Coupling mammalian cell surface display with somatic hypermutation for the discovery and maturation of human antibodies

Abstract

A novel approach has been developed for the isolation and maturation of human antibodies that replicates key features of the adaptive immune system by coupling in vitro somatic hypermutation (SHM) with mammalian cell display. SHM is dependent on the action of the B cell specific enzyme, activation-induced cytidine deaminase (AID), and can be replicated in non-B cells through expression of recombinant AID. A library of human antibodies, based on germline V-gene segments with recombined human regions was used to isolate low-affinity antibodies to human β nerve growth factor (hβNGF). These antibodies, initially naïve to SHM, were subjected to AID-directed SHM in vitro and selected using the same mammalian cell display system, as illustrated by the maturation of one of the antibodies to low pM K_D. This approach overcomes many of the previous limitations of mammalian cell display, enabling direct selection and maturation of antibodies as full-length, glycosylated IgGs.

The immune system is exquisitely adapted to generate a high frequency of functional antibodies starting from a small number of variable region genes. In the absence of antigen, pre-B cells undergo V(D)J recombination and express IgM on their surface. Complementarity-determining region 3 (CDR3) sequences encoded through such gene recombination are critical for antigen recognition by unmutated B cell receptors, and may be largely responsible for the primary repertoire (1–3). Naïve B cells use high avidity with surface IgM to facilitate low-affinity binding, with subsequent activation-induced cytidine deaminase (AID)-dependent somatic hypermutation (SHM) and class switching to generate high-affinity antibody responses (Fig. 1A). AID is essential for the initiation of SHM in B cells by the deamination of cytidine residues directly in Ig genes (4, 5). To achieve this, AID is targeted to V-region DNA sequences, termed hotspots (e.g., WRCH) that result in mutations and amino acid substitutions, which are frequently in positions biased to modulate antigen binding (6). Expression of AID alone has been shown to be sufficient to reproduce the salient features of SHM in both B cells and other mammalian cells (4, 7, 8).

Fig. 1.

Design of the ABELmAb library for mammalian surface display (A) Pro-B cells initially recombine V, D and J regions to express a naïve repertoire of avid IgM antibodies. Antigen binding to specific clones stimulates B-cell maturation, including class switching (IgG), proliferation, and AID-mediated SHM to produce antigen specific, high-affinity antibodies secreted by plasma cells and presented on circulating memory cells. (B) CDR3/FW4 IgG and IgM diversity was isolated from pooled PBMCs and grafted into selected V regions to produce a library of germline full-length Abs. Antigen-coated beads were used to isolate low-affinity antibodies, with subsequent FACS selection and AID-mediated maturation of high-affinity antibodies. (C) Design of the ABELmAb full-length library is shown highlighting placement of restriction sites that were used to graft CDR3/FW4 diversity (red, green, and orange regions), amplified from a pool of PBMCs from seven normal donors, into selected IgH (IGHV1-2, 1-69, 3-7, 3-23, 3-30-3, 4-34, 4-59, 5-51, and 6-1), IgK (IGKV4-1, 3-20, 2D-30, 1D-39, and 1-33) and IgL (IGLV1-40, 2-11, 3-21, and 7-43) synthesized V regions. Constant domains were also synthesized, and a transmembrane (blue) and cytoplasmic domain (red) were added to the C-terminal end of the HC.

Recombinant antibodies represent the fastest growing class of new medicines, and generation of antibodies that meet specific criteria is increasingly important for therapeutic applications. At present, there are two predominant methodologies for therapeutic antibody generation: immunization-based and surface display–based approaches. Immunization of wild type or transgenic animals (9, 10) is an effective method for generating antibodies to many antigens, and has led to the majority of Food and Drug Administration–approved therapeutic antibodies (11). Nevertheless, immune tolerance may lead to difficulties generating neutralizing antibodies when antigens are well conserved or are toxic upon administration to animals. Specific, immunodominant epitopes may be preferentially selected, making it difficult to identify functional antibodies (12).

Display technologies such as phage, yeast and ribosome display are based on the in vitro selection of antibody fragments from libraries (13) and overcome limitations of immune tolerance or epitope dominance in vivo. Selection from such libraries may not always generate high-affinity antibodies without subsequent affinity maturation (14). Furthermore, antibody fragments isolated from microbial display systems are not always easily reformatted to produce well-expressed IgGs, soluble enough to be formulated for subcutaneous delivery (15).

Mammalian cell expression systems offer a number of potential advantages for therapeutic antibody generation, including the ability to coselect for key manufacturing-related properties such as high-level expression and stability, while displaying functional glycosylated IgGs on the cell surface. However, mammalian cell display has been hampered by the smaller library sizes that can be screened, making direct isolation of high-affinity binders from naïve libraries improbable. Although small libraries biased toward a particular antigen have been used successfully (16–19), a more generalized approach to generate high-affinity human antibodies from immunologically naïve libraries has not been reported.

By mimicking in vitro features such as germline recombination of V-gene fragments, recovery of low-affinity initial binders, and subsequent AID-mediated SHM, we have carried out de novo discovery and maturation of an anti-human β nerve growth factor (hβNGF) antibody to high affinity.

Results

We sought to reproduce in vitro features of the adaptive immune system that are critical for generation of diverse high-affinity antibodies, including a discrete starting repertoire of human rearranged immunoglobulin variable region sequences, mammalian cell-surface presentation, high avidity for isolation of weak binders, and AID-mediated SHM (Fig. 1B). Although this approach is potentially applicable to any mammalian cell line, the HEK293 cell line was chosen because of its ease of transfection, handling, and rapid cell growth. A robust system for coexpressing IgG and AID in mammalian cells was created using episomal vectors expressing the heavy chain (HC), light chain (LC), and AID, each also encoding a separate antibiotic selectable marker (20, 21). To display immunoglobulin on the surface of these cells, full-length human IgG1 HC was modified by addition of a C-terminal transmembrane domain (22).

Construction of a Fully Human Library of IgGs.

The IgG library was based on germline sequence V-gene segments joined to prerecombined regions isolated from a panel of human donors. A total of nine HC and nine LC (5κ and 4λ) human germline V-gene segments were selected for the library based on the frequency of in vivo germline usage (23, 24) (Fig. 1C). V regions were chemically synthesized and fused to region sequences (encoding CDR3 and FR4 diversity) isolated by PCR from pooled peripheral blood mononuclear cells (PBMCs) of normal donors. Full-length V regions for HC and LC were assembled with human HC and LC constant regions and transfected into HEK293 cells. A sampling of the stably selected library by high-throughput sequencing (HTP) provided a lower estimate of 6 × 10⁷ total diversity of combinatorially expressed antibody sequences (25). The library was designed to provide multiple initial candidates with germline V-gene segments for further maturation by SHM, and is termed ABELmAb (AnaptysBio Evolving Library of monoclonal Antibodies).

Isolation of Novel Human Antibodies to hβNGF.

A human cytokine, hβNGF, was selected as a target for antibody discovery because of its well-described role in modulating pain sensation following tissue injury and inflammation (26). βNGF binds and activates its cognate receptor, tropomyosin-related kinase A receptor (TrkA), up-regulating the expression and activity of pathways that enhance acute and chronic pain. Antagonism of the βNGF/TrkA signaling pathway has been shown in animal and clinical studies to be a potent means of attenuating pain sensation in a number of clinical indications (27, 28).

The transfected library was expanded to 10⁹ cells, and subjected to four rounds of negative selection against streptavidin (SA)-coupled magnetic beads, followed by a single round of positive selection against SA-coupled magnetic beads coated with biotinylated βNGF (Fig. 1B). Positively selected cells were expanded, and two rounds of fluorescence-activated cell sorting (FACS) selection were performed under high avidity conditions. Single cell clones (SCCs) were isolated with the second round of FACS selection, sequenced, and each characterized for binding to βNGF and the ability of soluble TrkA-Fc receptor to compete this binding (Fig. 2 A–C, Table S1, and SI Materials and Methods). Of 37 isolated round 2 SCCs, six unique clones were chosen for affinity maturation and stably transfected with AID. Three rounds of FACS selection were performed using progressively lower concentrations of fluorescently labeled hβNGF antigen. Selected antibody-expressing cells exhibited improved hβNGF binding by the third round of SHM, and HC and LC sequencing of each selected population revealed enriched mutations, contained primarily in HC CDR regions.

Fig. 2.

Flow cytometry antigen-binding analyses of clone C10A, S1 and S2 affinity maturation strategies scattergrams showing βNGF binding to isolated ABELmAb cell clone C10A (A–C) and subsequent affinity maturation in strategies S1 (D–F) and S2 (G–I). Approximately 5 × 10⁷ cells were sorted per round. (A) C10A cells did not bind 50 nM irrelevant myc-tagged antigen, (B) bound 50 nM myc-βNGF under identical SA-generated avidity conditions, and (C) 50 nM βNGF binding was competed in the presence of equimolar concentration of 50 nM TrkA-Fc receptor. FACS scattergrams are shown for (D) round 1 (10 nM WFP-βNGF), (E) round 7 (50 pM Dyl-NGF), and (F) round 12 (35 pM Dyl-NGF) from strategy S1, and (G) round 1 (10 nM WFP-βNGF), (H) round 5 (500 pM Dyl-NGF) and (I) round 8 (150 pM Dyl-NGF) for strategy S2. Emerging populations of βNGF binding cells expressing CDR1 insertions GDTFSNYA and A were evident during intermediate rounds of selection in strategies S1 (E) and S2 (H), respectively (arrows).

Affinity Maturation of a hβNGF-Specific Antibody.

Expression in the HEK293 cells is stably maintained using an episomal system that supports a low copy number (3–5 per cell) of each vector (21). Unique HC and LCs from each SCC were cloned, combinatorially paired, expressed in HEK293 cells, and assessed in Biacore and flow cytometry-based antigen-binding assays. The HC/LC pair from each SCC providing the best binding to βNGF were retransfected with AID for further maturation. The sole HC isolated from SCC C10A contained an enriched mutation, S31N, in CDR1. One of three distinct germline LC sequences recovered from SCC C10A, in combination with this HC, was used for further maturation (APE391).

Affinity maturation of APE391 was carried out utilizing two independent but initially identical cell populations, strategies S1 and S2. Rounds 1–3 of FACS selection for each strategy were performed using low nM concentrations of βNGF fused to wasabi fluorescent protein (WFP), each round selecting approximately 0.2–0.5% of the brightest cells. Subsequent rounds of FACS selection used low pM concentrations of fluorochrome-labeled βNGF in combination with unlabeled βNGF competition. Sequences were monitored by the sequencing of approximately 40 HC and LC after each round. HTP sequencing was also employed to monitor enriching mutations during early rounds of FACS selection of strategy S1 (Table S2 and Fig. S1). Affinity maturation was observed starting with the third round of FACS selection, continuing through the final rounds of affinity maturation (Round 12 for S1, Round 9 for S2), as evidenced by improved βNGF binding (Fig. 2 D–I), sequence data (Table 1, Table S2, and SI Materials and Methods), and Biacore analysis of isolated clones (Figs. 3 and 4 and SI Materials and Methods).

Table 1.

AID-mediated mutations and insertions observed during affinity maturation

All HC mutations and insertion events observed during the S1 and S2 evolution corridors are shown. Column one indicates the category of SHM event, and the second column indicates the strategy in which it was identified (S1 or S2). Columns three and four show the starting and ending nucleotide sequence surrounding the site of mutation (codon in bold, mutation underlined) or insertion (original sequence in bold, duplicated sequence underlined). Columns five through seven highlight the starting and ending nucleotide and resulting mutation, referenced by Kabat. Seven of the 19 mutations involved nucleotide transversions, while the remaining 12 involved nucleotide transitions. Sixteen of the 19 mutations were initiated at the nucleotides G or C, three at nucleotide A, and thirteen of 19 occurred at known AID hotspots (WRCH) located primarily within CDRs 1, 2, or 3

Fig. 3.

Biacore and in vitro analysis of affinity maturing S1 strategy anti-βNGF antibodies. Selected sensorgrams and in vitro assays show the progression of antigen-binding kinetics to antibodies produced by HEK293 clones selected over the course of the S1 affinity maturation strategy (SI Materials and Methods). (A) Incorporation of mutation S31N (APE391) in the HC CDR1 results in detectable binding on Biacore. RU, resonance units. (B) Addition of mutation L45F in CDR2 of HC (APE583) improves affinity to low micromolar affinity (k_on = 9.5 × 10⁴ M^-1 s^-1, k_off = 0.08 s^-1). Incorporation of insertion CDR1 GDTFSNYA and a CDR3 mutation D100A (APE803) resulted in an affinity of 760 pM (k_on = 7.8 × 10⁶ M^-1 s^-1, k_off = 6.0 × 10^-3 s^-1). (C) Introduction of the remaining HC mutations into a combinatorial library resulted in the identification of an antibody (APE925) with a K_D = 25 pM (k_on = 1.4 × 10⁷ M^-1 s^-1, k_off = 3.5 × 10^-4 s^-1). (D) Antibodies originating from the S1 strategy and an antibody to an irrelevant antigen (APE273) were characterized in an HTRF assay for their ability to inhibit binding of tanezumab (APE081), an anti-hβNGF antibody. (E) Affinity matured antibodies APE803, 925, and 928 were able to inhibit binding of hβNGF to its high-affinity receptor TrkA-Fc, (F) and demonstrated potent inhibition of βNGF-dependent ERK1/2 phosphorylation in neuronal PC12 cells (42).

Fig. 4.

Sequence and kinetic analysis of affinity maturation trajectories. (A) The kinetic properties of affinity matured anti-βNGF antibodies, secreted from SCCs in strategies S1 and S2 and monitored by SPR, are shown in a k_on vs. k_off plot. Unsequenced cell clones recovered during the S1 and S2 strategies are shown in gray. (B) HC chain sequence variants corresponding to the affinity maturing clones shown in A are displayed, starting at Kabat position 22, with mutations highlighted in orange and codon insertions shown in blue.

Mutational Trajectories Accompanying Affinity Maturation of C10A.

Maturation of APE391 in strategy S1 continued with the observation of the L45F mutation in round 3 (out of 32 sequences), with all recovered sequences containing this mutation by round 4 (APE583, Fig. 4). Kinetic characterization of this mutation was associated with a significant improvement in binding affinity for hβNGF. Two HC FW1/CDR1 insertions, GDTFSNYA (position 26) and TFSNYAI (position 28), were found to be enriched in the context of APE583 to a majority of recovered sequences from rounds 6 and 7 (APE659 and APE646, respectively), each associated with an approximately 20-fold improvement in binding affinity. An HC CDR3 mutation, D100A, was observed in round 12 in the context of the APE646 variant, improving k_on 100-fold and leading to an affinity of 760 pM (APE803, Fig. 4).

Affinity maturation of APE391 in the S2 strategy commenced in the HC with the incorporation of a single amino acid insertion, A24 (APE608), coincident with incorporation of mutations G44R and G55D in round 3, each associated with significant improvements in affinity (Fig. 4). APE608 became the predominate HC sequence by round 5, with additional CDR1, FW3 and CDR3 HC mutations, identified in subsequent rounds. L45F, identified early in the S1 evolution strategy, was not observed in strategy S2 until round 10 of affinity maturation. Other affinity maturation events in the S1 strategy, such as D100A and the longer CDR1 codon insertions, were not detected in the S2 strategy.

Sequence analysis identified a total of 18 enriched mutations within the HC that improved antigen-binding kinetics (Table 1 and Fig. 4B). Seven of the 18 enriched mutations were observed in both S1 and S2 strategies; five of the remaining mutations involved sequence changes at the same amino acid positions (e.g., D100A vs. D100G) and were also found in both S1 and S2 strategies. Eleven substitutions were cataloged early in the S1 strategy (Fig. S2), nine of which were identical or related to those observed in later rounds. Enriching mutations were also identified by HTP sequencing of the LC, one of which (A60T) was incorporated into the affinity matured antibodies APE925 and APE928.

Four independent AID-mediated codon insertions were observed in the maturing HC sequence near the junction of FW1 and CDR1 (Table 1), with insertions ranging from a single alanine to segments containing 7, 8, and 12 amino acids. All four insertions originate from local sequence duplications, three of the four contained secondary AID mutations in combination with the insertion, and each was associated with a significant improvement for binding affinity for the antigen (Fig. 4), all hallmarks of insertions observed in in vivo derived antibodies (29–31).

Mutations observed in both strategies not utilized by APE803 were recombined by overlap extension PCR in a combinatorial library, each HC/LC pair expressed, and characterized on Biacore utilizing direct capture of secreted antibodies (Fig. 4). Variants identified with dissociation constants in the low pM range were further characterized in Biacore experiments utilizing normalized concentrations of purified antibodies captured at low densities (R_max < 30), with APE925 possessing a K_D of 25 pM (Fig. 3C).

Antibodies spanning the S1 affinity maturation trajectory were characterized for their ability to antagonize binding of βNGF to TrkA, compete for binding to βNGF with known anti-βNGF antibodies, and inhibit βNGF dependent TrkA signaling in a rat pheochromocytoma-derived cell line (PC12). C10A-derived antibodies demonstrated a dose-dependent ability to inhibit binding of an anti-βNGF antibody (Tanezumab) to βNGF in a homogenous time-resolved fluorescence (HTRF) assay that correlated well with kinetic characterization of each clone (Fig. 3D; compare with Fig. 4). Antibodies APE925 and APE928 demonstrated a dose-dependent inhibition of TrkA-Fc binding to βNGF (Fig. 3E) and inhibited βNGF-induced TrkA signaling and activation of extracellular regulated kinase (ERK)1/2 in PC12 cells (Fig. 3F). These results demonstrate that the fully human antibodies generated from this in vitro mammalian cell affinity maturation strategy bind βNGF with low pM dissociation constants and exhibit potent and functional inhibition of βNGF binding to its cognate receptor.

Discussion

By combining the critical features of adaptive immunity, we have developed an in vitro system for generating potent, biologically active antibodies in a mammalian cell context. This approach makes possible both de novo discovery and maturation from a naïve antibody library and may also be used for maturation of preexisting antibodies. In vitro affinity maturation by SHM has been described in human B cell lymphoma lines (32), chicken DT40 cells (33, 34), CHO cells (4), and 18–81 cells (35). However, these cell lines are difficult to transfect at efficiencies suitable for use with diverse libraries. Nevertheless, the potential of utilizing SHM in vitro has been apparent since Cumbers et al. (7) were able to evolve the endogenous IgM in Ramos cells to recognize SA with an apparent affinity of 11 nM after 19 rounds of FACS sorting.

Overlapping but distinct sets of mutations were identified in two affinity maturation strategies (Table 1). Early or rare SHM events (e.g., L45F in strategy S1 vs. G44R in strategy S2) may direct the antibody to alternative evolutionary fates that are equally consistent with high-affinity binding and biological activity, hinting at the nondeterministic nature of SHM in vitro and in vivo. HTP sequence analysis of early evolving populations provides a powerful method for the identification of large, unbiased sets of enriched mutations. Many of the beneficial mutations observed by HTP sequencing were not identified by Sanger sequencing of small numbers of templates until later FACS rounds, or were never observed (e.g., V89L HC and A60T LC). Antagonistic pleiotropy has been shown to slow the compilation of advantageous mutations (36) within a protein by limiting the number of evolutionary routes available for maturation. Early combination of mutations identified by sequencing into arrayed libraries, paired with kinetic screening of secreted antibodies, is likely to further speed the identification of optimal antigen-binding solutions.

SHM events observed in our heterologous system mirror those observed during in vivo affinity maturation. Nucleotide transversions and transitions, double mutation events (Table 1 and Table S2), and the nonsynonymous mutations that result, suggested that AID activity alone is sufficient to generate functional amino acid substitutions (37). Mutation patterns and amino acid diversification produced by AID on HCs and LCs in this in vitro system have been examined in the absence of functional selection, and closely resemble that seen in antibody sequences that have undergone in vivo affinity maturation (Fig. S1).

Codon insertions and deletions, particularly within CDR regions, are a rare but important feature of in vivo SHM that significantly expands the sequence space that may be sampled during affinity maturation (30, 38, 39). It has been estimated that approximately 6.5% of native human antibodies contain insertions or deletions within the variable domains as a consequence of SHM (40), and a number of potential mechanisms for their generation via AID-mediated mutagenesis have been proposed (39, 41). The identification of multiple, localized codon insertions within CDR regions in this and other affinity maturation programs substantiates this as a robust feature of AID-directed SHM, and another aspect in which the in vivo process can be reproduced in vitro.

The use of AID-mediated mutagenesis and flow cytometry to evolve antibodies against fluorescent-labeled antigen(s) allows complex selection methodologies to be applied to select antibodies evolved toward multiple properties of interest. We utilized FACS selection in two dimensions to ensure that only well-expressed human IgGs with high binding affinity were selected. FACS-based selection may be carried out using multiple parameters with control of stringency used to exert selection pressure to drive maturation to a desired endpoint. The simultaneous surface display and secretion of full-length antibodies allows for the direct kinetic characterization of maturing cell clones or populations, enabling parallel screening for desired functional or physiochemical properties.

Combining key elements of adaptive immunity in a single in vitro system allows selection and maturation of antibodies to high-affinity, while avoiding the intrinsic limitations of the in vivo immune response. A library of rearranged immunoglobulins with germline V-gene segments was displayed on the surface of HEK293 cells as a means to mimic the initial recombined repertoire presented on the surface of B cells prior to affinity maturation. The combination of the use of high binding avidity to isolate initial binders, and AID-induced SHM enables selection and affinity maturation of high-affinity antibodies utilizing a minimum of mutations from the germline sequence and offers the flexibility to select for antibody characteristics, such as multiple antigen or epitope binding, high expression level in mammalian cells and stability not achievable using other antibody discovery systems.

Materials and Methods

Transfection, Stable Expression and Selection of HEK293 c18 Cells.

Stable episomal HEK-293 c18 cell lines expressing IgG HCs modified with a C-terminal transmembrane domain (28) or LCs, together with AID were generated by seeding a T75 culture flask with 3 × 10⁶ HEK293 c18 cells in 10 mL Dulbecco’s modified Eagle’s medium (DMEM) containing 10% FBS (Life Technologies). Plasmids were transfected using 500 μL OptiMEM (Life Technologies) and 20 μL HD-Fugene (Roche Diagnostics). Three days posttransfection, cell growth medium was exchanged with 10 mL DMEM containing 10% FBS, 50 μg/mL Geneticin, 10 μl/mL Antibiotic-Antimycotic Solution, 1.5 μg/mL puromycin, 15 μg/mL blasticidin, and/or 350 μg/mL hygromycin (all from Life Technologies), and the cells were incubated for approximately 4 weeks with periodic reseeding and exchange of the cell culture medium.

ABELmAb Library Assembly and Transfection.

CDR3 diversity was amplified from PMBCs from a pool of normal donors and inserted into nine HC and nine LC synthesized germline V regions (DNA2.0) selected based on their representative use in vivo. Constant domains for the IgHC-γ1, IgKC, and IgLC3 chains were also synthesized. A type I transmembrane and cytoplasmic region were added to the C-terminal end of the HC to enable presentation on the cell surface of HEK293 cells. The synthesized HC and LC constant region gene fragments were assembled with the respective V regions in episomal vectors (ABELmAb library). The ABELmAb library DNA was transfected as described above as separate IGHV sublibraries, for each of the nine germline IGHV and IGKV templates. A random sampling of the stably selected library by HTP provided a lower estimate of complexity of at least 50,000 unique HC sequences and at least 1,200 unique LC sequences, for a lower estimate of 6 × 10⁷ total diversity (25) of combinatorially expressed antibody sequences.

Target Cell Isolation Using Dynal Magnetic Biotin Binder Beads.

Nine HC sublibraries were transfected in combination with the pooled LC library into HEK293 c-18 cells and used to isolate de novo βNGF binders. Cells (1 × 10⁸ from each sublibrary) were subjected to four rounds of negative selection by incubating with 2 × 10⁸ biotin binder beads (Dynabeads® Biotin Binder; Life Technologies) for 0.5 h and retaining unbound cells. Antigen-coated beads were prepared by washing 2 × 10⁷ beads in PBS, 0.1% BSA (Sigma-Aldrich) and incubating with 600 ng biotinylated βNGF (50 μL PBS, 0.1% BSA) for 1 h at 4 °C with rotation. Beads were washed twice, negatively selected HEK293 library cells were added and incubated for 2 h at 4 °C with rotation. Positively selected HEK293 library cells were resuspended in growth medium and expanded for AID transfections and FACS selection.

Antigen and Antibody Expression and Purification.

Antibody variants were generated as full-length secreted IgG molecules lacking the transmembrane and cytoplasmic domains, and expressed transiently in HEK293 c-18 cells. Transfected cell supernatants were loaded on a protein A/G agarose resin (Thermo Scientific), washed with 6 column-volumes of PBS, pH 7.4, and eluted with 100 mM glycine, pH 3.0. Human βNGF-His used in the ABELmAb library discovery and affinity maturation was expressed transiently in HEK293 c-18 cells and purified using standard his-tag affinity purification methodologies. FITC-labeled antigens utilized for FACS were prepared using standard amine coupling chemistry.

FACS Selection with Antigen Avidity.

Binding analysis: Antibody transfected HEK293 cells (5 × 10⁵ in 0.5 mL PBS, 0.1% BSA) were incubated with various concentrations of βNGF-WFP-Myc for 0.5 h at 4 °C. Unlabeled Goat anti-Myc (AbCam) was added at a 2∶1 (antigen:anti-Myc antibody) molar ratio, and cells were incubated for 0.5 hr at 4 °C. FITC-AffiniPure Fab Fragment Goat anti-Human IgG (H + L) (Jackson ImmunoResearch) was added (1∶2,000) for 0.5 h at 4 °C. Cells were then pelleted and resuspended in 0.3 mL 0.2 μg/mL DAPI in PBS, 0.1% BSA (Sigma-Aldrich) and analyzed for fluorescence on a BD Influx cell sorter (BD Biosciences). FACS: Antibody transfected HEK293 cells (5 × 10⁷ in 20 mL PBS, 0.1% BSA) were incubated with 25 nM βNGF-WFP for 0.5 h at 4 °C. Goat anti-Myc andFITC-Goat anti-Human IgG were added as above to the cells. Stained cells were resuspended in 1.0 mL 0.2 μg/mL DAPI in PBS, 0.1% BSA and sorted for the strongest antigen-binding cells on a BD Influx cell sorter.