigen   (e.g.   HEL-­‐Dylight488   conjugate)   and   time-­‐lapse   images   were   taken   to   measure   association   kinetics   (Figure   4.4L).   Association  kinetics   were   measured   one   chamber   at   a   time   to   obtain   sufficient   time   resolution;  however,   dissociation   kinetics   were   simultaneously   measured   in   multiple   chambers  (e.g.  all  chambers  in  a  single  column)  using  automated  scanning.  As  secreted  antibodies  were   irreversibly   captured   on   beads   for   several   days   (see   Chapter   2,   Figure   2.7),  measurements   of   antibody-­‐antigen   kinetics   and   selectivity   could   be   performed   for  several   days   after   initial   mAb   capture.   Reported   error   represents   the   calculated  standard  deviation  of  multiple  replicate  measurements.  Values  measured  only  once  are  reported  without  error  bars.     4.1.4 Sequencing   of   Antibody   Heavy   and   Light   Chain   Genes   and   Recombinant   Expression  of  Selected  mAbs  After  RT-­‐PCR  amplification,  reaction  products  were  analyzed  by  electrophoresis  on  1%  DNA  agarose  gels  stained  with  1X  SYBR  Safe  DNA-­‐binding  dye.  Amplicons  were  extracted  from  the  gels  by  a  scalpel  and  purified  using  spin  columns  as  per  the  Qiagen  MinElute  kit.  Samples  were  eluted  from  the  spin  columns  in  low  TE  buffer  and  stored  in  a   -­‐20°C   freezer.   DNA   samples   and   associated   primers  were   submitted   to   the   Nucleic  Acid   Protein   Service   Unit   (NAPS,   http://www.msl.ubc.ca/services/naps)   for   standard  Sanger  sequencing.  Each  sample  was  sequenced  using  both  forward  5’  (leader  region)  and   3’   reverse   (constant   region)   primers.   Heavy   and   light   chain   genes   from   selected  anti-­‐HEL  mAbs  were  compared  with  mouse  germline  antibody  sequences  using   IMGT   114 V-­‐quest   (http://www.imgt.org/IMGT_vquest/share/textes/).   Common   sequence  differences  found  between  germline  genes  and  amplified  genes  sequenced  using  both  5’  and   3’   primers   were   labeled   as   somatic   mutations.   To   ensure   that   assigned   somatic  mutations  were   not   generated   by   polymerase   errors   during   RT-­‐PCR,   heavy   and   light  chain   genes   were   amplified   and   sequenced   in   triplicate   (Figure   4.5).   A   commercial  vendor  (Synogene,  Inc.,  http://synogeneinc.com/)  performed  cloning  and  recombinant  expression   of   selected  mAbs.   Briefly,   heavy   and   light   chain   genes   for   selected  mAbs  were   cloned   into   transfection   vectors,   expressed   as   whole   IgG1  molecules   in   human  embryonic  kidney   (HEK293)  cells,   and  purified  by  Protein  G  affinity  chromatography.  Purified   recombinant   mAbs   were   tested   for   antigen   reactivity   using   the   microfluidic  fluorescence  bead  assay.       Figure  4.5           Heavy  chain  genes  from  four  single-­‐cell  selected  anti-­‐HEL  mouse  mAbs  amplified   in   triplicate   by   RT-­‐PCR.   All   amplicons   were   extracted   and   purified   for   DNA   sequencing.   Comparison   of   DNA   sequences   of   amplicons   was   performed   to   verify   that   assigned   somatic   mutations   were   not   generated   by   polymerase   errors   during   RT-­‐PCR.   RxCy   nomenclature   designates  the  row  and  column  address   for   the  microfluidic  chamber   from  which  the  cells  were   recovered.  Shown  is  a  1%  DNA  agarose  gel  with  100  bp  ladder.   Heavy chain Nested 2nd round Primers Replicate 1 } Replicate 2 } Replicate 3 } R1C14 R1C14 R7C13 R0C12 R2C06 R1C14 R7C13 R0C12 R2C06R7C13 R0C12 R2C06 115 4.2 Results   4.2.1 Kinetic   Screening   and   RT-­‐PCR   Amplification   of   Antibody   Genes   from   Single  Hybridoma  Cells  Single   cells   from   two  distinct  mouse  hybridoma  cells   lines,  D1.3  and  HyHEL-­‐5,  were  loaded  into  ~1  nL  microfluidic  chambers  in  a  single  device.  Binding  kinetics  and  affinities   of   cell-­‐secreted   mAbs   measured   from   multiple   single   cells   of   each   cell-­‐line  were   precise   to   within   30%   (Table   4.1),   and   were   accurate   when   compared   with  binding   data   for   purified   mAbs   (see   Chapter   2,   Table   2.1).     Importantly,   secreted  antibodies  were  detected   from  all  hybridoma  cells   loaded   into  microfluidic  chambers.  HyHEL-­‐5  cells  secreted  mAbs  that  bound  HEL  with  a  ~50-­‐fold  higher  affinity  than  mAbs  secreted   by  D1.3   cells,   as   a   result   of  ~5-­‐fold   faster   association   kinetics   and  ~10-­‐fold  slower  dissociation  kinetics.  Chambers  containing  D1.3,  HyHEL-­‐5  and  no  cells  were  sequentially  eluted  into  a  common  output  port  on   the  device,   recovered,  and  subjected   to  RT-­‐PCR  amplification  (Figure   4.6).   RT-­‐PCR   reactions   performed   on   recovered   samples   from   chambers  containing   cells   always   yielded   amplified   antibody   genes,   whereas   no-­‐cell   control  chambers  did  not  yield  any  amplification  products   and  DNA  sequencing  of   amplicons  confirmed  that  the  correct  heavy  and  light  chain  genes  were  amplified  from  chambers  containing   distinct   hybridoma   cells.   These   experiments   therefore   show   that   the  described   microfluidic   device   (Figure   4.3)   enables   highly   efficient   measurement   of  antibody-­‐antigen   binding   kinetics   and   affinities   and   selective   recovery   of   single  hybridoma  cells   for  robust  RT-­‐PCR  amplification  of  antibody  genes.  These  results  also  suggest  that  cells  were  stable     116 Table  4.1     Antibody-­‐antigen   binding   kinetics   and   affinities   from   single   D1.3   and   HyHEL-­‐5   hybridoma   cells   measured   by   a   microfluidic   fluorescence   bead   assay   using   a   HEL-­‐Dylight488   fluorescent   conjugate.     Results   represent   the   average   and   standard   deviation   of   replicate   measurements  performed  on  multiple  distinct  D1.3  (n  =  30)  and  HyHEL-­‐5  (n  =  5)  cells.       Hybridoma   kon  (M-­‐1s-­‐1)   koff  (s-­‐1)   Kd  (M)  D1.3   (1.98  ±  0.32)  ×  106   (2.62  ±  0.48)  ×  10-­‐3   (1.33  ±  0.32)  ×  10-­‐9    HyHEL-­‐5     (9.08  ±  1.13)  ×  106   (1.56  ±  0.44)  ×  10-­‐4   (1.71  ±  0.53)  ×  10-­‐11       Figure  4.6           Multiplex   RT-­‐PCR   amplification   of   antibody   heavy   and   light   chain   genes   from   eluted  chambers  containing  no  cells  (no-­‐template  control,  NTC),  a  D1.3  cell,  and  HyHEL-­‐5  cell.  No   amplification  was  observed   from  eluted  chambers  containing  no  cells.    Amplified  gene  products   were  extracted  and  purified   from  the  gel  and  sequenced  to  confirm  that   they  correspond  to   the   corresponding  hybridoma  cell-­‐line.  Shown  is  a  1%  DNA  agarose  gel  with  100  bp  ladder.     4.2.2 Microfluidic   Screening   and   Selection   of   mAbs   from   ASCs   Purified   from   Mice  Immunized  with  HEL  The  complete  microfluidic  pipeline  for  selecting  mAbs  from  single  ASCs  (Figure  4.1)  was  demonstrated  from  mice  immunized  with  the  model  antigen  hen  egg  lysozyme  (HEL).  Injection  of  BALB/c  strain  mice  with  HEL  resulted  in  a  large  expansion  of  FACS-­‐purified  CD138+  mouse  splenocytes  (50,000  –  150,000  cells  per  spleen  versus  <10,000  cells  per  spleen  from  unimmunized  mice).  ELISPOT  analyses  confirmed  that  up  to  50%   Chambers eluted in order from left-to-right NTC D1.3 HyHEL-5 NTC D1.3 HyHEL-5 117 of   CD138+   splenocytes   secreted   IgG   antibodies,   with   roughly   half   of   these   cells  secreting   mAbs   that   bound   HEL.   In   contrast   to   BALB/c   mice,   C57BL/6   strain   mice  immunized   with   HEL   produced   only   2,000-­‐3,000   CD138+   mouse   splenocytes,  consistent  with   the   fact   that  HEL   fails   to  elicit   the  necessary  T-­‐cell  help   to  generate  a  robust  antibody  response  in  this  strain  of  mice.147,148    CD138+   splenocytes   harvested   from   HEL-­‐immunized   BALB/c   mice   were  screened   in   the  microfluidic   device   for   production   of   anti-­‐HEL  mAbs.   By   sequentially  directing  flow  to  each  chamber  using  row  and  column  valves,  over  100  single  cells  were  loaded   into   the   chamber   array.   In   contrast,   using   stochastic   cell   loading   strategies,  nearly   two-­‐thirds  of  chambers  are   typically   left  with  either  zero  or  >1  cells.51,149  Cells  and  antibody-­‐capture  beads  were  incubated  in  the  microfluidic  chambers  for  1  hour  in  order   to   ensure   successful   detection   from   all   cells   secreting   anti-­‐HEL   mAbs.   The  detection  limit  of  the  fluorescence  bead  assay  is  ~8×104  mAb  molecules  (see  Chapter  2,  section  2.4.2);  thus,  the  minimum  antibody  secretion  rate  required  for  detection  after  a  1-­‐hour   incubation   is   20   mAb   molecules   per   second   assuming   all   secreted   mAbs   are  captured   on   a   single   bead.   In   practice,   both   the   antibody   secretion   rate   and   bead  capture  efficiency  are  unknown;  however,  ASCs  are  known  to  secrete  thousands  of  mAb  molecules   per   second   under   ideal   conditions,   suggesting   that   most   (if   not   all)   ASCs  secreting  anti-­‐HEL  mAbs  would  be  detected  after  1  hour  incubation  in  the  microfluidic  device.21,124,150   By   comparison,   mouse   hybridoma   cells   secrete   at   least   200  mAbs/second  when   incubated   in  microfluidic   devices   under   identical   conditions   (i.e.  room  temperature  and  identical  cell  culture  medium,  see  Chapter  2,  section  2.4.3).   118 Nearly   200   high-­‐affinity   anti-­‐HEL  mAbs   (190)   were   identified   by   microfluidic  single-­‐cell  screening  of  CD138+  splenocytes  from  six  different  HEL-­‐immunized  mice.  On  average,   30   ±   10%   of   chambers   contained   cells   secreting   anti-­‐HEL  mAbs,   consistent  with   frequencies   of   antigen-­‐specific   ASCs   measured   by   ELISPOT   (Figure   4.7).   The  number   of   chambers   containing   cells   secreting   anti-­‐HEL   mAbs   was   increased   by  iterative  re-­‐loading  of  cells  into  the  device.  Specifically,  after  fluorescent  scanning,  cells  were  flushed  out  of  all  chambers  negative  for  anti-­‐HEL  mAbs  and  these  chambers  were  re-­‐loaded  with  new  cells.  This  process  can  be  repeated  multiple   times   to  obtain  more  complete  occupancy  of  chambers  with  cells  secreting  anti-­‐HEL  mAbs.     In   this   manner,  chambers  were  loaded  with  cells  up  to  three  times  in  a  single  experiment,  yielding  70  anti-­‐HEL  mAbs  (63%  of  chambers)  in  a  single  device.         Figure  4.7           Identification  of  microfluidic  chambers  containing  single  cells  secreting  anti-­‐HEL   mAbs.     After   flushing   chambers  with   fluorescently   labeled   antigen   (i.e.   14.3nM  HEL-­‐Dylight488   conjugate),   high-­‐resolution   fluorescence   imaging   of   all   chambers   is   performed.   The   maximum   fluorescence   bead   intensity   in   each   chamber   is   measured   and   a   threshold   is   set   equal   to   2   standard   deviations   larger   than   the   average   fluorescence   of   no-­‐cell   control   chambers   (95%   confidence  interval).       0 2000 4000 6000 8000 10000 12000 14000 16000 18000 Fl uo re sc en ce in te ns ity (a .u .) Chamber index ~32% positive chambers (34 out of 108) Threshold 119 A  single  device  yielded  a  comparable  number  (~50)  of  antigen-­‐specific  mAbs  to  a   single   hybridoma   fusion   without   the   weeks   of   clonal   expansion   and   sub-­‐cloning  required  in  the  latter  process.  Indeed,  the  number  of  anti-­‐HEL  mAbs  identified  using  the  microfluidic   single-­‐cell   screening   (190)   is  approximately   two-­‐fold   larger   than  all  anti-­‐HEL  mAbs  generated  by  hybridoma  methods  and  previously  reported  reported   in   the  literature.114,146  In  addition,  the  microfluidic  single-­‐cell  screening  approach  was  found  to  identify  antigen-­‐specific   mAbs   from   single   ASCs   with   higher   efficiency   and   throughput   than  recently   described   single-­‐cell   micro-­‐well   screening   approaches.51,90   Papa   et   al.  identified  14   single   cells   secreting   antigen-­‐specific   (i.e.   anti-­‐ovalbumin,   anti-­‐OVA)   IgG  mAbs  out  of  18,000  single  cells  harvested  from  antigen-­‐immunized  mice  and  stimulated  in   culture   for   3   days   prior   to   screening   in   micro-­‐engraved   wells.90   Similarly,   using  micro-­‐fabricated  wells   and   a   fluorescence   immunospot   assay   (ISAAC),   27   single   cells  secreting   anti-­‐HEL   mAbs   were   detected   after   screening   10,000   ASCs   from   HEL-­‐immunized  mice.   The   ~100-­‐fold   difference   in  measured   frequency   of   ASCs   secreting  anti-­‐HEL  mAbs   (0.27%  of   cells  using   ISAAC  versus  30%   in   the  microfluidic  device)   is  attributed   to   increased   sensitivity   and   a   lower   detection   limit   of   the   microfluidic  fluorescence  bead  assay  as  compared  with  ISAAC.51,112  Consistent  with  this  notion,  the  ISAAC  method  detected  as  few  as  15%  of  single  hybridoma  cells  secreting  high  affinity  (Kd   ~   200   pM)   anti-­‐HEL   mAbs.51   It   is   unlikely   that   the   discrepancy   in   frequencies  measured  by  microfluidic  screening  versus  ISAAC  is  caused  by  differences  in  the  purity  of   antigen-­‐specific   ASCs   in   the   enriched   cell   fraction   as   both   approaches   screened   120 similar   cell   populations,   namely   CD138+   plasma   cells   from   the   spleen   of   HEL-­‐immunized  mice.  Anti-­‐HEL   mAbs   were   detected   by   microfluidic   screening   of   CD138+   mouse  splenocytes  when  stored  for  36  hours  post-­‐harvest  at  4°C  in  standard  cell  medium  (i.e.  Gibco   RPMI   1640  medium  with   10%   fetal   calf   serum   and   0.5%  mercaptoethanol).   A  similar   frequency   of   chambers   containing   antigen-­‐specific   ASCs   were   detected   when  comparing   CD138+   mouse   splenocytes   screened   either   4   hours   or   36   hours   post-­‐harvest   (30%   of   chambers   versus   21%,   respectively).   As   one   round   of   microfluidic  single-­‐cell  mAb  screening  can  be  performed  in  less  than  4  hours,  hundreds  of  antigen-­‐specific   mAbs   can   be   identified   from   a   single   animal   by   using   multiple   devices   to  repeatedly  screen  ASCs  stored  for  1-­‐2  days  post-­‐harvest.  Corti  et  al.  recently  reported  that   ASCs   can   be   maintained   in   culture   for   up   to   a   week   using   cell   medium  supplemented   with   exogenous   cytokines   (e.g.   IL-­‐6)   that   promote   plasma   cell  survival.150   Thus,   in   principle,   ASCs   harvested   and   maintained   in   culture   can   be  subjected  to  dozens  of  microfluidic  antibody  screens  over  several  days   in  order  select  thousands  of  antigen-­‐specific  mAbs  from  a  single  animal.     4.2.3 Antibody-­‐Antigen  Binding  Kinetics  and  Affinities  of  Novel  Anti-­‐HEL  Mouse   mAbs    Naturally-­‐occurring  antibodies  are  known  to  bind  target  antigens  with  affinities  spanning   up   to   five   orders   of   magnitude   (Kd   values   between   10   µM   and   100   pM);93  however,   most   mAbs   used   for   research   and   therapeutic   applications   bind   with  moderate  to  high  binding  affinities  (Kd  <  10  nM).151  Average  Kd,  kon  and  koff  values   for   121 anti-­‐HEL   mAbs   selected   using   the   microfluidic   single-­‐cell   approach   were   2.8nM,  3.4×106  M-­‐1s-­‐1  and  5.3×10-­‐3  s-­‐1  (Table  4.2),  respectively.  Notably,  over  80%  and  90%  of  anti-­‐HEL  mAbs   (60   to   65   out   of   70)   characterized   in   the  microfluidic   device   had   Kd  values   less   than   10   nM   and   on-­‐rate   constants   greater   than   106   M-­‐1s-­‐1,   respectively  (Figure  4.8).  By  contrast,  in  a  sample  of  23  hybridoma-­‐generated  anti-­‐HEL  mAbs,  only  2  mAbs   (<10%)  were   found   to   bind  HEL  with   on-­‐rate   constants   greater   than   106  M-­‐1s-­‐ 1.114,152  Microfluidic   selection   of   anti-­‐HEL   mAbs   was   biased   against   Kd   values   greater  than   10nM   (i.e.   low-­‐affinity   mAbs)   by   flushing   the   chamber   array   with   a   low  concentration   of   fluorescent   antigen   (e.g.   214   ng/mL   or   14.3   nM   of   HEL-­‐Dylight488  conjugate)   prior   to   imaging   (see   Chapter   2,   equation   2.2).   Similarly,   by   flushing  chambers   with   a   14.3   nM   concentration   solution   of   HEL-­‐Dylight488   conjugate   for   a  short  time  period  (e.g.  5  min),  selection  was  biased  against  anti-­‐HEL  mAbs  with  on-­‐rate  constants  less  than  106  M-­‐1s-­‐1  (see  Chapter  2,  equation  2.3).  Thus,  by  carefully  selecting  the   fluorescent   antigen   concentration   and   loading   time,   mAb   selection   was   biased  toward   particular   kinetic   and   equilibrium   binding   constants.   Importantly,   the   above-­‐reported   frequency   of   antigen-­‐specific   ASCs   (~30%)   is   likely   an   underestimate   as   it  accounts  only  for  cells  secreting  moderate  to  high-­‐affinity  anti-­‐HEL  mAbs  (Kd  <  10  nM).  Goldbaum   et   al.   reported   that   approximately   20%   of   hybridoma-­‐generated   anti-­‐HEL  mouse  mAbs  bound  HEL  with  equilibrium  dissociation  constants  (Kd)  ranging  from  10  nM   to   ~100   nM.114   Thus,   accurate   frequencies   of   antigen-­‐specific   ASCs   can   be  determined  by  using  microfluidic  single-­‐cell  screening  with  high  antigen  concentrations   122 (e.g.   ~100   nM)   for   longer   periods   of   time   (e.g.   1   hour)   to   detect   both   low   and   high  affinity  anti-­‐HEL  mAbs.    Foote  and  Milstein  proposed  that  clonal  selection  of  B  cells  is  influenced  by  both  kinetic   and   equilibrium   binding   properties,   with   a   premium   placed   on   B   cells  expressing  mAbs  that  bind  target  antigens  rapidly  (i.e.  high  kon).20  For  this  reason,  it  is  often  assumed  that  koff  values  are  highly  correlated   to  equilibrium  Kd   values,  with  on-­‐rate   constants   confined   to   a   narrow   range   of   values.153   In   sharp   contrast   to   this  assumption,   no   correlation  was   observed  between  Kd,  kon   and  koff   values   for   anti-­‐HEL  mAbs   selected   using   the  microfluidic   single-­‐cell   approach   (Figure   4.9)   Selected   anti-­‐HEL  mAbs  exhibited  equilibrium  and  kinetic  rate  constants  both  spanning  ~2  orders  of  magnitude,   with   Kd   values   ranging   from   ~100   pM   -­‐   10   nM,   kon   values   ranging   from  ~105-­‐107  M-­‐1s-­‐1,  and  koff  values  ranging  from  ~5×10-­‐4-­‐10-­‐2  s-­‐1  (Table  4.2).  Furthermore,  this   large   diversity   of   kinetic   (on-­‐rate   and   off-­‐rate)   and   equilibrium   dissociation  constants  was  observed   in  mAbs   from  multiple  different  HEL-­‐immunized  mice   (Table  4.3).   Collectively,   anti-­‐HEL   mouse   mAbs   selected   by   microfluidic   screening   and  hybridoma  methods  exhibit  a  2  to  3  order  of  magnitude  range  in  kon   (104  -­‐  107  M-­‐1s-­‐1),   koff   (10-­‐2   -­‐   10-­‐4   s-­‐1)     and   Kd   (107   -­‐   1010   M)   values,114   demonstrating   that   affinity  maturation  can  significantly  alter  both  on-­‐rate  and  off-­‐rate  kinetics.    The  highest-­‐affinity  anti-­‐HEL  mAbs  selected  by  microfluidic  single-­‐cell  screening  exhibited  Kd  values  as  low  as  ~100  pM,  consistent  with  the  theoretical  affinity  “ceiling”  resulting   from   physiological   constraints   of   the   adaptive   immune   system.93   The  maximum   on-­‐rate   constants   of   antibody-­‐antigen   interactions   are   governed   by   the  diffusion  limit  and  stringent  geometric  requirements  for  antibody-­‐antigen  binding.93,96   123 For   antibody   interactions  with   small  molecule   haptens,   on-­‐rate   constants   as   large   as  108  M-­‐1s-­‐1  have  been  observed.20,93,154  As  protein  molecules  are  approximately  ten-­‐fold  larger   in   size   than   hapten   molecules   and,   thus,   have   ten-­‐fold   smaller   diffusion  constants,118,155   diffusion-­‐limited   on-­‐rate   constants   of   antibody   interactions   with  protein  antigens  would  be  approximately  107  M-­‐1s-­‐1  (see  Chapter  2,  equations  2.5  and  2.6).  Indeed,  three  microfluidic-­‐selected  mAbs  bound  HEL  with  kon  values  ranging  from    1×107   M-­‐1s-­‐1   –   2.5×107   M-­‐1s-­‐1,   comparable   to   the   fastest   binding   anti-­‐HEL   mAbs  previously  generated  by  hybridoma  methods  (HyHEL-­‐5  and  HyHEL-­‐10).156    As  cellular  endocytosis   typically  occurs  on   the   time  scale  of   several  minutes,   it  has   been   postulated   that   the   clonal   selection   process   cannot   discriminate   between  B  cells  producing  antibodies  that  bind  antigen  with  interaction  half-­‐lives  greater  than  one  hour   (i.e.   koff   <   10-­‐4   s-­‐1).93   Consistent   with   this   notion,   the   smallest   off-­‐rate   constant  observed  amongst  all  selected  anti-­‐HEL  mAbs  was  5.7  ×  10-­‐4  s-­‐1,  corresponding  to  a  half-­‐life   of  ~20  min   for   the   antibody-­‐antigen   complex.     Of   the   200   -­‐   300   anti-­‐HEL  mouse  mAbs  generated  by  both  microfluidic  single-­‐cell  screening  and  the  hybridoma  method,  less  than  a  handful  of  mAbs  bind  HEL  with  an  off-­‐rate  constant  less  than  10-­‐4  s-­‐1.105,114  The   rarity   of   such   mAbs   is   consistent   with   the   fact   that   B   cells   can   stochastically  generate  mAbs  with  very  low  off-­‐rate  constants,  but  that  the  adaptive  immune  system  does   not   positively   select   these   cells   over   other   B   cells   producing   high-­‐affinity  mAbs  with  koff  >  10-­‐4  s-­‐1.93       124   Table  4.2     Measured  antibody-­‐antigen  binding  kinetics  and  affinities  from  over  70  anti-­‐HEL   mAbs  selected  using  the  microfluidic  single-­‐cell  screening  approach.       kon  (M-­‐1s-­‐1)   koff  (s-­‐1)   Kd  (M)  Average   3.4  ×  106   5.3  ×  10-­‐3   2.8  ×  10-­‐9  Maximum   2.5  ×  107   3.0  ×  10-­‐2   1.7  ×  10-­‐8  Minimum   2.7  ×  105   5.7  ×  10-­‐4   1.4  ×  10-­‐10  Fold-­‐variation  (Max/Min)   93   51   120     Table  4.3     Range  of  kinetic  and  equilibrium  rate  constants  for  anti-­‐HEL  mAbs  selected  using   the  microfluidic  single-­‐cell  screening  approach  from  three  different  HEL-­‐immunized  BALB/c  mice.       kon  (M-­‐1s-­‐1)   koff    (s-­‐1)   Kd  (M)   Mouse   Max     Min     Max     Min     Max     Min    M1   6.9  ×  106   1.7  ×  106   4.0  ×  10-­‐3   5.8  ×  10-­‐4   7.8  ×  10-­‐10   1.4  ×  10-­‐10  M2   2.5  ×  107   5.5  ×  105   3.0  ×  10-­‐2   7.8  ×  10-­‐4   2.9  ×  10-­‐9   7.8  ×  10-­‐10  M3   7.7  ×  106   2.7  ×  105   2.3  ×  10-­‐2   9.5  ×  10-­‐4   1.7  ×  10-­‐8   2.7  ×  10-­‐9   125   Figure  4.8     Measured  binding  kinetics  and  affinities  from  ~70  anti-­‐HEL  mAbs  selected  by  microfluidic  single-­‐cell  screening.  Equilibrium   dissociation  constants  (A),  on-­‐rate  constants  (B),  and  off-­‐rate  constants  (C)  plotted  in  rank  order  of  affinity,  as  well  as  histograms  of  these   binding  constants  (D-­‐F).         0 5 10 15 20 0.10 0.25 0.50 0.75 1.00 2.50 5.00 7.50 10.00 Fr eq ue nc y Kd (× 10-9 M) 1.00E+05 1.00E+06 1.00E+07 1.00E+08 k o n ( M -1 s-1 ) 1.00E-10 1.00E-09 1.00E-08 1.00E-07 K d (M ) 0 5 10 15 20 25 0.25 0.50 0.75 1.00 2.50 5.00 7.50 10.00 25.00 Fr eq ue nc y ko (× 103 s-1) 0 5 10 15 20 25 30 0.25 0.50 0.75 1.00 2.50 5.00 7.50 10.00 25.00 Fr eq ue nc y kon (× 106 M-1s-1) D F E A B C 1.00E-04 1.00E-03 1.00E-02 1.00E-01 k o (s -1 ) 126     Figure  4.9     No   correlation   observed   between   equilibrium   and   kinetic   binding   rate   constants   for   over   70   anti-­‐HEL   mAbs   selected   by   microfluidic  single-­‐cell  screening.  R2  values  correspond  to  linear  regression  of  the  plotted  data.     R² = 0.08977 1.00E-10 1.00E-09 1.00E-08 1.00E-07 1.00E-04 1.00E-03 1.00E-02 1.00E-01 K d (M ) ko (s-1) R² = 0.11907 1.00E-11 1.00E-10 1.00E-09 1.00E-08 1.00E-07 1.00E+05 1.00E+06 1.00E+07 1.00E+08 K d (M ) kon (M-1s-1) R² = 0.38488 1.00E-04 1.00E-03 1.00E-02 1.00E-01 1.00E+05 1.00E+06 1.00E+07 1.00E+08 k o (s -1 ) kon (M-1s-1) A B C 127 4.2.4 Analysis  of  Heavy  and  Light  Chain  Genes  from  Novel  Anti-­‐HEL  Mouse  mAbs    ASCs   secreting   anti-­‐HEL   mAbs   were   selectively   recovered   from   microfluidic  devices   in   order   to  perform   single-­‐cell  RT-­‐PCR  amplification  of   heavy   and   light   chain  genes   (Figure   4.10).   In   total,   75   ASCs   secreting   anti-­‐HEL   mAbs   were   selectively  recovered  from  microfluidic  devices  and  subjected  to  RT-­‐PCR  for  both  heavy  and  light  chain  genes.  These  reactions  yielded  48  Igκ  and  24  IgH  genes,  corresponding  to  RT-­‐PCR  success   rates   of   64%   and   32%,   respectively.   Both   heavy   and   light   chain   genes   were  amplified  for  18  anti-­‐HEL  mAbs  (24%  of  recovered  ASCs).  These  RT-­‐PCR  success  rates  were   comparable   to   previously   reported   success   rates   (20-­‐40%)   for   amplification   of  antibody  genes  from  single  mouse  B  cells  in  RT-­‐PCR  tubes.58  VDJ  gene  usage  was  analyzed  for  heavy  and  light  chain  genes  from  selected  anti-­‐HEL  mAbs.  Kappa  chains  were  encoded  by  22  unique  Vκ  genes,  representing  9  out  of  the  16   distinct   mouse   Igκ   gene   families58   (Figure   4.11).   Heavy   chain   gene   usage   was  considerably   less   diverse   than   kappa   chain   gene   usage,   with   14   unique   VH   genes  amplified,   representing   only   4   out   of   16   out   of   mouse   IgH   gene   families58.   A   large  fraction   of   amplified   heavy   chain   genes   (35%)   belonged   to   the   VH1   gene   family,  consistent  with   a   previous   report   that   this   gene   family   accounted   for   nearly   75%   of  heavy   chain   genes   produced   in   mouse   B   cells.58   The   diversity   (D)   and   junction   (J)  regions  of  selected  anti-­‐HEL  mAbs  were  encoded  by  4  out  of  10  D  gene  families,  and  all  four  JH  and  Jκ  gene  families,  respectively.   128   Figure  4.10     Single-­‐cell   RT-­‐PCR   amplification   of   antibody   heavy   and   light   chain   genes   from   ASCs  secreting  anti-­‐HEL  mAbs.  Cells  are  sequentially  recovered  from  the  microfluidic  device  (left-­‐ to-­‐right).   RxCy   nomenclature   designates   the   row   and   column   address   for   the   microfluidic   chamber  from  which  the  cells  were  recovered.  Shown  is  a  1%  DNA  agarose  gel  with  100  bp  ladder.     R0C04 R2C04 R2C07 R3C06 R5C12 R5C14 R6C01 R6C10 Chambers eluted in order from left-to-right Chambers eluted in order from left-to-right Kappa Chain Genes Heavy Chain GenesA B R0C04 R2C04 R2C07 R3C06 R5C12 R5C14 R6C01 R6C10 129   Figure  4.11     Light   (A)   and   heavy   (B)   chain   gene   usage   for   anti-­‐HEL  mouse  mAbs   selected   by   microfluidic  single-­‐cell  screening.     Although  selected  anti-­‐HEL  mAbs  were  encoded  by  a  greater  diversity  of  kappa  chain  genes,  heavy  chain  genes  were  more  highly  mutated  than  kappa  chain  genes.  The  mean   homologies   of   kappa   and   heavy   chain   sequences   to   their   respective   germline  genes  were  96.7%  and  93.2%,  corresponding  to  approximately  9  and  20  somatic  DNA  mutations  per  gene,  respectively.  These  mutations  produced  an  average  of  2.6  and  7.3  amino   acid   substitutions   per   kappa   and   heavy   chain,   respectively.  Whereas   the  most  heavily  mutated  heavy   chain   contained  15  amino  acid   substitutions,   the  most  heavily  mutated  kappa  chain  contained  only  5  substitutions.  Surprisingly,  20%  of  all  amino  acid  substitutions   involved   serine   substitutions   to   asparigine   (N),   arginine   (R),   or   glycine   0.0% 5.0% 10.0% 15.0% 20.0% 25.0% 30.0% 35.0% 40.0% 1 2 3 5 Fr eq ue nc y VH gene family n = 27 0.0% 5.0% 10.0% 15.0% 20.0% 25.0% 30.0% 3 4 5 6 8 9 10 12 19 Fr eq ue nc y Vκ gene family n = 40 A B 130 (G).   It   is   unclear  whether   somatic  mutation   favoured   serine   substitutions  or  whether  these  substitutions  are   functionally  significant   for   facilitating  mAb  binding   to   the  HEL  protein.   Serine   substitutions   have   previously   been   used   to   to   modulate   the  hydrophobicity  of  the  antibody  binding  pocket.157    Interestingly,   no   correlation  was   observed   between   the   degree   of  mutation   of  antibody   heavy   and   kappa   chains   and   the   binding   kinetics   or   affinities   of   the  corresponding   anti-­‐HEL  mAbs   (Table   4.4).   As   a   corollary,   multiple   high-­‐affinity   anti-­‐HEL   mAbs   consisted   of   heavy   or   light   chains   completely   homologous   to   germ-­‐line  sequences.   Similarly,   comparing   two   anti-­‐HEL   mAbs   previously   generated   by  hybridoma  methods  (HyHEL-­‐8  and  HyHEL-­‐10),  HyHEL-­‐8  mAb  binds  HEL  with  a  5-­‐fold  lower   affinity   than   HyHEL-­‐10   despite   being   encoded   by   the   same   kappa   and   heavy  chain   genes   and  having  more   than   twice   the  number   of   amino   acid   substitutions   (17  versus  7,   respectively).152   It   thus  appears   that  HEL-­‐immunized  BALB/c  mice  generate  several   different   (near)   germ-­‐line-­‐encoded   antibodies   that   bind   antigen   with   high  affinity.   This   observation   is   consistent   with   past   studies   that   found   that   the   average  avidity   of   anti-­‐HEL   mAbs   did   not   increase   during   the   immune   response   in   HEL-­‐immunized  mice,   and   that   affinities   and   association   rate   constants   of   anti-­‐HEL  mAbs  were   independent   of   the   number   of   immunizations   and   immunization   dosages.114,146 131 Table  4.4   Binding   kinetics,   affinities,   VDJ   gene   usage   and   number   of   amino   acid   (AA)   substitutions   in   kappa   and   heavy   chain   gene   sequences  for  select  subset  of  selected  anti-­‐HEL  mAbs.    (n/a  =  not  amplified,  i.e.  the  corresponding  kappa  or  light  chain  gene  did  not  amplify   by  RT-­‐PCR).  [continued  on  next  page]     mAb   kon     (×106  M-­‐1s-­‐1)   koff    (×10-­‐3  s-­‐1)   Kd    (×10-­‐9  M)   Vk  gene   Jk  gene   AA  (Kappa)   AA  (total)  M3_R03C04   3.6   1.0   0.3   IGKV5-­‐48*01   IGKJ1*01   2   -­‐  M3_R06C03   2.1   0.6   0.3   IGKV3-­‐2*01   IGKJ2*01   0   -­‐  M1_R05C14   3.8   2.8   0.7   IGKV5-­‐43*01   IGKJ2*01   3   10  M1_R06C01   3.6   2.9   0.8   IGKV5-­‐39*01   IGKJ2*01   1   8  M2_R00C03   4.4   3.8   0.9   n/a   n/a   -­‐   -­‐  M2_R07C08   2.4   2.3   1.0   IGKV4-­‐74*01   IGKJ1*01   3   13  M2_R07C06   6.5   8.3   1.3   IGKV4-­‐74*01   IGKJ1*01   4   19  M2_R01C08   0.6   0.8   1.3   IGKV3-­‐12*01   IGKJ2*01   5   10  M2_R06C08   2.6   4.9   1.9   IGKV19-­‐93*01   IGKJ2*01   4   10  M1_R06C10   1.8   4.5   2.5   IGKV5-­‐43*01   IGKJ2*01   1   7  M1_R00C04   1.7   4.3   2.6   IGKV10-­‐96*01   IGKJ4*01   2   9  M1_R02C07   3.0   1.9   6.3   IGKV8-­‐21*01   IGKJ1*01   3   -­‐   132 Table  4.4  [continued  from  previous  page]  –  For  clarity,  information  on  binding  kinetics,  affinities,  and  total  AA  substitutions  is  repeated  on   both  pages.     mAb   kon     (×106  M-­‐1s-­‐1)   koff    (×10-­‐3  s-­‐1)   Kd    (×10-­‐9  M)   VH  gene   DH  gene   JH  gene   AA  (Heavy)   AA  (total)  M3_R03C04   3.6   1.0   0.3   n/a   n/a   n/a   -­‐   -­‐  M3_R06C03   2.1   0.6   0.3   IGHV3-­‐8*02   -­‐   IGHJ4*01   -­‐   -­‐  M1_R05C14   3.8   2.8   0.7   IGHV3-­‐8*02   IGHD4-­‐1*01   IGHJ2*01   7   10  M1_R06C01   3.6   2.9   0.8   IGHV2-­‐9-­‐1*01   IGHD4-­‐1*02   IGHJ4*01   7   8  M2_R00C03   4.4   3.8   0.9   IGHV3-­‐8*02   -­‐   IGHJ3*01   3   -­‐  M2_R07C08   2.4   2.3   1.0   IGHV3-­‐8*02   -­‐   IGHJ3*01   10   13  M2_R07C06   6.5   8.3   1.3   IGHV1S81*02   -­‐   IGHJ1*01   15   19  M2_R01C08   0.6   0.8   1.3   IGHV3-­‐8*02   -­‐   IGHJ3*01   5   10  M2_R06C08   2.6   4.9   1.9   IGHV3-­‐8*02   -­‐   IGHJ3*01   6   10  M1_R06C10   1.8   4.5   2.5   IGHV3-­‐8*02   IGHD4-­‐1*02   IGHJ2*01   6   7  M1_R00C04   1.7   4.3   2.6   IGHV1S130*01   IGHD2-­‐4*01   IGHJ4*01   7   9  M1_R02C07   3.0   1.9   6.3   n/a   n/a   n/a   -­‐   -­‐   133 Several  VH  and  Vκ    genes  encoded  multiple  distinct  anti-­‐HEL  mAbs.  Most  notably,  the  VH3-­‐8  gene  accounted   for  80%  of   sequences  assigned   to   the  VH3  gene   family   and  nearly  30%  (8  out  of  27)  of  all  amplified  heavy  chain  genes.  In  contrast,  the  entire  VH3  gene  family  accounted  for  only  ~5%  of  heavy  chains  amplified  from  hundreds  of  B  cells  harvested   from   unimmunized   mice.58   Similarly,   multiple   selected   anti-­‐HEL   mAbs  contained   light  chains  encoded  by   the  Vκ3-­‐2,  Vκ4-­‐61,  Vκ4-­‐74,  Vκ5-­‐39,  Vκ5-­‐43,  Vκ10-­‐96,  and  Vκ19-­‐93  genes.  Of  these,  the  Vκ5-­‐43  encoded  light  chains  for  the  greatest  number  of  distinct   anti-­‐HEL  mAbs   (5).   Anti-­‐HEL  mAbs  with   the   VH3-­‐8   heavy   chain  were   paired  with  kappa  chains  encoded  by   five  different  Vκ  gene  families,  whereas  anti-­‐HEL  mAbs  encoded  by  the  Vκ5-­‐43  light  chain  gene  were  paired  with  heavy  chains  encoded  by  three  different  heavy  chain  genes  (VH1-­‐9,  VH3-­‐8,  and  VH5-­‐4).  Interestingly,   the  VH3-­‐8  and  Vκ5-­‐43  genes  also  encode  multiple  anti-­‐HEL  mAbs  previously  generated  by  hybridoma  methods  (Table  4.5).  The  VH3-­‐8  and  Vκ5-­‐43  genes  encode  respective  heavy  and   light  chains   for   the  HyHEL-­‐10,  HyHEL-­‐8,  HyHEL-­‐26,  and  HyHEL-­‐63  anti-­‐HEL  mAbs.152  The  Vκ5-­‐43  gene  encodes  an  additional   two  hybridoma-­‐generated   anti-­‐HEL  mAbs   (D44.1   and  F10.6.6)  paired  with   a  heavy   chain   encoded  by  the   VH1-­‐9   gene     (Table   4.5).   The   significant   over-­‐representation   of   anti-­‐HEL   mAbs  encoded  by  one  or  both  of  the  Vκ5-­‐43  and  VH3-­‐8  genes  suggests  that  the  corresponding  heavy   and   light   chains   serve   as   structurally   complementary   scaffolds   to   the   HEL  protein,  thus  yielding  high-­‐affinity  anti-­‐HEL  mAbs.  Anti-­‐HEL  mAbs  encoded  by  identical  Vκ5-­‐43  and  VH3-­‐8  genes  displayed  a  diverse  range  of  binding  kinetics  and  affinities  despite  being  90-­‐95%  homologous  in  amino  acid  sequence  (Figure  4.12  and  Table  4.6).  For  example,  the  HyHEL-­‐10  and  HyHEL-­‐26  mAbs   134 bind  HEL  with  a  10-­‐fold  difference   in  binding  affinity,  despite   the   fact   that  HyHEL-­‐26  contains  identical  amino  acid  residues  to  HyHEL-­‐10  at  all  but  one  position  (in  the  kappa  chain   CDR1   region)   known   to   contact   HEL   at   the   HyHEL-­‐10/HEL   binding   interface  (Figure   4.12).152   Similarly,   the   M1_R05C14   and   M1_R06C10   mAbs,   which   are   also  encoded  by   the  VH3-­‐8  and  Vκ5-­‐43  genes,  bind  HEL  with  a  3-­‐fold  difference   in  binding  affinity,   while   having   only   four   total   amino   acid   differences   in   both   heavy   and   light  chain  CDR  regions.  The  existence  of  multiple  amino  acid  substitutions  in  the  1st  and  2nd  heavy  chain  framework  regions  of  otherwise  homologous  anti-­‐HEL  mAbs  with  distinct  binding   affinities   and   kinetics   (Figure   4.12   and   Table   4.5)   indicates   that   residues   at  these   positions   may   either   be   present   at   the   antibody-­‐antigen   contact   interface   or  induce   functionally-­‐significant   changes   to   the   antibody   structural   scaffold.117   Taken  together,  these  observations  are  consistent  with  the  fact  that  relatively  few  amino  acid  substitutions   can   significantly   alter   antibody-­‐antigen   binding   kinetics,   affinities,   and  specificities.114,152   Indeed,   a   single   amino   acid   substitution   can   theoretically   alter  antibody-­‐antigen   binding   affinity   by   up   to   3   orders   of   magnitude   if   it   abolishes   or  generates  a  new  hydrogen  bond  at  the  binding  interface.114    Many  amino  acid  substitutions  in  both  framework  and  CDR  regions  were  found  to  be  common  between  anti-­‐HEL  mAbs  encoded  by   identical  Vκ  and  VH  genes   (Figure  4.12).  For  example,  the  glycine  (G)  to  aspartic  acid  (D)  substitution  observed  at  position  37  (IMGT)  in  H-­‐CDR1  of  HyHEL-­‐10,  HyHEL-­‐8,  and  HyHEL-­‐26  mAbs  was  observed  in  half  (4  out  of  8)  of  newly  selected  anti-­‐HEL  mAbs  encoded  by  the  VH3-­‐8  gene.  Similarly,  the  serine  (S)  to  glycine  (G)  conversion  at  position  36  (IMGT)  in  the  K-­‐CDR1  of  HyHEL-­‐10  and  HyHEL-­‐8  mAbs  was  conserved  in  2  out  of  4  newly  selected  mAbs  encoded  by  the   135 Vκ5-­‐43   gene.   Interestingly,   identical   somatic   mutations   even   in   highly   homologous  antibodies   had   significantly   different   effects   on   mAb   binding   affinity.   Of   the   newly  selected   high-­‐affinity   anti-­‐HEL   mAbs   encoded   by   the   VH3-­‐8   gene,   half   (4   out   of   8)  contained   the   germ-­‐line   alanine   (A)   residue   at   position   103   in   the   H-­‐CDR3   region,  whereas  the  other  half  contained  a  mutated  aspartic  acid  (D)  residue  also  found  in  the  HyHEL-­‐10  mAb   (Figure  4.12).   Reversion   of   the  D   to  A   amino   acid   substitution   in   the  HyHEL-­‐10  mAb  was  previously  shown  to  reduce  its  binding  affinity  for  HEL  by  nearly  4  orders   of   magnitude   (9000-­‐fold).158   In   contrast,   newly   selected  mAbs   containing   the  alanine   germ-­‐line   residue   were   capable   of   binding   HEL  with   sub-­‐nanomolar   binding  affinities;  for  instance,  the  M3_R06C03  mAb  binds  HEL  with  Kd  equal  to  ~300  pM  (Table  4.5).  Thus,  mAbs  encoded  by  identical  Vκ  and  VH  genes  may  contain  similar  heavy  and  kappa  chain  scaffolds,  but  likely  have  functionally  distinct  binding  epitopes  governed  by  context-­‐specific  somatic  mutations  unique  to  each  mAb.  As  the  heavy  chain  CDR3  (H-­‐CDR3)  region  is  the  site  for  VDJ  recombination,  this  region   typically   contains   the   greatest   sequence   diversity   in   the   antibody  molecule.159  Consistent  with  this  fact,  the  H-­‐CDR3  region  contained  the  greatest  sequence  diversity  in  all  anti-­‐HEL  mAbs  encoded  by  the  VH3-­‐8  gene  (Figure  4.12).  The  highest  affinity  anti-­‐HEL   mAb   selected   by   microfluidic   single-­‐cell   screening   contained   an   identical   CDR3  region   to   the   high-­‐affinity   HyHEL-­‐10   mAb.   In   contrast,   the   CDR3   sequences   most  divergent   from   the   HyHEL-­‐10  mAb  were   found   in   the   lowest   affinity   anti-­‐HEL  mAbs  (HyHEL-­‐26  and  D44.1)  and  the  X25  mAb,  which  is  encoded  by  the  same  Vκ  and  VH  genes  but  binds   the  hapten  dinitrophenyl   (DNP)  and  not  HEL.  The  X25  mAb  contains  germ-­‐ 136 line  residues  in  all  CDR  regions  other  than  H-­‐CDR3,  implicating  sequence  differences  in  the  H-­‐CDR3  region  as  a  primary  determinant  for  mAb  binding  specificity.152,159  The   discovery   of   nearly   identical   mAbs   generated   from   different   HEL-­‐immunized   mice   was   a   particularly   surprising   result.   Of   the   eight   different   mAbs  containing  VH3-­‐8-­‐encoded  heavy  chains  generated  from  three  different  mice,  six  mAbs  were   recombined  with  an   identical  D  allele   (IGHD4-­‐1*01)  and  either   the   IGKJ2*01  or  IGKJ3*01  J  allele  (Table  4.6).  Anti-­‐HEL  mAbs  with   identical  VDJ  usage  were  generated  by   multiple   independent   recombination   events   in   different   ASCs,   as   confirmed   by  analysis  of  nucleotide  sequences  in  the  heavy  chain  junction  region  of  these  mAbs.  VDJ  recombination   is   imprecise  due   to  addition  or  deletion  of  untemplated  nucleotides   to  the   junctional   N   regions   by   terminal   deoxyribonucleotidyl   transferase   (TdT)   and  exonuclease   enzymes,   respectively;   thus,   the   junction   sequence   serves   as   a   unique  signature   of   a   particular   recombination   event.2   All   four   anti-­‐HEL  mAbs   from   a   single  mouse  (M2)  were  derived  from  a  single  recombination  event,  whereas  four  additional  anti-­‐HEL  mAbs  were  generated  by  four  distinct  recombination  events   in  two  different  mice   (M1   and   M3)   (Table   4.6).   Of   these,   two   mAbs   (M1_R05C14   and   M1_R06C10)  derived   from   independent   heavy   chain   recombination   events   in   a   single  mouse  were  paired  with   the   identical  kappa  chain  (Vκ5-­‐43)  as  multiple  hybridoma-­‐generated  anti-­‐HEL  mAbs  (HyHEL-­‐10,  HyHEL-­‐8,  HyHEL-­‐26,  and  HyHEL-­‐63)  (Table  4.5).  The   fact   that  these  mAbs  were  generated  by  different  methods  in  distinct  mice  many  decades  apart  indicates  that  the  mouse  antibody  immune  response  to  HEL  exhibits  preferential  V  gene  usage,   VDJ   recombination   and   heavy/light   chain   pairing. 137   Figure  4.12     Amino   acid   sequences   of   anti-­‐HEL   mAbs   encoded   by   IgHV3-­‐8   heavy   (A)   and   IgK5-­‐43   kappa   (B)   chain   genes.     HyHEL-­‐10,   HyHEL-­‐26,   HyHEL-­‐8,   HyHEL-­‐63,   F10.6.6,   and   D44.1   are   hybridoma-­‐generated   anti-­‐HEL   mAbs,   whereas   X25   is   an   anti-­‐DNP   mAb.   Boxed   residues   are   those   that   contact  HEL   in   the  HyHEL-­‐10/HEL   complex.   Sequences   are   aligned   and   clustered  using  ClustalX.  Truncated  heavy   chain  sequences  (i.e.  M2_R01C08  and  M3_R01C03)  were  not  used  for  clustering.  CDR  regions  are  highlighted  by  shaded  boxes.     ********* ** * ***********: .**.*****.**.:* * *::.**.*:*:***:.******* *** :*. ***.:*:****** ** ** :* IGHV3-8*02 EVQLQESGPSLVKPSQTLSLTCSVTGDSITSGYWNWIRKFPGNKLEYMGYISYSGSTYYNPSLKSRISITRDTSKNQYYLQSNSVTTEDTATYYCARTGT--AYWGQGTLVT--- 110 X25 DVQLQESGPSLVKPSQTLSLTCSVTGDSITSGYWNWIRKFAGNKLEYMGYISYSGSTYYNPSLKSRISITRDTSKNQYYLQLNSVTTEDTATYYCARYRTTFDYWGQGTTLT--- 112 HyHEL-26 EVQLQESGPSLVKPSQTLSLTCSVTGDSITSDYWSWIRKFPGNKLEYMGYISYSGSTYYNPSLKSRISITRDASKNQYYLQLNSVTAEDTATYYCARWEM--DYWGQGTSVT--- 110 HyHEL-63 EVQLQESGPSLVKPSQTLSLTCSVTGDSVTSDYWSWIRKFPGNKLEYMGYISYSGSTYYHPSLKSRISITRDTSKNQYYLQLNSVTTEDTATYYCASWGG--DVWGAGTTVTVSS 113 M1_R06C10 EVQLQESGPSLVKPSQTLSLTCSVTGDSITRGYWNWIRKFPGNKLECMGYISYSGGTYYNPSLKSRISITRDIFKNQYYLQLHSVTTEDTATYYCATWDG--DYWGQGTTLTVSS 113 M1_R5C14 EVQLQESGPSLVKPSQTLSLTCSVTGDSITNDYWSWIRKFPGNKLEYMGYISYSGSTYYNPSLKGRISITRDTSKNQCYLQLISVTPEDTATYYCANWDG--DCWGQGTTLTVSS 113 M2_R06C08 EVQLQESGPSLVKPSQTLSLTCSVTGDSITGGYWSWIRKFPGNKLEYMGYITYSGSTFYNPSLKSRISITRDTSKNQYFLQLNSVTTDDTATYYCADWDG--AYWGQGTLVTVSA 113 M2_R07C08 EVQLQESGPSLVKPSQTLSLTCSVTGDSITSDYWSWIRKFPGNKFEYMGYITYSGSTYYNPSLRGRISITRDTSKNQYYLHLSSVTTDDSATYYCADWDG--AYWGQGTLVTVSA 113 M2_R00C03 EVQLQESGPSLVKPSQTLSLTCSVTGDSITSGYWSWIRKFPGNELEYMGYMSYSGSTYYNPSLKSRISITRDTSKNQYYLQLNSVTTEDTATYYCADWDG--AYWGQGTLVTVSA 113 HyHEL-10 DVQLQESGPSLVKPSQTLSLTCSVTGDSITSDYWSWIRKFPGNRLEYMGYVSYSGSTYYNPSLKSRISITRDTSKNQYYLDLNSVTTEDTATYYCANWDG--DYWGQGTLVT--- 110 HyHEL-8 EVQLQESGPSLVKPSQTLSLTCSVTGDSIISDYWSWIRKFTGNKLEYMEYISFSGNTFYNPSLKSRISITRDTSKNQHYLQLSSVTTEDTATYYCANWDG--TYWGQGTLVT--- 110 M3_R06C03 PPQLQESGPSLXKPXQXLSLTCSVTGDSITSDYWSWIRKFPGNRLEYMGYISHSGNTFYNPSLKSRISITRDTSKNQFYLQLNSVTTEDTATYYCANWDG--DYWGQGTTLTVSS 113 M3_R01C03 ----------------------------IXSGYWRGIRKFPGNKLEYMGYMSXSGSTYYNPSLKSRISITRDTSKNXPPLQLNSVTTEDTATYYCATWDG--DYWGQGTTLTVS- 84 M2_R01C08 -----------------LSLTCSVTGDSITSDYRGWIRKFPGNKLEYMGYISYSGNTYYNPSLKSRVSITRDTSKNQYYLQLNSVTTEDTATYYCADWDG--AYWGQGTLVTVSA 96 CDR1 CDR2 CDR3 1.......10........20........30........40........50........60........70........80........90.......100.......110..... ** ***.* *******:************.********************.*** ***************** :: :*******:******..** : D44.1 DIELTQSPATLSVTPGDSVSLSCRASQSISNNLHWYQQKSHESPRLLIKYVSQSSSGIPSRFSGSGSGTDFTLSINSVETEDFGMYFCQQSNSWPRTFGGGTKLEIKR--- 108 F10.6.6 DIELTQSPATLSVTPGDSVSLSCRASQSISNNLHWYQQKSHESPRLLIKYTSQSMSGIPSRFSGSGSGTDFTLSINSVETEDFGVYFCQQSGSWPRTFGGGTKLDIKR--- 108 X25 DIVLTQSPATLSVTPGDSVSLSCRASQSISNNLHWYQQKSHESPRLLIKYASQSISGIPSRFSGSGSGTDFTLSINSVETEDFGMYFCQQSNSWPFTFGSGTKLEIK---- 107 HyHEL-10 DIVLTQSPATLSVTPGNSVSLSCRASQSIGNNLHWYQQKSHESPRLLIKYASQSISGIPSRFSGSGSGTDFTLSINSVETEDFGMYFCQQSNSWPYTFGGGTKLEIK---- 107 HyHEL-8 DIVLTQSPATLSVTPGDSVSLSCRASQSIGNNLHWYQQKSHESPRLLIKYASQSISGIPSRFSGSGSGTDFTLTINTVETEDFGMYFCQQSNNWPYTFGGGTKLEIK---- 107 HyHEL-26 DIVLTQSPATLSVTPGDSVSLSCRASQSISNNLHWYQQKSHESPRLLIKYASQSISGIPSRFSGSGSGTDFTLSVNSVETEDFGMYFCQQSNSWPYTFGGGTKLEIK---- 107 HyHEL-63 DIVLTQSPATLSVTPGDSVSLSCRASQSISNNLHWYQQKSHESPRLLIKYASQSISGIPSRFSGSGSGTDFTLSINSVETEDFGMYFCQQSNSWPYTFGGGTKLEIKR--- 108 M1_R5C14 DIVLTQSPDTLSVTPGDSVSLSCRASQSISNNLHWYQQKSHESPRLLIKYASQSISGIPSRFSGSGSGTDFTLSIISVETEDFGMYFCQQSNNWPYTFGGGTKXX------ 105 M1_R06C10 DIVLTQSPATLSVTPGDSVSLSCRASQSIGNNLHWYQQKSHESPRLLIKYASQSISGIPSRFSGSGSGTDFTLSINSVETEDFGMYFCQQSNSWPYTFGGGTKXXXXXXXX 111 M3_R05C04 DIVLTQSPATLSVTPGDSVSLSCRASQSISNNLHWYQQKSHESPRLLIKYASQSISGIPSRFSGSGSGTDFTXSINSVETEDFGMYFCQQSSSWPRTFGXXTKXXXX---- 107 IGKV5-43*01 DIVLTQSPATLSVTPGDSVSLSCRASQSISNNLHWYQQKSHESPRLLIKYASQSISGIPSRFSGSGSGTDFTLSINSVETEDFGMYFCQQSNSWPVHVRRGDQAGNK---- 107 M3_R06C13 DIVLTQCPATLSVTPGDSVSLSCRASQSISNNLHWYQQKSHESPRLLIKYASQSISGIPSRFSGSGSGTDFTLSINSVETEDFGMYFCQQSNSWPYTSEXX-QAE------ 104 M3_R05C07 DIVLTQSPATLSVTPGDSVSLSCRASQNIGNNLHWYQQKSHESPRLLIKYASQSISGIPSRLSGSGSGTDFTXSINNVETEDFGMXXX------NRVTAGRTRSEXXQ--- 102 1.......10........20........30........40........50........60........70........80........90.......100.......110. CDR1 CDR2 CDR3 A B 138 Table  4.5     Binding  kinetics,  affinities,  VDJ  gene  usage  and  CDR  sequences   for  anti-­‐HEL  mAbs  encoded  by   the  Vκ5-­‐43  and  VH3-­‐8  genes.     (n/a  =  not  amplified,  n/r  =  not  reported).  mAbs  are  listed  in  order  of  binding  affinity  to  HEL  (highest  affinity  at  the  top).  mAbs  marked  with   an  asterisk  are  encoded  by  both  Vκ5-­‐43  kappa  and  VH3-­‐8  heavy  chains.  The  heavy  chain  diversity  (D)  region  of  some  mAbs  was  not  identified.   [continued  on  next  page]     mAbs   kon  (M-­‐1s-­‐1)   koff  (s-­‐1)   Kd  (M)   Vk  gene   Jk  gene   K-­‐CDR1   K-­‐CDR2   K-­‐CDR3  Germline   -­‐   -­‐   -­‐   IGKV5-­‐43*01   IGKJ2*01   QSISNN   YAS   QQSNSWPVH  F10.6.6   7.2E+06   7.0E-­‐04   1.0E-­‐10   IGKV5-­‐43*01   IGKJ1*01   QSISNN   YTS   QQSGSWPRT  HyHEL-­‐10*   2.3E+05   5.2E-­‐05   2.2E-­‐10   IGKV5-­‐43*01   IGKJ2*01   QSIGNN   YAS   QQSNSWPYT  M3_R06C03   2.1E+06   6.4E-­‐04   3.0E-­‐10   IGK3-­‐2*01   IGKJ2*01   ESVDNYGISF   AAS   QQSKEVPYT  M3_R01C03   2.2E+06   9.3E-­‐04   4.2E-­‐10   n/a   n/a   -­‐   -­‐   -­‐  M1_R05C14*   3.8E+06   2.8E-­‐03   7.4E-­‐10   IGKV5-­‐43*01   IGKJ2*01   QSISNN   YAS   QQSNNWPYT  M2_R00C03   4.4E+06   3.8E-­‐03   8.7E-­‐10   n/a   n/a   -­‐   -­‐   -­‐  M2_R07C08   2.4E+06   2.3E-­‐03   9.7E-­‐10   IGKV4-­‐74*01   IGKJ1*01   SSVSSSF   STS   HQYHRSPPT  HyHEL-­‐8*   1.9E+05   2.2E-­‐04   1.1E-­‐09   IGKV5-­‐43*01   IGKJ2*01   QSIGNN   YAS   QQSNNWPYT  M2_R01C08   6.1E+05   7.8E-­‐04   1.3E-­‐09   IGKV3-­‐12*01   IGKJ2*01   KSVSTSGYSY   LVS   QHIRELT  M2_R06C08   2.6E+06   4.9E-­‐03   1.9E-­‐09   IGKV19-­‐93*01   IGKJ2*01   QDINKY   YTS   IQYDNXPYTF  M1_R06C10*   1.8E+06   4.5E-­‐03   2.5E-­‐09   IGKV5-­‐43*01   IGKJ2*01   QSIGNN   YAS   QQSNSWPYT  HyHEL-­‐26*   1.7E+05   4.9E-­‐04   2.9E-­‐09   IGKV5-­‐43*01   IGKJ2*01   QSISNN   YAS   QQSNSWPYT  D44.1   4.2E+04   2.9E-­‐03   1.4E-­‐07   IGKV5-­‐43*01   IGKJ1*01   QSISNN   YVS   QQSNSWPRT   139 Table   4.5   [continued   from   previous   page]   –   For   clarity,   information   on   binding   kinetics   and   affinities   is   repeated   on   both   pages.   All   hybridoma-­‐generated  mAbs   from   the  HyHEL   series   have   ~10-­‐fold   lower   on-­‐   and   off-­‐rate   constants   compared   to   the  microfluidic-­‐selected   anti-­‐HEL  mAbs;  it  is  possible  that  this  discrepancy  resulted  from  methodological  differences  (e.g.  previous  measurements  were  performed  by   SPR  using  Fab  fragments  whereas  measurements  in  this  study  were  performed  using  the  microfluidic  bead  assay  and  full-­‐length  IgG  mAbs).     mAbs   kon  (M-­‐1s-­‐1)   koff  (s-­‐1)   Kd  (M)   VH  gene   DH  gene   JH  gene   H-­‐CDR1   H-­‐CDR2   HCDR3  Germline   -­‐   -­‐   -­‐   IGHV3-­‐8*02     IGHJ3*01   GDSITSGY   ISYSGST   ARTGTAY  F10.6.6   7.2E+06   7.0E-­‐04   1.0E-­‐10   IGHV1-­‐9*01     IGHJ2*01   GYTFTTYW   ILPGSDST   ARGDGFYVY  HyHEL-­‐10*   2.3E+05   5.2E-­‐05   2.2E-­‐10   IGHV3-­‐8*02     IGHJ3*01   GDSITSDY   VSYSGST   ANWDGDY  M3_R06C03   2.1E+06   6.4E-­‐04   3.0E-­‐10   IGHV3-­‐8*02   IGHD1-­‐1*01   IGHJ4*01   GDSITSDY   ISHSGNT   ANWDGDY  M3_R01C03   2.2E+06   9.3E-­‐04   4.2E-­‐10   IGHV3-­‐8*02   IGHD3-­‐3*01   IGHJ2*01   IXSGY   MSXSGST   ATWDGDY  M1_R05C14*   3.8E+06   2.8E-­‐03   7.4E-­‐10   IGHV3-­‐8*02   IGHD4-­‐1*01   IGHJ2*01   GDSITNDY   ISYSGST   ANWDGDC  M2_R00C03   4.4E+06   3.8E-­‐03   8.7E-­‐10   IGHV3-­‐8*02   IGHD4-­‐1*01   IGHJ3*01   GDSITSGY   MSYSGST   ADWDGAY  M2_R07C08   2.4E+06   2.3E-­‐03   9.7E-­‐10   IGHV3-­‐8*02   IGHD4-­‐1*01   IGHJ3*01   GDSITSDY   ITYSGST   ADWDGAY  HyHEL-­‐8*   1.9E+05   2.2E-­‐04   1.1E-­‐09   IGHV3-­‐8*02     IGHJ3*01   GDSIISDY   ISFSGNT   ANWDGTY  M2_R01C08   6.1E+05   7.8E-­‐04   1.3E-­‐09   IGHV3-­‐8*02   IGHD4-­‐1*01   IGHJ3*01   GDSITSDY   ISYSGNT   ADWDGAY  M2_R06C08   2.6E+06   4.9E-­‐03   1.9E-­‐09   IGHV3-­‐8*02   IGHD4-­‐1*01   IGHJ3*01   GDSITGGY   ITYSGST   ADWDGAY  M1_R06C10*   1.8E+06   4.5E-­‐03   2.5E-­‐09   IGHV3-­‐8*02   IGHD4-­‐1*01   IGHJ2*01   GDSITRGY   ISYSGGT   ATWDGDY  HyHEL-­‐26*   1.7E+05   4.9E-­‐04   2.9E-­‐09   IGHV3-­‐8*02     IGHJ4*01   GDSITSDY   ISYSGST   ARWEMDY  D44.1   4.2E+04   2.9E-­‐03   1.4E-­‐07   IGHV1-­‐9*01     IGHJ2*01   GYTFSTYW   ILPGSGST   ARGDGNYGY   140 Table  4.6   Nucleotide  sequences  of  the  heavy  chain  junction  region  for  anti-­‐HEL  mAbs  encoded  by  the  VH3-­‐8  gene.    mAbs  marked  with  an   asterisk  also  utilize  the  same  kappa  chain  gene  (Vκ5-­‐43).     mAb   N1-­‐REGION   P5'D   D-­‐REGION   P3'D   N2-­‐REGION   P5'J   5'J-­‐REGION  M1_R06C10*   c   -­‐   ctgggac   -­‐   ggg   -­‐   gactactgg  M1_R05C14*   -­‐   -­‐   actgggac   -­‐   ggg   -­‐   gactgctgg  M2_R06C08   gg   -­‐   actgggac   -­‐   gg   -­‐   tgcttactgg  M2_R00C03   gg   -­‐   actgggac   -­‐   gg   -­‐   tgcttactgg  M2_R07C08   gg   -­‐   actgggac   -­‐   gg   -­‐   tgcttactgg  M2_R01C08   gg   -­‐   actgggac   -­‐   gg   -­‐   tgcttactgg  M3_R06C03   -­‐   -­‐   attgggatgg   -­‐   c   -­‐   gactactgg  M3_R01C03   -­‐   -­‐   cttgggatg   -­‐   gc   -­‐   gactactgg   141 4.2.5 Cloning  and  Expression  of  Novel  Anti-­‐HEL  Mouse  mAbs  To  validate  the  microfluidic  single-­‐cell  antibody  selection  method,  a  single  high-­‐affinity   anti-­‐HEL   mAb   was   commercially   produced   by   cloning   and   recombinant  expression  as  a   full-­‐length  mouse   IgG  antibody.  The   recombinant  mAb  was   tested   for  binding  to  HEL  using  the  microfluidic  fluorescence  bead  assay.  The  cloned  mAbs  bound  HEL  with  similar  binding  kinetics  and  affinities  as  observed  when  selected  from  mouse  ASCs  (Table  4.7  and  Figure  4.13).  Although  recombinant  production  of  additional  mAbs  is   required   to   quantify   overall   success   rates,   this   experiment   demonstrates   that,   in  principle,  mAbs  selected  from  the  microfluidic  platform  retain  their  antigen  specificity,  binding  kinetics,  and  affinities  when  produced  by  standard  recombinant  methods.     Figure  4.13     Sample   association   and   dissociation   curves   of   recombinant   M1_R6C01   mAb   binding   to   HEL-­‐Dylight488   fluorescent   conjugate   (14.3   nM   concentration)   measured   using   the   microfluidic   fluorescence   bead   assay.   Solid   line   represents   experimental   fit   using   mass-­‐action   equations  (Chapter  2,  equations  2.7a-­‐c)       0 0.2 0.4 0.6 0.8 1 1.2 0 5 10 15 20 25 N or m al iz ed F lu or es ce nc e In te ns ity (a .u .) Time (min) Recombinant M1_R6C01 142 Table  4.7   Comparison  of  binding  kinetics  of  M1_R06C01  anti-­‐HEL  mouse  mAb  selected  from   single   ASC   and   produced   by   recombinant   expression   in   mammalian   cells.   Reported   error   represents   the   calculated   standard   deviation   of   multiple   replicate   measurements.   Values   measured  only  once  are  reported  without  error  bars.     mAb   kon  (M-­‐1s-­‐1)   koff  (s-­‐1)   Kd  (M)  M1_R06C01   3.6×106   2.9×10-­‐3   7.9×10-­‐10  Recombinant  M1_R06C01     (4.5  ±  1.3)  ×106   (4.6  ±  0.7)  ×10-­‐3   (1.0  ±  0.3)  ×10-­‐9     4.3 Conclusions  The  microfluidic  single-­‐cell  screening  method  described  in  this  chapter  provides  a   rapid   and   high-­‐throughput   route   for   screening,   selection   and   production   of   high  affinity   antigen-­‐specific   monoclonal   antibodies   (mAbs).   The   technology   facilitates  selection   of   mAbs   based   on   both   kinetic   and   equilibrium   binding   parameters,   which  may  be  particularly  useful   for  the  selection  of  therapeutic  mAbs  that  bind  their  target  antigens  with  Kd  values   less  than  10  nM.12  Selection  of  mAbs  based  on  kinetic  off-­‐rate  constants  (i.e.,  very  long  interaction  half-­‐lives)  may  also  produce  therapeutic  mAbs  that  can  be  administered  less  frequently  or  in  smaller  dosages,  thus  reducing  both  the  cost  and   side   effects   of   administering   these   therapies.   Methods   for   laboratory-­‐directed  evolution  of   antibodies   (e.g.  phage  display,   yeast  display,   etc.)   typically   select  binding  candidates   libraries   based   on   an   equilibrium   binding   or   kinetic   dissociation-­‐based  screen.160  As  a  result,  laboratory-­‐evolved  variants  of  mAbs  often  have  a  large  diversity  of  off-­‐rate  constants  (e.g.  3-­‐  to  4-­‐orders  of  magnitude)  but  a  relatively  small  range  of  on-­‐rate   constants   (e.g.   10-­‐fold).102,153   Thus,   the  microfluidic   single-­‐cell-­‐screening  method  may  be  uniquely  suited  to  on-­‐rate  selection  of  mAbs  by  screening  the  large  diversity  of  on-­‐rate   constants   produced  by   the  natural   immune   system.  On-­‐rate   selection  may  be   143 particularly   useful   for   anti-­‐viral   therapies,   in  which  mAbs  with   large   association   rate  constants  may  serve  as  entry  inhibitors  for  potent  virus  neutralization.100    Selection  of  mouse  anti-­‐HEL  mAbs  selected  by  both   the  microfluidic   single-­‐cell  screening  and  hybridoma  methods  revealed  several  interesting  features  of  the  adaptive  immune  system.  Firstly,  no  correlation  was  observed  between  the  numbers  of  somatic  coding   mutations   and   binding   kinetics   and   affinities   of   associated   anti-­‐HEL   mAbs,  consistent  with  the  idea  that  affinity  maturation  increases  the  number  and  diversity  of  mAbs   rather   than   an   increase   in   the   average   affinity   of   antigen-­‐specific   mAbs.146  Analysis   of   gene   sequences  of   anti-­‐HEL  mAbs   also   revealed   that  both   repertoire  drift  (i.e.  point  mutations  to  antibody  genes)  and  repertoire  shift  (i.e.  different  V  gene  usage)  are   capable   of   generating   large   diversity   in   antigen-­‐antibody   binding   kinetics   and  affinities.  For  example,   the  D44.1  and  F10.6.6  mAbs  vary  by  3  orders  of  magnitude   in  HEL-­‐binding   affinity   (102   pM   versus   144   nM)   despite   differing   by   a   total   of   only   5  amino  acid  residues  in  both  heavy  and  light  chain  CDR  regions.  Similarly,  a  single  amino  acid   substitution   between   the   HyHEL-­‐10   mAb   and   its   nearest   germ-­‐line   sequence  altered  binding  of  HyHEL-­‐10  to  HEL  by  nearly  4  orders  of  magnitude  (i.e.  9000-­‐fold).158  This   diversity   in   binding   affinities   exceeds   the   range   of   binding   affinities   of   all  microfluidic-­‐selected   anti-­‐HEL  mAbs   encoded   by   one   to   two   dozen   unique   Vκ   and   VH  genes,  respectively.  By  contrast,  based  on  the  mouse  immune  response  to  the  2-­‐phenyl-­‐oxazolone  (phOx)  hapten,  Foote  and  Milstein  concluded  that  repertoire  drift  resulted  in  more   modest   (<10-­‐fold)   changes   to   antibody-­‐antigen   binding   kinetics   and   affinities  when   compared   to   repertoire   shift.20   Thus,   it   appears   that   the   relative   importance  of  repertoire  drift  and  shift  in  generating  mAbs  with  diverse  binding  kinetics  and  affinities   144 may  vary   in   the   immune   response   to  different   antigens.  The   relative   improvement   in  antibody  on-­‐  and  off-­‐rate  constants  by  affinity  maturation  may  also  vary  in  the  immune  response  to  different  antigens.93    Early   reports   of   antibody  V   gene  usage   indicated   that   fetal  mice  preferentially  use  particular  V  gene  segment  based  on  particular  chromosomal  organization,  but  that  this  preferential  usage  is  lost  throughout  development.161  Thus,  V  gene  segment  usage  in  antibodies  produced  by  adult  mice  is  dependent  on  the  strain-­‐specific  V  gene  family  size.161   By   sequencing   the   entire   antibody   repertoire   in   both   immature   and   adult  zebrafish,   Quake   and   coworkers   have   also   identified   a   “stereotyped”   development,   in  which   particular   VDJ   usage   is   highly   enriched   between   different   individuals.162  However,   these   studies   have   not   shed   light   on   whether   stereotyped   antibody  repertoires   tend   to   bias   antigen-­‐specific   immune   responses   to   particular   V   gene  segments.    Anti-­‐HEL   mAbs   selected   by   microfluidic   single   cell   screening   and   hybridoma  methods   from  distinct  mice  many  decades   apart  displayed  an  unprecedented   level   of  stereotypy,  consisting  of  preferential  V  gene  usage,  VDJ  recombination  and  heavy/light  chain  pairing.  Epitope  mapping  studies  and  crystallographic  structures  of  mouse  mAbs  in  complex  with  the  HEL  protein  suggest  that  the  stereotyped  anti-­‐HEL  mAb  response  is  functionally   significant.146,163  Crystal   structures  of  multiple  mAbs   in   contact  with  HEL  indicate  that  both  heavy  and  light  chain  amino  acid  residues  make  contact  with  HEL  at  the   binding   interface.152,163   Preferential   heavy/light   chain   pairing   among   anti-­‐HEL  mAbs   may   thus   reflect   positive   selection   in   the   adaptive   immune   system   for   mAbs  heavy  and  light  chains  that  jointly  form  effective  HEL-­‐binding  domains.     145 Stereotypy  in  the  anti-­‐HEL  mAb  response  also  appears  to  be  caused  by  particular  (immunodominant)   regions   of   the   HEL   protein.   Smith-­‐Gill   and   coworkers   previously  reported   that   anti-­‐HEL  mouse  mAbs   can  be   sub-­‐divided   into   three   “complementation  groups”   that   bind   three   non-­‐overlapping   regions   on   the   HEL   protein.146   These  complementation   groups   were   functionally   defined   by   competitive   binding  experiments,   where   mAbs   from   different   complementation   groups   simultaneously  bound  different   regions   of   the  HEL  protein,  while   co-­‐binding  of  mAbs   from   the   same  complementation  group  was  inhibited  (Figure  4.14).  Using  this  approach,  D1.3,  HyHEL-­‐5,   and   HyHEL-­‐10   mAbs   were   identified   as   representative   members   of   the   three  different  complementation  groups  and  anti-­‐HEL  mAbs  encoded  by  identical  heavy  and  light   chain   genes  were   found   to  belong   to   the   same   complementation  group.146  Thus,  the  enrichment  of  anti-­‐HEL  mAbs  with  identical  V  gene  usage,  VDJ  recombination  and  heavy/light  chain  pairing  as  the  HyHEL-­‐10  mAb  suggest  that  these  mAbs  all  bind  to  an  immunodominant   region   of   the   HEL   protein.   Several   anti-­‐HEL   mAbs   selected   by  microfluidic   single-­‐cell   screening  were   encoded   by   only   one   of   the  Vκ5-­‐43   and  VH3-­‐8  genes    (Table  4.5).    In  addition,  one  selected  anti-­‐HEL  mAb  was  encoded  by  the  same  VH  gene   as   D1.3   (VH2-­‐6-­‐7),   while   another   mAb   was   encoded   by   the   same   Vκ   gene   as  HyHEL-­‐5  (Vκ4-­‐59).  Competitive  binding  experiments  of  these  mAbs  will  be  required  to  investigate  whether   the  complementation  group  of  a  particular  mAb  can  be  predicted  solely  based  on  only  heavy  or  light  chain  gene  usage.     146   Figure  4.14   The  HEL  protein  can  be  sub-­‐divided  into  three  non-­‐overlapping  regions  that  bind   to   distinct   (“complementation”)   groups   of   mAbs.146   D1.3,   HyHEL-­‐5,   and   HyHEL-­‐10   are   representative  members  of  the  three  different  complementation  groups.  Image  reproduced  from   Batista  et  al  (Cell  Press,  1998).115    Though  not  typically  observed  across  multiple  distinct  animals,  restricted  V  gene  usage   has   also   been   previously   reported   in   the  mouse   antibody   response   to   haptens  phOx   and   phosphorylcholine,   the   amyloid   β   peptide,   polysaccharides,   as   well   as   the  human   antibody   response   to   infectious   bacteria   [e.g.   haemophilus   influenza]   and  viruses   [e.g.   human   immunodeficiency   virus   (HIV),   influenza   virus,   cytomegalovirus  (CMV)].20,164–170   Thus,   antibody   stereotypy  may   be   a   common   feature   of   the   antigen-­‐specific  adaptive  immune  response  in  both  mice  and  humans.  Antibody  stereotypy  may   147 have  clinical  and  therapeutic  implications  for  the  human  adaptive  immune  response  to  natural   pathogens.   For   example,   recent   studies   of   antibodies   produced   by   humans   in  response   to   the   2009   pandemic   H1N1   influenza   virus   infection   or   vaccination   have  revealed   an   abundance   of   mAbs   that   bind   to   a   conserved   “stem”   region   of   the   viral  hemagglutinin   (HA)   protein   and,   hence,   broadly   cross-­‐react   with   HA   proteins   from  multiple   influenza   sub-­‐types   (e.g.   H1,   H3,   H5).49,171,172   These   broadly   cross-­‐reacting  mAbs  are  reportedly  enriched  for  heavy  chains  encoded  by  the  VH1-­‐69  gene,  which  has  also  been  found  in  human  mAbs  against  other  viruses,  such  as  HIV  and  Hepatitis  C  virus  (HCV).169,172   It   remains   to  be  seen  whether  other  V  genes  and  heavy/light  chain  pairs  are  preferentially  enriched  in  broadly  cross-­‐reacting  anti-­‐influenza  mAbs.  Microfluidic  screening  of  mAbs  from  multiple  infected  humans  could  shed  light  on  whether  broadly  cross-­‐reacting  mAbs  are  produced  by  all  infected  humans.  If  not,  it  will  be  interesting  to  study   whether   individuals   with   antibody   repertoires   deviating   from   a   stereotyped  response   exhibit   different   clinical   outcomes   in   response   to   a   common   antigenic  challenge.    As  described   in  Chapter  5,  microfluidic   single-­‐cell   screening   can  be  adapted   to  the  selection  of  mAbs  based  on  other  functional  binding  properties  (e.g.  selectivity,  viral  neutralization,   etc.)   from   several   different   animal   species   (e.g.   rabbits,   humans).   This  chapter   will   also   describe   approaches   to   increase   the   number   of   ASCs   that   can   be  analyzed  in  a  single  microfluidic  device,  as  well  as  methods  to  increase  the  success  rate  of  single-­‐cell  RT-­‐PCR  reactions.   148 Chapter    5: Conclusions  and  Future  Work  The  previous  chapters  describe   the  development  of  a  microfluidic  platform   for  screening   and   selection   of   high   affinity   monoclonal   antibodies   (mAbs)   from   single  antibody-­‐secreting   cells   (ASCs).   This   technology   was   used   to   precisely   measure   the  binding   kinetics   of   antigen   and   antibodies   secreted   by   single   ASCs   (Chapter   2),   to  recover  and  amplify  heavy  and   light  chain  genes   from  single  ASCs  (Chapter  3),  and  to  screen   and   select   novel   high   affinity   antigen-­‐specific   mAbs   from   primary   ASCs  harvested   from   immunized   animals   (Chapter   4).     As   a   validation   of   the   technology,  nearly   200   high   affinity   mouse   mAbs   to   the   model   antigen   hen   egg   lysozyme,  representing   a   10-­‐fold   increase   in   the   number   of   high   affinity   anti-­‐HEL  mouse  mAbs  previously   generated   using   recent   micro-­‐well   technologies   and   conventional  hybridoma   methods.     The   microfluidic   mAb   selection   technology   has   also   provided  interesting   insights   into   affinity   maturation,   as   well   as   stereotypy   and   immuno-­‐dominance  in  the  adaptive  immune  system.    Specifically,  the  binding  constants  and  gene  sequences  of  anti-­‐HEL  mAbs  selected  by  microfluidic  screening  support  the  hypothesis  that   affinity  maturation   acts   to   increase   the   number   and   diversity   of   antigen-­‐specific  mAbs,   rather   than   increase   the   average   affinity   of   individual   antibodies.146   Anti-­‐HEL  mouse  mAbs  selected  using  both  microfluidic  and  hybridoma  methods  also  revealed  an  unexpected   level   of   convergent   evolution   across   different   mice,   consisting   of  preferential  V  gene  usage,  VDJ  recombination,  and  heavy-­‐light  chain  pairing.  The  mAbs  enriched   in   these   “stereotyped”   antibody   responses   are   generated   in   response   to  immunodominant  epitopes  in  the  HEL  protein.           149 This  chapter  focuses  on  future  extensions  of  the  microfluidic  single  cell  antibody  screening  technology  and  can  be  divided  into  three  sections:  selection  of  mAbs  based  on  multiple   different   functional   properties   (e.g.   antigen   binding   affinity,   kinetics,  selectivity,  viral  neutralization,  etc.),  increasing  the  capacity  to  screen  larger  numbers  of  single  ASCs,  and  selection  of  mAbs  from  different  animal  species  and  cell  types.     5.1 Selection  of  mAbs  for  Multiple  Functional  Binding  Properties    Although  the  present   thesis   focused  on  screening  mAbs  based  on   their  antigen  binding   kinetics   and   affinities,   antibodies   may   also   be   selected   for   other   important  functional   binding   properties,   such   as   specificity   or   cross-­‐reactivity   to   antigenic  variants,  binding  to  particular  antigenic  epitopes,  viral  neutralization,  and  inhibition  of  cell   growth  and   signaling.49,146,173,174  Antibodies  with   these  properties   can  be   selected  by  integrating  a  variety  of  different  high-­‐resolution  optical  assays  with  the  microfluidic  mAb  selection  technology.  For  example,  mAbs  that  bind  one  or  more  antigenic  variants  can  be   selected  by   flushing  bead-­‐captured  mAbs  with   antigenic   variants   each   labeled  with   a   spectrally   unique   fluorophore.   Subsequently,   microfluidic   chambers   can   be  fluorescence   imaged   with   multiple   filter   cubes   corresponding   to   each   labeled  fluorophore.   In   principle,   optical   multiplexing   can   be   performed   using   five   to   ten  distinct   fluorophores,   though   more   sophisticated   multiplexing   strategies   using   both  wavelength-­‐  and   intensity-­‐multiplexing  with  quantum  dots175  may   facilitate  screening  of  a  large  number  of  antigenic  variants  in  this  manner.      Selection   of   mAbs   that   bind   particular   epitopes   may   also   be   performed   by  competitive   binding   assays   using   cell-­‐secreted  mAbs   and   previously-­‐generated  mAbs   150 known  to  bind  particular  antigenic  epitopes.146,172,176  This  may  be  particularly  useful  for  biasing   antibody   selection   against   immunodominant   epitopes.   For   instance,   this  approach   may   facilitate   selection   of   mAbs   to   the   conserved   “stem”   region   of   the  influenza  hemagglutinin  (HA)  protein  rather  than  the  “head”  region  that  appears  to  be  the   dominant   target   of   antibodies   produced   in   response   to   seasonal   influenza  vaccines.171,177  This  approach  may  therefore  enable  the  selection  of  mAbs  that  broadly  cross-­‐react  with  different  influenza  viral  strains.49      Optical   assays   for   antibody   viral   neutralization   may   also   be   integrated   with  microfluidic   single   cell   antibody   screening.     For   example,   single   ASCs   can   be   pre-­‐incubated  with   pseudo-­‐typed   viruses   containing   fluorescent   transgene   reporters   (e.g.  GFP)   prior   to   mixing   of   virus   with   a   reporter   cell-­‐line.178,179   Fluorescence  measurements  of  the  reporter  cell-­‐line  can  then  be  used  to  screen  whether  cell-­‐secreted  mAbs  altered  viral   transduction  efficiencies.   Similarly,   cell-­‐secreted  mAbs  can  be  pre-­‐incubated  with  influenza  viruses  prior  to  mixing  with  red  blood  cells  in  order  to  select  mAbs   that   inhibit   bind   and   inhibit   hemagglutination   (i.e.   the   hemagglutination  inhibition  assay).172,180    A   variety   of   optical   cell   growth   and   signaling   assays   can   be   used   to   identify  therapeutic   mAbs   for   the   treatment   of   different   cancers   (breast,   colorectal)   and  immunological  disorders   (rheumatoid  arthritis,   lupus).41  For   instance,   the   therapeutic  antibody  Herceptin   is   known   to   inhibit   in   vitro  proliferation   of   human   breast   cancer  cells.174  Thus,   antibody-­‐secreting  cells   can  be  co-­‐incubated  with  breast   cancer   cells   in  microfluidic   chambers   in   order   to   select   mAbs   that   similarly   inhibit   cancer   cell  proliferation.  Humira,   a   clinically   approved  antibody   for   the   treatment  of   rheumatoid   151 arthritis,   is   known   to   target   the   inflammatory   cytokine   tumor   necrosis   factor   alpha  (TNF-­‐α).  Tay  et  al.  recently  used  a  microfluidic  fluorescence  imaging  assay  to  study  the  effects   of   in   vitro   TNF-­‐α   stimulation   in   an   engineered   mouse   fibroblast   (3T3)   cell-­‐line.181   By   extension,   co-­‐incubation   of   ASCs  with   this   reporter   cell-­‐line  may   facilitate  selection   of   mAbs   that   inhibit   TNF-­‐α   cell   signaling.   Screening   of   mAbs   based   on  aggregation   properties   and   solution   stability   may   also   be   facilitated   by   microfluidic  methods.     5.2 Increasing  Capacity  to  Screen  Larger  Numbers  of  ASCs  By  conventional  wisdom,  high  affinity  antibodies  are  rare,  suggesting  that  higher  quality  antigen-­‐specific  mAbs  will  identified  by  screening  more  single  cells.  Despite  only  screening  ~0.1%  of  the  total  CD138+  splenocytes  harvested  from  HEL-­‐immunized  mice  (i.e.  ~100  out  of  over  100,000  cells  per  spleen),  most  of  the  characterized  mAbs  (>80%)  had  Kd  values  less  than  or  equal  to  1  nM,  with  the  highest  affinities  reaching  ~100  pM  (Chapter   4,   Table   4.2).   However,   selection   of   rare   mAbs   with   exceptionally   low  dissociation   rates   (koff   <  10-­‐4   s-­‐1)   and  equilibrium  dissociation  constants   less   than   the  100   pM   affinity   ceiling   may   require   improvements   to   the   capacity   of   the   of   the  microfluidic   system   to   screen   larger   number   of   ASCs.   Indeed,   using   a   combination   of  ELISA  analysis  on  binned  cell  populations  and   the  SLAM  approach   (Chapter  1,  Figure  1.8),  Babcook  and  coworkers  screened  a  large  number  of  B  cells  to  identify  mAbs  that  bound  antigen  (i.e.  human  IL-­‐8)  with  picomolar  affinities.50,182  The  current  microfluidic  architecture  is  limited  to  analyzing  up  to  1000  cells  per  device,   due   to   practical   limitations   imposed   by   individual   addressability   in   two-­‐ 152 dimensional   microfluidic   chamber   arrays.87   Further   increases   in   cell   capacity   can   be  obtained  by  integrating  the  developed  microfluidic  technology  with  upstream  assays  of  binned   cell   populations.   For   example,   up   to   100   ASCs   can   be   FACS-­‐sorted   into  individual  wells  and   the  cell   supernatant   from  each  well   can  be  screened   for  antigen-­‐specific   mAbs   using   enzyme-­‐linked   immunosorbent   assays   (ELISA).   Selected   wells  containing   antigen-­‐specific   antibodies   can  be   subsequently   screened  at   the   single-­‐cell  level   using   the   developed   microfluidic   technology.   In   this   manner,   hundreds   of  thousands   of   ASCs   can   be   screened   using   only   ten   96-­‐   or   384-­‐well   plates   and   an  optional  number  of  microfluidic  devices  depending  upon  the  number  of  selected  wells.  Alternatively,   larger   number   of  ASCs  may  be   screened  by   integrating  highly   sensitive  bead-­‐based   detection  with   previously   reported  micro-­‐well   or   droplet-­‐based  methods  capable   of   encapsulating   tens   to   hundreds   of   thousands   of   cells,   thus   eliminating   the  need   for   addressable   microfluidic   chamber   arrays.14,51,91,92   One   embodiment   of   this  approach  involves  the  co-­‐encapsulation  of  single  ASCs  with  multi-­‐functionalized  beads  capable   of   capturing   both   secreted   antibodies   and   antibody-­‐encoding   mRNA   (Figure  5.1).  By  co-­‐incubating  cells  and  beads  in  droplets  and  subsequent  cell  lysis  by  merging  droplets   containing   mild   detergent   (e.g.   0.5%   NP-­‐40),   both   secreted   mAbs   and  antibody-­‐encoding   mRNA   from   single   ASCs   can   be   captured   on   the   bead   surface.  Subsequent  recovery  and  FACS-­‐sorting  of  beads  from  the  droplet  emulsion  can  be  used  to   enrich   beads   containing   antigen-­‐specific   mAbs.   Finally,   RT-­‐PCR   amplification   of  sorted   beads   can   be   used   to   amplify   the   heavy   and   light   chain   genes   for   subsequent  cloning  and  expression  of  selected  mAbs.   153 Importantly,   the   throughput   of   all   single   cell   antibody-­‐screening   methods,  including   the   described   microfluidic   technology,   is   constrained   by   low   amplification  efficiencies   of   heavy   and   light   chain   genes   from   single   cells.   RT-­‐PCR   amplification  successfully   identified  approximately  20%  of  paired  heavy  and   light  chain  genes   from  selected  ASCs  both  in  this  work  (see  Chapter  4,  section  4.2.4)  and  other  state-­‐of-­‐the-­‐art  studies.58   Improved   primer   design   may   lead   to   significantly   higher   amplification  efficiencies   at   low   template   concentrations,   as   found   when   comparing   two   different  primer   design   strategies   (Chapter   3).   Alternative   amplification   strategies,   such   as   5’-­‐RACE   and   single-­‐cell   whole   transcriptome   amplification   (WTA),   obviate   the   need   for  design  of  primers   to   the  antibody  variable   region,  but   it   is  presently  unclear  whether  these   methods   will   result   in   higher   success   rates   than   single-­‐cell   RT-­‐PCR  amplification.183,184  Recent  advances  in  microfluidic  technologies  have  facilitated  robust  RT-­‐PCR   amplification   of  mRNA   transcripts   from   single   cells  with   sensitivity   down   to  single  molecules.75,81  Thus,  integration  of  single-­‐cell  RT-­‐PCR  amplification  and  recovery  into   microfluidic   devices   may   yield   significantly   greater   numbers   of   heavy   and   light  chain   genes   from   selected   ASCs.79 154   Figure  5.1           Bi-­‐functionalized  beads  for  the  simultaneous  capture  of  mAbs  and  antibody-­‐encoding  mRNA  from  single  cells.    (Top)  Scheme   for   chemical   conjugation   of   secondary   mAbs   and   oigo(dT)     to   beads   using   carbodiimide   chemistry.   (Bottom)   Microscope   images   of   bi-­‐ functionalized  beads  trapped  by  a  microfluidic  sieve  valve  (A).  Captured  on  the  bead  surface  are  fluorescently  labeled  synthetic  DNA  (B)  and   fluorescently   labeled   mouse   mAbs   (C).   (Bottom)   Measurement   of   binding   kinetics   of   antigen   and   single   cell-­‐secreted   antibodies   on   bi-­‐ functionalized  beads  as  described  in  Chapter  2.  Figure  adapted  from  US  Patent  Application  2012/0015347  A1.185  [continued  on  next  page]       ! 155   Figure  5.1     [continued   from   previous   page]   Microscope   image   of   hybridoma   cell   adjacent   to   antibody   capture   beads   in   microfluidic   device  (left)  and  measurement  of  binding  kinetics  of  antigen  and  single  cell-­‐secreted  antibodies  on  bi-­‐functionalized  beads  as  described  in   Chapter   2   (right).   Figure   adapted   from   US   Patent   Application   2012/0015347   A1.185 156 5.3 Selection   of   mAbs   from   Other   Animal   Species   (e.g.   Humans,   Rabbits,   etc.)   and  Cell  Types  As   microfluidic   single   cell   antibody   screening   circumvents   the   need   for  hybridoma   generation   by   fusion   with   a   partner   cancer   cell-­‐line,   the   technology   can  select  mAbs   from  many  different  species,   including  humans  and  rabbits.  Human  ASCs  can   be   readily   purified   based   on   CD19   and   CD38   cell-­‐surface   markers,   and   several  different   primer   designs   have   been   previously   reported   for   single-­‐cell   RT-­‐PCR  amplification  of  both  human  and  rabbit  heavy  and  light  chains  genes.19,23,50,172  Thus,  in  addition   to   facilitating   selection   of   research-­‐grade   mouse   mAbs,   the   microfluidic  technology  can  be  used  to  select  fully  human  therapeutic  mAbs  that  bind  target  antigen  with  high  affinity.      Extension   of   the  microfluidic   technology   for  mAb   selection   from   species   other  than   mice   and   humans   will   require   advances   in   the   identification   of   cell-­‐surface  markers   to   purify   and   enrich  ASCs   from   immunized   animals.31,50   As   described   above,  the  current  microfluidic  architecture  is  limited  to  screening  several  hundred  cells  from  a   single   animal;   thus,   successful   mAb   selection   using   this   approach   requires   that  antigen-­‐specific   ASCs   comprise   >1%  of   the   purified   cell   populations.   Increases   in   the  capacity  to  screen  larger  numbers  of  ASCs  screened  will  reduce  the  necessary   level  of  enrichment  for  screening  ASCs  from  all  species,  as  well  as  facilitate  the  selection  of  rare  mAbs.   Interestingly,   even   though   up   to   70%   of   enriched   human   ASCs   secrete   anti-­‐influenza  mAbs,   Corti   et   al.   screened   100,000   plasma   cells   in   order   to   select   a   single  mAb  that  cross-­‐reacts  with  all  influenza  strains.23,49   157 In   addition   to   screening   antibody-­‐secreting   cells   (ASCs)   harvested   from  immunized   animals,   the   microfluidic   technology   can   be   extended   to   selection   of  antibodies   from  memory  B   cells   and   sub-­‐cloning  of  hybridoma  and   recombinant   cell-­‐lines,  such  as  Chinese  hamster  ovary  (CHO)  cells  and  human  embryonic  kidney  (HEK)  cells   transfected   with   expression   vectors   (e.g.   phage,   viruses)   for   production   of   full-­‐length   antibodies   (e.g.   IgG)   or   antibody   fragments   (e.g.   scFv,   Fab,   Fv,   bispecific  antibodies,   minibodies,   antibody-­‐drug   conjugates   ADCs,   etc.).28,45,50,186,187   Memory   B  cells  can  be  induced  in  vitro  to  secrete  antibodies  by  immortalization  with  Epstein-­‐Barr  virus   (EBV)   and/or   stimulation   with   toll-­‐like   receptor   (TLR)   agonists   (i.e.   CpG  oligonucleotides).     The   induced   memory   B   cells   can   thus   be   screened   in   a   manner  analogous  to  antibody-­‐secreting  cells.      In  contrast  to  selecting  cells  secreting  antibodies  with  desired  antigen-­‐binding  kinetics  or   specificity,   time-­‐course  measurements  of   the  amount  of  bead-­‐captured  antibodies  (see  Figure  2.12)  can  be  used  to  select  clones  that  produce   large   amounts   of   antibody   for   recombinant   production.     Finally,   the  microfluidic  technology  can  be  used  to  select  recombinant  cells  and  other  cell  types  (e.g.  T-­‐cells)   producing   bio-­‐molecules,   such   as   T-­‐cell   receptors   (TCRs),   cytokines,  recombinant  proteins,  carbohydrates,  lipids,  and  small  molecules.       5.4 Other  Insights  into  the  Adaptive  Immune  System     Application   of   the   developed   microfluidic   technology   for   selecting   anti-­‐HEL  mAbs   yielded   interesting   insights   into   “convergent   evolution”   of   antibodies   in   mice;  that   is,   anti-­‐HEL   mAbs   are   preferentially   encoded   by   particular   V   genes,   VDJ  recombination,  and  heavy-­‐light  chain  pairing  in  response  to  immunodominant  epitopes   158 on  the  HEL  protein.  Future  extensions  of  this  approach  may  assist  in  vaccine  design  by  assigning   enriched   antibody   populations   to   immunodominant   epitopes   in   common  human  pathogens,   such  as  human   immunodeficiency  virus   (HIV),   influenza  virus,   and  human   cytomegalovirus   (CMV).169,170,172  Moreover,  microfluidic   selection   can   provide  both  functional  and  sequence  information  for  antigen-­‐specific  mAbs  that  can  be  used  to  query  fully-­‐sequenced  antibody  repertoires  in  order  to  obtain  more  global  insights  into  antigen-­‐specific  antibody  “stereotypy”.18,162  This  approach  may  facilitate  the  diagnostic  and/or   prognostic   identification   of   antibodies   that   are   over-­‐   or   under-­‐expressed   in  particular  states  of  human  health,  including  bacterial  and  viral  infections,  autoimmune  disorders   (e.g.   systemic   lupus   erythematosus,   rheumatoid   arthritis,   cystic   fibrosis,  multiple   sclerosis,   Crohn’s   disease,   etc.),   allergic   responses,   osteoporosis,   cancer   (e.g.  breast,   colorectal,   lymphoma,   leukemia,   melanoma,   etc.),   neurodegenerative   diseases  (e.g.  Alzheimer’s  disease,  prion  diseases),  and  other  diseases.           159 References  1.   Da   Costa,   D.   J.  Microfluidic   Technology   for   Screening   and   Selection   of   Monoclonal   Antibodies  from  Single  Cells.  43  (University  of  British  Columbia:  2011).  2.   Murphy,  K.  M.,  Travers,  P.  &  Walport,  M.  Janeway’s  Immunobiology.  (Garland  Science  /  Taylor  &  Francis  LLC:  2011).  3.   Kantor,  A.  &  Herzenberg,  L.  Origin  of  Murine  B-­‐Cell  Lineages.  Annu.  Rev.   Immunol.   11,  501–538  (1993).  4.   Ackermann,   B.   L.   &   Berna,   M.   J.   Coupling   immunoaffinity   techniques  with  MS   for  quantitative   analysis   of       low-­‐abundance   protein   biomarkers.   Expert   Rev.   Proteomics  4,  175–186  (2007).  5.   Bradbury,   A.   et   al.   Antibodies   in   proteomics   II:   screening,   high-­‐throughput  characterization       and   downstream   applications.  Trends   Biotechnol.  21,   312–317  (2003).  6.   Foon,  K.  &  Todd,  R.  Immunological  Classification  of  Leukemia  and  Lymphoma.  Blood   68,  1–31  (1986).  7.   Picker,   L.   J.,   Weiss,   L.   M.,   Medeiros,   L.   J.,   Wood,   G.   S.   &   Warnke,   R.   A.  Immunophenotypic   criteria   for   the   diagnosis   of   non-­‐Hodgkin’s   lymphoma.   Am   J   Pathol  128,  181–201  (1987).  8.   Uhlen,  M.  Affinity  as  a  tool  in  life  science.  Biotechniques  44,  649–654  (2008).  9.   Martin-­‐Moe,   S.   et   al.   The   Structure   of   Biological   Therapeutics.   Formulation   and   Process   Development   Strategies   for   Manufacturing   Biopharmaceuticals   1–40  (2010).at    10.  Zider,  A.  &  Drakeman,  D.  L.  The  future  of  monoclonal  antibody  technology.  MAbs  2,  361–364  (2010).  11.  Reichert,  J.  M.  Monoclonal  antibodies  in  the  clinic.  Nat  Biotech  19,  819–822  (2001).  12.  Carter,   P.   J.   Potent   antibody   therapeutics   by  design.  Nat  Rev   Immunol  6,   343–357  (2006).  13.  Little,  M.  Antibody  Libraries:  How  to  Overcome  the  Discovery  Log-­‐Jam.  2006  21,  54–58  14.  Love,   J.   C.,   Ronan,   J.   L.,   Grotenbreg,   G.   M.,   van   der   Veen,   A.   G.   &   Ploegh,   H.   L.   A  microengraving   method   for   rapid   selection   of   single   cells   producing   antigen-­‐specific  antibodies.  Nat  Biotech  24,  703–707  (2006).  15.  Love,   J.   C.   Making   antibodies   from   scratch.  Nature   Biotechnology   28,   1176–1178  (2010).  16.  Braden,  B.  C.  &  Poljak,  R.  J.  Structural  features  of  the  reactions  between  antibodies  and  protein  antigens.  FASEB  J.  9,  9–16  (1995).  17.  Ganusov,  V.  V.  &  Boer,  R.   J.  D.  Do  most   lymphocytes  in  humans  really  reside  in  the  gut?  Trends  in  Immunology  28,  514–518  (2007).  18.  Arnaout,  R.  et  al.  High-­‐Resolution  Description  of  Antibody  Heavy-­‐Chain  Repertoires  in  Humans.  PLoS  ONE  6,  e22365  (2011).  19.  Tiller,  T.  et  al.   Efficient  generation  of  monoclonal   antibodies   from  single  human  B  cells  by  single  cell  RT-­‐PCR  and  expression  vector  cloning.  J.  Immunol.  Methods  329,  112–124  (2008).   160 20.  Foote,  J.  &  Milstein,  C.  Kinetic  maturation  of  an  immune  response.  ,  Published  online:   08  August  1991;  |  doi:10.1038/352530a0  352,  530–532  (1991).  21.  Niels   Jerne   The   Generative   Grammar   of   the   Immune   System.   at    22.  Fooksman,   D.   R.   et   al.   Development   and   Migration   of   Plasma   Cells   in   the   Mouse  Lymph  Node.  Immunity  33,  118–127  (2010).  23.  Wrammert,   J.   et   al.   Rapid   cloning   of   high-­‐affinity   human   monoclonal   antibodies  against  influenza  virus.  Nature  453,  667–671  (2008).  24.  McHeyzer-­‐Williams,  L.  J.  &  McHeyzer-­‐Williams,  M.  G.  Antigen-­‐Specific  Memory  B  Cell  Development.  Annual  Review  of  Immunology  23,  487–513  (2005).  25.  von  Behring,  E.  &  Kitasato,  S.  [The  mechanism  of  diphtheria  immunity  and  tetanus  immunity  in  animals.  1890].  Mol.  Immunol.  28,  1317,  1319–1320  (1991).  26.  Ehrlich,   P.   Partial   Cell   Functions.   Scandinavian   Journal   of   Immunology   31,   4–13  (1990).  27.  Pier,  G.  B.  &  Lyczak,  J.  B.  Immunology,  Infection,  and  Immunity.  (ASM  Press:  2004).  28.  KOHLER,  G.  &  MILSTEIN,  C.  Continuous  cultures  of  fused  cells  secreting  antibody  of  predefined  specificity.  Nature  256,  495–497  (1975).  29.  Davis,   J.  M.,  Pennington,   J.  E.,  Kubler,  A.-­‐M.  &  Conscience,   J.-­‐F.  A  simple,  single-­‐step  technique  for  selecting  and  cloning  hybridomas  for  the  production  of  monoclonal  antibodies.  Journal  of  Immunological  Methods  50,  161–171  (1982).  30.  Joyce,   J.   G.   &  Meulen,   J.   ter   Pushing   the   envelope   on  HIV-­‐1   neutralization.  Nature   Biotechnology  28,  929–931  (2010).  31.  Spieker-­‐Polet,   H.,   Sethupathi,   P.,   Yam,   P.   C.   &   Knight,   K.   L.   Rabbit   monoclonal  antibodies:   generating   a   fusion   partner   to   produce   rabbit-­‐rabbit   hybridomas.   Proceedings  of  the  National  Academy  of  Sciences  of  the  United  States  of  America  92,  9348–9352  (1995).  32.  Jones,   P.   T.,   Dear,   P.   H.,   Foote,   J.,   Neuberger,   M.   S.   &   Winter,   G.   Replacing   the  complementarity-­‐determining   regions   in   a   human   antibody   with   those   from   a  mouse.  321,  522–525  (1986).  33.  Riechmann,   L.,   Clark,  M.,  Waldmann,  H.  &  Winter,  G.  Reshaping  human  antibodies  for  therapy.  332,  323–327  (1988).  34.  Winter,  G.  &  Milstein,  C.  Man-­‐made  antibodies.  Nature  349,  293–299  (1991).  35.  Beck,   A.,  Wurch,   T.,   Bailly,   C.   &   Corvaia,   N.   Strategies   and   challenges   for   the   next  generation   of   therapeutic   antibodies.   Nature   Reviews   Immunology   10,   345–352  (2010).  36.  Karpas,   A.,   Dremucheva,   A.   &   Czepulkowski,   B.   H.   A   human   myeloma   cell   line  suitable   for   the   generation   of   human   monoclonal   antibodies.   Proceedings   of   the   National   Academy   of   Sciences   of   the   United   States   of   America   98,   1799–1804  (2001).  37.  Green,   L.   L.   et   al.   Antigen–specific   human   monoclonal   antibodies   from   mice  engineered  with  human   Ig  heavy  and   light   chain  YACs.  Nature  Genetics  7,   13–21  (1994).  38.  Lonberg,   N.   et   al.   Antigen-­‐specific   human   antibodies   from   mice   comprising   four  distinct  genetic  modifications.  368,  856–859  (1994).   161 39.  Kellermann,   S.-­‐A.  &  Green,   L.   L.  Antibody  discovery:   the  use  of   transgenic  mice   to  generate   human   monoclonal   antibodies   for   therapeutics.   Current   Opinion   in   Biotechnology  13,  593–597  (2002).  40.  Lonberg,   N.   Fully   human   antibodies   from   transgenic   mouse   and   phage   display  platforms.  Current  Opinion  in  Immunology  20,  450–459  (2008).  41.  Nelson,   A.   L.,   Dhimolea,   E.   &   Reichert,   J.   M.   Development   trends   for   human  monoclonal  antibody  therapeutics.  Nat  Rev  Drug  Discov  9,  767–774  (2010).  42.  L’Haridon,   R.   M.,   Bourget,   P.,   Lefevre,   F.   &   La   Bonnardiere,   C.   Production   of   an  Hybridoma  Library   to  Recombinant   Porcine  Alpha   I   Interferon:  A  Very   Sensitive  Assay  (ISBBA)  Allows  the  Detection  of  a  Large  Number  of  Clones.  Hybridoma  10,  35–47  (1991).  43.  Sogut,   I.,   Hatipoglu,   I.,   Kanbak,   G.   &   Basalp,   A.  Monoclonal   Antibodies   Specific   for  Hepatitis   B   e   Antigen   and   Hepatitis   B   Core   Antigen.   Hybridoma   30,   475–479  (2011).  44.  Pasqualini,   R.   &   Arap,   W.   Hybridoma-­‐free   generation   of   monoclonal   antibodies.   Proceedings   of   the   National   Academy   of   Sciences   of   the   United   States   of   America   101,  257–259  (2004).  45.  Lanzavecchia,   A.,   Corti,   D.   &   Sallusto,   F.   Human   monoclonal   antibodies   by  immortalization  of  memory  B  cells.  Current  Opinion  in  Biotechnology  18,  523–528  (2007).  46.  Hoogenboom,   H.   R.   Selecting   and   screening   recombinant   antibody   libraries.   Nat   Biotech  23,  1105–1116  (2005).  47.  Harding,   F.   A.,   Stickler,   M.   M.,   Razo,   J.   &   DuBridge,   R.   B.   The   immunogenicity   of  humanized  and  fully  human  antibodies.  MAbs  2,  256–265  (2010).  48.  Bartelds,  G.  M.  et  al.  Development  of  Antidrug  Antibodies  Against  Adalimumab  and  Association   With   Disease   Activity   and   Treatment   Failure   During   Long-­‐term  Follow-­‐up.  JAMA:  The  Journal  of  the  American  Medical  Association  305,  1460–1468  (2011).  49.  Corti,   D.   et   al.   A   Neutralizing   Antibody   Selected   from   Plasma   Cells   That   Binds   to  Group  1  and  Group  2  Influenza  A  Hemagglutinins.  Science  333,  850–856  (2011).  50.  Babcook,   J.   S.,   Leslie,   K.   B.,   Olsen,   O.   A.,   Salmon,   R.   A.   &   Schrader,   J.   W.   A   novel  strategy   for   generating  monoclonal   antibodies   from   single,   isolated   lymphocytes  producing  antibodies  of  defined  specificities.  Proceedings  of  the  National  Academy   of  Sciences  of  the  United  States  of  America  93,  7843–7848  (1996).  51.  Jin,   A.   et   al.   A   rapid   and   efficient   single-­‐cell   manipulation   method   for   screening  antigen-­‐specific   antibody-­‐secreting   cells   from   human   peripheral   blood.  Nat   Med   15,  1088–1092  (2009).  52.  Owens,  G.  P.  et  al.  Single-­‐cell  repertoire  analysis  demonstrates  that  clonal  expansion  is   a   prominent   feature   of   the   B   cell   response   in  multiple   sclerosis   cerebrospinal  fluid.  J.  Immunol  171,  2725–2733  (2003).  53.  Wang,  X.  &  Stollar,  B.  D.  Human   immunoglobulin  variable   region  gene  analysis  by  single  cell  RT-­‐PCR.  J.  Immunol.  Methods  244,  217–225  (2000).  54.  Sehgal,  D.,  Schiaffella,  E.,  Anderson,  A.  O.  &  Mage,  R.  G.  Analyses  of  single  B  cells  by  polymerase   chain   reaction   reveal   rearranged   VH   with   germline   sequences   in  spleens  of  immunized  adult  rabbits:  implications  for  B  cell  repertoire  maintenance  and  renewal.  J.  Immunol.  161,  5347–5356  (1998).   162 55.  Ali,   M.,   Hitomi,   K.   &   Nakano,   H.   Generation   of   monoclonal   antibodies   using  simplified  single-­‐cell  reverse  transcription-­‐polymerase  chain  reaction  and  cell-­‐free  protein  synthesis.  J.  Biosci.  Bioeng.  101,  284–286  (2006).  56.  Küppers,  R.,  Zhao,  M.,  Hansmann,  M.  L.  &  Rajewsky,  K.  Tracing  B  cell  development  in  human   germinal   centres   by   molecular   analysis   of   single   cells   picked   from  histological  sections.  EMBO  J  12,  4955–4967  (1993).  57.  Poulsen,  T.  R.,  Jensen,  A.,  Haurum,  J.  S.  &  Andersen,  P.  S.  Limits  for  Antibody  Affinity  Maturation  and  Repertoire  Diversification  in  Hypervaccinated  Humans.  J  Immunol   187,  4229–4235  (2011).  58.  Tiller,  T.,  Busse,  C.  E.  &  Wardemann,  H.  Cloning  and  expression  of  murine  Ig  genes  from  single  B  cells.  Journal  of  Immunological  Methods  350,  183–193  (2009).  59.  Poulsen,  T.  R.,  Meijer,  P.-­‐J.,  Jensen,  A.,  Nielsen,  L.  S.  &  Andersen,  P.  S.  Kinetic,  Affinity,  and   Diversity   Limits   of   Human   Polyclonal   Antibody   Responses   against   Tetanus  Toxoid.  J  Immunol  179,  3841–3850  (2007).  60.  Smith,  K.  et  al.  Human  monoclonal  antibodies  generated  following  vaccination  with  AVA   provide   neutralization   by   blocking   furin   cleavage   but   not   by   preventing  oligomerization.  Vaccine  30,  4276–4283  (2012).  61.  Wrammert,  J.  et  al.  Rapid  and  Massive  Virus-­‐Specific  Plasmablast  Responses  during  Acute  Dengue  Virus  Infection  in  Humans.  J.  Virol.  86,  2911–2918  (2012).  62.  Niro,   R.   D.   et   al.   Rapid   Generation   of   Rotavirus-­‐Specific   Human   Monoclonal  Antibodies  from  Small-­‐Intestinal  Mucosa.  J  Immunol  185,  5377–5383  (2010).  63.  Whitesides,  G.  M.  The  origins  and  the  future  of  microfluidics.  Nature  442,  368–373  (2006).  64.  Manz,   A.   et   al.   Planar   chips   technology   for   miniaturization   and   integration   of  separation   techniques   into   monitoring   systems:   Capillary   electrophoresis   on   a  chip.  Journal  of  Chromatography  A  593,  253–258  (1992).  65.    Agilent   |  Agilent  Technologies’  Microfluidics  System  Used   in  Breakthrough  Breast  Milk   Study.   (2010).at    66.  Jason-­‐Moller,   L.,   Murphy,   M.   &   Bruno,   J.   Overview   of   Biacore   Systems   and   Their  Applications.   Current   Protocols   in   Protein   Science   (2001).at    67.  Metzker,  M.   L.   Sequencing   technologies   -­‐   the   next   generation.  Nat.   Rev.   Genet.  11,  31–46  (2010).  68.  Maerkl,   S.   J.  &  Quake,   S.   R.  A   Systems  Approach   to  Measuring   the  Binding  Energy  Landscapes  of  Transcription  Factors.  Science  315,  233–237  (2007).  69.  Maerkl,   S.   J.   Next   generation   microfluidic   platforms   for   high-­‐throughput   protein  biochemistry.  Current  Opinion  in  Biotechnology  22,  59–65  (2011).  70.  Einav,  S.  et  al.  Discovery  of  a  hepatitis  C  target  and  its  pharmacological  inhibitors  by  microfluidic  affinity  analysis.  Nat  Biotech  26,  1019–1027  (2008).  71.  Lee,   C.-­‐C.   et   al.   Multistep   Synthesis   of   a   Radiolabeled   Imaging   Probe   Using  Integrated  Microfluidics.  Science  310,  1793–1796  (2005).  72.  Anderson,   M.   J.,   Hansen,   C.   L.   &   Quake,   S.   R.   Phase   knowledge   enables   rational  screens  for  protein  crystallization.  Proceedings  of  the  National  Academy  of  Sciences   103,  16746–16751  (2006).   163 73.  Hansen,   C.   L.,   Skordalakes,   E.,   Berger,   J.   M.   &   Quake,   S.   R.   A   robust   and   scalable  microfluidic  metering  method  that  allows  protein  crystal  growth  by  free  interface  diffusion.  Proceedings   of   the  National   Academy   of   Sciences   of   the  United   States   of   America  99,  16531–16536  (2002).  74.  Hansen,   C.   L.,   Sommer,  M.  O.  A.  &  Quake,   S.   R.   Systematic   investigation   of   protein  phase   behavior   with   a   microfluidic   formulator.   Proceedings   of   the   National   Academy  of  Sciences  of  the  United  States  of  America  101,  14431–14436  (2004).  75.  Heyries,  K.  A.  et  al.  Megapixel  digital  PCR.  Nature  Methods  8,  649–651  (2011).  76.  Kartalov,   E.   P.   et   al.   High-­‐throughput   multi-­‐antigen   microfluidic   fluorescence  immunoassays.  BioTechniques  40,  85–90  (2006).  77.  Diercks,  A.  H.  et  al.  A  microfluidic  device  for  multiplexed  protein  detection  in  nano-­‐liter  volumes.  Anal.  Biochem  386,  30–35  (2009).  78.  Falconnet,  D.  et  al.  High-­‐throughput   tracking  of   single  yeast   cells   in  a  microfluidic  imaging  matrix.  Lab  Chip  11,  466–473  (2011).  79.  Leung,   K.   et   al.   A   programmable   droplet-­‐based   microfluidic   device   applied   to  multiparameter   analysis   of   single   microbes   and   microbial   communities.   PNAS  (2012).doi:10.1073/pnas.1106752109  80.  Lecault,   V.   et   al.   High-­‐throughput   analysis   of   single   hematopoietic   stem   cell  proliferation  in  microfluidic  cell  culture  arrays.  Nat  Meth  8,  581–586  (2011).  81.  White,   A.   K.   et   al.   High-­‐throughput   microfluidic   single-­‐cell   RT-­‐qPCR.   PNAS  (2011).doi:10.1073/pnas.1019446108  82.  Lecault,  V.,  White,  A.  K.,  Singhal,  A.  &  Hansen,  C.  L.  Microfluidic  single  cell  analysis:  from   promise   to   practice.   Current   Opinion   in   Chemical   Biology  doi:10.1016/j.cbpa.2012.03.022  83.  Marcus,  J.  S.,  Anderson,  W.  F.  &  Quake,  S.  R.  Microfluidic  Single-­‐Cell  mRNA  Isolation  and  Analysis.  Analytical  Chemistry  78,  3084–3089  (2006).  84.  Duffy,  D.  C.,  McDonald,  J.  C.,  Schueller,  O.  J.  &  Whitesides,  G.  M.  Rapid  Prototyping  of  Microfluidic   Systems   in   Poly(dimethylsiloxane).   Anal.   Chem.   70,   4974–4984  (1998).  85.  McDonald,  J.  C.  et  al.  Fabrication  of  microfluidic  systems  in  poly(dimethylsiloxane).   Electrophoresis  21,  27–40  (2000).  86.  Unger,   M.   A.,   Chou,   H.-­‐P.,   Thorsen,   T.,   Scherer,   A.   &   Quake,   S.   R.   Monolithic  Microfabricated   Valves   and   Pumps   by   Multilayer   Soft   Lithography.   Science   288,  113–116  (2000).  87.  Thorsen,  T.,  Maerkl,  S.  J.  &  Quake,  S.  R.  Microfluidic  Large-­‐Scale  Integration.  Science   298,  580–584  (2002).  88.  NOSSAL,   G.   J.   &   LEDERBERG,   J.   Antibody   production   by   single   cells.  Nature   181,  1419–1420  (1958).  89.  Viret,  C.  &  Gurr,  W.  The  Origin  of  the  ‘One  Cell-­‐One  Antibody’  Rule.  J  Immunol  182,  1229–1230  (2009).  90.  Story,   C.   M.   et   al.   Profiling   antibody   responses   by   multiparametric   analysis   of  primary   B   cells.   Proceedings   of   the   National   Academy   of   Sciences   105,   17902–17907  (2008).  91.  Köster,  S.  et  al.  Drop-­‐based  microfluidic  devices  for  encapsulation  of  single  cells.  Lab   on  a  Chip  8,  1110  (2008).   164 92.  Debs,  B.  E.,  Utharala,  R.,  Balyasnikova,  I.  V.,  Griffiths,  A.  D.  &  Merten,  C.  A.  Functional  single-­‐cell   hybridoma   screening   using   droplet-­‐based   microfluidics.   PNAS  (2012).doi:10.1073/pnas.1204514109  93.  Foote,   J.   &   Eisen,   H.   N.   Kinetic   and   affinity   limits   on   antibodies   produced   during  immune  responses.  Proceedings  of   the  National  Academy  of  Sciences  of   the  United   States  of  America  92,  1254–1256  (1995).  94.  Trisler,  K.  et  al.  A  metalloantibody  that   irreversibly  binds  a  protein  antigen.   J.  Biol.   Chem.  282,  26344–26353  (2007).  95.  Lauffenburger,  D.  A.  &  Linderman,  J.  Receptors:  Models   for  Binding,  Trafficking,  and   Signaling.  (Oxford  University  Press,  USA:  1996).  96.  Northrup,  S.  H.  &  Erickson,  H.  P.  Kinetics  of  protein-­‐protein  association  explained  by  Brownian  dynamics  computer  simulation.  Proceedings  of  the  National  Academy  of   Sciences  of  the  United  States  of  America  89,  3338–3342  (1992).  97.  Shick,  K.  A.,  Xavier,  K.  A.,  Rajpal,  A.,  Smith-­‐Gill,  S.  J.  &  Willson,  R.  C.  Association  of  the  anti-­‐hen   egg   lysozyme  antibody  HyHEL-­‐5  with   avian   species   variant   and  mutant  lysozymes.  Biochimica  et  Biophysica  Acta   (BBA)   -­‐  Protein  Structure  and  Molecular   Enzymology  1340,  205–214  (1997).  98.  Edwards,  B.  M.  et  al.  The  Remarkable  Flexibility  of  the  Human  Antibody  Repertoire;  Isolation   of   Over   One   Thousand   Different   Antibodies   to   a   Single   Protein,   BLyS.   Journal  of  Molecular  Biology  334,  103–118  (2003).  99.  Sampson,   J.   H.   et   al.   Unarmed,   Tumor-­‐Specific   Monoclonal   Antibody   Effectively  Treats  Brain  Tumors.  PNAS  97,  7503–7508  (2000).  100.  Roost,  H.  P.  et  al.  Early  high-­‐affinity  neutralizing  anti-­‐viral   IgG  responses  without  further  overall   improvements  of  affinity.  Proc  Natl  Acad  Sci  U  S  A  92,  1257–1261  (1995).  101.  Klasse,   P.   J.   &   Sattentau,   Q.   J.   Occupancy   and   Mechanism   in   Antibody-­‐Mediated  Neutralization  of  Animal  Viruses.  J  Gen  Virol  83,  2091–2108  (2002).  102.  Boder,   E.   T.,   Midelfort,   K.   S.   &   Wittrup,   K.   D.   Directed   evolution   of   antibody  fragments   with   monovalent   femtomolar   antigen-­‐binding   affinity.   Proceedings   of   the  National  Academy  of  Sciences  of  the  United  States  of  America  97,  10701–10705  (2000).  103.  Rich,   R.   L.   &   Myszka,   D.   G.   Higher-­‐throughput,   label-­‐free,   real-­‐time   molecular  interaction  analysis.  Analytical  Biochemistry  361,  1–6  (2007).  104.  Xavier,   K.   A.,  McDonald,   S.  M.,  McCammon,   J.   A.   &  Willson,   R.   C.   Association   and  dissociation   kinetics   of   bobwhite   quail   lysozyme   with   monoclonal   antibody  HyHEL-­‐5.  Protein  Eng.  12,  79–83  (1999).  105.  Xavier,  K.  A.  &  Willson,  R.  C.  Association  and  dissociation  kinetics  of  anti-­‐hen  egg  lysozyme   monoclonal   antibodies   HyHEL-­‐5   and   HyHEL-­‐10.   Biophys   J   74,   2036–2045  (1998).  106.  Bornhop,   D.   J.   et   al.   Free-­‐Solution,   Label-­‐Free  Molecular   Interactions   Studied   by  Back-­‐Scattering  Interferometry.  Science  317,  1732–1736  (2007).  107.  Homola,   J.,   Yee,   S.   S.   &   Gauglitz,   G.   Surface   plasmon   resonance   sensors:   review.   Sensors  and  Actuators  B:  Chemical  54,  3–15  (1999).  108.  D’Ulivo,   L.,   Saint-­‐Guirons,   J.,   Ingemarsson,   B.   &   Riekkola,   M.-­‐L.   Quartz   crystal  microbalance,   a   valuable   tool   for   elucidation   of   interactions   between   apoB-­‐100   165 peptides   and   extracellular   matrix   components.   Analytical   and   Bioanalytical   Chemistry  396,  1373–1380  (2010).  109.  Pope,  M.  E.,  Soste,  M.  V.,  Eyford,  B.  A.,  Anderson,  N.  L.  &  Pearson,  T.  W.  Anti-­‐peptide  antibody   screening:   Selection   of   high   affinity   monoclonal   reagents   by   a   refined  surface  plasmon  resonance  technique.  Journal  of  Immunological  Methods  341,  86–96  (2009).  110.  Leonard,   P.   et   al.   High   throughput   ranking   of   recombinant   avian   scFv   antibody  fragments   from   crude   lysates   using   the   Biacore   A100.   Journal   of   Immunological   Methods  323,  172–179  (2007).  111.    Biacore   Life   Sciences   >   Product   Range   >   Systems   Overview   >   Biacore   3000   >  System   Information.   at    112.  Singhal,   A.,   Haynes,   C.   A.   &  Hansen,   C.   L.  Microfluidic  measurement   of   antibody-­‐antigen  binding  kinetics  from  low-­‐abundance  samples  and  single  cells.  Anal.  Chem   82,  8671–8679  (2010).  113.  Phillips,   D.   C.   Symposium   on   Three-­‐Dimensional   Structure   of  Macromolecules   of  Biological  Origin.  by  Invitation  of  the  Committee  on  Arrangements  for  the  Autumn  Meeting.   Presented   before   the   Academy   on   October   19,   1966.   Chairman,  Walter  Kauzmann:   THE   HEN   EGG-­‐WHITE   LYSOZYME   MOLECULE.   Proceedings   of   the   National  Academy  of  Sciences  of  the  United  States  of  America  57,  483  (1967).  114.  Goldbaum,   F.   A.   et   al.   Lack   of   significant   differences   in   association   rates   and  affinities   of   antibodies   from   short-­‐term   and   long-­‐term   responses   to   hen   egg  lysozyme.  J.  Immunol  162,  6040–6045  (1999).  115.  Batista,   F.   D.   &   Neuberger,   M.   S.   Affinity   Dependence   of   the   B   Cell   Response   to  Antigen:  A  Threshold,  a  Ceiling,  and  the  Importance  of  Off-­‐Rate.  Immunity  8,  751–759  (1998).  116.  Ito,  W.,   Yasui,   H.   &  Kurosawa,   Y.  Mutations   in   the   Complementarity-­‐determining  Regions  do  not  cause  Differences  in  Free  Energy  during  the  Process  of  Formation  of   the   Activated   Complex   between   an   Antibody   and   the   Corresponding   Protein  Antigen.  Journal  of  Molecular  Biology  248,  729–732  (1995).  117.  England,   P.,   Nageotte,   R.,   Renard,   M.,   Page,   A.   L.   &   Bedouelle,   H.   Functional  characterization  of  the  somatic  hypermutation  process  leading  to  antibody  D1.3,  a  high   affinity   antibody   directed   against   lysozyme.   J.   Immunol   162,   2129–2136  (1999).  118.  Tyn,   M.   T.   &   Gusek,   T.   W.   Prediction   of   diffusion   coefficients   of   proteins.   Biotechnology  and  Bioengineering  35,  327–338  (1990).  119.  He,   L.-­‐Z.   &   Niemeyer,   B.   A   Novel   Correlation   for   Protein   Diffusion   Coefficients  Based  on  Molecular  Weight  and  Radius  of  Gyration.  Biotechnol.  Prog.  19,  544–548  (2003).  120.  Dubin,  S.  B.,  Clark,  N.  A.  &  Benedek,  G.  B.  Measurement  of  the  Rotational  Diffusion  Coefficient   of   Lysozyme   by   Depolarized   Light   Scattering:   Configuration   of  Lysozyme  in  Solution.  The  Journal  of  Chemical  Physics  54,  5158–5164  (1971).   166 121.  Abdiche,   Y.,   Malashock,   D.,   Pinkerton,   A.   &   Pons,   J.   Determining   kinetics   and  affinities  of  protein  interactions  using  a  parallel  real-­‐time  label-­‐free  biosensor,  the  Octet.  Analytical  Biochemistry  377,  209–217  (2008).  122.  Bates,   S.   R.   &   Quake,   S.   R.   Highly   parallel   measurements   of   interaction   kinetic  constants   with   a   microfabricated   optomechanical   device.   Appl.   Phys.   Lett.   95,  073705  (2009).  123.    Biacore   Life   Sciences   >   Technology   >   Introduction   >  Data   from   an   interaction   >  Single-­‐Cycle   Kinetics.   at    124.  McKinney,  K.  L.,  Dilwith,  R.  &  Belfort,  G.  Optimizing  antibody  production   in  batch  hybridoma  cell  culture.  Journal  of  Biotechnology  40,  31–48  (1995).  125.  Tan,   W.-­‐H.   &   Takeuchi,   S.   A   trap-­‐and-­‐release   integrated   microfluidic   system   for  dynamic  microarray       applications.  Proc.  Natl.   Acad.   Sci.  U.   S.   A.  104,   1146–1151  (2007).  126.  Stevens,   R.   H.   Immunoglobulin   heavy   chain   mRNA   in   mitogen-­‐stimulate   B   cells.   European  Journal  of  Immunology  5,  47–53  127.  Arbabi-­‐Ghahroudi,   M.,   Tanha,   J.   &   MacKenzie,   R.   Prokaryotic   expression   of  antibodies.  Cancer  and  Metastasis  Reviews  24,  501–519  (2005).  128.  Chiba,   Y.   &   Akeboshi,   H.   Glycan   Engineering   and   Production   of   ‘Humanized’  Glycoprotein   in   Yeast   Cells.   Biological   and   Pharmaceutical   Bulletin   32,   786–795  (2009).  129.  Verma,  R.,  Boleti,  E.  &  George,  A.  J.  .  Antibody  engineering:  Comparison  of  bacterial,  yeast,   insect   and   mammalian   expression   systems.   Journal   of   Immunological   Methods  216,  165–181  (1998).  130.  Ma,   J.   K.-­‐C.,   Drake,   P.   M.   W.   &   Christou,   P.   The   production   of   recombinant  pharmaceutical  proteins  in  plants.  Nat  Rev  Genet  4,  794–805  (2003).  131.  Lagerkvist,  A.   C.,   Furebring,   C.  &  Borrebaeck,   C.  A.   Single,   antigen-­‐specific  B   cells  used  to  generate  Fab  fragments  using  CD40-­‐mediated  amplification  or  direct  PCR  cloning.  BioTechniques  18,  862–869  (1995).  132.  de  Wildt,  R.  M.  et  al.  A  new  method  for  the  analysis  and  production  of  monoclonal  antibody   fragments   originating   from   single   human   B   cells.   J.   Immunol.   Methods   207,  61–67  (1997).  133.  Heilmann,  C.  &  Barington,  T.  Distribution  of  kappa  and  lambda  light  chain  isotypes  among   human   blood   immunoglobulin-­‐secreting   cells   after   vaccination   with  pneumococcal  polysaccharides.  Scand.  J.  Immunol.  29,  159–164  (1989).  134.  Wang,  Z.  et  al.  Universal  PCR  amplification  of  mouse  immunoglobulin  gene  variable  regions:  the  design  of  degenerate  primers  and  an  assessment  of  the  effect  of  DNA  polymerase   3’   to   5’   exonuclease   activity.   Journal   of   Immunological  Methods  233,  167–177  (2000).  135.  Saiki,   R.   et   al.   Primer-­‐Directed   Enzymatic   Amplification   of   Dna   with   a  Thermostable  Dna-­‐Polymerase.  Science  239,  487–491  (1988).  136.  Ririe,  K.  M.,  Rasmussen,  R.  P.  &  Wittwer,  C.  T.  Product  differentiation  by  analysis  of  DNA  melting   curves  during   the       polymerase   chain   reaction.  Anal.  Biochem.  245,  154–160  (1997).   167 137.  Rychlik,   W.   &   Rhoads,   R.   E.   A   computer   program   for   choosing   optimal  oligonudeotides   for   filter   hybridization,   sequencing   and   in   vitro   amplification   of  DNA.  Nucl.  Acids  Res.  17,  8543–8551  (1989).  138.  Owczarzy,   R.   Melting   temperatures   of   nucleic   acids:   Discrepancies   in   analysis.   Biophysical  Chemistry  117,  207–215  (2005).  139.  Bustin,   S.   A.   Absolute   Quantification   of   mRNA   Using   Real-­‐Time   Reverse  Transcription   Polymerase   Chain   Reaction   Assays.   J   Mol   Endocrinol   25,   169–193  (2000).  140.  Don,  R.  H.,  Cox,  P.  T.,  Wainwright,  B.  J.,  Baker,  K.  &  Mattick,  J.  S.  ‘Touchdown’  PCR  to  Circumvent   Spurious   Priming   During   Gene   Amplification.   Nucl.   Acids   Res.   19,  4008–4008  (1991).  141.  Hecker,   K.   &   Roux,   K.   High   and   low   annealing   temperatures   increase   both  specificity  and  yield       in   touchdown  and  stepdown  PCR.  Biotechniques  20,  478–&  (1996).  142.    Invitrogen   Dynabeads   Oligo(dT)25   datasheet.   at    143.  Phillips,   R.,   Kondev,   J.   &   Theriot,   J.  Physical   Biology   of   the   Cell.   (Garland   Science:  2008).  144.  Subkhankulova,  T.  &  Livesey,  F.  J.  Comparative  evaluation  of  linear  and  exponential  amplification   techniques   for   expression   profiling   at   the   single-­‐cell   level.  Genome   Biology  7,  R18  (2006).  145.  Linke,   D.   Detergents:   An   Overview.   Guide   to   Protein   Purification,   Second   Edition   463,  603–617  (2009).  146.  Newman,  M.  A.,  Mainhart,  C.  R.,  Mallett,  C.  P.,  Lavoie,  T.  B.  &  Smith-­‐Gill,  S.  J.  Patterns  of   antibody   specificity   during   the   BALB/c   immune   response   to   hen   eggwhite  lysozyme.  J.  Immunol  149,  3260–3272  (1992).  147.  Gammon,   G.   et   al.   The   Choice   of   T-­‐Cell   Epitopes   Utilized   on   a   Protein   Antigen  Depends   on   Multiple   Factors   Distant   from,   as   well   as   at   the   Determinant   Site.   Immunological  Reviews  98,  54–73  (1987).  148.  Brink,  R.,  Phan,  T.  G.,  Paus,  D.  &  Chan,  T.  D.  Visualizing  the  effects  of  antigen  affinity  on   T-­‐dependent   B-­‐cell   differentiation.   Immunology   and   Cell   Biology   86,   31–39  (2008).  149.  Lecault,   V.   et   al.   High-­‐throughput   analysis   of   single   hematopoietic   stem   cell  proliferation   in   microfluidic   cell   culture   arrays.   Nature   Methods   8,   581–586  (2011).  150.  Corti,   D.,   Sallusto,   F.   &   Lanzavecchia,   A.   High   throughput   cellular   screens   to  interrogate  the  human  T  and  B  cell  repertoires.  Current  Opinion  in  Immunology  23,  430–435  (2011).  151.  Chan,  T.  D.  &  Brink,  R.  Affinity-­‐based  selection  and  the  germinal  center  response.   Immunological  Reviews  247,  11–23  (2012).  152.  Lavoie,   T.   B.   et   al.   Structural   differences   among   monoclonal   antibodies   with  distinct   fine   specificities   and   kinetic   properties.   Mol.   Immunol   36,   1189–1205  (1999).   168 153.  Schwesinger,   F.   et   al.   Unbinding   forces   of   single   antibody-­‐antigen   complexes  correlate   with   their   thermal   dissociation   rates.   Proc.   Natl.   Acad.   Sci.   U.S.A.   97,  9972–9977  (2000).  154.  Eisen,   H.   N.   &   Siskind,   G.   W.   Variations   in   Affinities   of   Antibodies   during   the  Immune  Response*.  Biochemistry  3,  996–1008  (1964).  155.  Banks,   D.   S.   &   Fradin,   C.   Anomalous   Diffusion   of   Proteins   Due   to   Molecular  Crowding.  Biophys  J  89,  2960–2971  (2005).  156.  Xavier,  K.  A.  &  Willson,  R.  C.  Association  and  dissociation  kinetics  of  anti-­‐hen  egg  lysozyme   monoclonal   antibodies   HyHEL-­‐5   and   HyHEL-­‐10.   Biophys.   J   74,   2036–2045  (1998).  157.  Ofek,   G.   et   al.   Relationship   between   Antibody   2F5   Neutralization   of   HIV-­‐1   and  Hydrophobicity  of   Its  Heavy  Chain  Third  Complementarity-­‐Determining  Region.   J   Virol  84,  2955–2962  (2010).  158.  Lavoie,   T.   B.,   Drohan,   W.   N.   &   Smith-­‐Gill,   S.   J.   Experimental   Analysis   by   Site-­‐Directed  Mutagenesis  of  Somatic  Mutation  Effects  on  Affinity  and  Fine  Specificity  in  Antibodies  Specific  for  Lysozyme.  J  Immunol  148,  503–513  (1992).  159.  Xu,   J.   L.  &  Davis,  M.  M.  Diversity   in   the  CDR3  Region  of  VH   Is  Sufficient   for  Most  Antibody  Specificities.  Immunity  13,  37–45  (2000).  160.  Boder,  E.  T.  &  Wittrup,  K.  D.  Optimal  Screening  of  Surface-­‐Displayed  Polypeptide  Libraries.  Biotechnology  Progress  14,  55–62  (1998).  161.  Yancopoulos,   G.   D.,   Malynn,   B.   A.   &   Alt,   F.   W.   Developmentally   Regulated   and  Strain-­‐Specific   Expression  of  Murine  VH  Gene  Families.   J   Exp  Med  168,   417–435  (1988).  162.  Jiang,  N.  et  al.  Determinism  and  Stochasticity  During  Maturation  of   the  Zebrafish  Antibody  Repertoire.  PNAS  108,  5348–5353  (2011).  163.  Braden,  B.  C.,  Goldman,  E.  R.,  Mariuzza,  R.  A.  &  Poljak,  R.  J.  Anatomy  of  an  antibody  molecule:   structure,   kinetics,   thermodynamics   and   mutational   studies   of   the  antilysozyme  antibody  D1.3.  Immunol.  Rev  163,  45–57  (1998).  164.  Lötscher,   M.,   Amstutz,   H.,   Heusser,   C.   H.   &   Blaser,   K.   Fine   specificity   and   heavy  chain  V  gene  usage  of  antibodies  to  the  phosphorylcholine  hapten.  Mol.   Immunol.   27,  369–378  (1990).  165.  Lötscher,   M.,   Blaser,   K.   &   Heusser,   C.   H.   Specific   V   gene   usage   in   anti-­‐phosphorylcholine  IgE  antibodies.  Mol.  Immunol.  30,  1573–1582  (1993).  166.  Robert,   R.   et   al.   Restricted   V   gene   usage   and   VH/VL   pairing   of   mouse   humoral  response   against   the   N-­‐terminal   immunodominant   epitope   of   the   amyloid   β  peptide.  Mol.  Immunol.  48,  59–72  (2010).  167.  Fernandez-­‐Sanchez,   A.,   Garcia-­‐Ocana,  M.  &   de   los   Toyos,   J.   R.  Mouse  monoclonal  antibodies  to  pneumococcal  C-­‐polysaccharide  backbone      show  restricted  usage  of  VH-­‐DH-­‐JH  gene   segments   and   share   the   same  kappa       chain.   Immunol.   Lett.  123,  125–131  (2009).  168.  Adderson,  E.  E.,  Shackelford,  P.  G.,  Quinn,  A.  &  Carroll,  W.  L.  Restricted  Ig  H  chain  V  gene   usage   in   the   human   antibody   response   to   Haemophilus   influenzae   type   b  capsular  polysaccharide.  J.  Immunol.  147,  1667–1674  (1991).  169.  Breden,  F.  et  al.  Comparison  of  antibody  repertoires  produced  by  HIV-­‐1  infection,  other  chronic  and  acute  infections,  and  systemic  autoimmune  disease.  PLoS  ONE  6,  e16857  (2011).   169 170.  McLean,   G.   R.   et   al.   Recognition   of   human   cytomegalovirus   by   human   primary  immunoglobulins  identifies  an  innate  foundation  to  an  adaptive  immune  response.   J.  Immunol.  174,  4768–4778  (2005).  171.  Li,   G.-­‐M.   et   al.   Pandemic   H1N1   Influenza   Vaccine   Induces   a   Recall   Response   in  Humans   That   Favors   Broadly   Cross-­‐Reactive   Memory   B   Cells.   PNAS  (2012).doi:10.1073/pnas.1118979109  172.  Thomson,   C.   A.   et   al.   Pandemic   H1N1   influenza   infection   and   vaccination   in  humans   induces   cross-­‐protective   antibodies   that   target   the   hemagglutinin   stem.   Front.  Immun.  3,  87  (2012).  173.  Hangartner,  L.,  Zinkernagel,  R.  M.  &  Hengartner,  H.  Antiviral  antibody  responses:  the   two   extremes   of   a   wide   spectrum.  Nature   Reviews   Immunology   6,   231–243  (2006).  174.  Hudziak,  R.  M.  et  al.  p185HER2  monoclonal  antibody  has  antiproliferative  effects  in  vitro   and   sensitizes   human   breast   tumor   cells   to   tumor   necrosis   factor.  Mol   Cell   Biol  9,  1165–1172  (1989).  175.  Han,   M.   Y.,   Gao,   X.   H.,   Su,   J.   Z.   &   Nie,   S.   Quantum-­‐dot-­‐tagged   microbeads   for  multiplexed  optical  coding  of      biomolecules.  Nat.  Biotechnol.  19,  631–635  (2001).  176.  Lelli,   D.   et   al.   West   Nile   virus:   characterization   and   diagnostic   applications   of      monoclonal  antibodies.  Virol.  J.  9,  (2012).  177.  Wrammert,  J.  et  al.  Broadly  Cross-­‐Reactive  Antibodies  Dominate  the  Human  B  Cell  Response   Against   2009   Pandemic   H1N1   Influenza   Virus   Infection.   J   Exp   Med  (2011).doi:10.1084/jem.20101352  178.  Heyndrickx,  L.  et  al.   International  Network   for  Comparison  of  HIV  Neutralization  Assays:  The  NeutNet  Report  II.  PLoS  ONE  7,  e36438  (2012).  179.  Rimmelzwaan,  G.  F.  et  al.  Use  of  GFP-­‐expressing  influenza  viruses  for  the  detection  of  influenza  virus  A/H5N1  neutralizing  antibodies.  Vaccine  29,  3424–3430  (2011).  180.  Katz,   J.   M.,   Hancock,   K.   &   Xu,   X.   Serologic   assays   for   influenza   surveillance,  diagnosis  and  vaccine  evaluation.  Expert  Review  of  Anti-­‐infective  Therapy  9,  669–683  (2011).  181.  Tay,   S.   et   al.   Single-­‐cell   NF-­‐κB   dynamics   reveal   digital   activation   and   analogue  information  processing.  Nature  466,  267–271  (2010).  182.  Rathanaswami,   P.   et   al.   Demonstration   of   an   in   vivo   generated   sub-­‐picomolar  affinity   fully   human   monoclonal   antibody   to   interleukin-­‐8.   Biochemical   and   Biophysical  Research  Communications  334,  1004–1013  (2005).  183.  Ozawa,  T.,  Kishi,  H.  &  Muraguchi,  A.  Amplification  and  analysis  of  cDNA  generated  from  a  single  cell  by  5’-­‐RACE:  application  to  isolation  of  antibody  heavy  and  light  chain  variable  gene  sequences  from  single  B  cells.  BioTechniques  40,  469–470,  472,  474  passim  (2006).  184.  Lao,  K.  Q.  et  al.  mRNA-­‐Sequencing  Whole  Transcriptome  Analysis  of  a  Single  Cell  on  the  SOLiDTM  System.  J  Biomol  Tech  20,  266–271  (2009).  185.  Singhal,  A.,  Hansen,  C.  L.  G.,  Schrader,  J.  W.,  Haynes,  C.  A.  &  Costa,  D.  J.  D.  Methods  for   Assaying   Cellular   Binding   Interactions.   at    186.  Birch,   J.  R.  &  Racher,  A.   J.  Antibody  production.  Adv.  Drug  Deliv.  Rev.  58,  671–685  (2006).   170 187.  Holliger,   P.  &  Hudson,  P.   J.   Engineered   antibody   fragments   and   the   rise   of   single  domains.  Nat.  Biotechnol.  23,  1126–1136  (2005).     171 Appendices   Appendix  A    -­‐  Primer  Designs  for  Amplifying  Mouse  Antibody  Genes   A.1 Highly  Degenerate  Primer  Set  for  Amplifying  Mouse  Heavy  and  Light  Chain   Antibody  Genes.    Primers  sequences  taken  from  Wang  et  al.134  Primer  Name   Sequence   CH  Region  (3')   IgA   gga  aga  tct  GAT  GGT  GGG  ATT  TCT  CGC  AGA  CTC   IgE   gga  aga  tct  TAA  GGG  GTA  GAG  CTG  AGG  GTT  CCT  G   IgM   gga  aga  tct  GAC  ATT  TGG  GAA  GGA  CTG  ACT  CTC   IgG1   gga  aga  tct  ATA  GAC  AGA  TGG  GGG  TGT  CGT  TTT  GGC   IgG2A   gga  aga  tct  CTT  GAC  CAG  GCA  TCC  TAG  AGT  CA   IgG2B   gga  aga  tct  AGG  GGC  CAG  TGG  ATA  GAC  TGA  TGG   IgG3   gga  aga  tct  AGG  GAC  CAA  GGG  ATA  GAC  AGA  TGG   VH  Region  (5')   5'  MH1   ctt  ccg  gaa  ttc  SAR  GTN  MAG  CTG  SAG  SAG  TC   5'  MH2   ctt  ccg  gaa  ttc  SAR  GTN  MAG  CTG  SAG  SAG  TCW  GG   5'  MH3   ctt  ccg  gaa  ttc  CAG  GTT  ACT  CTG  AAA  GWG  TST  G   5'  MH4   ctt  ccg  gaa  ttc  GAG  GTC  CAR  CTG  CAA  CAR  TC   5'  MH5   ctt  ccg  gaa  ttc  CAG  GTC  CAA  CTV  CAG  CAR  CC   5'  MH6   ctt  ccg  gaa  ttc  GAG  GTG  AAS  STG  GTG  GAA  TC   5'  MH7   ctt  ccg  gaa  ttc  GAT  GTG  AAC  TTG  GAA  GTG  TC   Cκ  Region  (3')   3'  Kc   ggt  gca  tgc  GGA  TAC  AGT  TGG  TGC  AGC  ATC   Vκ  Region  (5')   5'  Mk   gg  gag  ctc  GAY  ATT  GTG  MTS  ACM  CAR  WCT  MCA   172 A.2 Low  Degeneracy  Nested  PCR  Primer  Set   for  Amplifying  Mouse  Heavy  and   Light  Chain  Antibody  Genes.    (continued  on  next  page)  Primer  sequences  taken  from  Tiller  et  al.134       Primer  Name   Sequence   IgH  1st  round  PCR   5′  MsVHE     GGGAATTCGAGGTGCAGCTGCAGGAGTCTGG   3′  Cμ  outer     AGGGGGCTCTCGCAGGAGACGAGG     3′  Cγ1  outer     GGAAGGTGTGCACACCGCTGGAC   3′  Cγ2c  outer     GGAAGGTGTGCACACCACTGGAC     3′  Cγ2b  outer     GGAAGGTGTGCACACTGCTGGAC     3′  Cγ3  outer     AGACTGTGCGCACACCGCTGGAC   3′  Cα  outer     GAAAGTTCACGGTGGTTATATCC   IgH  2nd  round  PCR   5′  MsVHE     GGGAATTCGAGGTGCAGCTGCAGGAGTCTGG   3′  Cμ  inner     AGGGGGAAGACATTTGGGAAGGAC   3′  Cγ1  inner     GCTCAGGGAAATAGCCCTTGAC   3′  Cγ2c  inner     GCTCAGGGAAATAACCCTTGAC   3′  Cγ2b  inner     ACTCAGGGAAGTAGCCCTTGAC   3′  Cγ3  inner     GCTCAGGGAAGTAGCCTTTGAC   3′  Cα  inner     TGCCGAAAGGGAAGTAATCGTGAAT     173 A.3 (continued   from   previous   page)   Low  Degeneracy   Nested   PCR   Primer   Set   for  Amplifying  Mouse  Heavy  and  Light  Chain  Antibody  Genes.    Primer  sequences  taken  from  Tiller  et  al.134       Primer  Name   Sequence       Igκ  1st  round  PCR   5′  L-­‐Vκ_3     TGCTGCTGCTCTGGGTTCCAG   5′  L-­‐Vκ_4     ATTWTCAGCTTCCTGCTAATC   5′  L-­‐Vκ_5     TTTTGCTTTTCTGGATTYCAG   5′  L-­‐Vκ_6     TCGTGTTKCTSTGGTTGTCTG   5′  L-­‐Vκ_6,8,9     ATGGAATCACAGRCYCWGGT   5′  L-­‐Vκ_14     TCTTGTTGCTCTGGTTYCCAG   5′  L-­‐Vκ_19     CAGTTCCTGGGGCTCTTGTTGTTC   5′  L-­‐Vκ_20     CTCACTAGCTCTTCTCCTC   3′  mCκ     GATGGTGGGAAGATGGATACAGTT       Igκ  2nd  round  PCR   5′  mVkappa   GAYATTGTGMTSACMCARWCTMCA   3′  BsiWI  P-­‐mJK01     GCCACCGTACGTTTGATTTCCAGCTTGGTG   3′  BsiWI  P-­‐mJK02     GCCACCGTACGTTTTATTTCCAGCTTGGTC   3′  BsiWI  P-­‐mJK03     GCCACCGTACGTTTTATTTCCAACTTTGTC   3′  BsiWI  P-­‐mJK04     GCCACCGTACGTTTCAGCTCCAGCTTGGTC   174 Appendix  B    -­‐  Labview  Software  for  Hardware  Automation  and  Image  Analysis   B.1 Custom  LabView  Software   to  Automate  CCD  Camera,  Brightfield   Illumination,  Microscope   Stage  Control,   and   Microfluidic  Valve  Operation.      Image  prepared  by  Daniel  Da  Costa.   175 B.2 Custom  LabView  Software  for  Automated  Analysis  of  Images.         The  user  selects  a  region-­‐of-­‐interest  and  the  program  calculates  the  maximum  fluorescence  intensity  within  this  region  in  all   images.  This   program  was  used  to  measure  fluorescence  of  all  chambers  as  well  as  binding  kinetics  of  individual  chambers.  Binding  rate  constants  were   determined  by   fitting   the  data   to   first-­‐order  mass-­‐action  equations   (Chapter  2,   Section  2.1.1)  using  nonlinear   least   squares  minimization.   Image  prepared  by  Daniel  Da  Costa.    """@en ; edm:hasType "Thesis/Dissertation"@en ; vivo:dateIssued "2012-11"@en ; edm:isShownAt "10.14288/1.0071837"@en ; dcterms:language "eng"@en ; ns0:degreeDiscipline "Chemical and Biological Engineering"@en ; edm:provider "Vancouver : University of British Columbia Library"@en ; dcterms:publisher "University of British Columbia"@en ; dcterms:rights "Attribution-NonCommercial-NoDerivatives 4.0 International"@en ; ns0:rightsURI "http://creativecommons.org/licenses/by-nc-nd/4.0/"@en ; ns0:scholarLevel "Graduate"@en ; dcterms:title "Microfluidic technologies for rapid, high-throughput screening and selection of monoclonal antibodies from single cells"@en ; dcterms:type "Text"@en ; ns0:identifierURI "http://hdl.handle.net/2429/43575"@en .