UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Microfluidic technologies for rapid, high-throughput screening and selection of monoclonal antibodies.. 2012

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata

Download

Media
ubc_2012_fall_singhal_anupam.pdf [ 29.89MB ]
Metadata
JSON: 1.0071837.json
JSON-LD: 1.0071837+ld.json
RDF/XML (Pretty): 1.0071837.xml
RDF/JSON: 1.0071837+rdf.json
Turtle: 1.0071837+rdf-turtle.txt
N-Triples: 1.0071837+rdf-ntriples.txt
Citation
1.0071837.ris

Full Text

 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