ENGINEERING DEEP SEQUENC ING - GUIDED PLATFORMS TO EVALUATE SEQUENCE - FUNCTION RELATIONSHIPS BETWEEN PROTEINS FOR THE DEVELOPMENT OF THERAPEUTIC ANTIBODIES By Angélica V . Medina - Cucurella A DISSERTATION Submitted to Michigan State Univer sity i n partial fulfillment of the requirements f or the degree of Chemical Engineering Doctor of Philosophy 201 9 A BSTRACT ENGINEERING DEEP SEQUENCING - GUIDED PLATFORMS TO EVALUATE SEQUENCE - FUNCTION RELATIONSHIPS BETWEEN PROTEINS FOR THE DEVELOPM E NT OF T HERAPEUTIC ANTIBODIES By Angélica V . Medina - Cucurella Over the past two decades, monoclonal antibodies (mAbs) have been used as a major class of therapeutic treatments for cancer and autoimmune diseases given their high specificity against a given t a rget a ntigen. mAbs can work as antagonists by blocking the downstream signaling pathway through receptors or as agonists by boosting the immune system response to direct tumor cell apoptosis. The understanding of the antibody - mediated recognition o f patho g ens re veals valuab le information related to the immune - protective responses within the host organism. Such information has led scientists to develop new effective vaccines and therapeutics. Nevertheless, u nderstanding the physical basis of affinity and sp e cifici ty in these interactions is a theoretical and experimental challenge. Subsequently, researchers have developed multiple high - throughput approaches , like deep mutational scanning , to identify the relative binding contribution of individual ami no acid residu es towards t he overall antibody:antigen complex . In this dissertation, I present the successful application of deep sequencing - guided engineering platforms to address numerous aims relevant to the protein engineering and antibody discovery fi eld inc l uding the understanding of sequence - function relationships between proteins, antibody conformational epitope mapping, and the development of antibody therapeutic s. First, we use our pipeline utilizing comprehensive mutagenesis, yeast surface displ ay, and deep s equencing to gain insights on the interactions between i nterleukin - 31, a cytokine involved in chronic skin inflammations, and its receptors . I dentification of the binding sites on interleukin - 31 by its receptors allows the development of anta gonist m Abs to inhibit the downstream signaling pathway. In fact, the mappe d conformational epitope of a candidate mAb shows that it inhibit s the signaling pathway by bi nding an overlapping site shared between receptors . A significant limitation of sequenc e - funct i on map ping by the above method is the requirement that the yeast surface displayed target protein be in a conformation recognizable by the antibody. For example, s ome proteins such as the neurotrophin family display on the yeast surface in a mostly misfol d ed or inactive conformation . Consequently, we developed a deep sequencing - guided protein engineering workflow to increase the production of folded canine nerve growth factor, a neurotrophin involved in multiple chronic pain conditions. W e identif ie d benef i cial m utations with in the pro - region of the protein that improved the display of mature , conformationally sensiti v e protein that enabled the determinatio n of conformational epitope s for multiple antagonist mAbs . Two fundamental limitations in the creatio n of la rge mutagenesis libraries using current template - based mutagenesis is the overrepresentation of specific nucleobases and the difficulty of constructing user - defined libraries beyond single site comprehensive codon scanning . We improve on curr ent met h ods by using unpurified oligo pools to prepare user - define single and double mutagenesis libraries from plasmid DNA. Results demon strated a near - complete coverage of desired mutations with even representation of nucleobases and few off - target mutat ions. L astly, we present a new method guided by next - generation sequencing for the selection in cell lysate of agonists mAb for OX40, a costimulatory immune receptor . This project was performed as an industrial internship during Summer 2018. Synthesized O X40 ant i bodies after deep sequencing selection with cell lysate showed higher therapeutic potentials compared to antibodies enri ched by the traditional soluble selection method . iv ess is t he doi ng, not the getting; in the trying, not the triumph. Success is a personal standard, reaching for the highest that is in us, becoming all that we can be . Zig Ziglar I dedicate this thesis to my family every little bit of me is a piece of you . Thank s for always supported and encouraged me to keep nourishing my passion for science. v ACKNOWLEDGMENTS I would like to acknowledge the extraordinary mentorship of my advisor, Dr. Tim Whitehead. Without his valuable guidance this dissert a tion w ould not have been possible. Thanks for all the experiences I was given through all these years to become a great researcher and quantitative thinker. To the past and present lab mates Dr. Caitlin Kowal s ky Stein , Dr. Justin Klesmith, Dr. E mily Wr e nbeck, and Matthew Faber thanks for willingness to help and guide me to success in all m y projects. Thanks for conveyed me a day - to - day spirit of adventure regarding research. I would also like to express the deepest appreciation to the Plant Bio technol o gy for Health and Sustainability graduate training program and the Kinesis Foundation for the financial support granted through my graduate studies. Undoubtedly, these supports gave me valuable skills and experiences that have influenced in my futu re ende a vors a nd career goals. I want to thank my graduate school friends, for listening all my fr ustrations and sharing laughs. Lastly, and most importantly, huge thanks to my eternal cheerleaders - my parents, my siblings, and my fiancé, William - for th eir nev e r - endi ng long - distance patience, love, and support through my graduate studies and for alw ays believe in me. vi TABLE OF CONTENTS LIST OF TABLES ................................ ................................ ................................ ....................... ix LIST OF FIGURES ................................ ................................ ................................ ...................... x KEY TO ABBREVIATIONS ................................ ................................ ................................ ... xiii CHAPTER 1 ................................ ................................ ................................ ................................ .. 1 1. Introducti o n to antibodies, epitope mapping, and deep sequencing - guided platforms ................ 1 1.1. Introduction ................................ ................................ ................................ .......................... 2 1.2. Background ................................ ................................ ................................ .......................... 2 1.2.1. Monoclonal Antibodies ................................ ................................ ................................ . 2 1.2.2. Epitope Mapping ................................ ................................ ................................ ........... 3 1.2.3. Deep Mutational Scanning ................................ ................................ ............................ 3 1.2.4. Next - generation sequencing in antibody discovery applications ................................ .. 8 CHAPTER 2: ................................ ................................ ................................ ............................... 11 2 . Feline Interleukin - 31 shares overlapping epitope s with Oncostatin M Receptor and IL - 31RA ................................ ................................ ................................ ................................ ....................... 11 2.1. Abstract ................................ ................................ ................................ .............................. 12 2.2. Introduction ................................ ................................ ................................ ........................ 13 2.3. Materials and methods ................................ ................................ ................................ ....... 15 2.3.1. Strains ................................ ................................ ................................ .......................... 15 2.3.2. Plasmid Constr ucts ................................ ................................ ................................ ...... 16 2.3.3. Preparation of fIL - 31, fOSMR, fIL - 31RA constructs, and mAb#1 ............................ 16 2.3.4. Analytical Size Exclusion Chromatography ................................ ................................ 16 2.3.5. Surface Plasmon Resonance binding assays ................................ ............................... 17 2.3.6. ELISA ................................ ................................ ................................ .......................... 17 2.3.7. Yeast Surface Displ a y Expression and Bindi ng Activity ................................ ............ 17 2.3.8. Competition Binding Assays ................................ ................................ ....................... 18 2.3.9. Cell - based assays ................................ ................................ ................................ ......... 18 2.3.10. Preparation of Mutagenesis Libraries ................................ ................................ ........ 18 2.3.11. Determina tion of receptor binding sites and conformational epitope of mAb#1 ...... 19 2.3.12. Data Analysis ................................ ................................ ................................ ............. 19 2.3.13. Data Availability ................................ ................................ ................................ ........ 20 2.4. Results ................................ ................................ ................................ ................................ 20 2.4.1. fOSMR and fIL - 31RA bind independently to f IL - 31 ................................ ................. 20 2.4.2. fOSMR partially, but not completely, inhibits binding of fIL31RA - 1FNIII to fIL - 31 22 2.4.3. fIL - 31 can be specifically inhibited by fOSMR - CBD in cell - based assays ................ 24 2.4.4. Structural homology model for fIL - 31 ................................ ................................ ........ 24 2.4.5. YSD and saturation mutagenesis reveals partially overlapping binding sites for fOSMR and fIL - 31RA ................................ ................................ ................................ ........... 25 vii 2.4.6. mAb#1 conformational epitope reveals nature of inhibiti o n ................................ ....... 29 2.5. Discussion and concl usion ................................ ................................ ................................ . 30 CHAPTER 3 ................................ ................................ ................................ ................................ 33 3. Pro region engineeri n g of nerve growth factor by deep mutational scanning enables a yeast platform for conformational epitope mapping of anti - NGF monoclonal antibodies .................... 33 3.1. Abstract ................................ ................................ ................................ .............................. 34 3.2. Intr oduction ................................ ................................ ................................ ........................ 35 3.3. Materials and Methods ................................ ................................ ................................ ....... 37 3.3.1. Plasmid Constructs ................................ ................................ ................................ ...... 37 3.3.2. Preparation of anti - NGF mAbs ................................ ................................ ................... 37 3.3.3. TF - 1 Cell Prolifer ation Assay ................................ ................................ ...................... 38 3.3.4. Surface Plasm o n Resonance ................................ ................................ ........................ 39 3.3.5. Yeast Surface Display Expression and Binding ................................ .......................... 39 3.3.6. Preparation of Mutagenesis Libraries ................................ ................................ .......... 40 3.3.7. Screening of Pro - - cNGF Libraries ................................ ................ 40 3.3.8. Determination of Conformational Epitopes ................................ ................................ 41 3.3.9. Deep Sequencing Preparatio n ................................ ................................ ...................... 41 3.3.10. Data Analysis ................................ ................................ ................................ ............. 41 3.3.11. Data Availability ................................ ................................ ................................ ........ 42 3.4. Results ................................ ................................ ................................ ................................ 42 3.4.1. Initial cNGF constructs are improperly folded on the yeast surface ........................... 42 3.4.2. Comprehensive analysis of pro mutations that improve cNGF folding ...................... 45 3.4.3. Combining single mutants improves the amount of displayed, folded cNGF ............. 49 3.4.4. Small amount of pro - cNGF display s on the yeast surface ................................ .......... 49 3.4.5. Conformational epito pes reveal similar profiles, but distinct epitopes, for tanezumab and all three mAbs ................................ ................................ ................................ ................. 51 3.5. Discussion and Conclusion ................................ ................................ ................................ 55 3.5.1. Pro - region engineering pipeline ................................ ................................ .................. 55 3.5.2. Pro - region sequence - function relationships ................................ ................................ 57 3.5.3. Mapp ing epitopes targeted by anti - cNGF mAbs ................................ ......................... 58 CHAPTER 4 ................................ ................................ ................................ ................................ 60 4. User - def ined single pot mutagenesis using unpurified oligo pools ................................ .......... 60 4.1. Abstract ................................ ................................ ................................ .............................. 61 4.2. Introduction ................................ ................................ ................................ ........................ 61 4.3. Materials and Methods ................................ ................................ ................................ ....... 65 4.3.1. Stra ins ................................ ................................ ................................ .......................... 65 4.3.2. Plasmid Constructs ................................ ................................ ................................ ...... 65 4.3.3. Degenerate Oligos and Oligo Pool Design ................................ ................................ .. 66 4.3.4. Preparat ion of Mutagenesis Libraries ................................ ................................ .......... 66 4.3.5. Deep Sequencing Preparation and Data Analysis ................................ ....................... 66 4.3.6. Yeast two - hybrid screening ................................ ................................ ......................... 67 CHAPTER 5 ................................ ................................ ................................ ................................ 68 5. Summary and Future Approaches ................................ ................................ ............................. 68 viii 5.1. Summary ................................ ................................ ................................ ............................ 69 5.2. Future Approaches ................................ ................................ ................................ ............. 70 APPENDICES ................................ ................................ ................................ ............................. 73 APPENDIX A : C ha racterizing protein - protein interactions using deep sequencing coupled to yeast surface display ................................ ................................ ................................ ..................... 74 APPENDIX B : P referential Identification o f Agonistic OX40 Antibodies by Using Cell Lysate to Pan Natively Paired, Humanized Mouse - Derived Yeast Surface Display Libraries ................ 98 APPENDIX C : Supplementary Notes ................................ ................................ ................... 123 APPENDIX D : Supplementary Figures ................................ ................................ .................. 130 APPENDIX E : Supplementary Tables ................................ ................................ ................... 151 REFERENCES ................................ ................................ ................................ .......................... 160 ix LIST OF TABLES Table 1. Summary of experimental determined dissociation constants ( KD) using yeast surface display and surface plasmon resonance. ................................ ................................ ....................... 22 Table 2. Signal:noise ratio for Pro - cNGF constructs defined as the MFI of yeast cells labeled with specified mAb over MFI of unla b eled cells. ................................ ................................ ......... 49 Table A 1.Gene amplification and Illumina adapter primers to prepare samples for deep sequencing . 91 Table B 1. Functional characteristics of 41 scFv binders converted into full - length IgG1 mAbs 117 Table E 1. Sorting conditions and FACS collection statistics for fIL - 31 libraries .. 152 Table E 2. Primers for deep sequencing. ................................ ................................ .................... 152 Table E 3. fIL - 31 library statistics results ................................ ................................ ................... 153 Table E 4. FACS collection statistics for Pro - - cNGF library screening experiments ................................ ................................ ................................ ................................ . 154 Table E 5. Summary of Average Dissociation Constant, K D values using Surface Plasmon Resonance (SPR) for human pro - NGF and canine NGF and Yeast Surface Display (YSD) for pro.v4 - cNGF, and sorting conditions for library screening using pro.v4 - cNGF. ....................... 154 Table E 6. FACS collection statistics f or cNGF libraries ................................ ........................... 155 Table E 7. Primers set for deep sequencing. ................................ ................................ ............... 155 Table E 8. Libraries Statistics Results for cNGF constructs ................................ ....................... 157 Table E 9. Primers set for deep sequencing. ................................ ................................ ............... 158 Table E 10. Summary of statistics for single site saturation mutagenesis (SSM) and doub le site saturation mutagenesis (DSM) libraries ................................ ................................ ...................... 159 x LIST OF FIGURES Figure 1. A streamlined process required for PPI characterization using deep sequencing and mutagenesis analysis. ................................ ................................ ................................ ...................... 5 Figure 2. Interactions between IL - 3 1 and its receptors. ................................ ................................ 14 Figure 3. Binding profiles of fel ine IL - 31 (fIL - 31) with all different constructs of fIL - 31RA, fOMSR - ECD, and mAb#1. ................................ ................................ ................................ ........... 21 Figure 4. Competition binding assays demonstrated that fOSMR - ECD, fIL - 31RA, and mAb#1 share an overlapping ep itope using SPR and YSD. ................................ ................................ ...... 23 Figure 5. Effect of fIL - 31 to induce STAT3 phosphorylation in macrophage cell line FCWF - 4 (a.) and fOSMR - ECD inhibition of pSTAT signaling induced by fIL - 31 (b.). ............................ 25 Figure 6. Determination of the fIL31RA - 1FNIII binding site using deep sequencing. ................ 27 Figure 7. Determination of the fOSMR - ECD binding site using deep sequencing. ..................... 28 Figure 8. Determination of the mAb#1 conformational epitope using deep sequencing. ............ 29 Figure 9. Specific bind ing sites revealed for fIL31RA - 1FNIII and fOSMR - ECD. ...................... 31 Figure 10. cNGF yeast display constructs are mostly misfolded as probed by conformationally sensitive mAbs. . ................................ ................................ ................................ ............................ 44 Figure 11. Identifying sequence - function relationships for pro region engineering using deep sequencing. ................................ ................................ ................................ ................................ .... 46 Figure 12. Per - position heatmap of enrichment scores for pro - cNGF mutants after 2 sorts with tanezumab. ................................ ................................ ................................ ................................ .... 48 Figure 13. Pro - cNGF variants with multiple point mutations show improved cNGF folding. .... 50 Figure 14. Binding titrations for surface - displayed pro.v4 - cNGF for various mAbs. .................. 51 Figure 15. Determination of the cNGF:tanezumab co nformational epitope using deep sequencing. ................................ ................................ ................................ ................................ .... 54 Figure 16. Correlation in Shannon entropy of the binding populations between tanezumab and other mAbs. ................................ ................................ ................................ ................................ ... 55 Figure 17.Correlation between the freque ncy of (a.) AmiE, (b.) PYR, and (c.) UCA9 mutants xi between replicates. ................................ ................................ ................................ ........................ 63 Figure 18. A single unpurified oligo pool combined with nicking mutagenesis shows a near - complete coverage of all programme d mutations for user - defined single and double mutagenic libraries. . ................................ ................................ ................................ ................................ ....... 64 Figure A 1. Essential considerations needed for preparing Site Saturation Mutagenesis (SSM) Libr aries . 81 Figure A 2. Sorting gates used for library screening. ................................ ................................ ... 87 Figure A 3. PCR steps performed for deep sequencing preparation of SSM libraries. ................ 89 Figure A 4. Deep sequencing results and data analysis used to determi ne the conformational epitope. ................................ ................................ ................................ ................................ .......... 94 Figure A 5. Flow - chart of analysis used to determine the conformational epitope. ..................... 94 Figure B 1. Overview of the generation and screening of scFv librari es derivates from B cells from humanized mice with either soluble OX40 or cells and DNA expressing OX40 109 Figure B 2. scFv libraries from immunized mice subjected to FACS selection for OX40 . ....... 111 Figure B 3. Overlapping clones in the pre - and post - sort populations obtained from each experimental parameter. ................................ ................................ ................................ .............. 114 Figure B 4. Cl onal cluster plot of anti - OX40 clones with frequencies higher than 0.1% in the post - sorted populations. ................................ ................................ ................................ .............. 116 Figure D 1. Mass spectrometry and analytical SEC for feline IL - 31 construct 131 Figure D 2. Structural model of fIL - 31. ................................ ................................ ..................... 132 Figure D 3. Determination of the binding sites for fIL - 31/fIL31 RA - 1FNIII interaction. .......... 133 Figure D 4. Determination of the binding si tes for fIL - 31 - fOSMR - ECD interaction. ............... 134 Figure D 5. Determinatio n of the conformational epitope for fIL - 31/mAb#1 interaction. . ....... 135 Figure D 6. Effects of anti - NGF mAbs on canine - NGF Induced Proliferation of TF - 1 Cells (representative curves). ................................ ................................ ................................ ............... 136 Figure D 7. Per - position heatmap of enrichment scores for pro - cNGF mutants after 1 sort with tanezumab. ................................ ................................ ................................ ................................ .. 137 Figure D 8. Per - position heatmap of enr ic hment scores for pro - cNGF mutants after 1 sort with mAb #1. ................................ ................................ ................................ ................................ ....... 138 xii Figure D 9. Per - - cNGF mu tants after 1 sort with tanezumab. ................................ ................................ ................................ .......................... 139 Figure D 10. Per - - cNGF mutants afte r 1 sort with mAb #1. ................................ ................................ ................................ .............................. 139 Figure D 11. Per - position heatmap of enrichment scores for pro - cNGF mutants after 2 sorts with mAb #1. ................................ ................................ ................................ ................................ ....... 140 Fig ure D 12. Mean fluorescence intensities for individual point mutants compared with wild - type pro - cNGF. ................................ ................................ ................................ ................................ ... 141 Figure D 13. Determination of conformat ional epitope for cNGF_tanezumab. ......................... 142 Figure D 14. Determination of conformational epitope for cNGF_mAb #1. ............................. 143 Figure D 15. Determination of conformational epitope for cNGF_mAb #2. ............................. 144 Figure D 16. Determination of conformational epitope for cNGF_mAb #3. . ............................ 145 Figure D 17. Correlation between counts in the displayed popu lation rel ative to the counts in the unselected population for cNGF. ................................ ................................ ................................ 146 Figure D 18. Per - position heatmap of sequencing counts for AmiE mutants i n (a.) replicate 1 and (b.) replicate 2 using the oligo pool. ................................ ................................ ........................... 147 Figure D 19. Per - position heatmap of sequencing counts for AmiE mutants using degenerate ................................ ................................ ................................ ............................. 148 Figure D 20. Per - position heatmap of sequencing counts of UCA9 mutants in (a.) replicate 1 and (b.) r eplicate 2 using the oligo pool. ................................ ................................ ........................... 149 Figure D 21. Per - position heatmap of sequencing counts for PYR1 mutants (a.) replicate 1 and (b.) replicate 2 using the oligo pool. ................................ ................................ ........................... 150 xiii KEY TO ABBREVIATIONS FACS fluorescence activated cell sorter IL - 31 interleukin - 31 IL - 31RA gp - 130 like receptor K D dissociation constant mAbs monoclon al antibodies NGF nerve growth factor OSMR oncostatin - M receptor PBS - BSA phosphate buff ered saline with bovine serum albumin PPIs protein - protein interactions scFv single - chain fragment variable SSM single - site saturation mutagenesis YSD yeast surfa ce display 1 CHAPTER 1 1. I ntroduction to antibodies , epitope mapping, and d eep sequencing - guided platforms Portions of this chapter Characterizing protein - prote in interactions using deep sequencing coupled to yeast surfac in M ethods in Molecular Biology ( 2018 ) 1764:101 - 121 2 1.1. Introduct ion Antibody - antigen interactions are central to understand ing humoral immune responses. These non - covalent interactions in volve hydrogen bonds, hydrophobic interaction s, e lectrost ati c interactions, and van der Waals forces . The surface conformational sha pe, the residues at the interface, and the sensitivity to the environmental conditions are some examples of the physical characteristic s that distinguish individual anti body paratop es 1 . The understanding of the antibody - mediated re cognition of pathogens reveals va luable information related to protective immune responses within the host organism. Such information may lead to the development of new effective vaccines and therape utics agains t m ultiple human diseases . 1.2 . Background 1.2.1. Monoclonal Antibodies D u rin g the past decades, monoclonal antibodies ( mAbs ) have been used as a major class of therapeutic treatment s against cancer and autoimmune diseases in addition to diagn ostic and re sea rch applications 2 . m Abs maintain efficacy b y binding at unique location s within the targe t antigen . Since 1986, a total of forty - seven thera peutic mAbs have been approved for commercial sale in the Unit ed State s and Eu rope with a worldwide sale of approximately 100 billion by the end of 2017 3 . Given the stead ily growing demand, it is expected that the mar ket size for mAb s will be worth $1 30 - 200 billion by 2022 3 . The low risk of unex pected safet y concern s and high - specificity towards targets make them an e nhanced thera peu tic o ver other alternative options 2 . These molecules can work as antagonist s by block ing the signaling pathway through receptor s or as agoni st s by boost ing the immune system response to direct tumor cell apoptosis 4 . Nevertheless , adequate experimental data is required to consider that a the rapeutic mAb is capable of mollifyi ng the disease being treated. 3 1.2.2. Ep itope Mapping A n important step for candidate mAb evaluation is to determine the epitope that they target. An epitope is the site of the antigen that is recognize d by the antibod y . T he epitopes can be presented as a linear or as a conformational epitope. A linear epit ope is composed of res idues that are adjacent i n the polypeptide chain while a conformational epitope is comp rise d of residues located in different fragments of the prot ein . M ultipl e epitope mapping studies suggest that natural antibodies reco gnize a confo rma tional epitope as linear epitopes are not always expose d in the quaternary structure of the antigen 5,6 . Th e epitope characterization can be used to provi de a basis for structural va cci ne design , to engineer and improve antibody affinity maturation , to develop agonist molecules for cancer therapy and to predict antigen structure, among others 7 9 . Des pite multipl e studies over the years, the u nderstanding of the physical ba sis of affini ty and specificit y in protein - protein interactions ( PPIs ) is a theoretical and experimental challenge. Not all residues contribute to the same magnitude in the binding i nteraction. The residues that show a strong binding interaction give guida nce to unders tan d how antibodies can potentially neutralize an infection a key component in the development of therapeutic mAbs. Consequently, researchers have developed multiple i nnovative te chniques to identify neut ralizing epitopes in faster , inexpens ive, and high th roughput approaches 10 12 . Indeed, most of these methods rel y on a powerful high - throughput technology called deep mutational scanning 13 to examine the sequence - function relations between PPIs . 1.2. 3 . Deep M utati o nal S cannin g D elineating the sequence determinants of stability, affinity , and specifi cit y of PPI s ha s been a major research goal for deca de s. The classical approach to study PPI sequence - function - hich residue s are individually mutated 4 to alanine and assayed by assorted biophysical t ech niques 14,15 . The change of binding affinity upon mutation gives a reasonable measure of the importance of the perturbed residue. However, such classical mutagenesis and screening studies are extremely labor - intensive . F or example, these methods are limited to the characterization of each mutant separately 14 . Recently, transformative methods utilizing large - scale mutagenesis, surface display, and next generation sequencing , NGS , have been developed to obtain relative binding contributions of individual residues by testing thousands of PPI mutants in a single experiment 8,10 13,16 . The change of binding affinity upon a mutation gi ves a reasonable measure of the importance of the perturbed residues. All methods share a similar framework: a population containing mutants of one PPI partner is prepared and cloned into a surface - display vector. The population is selected and/or screened for positive or negative b inding to the other partner, and then the selected and unselected populations are deep sequenced and analyzed. Finally, the change in frequency for each library member is calculated and converted to a relative binding sco re 16 . In the method developed by our lab 12 we utilize yeast surface display ( YSD ) 17,18 as it affords quantitative screening via fluo rescence - activated cell sorting ( FACS ) 5,10,11 . Yeast display has become one of the powerful platforms used for the isolatio n, the engineering, and the epitope mapp ing of antibodies and other proteins 17 . In the most common YSD technique, t he target protein is displayed as an Ag a 2 fusion protein, which is attached to yeast cell wall t o an Aga1 protein through disulfide bonds ( Figure 1a ) . Then, the protein expression is typically detected using a fluorescent conjugated antibody against either hemagglutinin or a c - cmyc tag using flow cytometry . Then, the interaction wi t h a biotinylated secondary protein is detected with a secondary fluorescent conjugated antibody again st biotin ( Figure 1b ) . This scheme of two - color labeling reaction identifies the stability and affinity of a protein - protein interaction simultaneously, one of the many advantages that yeast 5 Figure 1 . A streamlined process required for PPI characterization using deep sequencing and mutagenesis analysis . ( a . - c . ) Requirements for the pipeline: the binding activity of two proteins is measured using yeast surface display coupled to flow cytometry, and the relative dissociation c onstant is determined using yeast clonal titrations. The top panel is adapted from Chao et al. 17 ( d . ) The workflow covered in this chapter to characterize pro tein - protein interactions. ( e . - f . ) Deep sequencing results can be visualized as a heatmap and used to determine the conformational epitope of one member of the interaction. s urface display offers. Furthermore, the proteins can bind at different concentrati ons allowing the researchers to determine the dissociation constant , K D , following Hill k inetics ( Figure 1c ). Compared with competing methods using YSD 10,11,19 , our approach is faster and less expensive albeit with a limited d ynamic range of approximately 10 - fold change in binding affinity centered about the wild - type sequence. Accordingly, our method is suitable for fine maturati on of PPI affinity and specificity, or to determine fine conformational epitopes. As an example, an d as mentioned before , this method can be used to determine if a mAb can function as an antagonist to inhibit the downstream signaling pathway of an antigen . This goal can be achieved by understa nd ing the interactions between a target antigen and its recep tors. Chapter 2 presents a 6 study that aims to measure the intermolecular interactions between a cytokine involve s in chronic skin inflammation, interleukin - 31 (IL - 31) , and its receptors . To evaluate these interactions, we determined the binding sites of ea ch receptor using a predicted feline IL - 31 structure model combined with o ur workflow . Interesting, these sites largely overlapped with the previous sites determined by computational analysis and alanine scanning mutagenesis 20 . However, our binding sites revealed a new overlapping site between both receptors no t describe d before. We also mapped the conformational epitope for an anti - interleukin - 31 mAb. Our constructed epitope revealed that this mAb could potentially antagonize feline IL - 31 signaling pathway by (i.) binding to a shared site between both receptors, and (ii. ) decreasing the binding signals of both receptors . This work is a further evidence of the power of our method for studying the sequence - function relationship between PPI s and for mapping conformational epitopes of potential neutralizing antibody candidate s. Although yeast surface display is an extraordinary platform to evaluate multiples PPIs, a major limitation of this method is the requirement that the displayed protein must be in a conformation recognizable by the antibody. Some complicated protein s di splay on the surface of the yeast in a mostly inactive conformation , making conformational epitope m apping measurements impossible. As an example, a previous study showed that a member of the neurotrophin family, brain - derived neurotrophic factor, displaye d on the yeast surface in a mostly inactive conformation 21 . Chapter 3 presents a deep sequencing - guided protein engineering workflow developed to improve the production of folded, disp layed nerve growth factor ( NGF ) , a related neurotrophin involved in multiple chronic pain conditions 22 . This protein displayed on the yeast surface but was barely recognized by conformationally sensitive mAbs. Th us, we created m utational libraries in the neurotrophin pro region using an engineering pipeline to enhance the 7 production of folded, display ed canine NGF. Such libraries revealed new insight s into the sequence - function relationships of the neurotrophin pr o region and allow ed us to generate conformational epitopes maps of multiple anti - NGF mAbs using an engineered NGF construct . This study demo nstrate d the potential of deep sequencing - guided engineering pipelines to assign a functional effect for every poss ible nonsynonymous point mutant and to map the conformational epitopes for potential anti - NGF mAbs. O ur pipeline could be a promising platform to increase the production of other member s of neurotrophin family and other complicated proteins. Another limit ation of these methods is the design of oligonucleotides (oligos) for user - defined mutations used in the synthesis of large - scale mutagenesis libraries. These libraries are often produced by using degenerate oligos which can make up to 63 different mutatio ns per codon substitution . However, there are some experimental biases. For example, solid phase synthesis of degenerate oligonucleotides often introduces an overrepresentation of nucleobases. Multiple vendors have recently developed new microarray - synthes ized oligo pool s technolog ies which can be used for synthetic biology designs . Although these pools are synthe sized at low attomole amounts, given the very low oligo concentration needed in our mutagenesis protocol, we speculate that oligo pools could be d irectly used in the reaction. Chapter 4 presents the integration of a single oligo pool with nicking mutagenes is 23 to construct user - defined single and double mutagenesis libraries for three different targeted proteins with low off - target rates . These oligo pools can be designed to encode every possible single mutation in a more precise and h omo geneous manner . Indeed, the results showed that the oligo pool librar ies had a much more even representation of all 20 amino acid substitutions compared with the traditional degenerate oligos. The constructed libraries show a near - complete coverage o f the programmed mutations with a small percent of non - programmed mutati ons. T his technology can be used for 8 further studies including conformational epitope mapping, protein engineering directed evolution pipelines, and local fitness landscape evaluation . A ppendix A presents our step - by - step protocol to determine the affinity and specificity for full - length protein binders which can be used to determine fine conformational epitopes 24 . This me thod combines the creation of single site saturation mut agenesis ( SSM ) libraries using nicking m utagenesis 23 , the transformation into yeast, the screening of SSM libraries using FACS, the DNA preparation for deep seq uencing, and the data analysis to determine conformational epitopes as shown in Figure 1 . After deep sequencing, a relative binding term for each mutant is derived from the change in frequency of the bound population compared with the unselected population 25 . These binding terms are often visualized as heatmaps were a positive value represents a beneficial mutation, a value of zero represent a neutral activity, and a negative v alue represents a deleterious mutation (Figure 1 e ) . Then, t he Shannon entropy, a measure of sequence conservation, is calculated for each position. First, we remove the structurally conserved positions. These positions have an entropy less than or equal to the midpoint for the displayed population. Next, we map th e conformational epitope by discriminating between conserved and non - c onserved positions. A conserved or epitope position will have the Shannon entropy of the bound population less than or equal to the midpoint while non - conserved positions will hav e an en tropy value higher than the midpoint. 1.2.4. Next - g eneration sequencing in antibody discovery applications I n addition to academ ic research efforts, biopharmaceutical companies are focused on improv ing the techniques used to discover human mAbs for therap eutic applications 26 . In the conventional approach of hybridoma screening, a mouse i s injected with the target antigen, follow by the collection of produced B cells and fusion with myeloma cells to produce hybridoma s . 9 Although this technology preserves the natural antibody DNA sequence and pa i ring, the screening process remains inefficient 27 . Alternative approaches are now available through the recent adva nces offered by next - generation sequencing 26 . One example is the comprehens ive antibody discovery pipeline based on a parallel single cell platform created by the biopharmaceutical company GigaGen. Their workflow combines microfluidic methods and multiplex PCR to obtain natively paired heavy and light chain single - chain variable fragment ( scFv ) libraries from isolated B cells, follow by yeast display of s cFv libraries with FACS and deep se quencing t o identify clones with the highest affinity 28 31 . Appendix B present s a developed method to screen humanized mouse - derived yeast scFv libraries using recombinant OX40 protein ( tumor necrosis factor receptor superfamily member 4, TNFRS4 ) in cell lysate 31 . Their novel pipeline was used to evaluate two method s f or mice immunization and two methods for the selection of OX40 agonists. This project was performed at GigaGen in San Francisco, California as an industrial internship during Summer 2018. While I was conducting my graduate internship at this company, I was responsible for the scree ning of humanized mouse - derived yeast scFv libraries using recombinant OX40 protein in cell lysate and with soluble OX40 protein. Then, deep sequencing was used to compare the pre - and post - sorted yeast scFv libraries obtained fro m both selection methods. After that, was responsible for the expression and characterization of forty - one enriched monoclonal antibodies , and further experiments including cell surface binding, kinetics assays, and in vitro activity. This s tudy demonstrated that the cell lysate selection methods yielded OX40 antibodies with higher therapeutic potential compared to the antibodies enriched by the soluble selection method 31 . 10 In the following section s , we will show the successful application of deep sequencing - guided engineering methods to address numerous aims relevant to the protein engineering and antibody di scovery field. These methods were applied to improve the understanding of protein sequence - f unction relationships, antibody - epitope mapping, development of antibody therapeutics, and other end uses. The mutational - point analysis obtained from deep sequenci ng was suitable to identify nearly all beneficial point mutants in a protein in a simplified workflow and to find the optimal conditions for multiple applications. These contributions prove that effective strategies exist to overcome significant limitation s in the field of deep mutational scanning . 11 CHAPTER 2: 2. Feline Interleukin - 31 shares overlapping epitopes with Oncostatin M Receptor and IL - 31RA 12 2.1 . A bstract Interleukin - 31 (IL - 31) is a major protein involved in sever e inflammatory skin disorders. Its signaling pathway is mediated through two type I cytokine receptors, IL - 31RA (also known as gp130 - like receptor) and Oncostatin M receptor (OSMR). Understanding m olecular details in these interactions is crucial for the d e velopment of antagonist anti - IL - 31 monoclonal antibodies (mAbs) as potential therapies. Previous immunoprecipitation studies suggest that human IL - 31 binds to IL - 31RA and then recruits OSMR to for m a ternary complex. In this model, OSMR cannot interact wi t h IL - 31 in the absence of IL - 31RA. In this work, we show that feline IL - 31 (fIL - 31) binds independently with feline OSMR using surface plasmon resonance, ELISA, and yeast surface display. Moreover , competition experiments suggest that OSMR shares a partia l ly overlapping epitope with IL - 31RA. To map the binding sites of both receptors on fIL - 31, we used deep mutational scanning combining comprehensive mutagenesis of yeast surface displayed fIL - 31, f luorescence activated cell sorting, and deep sequencing. Th e constructed binding site for IL - 31RA contains fIL - 31 positions E20 and K82, while the binding site for OSMR comprises the ith the previous sites identified for human IL - 31 and its r e ceptors determined by computational analysis and alanine scanning mutagenesis. However, our results also revealed a new overlapping site, composed of positions R69, R72, P73, D76, D81, and E97, be tween both receptors to which we shared site T he conformational epitope of an anti - feline IL - 31 mAb that inhibits both OSMR and IL - 31RA binding and signaling also mapped to this shared site. Together, our results show conclusively that in fe lines , IL - 31 binds IL - 31RA and OSMR independently through a partially shared epitope. These results suggest reexamination of the putative canonical mechanisms for IL - 31 signaling in higher animals. 13 2.2. Introduction Cytokines comprise a large family of sma ll proteins that play a critical role in the development an d control of the immune response. Certain cytokines are associated with the initiation and the persistence of pathological pain behavior including nerve and skin injuries. A more recently discovere d cytokine, interleukin - 31 (IL - 31), has been linked to the i nduction of chronic skin inflammation 32 . Human and murine data have shown high expression of IL - 31 associated with severe inflammatory skin disorders including pruritis, alopecia, skin lesion, and atopic dermatitis (AD) and with other regulated allergic dise a ses such as asthma 32 38 . An experi mental animal model for human AD reported a strong correlatio n between itch - associated scratching behavior in NC/Nga mice and expression of IL - 31 mRNA 36 . Elevated IL - 31 serum levels were found in adult pat ients with AD compare to healthy control subjects 39 and in pediatric patients during AD flare and quiescence 40 . Together, these data suggest that IL - 31 represents an important target for the development of treatments against such skin inflammatory diseases. Accordingly, antagonist anti - IL - 31 mAbs are current l y in development for human health 41,42 and animal health 43 . For example, an anti - hIL - 31RA mAb, CIM331, binds to IL - 31RA, inhibits IL - 31 signaling a nd reduces severe pruritus 42 . In veterinary medicine n ized anti - IL - 31 mAb, Lokivetmab, showed efficacy in clin ical trials for canine pruritis and is currently approved as an AD therapy for dogs 43 . IL - 31 is a member of the IL - 6 cytokine superfamily produced preferentially by T helper type 2 cells 32 . Mature human IL - 31 (hIL - 31) is composed of 141 amino acids 32 with a predicted topology of four antiparallel helices 20 . The IL - 31 signaling pathway is thought to be mediated through a gp130 - like type 1 cytokine receptor (IL - 31RA , also known as GPL ) and oncostatin M receptor (OSMR) 20,32,44,45 . Both receptor s belong to the type I cytokine receptor family which shar e 14 a common cytokine bindin g domain (CBD) formed by two fibronectin type III - like domains 46 . Previous studies supplied immunoprecipitation evidence that human IL - 31RA (hIL - 31RA) bin ds directly to hIL - 31. In these same studies immunoprecipitation results failed to detect direct human OSMR (hOSMR) binding to hIL - 31 20,44 . However, an increase in binding was distinguished when hIL - 31RA and h OMSR were combined , suggesting that hIL - 31 binds first to hIL - 31RA, at which time h OMSR is recruited to form the ternary c omplex 20,44 . In this model, the ternary complex activates numerous downstream signali ng pathways 20,32,44,47,48 ( Figure 2a ). Figure 2 . Interactions between IL - 31 and its receptors. Adopted from Le Saux et. al 20 . The cartoon model of IL - 31 structure was generated using the felin e s equence. ( a. ) The ternary complex formed between IL - 31, IL - 31RA, and OSMR. D ifferent domains of both receptors are indicated in the figure legend. ( b. ) Binding sites for the interaction between human IL - 31 (hIL - 31) and its receptors as described by Le S a ux et at 20 . Residues corresponding to the hIL - 31 sequence are ind i cat ed in parent hesis. S pecific atomistic knowledge of the binding interactions between IL - 31 and its receptors is crucial for the development of antagonist therapeutic mAbs. Based on the structure of IL - 6/ IL - - Receptor/gp130 c omplex 49 , t he IL - 6 cytokine superfamily family is thought to interact with their receptors through t hree different contact binding sites (sites I, I I, and III) . Le Saux, et al . used 15 computational analysis and sparse alanine scanning to delineate sites II and III only as critical binding sites for the interaction between hIL - 31 and its receptors 20 . In particular, Glu44, Glu106, and His110 were identified as critical res idues for binding site II while Lys134 was identified within the binding site III ( Figure 2b ) . The aim of this study is to gain insights into the interactions between fe line IL - 31 (fIL - 31) and its feline receptors fOSMR and fIL - 31RA and to map the conform ational epitope for an anti - fIL - 31 mAb (mAb #1) . In contrast to previous studies conducted with homologs, we show through multiple biophysical methods that fOSMR directly binds fIL - 31 and partially interferes with fIL - 31RA binding. We identified the potent ial binding sites for fOSMR, fIL - 31RA, and an ant i - fIL - 31 mA b (mAb#1) using a predicted fIL - 31 structure model combined with fine epitope mapping 24 using yeast surface display 17 , nicking mutagen esis 23 , and deep sequencing 50 . The c onstructed binding sites agree d with the sites previously found by L e Saux et al 20 an d showed an additional overlapping site between both receptors to which we terme . Finally, mAb#1 also bound this shared site between both receptors demonstrating its potential to inhibit the signaling pathway. Together, these results s uggest that the current model for IL - 31 mediated signaling in higher mammals is incomplete and suggest efficient therapeutic strategies to antagonize IL - 31 - mediated signaling. 2.3. Materials and methods 2.3.1. Strains T he Escherichia coli strain used in t his study was XL1 - Blue (Agilent, Santa Clara, CA ) endA1 supE44 thi - 1 hsdR17 recA1 gyrA96 relA1 lac [F ' proAB lacI1 q Z Li M15 Tn10 (Tet r )] . The Saccharomyces cerevisiae strain used in this study was EBY100 (American Type Culture Collection, Manassas, VA) MATa AGA1::GAL1 - AGA1::URA3 ura3 - 52 trp1 leu2 - delta200 his3 - 16 delta200 pep4::HIS3 prb11.6R can1 GAL . 2.3.2. Plasmid Construct s p ETconNK_fIL - 31 was created by inserting a codon - optimized gene encoding the Ser1 Gln136 from the mature portion of feline interleukin - 3 1 (fIL - 31) (GenScript, Piscataway, NJ) into yeast display vector pETconNK 23 (Addgene plasm i d # 81169 ) using standard restriction cloning. Sequences were verified by Sanger sequencing (Genewiz, South Plainfield, NJ). The full sequence is list ed in Note C 1 . 2.3.3. Preparation of f IL - 31, f OSMR, f I L - 31RA constructs, and mAb#1 F IL - 31 was produced rec o mbinantly in E. coli and CHO cells with a C - terminal His - tag and purified by Ni - NTA affinity chromatography . Other proteins were prepared as Fc fusions and produced in mammalian cell culture and purified by protein A affinity chromatography. Each protein w as prepared in phosphate - buffered saline (PBS) at a concentration of at least 0.082 mg/mL by Zoetis. Proteins were biotinylated at a molar ratio of 1:20 protein: biotin using the EZ - link NHS - l ogies, Carlsbad, CA). Biotinylated proteins were desalted using Zeba Spin desalting columns (Thermo Fisher, W altham, MA) and stored at 4ºC. 2.3.4. Analytical Size Exclusion Chromatography S ize exclusion chromatography was performed using a TSK SuperSW3000 4.6 x 30 mm gel permeation column. 50 g of fIL - 31 was injected at a flow rate of 0.25 mL/min for 25 minutes with a mobile phase of 200 mM sodium phosphate, pH 7.2. 17 2.3.5. Surface Plasmon Resonance binding assays S urface Plasmon Resonance was performed on a Biacore T200 (GE Healthcare, Pittsbu rgh, PA) to measure binding affinities of fOSMR - ECD and mAb#1 to fIL - 31 . fIL - 31 immobilization on CM5 sensor and the binding measurements were conducted as previously described 51 . Data was analyzed with Biacore T200 Evaluation software by using the method of double referencing. The resulting curve was fitted with the 1:1 binding model. 2.3.6. ELISA T he plate was coated with feline IL - 31 overnight at 4°C in carbonate - bicarbonate buffer, pH 9.6 (Sigma - Aldrich, St. Louis, MO) follow by a blocking step with 5% skim milk in PBS with 0.05% TWEEN 20 for 1 hour at room temp. I ndividual proteins were diluted at different concentrations in blocking buffer and were added to the coated plate for 2 hrs at room temp. Next, HRP conjugated secondary antibodies were added to appropriate wells at 1:10000 in blocking buffer for 1 hr at ro om temp (goat anti - human Fc for receptors, jackson goat anti - cat Fc for mAb#1, KPL anti - mouse). Finally, KPL SureBlue TMO (VWR) was added to develop absorbance at 450 nm. 2.3. 7 . Yeast Surface Display Expression and Binding Activity M ean fluorescence intens ities (MFI) were measured using a BD Accuri C6 flow cytometer. The expression of fIL - 31 on the yeast surface was detected using anti - c - myc - FITC (Miltenyi Biotec, San Diego, CA) and the binding interaction with biotinylated proteins was detected using strep tavidin - R - phycoerythrin conjugate (Thermo Fisher ) . Dissociation constant (K D ) values were determined according to Chao et al. 17 Titrations were performed at triplicates on at least two separate days and MFI values were used to calculate the experimental K D using one site specific binding equation (Hill c oefficient of 1) in Grap h Pad Prism software. Labeling concentrations 18 tested vary from 0.064 nM to 262.1 nM for f OSMR and mAb#1, and from 2.05 nM to 524.3 nM for f IL31RA - 1FNIII. 2.3. 8 . Competition Binding Assays F or YSD 1x10 5 yeast cells were labeled at twelve times their K D val ues with either non - biotinylated f OMSR or non - biotinylated mAb#1 for 30 mins at room temperature in PBS with 1g/liter of bovine serum albumin (PBS - BSA). After spinning down and washing with 200 l of P BS - BSA, cells were labeled at 1x K D values with either biotinylated mAb#1, f OSMR, or f IL31R A - 1FNIII for 30 mins at room temperature in PBS - BSA. Then, after a second step of - cymc - FITC streptavidin - R - phycoerythrin conjugate in 49.15 l of PBS - BSA for 10 mins at 4°C. Cells were washed thoroughly with PBS - BSA and read on a flow cytometer. 2.3.9. Cell - based assays E . coli - derived or CHO - derived fIL - 31 at various conc entrat ions was incubated with the adherent feline macrophage cell line FCWF - 4 (ATCC CRL - 2787) for 1 hr at 37 o C. Activation of pSTAT signaling was determined using an AlphaLISA SureFire Ultra pSTAT3 kit (PerkinElmer). For inhibition experiments, fOSMR - ECD w as pre - incubated with 10 g/mL fIL - 31 for 1 hr at 37 o C prior to stimulation before testing for activation of pSTAT signaling. 2.3. 10 . Preparation of Mutagenesis Libraries T wo comprehensive single - site saturation mutagenesis libraries were constructed using nicking mutagenesis as described 23 . All igned using Quick Change Primer Design Pr ogram (www.agilent.com) and were ordered from Integrated DNA Technologies (Coralville, IA). Library 1 covered residues Ser1 Phe68 and library 2 covered Arg69 Gln136 of the mature fIL - 31. mid was transformed into 19 chemically compe tent Saccharomyces cerevisiae EBY100 cells, and cells were grown and stored at a concentration of 1x10 7 cells per ml in yeast storage buffer at - 80°C according to published protocols 24 . 2.3. 11 . Determination of receptor binding sites and conformational epitope of mAb#1 T he library screening through FACS and deep sequencing preparation was performed exactly as previously d escribed 24 . The sorting wa s done on an Influx Cell Sorter at the Michigan State University Flow facility. After prepari ng the plasmid DNA for deep sequencing, libraries were pooled and sequenced on an Illumina MiSeq using 2 x 250bp pair - end reads at the Michigan State University Ge nomic Sequencing Core facility or the University of Illinois at Chicago DNA Service facility. Appendix E contains the sorting conditions ( Table E 1 ), the primers used for deep sequencing ( Table E 2 ), and the summary table of statistics ( Table E 3 ). 2.3.1 2 . Data Analysis A modified version of Enrich 0.2 software as described in Kowalsky et al. 25 was used to compute enrichment ratios from i ) for variant i was defined as: , 2.1 where is the mean fluorescence of variant i and is the mean fluorescence of wild type fIL - 31. This equation can be written in terms of experimental observables according to: 2.2 where is t he log normal fluorescence standard de viation of the clonal population, enrichment ratio and is the percentage of cells collected by the sorting gate on the flow 20 cytometer. Custom python scripts available at Github (user: JKlesmith) were used t o normalize the relative fluorescence, and to calculate the Shannon Entropy and overall statistics 12 . 2.3.1 3 . Data Availability R aw sequencing reads for this work have been deposited in the SRA (SAMN11289369 72, 7 9, 81 - 83, 90 - 91, SAMN11289422 - 23 ). 2.4. Results 2.4.1. f OSMR and f IL - 31 R A bind independen tly to fIL - 31 W e assessed the binding activity of fIL - 31 against soluble f OSMR extracellular domain (fOSMR - ECD), four versions of soluble fIL - 31RA (fIL31RA - CBD comprising the CBD only, two fIL31RA - 1FNIII iso forms comprising CBD and 1 FNIII domain, and fIL3 1RA - 2FNIII comprising CBD and two FNIII domains, Figure 3 a ), and an anti - fIL - 31 mAb (mAb#1) ( Figure 3 ) . All receptors were expressed as fusion proteins with a C - terminal Fc. Recombinant proteins were produce d by transient expression in CHO cells and purifi ed by affinity chromatography using Protein A resin . fIL - 31 was at the expected molecular weight ( Suppl Figure D 1a ) after PNGase treatment. fIL - 31 was monomeric with a small dimeric peak as judged by analytical gel permeation chromatography ( Suppl Figure D 1b ). We first assessed the ability of f OSMR - ECD and fIL31 - R A to independently recognize soluble fIL - 31 using ELISAs. fIL - 31 was coated and then receptors or mAb were incubated at indicated concentrations. After washing, receptors were detected with an ant i - human IgG. fOSMR - ECD bound fIL - 31 in a dose dependent fashion, whereas the fIL - 31RA CBD alone did not recognize fIL - 31 at any concentration tested ( Figure 3b ). However, both fIL31RA - 1FNIII isoforms recognized fIL - 31. We also tested mAb#1, which could als o bind fIL - 31 ( Figure 3b ). S urface plasmon resonance (SPR) measurements using immobilized fIL - 31 showed similar 21 binding results for fOSMR - ECD and mAb#1 . ( Figure 3c ). Binding equilibrium measurements were determined using kinetic fitting with 1:1 binding mo de, revealing low - nM affinities for both fIL - 31 - fOSMR - ECD and fIL - 31 - mA b #1 interactions. Figure 3 . Binding profiles of feline IL - 31 (fIL - 31) with all different constructs of fIL - 31RA, fOMSR - ECD, and mAb#1 . ( a. ) f IL - 31R A and fOSMR - ECD constructs tested in this work. ( b . ) Surface plasmon resonance sensograms and ( c . ) ELISA plot s showing the binding of fIL - 31 with all constructs. Done by Zoetis . ( d. ) Flow cytograms showing the increase in fluorescence in the fIL - 31 binding channel wi t h all constructs. ( e. ) Binding titrations curves for yeast surface - displayed fIL - 31 with mAb#1, fOSMR - ECD , and fIL31R A - 1FNIII. Titrations were performed at W e also measured these intermolec u lar interactions usin g yeast surface display (YSD) , where fIL - 31 was displayed on the yeast surface with an N - terminal yeast surface protein Aga2p and a C - terminal c - myc epitope tag. Saturating amounts of biotinylated soluble fIL - 31RA constructs, fOSMR - EC D , and mAb#1 were incu bated with yeast cells displaying fIL - 31, followed 22 by secondary labeling ( Figure 3 d ) . fOSMR - ECD bound fIL - 31 with nanomolar affinity ( Figure 3 d - e ). While f IL31RA - CBD did not bind, both fIL31RA - 1FNIII isoforms and fIL31RA - 2FNIII were a b le to recognize fIL - 31, consistent with ELISA and SPR data. Given that the fIL31RA - 1FNIII isoform obtained the highest mean fluorescence intensity (MFI) under saturating amount s of receptor, we used this receptor for the remainder of this work. Dissociati o n constants ranged from 0.26 ± 0.01 nM for mAb#1 to 106 ± 2.7 nM for IL31R A - 1FNIII ( Table 1) . For all interactions including mAb #1, the binding on the yeast surface could be m odeled using 1:1 binding kinetics (best fits of Hill coefficient 1.0 ± 0.04), consistent with fIL - 31 and receptors being mostly monomeric. Table 1 . Summary of experimental determined dissociation constants (KD) using yeast surface display and surface plasmon resonance. Error bars represent standard error o f at least 3 independent meas u rements. ND, not determined. Average K D values [nM] Name Yeast Surface Display Hill Coefficient Surface Plasmon Resonance Fit f OSMR - ECD 10.3 ± 0.5 0.99 ± 0.04 3.8 1:1 fIL31RA - 1FNIII 10 6 ± 2.7 1.00 ± 0.01 ND 1:1 mAb # 1 0.2 6 ± 0. 01 0.99 ± 0.04 3.1 1:1 2 .4.2. fOSMR partially, but not completely, inhibits binding of fIL31RA - 1FNIII to fIL - 31 T o determine whether the receptors bind to independent sites on fIL - 31 or share overlapping sites we performed c ompetition binding assay s using SPR and YSD. First, we set - up an SP R - based competitive binding assay where fIL - 31 is bound and then either fOSMR - ECD or mAb#1 is captured at saturating amounts. Next, either fOSMR - ECD or mAb#1 is injected. If receptor and mAb bind at different epit o p es then there should be no difference in response. When fOSMR - ECD was capture on the surface, mAb#1 binding signal decreased indicating some degree of competition. Similarly, the captured mAb#1 decreased the binding signal of fOSMR - ECD by approximately 5 0 % ( Figure 4 a ), indicating at least partial ly overlapping binding footprints. 23 Next, using our YSD set - up, fIL - 31 yeast cells were labeled with either non - biotinylated fO S M R - ECD or mAb#1 at 12x their experimentally determined K D values, washed, and subsequ e n tly labeled with either biotinylated fOSMR - ECD, fIL31RA - 1FNIII or mAb #1 at 1xK D . At such labeling concentrations and considering 1:1 binding kinetics, it was expected that the non - biotinylated proteins inhibit the binding signal of the biotinylat ed prot e i ns by approximately 93% if proteins occupy overlapping binding sites and 0% if binding sites are completely non - Figure 4 . Competition binding assays demonstrated that fOSMR - ECD, fIL - 31RA, and mAb#1 share an overlappi ng epitope using SPR and YSD. ( a. ) SPR sensograms showing a decrease in the binding signal when fOMSR - ECD and mAb#1 were capture d on the surface. Done by Zoetis . ( b . ) Percentage of inhibition for fIL - 31 yeast cells labeled with either non - biotinyla ted mAb# 1 or non - biotinylated f OSMR - ECD at twelve times their respective Kd values and subsequently labeled with mAb#1, f OSMR - ECD , and f IL31R A - 1 FNIII at Kd values. (error bars, o verlapping. The non - biotinylated mAb #1 decr eases binding signal of itself, fOSMR - ECD, and fIL31RA - 1FNIII by at least 90% ( Figure 4 b ) . These results demonstrated that mAb#1 binds an 24 overlapping epitope or otherwise sterically prevents binding to the two receptors. Non - biotinylated fO S M R - ECD decreases binding to fOSMR - ECD and fIL31RA - 1FNIII by 97% and 65%, respectively ( Figure 4 b ). The c ollected data from both measurements suggests that fOSMR - ECD and fIL31RA - 1FNIII share partially overlapping binding epitopes on fIL - 31 and that mAb#1 blocks both receptors from binding. 2.4.3. fIL - 31 can be specifically inhibited by fOSMR - CBD in cell - base d assays T o determine whether recombinant fIL - 31 could functi on in cell - based signaling, we incubated varying amounts of either E. coli - derived or CHO - derived fIL - 31 with the feline macrophage cell line FCWF - 4 and determined pSTAT signaling using AlphaLISA pSTAT3 kit. Both recombinant IL - 31s activated pSTAT signalin g with an EC 50 < 0.1 g/mL (Figure 5 a) , demonstrating fIL - 31 functionality. To determine whether fOSMR - CBD can inhibit IL - 31, we first pre - incubated receptor with 10 g/ mL fIL - 31 for 1 hr at 37 o C prior to stimulation. Increasing amounts of fOSMR - CBD could directly block fIL - 31 - mediated pSTAT signaling (Figure 5 b ) , indicating that blocking the IL - 31 OSMR - binding epitope is sufficient to block signaling. 2.4.4. Structural homology model for fIL - 31 M apping the fIL - 31 binding sites and the conformational epitope requires a reasona bly accurate homology model. Given th at there is no IL - 31 crystal structure available, we generated our initial models using I - TASSER 52,53 . While all models contained the predicted four helix up - up - down - down topology consistent with a previous stru ctural model from LeSaux et al. 20 , none were likely to be complete ly accurate. All models contained several buried hydrophilic loop residues and surface exposed hydrophobic residues, and only some models properly p aired the disulfide bond between Cys49 - Cys132 (data not shown). Based on the initial set of models, we re - ra n ITASSER using additional constraints based on putative hydrophobic contacts and requiring a 25 distance restraint of 2Å for Cys49 - Cys123. Then, we us ed Rosetta to identify the lowest - scoring variant and refine all atoms of the structure. The resulting homol ogy model of the mature protein has a mostly hydrophobic core formed from the four antiparallel helices ( Supp Figure D 2 ) . Molprobity analysis 54 shows a structure with minimal clashes, over 99 % favored rotamers, and only 2 Ramachandran outliers (Glu41, Ser42) on a loop covering the interface between the A and D helix. Whil e there will certainly be deviations between this model and the true structure, we judged this model sufficient to evaluate epitope mapping experiments. Figure 5 . Effect of fIL - 31 to induce STAT3 phosphorylation in macrophage cel l line FCWF - 4 ( a. ) and f OSMR - ECD inhibition of pSTAT signaling induced by fIL - 31 ( b. ) . Done by Zoetis . 2 .4.5. YSD and satur at ion mutagenesis reveals partially overlapping binding sites for f OSMR and fIL - 31R A W e determined comformational epitopes on fIL - 31 for its binding partners using our previously published method combining yeast surface display, nicking mutagenesis, and de ep sequencing 12,24 . The general concept behind our method is that mutations that result in loss of binding will predominantly map to the epitope. To that e n d, we created two comprehensive s ingle - 26 site saturation mutagenesis (SSM) libraries for fIL - 31 using nicking mutagenesis 23 , transformed these libraries into S. cerevisiae EBY100, and deep sequenced the population. We observ ed an average coverage of 79.5 % for every possible single missense and nonsense nonsynonymous substitution (2,161 out of 2 ,7 2 0 mutations; Supplementary Tab le E 3 ). The two SSM fIL - 31 libraries were expressed on the surface of yeast and each labeled with biotinylated fOSMR - ECD, fIL31RA - 1FNIII or mAb #1 at half of the experimentally determined dissociation constant ( Supplementar y T able E 1 ). Libraries were sorted by FACS into two distinct populations: one population corresponding to approximately the top 7% by fluorescence for the channel corresponding to the biotinylated protein (bound population), and a reference population of y ea s t cells that passed throu gh the cell sorter. Plasmid DNA was extracted and deep sequenced. For each variant we calculated the relative fluorescence values based on the change in frequency between the bound and reference populations 12 . We also determined the per - p os i tion Shannon entropy (SE) , a measure of sequence conservation, in order to determine the epitope. Summary statistics and complete per - position fIL - 31 heatmaps are given in Table E 3 and Figures D 3 - D 5 , respectively . A subset of the heatmaps for both recep to r s and mAb#1 are shown in Figure 6 - 8 . Positions with less than 25% accessible surface area (ASA) are removed from analysis, as mutations at these positions often result in misfolded protein. Applied to the structural model of fIL - 31, this analysis remove d 6 0/136 positions; we define these posi tions as structurally conserved. The SE values for these positions are significantly lower than for surface exposed positions (2.17 vs. 1.78 for fOSMR dataset, 2.24 vs, 1.93 for fIL31R A - 1FNIII dataset, and 2.30 vs. 2 .0 0 for mA b #1, p - values for all < 10 - 5 ). Posit ions with insufficient data for more than ten mutations were also excluded from the analysis, leaving 68 fIL - 31 positions for further analysis. Surface exposed residues with lower then midpoint SE values were de emed epi tope positions. Of the remaining 27 non - conserved positions, ones with the highest 10% SE were classified as completely non - conserved. Figure 6 . Determination of the fIL31RA - 1FNIII binding site using deep sequencing. Shann on entropy with its respective cut - off (dashed lines) is plotted below the heatmap as well as the structural homology model with the determined binding footprint . F igure 6 shows a restricted per - position heatmap and IL - 31 structure for the fIL31RA - 1FNIII i nt eraction . 16 IL - 31 residues were identified as belonging to the fIL - 31RA epitope. These residues form a semi - contiguous patch surrounding the previously described site II from Le Saux et al for hIL31 20 . In particular, E20 on helix A and K82 on helix C were 2/3 previously identified epitope positions on Site II from a la ni ne scann ing experiments by Le Saux et al. The third identified residue, Q86, does show reduced binding upon alanine mutation, although there is insufficient data for other mutations at that positions to make a definitive epitope determination. 28 Unexpect ed ly , an adj acent contiguous patch on helix B (R69, R72, P73), BC loop (D76), helix C (D81) and CD loop (E97) also appears part of the f IL - 31RA epitope, as most mutations at these positions were strongly depleted in the bound library. Other positions ident if ie d as bel onging to the epitope (L19, L50, and I59) are discontinuous with the rest of the epitope positions and may represent structural conserved positions that reflect inadequacies with the structural model. Figure 7 . Determi nation of the fOSMR - ECD binding site using deep sequencing . Shannon entropy with its respective cut - off (dashed lines) is plotted below the heatmap as well as the structural homology model with the determined binding footprint. F igure 7 shows the f OSMR - fI L 3 1 heatma p and structural model. Overall, 17 positions mapped to the binding site while 28 were completely non - conserved. The perceived epitope covers the expected patch on site III, including G38 and K110 previously identified by Le Saux et al. 20 for the hIL31 - hOSMR interaction. This epitope is characterized by strong bi nding at the 29 - K110) at the beginning of the D helix. However, we also observed low SE and thus strong conservation for the same contiguous patch ( R69, R72, P73, D76, D81, E97) as for the fIL - 31RA interaction. We deemed this region t he shared site. Although the sequence entropy of L50 is slightly above the cutoff, the discontinuous L19, L50, and I59 residues observed in the fIL - 31RA binding maps are still conserved in this structure and most likely represent structurally conserved re si dues. Figure 8 . Determination of the mAb#1 conformational epitope using deep sequencing. Shannon entropy with its respective cut - off (dashed lines) is plotted below the heatmap as well as the structural homology mod el with the d et ermined binding footprint. 2 .4.6. mAb#1 conformational epitope reveals nature of inhibition I n our final approach, we determined the conformational epitope for the anti - fIL - 31 mAb #1 using the same procedure. 20 positions mapped to the epitope ( Fig 8, F ig D5 ). mA b#1 binds to all the positions within the shared site between fOSMR and fIL - 31RA as well as adjacent positions K77 and N78. However, high SE was observed at both canonical site II (E20, K82) and site III 30 (G38, K110) positions, su ggesting that mAb #1 does no t directly inhibit either canonical IL6 - like receptor binding sites. Based on these results, mAb#1 potentially inhibits the signaling of both receptors through the shared site. 2.5. Discussion and conclusion In this study we have explor ed the bi oc hemistry for intermolecular interactions between fIL - 3 1 , its receptors , and an anti - fIL - 31 mAb . Additionally, we used our established deep mutational scanning pipeline with an improved predicted fIL - 31 atomic structure to map the bind ing sites of fIL - 31 re ceptors and the conformational epitope of a potential mAb on fIL - 31 . We found that, as expected, Site II residues contributed to fIL - 31RA recognition while Site III residues were important in f OSMR binding. However, there were two rel ative surprises resu lt ing from the current work. First, in c ontrary to previous results with human orthologs 20,44 , our work shows that fOSMR - ECD can directly recognize fIL - 31 without the presence of fIL - 31RA. Importantly, we used three different independent biophysical methods for assessing binding. We also show that fIL - 31 used in binding experi me nts is f unctional in cell - based assays and can be specifically inhibited by fOSMR - ECD. Taken together, the biochemi cal evidence for this interaction is unambiguous. However, we note that our experiments were all performed with feline IL - 31 and receptors; t hus, it remains to be seen whether there are species - specific differences between IL - 31 signaling pathways. In part icular, given that the lack if interaction between human OSMR and IL - 31 was determined by immunoprecipitation, it would be of interest to p er form mor e stringent biochemical validation on the human orthologs to assess whether the mechanism of interaction is conserved across higher animals. Also importantly, we did not establish in the present work exactly how signaling occurs are both OSMR a nd IL - 31RA required in felines, or is binding to one 31 receptor sufficient? Figure 9 . Specific binding sites revealed for fIL31RA - 1FNIII and fOSMR - ECD. The shared site (orange) is composed of B - helix, BC loop, C - helix, an d CD loop . The binding site for fIL31R A - 1FNIII (green), site II, includes A helix and C helix. The binding site for fOSMR - ECD (purple), site III, comprises the N - terminal of AB loop and D - helix. S econd, we found that fOSMR and fIL 31 - RA could compete for bi n ding on fIL - 31, and fine epitope mapping using yeast display pipeline revealed a contiguous patch we deemed the shared site between both receptors ( Figure 9 ). There are two cavea ts with an unambiguous determination of this shared site as an epitope for O SMR and IL - 31RA recognition. First, we relied on a homology model of IL - 31. Thus, it is possible that these shared site positions may not be surface exposed in the monomer. However , we think this unlikely given the contiguous patch revealed by the epitope mapping and the relative hydrophilicity of the involved side - chains. Second, although fIL - 31 was mainly produced as a monomeric protein, size exclusion chromatography shows a small percentage of recombinant fIL - 31 is dimeric. Such results lead us to ask w h ether th is shared site is exposed on the surface or if it is buried in the interface between 32 each monomer. While unlikely, this is a real possibility and points to the limitations of fine epitope mapping without a high resolution experimentally determined structur e. For over the past few years, IL - 31 and its signaling pathway have been identified as one of the central causes of severe inflammatory skin disorders. Fine epitope infor mation presented here may lead to the development of antagonist mAbs that i n hibit th e downstream signaling pathway. 33 CHAPTER 3 3 . Pro region engineering of nerve growth factor by deep mutational scanning enables a yeast platform for conformational epitope mapping of anti - NGF monoclonal antibodies Porti ons of this chapter were adapted from Pro region engineering of nerve growth factor by deep mutational scanning enables a yeast platform for conformational epitope mapping of anti - NGF Biotech nology and Bioeng ineering ( 2018 ) 115:1 925 - 1937 34 3 .1 . Abstract Nerve growth factor (NGF) plays a central role in multiple chronic pain conditions. As such, anti - NGF mAbs that function by antagonizing NGF downstream signaling are leading drug candidates for non - opioid pain relief. To e val uate ant i - canine NGF (cNGF) mAb s we sought a yeast surface display platform of cNGF. Both mature cNGF and pro - cNGF displayed on the yeast surface but bound conformationally sensitive mAbs at most 2.5 - fold in mean fluorescence intensity above background, su ggesting that cNGF was mostly m isfolded. To improve the amount of folded, displayed cNGF, we used comprehensive mutagenesis, FACS, and deep sequencing to identify point mutants in the pro - region of canine NGF that properly enhance the folded protein dis pla yed on t he yeast surface. Out o f 1,737 tested single point mutants in the pro region, 49 increased the amount of NGF recognized by conformationally sensitive mAbs. These gain - of - function mutations cluster around residues A - 61 P - 26. Gain - of - function mu tan ts were additive, and a constru ct containing three mutations increased amount of folded cNGF to 23 - fold above background . Using this new cNGF construct, fine conformational epitopes for tanezumab and three anti - cNGF mAbs were evaluated. The epitope re vea led by t he yeast experiments largely overlapped with the tanezumab epitope previously determined by X - ray crystallography. The other mAbs showed site - specific differences with tanezumab. As the number of binding epitopes o f functionally neutralizing ant i - N GF mAbs on NGF are limited, subtle differences in the individual interacting residues on NGF that bind each mAb contribute to the understanding of each antibody and variations in its neutralizing activity. These results de monstrate the potential of deep se quencing - guided protein engineering to improve the production of folded surface - displayed protein , and the resulting cNGF construct provides a platform to map conformational epitopes for other anti - neurotrophin mAbs. 35 3 .2. Introduction Nerve growth fact or (NGF) wa s the first discovered member of the neurotrophi n family , which also includes brain - derived neurotrophic factor (BDNF), neurotrophin - 3, and neurotrophin - 4. This family of proteins regulates the development, functio n, and survival of neurons in t he peripher al and central nervous systems 22,55 . Neurotrophins activate downstream signaling pathways by binding the pan - neurotrophin receptor, p75NTR , and to the family of tropomyosin recep tor kinase s ( TrkA, TrkB, and Trk C ) with various affinities . NGF is synthesized as a pre - pro protein. The N - terminal pre sequence is released during translocation to the endoplasmic reticulum while the pro - peptide is often but not always cleaved by proprote in converta ses prior to secretion . The NGF mature domain is approximately 120 amino acids and arranged as noncovalent homodimers where each monomer conformation possess es a cysteine knot created by three disulfide bonds 56,57 . M ultiple stud ies have demonstrated the high levels of NGF during peripheral nerve injury, inflammation, and chronic pain conditions 58 ,59 . As a consequence, researchers have developed anti - NGF monoclonal antibodies (mAbs) as potential medicines to modulate chronic pain and many oth er conditio ns. These mAbs function by interfering with binding to p75NTR and/or TrkA. Humanized Tanezumab 60 is furthest alo ng in a pha se III clinical trial . Comprehensive reviews of clinical studies with tanezumab and others mAb s are found elsewhere 61 ,62 . As these signaling pathways are highly conserved in higher mammals, here we have investi gated a panel of mAbs against canine NGF (cNGF ). Canine NGF was used for analysis as this is of interest to Zoetis Animal Health and varies from human NGF by only three amino acids. An important step in evaluating the neutralizing capacity of antibodies is to determine the epitope s that they target . To that end, yeast surface display 17 is a validated platform to determine 36 fine conformational epitopes for complicated proteins 11,12,19 . Typically, a set of mutants of the target protein is displayed on the surface of yeast, assessed for antibody bindi ng, with lo ss of binding mutants mapping to the epitope. A major limitation of the method is the requirement that the displayed target protei n be in a conformation recognizable by the antibody. This is an issue with cNGF and consistent with a previous stud y s howing t hat the related neurotrophin BDNF displayed on the surface of yeast in a mostly inactive conformation 21 . A previous directed evolution study showed that mutations to the pr o r egion co uld enhance the folding of the related neurotrophin human BDNF in S. cerevisiae 63 . Evidence suggests that neurotrophin pro region s act as chaperone s to assist folding of a mature neurotroph in as they pass through the secretory pathway 64 67 . The pro peptid e i s monome ric and highly flexible as shown by the lack of electron density in a solved structure of a pro NGF complex 68 and biophysical analysis in vitro 69 . Two domains are sufficient to process and express active mouse NG F 7 0 (Box 3 a nd Box 5 , shown for canine and human NGF see Fig 10a ) . Three dibasic sites are proteolytically cleaved during processing of mature NGF through the secretory pathway 65,71 . In the current study we developed a yeast display platform for the production of folded cNGF. We used yeast surface display , saturation mutagenesis, flu ore scence a ctivated cell sorting (FACS), and deep sequencing to identify mutations in the pro - region that enhanced display of folded cNGF. Mutational libraries created using this engineering pipeline revealed new insight into the role of the neurotroph in p ro region. Combinations of mutations yielded constructs with a 23 - fold increase in the signal to noise ratio of display of folded cNGF over background as measured by mean fluorescence intensity (MFI) . This engineered pro - cNGF allow ed us to generate co nfor mat ional ep itope maps of multiple anti - cNGF antibodies . All anti - cNGF mAbs had an overlapping footprint with tanezumab but each had several site - specific differences . This research 37 improves our understanding of sequence - function relationships in pro sequen ces for neu rotrophins and highlights the power of deep sequencing to augment classical directed evolution experimental pipelines 72 . 3 .3 . Materials and Methods 3 .3.1 . Plasmid Constructs p ETconNK_cNGF, pETconNK_Aga2_cNGF, pETconNK_procNGF, and pETconNK 1,2 - cNGF plasmids were prepared by cloning custom codon - optimized genes (GenScrip t, Piscataw ay, NJ) into pETconNK 23 (Addgene plasmid # 81169 ) using standard restriction clon ing . Sequen ces were verified by Sanger sequencing (Genewiz, South Plainfield, NJ), with full sequences listed in Note C2. 3.3.2. Preparation of anti - NGF mAbs T anezumab, a humanized anti - NGF mAb, was expressed recombinantly in CHO cells based upon the p ubl ished se quences 60 on hIgG2/kappa constant regions. This antibody was purified using Protein A resin, dialyzed into PBS, and sterile filtered. Three different caninized antibodies (mAb #1, mAb #2, and mAb #3) were supplied by Zoetis Inc. These antibod ies were also expressed in CHO cells, affinity purified using Protein A resin, and dialyzed into either PBS or 20 mM Na Acetat e, 150 nM N aCl, pH 7.4 buffer. Concentrations were assessed via A280 absorbance using the Edelhoch method and ranged from 1.5 mg/m L to 6.45 mg/mL final concentration. At least 1.5 mg/mL mAbs in PBS were biotinylated at a molar ratio of 1:20 mAb: biotin usi ng the EZ - l ink NHS - Technologies). Biotinylated mAbs were then desalted i nto PBS using Zeba Spin desalting columns o C. 38 3.3.3. TF - 1 Cell P roliferation Assay F unctional potency of antibodies against cNGF was evaluated in a cell proliferation assay utilizing the TF - 1 cell line which expresses human TrkA 73 . TF - 1 cells (American Type Culture Collection [ATCC], Rockville, MD) were maintained in ATCC modified RPMI 1640 m edi um (Life Technologies, Carlsbad, CA) supplemented with 10% FBS (Life Technologies, Carlsbad, CA) and 2 ng/ml recombinant human GM - CSF (R&D Systems Inc., Minneapolis, MN) and incubated at 37 °C with 5% CO 2 . On the day of experiment, TF - 1 cells were wash ed twice wi th DPBS (Life Technologies, Carlsbad, CA) before resuspending in proliferation assay medium: ATCC mod ified RPMI 1640 supplemented with 1% FBS and 10 µg/ml gentamicin. The TF - 1 proliferation assay was performed in 96 well microplates (Corning Inc ., Corning, NY) by incubating 15,000 cells per well with anti - NGF antibodies at concentrations indicated and 2 n g/ml recombinant cNGF. cNGF was generated at Zoetis in stable CHOK1 cells. After a 72 hour s culture period, a CellTiter - GLO luminescent assay ki t ( Promega, Madison, WI) was employed to evaluate the effects of anti - - NGF induced cel lular proliferation. After addition of CellTiter - to a white 96 well Opt iplate (Perkin Elmer, Waltham, MA) before reading luminescence on a Spectromax M5e microplate reader (Molecular Devices, San Jose, CA). Maximal response in the assay is defined proliferation in the presence of cNGF only (no antibody). Minimal re spo nse is d efined as measured proliferation without cNGF. Calculated inhibition (NGF neutralization) values for anti - cNGF antibodies are expressed as a percentage of minimal and maximal responses. The resulting percent inhibition/neutralization data was pl ott ed with GraphPad Prism 7 (GraphPad software, San Diego, CA) for IC50 determination using a 4 - parameter curve fit. 39 3.3.4. Surface Plasmon Resonance S urface Plasmon Resonance was performed on a Biacore T200 (GE Healthcare, Pittsburgh, PA) to measure bi ndi ng affin ities of each antibody to nerve growth factor (NGF in 10 mM Sodium Acetate pH 4 (GE Healthcare, BR - 1003 - 49), 5 µg/ml human NGF (R&D Systems Cat #256 - GF/CF) in 10 mM Sodium Aceta te pH 4 and 1 µg/ml human proNGF (Alomone labs Cat. # N - 280) in 10 mM Sodium Acetate pH 5 ( GE Healthcare , BR - 100 3 - 51 ) was immobilized by amine coupling using EDC/NHS for a final density ~250 RU (resonance unit) on CM5 sensor flow cells 2 - 4, respectively. Flow cell 1 is used as an internal reference to correct running buffer effects. Antibody binding was measured at 15ºC with a contact time of 250 seconds and flow rate of 30 l/min. The dissociation period was 300 seconds. Regeneration was performed with regeneration buffers (10 mM Glycine pH1.5 and 10 mM NaOH) and flow rate at 20µl/min for 60 seconds each. Running/diluti on buffer (1X HBS - EP, GE Healthcare, BR - 1006 - 69, 10X inc luding 100 mM HEPES, 150 mM NaCl, 30 mM EDTA and 0.5% v/v surfactant P20, pH7.4, 1:10 in filtered MQ H2O) was used as negative control at the same assay format. Data was analyzed with Biacore T200 Eva luation software by using the method of double referenci ng. The resulting curve was fitted with the 1:1 binding model. 3.3.5. Yeast Surface Display Expression and Binding C ellular fluorescence was measured using a BD Accuri C6 flow cytometer. Yeast cells displaying cNGF variants were detected using anti - cymc - FITC (Miltenyi Biotec, San Diego, CA) and an anti - FLAG tag alexa fluor 647 - conjugated antibody (R&D System, Minneapolis, MN). Binding to biotinylated mAbs was detected using streptavidin - R - phycoeryt hrin conjugate (Thermo Fisher, Waltham, MA ). Apparent di ssociation constants were determined according to Chao et al . 17 by titratin g mAb at labeling concentrations from 0.064 nM to 262 nM. Titrations 40 were performed in triplicate on at least two separate days. 3.3.6. Preparation of Mutagenesis Libraries C omprehensive single site s aturation mutagenesis (SSM) libraries were constructed u sing nicking mutagenesis exactly as described 23 . All mutagenic oligos were designed using Q uik Change primer design Program (Agilent) and were ordered from IDT (Coralville, IA). For pro - cNGF, two separate libraries were prepared: library 1 covered residues Glu - 102 Asn - 52 and library 2 cov ered residues Ile - 51 Arg0. For pro 1,2 - cNGF, the libra ry covered residues Gln - 55 Arg0. For conformational epitope mapping pro.v 4 - cNGF was split into two libraries, with library 1 covering residues Ser1 Asp60 and library 2 covering residues Pro61 Ala120. Library plasmid DNA was transformed into chemicall y competent Saccharomyces cerevisiae EBY100, grown, and stored in yeast sto rage buffer at - 80°C exactly according to published protocols 24 (see Appendix A ). 3.3.7. Screening of Pro - cNGF and Pro - cNGF Libraries 1 x10 7 cells were grown from freezer stocks in 1 ml of SDCAA for 6 hrs at 30°C and re - inoculated at OD 600 = 1.0 in 1 ml of SGCAA at 18°C for 16 hrs. 2x10 7 yeast libraries were labeled with either biotinylated tanezumab or mAb #1 at 5 nM for 30 mins at room temperature in PBS - - cymc - - R - phycoerythrin conjugate in 1.89 ml of PBS - BSA for 10 mins at 4°C. Sorting was done on a BD Influx Cell Sorter at the Michigan State University Flow facility. For each sort 200,000 cells were collected (approx. 100 - fold the theoretical diversity at the amino acid level) using a diagonal gate set to collect the top 2 - 3% of the displaying popul ation (full statistics in Table E4 ). Collected cells from each population were recovered at 30°C for 30 hrs in of penicillin - streptomycin, washed, and then stored in 1 ml of yeast 41 storage buffer at a concentration of 4x10 7 cells per ml at - 80°C. 3.3.8. Determinat ion of Conformational Epitopes N GF conformational epitopes for all four mAbs were determined using ye ast surface display, comprehensive mutagenesis, FACS, and deep sequencing exactly as previously described 24 . Appendix E contains the average dissociation constant values ( Table E5 ), the percentage collected from library screening ( Table E6 ), primers used for deep sequencin g ( Table E7 ), and statistics results ( Table E8 ). 3.3.9. Deep Sequencing Preparatio n L ibraries were prepared for deep sequencing according to Kowalsky et al. (2015a) 25 using Method B with both PCR reactions set to 14 cycles. The libraries were pooled and seque nced on an Illumina MiSeq using 2 x 250bp pair - end reads at the Michigan State University Genomic Sequencing Core facility or the University of Illinois at Chicago DNA S ervice facility. Primer sequences used for each library are listed in Table E7 . 3.3.10. Data Analysis A modified version of Enrich 0.2 software as described in Kowalsky et al. (2015a) 25 was used to compute enrichment ratios from the raw sequencing files. Custom python scripts available at Github (user: JKlesmith) were used to normalize the enrichment ratios (ER i ) defined as: 3.1 where f i,sel is the frequency of variant i in the selected population, and f i,ref is the frequency of variant i in the reference population. Libraries statistics results are listed in Table E4 . For the pr o region sorting experiments, we define an enrichment score (ES i ) for each mutant i as the enrichment ratio of the selected mutant minus the wild - type enrichment ratio: 42 3.2 For conformational epitope mapping experiments, we define a relative bindi ng term for each mutant as the log transform of the mean fluorescence for variant i , , normalized to the relative mean fluorescence of the wild - type construct, : 3. 3 This equation can be written in terms of experimental observable s according to: 3. 4 where is the log normal fluorescence standard deviation of the clonal population, and is the percentage of cells colle cted by the sorting gate on the flow cytometer ( Table E6 ) 25 . 3.3.11. Data Availability F ull datasets including normalized fitness metrics, pre - and post - selection read counts, and raw log base two enrichment scores for each variant can be found in Med ina - Cucurella et al. 51 . Raw sequencing reads for this work have been deposited in the SRA (SAMN07693504 SAMN07693526 ). 3.4 . Results 3.4.1 . Initial cNGF constructs are improperly folded on the yeast surface W e sought a yeast surface display platform of cNGF to evaluate binding of candidate anti - cNGF mAbs . Mature cNGF is 97.5% pairwise identical to human NGF (hNGF), with an additional 17 substitutions (out of 103 total residues) on the pro - sequence ( Figure 10a ). We assessed proper folding of c NGF on the yeast surface using two conformationally sensitive mAbs . Soluble 43 tanezumab an anti - hNGF mAb and an anti - cNGF mAb (mAb#1) recognized soluble recombinant cNGF as shown by surface plasmon resonance ( Figure 10b , Table E5 ). However, neither tanez umab nor mAb#1 recognized denatured cNGF as demonstrated by lack o f signal by Western blotting (data not shown). Recombinant cNGF increased proliferation of TrkA - expressing TF - 1 cells 73 , and this proliferation could be blocked by the anti - cNGF mAbs ( F igure D 6 ). Thus, both tanezumab and mAb#1 recognize a conformational epitope on cNGF and all mAbs function as cNGF antagonists. This biochemi cal data is corroborated with a previously published co - structure of tanezumab with hNGF 60 that reveals binding mainly at the homodimer interface between subunits in a conformation that requires properly folded hNGF. In the remainder of this ribe cNGF that is recognized by these conformationally s ensitive mAbs. We tested folding of four different cNGF yeast display constructs ( Figure 10c ). First, mature cNGF was fused with a N - terminal Aga2p domain , a N - terminal (G 4 S) 3 linker and a C - termina l c - myc epitope tag (cNGF; pETconNK_cNGF). Second, the mature cNGF was displayed with an N - terminal Aga2p pre sequence, an N - terminal FLAG epitope tag, a C - terminal (G 4 S) 3 linker, a C - terminal Aga2p, and a terminal c - myc epitope tag (Aga2 - cNGF; pETconNK_Ag a2_cNGF). Our third construct was identical to Aga2 - cNGF but included the full - length pro region between the C - terminus of the Aga2 pre sequence and the N - terminus of the mature cNGF (pro - cNGF; pETconNK_Pro - cNGF). Finally, a classical study defined sectio ns of the basis of sequence conservation 70 ( Figure 10a ). Truncation experiments showed that only Box 3 and Box 5 of the pro sequence were n ecessary and sufficient to produce active mouse NGF. Thus, our fourth construct p - cNGF was identical to pro - cNGF except Box 1 and Box 2 were deleted from the pro sequence. Although all 44 variants displayed on the yeast surface, yeast cells labeled with saturating amounts of mAbs yielded signals of only 1.1 (cNGF) to 2.2 (pro - cNGF) in the signal:noise ratio ( Fig 10d - e ). This signal:noise ratio was calculated by measuring the ratio of sample MFI over the MFI in t he absence + and fsc/ssc + cells (to ensure measurement of individual yeast cells). Sin ce similar experiments from our research group show 50 to over 100 - fold above background for diverse protein - protein interactions 12,74 , we conclude that cNGF surface displays in a mostly misfolded form. Figure 10 . cNGF yeast display const ructs are mostly misfolded as probed by conformationally sensitive mAbs. ( a. ) Sequence alignment of the canine and human pro regions of NGF. Domain boundaries and dibasic protease cleavage sites are shown. ( b. ) Surface plasmon resonance sensorgrams of cNGF:conformational mAb binding. cNGF was immobilized on a CM5 surface by amine coupling and either tanezumab or mAb#1 was injected and flowed over surface at various concentrations starting at 100nM and titrating down with 3 - fold dilutions , flowed over the chips. Done by Zoetis . ( c. ) Four different cNGF constructs tested in t he present work. ( d. - e. ) Flow cytograms ( d. ) and bar charts ( e. ) showing increase in fluorescence in cNGF binding channel probed by tanezumab and mAb #1 (error bars, sta ratios were obtained by calculating th e ratio of the MFI of the sample to the MFI in the absence of biotinylated mAb. 45 3 .4.2. Comprehensive analysis of pro mutations that improve cNGF folding A previous direc ted evolution study showed that mutations within the pro sequence could enhance proper folding of the mature neurotrophin BDNF on the yeast surface 63 . Based on this precedent, we sought mutants that improve the expression of folded cNGF . A flowchart of the experimental pipeline utilizing comprehens ive mutagenesis, FACS , and deep sequencing is shown in Figure 11a . Comprehensive s ingle - site saturation mutagenesis (SSM) libraries for pro - cNGF - cNGF were prepared by nicking mutagenesis 23 and transformed into S. cerevisiae EBY100. The SSM librar ies covered an average of 84.3% of all possible si ngle missense and nonsense mutations ( 1 737 mutations for pro - cNGF, 943 mutations for - cNGF ; Table E8 ) . Libraries were labeled with either biotinylated tanezumab or mAb #1 at 5 nM and sorted by FACS . Tanezumab was chosen as structural information of the mAb - NGF complex is known, while mAb #1 was chosen as a representative anti - cNGF a s it had the highest initial signal:noise ratio. We collected the top 3% by cell fluorescence in the mAb channel, along with a reference population of yeast cells that passed through the cell sorter. After each sort, plasmid DNA was isolated, prepared, and deep sequenced. We evaluated each mutant by a relative enrichment score ( ES ) defined as the enrichment ratio of the mutant in the sorted population minus t he enrichment ratio of the wild - type sequence. In this scoring system, a mutant with positive ES imp roves cNGF folding relative to the wild - type sequence. After the first sort, bulk populations from all three libraries showed increased fluorescence associ ated with cNGF folding ( Figure 11b ). We made a number of observations based on these near - comprehens ive datasets after the first sort. First, most mutations, including, premature stop codons, centered near an ES of 0 (for pro - cNGF: missense - 0.12 ± 0.52 , nonsense - 0.53 ± 0.36; pro 1,2 - cNGF: missense - 0.14 ± 0.45 , nonsense - 0.03 ± 0.44 ; mean values at 1 s.d. ). This low signal 46 to noise for loss of function mutations is not surprising after the first sort because the starting constructs have a fluorescence intensit y in the binding channel barely above background. Second, for each construct the correlation between each mAb was high: R 2 =0.69 for pro 1,2 - cNGF and R 2 =0.77 for pro - cNGF ( Fig 11c - d ). Because the observed reproducibility is similar to that seen in replicates using this deep sequen c ing methodology 75 , we cannot differentiate between experimental noise inherent in the deep sequencing pipeline and true biological differences of folding probed by individual mAbs. Third, correlation between pro - cNGF and pro 1,2 - cNGF was Figure 11 . Identifying sequence - function relationships for pro region engineering using deep sequencing . ( a. ) Comprehensive site - saturation mutagenesis libraries were constructed for the pro region and sorted twice b y FACS to collect the top 3% of cells using a diagonal gate set for fluorescent channels corresponding to mAb binding and surface display. Collected populations were deep sequenced, compared with the reference population and converted to an enrichment scor e. ( b. ) Increase in fluorescence channel associated with mAb binding for libraries before and ( c. g. ) Correlation in enrichment scores for each mutation for different mAbs and i nitial constructs . 47 s tati st ically significant (p - value 3.1x10 - 05 tailed paired t - test) but comparatively and unexpectedly lower ( Fig 11e - f ). Thus, mutations that confer differences in cNGF folding are, to a certain extent, context sensitive and suggest that higher order models wou ld be needed to capture the chaperone function of the pro sequence . Fourth, there are many mutations that improve cNGF folding. Using an ES cut - off of three standa rd deviations above 0 we identified 49 (2.8%) pro - cNGF mutations that improve folding. Finall y, because pro - cNGF had more beneficial mutations and a higher bulk population fluorescence than pro 1,2 - cNGF ( Figure 11b ), we moved ahead with pro - cNGF alone for the next round of sorting . Whereas after the first sort libraries showed only a modest incre as e in cNGF folding, libraries after sort 2 obtained 3.4 and 7.7 increase in the signal:noise ratio for tanezumab and mAb#1, respectively ( Figure 11b ). The deep se quencing results were very similar between mAbs, with a R 2 =0.82 for the entire dataset betwee n conformational antibodies and all 27 tanezumab mutations with an ES above 2 matched in the mAb#1 dataset ( Figure 11g ). A full - length heatmap of the pro - region sh owing site - specific preferences probed by tanezumab binding is shown in Figure 1 2 (full datas et s for all constructs and sorts are listed in Figures D 7 - D 1 1 ). Consistent with Suter et al. (1991), Box 3 is more conserved than the rest of the sequence (mean ES missense mutations - 0.23 vs. 0.01; p - value 2.9x10 - 11 tailed paired t - test). In fact, for mos t positions outside of Box 3 a large majority of missense mutations are tolerated, highlighting the inessentiality of specific sequences for the majority o f the pro region. 48 A number of intriguing sequence - function relationships were revealed by inspectio n of mutations enhancing cNGF folding. Of the 27 mutations with an ES score above 2, all were located either in Box 3 or close to the domain boundary in Bo x 2. Positive substitutions at Box 2 include substitutions at Ala - 61, Thr - 57, and Arg - 53. Most subs t i tutions at Ala - 61 have positive ES, especially aliphatics and aromatics. We note that the nonsense codon has a slightly positive ES at this position: si nce a premature stop codon is unlikely to result in displayed cNGF, calculations for some slightly po s i tive ES reported here may be within noise of the measurement. All substitutions at Thr - 57 had positive ES, including major gains for mutations to Tyr, M et, Leu, Val, and the polar Gln. While Box 3 is more conserved than other pro regions, certain positi o n s showed beneficial mutations. Most notably, most substitutions in the strongly conserved protease ng very high ES for charge reversal mutations at Lys - 43Asp and Arg - 40Asp. Figure 12 . Per - position heatmap of enrichment scores for pro - cNGF mutants after 2 sorts with tanezum ab. 49 3 .4.3 . Combining single mutants improves the amount of displayed, folded cNGF T o improve display of folded cNGF, we first made three isogenic construc ts of the best mutations (ES values higher than 4.6) identified from the deep sequencing experiment (Box 2: A - 61 W, T - 57Q; Box 3: K - 43D ) and tested their ability to recognize biotinylated tanezumab. In the yeast surface display context, all three chosen mut an ts showed improved MFI relative to pro - cNGF ( Fig ure D1 2 ). Next, we reasoned that combining mutations would result in higher amounts of folded cNGF. Pro.v1 - cNGF Pro.v3 - cNGF were double mutants, while Pro.v4 - cNGF contained all three mutations ( Table 2 ) . Ea ch was tested for the expression of the folded protein by labeling with tanezumab, mAb #1, or two additional anti - cNGF mAbs (mAb #2, mAb #3). While all constructs demonstrated at least 12 - fold increase in signal:noise ratio, Pro.v3 - cNGF and Pro.v4 - cNGF s ho wed between a 19 and 23 increase in signal:noise ratio depending on the probe mAb ( Fig 13a - b ). Improved pro constructs gained at most a 1.9 - fold increase in surface expression over pro - cNGF ( Figure 13c ), showing that most of the effect of mutations cente re d on improving the folding of displayed cNGF. Table 2 . Signal:noise ratio for Pro - cNGF constructs defined as the MFI of yeast cells labeled with specified mAb over MFI of unlabeled cells. Errors bars represent the standard error of the L ibrary Name Mutations tanezumab mAb #1 mAb #2 mAb #3 Pro.v1 - cNGF T - 57Q - K - 43D 6.5 ± 1.5 7.6 ± 0.7 2.8 ± 1.6 5.3 ± 1.2 Pro.v2 - cNGF A - 61W - T - 57Q 10.1 ± 2.5 11.7 ± 1.3 4.3 ± 2.6 5.8 ± 1.0 Pro.v3 - cNGF A - 61W - K - 43 D 14.3 ± 3.7 19.0 ± 2.8 6.3 ± 4.4 7.3 + 1.8 Pro.v4 - cNGF A - 61W - T - 57Q - K - 43D 14.3 ± 3.4 20.5 ± 4.2 23.2 ± 2.6 8.7 ± 1.7 3 .4.4. Small amount of pro - cNGF displays on the yeast surface T o determine whether pro - cNGF displays on the yeast surface or whether the pro sequence is processed, wild - type cNGF, pro - cNGF and pro.v1 - v4 - cNGF were labeled with fluorescence conjugated antibodies against the C - terminal c - myc epitope tag and the N - terminal FLAG epitope 50 tag. The percentage of cleavage was estimated by measuring the FLAG/c - myc ratios of p ro - cNGF and pro.v1 - v4 - cNGF usi ng the fluorescence mean values. As a control, these samples were compared with the ratio of cNGF 63 labeled exactly the same. In all cases, cells positive for c - myc were slightly positive for F LAG binding ( Figure 13d ), indicating that at least a portion of the pro sequ ence is proteolytically cleave d before display and suggesting that mature cNGF is displayed. All constructs, Pro.v1 - v4 - cNGF, showed from 88 - 96 % of cleavage, meaning that a low per centage of the full - length pro - region is displayed on the yeast surface. F igure 13 . Pro - cNGF variants with multiple point mutations show improved cNGF folding. ( a. ) Flow cytograms for pro.v4 - cNGF versus pro - cNGF. ( b. ) Increases in fluorescent channel probed by mAb#2 and mAb #3 binding for pro.v4 - cNGF com pared with pro - cNGF. ( c. ) Bar charts showing the increase of all pro - cNGF variants in surface expression probed by labeling with a fluorescently conjugated anti - cmyc m Ab. ( d. ) Percentage of displayed, folded cNGF on the yeast surface (error bars, standard 51 3 .4. 5. Conformational epitopes reveal similar profiles, but distinct epitopes, for tanezumab and all three mAbs U sing the pro.v4 - cNGF yeast d isplay construct we determined dissociation constants of tanezumab and mAbs #1 - 3 to cNGF . Yeast cells were incubated wi th varying amounts of mAb, washed, and labeled with secondary reagents prior to analysis by flow cytometry. Binding dissociation constant s ranged from 143 ± 44 pM for mAb #3 to 800 ± 1 64 (1 s.d., n 3) for tanezumab ( Figure 14 , T able E5 ). Interestingly, best fits of the Hill coefficient for these mAbs are all significantly below 1 (p - value <0.0106 one tail t - test Table E5 ), indicating potential negative cooperativity between the dimeric displayed cN GF and dimeric mAb. Alternative l y, Hill coefficients of less than 1 can arise for binding of non - equivalent binding sites. Since cNGF pr esumably exists in a range of folded conformations on the cell surface, both alternatives are plausible 76 . Figure 14 . Binding titrations for surface - displayed pro .v4 - cNGF for various mAbs. Each mAb was titrated at least three separat e N ext, we determined the fine conformational epitopes of tanezumab and the three anti - cNGF mAbs using a previously developed method involving yeast sur face display, nicking mutagenesis, and deep sequencing 12,24 . The principle behind this method is that antigenic mutations that disrupt binding will map predominantly to the epitope positions recognized by th e antibody. Compared with other epitope mapping strategies that collect the popul ation that no 52 longer binds antibody 5,15 , for our approach we collect the binding population. We then deep sequence the reference and binding populatio n s. For each mutant , a relative binding term can be derived from the change in frequency of the bound population compared with a reference popula tion. Shannon entropy (SE) , a measure of sequence conservation, is then calculate d. P ositions with SE values le s s than or equal to the midpoint of the SE range are defined as belonging to the epitope 12 . In this experiment, two SSM libraries of cNGF were constructed by nicking mutagenesis, labeled with biotinylated mAb at half of the experimentally determined dissociation c o nstant ( Table E5 ), sorted by FACS, and deep sequenced. SSM libraries covered an average of 95.6% of all possible single nonsynonymous mutations in cNGF (1,115/1,200 for library 1 covering cNGF positions 1 - 60 and 1,178/1,200 for library 2 covering position s 61 - 120). Full statistics of library coverage is given in Tables E8 , and per - position cNGF heatmaps for all mAb epitopes are given in Figure 15a and Figures D1 3 - D1 6 . In our epitope mapping method, we typically start by removing positions that are struct u rally conserved from further analysis. We identify structurally conserved positions by col lecting the population that surface displays the C - terminal c - myc epitope tag; structurally conserved positions are those that have a Shannon entropy less than or eq u al to the midpoint for the surface displayed population. However, our results show that mu tations at most positions did not result in a change in the displayed protein ( Figure D1 7 ). These results are roughly consistent with the initial results showing mis f olded cNGF still displaying on the surface. Thus, knowing the mature NGF conformation, we reasoned that structurally conserved positions are mainly at the core or the homodimer interface and can be identified as positions with less than 25% accessible sur f ace area (ASA) evaluated from solved NGF structures. This analysis excluded 48 out of 120 53 residues ( Figure D1 3 ). The average Shannon entropy for the tanezumab binding population for these buried positions was significantly lower than for solvent accessibl e positions (1.75 vs. 2.31, p - value <10 - 11 tailed paired t - test), indicating that conservation of the folded state of cNGF is essential to recognition by tanezumab. As a n initial control we compare d our experimentally determined epitope with the previous l y published X - ray crystal structure of tan ezumab Fab bound to h NGF 60 . 11/120 positions mapped to the tanezumab epitope (S1 9 , W21, K32, G33, K34, F49, Y52, K88, A97, W99, and R100) while 38/120 positions were completely non - conserved ( Fig D 1 3 and 15 a ). As shown in Figures 15 b - h , epitope positions form a contiguous patch that largely maps to the binding footprint of tanezumab p r eviously described by La Porte et al. (2014) 60 . Of the eighteen NGF positions within 4Å of tanezumab and with C - C vector s pointing towards the antibody, seven were identified as hits in our pipeline, while six were structurally conserved. Of the remainder, only one epitope position (R114) was identified as nonconserved; this Arg makes hydrogen bonding interactions wi th the m ain chain of tanezumab . Our deep sequencing approach identifies strong conservation of the cNGF loop centered around K32 - K34 ( Figure 15e ) and positions central to the interface (S19, W21) ( Figure 15f ) that presumably make strong van der Waals conta cts . Th e remaining epitope hits were second shell residues (e.g. W99, R100) buttressing these original contacts. However, many epitope positions at the homodimer interface where tanezumab binds to hNGF (e.g. F54, T56, T85, F86, T106, A107, and C108) are in visible to our method since they are structurally conserved ( Figure 15g ). Another shortcoming is that positions recognized by the antibody using main chain contacts (e.g. G10, E11, F12, and H84) are also invisible, as mutations to other amino acids will no t disru p t antibody binding. Tanezumab binds to a ridge on NGF with side chains of residues R9 - F12, V111, and R114 pointing directly away from the antibody ( Figure 54 15h ). All of these positions are completely non - conserved ( Figure 15a ). Finally, we evaluated the binding interaction with three canine anti - cNGF mAbs. Shannon entropy results were very similar between tanezumab and each mAb, with a R 2 =0.88, 0.89, and 0.77 for mAb #1, mAb #2, and mAb #3, respectively ( Figure 16 ). All mAbs shared 5/11 cons erved position s as the tanezumab epitope, but possessed different sequence - binding profiles and therefore unique epitopes ( Fig D1 4 - D1 6 ) . Figure 15 . Determination of the cNGF:tanezumab conformational epitope using deep sequencing. ( a. ) A subset of the fitness metric per position heatmap of the top 7% bound population vs. the unselected population for tanezumab. Shannon entropy is plotted beneath with the midpoint cut - off shown as a dash ed line. ( b. - c. ) Tanezumab epitope view at the interface cor e of each monomer ( b. ) and as a homodimer mature structure ( c. ) . Positions colored in orange, blue, gray, and white are epitope, completely non - conserved, buried in the core, and non - conserved pos itions, respectively. ( d. - h. ) Close - up views of specific tan e zumab:cNGF sidechain interactions. Molecular representation of the interaction was modeled on the solved tanezumab:hNGF structure (PDB ID: 4EDW). I nteresting, there were two mutations, V6M and M9 2F that show increased relative 55 fluorescence in the binding p opulation for all four mAbs ( Fig D1 3 - D1 6 ). Our epitope mapping method cannot discriminate whether such mutations improve the binding affinity to a given mAb or whether the mutation increases the amount of folded cNGF. However, we speculate that the latter is the case as these two residues are surface exposed positions located distal to the tanezumab epitope, and all four mAbs bind in overlapping but atomically different epitopes. Figure 16 . Correlation in Shannon entropy of the binding populations between tanezumab and other mAbs. 3 .5. Discussion and Conclusion In this study we engineered a pro - region to enhance folding of cNGF on the yeast surface. This yeast platform enabled our team to map conformational epitopes for tanezuma b and a number of anti - cNGF mAbs. During the course of this research we showcased the power of deep sequencing to augment directed evolution workflows to improve protein properties 72 , gained new insight into sequence - function relationships for neurotrophin pro regions, and identified fine conformational epitopes for anti - cNGF mAbs. 3.5.1 . Pro - region engineering pipeline c NGF and pro - cNGF displayed on the yeast surface but in mo stly misfolded forms. This adds to a growing body of literature showing that while the quality control machinery in the S. cerevisiae endoplasmic reticulum 77 can impact the overall amount of protein displayed on the surface 8,75 , for a ny given grossly misfolded protein some will still pass through quality control checkpoints in the secretory pathway and display on the yeast surface 8,12,74,78 . 56 We used an engineering pipeline involving comprehensive site - saturation mutagenesis, FACS, and deep sequencing to improv e surface displayed, folded cNGF. This engineering pipeline enabled us to evaluate the functional effect for over 2,000 individual point mutants to the pro - cNGF and pro 1,2 - cNGF constructs. We then created multi - site constructs by combining the best mutati o ns cherry picked from the deep sequencing datasets. There are several advantages of this workflow compared with traditional directed ev olution approaches. First, precise single - site mutagenesis libraries constructed from nicking mutagenesis allow us to as s ign a functional effect to a single mutation; libraries constructed by error - prone PCR often have multiple mutations per gene, which th en need to be deconvoluted. Second, deep sequencing allows us to perform experiments in parallel: we performed duplicate sorts using different conformational mAbs, allowing one to identify mutations beneficial to both mAbs. Third, the small library size al lows us to simplify steps for yeast transformation and sample handling, shorten FACS times, and allow us to complete our screening in 2 sorts compared 4 - int mutants is the chance of missing epistatic mutations but in the present work identified beneficial pro mutants could be combined additively. Our best construct identified by deep sequencing greatly improved the folding of displayed cNGF and was suffici ent for the conformational epitope mapping performed. However, we suspect that there is room for further optimization of the yeast display of folded cNGF as ( i .) the max signal:noise ratio of pro.v4 - cNGF is still lower than that seen for other antibody - ant igen interactions probed by our lab; ( ii .) two mutations on mature cNGF could presumably improve folding of the pro.v4 - c N GF construct; and ( iii .) the Hill coefficient for all antibodies was less than one, which is consistent with cNGF existing in a range o f folded and partially misfolded configurations on the yeast surface. Further optimization could be done by selecting si m ilar 57 beneficial point mutants identified from the deep sequencing datasets ( Figure 12 ). We speculate that the combination of an additio nal subset of these point mutations would facilitate production of mature cNGF from yeast 63 , although this is beyond the scope of the present work. 3.5.2 . Pro - region sequence - function relationships A lthough the classic work from Suter et al . (1991) 70 showed that Box 3 and 5 of the pro - cNGF did not improv e folding as noticeably as fo r pro - cNGF. Given that many of the strongest beneficial mutations were in Box2, we speculate that Box2 encodes some chaperone activity necessary for complete cNGF folding. However, further studies should be performed to underst and domain boundaries and fun ctional differences between constructs. C onserved and beneficial mutations cluster around Box 3 as was suggested for mouse NGF 70 . Not able lack of conservation in B ox 3 centered around a cleavage site for proprotein convertases at Lys - 43 Arg - 40. The se four positions demonstrate d improved profiles for almost all the mutations. Indeed, Lys - 43A sp was one of the mutations introd uced in our engineered pro - cNGF that in turn, disrupt recognition of this KR dipeptide by endogenous Kex2 protease in S. cerevisiae . Interestingly, both pro - cNGF and prov.4 - cNGF display a mixture of full - length pro - cNGF and at least partially cleaved cNGF, suggesting that enhanced folding conferred by the Lys - 43Asp mutation does not result from differential protease cleavage. A previous study utilizing hydrog en - deuterium exchange with the pro region and mature NGF determined that Trp - 83 Ala - 63 in the pro sequence ( Trp - 84 Gly - 64 canine numbering) was involved in the intramolecular chaperone - like interactions with mature human NGF 64 . Inconsistent with this experiment, we obse rved all positions in this region to be non - conserved, with most mutations centered around an ES of 0. Our deep sequencing experiments probe folding 58 of cNGF but do not directly assess the chaperone function of specific pro sequence variants, so further wor k is nece ssary to reconcile this apparent discrepa ncy. 3.5.3. Mapping epitopes targeted by anti - cNGF mAbs W e were able to determine conformational epitopes for tanezumab and three potential canine mAbs. Our experimentally determined tanezumab epitope larg ely overl apped with the previously published tanezumab - mNGF structure 60 , although many epitope positions (i.) at the dimeri zation in terface, and (ii.) not participating in side chain contacts with antibody were invisible to our deep sequencing method. These shortcomings will be shared with all mutational - based epitope mapping methods. The three anti - cNGF mAbs had different seq uence - bin ding profiles at the epitope. We conclude from these results that our yeast display platform for cNGF is able to map fine conformational epitopes for candidate anti - NGF mAbs. Although our epitope mapping workflow cannot directly measure if cNGF is displayed as a monomer or dimer on the yeast surface, the results strongly suggests that our best constructs display as mature dimer cNGF. Mutations at the homodimer interface core aw ay from the epitope will disrupt the folded state of cNGF necessary fo r mAb recognition, and indeed these are depleted conformational epitope tumor necrosis facto r (TNF) - infliximab 12 , where mutations at the TNF homo - trimeric interface disrupt ed antibody binding. However, we note that the solution - based measurements of cNGF - mAb binding are high fM - low pM, whereas measurements from the yeast surface are mid - to high - pM. While Gai and Wittrup have shown the rough equivalence in affinity between yeast - and solution measurements 79 , their dataset included only 1 fM binder and thus we do not necessarily expect eq uivalence for these high affinity binders. 59 In conclusion, we used a deep sequencing pipeline to develop a yeast display platform for folded cNGF. This contribution highlights the power of deep sequencing to identify nearly all beneficial point mutants in a protein in a simplified workflow. Since a major limitation of yeast display is proper folding of complicated mammalian proteins, this work complements a recent directed evolution study 63 to show that effective strategies exist to overcome such limitations. O ur pipeline could be a promising plat form to increase the production of highly active titers of the other m embers of the neurotrophin family, to determine the specificity and affinity to their respective receptors, and to enable the epitope mapping for therapeutics against neuronal diseases such as the brain disorders caused by BDNF . 60 CHA P TER 4 4. User - defined single pot mutagenesis using unpurified oligo pools - defined single pot mutagenesis using unpurified PEDS ( under review ) 61 4.1 . A bstract User - defined mu tagenic libraries are fundamental for applied protein engineering workflows . Here we show that unpurified oligo pools can be used to prepare libraries from plasmid DNA with near - complete coverage of desired mutations and few off - target mutations. We fin d t hat oligo pools yield higher quality libraries when compared to in dividually synthesized degenerate oligos. We also show that multiple libraries can be multiplexed into a single oligo pool, making preparation of multiple libraries less expensive and mor e c onvenient. 4.2 . Introduction Directed mutagenesis is foundational for synthetic biology and protein engineering. Recent methods support the creation of large libraries of user - defined mutations in a single reaction 23,80,81 . Such protocols rely on annealing a short oligonucleotide to a parental template, wherein the oligo encodes a mutation by template mismatch. The complementary strand encoding the desi red mutation is synthesized, after which the parental template strand is specifically destroyed. For large libraries the mismatch is encoded using degenerate nucleotides, such that a single oligo can make up to 63 different mutations per codon substitution . However, degenerate oligonucleotides r equire hand - mixing to avoid overrepresentation of nucleobases 80 and are often unable to encode a desired subset of amino acids. Microarray - synthesized oligo pool technology has recently found use in synthetic biology 82 , wi th multiple vendors offering relatively long oligos at moderate error rates. Clever techniques have emerged to use these pools for gene synthesis 83 , but the low femtomolar co ncentrations of individual oligos usually necessitates amplification and further processing from pools, limiting us ability. We recently described the Nicking Mutagenesis (NM) 23 method to construct user - def ined libraries in one pot using routinely prepared plasmid DNA. Because N M uses very low oligo 62 concentrations, with the template in large excess to the mutational prim er, we hypothesized that unpurified single stranded oligonucleotides from microarrays cou ld be used directly in the reaction . To test whether unpurified oligo pools are compatible with NM, we synthesized a single custom oligo nucleotide pool (Agilent technologies) comprising oligos encoding all missense and nonsense mutations from positions 1 - 6 9 of the bacterial aliphatic amidase Ami E 84 (1449 oligos of leng th 33 - 60 nts) , all possible single nucleotide polymorphisms covering residues 1 5 - 114 for the anti - Influenza human antibody variable heavy gene UCA9 85 ( 1000 57 - nt oligos) , and targeted mutagen ic oligos for the Arabidopsis thaliana abscisic acid receptor PYR1 86 ( 185 51 - nt oligos printed with 6 replicates). Full oligo sequences are given in Medina - Cucurella et al ( under review , PEDS ) . The lyophilized oligo pool was solubilized in 40 L TE at a total concentration of 200 nM, phosphorylated, and diluted directly so as to contain 1.9 n M AmiE - specific oligos. This dilution was used directly in a standard nicking saturation mutagenesis protocol with an A miE - encoding plasmid as a template (see Supporting Note C4 ) . Two replicates were performed. We also oligos that were individually synthesized (IDT) covering the same stretch of the gene. Libraries were sequenced on an Illumina MiSeq in 250 bp paired end mode and processed using PACT 87 . 100% (1380/1380) of the desired mutations were incorporated for each replicate (see Table E10 for full library statistics of all sequences; processed datasets are sho wn as heatmaps in Figure D 1 8 ). The frequency of specific mutants had a correlation of 0. 88 between replicates, demonstrating the repeatability of the mutagenesis protocol ( Fig ure 1 7 a ). Importantly, the oligo pool library had a much more even representation of all 20 amino acids compared with the degenerate oligos, with a mean coefficient of v ariation (CV) of 0.12 compared with 0.59, respectively ( Fig 1 8 a and Fig D 19 ). However, the cumulative 63 Figure 17 . Correlation between the freque ncy of (a.) AmiE, (b.) PYR , and (c.) UCA9 mutants between replicates. d istributions of libraries - normalized to 200 - fold coverage were broadly similar ( Fig ure 1 8 b ). To demonstrate that multiple libraries can be prepared from the same oligo pool, we sou ght to construct a library of all single nucleotide polymorphisms on th e majority of the UCA9 by performing NM with an UCA9 - encoding plasmid as a template with the unpurified oligo pool. Sequencing confirmed an average of 98% ( 5 94 /612 and 606/612 nonsyn ony mous mutations from replicate 1 and replicate 2, respectively) of coverage of the desired mutations ( Fig 1 8 c, Fig D2 0 ) with a correlation of 0.99 and similar cumulative distributions between replicates ( Fig 1 7 b ). Notably, mutations were specifica lly pro gra mmed, as there were on average only 5.8% (86/1,488) off - target mutations observed in the read window ( Fig ure 1 8 c ). O ligo pools offer user - defined mutations relieved from the constraints of degenerate codon compatibility . We next sought to const ruct a lib rary of 185 designed mutations at 17 positions in the PYR1 86 receptor by performing NM using the unpurified oligo pool with a PYR1 - encoding plasmid as a template. All 185 mutations were encoded in the library ( Fig ure 1 8 d ) and with a correlation of 0.93 between replicates ( Fig 1 7 c ). Specifically prog ram med mutations were found on average at 91 - fold higher frequencies than the other 1556 potential single non - synonymous mutations in the 2 61 - nt Illumina sequencing window (median 1038 counts vs 0; mean 1198 counts 64 vs. 13 for encoded vs. non - encoded posit ion s) ( Table E10 and Fig D2 1 ) . Figure 18 . A single unpurified oligo pool combined with nicking mutagenesis shows a near - complete coverage of all programmed mutations for user - defined single and double mutagenic libraries. ( a. ) Th e re lative amino acid substitution frequencies for AmiE libraries rep = replicate). ( b. ) The cumulative distribution function of AmiE libraries as a function of the numb er o f counts normalized to a 200 - fold depth of coverage. ( c. ) Bar charts showing the percentage of programmed versus non - programmed mutations for UCA9 li braries for the two different replicates. ( d. ) A subset of the per position heatmap showing the number of c ounts per mutations encoded in PYR1 single site libraries. Boxes framed in red represent the programmed mutations. ( e. ) Per position heatmaps compari ng the expected versus observed programmed mutations in the PYR1 double site mutagenic library. Boxes f rame d in orange represent the non - expected mutations due to nucleotide mismatches between oligos and DNA template. ( f. ) X - Gal staining of yeast colonies for PYR1 double mutants that interact with HAB1 in the presence of 1 µM Mandipropamid using an establis hed yeast - two hybrid system. DMSO is the negative control. I t is also possible to sequentially perform NM with the same oligo pool, creating mutagenic l ibraries with two mutations per gene. We tested this synthesis by performing NM again using the PYR1 si ngle mutagenesis library plasmid DNA as a template. We expected 10,845 out of the 1,462,000 possible double point mutants in the Illumina sequencing read window. Deep 65 sequencing recovered 13,904 double mutants with 2 or more read counts, of which 8,316 wer e sp ecifically programmed ( Fig 1 8 e ). Double mutations were depleted at near - adjacent positions ( Fig 1 8 e ), presumably because a second oligonucleotide containing mismatches would either not anneal or overwrite the mutation encoded from the first oligonucleo tide . To demonstrate utility, this library was screened against the non - native ligand mandipropamid usi ng a previously established yeast 2 hybrid screen 8 6 . We uncovered PYR1 mutants specific to and responsive at 1 M mandipropamid; sequencing of 10 constructs showed all 10 had the same specifically pro gram med F108A F159M double mutant, which outcompeted single mutants F10A and F159M present in the libra ry ( Fig 1 8f ). In summary , we have shown that oligo pool synthesis technology can be integrated with nicking mutagenesis to construct user - defined single and double mutagenesis libraries. We anticipate its incorporation into standard directed evolution exp eriments and its utility for more thorough evaluation of local protein fitness landscapes. 4.3. Materials and Methods 4.3.1. Strains T he Escherichia col i st rain used in this study was XL1 - Blue (Agilent, Santa Clara, CA ) endA1 supE44 thi - 1 hsdR17 recA1 gyrA96 relA1 lac [F ' proAB lacI q Z M15 Tn 10 (Tet r )] . The Saccharomyces cerevisiae strain used in this study was MaV99, MAT a SPAL10::URA3. 4.3.2. Plasmid Construct s T he pEDA3_ A miE plasmid was created as described in Wrenbeck et al. 2016 . Plasmid pBD_PYR1_BbvCI was constructed from pBD_PYR1 86 by inserting a BbvCI restriction site using standard cloning. pET con _ UCA9 was created by inse rting a codon - optimized V H gene encoding 66 Met 1 Ser 1 22 (Integrated DNA Technologies, Coralville, IA) into yeast display vector pET con (Addgene plasmid #41522) using standard restriction cloning. Sequences were verified by Sanger sequencing (Genewiz, South Plainfield, NJ) and listed in Note C3 . 4.3.3. Degen erate Oligos and Oligo Pool Design A ll degenerate Design Program (www.agilent.com) and were ordered fr om Integrated DNA Technologies . A sin gle 7,118 - member oligonucleotide library pool was custom synthesized by Agilent Technologies (sequences listed in Medina - Cucurella et al., under review , PEDS ). 4.3.4. Preparation of Mutagenesis Libraries S ingle and d ouble - site saturation mutagenesis libra ries were constructed using nicking mutagenesis as described in Wre nbec k et al. 2016 88 with the following changes. To conserve the 20:1 template to oligonucleotide ratio, the volume and concentration of oligos are determined using Sup plem e n tary Note C4 for the AmiE and PYR1 libraries. The UCA9 libraries were prepared using an additional 100 0:1 dilution of the oligo pool. The AmiE library covered residues Met 1 P ro69, the PYR1 targete d library covered Val81 Arg167, and the UCA9 library covered Pro15 Gln114. 4.3.5. Deep Sequencing Preparation and Data Analysis T he mutagenesis plasmids we re prepared for deep sequencing exactly as described in Kowalsky et al. 2015 , , libraries were pooled and sequenced on an Illumina MiSeq using 2 x 250bp pair - end reads by the BioFrontiers Sequencing Core at the University of Colorado, Boulder. P rimers used for deep sequencing are listed in Table E9 and a summary of statistical results are in Table E10 . The software package PACT 87 , freely available at GitHub ( https ://github.com/JKlesmith/PACT/ ) , was used to calculate the sequencing counts 67 obtained from raw FASTQ files. Raw sequencing reads for this work have been deposited in the SRA ( SAMN10992661 SAMN10992668 ). 4.3.6. Yeast two - hybrid screening T he PYR1 double m utant library was transformed into yeast two - hybrid reporter strain MaV99 pA CT - HABI and tested for responsiveness to 1 M mandipropamid (Sigma - Aldrich, St. Louis, MO) as previously described by Park et al. 2015 86 . 68 CHAPTER 5 5. Summary and Future Approaches 69 5.1 . Summary As demonstrated in this dissertation, next - generation sequencing confers the ability to assign the functional effect of thousa nd mutations in one protein partner before and after a high - throughput selection for function. During the course of all these stud ies, we presented a standardized pipeline of comprehensive mutagenesis, yeast surface display, FACS, and deep sequencing to ad dress fundamental aims and limitations in the field of protein engineering. The rich informational datasets allow to map conformat ional epitopes of potential mAbs, to determine binding sites of target antigen receptors, and to engineer antigens to express them in conformationally active forms. This technique can be extended to evaluate multiple protein complexes , enzyme engineering, antibody affinity, paratope mapping, and plant ligand - receptor modules, among others. In Chapter 2 we aimed to understand the interactions between interleukin 31, a cytokine involved in chronic skin inflammations, and its receptors. Such information can l ead to the development of antagonist anti - IL - 31 mAbs to inhibit the downstream signaling pathway. O ur binding sites largely ov erlapped with the previous sites described by Le Saux et al . 20 and also revealed a new overlapping site between both receptors no t described before. Furthermore, the mapped epitope of a candidate mAb suggested its efficacy by antagoniz ing IL - 31 signaling pathway by binding mostly to the overlapping site between receptors . Although our method is suitable for studying sequence - function relationships b etween PPIs, there were some limitations. Chapter 3 describe d a deep sequencing - guided protein engineering workflow developed to increase the production of displayed proteins in a conformation recognizable by the binder partner. In this case, we aimed to i ncrease the amount of folded canine nerve growth factor , a neurotrophi n involved in multiple chronic pain conditions 58 . 70 M utational libraries created using this engineering pipeline revealed new insight into the role of the neurotrophin pro region. A combination of beneficial mutations within the pro - region of NGF allow ed us to (i.) enhance the display of mature NGF on the y east surface by a 23 - fold above background and (ii.) determine the conformation epitopes of multiples anti - NGF mAbs 51 . This research proves that y east surface display platforms can be engineered even for complicated mammalian proteins. To continue improving deep mutational scanning, in Chapter 4 we presented the integr ation of oligo pool synthesis technology with nicking mutagenesis to prepare precise and focused mutagenic libraries for multiple proteins which was easily extended to double mutant libraries. This technology avoids the need of hand - mixing oligos which imp roved the even representation of nucleobases with a near - complete coverage of targeted mutations. We speculate that oligo pools can be used for fu ture protein science studies. All t hese improved throughput technologies will definiti vely contribute to the d esign of new strategies for the development of safe and efficient therapeutics and vaccines. 5.2 . Future Approaches Although we addressed multiple limitations from deep mutational scanning pipelines, there still some room for cost - effective and time - consu ming improvements. As described by Kowalsky et al. 12 , conformational epitope map s for new antibody - antigen interactions can be obtained in at least 3 - 5 weeks which is mostly limited by the design and preparation mutagenesis libraries, and scr eening th r ough FACS. To address such concerns, the computational modeling software Rosetta is useful to predict protein structures and folding mechanisms, and to model multiple protein - ligand interactions 89,90 . Given tw o partner proteins and a random initial orientation, the Rosetta Docking 91 algorithm predicts bound state structures and binding pockets of the interacting p roteins 71 with the lowest energy. It uses a score function composes of multiples energetic terms involved in common non - covalent interactions. Unfortunately, these computational models might not represent an optimal time - consuming improvement as users obtain thousands of docked models after a single run. Thus, researchers wil l need to evaluate each model to select the best ones based on individual assumptions. Nevertheless, these new designs allow us to design strategic point mutants for either additional mod eling or experiments. For example, given an accurate three - dimensiona l structure of the partner protein display ed on the yeast surface, we could design mutagenic oligos for only surface residues. In addition, and by having a predicted model for the bound s tate of the protein :protein complex, oligos could also be design only for positions that are expected to disrupt the interaction . As we showed in all these projects, mutational data sets allow us to the epitope positions f or multiple mAbs. O ur lab has appli ed similar tools for other antibody applications. First, we mapped the site of an antibody that binds to an antigen, the paratope, by displaying scFvs in the surface of the yeast. Second, we used our workflow to engineer antibody specificity. For a candida te mAb with affinity to two related targets, these workflows can id entify mutations that improve the specificity of a mAb for one target while decreas ing the binding affinity for the second target. Consequently, we are confident that similar techniques can also be applied for the evaluation of other antibody forms includi ng bispecific and polyclonal antibodies. Although mAbs have become the lead molecules for therapeutics and vaccines applications, there are some potential clinical and commercial limitatio ns including the selectivity and high costs. One prominent approach to address these limitations is by integrating therapeutic Abs into bispecific formats. Bispecific antibodies (bsAbs) were designed to simultaneously bind to two different antigens or diff erent epitopes within the same antigen. Some improvements could 72 be implemented to deep mutational scanning to screen multiple antigen constructs in a single reaction . One option is the incorporation of unique barcodes next to all genes to be display ed on t he yeast surface. Then, epitope s for each displayed antigen is dete rmined by following the same screening steps through FACS and deep sequencing. However, new custom scripts will need to be developed to determine the contribution of individual assemblies b ased on barcode sequences. Similar high - throughput approaches can b e used to delineate the bulk human antibody response upon a vaccination or an infection. Some viral diseases have no approved vaccines with efficacy for all age - groups . One example is Deng ue virus (DENV), a worldwide mosquito - borne viral disease 92,93 . The main challenge for development of Deng ue vaccines is the antibody - dependent enhancement 94 (ADE) of the infection where the pre - existing , poorly cross - reactive antibodies developed by the human immu ne response after primary infection tend to increase the secondary heterotypic dengue infection 95,96 . Thus, new throughput techn olog ies are need to deconvolut e complex immune responses against this kind of infection. N ew experiments could give us valuable information (i.) to trace the behavior of the complex polyclo nal response upon vaccination or infection , (ii.) to understand the distribution of epitopes targeted by individual antibodies and (iii.) to predict the immunodominant and subdominant responses for any given complex polyclonal antibody mixture. To that end , we expect that all these future applications will have an enormous impact in the biomedical field by offering innovative epitope binning and mapping methods. 73 APPENDICES 74 APPENDIX A : C haracterizing protein - prot ein interacti o ns using deep sequencing coupled to yeast surface display Characterizing protein - protein interactions using in Methods in Molecular Biology ( 2018 ) 1764:101 - 121 75 A.1 . Introduction In this appendix, we provide a detailed protocol to determine relative binding affinities and conformational epitope maps for PPIs (overview in Figure 1 ). We cover creation of singl e site saturation mutagenesis (SSM) l i b raries using nicking mutagenesis 23 , transformation of libraries into yeast by the metho d o f Gietz and Woods 97 , screening of t he SSM YSD library using FACS, DNA preparation for sequencing on an Illumina platform, and data analysis to determine a relative binding score and conformational epitope map. Relative binding calcula tions and estimated errors are carried out according to m ethods described in Kowalsky et al. 12 . Note: we assume the end - u ser has (i.) one PPI partner successfully induced and displayed in a YSD format with the other partner biotinylated; with (ii.) reproducible measurement of the apparent dissociation constant using p ro t o cols as described in Chao et al . 17 . A.2. Materials A.2.1. Yeast and Ba ct e r ia Strains and Plasmid 1. Y east Strain: Saccharomyces cerevisiae strain EBY100 is available at American Type Culture Collection and prepared to be chemically competent according to Gietz and Woods 97 . (see a shortened protocol in Section A. 3.1.4 ) 2. Bacterial Strain: Escherichia coli strain X L1 - B lue high - efficiency electrocompetent cells are available through Agilent Technologies. Other competent cells with at least 1x10 9 3. Yeast display vectors : The YSD vector used, pETconNK, is freely available on Ad d g ene (plasmid #81169) 75 . The gene of interest is inserted between NdeI and XhoI restriction sites. 76 A.2.2 . Nicking Site - Saturation Mutagenesis (SSM) Library Preparation A ll enzymes and buffers for SSM library preparation with Nicking Mutagenesis a re from New England BioLab s Inc. (NEB) unless noted otherwise. A.2.2.1 . Reagents, Media and Plates: 1. pETconNK plasmid containing gene of interest (freshly prepared from a dam + bacterial strain) 2. Nuclease - free water ( NFH 2 O , Integrated DNA Technologies) 3. Cust o m mutagenic primers (see Section A.3.1.1 ) 4. CAAGTCCTCTTCAGAAATAAGCTTTTGTTC 5. T4 Polynucleotide Kinase Buffe r 6. 10X CutSmart Buffer 7. 5X Phusion HF Buffer 8. 10 mM ATP 9. 50 mM DTT 10. 50 mM NAD + 11. 10 mM dNTPs 12. 50% v/v sterile glycerol solution usi n g deionized H 2 O 13. TB media : 4.76% w/v of TB powder (pre - mixed), and 0.8% v/v of glycerol. Sterilize by autoclaving. 14. LB agar plates : 2.5% w/v of LB powder (pre - mixed), and 1.5% w/v of agar. Sterilize by autoclaving. 77 * Add kanamycin to a final concentrat ion o f the transformation efficiency and the large bioassay dishes for SSM Libra ries (see Section A. 3.2.1 ). A.2.2.2 . Enzymes: 1. 10 U/ l T4 Polynucleotide Kinase 2. 10 U/ l Nt.BbvCI 3. 10 U/ l Nb.BbvCI 4. 100 U/ l E x onuclease III 5. 6. - Fidelity DNA Polymerase 7. 40 U/ l Taq DNA Ligase 8. 20 U/ l DpnI *Diluen t for all enzymes required for Section A. 3.2.1 is 1X NEB CutSmart Buffer. A.2.2.3 . Equipment and Materials: 1. Zymo Clean & Concen t rator - 5 kit (Zymo Research) 2. Corning square bioassay dishes, 245 mm x 245 mm x 25 mm (Sigma - Aldrich) A.2.3 . Chemically Competent Library Yeast Transformation A .2.3.1 . Yeast Solutions and Plates: 1. Growth Media: Synthetic Dextrose medium supplemented with Cas a mino acids (SDCAA): 2% w/v dextrose (D - glucose), 0.67% w/v yeast nitrogen base without amino acids (Sigma - Aldrich), 0.5% w/v bac to casamino acids technical (BD Biosciences), 0.54% w/v Na 2 HPO 4 , and 0.856% w/v Na 2 HPO 4 2 O. Filter sterilize. Add 1% v/v of 10 , 000 U/ml Penicillin - Streptomycin immediately prior to growth to prevent bacterial contamination. 78 2. Induction Media: Synthetic Gal actose medium supplemented with Casamino acids (SGCAA): prepare like SDCAA but with 2% w/v of galactose instead of dextrose. 3. S D CAA agar plate: 0.54% w/v Na 2 HPO 4 , 0.856% w/v Na 2 HPO 4 2 O, 18.2% w/v sorbitol, and 1.5% w/v agar. Sterilize by autoclaving. 2% w /v dextrose (D - glucose), 0.67% w/v yeast nitrogen base without amino acids, 0.5 % w/v bacto casamino acids technical. Sterilize b y filtrating. Add the filter sterilized solution into the cool autoclaved mix (approximately below 50°C) at 1:10 ratio. Store fo r up to 6 months at 4°C. 4. Yeast storage buffer: 20% w/v glycerol, 20 mM HEPES, and 150 mM NaCl pH 7.5. Filter sterilize. A.2.3. 2 . Reagents: 1. 10 mg/ml Salmon Sperm DNA (Invitrogen) 2. 50% w/v Polyethylene Glycol, PEG Filter sterilize. 3. 1 M Lithium Acetate, LiOAc A.2.4. Library Screening A .2.4.1. Buffers and Reagents: 1. Phosphate buffered saline, PBS, at pH 7.4: 0.8 w/v NaCl, 0.02% w/v KCl , 0.144% w/v Na 2 HPO 4 , and 0.024% w/v KH 2 PO4. Sterilize by filtrating. 2. Phosphate buffered saline with bovine serum albumin (PBS - BS A) at pH 7.4: prepare as PBS and supplemented with 0.1% w/v bovine serum albumin (BSA). Sterilize by filtrating. 3. Anti - c - myc - FIT C antibody, FITC (Miltenyi Biotec) 4. Streptavidin, R - Phycoerythrin Conjugate, SAPE (Thermo Fisher) 5. Biotinylated PPI partner protein (see Note A1 ). 79 A.2.5 . Deep Sequencing Preparation of Yeast DNA A .2.5.1. Buffers and Reagents : 1. TE media : 10 mM Tris - HCl at pH 8 .0 and 0.1 mM EDTA 2. 5 U/ l Zymolyase (Zymo Research) 3. 10X Lamba Nuclease buffer (NEB) 4. SYBR - Gold Nucleic Acid Gel Stain (Thermo F isher) 5. Agencourt AMPure XP (Beckman Coutler) 6. Quant - it Pico Green dsDNA Assay kit (Life Technologies) 7. 70% v/v Ethanol A.2.5.2 . E n zymes: 1. 5,000 U/ml Lambda Nuclease (NEB) A.2.5.3 . Equipment: 1. Zymo Research Yeast Plasmid Miniprep II kit 2. Qiagen mini - prep kit 3. 9 6 - well magnetic plate A.3. Methods A.3.1. Library Preparation: Site - Saturation Mutagenesis (SSM) B ecause a protein of 250 amin o acids is encoded by a 750 - bp gene, separate SSM libraries are prepared for the gene of interest ( Figure A 1 a ) to allow compatibi lity with 250 bp paired end (PE) Illumina M iSeq sequencing reads (see Note A2 for considerations for library preparation ) . A.3 . 1.1. Design of mutagenic oligonucleotides NNK to 80 cover all possible point mutations, where N represents any of the A/T/G/C, and K represents T/G. Mutagenic oligos are desig n ed to be complementary to the wild - type template sequence as determined by the orientation of the BbvCI restriction site on the pETc onNK vector ( Figure A 1 b - c; see Note A3 ). 1. Design your mutagenic oligos using Quick Change Primer Design Program (www.agilen t codon to cover all possible 20 amino acids at each codon position. 2. Order the mutagenesis oligos on a 5 00 picomole DNA Plate Oligo from Integrated DNA in TE, pH 8 . A.3.1.2. Preparation of S S M Libraries by Nicking Mutagenesis This protocol is exactly as described in Wrenbeck et al . 88 . All reactions sh o uld be prepared on ice unless otherwise stated. 1. primer. 2. Into a PCR tube, add 2 I ncubate the reaction mixture at 37°C for 1 hour. 3. 2 O eaction mixture at 37°C for 1 hour. 4. Store phosphorylated oligos at - 20°C. 5. The day of mutagenesis, dilute phosp horylated mutagenic primers 1:1000 and SEC_Rev primer 1:20 using NFH 2 O ( see Note A 4 ). 81 6. For the preparation of ssDNA template strand, in a PCR tu be, add 0.76 pmol of dsDNA plasmid (approximately 2 - Exon uclease III (final concentration of 10 U/ and NFH 2 7. Place the tube in a pre heated (37°C) thermal cycle with the following program: 60 minutes at 37°C, 20 minutes at 80°C, and hold at 4°C. Figure A 1 . Essential considerations needed for preparing Site Saturation Mutagenesis (SSM) Libraries. ( a . ) The ge ne of interest is segmented in multiple libraries containing contiguous sections of 200 - 250 bp. Here, sections o f 225 bp are shown for compatibility with 250 bp PE Illumina MiSeq sequencing. ( b . ) po ssible 20 amino acids. ( c . ) An Nt.BbvCI restriction enzyme (Nt) is used to create a nick on the sense strand. Mu tagenic oligos are designed to be complementary to the antisense ssDNA template 8. T o proceed with the comprehensive codon m utagenesis on the first strand, in each PCR tube, add 2 82 phosphorylate + Taq - Fidelity DNA Polymerase. Mix the tube content briefly. 9. Place the tube into a preheated (98°C) thermal cycler with the followin g program: 2 minutes at 98°C, 15x cycles of 30 seconds at 98°C, 45 seconds at 55°C and 7 min utes at 72°C, followed by a final incubation at 45°C for 20 minutes, and hold at 4° C . Add 10. Purify each r using NFH 2 O accord 11. To degrade the template strand, transfer 1 4 of Exonuclease I. 12. Place the reacti o n tube in a preheated (37°C) thermal cycle with the following program: 60 minute s at 37°C, 20 minutes at 80°C, and hold at 4°C. 13. To synthesize the 2 nd mutagenic strand, ad d 2 e + Taq High - Fidelity DNA Polymerase to the same reaction mixture . Mix the tube content briefly. 14. Place the tube in a preheated (98°C) therma l cycler with the following program: 30 seconds at 98°C, 45 seconds at 55°C, 10 m inutes at 72°C, 20 minutes at 45°C, and hold at 4°C. 83 15. ube and incubate the reaction for 60 minutes at 37°C to degrade methylated and hemi - methylated wild - type DNA. 16. Purify each reaction using a Zymo Clean and Concentra using NFH 2 17. - Blue following standard electrocompetent transformation protocol 98 . 18. After recovery, bring the final volume of the transformation to 2 - 2.5 mL with additional sterile media (TB media). 19. Prepare six 10 - efficiency, the next day c ount the section that contains between 10 - 100 colonies. It is important to obtai n at least 99.9% coverage of the theoretical diversity o f the library ( see Note A5 ). 20. Spread the remaining cells onto the prepared large BioAssay dishes. 21. Place in a 37°C humid ity - controlled incubator overnight when BioAssay dishes have dried. A.3.1.3. Ext raction of dsDNA SSM Library Plasmid 1. On the next day, sc rape the large plates using between 5 - 10 mL TB media and collect the cells in a 50 ml centrifuge tube. 2. Vortex the cell s uspension and extract the library plasmid DNA of a 1 ml aliquot of the cell susp ension using a Qiagen mini - prep kit. Additional m ini - pre ps can be done if large amounts of library DNA are required. 3. Store the rest of the cells at - 80°C by resuspending the p ellet in 3 ml of 50% v/v glycerol. 84 A. 3.1.4. Chemically Competent Library Yeast T ransformation Competent yeast can be prepared up to six months ahead of time. 1. Grow the EBY100 cells in 500 mL YPD to an OD 600 of 1.2 and are harvested at 4000xg for 5 minutes. 2. Resuspend the pelleted cells in 250 ml sterile H 2 O, and repellet. 3. Resuspend th e pelleted cells in 10 ml of 100 mM LiOAc and repellet. 4. Resuspend in 3.5 ml of 100 mM LiOAc and then, add 1.5 ml of 50% v/v glycerol and the mixture vortexed. 5. Prepare aliquots - 80 o C. Do not snap - freeze cells. 6. 7. Add 7 petent EBY100 cel ls. 8. Vortex hard until there is a uniform mixture. 9. Add 5 of library plasmid to the mixture and vortex briefly. 10. Incubate the mixture a t 30°C for 30 minutes. 11. Heat shock the cells by incubating at 42°C for 20 minutes. 12. Pellet the cells by s pinning at 14000 rpm for 30 seconds. 13. Resuspend the cells pellet in 1 ml of SDCAA media and let stand for 5 minutes. 14. Prepare six 10 - fold serial dilutions from the suspension and plate on SDCAA plates using for 2 3 days at 30°C to cal culate transforma tion efficiency ( see Note A 5 ). 15. Add the remaining culture into 100 ml of SDCAA media. Grow for 30 hours at 30°C and 250 rpm. 85 16. On the next day, resuspend the cell culture at OD 600 =1 in 50 ml of SDCAA media. 17. Grow overnight at 30°C and 250 rp m. 18. Prepare multip le cells stocks by pelleting, resuspending in yeas t storage buffer to an OD 600 =1, and storing in 1 mL aliquots (approximately 1x10 7 cell s) at - 80°C. Do not snap - freeze cells ( see Note A 6 ). A.3.2. Library Screening A .3.2.1. Preparation of Labeling Reaction s 1. For each PPI partner to analyze, thaw a 1 ml aliquot as prepared on previous section, spin down at 2500g for 3 minutes, and remove th e supernatant. 2. Resuspend the pellet in 1 ml SDCAA media and grow for 4 - 6 hours at 30°C and 250 rpm. 3. S pin down the cell s a t 2500g for 3 minutes and re - inoculate at OD 600 = 1.0 in 1 ml of SGCAA media. Induce overnight using the predetermined induction cond itions ( see Note A 7 ). 4. Spin down the cells at 2500g for 3 minutes, wash with 1 ml of ice - cold PBS - BSA, and spin down aga in. 5. Resuspend the cells in ice - cold PBS - BSA at an OD 600 = 2.0 . 6. In PBS - BSA, label 1 ml (2x10 7 ) cells with the biotinylated protein at h alf of the apparent dissociation constant and incubate at room temperature for 30 minutes using a table top mixer. Vary the total reaction volume to ensure that the number of biotinylated protein is at least 10 - fold higher than the PPI partner that is disp layed on the yeast cell surface. For example, assuming a 10:1 partner:displayed protein ratio at a typi cal PPI apparent dissociation constant of 10 nM, 2x10 7 cells (1 ml) should be labeled with 5 nM biotinylated partner protein (half of the apparent dissoc iation constant). The total reaction volume is 86 calcula ted following equation A. 3.1. Thus, label 1 ml of f PBS - BSA with A. 3.1 7. Spin down at 2500g for 5 minutes, wash the pellet with 5 ml of PBS - BSA and spin down and remove supernatant again . In this and subseque nt steps, PBS - BSA should be ice - cold, the tabl etop centrifuge should be refrigerated, and all tubes sho uld be kept on ic e and protected from light. 8. - BSA, vortex briefly and incubate the label ed cells on ice for 10 minutes. 9. Repeat step 7. 10. Leave the cell pellet on ice until ready to sort. A. 3.2.2 Sorting Conditions Set - up 1. Set Gate1, Gate2, and Gate3 on your cell sorter as shown in Figure A2 . 2. Add 4 ml of ice - cold PBS - BSA to the cell pellet, mix b y vortexing, and transfer to a FACS - compatible tube. 3. Obtain the reference population by sorting 240,000 cells ( see Note A 8 ) using Gate 1 + ( Figure A2 a ). 4. Obtain the displayed population by sorting 240,000 cells using Gate1 + /Gate2 + ( Figure A2 b ). 87 Figure A 2 . Sorting gates used for library screening . Yeast SSM libraries are labeled with biotinylated complementary protein at half of the apparent dissociation constant. Next, SSM libraries are sorted using three different gates as shown : ( a . ) Gate 1 set with the light scatter parameters for yeast, forwar d scatter/side scatter; ( b . ) Gate 2 set on the forward scatter and the fluoresce n ce channel for displayed protein (FITC); and ( c . ) Gate 3 set on the fluorescence channel for displayed pro tein and fluorescence channel for bound protein. Gate 3 is configured to collect the top 5 - 10% of the bound population. 5. O btain the bound population for each PPI by sorting 240,000 cells using Gate1 + /Gate2 + /Gate3 + ( Figure A2 c ). 6. Recover the collected cell s in 5 ml of SDCAA media for approximately 30 hours at 30°C and 250 rpm. 7. Prepare cells stocks by storing 1 ml of OD 600 =4 cell stocks in yeast storage buffer and at - 80°C. A.3.3. Deep Sequencing Preparation of Yeast DNA A . 3.3.1. Primer Design and Library Am plification Test Yeast DNA is prepared for deep sequencing using a 2 - step PCR amplification: the first step is with a gene - r Figure A3 ) . Inner primers are design ed to be complementary to adjacent an Illumina universal primer sequence ( Figure A3 a ). The 88 following rules needs to be considered to determine these regions: 1. The length of the segment section plus the librar y should not be longer than 250 base pairs. 2. Design the segment region to have a melting temperature of 53 - 56°C using the NEB Phusion melting point c a lculator using Phusion High - Fidelity Polymerase. 3. Once th e gene - specific sequence is designed, append the conserved primer sequence as shown in Table A 1 . 4. Upon receiving the inner primers, we recommend performing a PCR verification with wild - type plasmid a s a template to confirm a single band of the expected size . Further steps for yeast DNA deep sequencing pre paration requires the addition of universal primers to add the Illumina adapters and barcodes. Universal primers are designed using the TruSeq Small R NA Oligo Sequences. The forward primer is the same for al l preparations while the reverse primer contains an indexing barcode that allows multiplexing of samples on an Illumina lane ( Figure A 3 b; full sequences shown in Table A1 ). A.3 . 3.2. Yeast plasmid D N A preparation for Deep Sequencing 1. Thaw an aliquot of the stored yeast library by hand, spin down at 2500g for 3 minutes, and remove the supernatant. 2. Resuspend the pellet cells in 200 l of Solution 1 and add 5 l of 5U/ l of Zymolyase. 3. Incubate at 37°C for 4 hours and mix once per hour. 4. Perform one freeze - tha w cycle in dry ice/EtOH bath and 42°C incubation. 5. Ad d 200 l of Solution 2, mix briefly and incubate for 3 - 5 minutes at room temperature. 6. r 5 minutes. 7. Transfer the supernatant to a Qiagen mini - prep column and spin down for 1 minute at 17,000g. 89 Figure A 3 . PCR steps performed for deep sequencing preparation of SSM libraries. Sequential PCR reactions to amplify the ge nes of interest and attach the Illumina adapters are shown for SSM library 2 (gold). ( a . ) After extractin g the plasmid DNA from yeast cells, SSM libraries are amplified by PCR using a set of inner primers containing a segment that overlaps with the gene o f interest (light blue) and the Illumina universal sequence (purple). ( b . ) A second round of PCR is perfor med to attach the Illumina adapter sequence using a set of outer primers which contain an overlapping region to the Illumina universal sequence (purp l e), a unique barcode on the reverse prime r (green), and Illumina adapter sequences (yellow) . 8. A dd 700 l of PB buffer and spin down for 30 seconds at 17,000g. 9. Add 700 l of PE buffer and spin down for 30 seconds at 17,000g. 10. Repeat step 9. 11. Take out supernata nt and spin down again at 17,000g for 1 minute to dry the column. 12. Transfer the column to a new clean 1.5 ml microfuge tube, add 30 l of elution buffer and spin down for 1 minute at 17,000g. 90 13. Reload the column with the eluate and spin down again. Store 15 l of eluate and proceed with the remaini ng 15 l. 14. For the purification of plasmid from the yeast preparat buffer. 15. Place the mixture in a preheated (30°C) thermocycler with the following cycle: 90 minutes at 30°C, 20 minutes at 80°C, and hold a t 4°C. 16. Clean the PCR product following the standard procedure from Qiagen mini - prep PCR 17. Store 15 l of eluat e and proceed with the remaining 15 l. 18. Phusion HF buffer, 18.5 2 gh - Fidelity Polymerase. 19. Place the tube in a preheated (98°C) therm ocycler with the following cycle: 30 seconds at 98°C, 14x cycles of 5 seconds at 98°C, 15 seconds at 53°C and 15 seconds at 72°C, follow by a final i ncubation for 10 minutes at 72°C, and a hold at 4°C. 20. 21. Place the t ube back in the thermocycler with the following cycle: 30 minutes at 37°C, 5 minutes at 95°C, and a hold at 4°C. 22. 2 l of 10 mM 91 - Fidelity Polymerase. 23. Repeat the same PCR cycle used for the inner prime rs. 24. YBR - Gold. It is important to verify that you have single clear band before proceeding ( see Note A 9 ). 25. Purify and clean the PCR product using Agencourt AMPure XP following the instructions for the 96 - well format procedure. 26. Measure the concent ration of each sample. 27. Stored the purify product at - 20°C. Table A 1 . Gene amplification and Illumina adapter primers to prepare samples for deep sequencing . I nner Primers for Library Amplification Primer Name Sequence Inner_FWD - gttcagagttcta cagtccgacgatc - Inner_REV - ccttggcacccgagaattcca - Outer Primers to add the Illumina adapters and barcodes Illumina_FWD - aatgatacggcgaccacc gagatctacac gttcag agttctacagtccgacgatc - Illumina_REV - caagcagaagacggcatacgagat nnnnnn gtgactggagtt ccttggcacccgagaattcca - Magenta : Illumina Adapter, nnnnnn : Indexing Barcode (see Kowalsky et al . 25 for complete set) , and Purple : Illumina Universal Sequence A . 3.3.3. dsDNA Quantification using Quant - iT Pico Green A t this point, samples are ready for deep sequencing. Follow the instructions for the Illumina MiSeq 2x250 b p Submission from your Sequencing Facility. Usually, each Illumina MiSeq sequenc ing holds between 10 - 15 million reads per lane. Based on the read dep th and library size, calculate the amount of reads necessary for each sample our group uses approximately 500,000 reads per sample and multiplexes 20 - 30 samples per lane. Individual sam ples are quantified and mixed together in a single vial. The followin g procedure was adopted from the Invitrogen MP 07581 manual. The final yield should be about 1 - 4 ng in 40 l. 92 1. Allow the Quant - iT reagent to warm to room temperature while covered in foi l. 2. Prepare a 200 - fold dilute solution of Quant - iT into TE buffer usin g a foil covered culture ould be prepared and used the day of the experiment. 3. Beginning with a 50 ng/ml s tock of a kit - supplied Lambda DNA standard, prepare a blank and a 1:2 - well black plate. 4. In a black 96 w 5. Carry out extra dilutions as necessary if the concentration is too high. 6. f diluted Pi co Green solution to DNA samples and standard samples, mix briefly, and incubate for 5 minutes at room temperature covered with foil to protect from light. 7. Measure the fluorescence of the samples (excitation ~480 nm, emission ~520 nm). 8. Subtract the fluorescence value of the reagent blank from that of each of the samples. 9. Use the corrected data to ge nerate a standard curve of fluorescence versus DNA standard concentration s and calculate the concentrations of each sample. In our hands the final co ncentration is between 5 - 10. Mix equivalent mass amounts of samples in a single 1.5 ml Eppendorf tube and send to your sequencing facility. A. 3.4 . Data Analysis C ustom scripts used in the data analysis are available at Github (user: JKlesmith). Sa mple command lines and instructions are provided at the same source. 93 1. Use the modified version of Enrich 0.2 software as describe in Kowalsky et al. 25 to compute the enrichmen t ratios of individual mutants for the DNA sequencing results from Illumina MiSeq run ( see Figure A 4 and Note A 10 ). Enrich 0.2 16 documentation is available at http://depts.washington.edu/sfields/sof tware/enrich/docs/0.2/enrich.html. The output from Enrich 0.2 is required as input for the remaining steps. The wild - type pr otein sequence is also required as input for the following steps. 2. The relative binding of each variant on the displayed and bound po pulation is calculated using a custom Python script called QuickNormalize.py ( see Note A 11 ). The output from this script is a .csv file that can be read by multiple programs. In our lab we use Microsoft Excel to visualize the data as heatmaps and to carry out the data analysis ( see Note A 12 ). 3. Calculate the Shannon Entropy for each variant on the displayed and bound population u sing a custom script called FACSEntropy.py - the output file is a .csv. The entropy values are used to discriminate those residues t hat participate in the protein - protein interaction and to determine the conformational epitope following the cut - off analysi s flowchart as shown in Figure A5 ( also see Note A 13 ). 4. Calculate the reportable statistics using QuickStat.py script. Statistics will report the reads passing through enrich, the percentage of possible codon substitutions observed, the percent of reads wit h none, one, and multiple nonsynonymous mutations, and the coverage of possible single nonsynonymous mutations. A.4 . Notes 1. The PPI partner protein is chemically biotinylated following the instructions for EZ - Link NHS - Biotin Reagents (Thermo - Fisher). We pr efer chemical biotinylation to genetically encoded biot inylation (e.g. avi - tag) as the former has a higher fluorescence signal. If 94 Figure A 4 . Deep sequencing results and data analys is used to determine the conformational epit ope. ( a . ) DNA sequencing results are processed using Enrich 0.2 software 99 to calcul ate the frequency, F v,X , of each point mutant, v, for each position, x, in the primary sequence. ( b . ) The frequency data of each variant from different populations is transformed into heatmaps comparing the re l ative fluorescence of each variant in the disp layed population (top) and the bound population (bottom) against the unselected population. ( c . ) Heatmaps are used to calculate the Shannon entropy for each residue on the displayed (black) and bound populatio n s (turquoise). Next, the entropy is used to de termine the conserved and non - conserved positions which allow to identify the conformational epitope. Figure A 5 . Flow - chart of analysis used to determine the conformation al epitope. 95 p roteins are small, covalent labeling with multiple biotins may disrupt the structure; in such a case we recommend genetically fusing the PPI partner to a carrier like maltose binding protein or an IgG Fc . Anecdotally, we have noticed clean er r esults with PPI partners with mo novalent interactions , and for that reason recommend creating a Fab if the PPI partner is a mAb. 2. The following rules apply for preparing separate mutagenesis libraries: (i.) The length of each library should be divisibl e by three to avoid splitting a codo n; (ii.) the gene should be segmented into libraries with a maximum length of 225 base pairs for Illumina 250 bp paired - end sequencing (273 base pairs for 300 bp paired - end sequencing); and (iii.) libraries should be sim ilar in length (+/ - three nucleotide s). 3. In some cases, the gene sequence of interest also contains a BbvCI restriction site. If the site is in the same orientation as the site on the pETconNK plasmid, continue the protocol as usual. If the BbvCI site is in t he opposite direction as the sit e on the pETconNK plasmid, use the YSD plasmid pETCON (Addgene # 41522) as this plasmid does not contain a nicking site. The orientation of the BbvCI may not be in the same way as exists on the pETconNK plasmid. For exam ple, if the nicking site is in the o pposite direction from Figure A1 b , Nb.BbvCI (not Nt.BbvCI) should be used first to create the ssDNA wild - type template; otherwise follow the protocol as described in Section A.3.1.2 . 4. We recommend preparing phosphorylat ed o ligos no earlier than the day be fore the nicking mutagenesis procedure. Avoid repeated freeze/thaw cycles. 5. For a library with NNK SSM at 75 amino acids the theoretical library size is 2,400 nucleotide variants. The percentage theoretical coverage is de scri bed by the following 96 equation: . In the above case, 16,500 transformants will give 99.9% coverage. 6. At this point, cells could be inoculated in fresh SGCAA media fo r Library Screening Preparation (Section A . 3.2 ) or frozen aliquots can be prepared for long - term storage. We often re - inoculate the cells in SGCAA media to an OD 600 = 1 and induce at 22°C to co nfirm that the mutagenesis libraries display on the yeast surfac e and binds the PPI partner. We prepare between 20 - 48 aliquots for long - term storage. 7. Induction temperature should be the same as used to prepare the PPI partner in a YSD format according to C hao et al. 17 . For each new YSD protein our lab tests induction of surface d isplay at 18, 20, 22, and 30 o C. 8. It is important that the number of collected cells should be at least 100 - fold higher than the theoretical library size to avoid complexity bottlenecks. For example, at least 240,000 cells should be collected for each sorted population for a library with NNK SSM at 75 amino acids with a theoretical library size of 2,400 nucleotide variants. 9. If the correct band size was not obtained from the second PCR product, we recommend on SYBR - Gold to identif y which PCR amplification did not work. We recommend sta ining the gel on SYBR - Gold for at least 1 hour to resolve low intensity bands. Fewer or more cycles of each PCR could be used to improve the product. 10. Our group routinely analyzes the quality of the Illumina sequencing data using FastQC available online at http://www.bioinformatics.babraham.ac.uk/projects/fastqc . Poor quality reads can hinder the data analysis using Enrich 0.2. The quality of the Illumin a sequencing data is highest for the forward read and the first 150 bp. For issues where 97 quality is poor on the reverse read, perform Enrich only for the forward read. We have also performed Enrich for short segments of the reads where the quality is highe st. 11. i ) for variant i is defined as, , A . 4. 1 where is the mean fluorescence of variant i and is the mean fluorescence of wild type. There are a number of assumptions used to calculate relative binding see Kowalsky et al. 12 for further details. 12. P ositions with insuf ficient data at more than 10 substitutions should be excluded from analysis. 13. In the current experimental set - up, discriminating m utations that disrupt the interface and maintain the overall fold between those that destabilize the structure is difficult to determine, as unfolded mutants still predominantly display on the yeast surface 8,74 . However, a recent study confirms that destabilizing mutations display with few er copies on the yeast surface than stabilizing mutations, at least for proteins with >200 residues 75 for small proteins destabilizing mutants appear to display at the same rate as stable mutants (T.A.W. and A.M.C., unpublished data). To furth er identify mutations that stabilize larger proteins, a FACS protocol is used with a sort gate set t o collect the top 5% of the displaying population. For library screening, 2x10 6 yeast cells per ml , in PBS - BSA , are labeled with - c - myc - FITC per 2x10 5 yeast cells. The population is sorted using a gate that collects the top 5% of the displaying population. Shannon entropy obtained from this study is used to identify structurally conserved positions. 98 A PPENDIX B : Preferential Identif ication of Ag o nistic OX40 Antibodies by Using Cell Lysate to Pan Natively Paired, Humanized Mouse - Derived Yeast Surface Display Libraries Preferential Identification of Agonistic OX40 Antibodies by Using Cell L ysate to Pan N atively Paired, Humanized Mouse - Derived Yeast An tibodies ( 2019 ) 8: 17 *Supplementary Materials are available online at http://www.mdpi.com/2073 - 4468/8/1/17/s1 99 B.1. Abstract To discover therapeutically rel eva n t antibody candidates, many groups use mouse immunization followed by hybridoma generati on or B cell screening. One modern approach is to screen B cells by generating natively paired single chain variable fragment (scFv) display libraries in yeast. Suc h m e thods typically rely on soluble antigens for scFv library screening. However, many thera peutically relevant cell - surface targets are difficult to express in a soluble protein format, complicating discovery. In this study, we developed methods to screen hu m anized mouse - derived yeast scFv libraries using recombinant OX40 protein in cell lysate. We used deep sequencing to compare screening with cell lysate to screening with soluble OX40 protein, in the context of mouse immunizations using either soluble OX 40 o r OX40 - expressing cells and OX40 - encoding DNA vector. We found that all tested methods p roduce a unique diversity of scFv binders. However, when we reformatted forty - one of these scFv as full - length monoclonal antibodies (mAbs), we observed that mAbs i den t ified using soluble antigen immunization with cell lysate sorting always bound cell surf ace OX40, whereas other methods had significant false positive rates. Antibodies identified using soluble antigen immunization and cell lysate sorting were also sig nif i cantly more likely to activate OX40 in a cellular assay. Our data suggest that sorting w ith OX40 protein in cell lysate is more likely than other methods to retain the epitopes required for antibody - mediated OX40 agonism. B.2. Introduction Many antibod y d r ugs bind to disease targets expressed on cell surfaces. For example, antibodies may bind to the surface of tumor cells and induce antibody - dependent cellular cytotoxicity (ADCC). Conventionally, antibody drug discovery groups use either hybridomas 100 or phage display 101 to discover antibod y d r ugs. Hybri domas are typically screened for cell surface 100 binders using enzyme - linked immunosorbent assays (ELISAs) in 96 - well plates 102 . Hybridoma methods, therefore, require expensive robotics to screen thousands of antibody candidates. Phage display has a much higher throughput, bec aus e billions - diverse phage libraries can be panned against cells affixed to well plates 103 . However, most therapeutic antibodies have been discovere d i n mice 26 , perhaps due to difficulties with developability of artificial a nti b odies, suc h as low solubility binders discovered in phage display 104,105 . Recently, we invent ed a novel met hod for screening millions - diverse antibody repertoires using microfluidics, yeast display, and deep sequencing 28 30 . Our method leverages the developability advantages of naturally paired antibodies with the massively parallel throughput of display technologies. Other groups later further v ali d ated our w ork with similar methods 106,107 . However, our previously published methods required soluble antigen for both mouse immunization and fluorescence - activated cell sorting ( FACS). This li m itation excluded the possibility of using the method to identify antibodies against multi - pass transmembrane proteins, such as G - protein coupled receptors. Additionally, the requirement for soluble protein may lead to antibodies directed aga inst spurious e pitopes not present in the native conformation on the surface of target cells. OX40, or tumor necrosis factor receptor superfamily member 4 (TNFRSF4), is a costimulatory immune receptor transiently expressed on T cells which upregulates T ce ll activity up o n binding to its ligand, OX40L. Therapeutic agonism of OX40 may increase T cell differentiation and tumor killing functions 108 . Agonism requires a ligand binding to OX40 in a way that generates complexes of crosslinked OX40 molecules on cell surfaces 109 . Although the crystal structure of OX40 binding to OX40L has been resolved 110 , the specific epitopes required for agonism are not well understood. Developmen t of novel the r apeutic antibodies would benefit from 101 a method that generates large panels of antibodies directed against a variety of OX40 epitopes that are bioavailable at the cell surface. To improve OX40 antibody discovery, we adapted our previously pub lished methods 28 30 to test different immunization methods (cell s versus soluble antigen) and different antibody selection methods (cell lysate versus soluble antigen). The cell lysate selection method was adapted from prior work 111,112 , specifically by using a peptide tag rather than biotin to label the cell lysate. We synthesized forty - one monoclonal antibodies (mAbs) from the various methods and found that soluble OX40 soluble antigen immunization followed by s orting with ce l l lysate was most likely to identify antibodies that bind cell surface antigen and yielded more antibodies that activate OX40 in cellular assays. B. 3 . Materials and Methods B. 3 .1. Mouse Immunization and Sample Preparation A ll mouse work was performed at An tibody Solutions (Sunnyvale, CA, USA) and overseen by a licensed veterinarian. All experiments were performed using mice from Trianni (San Francisco, CA, USA), which are C57BL/6 that transgenically express a complete reper toire of fully huma n immunoglobuli n gamma (IgG) and immunoglobulin kappa (IgK) V(D)J genes, but retain mouse promoters, introns, and constant domains. For the soluble OX40 immunizations, five Trianni mice were immunized with recombinant His - tagged human OX 40 extracellular do main (Acro OX40 - H5224, Newark, DE, USA), using ALD/MDP (alhydrogel/muramyl dipeptide) as an adjuvant. 10 g of OX40 protein with adjuvant was injected into the footpad twice per week for three weeks . We assessed titer at Day 21 with ELISA, using a dilution series of anti gen, ranging from 1000 ng/mL to 1 ng/mL and goat anti - 102 mouse IgG - HRP (Jackson ImmunoResearch 115 - 035 - 071, Wes t Grove, PA, USA) ( Supplementary Figure S1 ). After assessing serum titer, two more footpad boosts of 10 g without adjuvant were admi nistered to eac h animal before sacrifice . For the cells/DNA OX40 immunizations, we first transfected Flp - In 3T3 cells (Ther mo Fisher Scientific, Waltham, MA, USA ) with a vector encoding un - tagged, full - length human OX40 ( Supplementary Figure S2 ). A pool of OX40 - positive cells was selected using Hygromycin B (Gemini Bio 400123, West Sacramento, CA, USA) for 2 weeks. Cells were treated with Mitomycin C before cryopreservation. One to two million cells were injected per mouse. A footpad injection was performed with three Tri anni mice on day 0 with cells, then days 3, 7, and 10 with 20 µg DNA plasmid encoding full - length, untagged human OX40, then day 14 with cells, day 17 and 21 with DNA, and final boosts on Days 24 and 27 with cells prior to tissue harvest. Be fore the final boosts, mouse serum titer was assessed with flow cytometry, using a dilution series of each starting at 1:200 and ending at 1:145, 000 ( Supplementary Figure S3 ). Briefly, the same 3T3 cells stably expressing OX40 were incubate d with the seru m dilution, washed, and then stained with goat anti - mouse IgG - rPE (Jackson ImmunoResearch, 115 - 116 - 071, West Grove, PA, USA). The final library was generated from two of the three mice, as the third mouse died prior to tissue harvest. We sur gically removed lymph nodes (popliteal, inguinal, axillary, and mesenteric) and spleens from the sacrificed animals. Single cell suspensions for spleen and lymph nodes were made by manual disruption followed by passage through a 70 m filter. We used the E P an - B Cell Isolation (Stemcell Technologies, Vancouver, Canada) negative selection kit to isolate B cells fro m the single cell suspensions. Cells were stained for viability using Trypan blue and 103 then quantified with a C - Chip hemocytometer (In cyto, Chungnam - do, Korea). We then diluted the cells to 6000 cells/ L in PBS with 12% (Si gma, St. Louis, MO, USA). The purified cell populations were used for microfluidic encapsulation as described below. B. 3 .2. Generatin g Paired Heavy and Light Chain Libraries A s described previously 28 30 , the generation of libraries comprised of three steps: (i) poly(A)+ mRNA capture, (ii) multiplexed over lap extension reverse transcriptase polymerase chain reaction (OE - RT - PCR), and (iii) nested PCR to remove artifacts and add adapter s equences for de ep sequencing or yeast display libraries. Briefly, we isolated 1.6 1.9 million B cells into fluorocarbon oil (Dolomite, Royston, UK) emulsion microdroplets ( Supplementary Table S1 ) with a lysis buffer (20 mM Tris pH 7.5, 0.5 M NaCl, 1 mM EDT A, 0.5% Tween - 2 0, and 20 mM DTT) and oligo(dT) beads (New England BioLabs, Ipswich, MA, USA ), using an emulsion droplet microfluidic chip 28 30 . We purified beads from the droplets using Pico - Break solution (Dolomite, Royston, UK). We then performed multiplex OE - RT - PCR in emulsions, using purified RNA - boun d beads as a te mplate, as described elsewhere 28 30 . The OE - RT - PCR product was gel purified and PCR was performed to add adapters for Illumina sequencing or yeast display; for sequencing, a randomer of seven nucleotides was added to increase base calling accuracy in subsequent next generation sequencing st eps. Nested PCR is performed with 2 × NEBNext High - Fidelity amplification mix (New England BioLabs, Ipswich, MA, USA ) with either Illumina adapter containing primers or primers for cloning into the yeast expression vector. 104 B. 3 .3. Yeast Library Screening S ac charomyces cere visiae EBY100 cells (ATCC, Manassass, VA, USA ) were electroporated (Bio - Rad Gene Pulser II; 0.54 kV, 25 uF, resistance set to infinity) with gel - purified nested PCR product and linearized pYD vector 28 30 for homologous recombination in vivo. Transformed cells were expanded and induced with galactose to ge nerate yeast scFv display libraries. For the soluble OX40 FACS experiments, human OX40 - His (described above) protein was biotinylated using the EZ - Link Micro Sulfo - NHS - LC - Biotinylation kit (Thermo Fisher Scientific, Waltham, MA, USA ). The bi otinylation rea gent was resuspended to 9 mM and added to the protein at a 50 - fold molar excess. The reaction was incubated on ice for 2 hours, and then the biotinylation reagent was removed using Zeba desalting columns (Thermo Fisher Scientific, Waltham, M A, USA ). The fi nal protein concentration was calculated with a Bradford assay. The scFv libraries were then stained with anti - c - Myc (Thermo Fisher Scientific A21281, Waltham, MA, USA ) and an AF488 - conjugated secondary antibody (Thermo Fisher Scientific A11 039). Biotinyla ted OX40 was added to the yeast culture (250 nM final concentration) and stained with APC - streptavidin (Thermo Fisher Scientific, Waltham, MA, USA ). Approximately two million cells were then flow sorted on a FACSMelody (BD, San Jose, CA, USA ) for double po sitive cells (AF488+/APC+). Populations of bi nder scFv clones were recovered, expanded, and then subjected to a second and third round of FACS with the same antigen at 250 nM final concentration. A fourth round of FACS was additionally perfo rmed on select samples ( Supplementary Figure S4 ). For the ce lls/DNA OX40 FACS experiments, we engineered an expression vector that expresses full - length human OX40 fused to a FLAG peptide at the N - terminus ( Supplementary Figure S2 ). This vector was used to stably transfe ct CHO cells via targeted genome integration. 105 Approximately 12.5 × 10 6 OX40 - positive transfected cells encoding full - length human OX40 were used to prepare the cell lysate for each staining condition. First, cells were harvested and washed t wice with 10 mL of ice - cold PBS. Second, cells were resuspen ded in a lysis buffer (PBS, 1% Triton X - 100, 2 mM EDTA, and 1 × protease inhibitor cocktail) to a final concentration of 5 × 10 7 cells/mL and were incubated, rotating for 30 mins at 4 °C 111 . Finally, cells were harvested an d the supernata nt (the detergent - solubilized cell lysate) wa s removed to a fresh tube and stored at 4 °C until use. The final total protein concentration in the lysate was calculated using a Bradford assay. The scFv yeast libraries were labeled with 250 µL of cell lysate and incubated, rotating, overnight at 4 °C. The next day, labeled yeast cells were stained with anti - c - Myc, an AF488 - conjugated secondary antibody, and APC anti - FLAG (clone L5, BioLegend 637308, San Diego, CA, USA). Approximately, four mill ion cells were flow sorted on a FACSMelody. As described abo ve, the collected populations of binder scFv clones were recovered, expanded, and subjected to two additional rounds of FACS using the same cell lysate concentration. B. 3 .4. Sequence Analysis L ib raries were seq uenced on a MiSeq (Illumina, San Diego, CA, U SA) using a 500 cycle MiSeq Reagent Kit v2, as described previously 28 30 . Sequencing was performed in two different runs. In the first run, we directly sequenced the scFv libraries to obtain a forward read of 357 cycles for the light chain comple mentarity - deter mining region (CDR)3 and V - gene, and a revers e sequence read of 162 cycles across the heavy chain CDR3 and part of the heavy chain V - gene. In the second run, we first used the scFv library as a template for PCR to independently amplify heavy and light chai n V - genes. We then obtained a forward read of 255 cycles and a reverse read of 255 cycles for the heavy and light chain Ig separately. The second run yields overlapping reads, which is useful for sequencing error correction. 106 We used previous ly published me thods for error correction, reading frame identification, and FR/CDR junction calls 28 30,113 . We discard reads wi th E > 1 (E is the expected number of errors) , retaining sequences for which the most probable number of base call errors is zero. We also discard singleton nucleotide reads to further improve confidence i n antibody sequ ences. In order to identify V and J gene families and calculate percent identity to germline, we aligned antibody nucleotide sequences with the IMGT database 114 . We defi ne "clones" conservatively, with an emphasis on sequence accuracy. First, we concatenated the CDR3K and CDR3H amino acid sequences from each scFv sequence into a single contiguous amino acid sequence. Next, we used USEARCH 115 to compute the total number of amino acid differences in all pairwise alignments between each concatenated sequence in each cid differences in the concatenated CDR3s were coun ted as a single clone. Finally, we used the majority amino acid identity at each residue position to generate the consensus amino acid sequence of the clone from sequences of the members of the group. To generate clonal cluster plots, we first used USEARCH 115 to generate all pairwise alignments across the complete set of FACS - sorted IgH and IgK scFv sequences ( Supplementa ry Tables S2 - S9 ). We then computed t he total number of amino acid diffe rences between each scFv sequence. We then generated clustering plots using the igraph R package 116 , using the of the nodes correspond s to the frequency of the antibody c lone in the FACS - sorted population: small (<2% frequency), medium (2 - 12% frequency), and large (>12% frequency). An "edge" (a conc atenated CDR3s. 107 B. 3 .5. Monoclonal An tibody Expression and Characterizat ion W e synthesized mAbs by cloning antibody sequences into a variant of the pCDNA5/FRT mammalian expression vector (Thermo Fisher Scientific, Waltham, MA, USA ), as described previously 28 30 . Expression constructs were prepared using a BioXP robotic wor kstation (SGI DNA, La Jolla, CA, USA). Human IgHG1 is otype was used for all constant domains. MAb plasmids were then transiently transfection into ExpiCHO cells (Thermo Fisher Scientific, Waltham, MA, USA ). Transfected cells were cultured in ExpiCHO medium for 7 9 days. An IgG ELISA kit (Abcam, Cambridge, UK ) was used to quantify the concentration of antibody in the supernatants. To measure cell surface binding, we first generated stable human OX40 - expressing Flp - In Chinese hamster ovary ( CHO) cells (Thermo Fisher Scientific, Waltham, MA, USA ). One million ce lls (1:1 mix of OX40 and irrelevant PD - 1 - expressing negative control cells) were stained with 1 g of anti - min at 4 °C. Cells were t hen co - stained with anti - human IgG Fc - PE (BioLegend clone M1310G05, San Diego, CA, USA) and anti - human PD - 1 - APC (BioLegend clone EH12.2H7, San Diego, CA, USA) antibodies for 30 min at 4 °C. We then used a FACSMelody (BD, San Jose, CA, USA ) quantify binding . We used FlowJo to determine the intensity of the OX40 - expressing cells versus the irrelevant negative controls ( Supplementary Figure S5 ). For measur ement of the kinetics of binding to soluble OX40, 5 g/mL antibodies were loaded onto a Protein A biosenso r using the Octet Red96 system (ForteBio, Fremont, CA, USA) by a contract research organization (Bionova, Fremont, CA, USA). Loaded biosensors were di pped into His - tagged OX40 extracellular domain (Acro OX40 - H5224, Newark, DE, USA) at 200 nM, 108 100 nM, and 5 0 nM, or 1600 nM, 800 nM, and 400 nM, depending on the strength of the response to the OX40 antigen binding the mAb. Kinetic analysis was performed us ing a 1:1 binding model and global fitting ( Supplementary Figure S6 ). To determine the a bility of each mAb to activate OX40 in vitro , we used a kit (Promega, assay in the presence of cells expressing Fc RIIB, which simulates the putative in vivo mecha nism of OX40 cross - linking 109 . On the day prior t o the assay, Fc RIIB/CHO - K1 cells were thawed into 95% RPMI 1640/5% FBS and plated into 96 - well plates. After incubating for 5 7 hours at 37 °C, 5% CO 2 , OX40 - expressing Jurkat cells were thawed and added to t he wells containing Fc RIIB/CHO - K1 cells. After incubating the cell mixtures overnight, antibodies were diluted in 95% RPMI 1640/5% FBS. The antibody dilutions were then added to the wells containing the cells. The c ell/antibody mixtures were incubated at 37 °C, 5% CO 2 for 5 h, after which we added Bio - Glo Reagent. Luminescence was read using a Spectramax i3x plate reader (Molecular Devices, San Jose, CA, USA). IC50 was calculated by plotting RLU (relative luminescenc e units) vs concentration using SoftMax Pro (Molecular Devices , San Jose, CA, USA ) ( Supp lementary Figure S7 ) . In - house produced pogalizumab was used as a positive control, and an antibody binding to an irrelevant antigen was used as a negative control. B. 4 . Results B. 4 .1. Overview of the Experimental Approach F irst, we stably expressed full - l ength human OX40 protein in mouse 3T3 cells. Next, we immunized transgenic humanized Trianni mice with either OX40 - expressing 3T3 cells or soluble OX40 extracellular do main using a rapid immunization p rotocol. The cohort of mice immunized with OX40 - express ing 3T3 cells was additionally boosted with a DNA vector driving expression 109 of full - length OX40 protein. All mice were checked for anti - OX40 serum titer ( Supplementary Figures S1, S3 ) and sacrificed af ter approximately four weeks. Spleen and lymph nodes we re disaggregated into single cell solutions, tissues from replicate animals were pooled, and B cells were isolated from the single cell solutions and cryopreserved. We then used droplet microfluidics 28 30 to isolate m illions of single cells from each experimental arm ( Supplementary Table S1 ) into aqueous - in - oil picoliter droplets. Cells were lysed inside the droplets, and mRNA from the single cells was bound to oligo(dT) beads. The oligo(dT) beads were then injected in to a second emulsion with multipl ex primers that amplify heavy and light chain Ig. The p rimers are designed with overlapping linker sequences that physically link heavy and light chain Ig into scFv expression constructs. The linked Ig libraries are subject ed to deep sequencing to quantify clonal antibody diversity ( Supplementary Table S1 ). Ea ch library was then electroporated into yeast for scFv display ( Figure B 1 ). Figure B 1 . Overview of the generation and screening of scFv libr aries derivates from B cells from humanized mice with either soluble OX40 or cells and DNA expressing OX40 . B cells are isolated from spleen and lymph nodes. Next, B cells are encapsulated into droplets with oligo - dT beads and a lysis solution to generate DNA amplicon that encodes the scFv libraries with native pairing heavy and light Ig. The scFv libraries are then transfected into yeast cells and labeled with either soluble OX40 or lysate from cells expressing OX40. Next, FACS is used to collect scFv with the highest FACS signal. Fin ally, deep sequencing is used to identify all clones in the pre - and post - sort populations. 110 N ext, each yeast scFv library was subjected to FACS using either lysate from FLAG - tagged OX40 - expressing CHO cells or soluble His - tagge d OX40 extracellular domain ( Figure B2 ). Cell lysate is prepared by lysing recombinant c ells in a buffer containing a surfactant (1% Triton X - 100) and quantified using a Bradford assay. The cell lysate was then incubated with each yeast scFv library, stain ed with an anti - FLAG secondar y antibody and anti - c - Myc staining to quantify expression o f scFv, and then subjected to FACS to pan for antigen - positive, c - Myc - positive binders. FACS with soluble OX40 was performed as described previously 28 30 . We performed either three or four rounds of FACS panning and de ep sequenced the binders. Fin ally, to compare the functional characteristics of antibodi es identified with each experimental method, we synthesized forty - one monoclonal antibodies from the panning experiments. We chose a sample of antibodies from distinct putative clonal lineages, wit h an emphasis on the most common clones from each experimen tal method ( Supplementary Tables S2 S9 ). We then used the full - length mAbs to perform kinetics measurements, cell surface binding assays, and in vitro cellular activati on assays. B. 4 .2. Analysis of Serum Titers S oluble OX40 antigen yielded consistently hi gh anti - OX40 serum titers in five replicate mice ( Supplementary Figure S1 ). We pooled the splenocytes or lymph nodes from these five animals to produce one yeast scFv l ibrary for each tis sue type, for a total of two soluble immunization OX40 libraries. Imm unizations using OX40 cells/DNA were less consistent across three replicate mice, generating a non - responder, a medium responder, and a high responder ( Supplementary Fi gure S3 ). Splenocyt es or lymph nodes from two of three animals were pooled to produce a single natively paired yeast scFv library for each tissue type (the medium responder 111 animal died prior to tissue harvest). Thus, we generated a total of four natively p aired yeast scFv l ibraries (2 tissues × 2 immunization methods = 4 libraries; Supplemen tary Table S1 ). Figure B 2 . scFv libraries from immunized mice subjected to FACS selection for OX40 . An anti - c - Myc (AF488) staining is used t o identify yeast di splaying scFv on the cell surface (x - axis). APC - streptavidin is used to identify yeast cells with biotinylated soluble OX40 bound, and anti - FLAG (APC) staining is used to identify yeast cells bound to lysate from cells expressing OX40 - FL AG (y - axis). A nega tive control is used to set a quadrangle gate for the FACS selection (upper right corner). The percentage in each quadrangle represents the proportion of c - Myc positive yeast cells that fell within the gate. ( a. ) An example of the solubl e immunized, cell l ysate sorted, spleen library subjected to three rounds of sorting to enrich the cells for positive antigen binding and scFv display on yeast surface. ( b. ) FACS plots showing the percentage of enriched antibodies after the three rounds of sorting under each condition. Note that the image used for soluble immunized, cell lysa te sorted, spleen is the same image as the 3rd sort from A. B . 4 .3. Selection of OX40 scFv Binders with FACS P rior publications have described protocols for panning yeas t surface scFv disp lay libraries with biotinylated, detergent - solubilized cell lysate 111,112 . We reasoned that biotinylation was 112 suboptimal because the method lab els all proteins in the cell lysate rather than only the target protein, leading to a lo ss of specificity in sorting and additional labor - intensive steps for every panning experiment. Therefore, we developed an approach based on an OX40 protein fused to a FLAG pep tide tag. Briefly, we engineered an expression vector that expresses full - length human OX40 fused to a FLAG peptide at the N - terminus ( Supplementary Figure S2 ). This vector was used to stably transfect CHO cells via targeted genome integration. Fre sh cell lysate was prepared for each panning experiment by lysing recombinant cells in a buffer containing a surfactant (1% Triton X - 100). In a typical experiment, we obtained around 7 - 8 mg/mL of total protein concentration per 5.0 × 10 7 cells/mL. Prior w ork has shown that the sensitivity and specificity of scFv binder discovery are function s of the molarity of a soluble target used during panning 28 30 . We, therefore, tested panning with four different concentrations of cell lysate ( Supplementary Figure S8 ). After two rounds of panning, the fraction of sc Fv binde rs was as low as 6.4% (0.475 mg/mL) and as high as 31% (3.8 mg/mL). The FACS plo ts at 1.9 mg/mL of cell lysate were qualitatively and quantitatively similar to our prior panning experiments using various soluble antigens at 7 70 nM 28 30 . Therefore, we used approximately 2 mg/mL of cell lysate for all subs equent cell lysate panning experiments. Although all tissues, from both immuniza tion methods, and with both FACS methods yielded scFv binders, there were qualitative and quantitative differences in the FACS plots ( Figure B2 ). After three rounds of panning, the fraction of scFv binders was as low as 16.0% (cells/DNA immunization, solub le antigen FACS) and as high as 78.2% (cells/DNA immunization, cell lysate FACS). On average, the soluble immunogen yielded a higher fraction of scFv binders than the c ells/DNA immunization (61.6% versus 54.4%, respectively), and the cell lysate FACS yield ed a 113 higher fraction of scFv binders than the soluble antigen FACS (70.6% versus 45.5%, respectively). Because the cells/DNA immunogen followed by soluble FACS yielded a lower fraction of scFv binders, we performed a fourth round of panning on these librar ies ( Supplementary Figure S3 ), which improved the fraction of scFv binders by as much as 58.3% (from 16.0% to 74.3%), suggesting an increase in specificity. In general, cell ly sate FACS produced a more significant shift in FLAG - APC fluorescence (antigen bi nding) than soluble antigen FACS. B. 4 .4. Sequence Characteristics of OX40 scFv Binders W e deep sequenced the yeast scFv libraries before and after FACS ( Supplementary Table S1 ), as described previously 28 30 . Note that we use extremely conservative error processing, which favors clone sequence quality ove r capturing the "long tail" of clonal diversity. Before FACS, the scFv libraries contained between 16,491 and 19,509 clones. After FACS, the scFv libraries were much mo re oligo clonal, containing between 61 and 238 clones. We analyzed the most common ( 0.1% frequency) scFv sequences to determine pre - versus post - FACS clonal enrichments achieved by each method. We did not observe statistically significant differences betwe en the m ean clone counts of soluble versus cells/DNA immunizations, or between soluble v ersus cell lysate sorts ( p > 0.01, t - test). The average pre - sort scFv clone abundance was 0.032%, with a range from 0% (not detected) to 0.71%. Sequences present in the post - so rt libraries were not detected in the pre - sort libraries for 54/268 (20.1%) of c lones, suggesting that many candidate binder clones were extremely rare in the mouse repertoires. The average enrichment between pre - and post - sort clone counts was 505 6 - fold, with a range from 2.2 - fold to 500,000 - fold. We note that prior work on Balb/c, S JL, and Medarex HuMAb mice 28,30 yielded similar levels of enrichment and clonal diversity both before and after FACS. 114 Figure B 3 . Ove rlapping clones in the pre - and post - sort populations obtained from each experimental parameter. Clone frequencies are represented by the blue heatmap, with unique clones aligned into rows across the repertoires. We only show clones that are present with frequencies of 0.1% or higher in at least one of the post - FACS repertoires. We organize the repertoires by pre - vs. post - FACS, tissue of origin (lymph nodes vs. spleen), immunizati on method (soluble antigen vs. cells/DNA), and panning condition (soluble antigen vs. cell lysate). Done by GigaGen. N ext, to det ermine w hether different methods discover the same scFv binders, we analyzed the most co mmon ( 0.1% frequency) scFv sequences for overlap between each post - FACS library ( Supp lementary Tables S2 S9 ). Only 6.8% of enriched clones (16/235 non - redundant, uniqu e clones ) were shared between at least two of the eight series of scFv panning series (2 tissues × 2 FACS methods × 2 immunization methods) ( Figure B3 ), suggesting different immunization methods and different FACS methods typically capture different sequen ces. Not ably, 81.3% (13/16) of the shared clones were generated using the soluble immuno gen and identified with both soluble and cell lysate FACS methods. The only three scFv clones that were shared between lymph node and spleen were generated using the ce lls/DNA immunization method, suggesting that cells/DNA induces a more systemic antigen r esponse than soluble immunization. 115 A clonal cluster plot of full - length IgH and IgK se quences from scFv binder clones highlights similarities and differences among scFv binde r sequences ( Figure B4 ). An "edge" (a ces in the concatenated CDR3s. However, most clonal clusters comprised only a single clone, i.e., no related sequences were detected. Only 14 clonal clusters were c omprised of five or more scFv clones (putative clonal lineages). Of those, 100% (14/14) comp rised clones derived from only the soluble immunogen (using either FACS method) or only cells/DNA immunogen (using either FACS method). In general, sequence analysi s suggests that each immunization method, FACS method, and tissue produces mostly unique clo nes. However, where there is overlap, FACS with different methods is more likely to generate similar clones than different tissues, and different immunization metho ds are least likely to generate similar clones. We did not find any significant differences in the sequence characteristics of the most common ( 0.1% frequency) scFv clones between each post - FACS library. Sequence identity to germline (%ID) was high acros s all methods, averaging 98.4% for IgKV and 97.6% for IgHV. This suggests low levels of affin ity maturation in vivo. Variable (V) - gene diversity was low across all methods ( Supplementary Figures S9 S10 ), for example, 172/268 (64.2%) of clones were some all e le of IgHV1, and 72.4% of clones were some allele of IgKV1. In general, scFv clone binders p reviously identified in Balb/c, SJL, and Medarex HuMAb mice 28,30 showed similarly high levels of germline %ID and similarly low levels of V - gene diversity. Though each immunization and panning method yielded distinct clo nes ( Figure B4 ), there was a strong bias toward a limited variety of V - Joining (J) combinations, with between - method Pearson correlation coefficients ranging from 0 .59 to 0.81 ( Supplementary Figure S11 ), all of which were significant (p < 0.001). We postul ate that such similarities arise due to the limited V - J diversity in the pre - sort repertoires 116 ( Supplementary Figures S9 S10 ), though we cannot rule out the possibi l ity that OX40 binding is more likely given particular V - J pairs. Figure B 4 . Clonal cluster plot of anti - OX40 clones with frequencies higher than 0.1% in the post - sorted populations . scFv isolated from lymph nodes and spleen a r e indicated with circles and squares, respectively. scFv isolated from solub le immunization and sorted with either soluble OX40 or cell lysate expressing OX40 are colored in green and orange, respectively. scFv isolated from cell and DNA immunization and s o rted with either soluble OX40 or cell lysate expressing OX40 are colored in in the plots. The size of the nodes corresponds to the frequency of the antibody clone in the FACS - so r t ed population: small (<2% frequency), medium (2 12% frequency), and large ( >12% frequency). acid differences in the concatenated scFvs. Done by GigaGen B . 4 .5. Functional Chara c t eristics of Monoclonal Antibody Binders N ext, to investigate the therapeuti c potential of scFv binders, we synthesized forty - one putative binder scFv as full - length mAbs ( Table B1 ), using methods described elsewhere 30 . We 117 chose at least the top two (range: 2 7 scFv) most common scFv enriched in each tissue, fro m each immunization and FACS method combination, along with several other str ong and weak enriched scFv. First, we used a FACS assay to assess the ability of each mAb to bind OX40 recombinantly expressed on cell surfaces. In total, 39% (16/41) of the mAbs b o und cell surface antigen, for a 61% false positive rate ( Supplementary Figu re S5; Supplementary Table S10 ). Notably, immunizing with soluble OX40 antigen followed by sorting with lysate from OX40 - expressing cells yielded a 0% false positive rate, i.e., a l l mAbs identified with this method bound cell surface OX40. The other method s yielded significantly higher false positive rates ( z - test for proportions, p < 0.01), ranging from 35% to 67%. We also observed two distinct positive binder peaks in the histogr a m s for 68.8% of cell surface binding scFv (11/16), for unknown reasons. Table B 1 . Functional characteristics of 41 scFv binders converted into full - length IgG1 mAbs. mAb ID Enriched? Soluble immunized, soluble sorted Enriched? S oluble immunized, lysate sorted Enriched? Cells/DNA immunized, soluble sorted Enriched? Cells/DNA immunized, lysate sor ted Binds cells? Promega in vitro assay EC50 (ug/mL) (agonist) KD (nM) [Octet, global fit] tOX40.2 Yes No No No No not tested not te st ed tOX40.4 Yes Yes No No Yes (2 peaks) 0.044 9 tOX40.15 Yes No No No Yes (1 peak) 1.276 6 tOX40.19 Yes Yes No No Yes (1 peak) 0.811 7.7 tOX40.20 Yes No No No No not tested not tested tOX40.21 Yes Yes No No Yes (1 peak) does not agonize 7 tOX40.22 Y es Yes No No Yes (2 peaks) 0.087 7.3 tOX40.23 Yes Yes No No Yes (2 peaks) 0.419 5.2 tOX40.24 Yes Yes No No Yes (2 peaks) 0.024 22.9 tOX40.28 Yes No No No No not tested not tested tOX40.31 Yes Yes No No Yes (2 peaks) 0.043 22.4 tOX40.33 Yes No No No No n ot tested not tested 118 Table B1 (cont d) tOX40.34 Yes No No No No not tested not tested tOX40.35 Yes No No No Yes (1 peak) 4.316 2.7 tOX4 0.36 Yes No No No No not tested not tested tOX40.37 Yes Yes No No Yes (2 peaks) 0.195 12.5 tOX40.38 Yes No No No No not tested not tested tOX40.39 Yes Yes No No Yes (2 peaks) 0.199 90 tOX40.40 Yes Yes No No Yes (2 peaks) 0.426 151 tOX40.41 No Yes No N o Yes (2 peaks) 0.068 58.4 tOX40.42 No No Yes No Yes (1 peak) does not agonize no binding tOX40.43 No No Yes No N o not tested not tested tOX40.44 No No Yes No No not tested not tested tOX40.45 No No Yes No No not tested not tested tOX40.46 No No Yes N o No not tested not tested tOX40.47 No No Yes No No not tested not tested tOX40.48 No No Yes No No not tested not tested tOX40.49 No No Yes No No not tested not tested tOX40.50 No No Yes Yes Yes (2 peaks) 0.091 29.8 tOX40.51 No No No Yes No not tested not tested tOX40.52 No No No Yes No not tested not tested tOX40.54 No No No Yes Yes (2 peaks) 0.321 184 tOX40.5 5 No No No Yes No not tested not tested tOX40.56 No No Yes No No not tested not tested tOX40.57 No No Yes No No not tested not tested tOX4 0.58 No No Yes No No not tested not tested tOX40.59 No No Yes No No not tested not tested tOX40.60 No No Yes No N o not tested not tested tOX40.61 No No No Yes No not tested not tested tOX40.62 No No No Yes No not tested not tested tOX40.63 No No No Ye s No not tested not tested We then tested the sixteen cell - surface binding mAbs for in vitro activation in a cellu lar assay. The avera ge EC50 was 0.59 g/ L , with a range from 0.024 to 4.3 g/ L . Cell surface binding was a good predictor of in vitro agonism, with 87.5% (14/16) of cell surface binders demonstrating agonism ( Table B1; Supplementary Figure S7; Supplement ary Table S10 ), for 119 a 77% (27/41) false positive rate overall. Again, immunizing with soluble OX40 antigen followed by sorting with OX40 - embedded cell lysate yielded the lowest false positive rate, at 9.1% (1/11). The other methods yie lded significantly hi gher false positive rates ( z - test for proportions, p < 0.01), ranging from 42.1% to 92.9%. The number of peaks in the cell surface binding assay was associated with the strength of agonism: the 1 - peak mAbs have an average EC50 of 2.1 ( with 2 mAbs not showi ng any agonist activ ity), whereas the 2 - peak mAbs have an average EC50 of 0.17 (with all mAbs having agonist activity). Note that the positive control benchmark (pogalizumab) is a 2 - peak binder with a strong EC50 (0.039 g/ L ). We spec ulate that mAbs in th e 2 - peak group compr ise a different epitope bin than the mAbs in the 1 - peak group. The two mAbs that failed to agonize (tOX40.21 and tOX40.42) also showed the weakest fluorescence shift in the flow cytometry cell surface binding experi ments ( Supplementary Figure S5 ). We did n ot observe significant differences among the protocols in in vitro agonism EC50 ( p > 0.01, t - test). We also tested the cell surface binding mAbs for affinity using Octet. Of the mAbs that bound cell surface antigen, 93.8% (15/16) also b ound soluble antigen ( Table B1; Supplementary Figure S6; Supplementary Table S10 ). The average K D was 41.1 nM, with a range from 2.7 to 184 nM. The pogalizumab positive control yielded a K D of 1.9 nM. One of the antibodies that bound c ell surface antigen ( tOX40.42) failed to bind antigen by Octet and also failed in vitro agonism. This mAb was discovered using the cells/DNA immunization and soluble antigen sorting method. Another mAb (tOX40.21) did not agonize OX40 in vitro but did bind soluble antigen (K D = 7nM). This antibody was among the weakest binders in the cell surface flow cytometry assay and was discovered using the soluble antigen immunization with both the soluble sort and cell lysate sort methods. We speculate that this mAb b inds non - specifically , resulting in high affinity 120 but weak agonism and cell surface binding. We did not observe significant differences among the protocols in K D ( p > 0.01, t - test). B. 5 . Discussion In this study, we adapted previously published methods 28 30 to test whether different immunizatio n methods (cells/DNA versus solu ble antigen) and dif ferent selection methods (cell lysate versus soluble antigen) yielded mAbs with higher potential as therapeutic OX40 agonists. Though all methods successfully identified anti - OX40 mAbs, using cell lysate for selection generally yielded mAbs that were more likely to bind to cell surface antigen and activate OX40 in cellular assays. We speculate that cell lysate contained OX40 trimers, whereas soluble antigen comprised OX40 monomers, perhaps leading to the i dentification of more physiologi cally relevant binde rs. Using massively parallel microfluidics and deep sequencing allowed us to rigorously characterize mouse responses to different types of immunogens. The large scFv repertoires generated from the animals also facilitated robust testing of FACS methods. Ot her methods, such as hybridomas, would have required significantly more effort to generate such a comprehensive data set. In this study, immunization with cells/DNA was inconsistent and yielded a low prop ortion of agonist mAbs. In futur e experiments, we co uld establish a titer cutoff and only make yeast libraries from animals with titers exceeding that cutoff. Still, there are many ways that cells/DNA immunization could be optimized in the future. For exam ple, we could test different con centrations of cells in the mouse immunizations, different adjuvants, different DNA vectors, or alternate cell lines. We could also ensure high levels of cell surface antigen expression by using FACS to isolate populations o f cells with the highest antigen expression, as desc ribed elsewhere 117 . A more aggressive cells/DNA immun ization schedule cou ld increase titer and reproducibility, for example, 121 through daily injections of cells for the first few days of the immunization protocol 117 , biweekly immunizations with cells for ten weeks 118 , or biweekly immuniz ations with DNA for eight weeks 119 . Though the cell lysate sorting method yielded the highest proportion of agonist mAbs in this study, there are many opportunities for furth er improvement. We tested severa l different cell lys ate concentrations, but a more rigorous optimization would require a more thorough analysis of the impact of lysate concentration on FACS sensitivity and specificity. We only tested the cell lysate sortin g methods with OX40, whereas oth er targets may yield different results. For example, certain targets may unfold in our lysis buffer, yielding antibodies less likely to bind to a properly folded protein target. Additionally, we might find that peptide tags other than FLAG (for example, Hi s tag) might yield b etter results, or that C - terminal tags are preferential for certain targets. Finally, further work might compare screening yeast scFv libraries with cell lysate versus screening phage scFv display librari es against cells affixed to plat es. To our knowledge , no other group has published in - depth studies of the antibody repertoire response of Trianni humanized mice to immunogens. Our work yielded a low diversity of light chain V - genes, for example, >70% of s cFv binder clones were IgKV1. Th is level of light ch ain Ig diversity after immunization and FACS selection is similar to results obtained in wild type Balb/c and SJL mice 2 8 and humanized Med arex HuMAb mice 30 . Additionally, the V sequences of scFv b inders from both libraries were ~98% identical to germline V sequences, suggesting little if any affinity maturation in vivo. Prior work on repertoires of mice administered various immunogens found only 2 - 5 amino acid substitutions per V - gene 28,30,120 122 . Futur e work should invest igate whether Trianni mice generate similar responses with other immunogens. 122 Our methods open up exciting directions for mAb discovery and development. For example, we could use cell lysates to select for mAbs that bind to a specific e pitope, or do not bi nd to a specific epitope. In one scenario, cells could be engineered that express OX40 protein wi th mutations in the amino acids required for binding OX40L. Then, scFv libraries could be sorted using the mutated OX40, perhaps identifyin g antibodies that bi nd outside the OX40:OX40L binding domain. Another intriguing approach would be to immunize mice w ith tumor cells, and then pan for scFv that bind to lysates from tumors but not to lysates from normal tissue. This approach could be used to find mAbs directe d against novel tumor - specific targets. B. 6 . Patents The OX40 full - length antibody sequences and the OX40 - enriched yeast scFv libraries described in this article are patent - pending subject matter in USPTO provisional patent application number 62/788687, pr iority date 4 January 201 9 . 123 APPENDIX C : Supplementary Notes 124 Note C 1 . Sequence for IL - 31 construct C1.1. Feline IL - 31 DNA Sequence A TGTCTCACATGGCTCCAGCACATAGATTACAACCA TCAGATATT A GAAAGATCATC TTG GAATT AAGACCAATGTCTAAAGGTTTGTTGCAAGATTATTTGAAGAAAGAAATC GGTTTACCAGAATCAAACCATTCTTCATTGCCATGTTTATCTTCAGATTCTCAATTGC CACATATCAACGGTTCAGCAATCTTGCCATACTTTAGAGCTATTAGACCATTGTCAG ATAAGAACACAATCGATAAGATCATCG AACAACTAGACAAGTTGAAGTTTCAAAGA GAACCAG AAGCAAAAG T TTCTATGCCAGCTG ATAAC TTCGAAAGAAAGAATTTCAT CTTGGCAGTTTTACAACAATTTTCAGCTTGTTTGGAACATGTTTTGCAATCTTTGAAT TCAGGTCCACAA 125 Note C 2 . Sequences for Pro - NGF constructs C2.1. cNGF DNA Sequence A TGTCTTCATCTCATCCAGTTTTTC ATAGAGGTG A ATTTTCTGTTTG TGATTCA GTTTC TGTTTGGGTTGGTGACAAGACTACAGCTACAGATATCAAGGGTAAAGAAGTTATGG TTTTGGGTGAAGTTAACATCAACAACTCAGTTTTCAAGCAGTATTTCTTTGAAACAA AATGTAGAGATCCAACTCCAGTTGATTCTGGTTGTAGAGGTATCGATTCAAAGCATT GGAACTCTTACTGTACTACAACTCATA CATTCGTTAAGGCATTGACTATGGA TGGTA AACA A GCTGCATGGAGA TTCATTA GAATTGATACTGCTTGTGTTTGTGTTTTATCTAG AAAAGCAGGTAGAAGAGCT C2.2. Pro - cNGF DNA Sequence G AACCACATCCAGAATCTCATGTTCCAGCAGGTCATGCTATTCCACATGCTCATTGG ACAAAGTTGCAACATTCATTGGATACTGCATTGAGAAGAGCTAGATCTGCTCCAGCA GGTGC TATTGCTGCAAGAGTTACAGGTCAA ACTAGAAAC A TCACAGTTGATC CAAA GTT GTTTAAGAAAAGAAGATTGAGATCACCAAGAGTTTTATTTTCTACTCATCCACC ACCAGTTGCTGCAGATGCACAAGATTTGGATTTGGAAGCAGGTTCAACAGCTTCTGT TAACAGAACTCATAGATCAAAGAGATCTTCATCTCATCCAGTTTTTCATAGAGGTGA ATTTTCTGTTTGTGATTCAGTTTCTGTTTGGGTT GGTGACAAGACTACAGCTACAGAT A TCAAGGGTA A AGAAGTTATGGT TTTGGGT GAAGTTAACATCAACAACTCAGTTTTC AAACAATACTTTTTTGAAACAAAATGTAGAGATCCAACTCCAGTTGATTCTGGTTGT AGAGGTATCGATTCAAAGCATTGGAACTCTTACTGTACTACAACTCATACATTCGTT AAGGCATTGACTATGGATGGTAAACAAGCTGCATGGAGATTCATTAGAATTGATACT GCTT GTGTTTGTGTTTTATCTAGAAAAGC AGGTAGAAG A GCT C2.3. P - cNGF DNA Sequence C AAACTAGAAACATCACAGTTGATCCAAAGTTGTTTAAGAAAAGAAGATTGAGATC ACCAAGAGTTTTATTTTCTACTCATCCACCACCAGTTGCTGCAGATGCACAAGATTT GGATTTGGAAGCAGGTTCAACAGCTTCTGTTAACAGAACTCATAGATCAAAGAGAT CTTCATCTCATCCAG TTTTTCATAGAGGTGAATTTTCTGT TTGTGATTCA GTTTCTGTT TGG GTTGGTGACAAGACTACAGCTACAGATATCAAGGGTAAAGAAGTTATGGTTTT GGGTGAAGTTAACATCAACAACTCAGTTTTCAAACAATACTTTTTTGAAACAAAATG TAGAGATCCAACTCCAGTTGATTCTGGTTGTAGAGGTATCGATTCAAAGCATTGGAA CTCTTACTGTACTACAACTCATACATTCGTTAAGGCATTGAC TATGGATGGTAAACA AGCTGCATGG AGATTCATTA GAATTGATACTG CTTGTGTTTGTGTTTTATCTAGAAAA GCAGGTAGAAGAGCT 126 Note C 3 . Sequence for AmiE, UCA9, and PYR1 constructs C3.1. AmiE DNA Sequence ATGAGACATGGCGATATTAGCTCGTCAAATGATACCGTAGGCGTAGCCGTGGTGAA TTACAAGATGCCGCGTTTACATACTGCTGC TGAAGTCCTGG ATAATGCCCGCAAAAT TGCGGAAATGATCGTTGGTATGAAGCAAGGTCTGCCGGGCATGGATCTGGTTGTGTT TCCTGAATATTCTTTACAGGGTATTATGTACGACCCTGCTGAAATGATGGAAACAGC CGTGGCGATTCCAGGCGAAGAAACGGAAATCTTTAGCCGTGCTTGTAGAAAAGCAA ATGTTTGGGGTGTGTTCTCCCTGACCGGCGAACGTCATGAAGAACACCCTAGAAAGG CA CCATACAACAC TCTGGTCTTGATCGATAACAACGGTGAAATCGTACAAAAGTAC AGAAAGATCATCCCATGGTGTCCGATTGAAGGCTGGTATCCAGGTGGCCAGACATA CGTCTCTGAAGGTCCGAAAGGCATGAAGATCTCATTAATTATCTGCGATGACGGTAA TTATCCGGAAATTTGGAGAGATTGTGCCATGAAGGGTGCGGAATTGATCGTTCGCTG CCAAGGCTATATGTACCCTGCTAAAGACCAAC AAGTTATGATG GCTAAGGCAATGG CCTGGGCGAATAACTGTTATGTCGCTGTAGCAAACGCTGCAGGTTTTGATGGCGTTT ATAGCTACTTCGGTCATAGTGCCATTATCGGTTTTGACGGCCGTACTCTGGGTGAAT GCGGCGAAGAAGAAATGGGCATTCAATACGCGCAGTTGTCTCTGTCACAAATCCGC GATGCCCGTGCGAATGACCAAAGTCAGAACCATTTGTTTAAAATCTTGCACAGAGGT TACTC CGGTTTGCAGG CTTCGGGCGATGGCGACCGTGGTCTGGCAGAATGCCCATTT GAATTCTACCGTACCTGGGTTACTGATGCTGAAAAGGCAAGAGAAAACGTGGAACG CCTGACTCGCTCCACAACAGGTGTCGCCCAATGCCCAGTAGGTCGTCTGCCGTATGA AGGCCTCGAG C3.2. UCA9 DNA Sequence ATGCAGGTGCAGCTGGTGCAGTCTGGGGCTGAGGTGAAGAAGCCTGGGTCCTCGGT GAAGGTCTCC TGCAAGGCTTCTGGAGGCACCTTCAGCAGCTATGCTATCAGCTGGGT GCGACAGGCCCCTGGACAAGGGCTTGAGTGGATGGGAGGGATCATCCCTATCTTTG GTACAGCAAACTACGCACAGAAGTTCCAGGGCAGAGTCACGATTACCGCGGACGAA TCCACGAGCACAGCCTACATGGAGCTGAGCA GCCTGAGATCTGAGGACACGGCCGT GTATTACTGTGCGAAAGGTAGTGGTTATCATGTCCGCGATT ACTTTGACTACTGGGG CCAAGGAACCCTGGTCACCGTCTCCTCA C3.3. PYR1 DNA Sequence A TGCCTTCGGAGTTAACACCAGAAGAACGATCGGAACTAAAAAACTCAATCGCCGA GTTCCACACATACCAACTCGATCCAGGAAGCTG TTCATCACTCCACGCGCAACGAAT CCACGCGCCTCCGGAACTCGTCTGGTCAATCGTACGACGATTCGACAAACCACAAA CATACAA ACACTTCATC AAATCCTGCTCC GTCGAACAAAACTTCGAGATGCGCGTCG GATGCACGCGCGACGTGATCGTCATCAGTGGATTACCGGCGAACACATCAACGGAA AGACTCGATATACTCGACGACGAACGGAGAGTTACCGGATTCAGTATCATCGGAGG CGAACAT AGGCTGACGAATTACAAATCCGTTACGACGGTGCATCGGTTCGAGAAAG AGAATCGGATCTGGACGGTGGTTTTGGAATCTTACGTC GTTGATATGC CGGAAGGTA ACT CGGAGGATGATACTCGTATGTTTGCTGATACGGTTGTGAAGCTTAATTTGCAGA 127 AACTCGCGACGGTTGCTGAAGCTATGGCTCGTAACTCCGGTGACGGAAGTGGTTCTC AGGTGACG 128 N ote C 4 . A protocol to determine the concentration of pooled mutagenic ol i gos for gene X used in the Nicking Saturation Mutagenesis (NSM) experiments Overview: These steps will allow the user to achieve a 20:1 molar ratio of dsDNA template: pooled oligo co mpatible with NSM experiments. We assume that the end user has a lyophili z ed oligo pool that has been resuspe nded to a total oligonucleotide concentration of 200 nM , with a certain number of oligos in the pool that are specific to gene X . The steps are listed below: a. Calculate the molar fraction of mutagenic oligos for gene X in the oligo pool b. Calculate the effective concentration of mutagenic oligos for gene X c. Calc ulate the concentration of the phosphorylated oligos using 20 L of the lyophilized oligo pool (step 1 from NSM , https://www.nature.com/protocolexchange/protocols/5125 ) d. Calculate the moles of mutagenic oligos needed to achieve a 20:1 molar ratio o f dsDNA template: pooled oligo e. Calculate the dilution of phosphorylated oligos needed after NSM step 1 Example: Assuming i) a total of 0.2µM pooled oligos and ii) 1,449 out of 7,188 pooled oligos are specific for gene X . a. Molar fraction of protein X speci fic oligos: b. Effective concentration of mutagenic oligos: c. Concentration of phosphorylated oligos (see step 1 from NSM): 129 d. Moles of oligos to achie ve a 20:1 molar ratio: Based on NSM experiments, we need 0.76 pmoles of dsDNA plasmid to prepare the ssDNA template strand (step 2). For a total reaction volu me of 20 µL, e. D ilution of phosphorylated oligos: So here, we would add of the phosphorylated oligos to the 20 nicking mutagenesis protocol. 130 APPENDIX D : S upplementary Figures 131 Figure D 1 . Mass s pec trometry and a nalytical SEC for f eline IL - 31 construct . ( a. ) Mass spec analysis of IL - 31 after PNGase treatment. Conditions: 50mM Tris, pH 8.0. ( b. ) Chromatograph ic conditions: TSK SuperSW3000, 4.6 x 30 mm, mobile phase 200 mM sodium phosphate, pH 7.2, flow rate 0.25 mL/min for 25 minutes, injection amount: 50 g . Done by Zoetis . Feline IL - 31 4°C, ~ 2 weeks Injected: 50 µg a. b. 132 Figure D 2 . Structural model of fIL - 31 . ( a . - b . ) Two different views sh owing packing of hydrophobic core. Aliphatic residues Phe , Tyr, Leu, Ile, Val, and Met are colored in green sticks, and the remain i ng residues are shown as white sticks. The Cys49 - Cys132 disulfide bond is shown as spheres. ( c . ) A closer view of the hydroph obic core of the structural model. a. b. c. 133 Figure D 3 . Determination of the binding si t es for fIL - 31/fI L31RA - 1FNIII interaction. Fitness Metric heatmap of the top 7% bound populati on vs the reference population. The Shannon entropy is plotted below with its respective cut - off midpoint (dashed line). 134 Figure D 4 . Deter mination of the binding sites for fIL - 31 - fOSMR - ECD interaction. Fitness Metri c heatmap of the top 7% bound po pulation vs the reference population. The Shannon entrop y is plotted below with its respective cut - off midpoint (dashed line). 135 Figure D 5 . Determination of the conformational epitope for fIL - 31/mAb#1 interac tion. Fitness Metric heatmap of the top 7% bound population vs the reference population. The Shannon entropy is plotted below with its respective c ut - off midpoint (dashed line). 136 Figure D 6 . Effects of anti - NGF mAbs on canine - NGF Induced Proliferation of TF - 1 Cells (representative curves ). Done by Zoetis. (error bars, standard deviation, n=2) 137 Figu re D 7 . Per - position heatmap of enrichment scores for pro - cNGF mutants after 1 sor t with tanezumab. 138 Figure D 8 . Per - position heatmap of enrichment scores for pro - cNGF m u tants after 1 so rt with mAb #1. 139 Figure D 9 . Per - position he atmap of enrichment scores for p - cNGF mutants after 1 sort with tanezumab. Figure D 10 . Per - - cNGF mutants after 1 sort with mAb #1. 140 Figure D 11 . Per - posi tion heatmap of enrichment scores for pro - cNGF mutants after 2 sorts with mAb #1. 141 Figure D 12 . Mean fluorescence intensities for individual point mutants compared with wild - type pro - cNGF . 142 Figure D 13 . Determination of conformational epitope for cNGF_tanezumab . Fitness metric heatmap of the top 7% bound population vs the unselected population. Shannon entropy is plotted below with its respectively cut - o ff (dashe d line). 143 Figure D 14 . Determination of conformational epitope for cNGF_mAb #1 . Fitness metric heatmap of the top 7% bound population vs the unselected populati on. Shannon entropy is plotted below with its respect i vely cut - off (dashed line). 144 Figure D 15 . Determination of conformational epitope for cNGF_mAb #2. Fitness metric heatmap of the top 7% bound population vs the unselected populatio n. Shanno n entropy is plotted b e low with its respectively cut - off (dashed line). 145 Figure D 16 . Determination of conformational epitope for cNGF_mAb #3. Fitness metric heatmap of the top 7% bound population vs the unselect ed popula tion. Shannon entropy i s plotted below with its respectively cut - off (dashed line). 146 Figure D 17 . Correlation between counts in the displayed population relative to the counts in the unselected population for cNGF. 1 10 100 1000 10000 1 10 100 1000 10000 Normalized Counts Displayed Population Normalized Counts Unselected Population Missense Mutations Nonsense Mutations 147 Figure D 18 . Per - position heatmap of sequencing counts for AmiE mutants in ( a. ) replicate 1 and ( b. ) replicate 2 using the oligo pool. a. b. 148 Figure D 19 . Per - posi tion heat map of sequencing coun t 149 Figure D 20 . Per - position heatmap of s equencing counts of UCA9 mutants in ( a. ) replicate 1 and ( b. ) replicate 2 using the oli go pool. a. b. 150 Fig u re D 21 . Per - position heatmap of sequencing counts for PYR1 mutants ( a. ) replicate 1 and ( b. ) replicate 2 using the oligo pool. a. b. 151 APPENDIX E : Supplementary Tables 152 Table E 1 . Sorting conditions a nd FACS collection statistics for fIL - 31 libraries A ntigen Amount of Collected Cells Labeling Concentration [nM] Percent Sorted Library 1 Percent Sorted Library 2 f OSMR - ECD 300,000 5.1 6.5% 6.7% f IL31RA - 1FNI II 300 ,00 0 52.8 8.5% 6.0% mAb # 1 300,000 0.1 3 7.6% 7.8% Table E 2 . Primers for deep sequencing. L1: Library 1, L2: Library 2, Blue : Illumina Universal Sequence, NNNNNN: Indexing Barcode, Purple : Illumina Adapter, UN: Unselected Popul ati on, and CMYC: Displayed Population N ame Sequence Inner Primer: f IL - 31_ L 1_FWD 5' - GTTCAGAGTTCTACAGTCCGACGATC AGGGTCGGCTAGCCATATG - 3' f IL - 31_ L 1_REV 5' - CCTTGGCACCCGAGAATTCCA TCTGACAATGGTCTAATAGCTCT - 3' f IL - 31_ L 2_FWD 5' - GTTCAGAGTTCTACAGTCCGAC GATC TC AGC AATCTTGCCATACTTT - 3' f IL - 31_ L 2_REV 5' - CCTTGGCACCCGAGAATTCCA ATAAGCTTTTGTTCGGATCCG - 3' Outer Primers: Illumina_FWD 5' - AATGATACGGCGACCACCGAGATCTACAC GTTCAGAGTTCTACAGTCCGACGATC - 3' Top 5% Displayed Experiment: f IL - 31_L1_UN _REV 5' - CAAGCAGA AGACGG CAT ACGAGAT CGTGAT GTGACTGGAG TTC CTTGGCACCC GAGAATTCCA - 3' fIL - 31_L 1_ CMYC _REV 5' - CAAGCAGAAGACGGCATACGAGAT ACATCG GTGACTGGAGTTC CTTGGCACCCGAGAATTCCA - 3' f IL - 31_L2_UN _REV 5' - CAAGCAGAAGACGGCATACGAGAT GCCTAA GTGACTGGAGTTC CTTGGCACCCGAGAATTCCA - 3' fIL - 31_L 2_CMYC _RE V 5' - CAAGCAGAAGACGGCA TAC GAGAT TGGTC A GTGACTGGAGTTC CTTGGCACCCGAGAATTCCA - 3' Labeled with mAb#1 or f OSMR - ECD : f IL - 31_L1_UN _REV 5' - CAAGCAGAAGACGGCATACGAGAT GATCTG GTGACTGGAGTTC CTTGGCACCCGAGAATTCCA - 3' fIL - 31_L 1_mAb#1_REV 5' - CAAGCAGAAGACGGCATAC GAGAT C TGA TC GTGACTGGAGTTC CTTGGCAC CCGAGAATTCCA - 3' fIL - 31_L 1_OSMR_REV 5' - CAAGCAGAAGACGGCATACGAGAT GTAGCC GTGACTGGAGTTC CTTGGCACCCGAGAATTCCA - 3' fIL31_L 2_U N _REV 5' - CAAGCAGAAGACGGCATACGAGAT TACAAG GTGACTGGAGTTC CTTGGCACCCGAGAATTCCA - 3' 153 Table E2 (cont d) fIL - 31_L 2_mAb#1_REV 5' - CAA GCAGAAGACGGCATACG AGAT GG AACT GTGACTGGAGTTC CTTGGCACCCGAGAATTCCA - 3' fIL - 31_L 2_OSMR_REV 5' - CAAGCAGAAGACGGCATACGAGAT GGACGG GTGACTGGAGTTC CTTGGCACCCGAGAATTCCA - 3' Labeled with f IL31RA - 1FNIII: fIL31_L1_UN _REV 5' - CAAGCAGAAGACGGCATACGAGAT CGAAAC GTGAC TGGAGT TC C TTGGCACCCGAGAATTC CA - 3 ' fIL - 31_L 1_IL31RA - 1FNIII_REV 5' - CAAGCAGAAGACGGCATACGAGAT CCACTC GTGACTGGAGTTC CTTGGCACCCGAGAATTCCA - 3' fIL31_L2_UN _REV 5' - CAAGCAGAAGACGGCATACGAGAT GCTACC GTGACTGGAGTTC CTTGGCACCCGAGAATTCCA - 3' fIL - 31_L 2_IL31RA - 1FNIII_REV 5' - CAA GCAGAAGACGGCATACG AGAT GC TCAT GTGACTGGAGTTC CTTGGCACCCGAGAATTCCA - 3' Table E 3 . f IL - 31 library statistics results f OSMR - ECD and mAb#1 f IL31RA - 1FNIII Top 5% Displayed Library 1 Library 2 Library 1 Library 2 Library 1 Lib rary 2 P ercent of possibl e codo n substitutions observed in the unselected population: 1 - base substitution 100% 100.3% 94.4% 95.8% 97.4 % 97.4 % 2 - base substitutions 52.0% 56.4% 39.9% 42.0% 38.5 % 42.8 % 3 - base substitutions 38.3% 41.5% 37.1% 38.0% 37 .3 % 38 .8 % Percent of unse lected reads with: No nonsynonymous mutations: 41.8% 45.6% 39.1% 45.3% 44.5% 45.9 % One nonsynonymous mutation: 52.6% 50.8% 58.6% 52.3% 48.3% 49.3 % Multiple nonsynonymous mutations: 5.5% 3.5% 2.3% 2.3% 7.2 % 4.8 % Coverage of pos sib le single nonsyno nymous amino acid mutations: 77.8% 83.2% 75.7% 80.0% 77.9 % 81.0 % 154 Table E 4 . FACS collection statistics for Pro - cNGF - cNGF library screening experiments Library Size (aa's) Sort Round 1 Sort Round 2 tanezumab mAb #1 tan ezumab mAb #1 Pro - cNGF_tanezumab library 1 51 2.9% 2.4% 3.1% 2.7% Pro - cNGF_tanezumab library 2 52 3.3% 3.1% 2.8% 2.8% - cNGF_tane zumab 56 3.3% 3.6% - - Table E 5 . Summary of Average Disso ciation Constant, K D values usin g Surface Plasmon Resonance (SPR) for human pro - NGF and canine NGF and Yeast Surface Display (YSD) for pro.v4 - cNGF, and sorting conditions for library screening using pro.v4 - cNGF. The K D values from SPR were obtained using t he 1:1 biding model. Error bars represent 1 standard deviation of the regression. One tail t - test assuming unequal variances was used to calculate p - values for Hill coef Surface Plasmon Resonance Data Yeast Surface Display Data mAb H uman Pro - NGF Human NGF Canine NG F Average K D values with Hill coefficient, H=1 [pM] Labeling Concentrations for Screening Libraries [pM] Average K D values varying Hill co efficient, H [pM] Hill coefficient values p - values for Hill coefficient Average K D values [pM] Chi2 Average K D val ues [pM] Chi2 Average K D values [pM] Chi2 tanezumab 1610 0.01 15.9 1.26 19 0.77 801 ± 164 400.6 1319 ± 200 0.66 ± 0.05 0.0033 mAb # 1 286 0.22 0.243 2.32 0.118 1.92 209 ± 65 104.6 189 ± 36 0.70 ± 0.18 0.0103 mAb #2 48 0.05 0.308 1.47 0.074 1.29 307 ± 173 153.6 461 ± 300 0.62 ± 0.09 0.0003 mAb #3 2500 0.07 1.24 5.36 0.179 4.92 143 ± 44 71.6 325 ± 328 0.58 ± 0.19 0.0106 155 Table E 6 . FACS collection statistics for cNGF libraries m Ab Amount of c ollected c ells cNGF Library 1 Library 2 tanezumab 250,000 8.01% 6.10% mAb #1 250,000 7.36% 6.33% mAb #2 250,000 7.28% 6.60% mAb #3 250,000 6. 59% 8.48% Table E 7 . Primers set for deep sequencing. L1: library 1, L2: library 2, Blue : Illumina Universal Sequence, NNNNNN : Indexing Barcode, and Green : Illumina Adapter N ame Sequence Inner Primers Pro - cNGF_L1_FWD 5' - GTTC AGAGTTCTACAGTCCGACGATC GATGACGACAAGCATATG - 3' Pro - cNGF_L1_REV 5' - CCTTGGCACCCGAGAATTC CA AACTTTGGATCAACTGTGAT - 3' Pro - cNGF_L2_FWD 5' - GTTCAGAGTTCTACAGTCCGACGATC TTACAGGTCAAACTAGAAAC - 3' Pro - cNGF_L2_REV 5' - CCTTGGCACCCGAGAATTCCA AACTGGATGAGATGAAGA - 3' Pro - cNGF_FWD 5' - GTTCAGAGTTCTACAGTCCGACGATC GATGACGACAAGCATATG - 3' - cNGF_REV 5' - CCTTGGCACCCGAGAATTCCA AACTGGA TGAGATGAAGA - 3' cNGF_L1_FWD 5' - GTTCAGAGTTCTACAGTCCGACGATC AACTGGATGAGATGAAGA - 3' cNGF_L1_REV 5' - CCTTGGCACCCGAGAATTCCA ATCAACTGGAGTTGG - 3' cNGF_L2_FWD 5' - GTTCAGAGTTCTACAGTCCGACGATC CTTTTTTGAAACAAAATGTAGAGAT - 3' cNGF_L2 _REV 5' - CCTTGGCACCCGAGAATTCCA G CCTCCTCCACC - 3' Foward Outer Primer Illumina_FWD - AATGATACGGCGACCACCGAGATCTACAC GTTCAGAGTTCTACAGTCCGACGATC - Reverse Outer Primer s Pro - cNGF_L1_Unsel - CAAGCAGAAGACGGCATACGAGAT TGACAT GTGACTGGAGTT CCTTGGCACCCGAGAATT CCA - Pro - cNGF_L1_Display 5 - CAAGCAGAAGACGGCATACGAGAT GGACGG GTGACTGGAGTT CCTTGGCACCCGAGAATTCCA - Pro - cNGF_L1_tanezumab - CAAGCAGAAGACGGCATACGAGAT CTCTAC GTGACTGG AGTT CCTTGGCACCCGAGAATTCCA - Pro - cNGF_L1_mAb #1 - CAAGCAGAAGACGGCATACGAGAT GCGGAC GTGACTGGAGTT CCTTGGCACCCGAGAATTCC A - 156 Table E7 (cont d) Pro - cNGF_L2_Unsel - CAAGCAGAAGACGGCATACGAGAT TTTCAC GTGACTGGAGTT CCTTGGCACCCGAGAATTCCA - Pro - cNGF_L2_Displa y - CAAGCAGAAGAC GGCATACGAGAT GGCCAC GTGACTGGAGTT CCTTGGCACCCGAGAATTCCA - Pro - cNGF_L2_tanezumab - C AAGCAGAAGACGGCATACGAGAT CGAAAC GTG ACTGGAGT T CCTTGGCACCCGAGAATTCCA - Pro - cNGF_L2_mAb #1 - CAAGCAGAAGACGGCATACGAGAT CGTACG GTGACTGGAGTT CCTTGGCACCCGAGAA TTCCA - - cNGF_Unsel - C AAGCAGAAGACGGCATACGAGAT CCACTC GTGACTGGAGTT CCTTGGCACCCGAGAATTCCA - - NGF_Display - CAAGC AGAAGACGGCATACGAGAT GCTACC GTGACTGGAGTT CCTTGGCACCCGAGAATTCCA - - cNGF_tanezumab - CAAGCAGAAGACGGCATACGAGAT AT CAGT GTGACTGG AGTT CCTTGGCACCCGAGAATTCCA - - cNGF_mAb #1 - CAAGCAGAAGACGGCATACGAGAT GCTCAT GTGACT GGAGTT CCTTGGCACCCGAGAATTCCA - Pro - cNGF_L1_S2_tanezumab - CAAGCAGAAGACGGCATACGAGAT AGGAAT GTGACTGGAGTT CCTTGGCACCCGAGAATTCCA - Pro - cNGF_L1_S2_mA b #1 - CA AGCAGAAGACGGCATACGAGAT CTTTTG GTGACTGGAGTT CCTTGGCACCCGAGAAT TCCA - Pro - cNGF_L2_S2_tanezumab - CAAGCAGAAGACGGCATACGAGAT TAG TTG GTGACTGGAGTT CCTTGGCACCCGAGAATTCCA - Pro - cNGF_L2_S2_mAb #1 - CAAGCAGAAGACGGCATACGAGAT CCGGTG GTGACTGGAGTT CCTTGGC ACCCGAGAATTC CA - cNGF_L1_Unsel - CAAGCAGAAGACGGCATACGAGAT CGTGAT GTGACTGGAGTT CCTTGGCACCCGAGAATTCCA - cNGF_L1_Display - CAAGCA GAAGACGGCATACGAGAT ACATCG GTGACTGGAGTT CCTTGGCACCCGAGAATTCCA - cNGF_L1_mAb #1 - CAAGCAGAAGACGGCATACGAGAT GCCTAA GTGAC TGGAGTT CCTTG GCACCCGAGAATTCCA - cNGF_L1_mAb #2 - CAAGCAGAAGACGGCATACGAGAT TGGTCA GTGACTGGAGTT CCTTGGCA CCCGAGAATTCCA - cNGF_L1_mAb #3 - CAAGCAGAAGACGGCATACGAGAT CACTGT GTGACTGGAGTT CCTTGGCACCCGAGAATTCCA - cNGF_L1_tanezumab - CAAGCAGAAGACGGCATA CGAGAT A TTGGC GTGACTGGAGTT CCTTGGCACCCGAGAATTCCA - cNGF_L2_Unsel - CAAGCAGAAGACGGCATACGAGAT CTGATC GTGA CTGGAGTT CCTTGGCACCCGAGAATTCCA - cNGF_L2_Display - CAAGCAGAAGACGGCATACGAGAT AAGCTA GTGACTGGAGTT CCTTGGCACCCGAGAATTCCA - cGF_L2_mAb #1 - CAAGC AGAAGACGGCAT ACGAGAT GTAGCC GTGACTGGAGTT CCTTGGCACCCGAGAATTCCA - cNGF_L2_mAb #2 - CAAGCAGAAGACGGCATACG AGAT TACAAG GTGACTGGAGTT CCTTGGCACC CGAGAATTCCA - cNGF_L2_mAb #3 - CAAGCAGAAGACGGCATACGAGAT TTGACT GTGACTGGAGTT CCTTGGCACCCGAGAATTCCA - cNGF_L2_tan ezumab - CAAGCAGAAGACGGCATACGAGAT GGAACT GTGACTGGAGTT CCTTGGCACCCGAGAATTCCA - 157 Table E 8 . Libraries Stat istics Results for cNGF constructs Pro - cNGF Pro 1,2 - cNGF NGF Tile 1 Tile 2 Tile 1 Tile 2 Percent of possible codon subsititions observed in the unselected population: 1 - base substitution 96.50% 95.10% 94.20% 98.70 % 99.80% 2 - base substitutions 5 9.50% 52.80% 53.20% 55.70% 60.00% 3 - base substi tutions 47.40% 43.20% 44.80% 44.50% 47.90% Percent of unselected reads with: No nonsynonymous mutations: 35.20% 30.80% 28.60% 39.50% 34.50% One nonsynonymous mutation: 56.00% 61.80% 56.60% 55.40% 57.1 0% Multiple nonsynonymous mutations: 8.80% 7.40 % 14.80% 5.10% 8.50% Coverage of possible single nonsynonymous amino acid mutations: 85.50% 83.20% 84.20% 92.90% 98.20% 158 Table E 9 . Primers set f or deep sequencing. DSM: double site mutant library, Rep 1: Replicate 1, Re p 2: replicate 2, Red : Illumina Universal Sequence, NNNNNN : Indexing Barcode, and Blue : Illumina Adapter N ame Sequence Inner Pri mers AmiE_FWD 5' - GTTCAGAGTTCTACAGTCCGACGATC TTAACTTTAAGAAGTTTTTATACAT - 3' AmiE_ REV 5' - CCTTGGCACCCGAGAATTCCA AA GCACGGCTAAAGAT - 3' PYR1 _ FWD 5' - GTTCAGAG T TCTA CAGTCCGACGATC ATGCACGCGCGAC - 3' PYR1 _ REV 5' - CCTTGGCACCCGAGAATTCCA CGCGAGTTTCTGCAA - 3' UCA 9 _ FWD 5' - GTTCAGAGTTCTACAGTCCGACGATC GGGCTGAGGTGAAGAAG - 3' UCA9_REV 5' - CCTTGGC ACCCGAGAATTCCA GGTGACCAGGGTTCC - 3' Fo r ward Outer Primer Illumina FWD - AAT GATACGGCGACCACCGAGATCTACAC GTTCAGAGTTCTACAGTCCGACGATC - Reverse Outer Primers AmiE_Rep 1 - CAAGCAGAAGACGGCATACGAGAT GATCTG GTGACTGGAGTT CCTTGGCACCCGAGAATTCCA - AmiE_Re p 2 - CAAGCAGAAGACGGCATACGAG AT TCAAGT GTGACTGGAGTT CCTTGGCACCCGAGAATTCCA - PYR1_Rep 1 - CAAGCAGAAGACGGCATACGAGAT CACTGT GTGACTGGAGTT CCTTGGCACCCGAGAATTCCA - PYR1_R ep 2 - CAAGCAGAAGACGGCATACGAGAT ATTGGC GTGACTGGAGTT CCTTGGCACCCGAGAATTCCA - PY R1_DSM - CAAGCAGAAGACGGCATAC GAGAT GCCTAA GTGACTGGAGTT CCTTGGCACCCGAGAATTCC A - 3 UCA9 _Rep 1 - CAAGCAGAAGACGGCATACGAGAT TTGACT GTGACTGGAGTT CCTTGGCACCCGAGAATTCCA - UCA 9_Rep 2 - CAAGCAGAAGACGGCATACGAGAT AGGAAT GTGACTGGAGTT CCTTGGCACCCGAGAATTCCA - AmiE _"NNN" - CAAGCAGAAGACGG CATACGAGAT CTCTAC GTGACTGGAGTT CCTTGGCACCCGAGA A TTCC A - 159 Table E 10 . Summary of statistics for single site satu ration mutagenesis (SSM) and double site saturation mutagenesis (DSM) libraries O ligo Pool Primers Degenerate "NN N" Oligos SSM DSM SSM AmiE Replicate 1 Ami E Replicate 2 PYR1 Replicate 1 PYR1 Replicate 2 UCA9 Replicate 1 UCA9 Replicate 2 PYR1 AmiE Percentage of reads with: No nonsynonymous mutations: 50.8% 54.6% 63.0% 55.8% 71 .1% 71.5% 60.0% 39 .0 % One nonsy nonymous mutation: 38.0% 34.9% 24.1% 29.0% 1 4 .3% 13.9% 32.0%* 48.0 % Multiple nonsynonymous mutations: 11.2% 10.5% 11.8% 15.2% 15.5% 14.6% 9. 0% 13.0 % Coverage of all possible nonsynonymous amino acid mutations: 100% 100% 1 00% 100% 96.7% 98% 79.2%* 100% * 1 to 2 nonsynonymous mutations More than 2 non synonymous mutations 160 REFERENCES 161 REF E RENCES 1. Schr eiber, G. & Fleishman, S. J. Computational design of protein protein interactions. Curr. Opin. Struct. Biol. 23, 903 910 (2013 ) . 2 . Ecker, D. M., Jones, S. D. & Levine, H. L. The therapeutic monoclonal antibody market. MAbs 7, 9 14 (2015). 3. Grilo, A. L. & Mantalaris, A. The Increasingly Human and Profit able Monoclonal Antibody Market. Trends Biotechnol. 37, 9 16 (2019). 4. Scot t , A . M., Allison, J. P. & Wolchok, J. D. Monoclonal antibodies in cancer therapy. Cancer Immun. 12, 14 (2012). 5. Mata - Fink, J. e t al. Rapid Conformational Epitope Mapping of Anti - gp120 Antibodies with a Designe d Mutant Panel Displayed on Yeast. J. Mol. B i ol. 4 25, 444 456 (2013). 6. Sukupolvi - Petty, S. et al. Structure and function analysis of therapeutic monoclonal antibodies again st dengue virus type 2. J. Virol. 84, 9227 39 (201 0). 7. Correia, B. E. et al. Pro of of principle for epitope - focused vaccine d esi gn . Nature 507, 201 206 (2014). 8. Whitehead, T. A. et al. Optimization of affinity, specificity and function of designed infl uenza inhibitors using deep sequencing. Nat. Biote chnol. 30, 543 548 (2012). 9. Wa ng, L. - F. & Yu, M. Epitope Identification an d Di sc overy Using Phage Display Libraries: Applications in Vaccine Development and Diagnostics. Curr. Drug Targets 5, 1 15 (2004). 10. Van Blarcom, T. et al. Precise and Efficient Antibody Epitope Determination t hrough Library Design, Yeast Display and Nex t - Ge ne ration Sequencing. J. Mol. Biol. 427, 1513 1534 (2015). 11. Doolan, K. M. & Colby, D. W. Conformation - Dependent Epitopes Rec ognized by Prion Protein Antibodies Probed Using M utational Scanning and Deep Sequ encing. J. Mol. Biol. 427, 328 340 (2015). 1 2 . K owalsky, C. A. et al. Rapid fine conformational epitope mappin g using comprehensive mutagenesis and deep sequencing. J. Biol. Chem. 290, 26457 70 (2015). 13. Fowler, D. M. & Fi elds, S. Deep mutational scannin g: a new style of protein science. Nat. Meth o ds 11, 801 807 (2014). 14. Weiss, G. A., Watanabe, C. K., Zhong, A., Goddard, A. & Sidhu, S. S. Rapid mapping of protein function al epitopes by combinatorial alanine scanning. Pro c. Natl. Acad. Sci. U. S. A. 97, 8950 4 (2000). 15. Chao, G., Cochran, J. R. & D ane Wittrup, K. Fine Epitope Mapping of anti - Epidermal 162 Growth Factor Receptor Antibodies Through Random Mutagenesis and Yeast Surface Display. J. Mol. Biol. 342, 539 550 (2004) . 16. Fowler, D. M. et al. High - resolution mapping of protein sequence - funct i on relationships. Nat. Publ. Gr. 7, 741 746 (2010). 17. Chao, G. et al. Isolating and engineering human antibodies using yeast su rface display. Nat. Protoc. 1, 755 768 (2006). 18. Deventer, J. A. Van & Wittrup, K. D. Yeast Surface Display for Antibody Iso l ati on: Library Construction, Library Screening, and Affi nity Mat uration. Methods Mol. Biol. 1131, 151 181 (2014). 19. Adams, R. M., Mora, T., Walczak, A. M. & Kinney, J. B. Measu ring the sequence - affinity lands cape of antibodies with massively parallel t i tra tion curves. Elife 5, 5980 5985 (2016). 20. Le Saux, S. et al. Molecular dissection of human interleukin - 31 - mediated signal tr ansduction through site - directed mutagenesis. J. B iol. Chem. 285, 3470 7 (2010). 2 1. Burns, M. L. et al. Directed evolution of bra in - derived neurotrophic factor for improved folding and expression in Saccharomyces cerevisiae. Appl. Environ. Microbiol. 80, 5732 42 (2014). 22. Huang, E. J. & Reichardt, L. F . Neurotrophins: Roles in Neuron al Development and Fun ction. Annu. Rev. Neur o sci . 24, 677 736 (2001). 23. Wrenbeck, E. E. et al. Plasmid - based one - pot saturation mutagenesis. Nat. Methods 13, 928 930 (2016) . 24. Medina - Cucurella, A. V. & Whitehead, T. A. C haracterizing Protein - Protein In teractions Using Deep Sequencing Coupled to Y eas t Surface Display. Methods Mol. Biol. 1764 , 101 - 121 (2018) . 25. Kowalsky, C. A. et al. High - Resolution Sequence - Function Mappin g of Full - Length Proteins. PLoS One 10, e0118193 (2015). 26. Nelson, A. L., Dhimo lea, E. & Reichert, J. M. Development trends for human monoclonal antibody therapeutics. Nat. Rev. Drug Discov. 9, 767 774 (2010). 27. Smith, S. A. & Crowe, Jr., J. E. Use of Human Hybridoma Technology To Isolate Human Monoc lonal Antibo dies. Microbiol. Spe ctr. 3, 141 156 (2015). 28. Adler, A. S. et a l. Rare, high - affinity mouse anti - PD - 1 antibodies that function in checkpoint blockade, discovered using microfluidics and molecul ar genomics. MAbs 9, 1270 1281 (2017). 29. Adler, A. S. et al . Rare, high - affinit y anti - pathogen antibodies from human repert o ire s, discovered using microfluidics and molecular genomics. MAbs 9, 1282 1296 (2017). 30. Adler, A. S. et al. A natively paired a ntibody library yields drug leads with higher sen sitivity and specificity than a randomly paired antibody library. MAbs 10, 4 3 1 4 43 (2018). 163 31. Medina - Cucurella, A. et al. Preferential Identification of Agonistic OX40 Antibodies by Using Cell Lysate to Pan Natively Paired, Humanized Mouse - Derived Yeast S urface Displ ay Libraries. Antibo dies 8, 17 (2019). 32. Dillon, S. R. et al. I nte rleukin 31, a cytokine produced by activated T cells, induces dermatitis in mice. Nat. Immunol. 5, 752 760 (2004). 33. Neis, M. M. et al. Enhanced expression levels of IL - 31 co rrelate with IL - 4 and IL - 13 in a topic and allergic contact dermatitis. J. Al l erg y Clin. Immunol. 118, 930 937 (2006). 34. Rabenhorst, A. & Hartmann, K. Interleukin - 31: A Novel Diagnostic Marker of Allergic D iseases. Curr. Allergy Asthma Rep. 14, 423 (2014) . 35. Cornel issen, C., Lüscher - F irzlaff, J., Baron, J. M. & Lüscher, B. Sign a lin g by IL - 31 and functional consequences. Eur. J. Cell Biol. 91, 552 566 (2012). 36. Takaoka, A. et al. Involvement of IL - 31 on s cratching behavior in NC/Nga mice with atopic - lik e dermatitis. Exp. Dermatol. 15, 161 167 (2006). 37. Sonkoly, E. et al. IL - 3 1 : A new link between T cells and pruritus in atopic skin inflammation. J. Allergy Clin. Immunol. 117, 41 1 417 (2006). 38. Lewis, K . E. et al. Interleukin (IL) 31 induces in cynomo lgus monkeys a rapid and intense itch response that can be inhibited by an I L - 31 neutralizing antibody. J. Eur. Acad. Dermatology Venereol. 31, 142 150 (2017). 39. Raap, U. et al. C orrelation of IL - 31 serum levels with severity of atopic dermatitis. J. All ergy Clin. Immunol. 122, 421 3 ( 2008). 40. Ezzat, M., Hasan, Z. & Shaheen, K . Se rum measurement of interleukin - 31 (IL - 31) in paediatric atopic dermatitis: elevated levels correlate with severity scoring. J. Eur. Acad. Dermatology Venereol. 25, 334 339 (201 1). 41. Pantazi, E., Valenza, G. , Hess, M. & Hamad, B. The atopic dermatitis mar ket. Nat. Rev. Drug Discov. 17, 237 (2017). 42. Nemoto, O. et al. The first trial of CIM331, a humani zed antihuman interleukin - 31 receptor A antibody, in healthy volunteers and patients with atopic dermatitis to evaluate safety, tolerability and pharma c oki netics of a single dose in a randomized, double - blind, placebo - co. Br. J. Dermatol. 174, 296 304 (201 6). 43. Michels, G. M. et al. A blinded, randomized, placebo - controlled, do se determination trial of lokive tmab (ZTS - 00103289), a caninized, anti - canin e IL - 31 monoclonal antibody in client owned dogs with atopic dermatitis. Vet. Dermatol. 27, 478 - e129 (201 6). 44. Diveu, C. et al. P redominant expression of the long isoform of GP13 0 - like (GPL) receptor is require d for interleukin - 31 signaling. Eur. Cytokin e Ne tw. 15, 291 302 (2004). 45. Zhang, Q., Putheti, P., Zhou, Q., Liu, Q. & Gao, W. Structures and biological functions of IL - 31 an d IL - 31 receptors. Cytokine Growth Factor Rev. 19 , 347 356 (2008). 46. Diveu, C. et al. GPL, a novel cytokine receptor relate d to GP130 and leukemia inhibito ry factor receptor. J. Biol. Chem. 278, 49850 9 (2003). 164 47. Dambacher, J. et al. Interleukin 31 med iates MAP kinase and STAT1/3 activation in intest inal epithelial cells and its ex pression is upregulated in inflammatory bowe l di sease. Gut 56, 1257 65 (2007 ). 48. Dreuw, A. et al. Characterization of the signaling capacities of the novel gp130 - like cytoki ne receptor. J. Biol. Chem. 279, 36112 20 (2004). 49. Boulanger, M. J., Chow, D., Brevnova, E. E. & Garcia, K. C. Hexameric S t ruc ture and Assembly of the Int erleukin - 6/IL - - Receptor/gp130 Complex. Science 300, 2101 2104 (2003). 50. Araya, C. L. & Fowler, D. M. Deep mutational scanning: assessing protei n function on a massive scale. T rends Biotechnol. 29, 435 42 (2011). 51. Medina - Cucurella, A. V., Zhu, Y., B owen, S. J., Bergeron, L. M. & Whitehead, T. A. Pro region engineering of nerve growth factor by deep mutational scanning enables a yeast platform f or conformational epitope mappin g of anti - NGF monoclonal antibodies. Biotechnol. Bioeng. 115, 1925 1937 (201 8). 52. Roy, A., Kucukural, A. & Zhang, Y. I - TASSER: a unified platform for automated protein structure and function prediction. Nat. Protoc. 5, 725 738 (2010). 53. Zhang, Y. et al . I - TASSER server for protein 3D structure prediction. BMC Bioinforma. 2008 91 59, 305 309 (2008). 54. Chen, V. B. et al. Biological Crystallography MolProbity: all - atom structure validation for macromolecular crystallograph y. Acta Crystallogr D Biol Cryst allogr. 66 , 12 - 21 (2010) . 55. Lessmann, V., Gottmann, K. & Malcangio, M. Neu r otrophin secretion: current facts and future prospects. Prog. Neurobiol. 69, 341 374 (2003). 56. Park, H. & Poo, M. Neurotrophin regulation of neur al circuit development and funct ion. Nat. Rev. Neurosci. 14, 7 23 (2013). 57. Rattenholl, A. et al. The pro - s equence facilitates folding of human nerve gr owth factor from Escherichia coli inclusion bodies. Eur. J. Biochem. 268, 3296 3303 (2001). 58. Watson , J. J., Allen, S. J. & Dawbarn, D. Targeting Nerve Growth Factor in Pain. BioDrugs 22, 349 359 (2008). 59. S losky, L. M., Largent - Milnes, T. M. & Vandera h, T. W. Use of Animal Models in Understanding Cancer - induced Bone Pain. Cancer Growth Metastasis 8, 4 7 62 (2015). 60. La Porte, S. L. et al. Generation of a high - fidelity antibody against nerve growth factor u s ing library scanning mutagenesis and validati on with structures of the initial and optimized Fab - antigen complexes. MAbs 6, 1059 1068 (2014). 61. C hang, D. S., Hsu, E., Hottinger, D. G. & Cohen, S. P. Anti - nerve growth factor in pain management: current e v idence. J. Pain Res. 9, 373 83 (2016). 62. Kumar, V. & Mahal, B. A. NGF - the TrkA to successful pain treatment. J. Pain Res. 5, 279 87 (2012). 165 63 . Burns, M. L. et al. Pro - region engineering for improved yeast display and secretion of brain derived neurot r ophic factor. Biotechnol. J. 11, 425 436 (2016). 64. Kliemannel, M., Golbik, R., Rudolph, R., Schwarz, E. & Lilie, H. The pro - peptide of proNGF: S t ructure formation and intramolec ular association with NGF. Protein Sci. 16, 411 419 (2007). 65. Nomoto, H., T akaiwa, M., Mouri, A. & Furukawa, S. Pro - region of neurotrophins determines the processing efficiency. Biochem. Biophys. Res. Commun. 356, 919 924 (2007). 66. Hauburger, A., Kliem annel, M., Madsen, P., Rudolph, R. & Schwarz, E. Oxidative folding of nerve g rowth factor can be mediated by the pro - peptide of neurotrophin - 3. FEBS Lett. 581, 4159 4164 (2007). 67. Rattenholl, A. et al. Pro - sequence assist e d folding and disulfide bond for mation of human nerve growth factor. J. Mol. Biol. 305, 523 533 (2001). 68. F eng, D. et al. Molecular and Structural Insight into proNGF Engagement of p75NTR and Sortilin. J. Mol. Biol. 396, 967 984 (2010). 69. Kliemannel, M . et al. The mature part of proN GF induces the structure of its pro - peptide. FEBS Lett. 566, 207 212 (2004). 70. Suter, U., Heymach, J. V, Shooter, E. M. & Shooter, E. M. Two conserved domains in the NGF propeptide are necessary and sufficient for the bio s ynthesis of correctly processed and biologically active NGF. EMBO J. 10, 2395 400 (1991). 71. Pagadala, P. C . , Dvorak, L. A. & Neet, K. E. Construction of a mutated pro - nerve growth factor resistant to degradation and suitable for biophysical and cellular utilization. Proc. Natl. Acad. S ci. U. S. A. 103, 17939 43 (2006). 72. Wrenbeck, E. E., Faber, M. S. & White h ead, T. A. Deep sequencing methods for protein engineering and design. Curr. Opin. Struct. Biol. 45, 36 44 (2017). 73. Kitamura, T. et al. Establis hment and characterization of a unique human cell line that proliferates dependently on GM - CSF , IL - 3, or ery t hropoietin. J. Cell. Physiol. 140, 323 334 (1989). 74. Kowalsky, C. A. & Whitehead, T. A. Determination of binding affinity upon mutation for type I dockerin - cohesin complexes fro m C lostridium thermocellum and C lostridium cellulolyticum us ing deep seque n cing. Proteins Struct. Funct. Bioinforma. 84, 1914 1928 (2016). 75. Klesmith, J. R., Bacik, J. - P., Wrenbeck, E. E., Michalczyk, R. & Whitehead, T. A. Trade - offs between enzyme fit ness and solubility illuminated by deep mutational scanning. P roc. Natl. Aca d . Sci. U. S. A. 114, 2265 2270 (2017). 76. Abeliovich, H. An Empirical Extremum Principle for the Hill Coefficient in Ligand - Protein Interactions S howing Negative Cooperativity. B iophys. J. 89, 76 79 (2005). 77. Ellgaard, L. & Helenius, A. Q uality control in the endoplasmic reticulum. Nat. Rev. Mol. 166 Cell Biol. 4, 181 191 (2003). 78. Park, S. et al. Limitations of yeast surface display in engineering proteins of high thermostability . Protein Eng. Des. Sel. 19, 211 217 (2006). 79. Gai, S. A. & Wittrup, K. D. Yeast surface display for protein engineering and characterization. Curr. Opin. Struct. Biol. 17, 467 473 (2007). 80. Firnberg, E. & Ostermeier, M. PFunkel: Efficient, Expansive, User - Defined Mutagenesis. PLoS One 7, e52031 (2012). 81. Cozens, C. & Pinhei r o, V. B. Darwin Assembly: fast, efficient, multi - site bespoke mutagenesis. Nucleic Acids Res. 46, e51 e51 (2018). 82. Kosuri, S. & Church, G. M. La rge - scale de novo DNA synthesis: technologies and applications. Nat. Methods . 11, 499 507 (2014). 83. Plesa, C., Sidore, A. M., Lubock, N. B., Zhang, D. & Kosuri, S. Multiplexed gene synthesis in emulsions for exploring protein functional landscapes. Scien ce 359, 343 347 (2018). 84. Wren beck, E. E., Azouz, L. R. & Whitehead, T. A. Single - mutation fitness landsca p es for an enzyme on multiple substrates reveal specificity is globally encoded. Nat. Commun. 8, 1 5695 (2017). 85. Pappas, L. et al. Rapid developme nt of broadly influenza neutrali zing antibodies through redundant mutations. Nature 516, 418 422 (2014). 86. Park, S. - Y. et al. Agrochemical control of plant water use using engineered abscisic acid recepto rs. Nature 520, 545 548 (2015). 87. Klesmith, J. R . & Hackel, B. J. Improved mutan t function prediction via PACT: Protein Analysis and Classifier Toolkit. Bio i nformatics . (2018). 88. Wrenbeck, E., Klesmith, J., Stapleton, J. & Whitehead, T. Nicking Mutage nesis: comprehensive single - site saturation mutage nesis. Protoc. Exch. (2016). 89 . Das, R. & Baker, D. Macromolecular Modeling with Rosetta. Annu. Rev. Bioch e m. 77, 363 382 (2008). 90. Leaver - Fay, A. et al. ROSETTA3: an object - oriented software suite for the simulation and design of macromolecules. Metho ds in enzymology 487, 545 574 (2 011). 91. Chaudhury, S. et al. Benchmarking and Analysis of Protein Docking Performance in Rosetta v3.2. PLoS One 6, e22477 (2011). 92. van der SchaarJan, H. M., Wilschut , J. C. & Smit, J. M. Role of antibodies in controllin g dengue virus infection. Immuno biology 214, 613 629 (2009). 9 3. Schwartz, L. M., Halloran, M. E., Durbin, A . P. & Longini, I. M. The dengue vaccine pipeline: Implications for the future of dengue contr ol. Vaccine 33, 3293 3298 (2015). 94. Katzelnick, L. C . et al. Antibody - dependent enha ncement of severe dengue disea se in humans. Science 358, 929 932 (2017). 167 95. Murrell, S., Wu, S. - C. & Butler, M. Review of dengue virus and the development of a vaccine. Biotechnol. Adv. 29, 239 247 (2011). 96. Dejnirattisai , W. et al. Dengue virus sero - cr oss - reactivity drives antibody - dependent enhancement of infection with zi ka virus. Nat. Immunol. 17, 1102 1108 (2016). 97. Gietz, R. D. & Schiestl, R. H. High - efficiency yeast transformation using the LiAc/SS carrier DNA/PEG method. Nat. Protoc. 2, 31 34 ( 2007). 98. Sambrook, J. & Russ ell, D. W. Transformation of E. coli by Ele ctr oporation. CSH Protoc. 2006, pdb.prot3933 (2006). 99. Fowler, D. M., Araya, C. L., Gerard, W. & Fields, S. Enrich: software for analysis of protein function by enrichment and deple tion of variants. Bioinformati cs 27, 3430 3431 (2011). 100. KÖHLER, G. & MIL STEIN, C. Continuous cultures of fused cells secreting antibody of predefined specificity. Nature 256, 495 497 (1975). 101. McCafferty, J., Griffith s, A. D., Winter, G. & Chiswell, D. J. Phage antibodies: filam entous phage displaying antibody variable d omains. Nature 348, 552 554 (1990). 102. Kohl, T. O. & As coli, C. A. Direct and Indirect Cell - Based Enzyme - Linked Immunosorbent Assay. Cold Spring Harb . Protoc. 2017, pdb.prot093732 ( 2017). 103. Even - Desrumeaux, K . & Chames, P. Phage display and selections on cells. Methods Mol Biol. 907 , 225 235 (2012). 104. Sp encer, S., Bethea, D., Raju, T. S., Giles - Komar, J. & Feng, Y. Solubility evaluation of murine hybridoma antibodies. MAbs 4, 3 19 325 (2012). 105. Jain, T. e t al. Biophysical properties of the clinica l - stage antibody landscape. Proc. Natl. Acad. Sci. U. S. A. 114, 944 949 (2017). 106. Rajan, S. et al. Recombinant human B cell repertoires enable scre ening for rare, specific, and na tively paired antibodies. Comm un. Biol. 1, 5 (2018). 107. Wang, B. et al. Functional interrogation and mining of natively paired human VH:VL antibody repertoires. Nat. Biotechnol. 36, 152 155 (2018). 108. Linch, S. N., McNam ara, M. J. & Redmond, W. L. OX40 Agonists and Combination Immu notherapy: Putting the Pedal to the Metal. Front. Oncol. 5, 34 (2015). 109. Willoughby, J., Griffiths, J., Tews, I. & Cragg, M. S. OX40: Structure and function What questions remain? Mol. Immu nol. 83, 13 22 (2017). 110. Comp aan, D. M. & Hymowitz, S. G. T he Crystal Structure of the Costimulatory O X40 - OX40L Complex. Structure 14, 1321 1330 (2006). 111. Tillotson, B. J., De Larrinoa, I. F., Skinner, C. A., Klavas, D. M. & Shusta, E. V. Antibody af finity maturation using yeast di splay with detergent - solubiliz ed membrane proteins as antigen sources. Pr otein Eng. Des. Sel. 26, 101 112 (2013). 112. Cho, Y. K. & Shusta, E. V. Antibody library screens using detergent - solubilized 168 mammalian cell lysates as antigen sources. Protein Eng. D es. Sel. 23, 567 577 (2010). 1 13. Edgar, R. C. & Flyvbjerg, H. Error filt ering, pair assembly and error correction for next - generation sequencing reads. Bioinformatics 31, 3476 3482 (2015). 114. Lefranc, M. - P. et al. IMGT(R) , the international ImMunoGeneTi cs information system(R). Nucl eic Acids Res. 37, D1006 D1012 (2009). 115. Edgar, R. C. Search and clustering orders of magnitude faster than BLAST. Bioinformatics 26, 2460 2461 (2010). 116. Csardi, G. & Nepusz, T. The igraph software package for complex ne twork research | BibSonomy. In terJournal, Complex Systems 1695 , 1 9 (2006 ). 117. Dreyer, A. M., Beauchamp, J., Matile, H. & Pluschke, G. An efficient system to generate monoclonal antibodie s against membrane - associated prot eins by immunisation with antige n - expressing mammalian cells. BMC Biotechnol. 10, 87 (2010). 118. Rezaei, M. & Ghaderi, A. Production of a Mouse Monoclonal Antibody Against Mortalin by Whole Cell Immunization. Monoclon. An tib. Immunodiagn. Immunother. 36, 169 175 (2017). 119. Tamura, T. & Chiba, J. Production of anti bodies against multipass membrane proteins expressed in human tumor cells using dendritic cell immunization. J. Biomed. Biotechnol. 2009, 673098 (2009). 120. Re ddy, S. T. et al. Monoclonal antib odies isolated without screening by analyzing the variable - gen e repertoire of plasma cells. Nat. Biotechn ol. 28, 965 969 (2010). 121. Saggy, I. et al. Antibody isolation from immunized animals: comparison of phage display and antibody discovery via V gene repertoire mining. Protein Eng. Des. Sel. 25, 539 549 (2012). 122. Wilson, J. R. et al. Diversity of the murine antibody response targeting influenza A(H1N1pdm09) hemagglutinin. Virology 458 459, 114 124 (2014).