A Sortase A programmable phage display format for improved panning of Fab antibody libraries

Abstract

Phage display of combinatorial antibody libraries is a versatile tool in the field of antibody engineering, with diverse applications including monoclonal antibody (mAb) discovery, affinity maturation, and humanization. To improve the selection efficiency of antibody libraries, we developed a new phagemid display system that addresses the complication of bald phage propagation. The phagemid facilitates the biotinylation of Fab antibody fragments displayed on phage via Sortase A catalysis, and the subsequent enrichment of Fab-displaying phage during selections. In multiple contexts, this selection approach improved the enrichment of target- reactive mAbs by depleting background phage. Panels of cancer cell line-reactive mAbs with high diversity and specificity were isolated from a naive chimeric rabbit/human Fab library using this approach, highlighting its potential to accelerate antibody engineering efforts and to empower concerted antibody drug and target discovery.

Introduction

The selection of mAbs from combinatorial phage antibody libraries is a critical tool in the field of antibody engineering.^1–3 It enables the rapid identification of antibodies with favorable binding properties from vast antibody repertoires, in a process termed panning.⁴ The M13 bacteriophage is the most widely used phage for panning combinatorial antibody libraries, due to its malleable assembly process which permits incorporation of non-wild type capsid proteins.⁵ Typically, Fab or scFv antibody fragments are displayed as N-terminal fusions to either full- length or N-terminally truncated M13 gene III protein (plll), a minor coat protein responsible for viral entry into host bacteria.⁶ These antibody fragment-pIII gene products are encoded in a plasmid that is encapsulated in the phage particle during assembly and thus propagates selected antibodies. Phagemid display systems, in which the antibody fragment is encoded using a minimal phagemid vector, support robust library production and efficient phage packaging.⁷ Such systems depend on superinfection with helper phage, which supplies additional M13 gene products for phage production.^{8, 9}

A known limitation of many phagemid display systems is the production of bald phage that do not display antibody fragments, which can account for up to 99% of total phage in a given preparation.^{10, 11} This observation has been attributed to a superior incorporation of helper phage-supplied wild type pIII compared to pIII fusions during assembly, which leads to the predominant production of bald phage containing exclusively wild type pIII. Due to nonspecific phage adherence, bald phage can still be present in the panning eluent, and thus propagate irrelevant phagemids which complicate the selection process. A few approaches have been developed to address this problem using engineered helper phage constructs.^12–17 Here, we present an alternative solution to this problem for Fab-phage antibody libraries that is compatible with wild type pIII-supplying helper phage. The approach, termed Fab-phage biotinylation & capture (FBC), relies on site-specific incorporation of a biotin capture moiety onto Fab fragments displayed on phage, which facilitates the capture of Fab-displaying phage during panning. In FBC phagemid pC3Csort, a Sortase A recognition sequence is recombinantly fused to the C-terminus of the Fab light chain. Prior to panning, Sortase A ligation is harnessed to incorporate biotin onto the Fab fragment displayed on phage. After elution of target-adherent phage during panning, streptavidin-coated magnetic beads are used to selectively capture Fab- displaying phage. As a proof of concept, a naive chimeric rabbit/human Fab library was generated in pC3Csort and panned on intact cancer cell lines. In the context of this complex target, the FBC approach more efficiently enriched cell surface-reactive mAbs compared to a conventional whole-cell panning procedure. These results highlight the potential of FBC as an improved strategy for phage antibody library selections.

Results

Nonspecific phage adherence independent of Fab display

Nonspecific binding of phage to target substrates is a known phenomenon and impediment in the panning of phage antibody libraries.^18–20 This can be either antibody-dependent, via the display of nonspecific or polyspecific antibody fragments, or antibody-independent, via an inherent nonspecific adherence of M13 phage. To understand the relative contribution of the latter possibility, we performed single round panning experiments using phage clone ∆Fab, which does not display an antibody fragment (Suppl. Fig. 1), and a previously reported Fab- phage library.²¹ These two phage samples were individually panned on immobilized recombinant human ROR1 protein and the mammalian cell line 63–12(ROR1), which ectopically expresses the human transmembrane protein ROR1. No significant difference in phage output was observed between the two phage samples for either target substrate (Fig. 1), suggesting that a significant portion of phage are selected independent of antibody fragment display.

Figure 1. Phage adhere to target substrates independent of antibody fragment display.

Monoclonal phage ΔFab, which lacks a displayed antibody fragment, and a Fab-displaying phage library in phagemid pC3C, were subjected to single rounds of panning on purified human ROR1 antigen or 63–12(ROR1) cells. An unpaired two-tailed student’s t-test was used to determine statistical significance. Output data were normalized to an input of 10¹² cfu for each condition.

Nonspecific phage selection was more pronounced on whole cells and was unaffected by various blocking buffers in this setting (Suppl. Fig. 2). Given that bald phage are the primary component in phagemid-derived phage preparations,^{10, 11} their abundance in early round phagemid-derived panning outputs is expected. As bald phage cannot form immune complexes, their propagation creates a pervasive background component which ultimately reduces panning efficiency.

Sortase A-catalyzed biotinylation of Fabs displayed on phage

To address the aberrant propagation of bald phage, we developed phagemid pC3Csort, which enables site-specific incorporation of biotin onto Fab fragments displayed on phage. We reasoned that this display format could facilitate avidin-mediated enrichment of Fab-displaying phage during panning. To generate phagemid pC3Csort, a derivative of pC3C,²² the Sortase A recognition sequence LPETGG was fused to the C-terminus of the Fab light chain (Fig. 2A). Our aim was to use Sortase A, an enzyme that has been previously used to site-specifically label N- and C-termini of polypeptides and proteins,^23–25 to catalyze Fab-phage biotinylation via a transpeptidation reaction involving LPETGG and a Gly₃-biotin co-substrate. To examine the tolerability of this approach for Fab display and biotinylation, we prepared the previously reported anti-human ROR1 clones ERR1–324,²¹ henceforth referred to as 324, as monoclonal phage from pC3C and pC3Csort phagemids. The two phage preparations bound human ROR1 and anti-human light chain polyclonal antibodies to similar extents (Fig. 2B), suggesting that LPETGG incorporation does not interfere with Fab display or antigen binding. These phage were then subjected to Fab-phage biotinylation using an engineered Sortase A variant with enhanced catalytic activity,²⁶ and their reactivity towards streptavidin was assessed. The phage derived from pC3Csort, but not pC3C, had strong streptavidin reactivity, indicating that Fab- phage biotinylation is dependent on Fab fusion to LPETGG (Fig. 2C).

Figure 2. Fab-phage biotinylation of phage by Sortase A ligation.

(A) Fusion of Sortase A recognition sequence G₄SLPETGG to the C-terminus Fab light chain in pC3C yields phagemid pC3Csort. Superinfection of pC3Csort-bearing bacteria with helper phage (ΗΦ) generates phage that display Fab antibody fragment-G₄SLPETGG fusions. Fab-phage biotinylation of phage is carried out by Sortase A in the presence of a Gly₃-biotin co-substrate. (B) Monoclonal 324 phage derived from pC3C and pC3Csort were analyzed for reactivity towards human ROR1 and an anti-human κ light chain polyclonal antibody by ELISA. (C) Phage were then subjected to Fab-phage biotinylation using Sortase A ligation, and then analyzed for reactivity towards streptavidin and an anti-human κ light chain antibody by ELISA. For (B) and (C), 10¹¹ phage virions were analyzed per well. Error bars represent the standard deviation of technical triplicates from a single representative phage preparation.

Depletion of bald phage by Fab-phage biotinylation & capture

To assess whether the pC3Csort format reduces the propagation of bald phage during panning, we established a selection strategy termed Fab-phage biotinylation & capture (FBC) (Fig. 3A). This approach relies on avidin-mediated capture of Fab-biotinylated phage, in a process termed Fab-phage capture, after low pH elution of phage from target substrates. It requires the retainment of biotin on the Fab fragment after selection, and is thus incompatible with trypsin- based elution of phage from target substrates. We reason that this approach offers maximal selection advantage since it selects for both target engagement and Fab display in a single round of panning. FBC was tested using a reporter system in which 324 monoclonal phage was prepared from chloramphenicol resistant phagemids (CamR) pC3C-Cam and pC3Csort-Cam. These phage were then diluted in a previously described Fab-phage library in phagemid pC3C,²¹ which harbors the AmpR gene that confers carbenicillin resistance. The orthogonally resistant phagemids permitted quantification of enrichment ratios between Fab-biotinylated 324 (CamR) and non-biotinylated pC3C library (AmpR) phage during panning. We elected to use intact 63–12 and 63–12(ROR1) cells as target substrates to test this reporter system. Due to the high background binding of phage to cell surfaces (Fig. 1), panning against intact mammalian cells is notoriously challenging and is thus a stringent testing ground for the FBC approach. Inclusion of Fab-phage capture conferred a ~400-fold enrichment boost for Fab-biotinylated 324 phage when panned on 63–12(ROR1) cells (Fig. 3B). This enhancement was manifested as a reduced propagation of AmpR resistant phage (Suppl. Fig. 3A), indicating that Fab-phage capture depletes non-biotinylated phage. Inclusion of Fab-phage capture resulted in only a ~2- fold reduction in recovery of Fab-biotinylated 324 phage (Suppl. Fig. 3B), indicating that Fab- displaying phage are robustly recovered by FBC. 324 phage recovery by Fab-phage capture was dependent on Fab-phage biotinylation, its preparation from pC3Csort-Cam, and target cell ROR1 expression. Additionally, 324 phage recovery in the absence of Fab-phage capture was unaffected by both Fab-phage biotinylation and its preparation from pC3Csort-Cam, supporting the notion that Fab-phage biotinylation by Sortase A ligation does not impair antigen binding or phage infectivity. Together, these results highlight the potential of the FBC strategy to improve the panning efficiency of Fab-phage libraries.

Figure 3. Fab-phage biotinylation & capture (FBC) procedure improves the enrichment of a Fab recognizing human ROR1.

(A) Overview of FBC. Fab-phage biotinylation introduces biotin onto Fabs displayed on phage. After phage selection for target binding (depicted here as whole cells) and subsequent washing, elution yields a mixture of Fab-displaying and bald phage. Fab-phage capture using streptavidin-coated magnetic beads recovers Fab-displaying phage and enables removal of bald phage by washing. Captured Fab-displaying phage can then be released by trypsin digestion and re-amplified for subsequent panning rounds. (B) 324 monoclonal phage was prepared in modified pC3C or pC3Csort phagemids bearing the CamR resistance gene, with and without Fab-phage biotinylation. These phage were diluted into a non- biotinylated Fab-phage library (AmpR) by a factor of ~10⁴. A total of ~2.5e12 cfu phage were used as input for single panning rounds on 63–12(ROR1) or 63–12 cells. Enrichment of 324 monoclonal phage before and after Fab-phage capture was determined by colony-forming assays. ND denotes conditions below the detection limit. Asterisk denotes condition in which 2 of 3 data points were ND. Statistical significance was determined using an unpaired two-tailed student’s t-test.

Examination of a Fab library in pC3Csort

To test the suitability of the FBC approach for enriching Fab-displaying phage in the setting of an antibody library, a naive chimeric rabbit/human Fab library consisting of 4.4 × 10⁸ individual rabbit mAbs was prepared in phagemid pC3Csort. Preliminary clonal Fab cassette PCR and DNA fingerprinting analyses indicated that the library lacked major production biases and was composed of mAbs of diverse sequence (Suppl. Fig. 4). After Fab-phage biotinylation, both κ and λ sub-libraries had strong streptavidin reactivity (Suppl. Fig. 5). To quantify the extent of biotinylation, the sub-libraries were directly subjected to Fab-phage capture (Fig. 4A). Fab-phage capture recovered ~1% of input phage previously subjected to Fab-phage biotinylation. Phage recovery was ~1000-fold lower in the absence of Fab-phage biotinylation, suggesting that Fab-phage capture stringently depletes non-biotinylated phage. To quantify the efficiency of the FBC approach with respect to Fab-displaying phage, the library was pre-selected for Fab display using polyclonal anti-light chain antibodies, and then subjected to Fab-phage capture. Approximately 20% of Fab-displaying phage were recovered when Fab-phage biotinylation was implemented, while only ~0.01% were recovered in its absence (Fig. 4B). As this measurement relies on Fab-phage capture, this 20% efficiency represents the lower bound of Fab-phage biotin incorporation. These results suggest that both the Fab-phage biotinylation and Fab-phage capture steps can be implemented in the context of a Fab-phage library to selectively deplete bald phage.

Figure 4. Analysis of pC3Csort library.

(A) pC3Csort κ and λ sub-libraries in the presence or absence of Fab-phage biotinylation were subjected to Fab-phage capture. Phage recovery was then analyzed by colony-forming assays. The phage input for each capture replicate was 10⁸ cfu, and an unpaired two-tailed student’s t-test was used to determine statistical significance. (B) Fab-displaying phage from the pC3Csort library were pre-selected by panning against anti human light chain antibodies. Phage titer was measured before and after Fab-phage capture to determine the fraction of Fab-displaying phage that are recoverable by FBC. Input library phage for each condition was 10⁹ cfu.

To ensure phagemid pC3Csort enables robust production of phage particles, colonyforming assays were conducted on the pC3Csort library and a comparable naive chimeric rabbit/human Fab library in phagemid pC3C.²¹ No substantial difference between the two libraries was observed (Suppl. Fig. 6), suggesting that the LPETGG motif does not impede phage production. To assess the specificity of Fab-phage biotinylation, pC3Csort κ and λ sublibraries were analyzed by immunoblotting using ExtrAvidin-Peroxidase. The major signal for both sub-libraries under non-reducing conditions occurred at ~70 kDa (Fig. 5), corresponding to full Fab-ΔρΙll fragments with proper interchain disulfide bond formation. Under reducing conditions, the primary signal was present at ~25 kDa, corresponding to a free light chain. No ExtrAvidin-Peroxidase signals were detectable in the absence of Fab-phage biotinylation. These data indicate that Fab-phage biotinylation, in the context of this library, is site-specific and enables proper Fab formation.

Figure 5. Fab-phage biotinylation by Sortase A is restricted to the Fab light chain.

pC3Csort sub-libraries were subjected to Fab-phage biotinylation, and then analyzed by immunoblotting using an HRP-conjugated anti-human F(ab’)₂ antibody (top), and then ExtrAvidin-Peroxidase (bottom). 10¹¹ phage virions were analyzed per condition, as determined by spectrophotometry.

Improved selection efficiency by FBC panning

Next, we sought to determine whether the FBC approach could improve the enrichment of target-reactive mAbs during panning. The Fab library in pC3Csort was thus subjected to panning experiments on intact, live H929, BxPC-3, and MDA-MB-468 cells, representing human multiple myeloma, pancreatic adenocarcinoma, and breast adenocarcinoma cell lines, respectively. For all three experiments, the first round of panning was conducted in a conventional fashion, without FBC. The three phage outputs from the first round were then reamplified and each was split into two parts. One duplicate phage sample for each experiment was re-panned against the appropriate target cell line in a conventional fashion, as in the first round. The other duplicate was panned using the FBC approach (Fig. 3A). Due to a drastic reduction in library diversity after one panning round, this split procedure allowed a more direct comparison between conventional and FBC panning in the second panning round. Selections conducted with FBC yielded output phage titers ~100-fold lower than selections conducted without (Table 1). This is consistent with the finding that Fab-phage capture recovers ~1% of Fab-biotinylated phage input (Fig. 4B), and highlights the stability of Sortase A-catalyzed biotin incorporation.

Table 1.

Metrics of pC3Csort whole-cell panning experiments.

Target cells	R2 panning approach	R2 output (cfu)	R2 mAb hitsa	R2 unique mAb hitsb	R2 output Fab expressionc
H929	FBC	8 × 104	16/94 (17%)	13	56/94 (60%)
BxPC-3	FBC	8 × 104	9/94 (10%)	7	65/94 (69%)
MDA-MB-468	FBC	2 × 105	10/94 (11%)	5	67/94 (71%)
H929	Conventional	2 × 107	1/94 (1%)	1	21/94 (22%)
BxPC-3	Conventional	8 × 106	0/94 (0%)	0	40/94 (43%)
MDA-MB-468	Conventional	1 × 107	0/94 (0%)	0	32/94 (34%)

^aDetermined by flow cytometry

^bDetermined by DNA sequencing of VH and VL domains

^cDetermined by ELISA

To assess whether the FBC approach enables the isolation of surface-reactive mAbs, clonal soluble Fab-ΔρΙll fragments were screened against target cells. For all three target cell substrates, the FBC second round panning outputs contained a higher frequency of positive mAbs compared to the conventional panning outputs (Table 1). This implies that the FBC approach generated a more efficient enrichment of target-reactive mAbs against these three target cell lines. Consistent with the assertion that FBC selects for clones that produce Fab- displaying phage, the frequency of output clones that express soluble Fab was higher for FBC panning compared to conventional panning.

We next assessed the diversity of output clones from these experiments, as obtaining mAbs of variable specificity is a highly desirable feature of antibody selections. Clonal DNA sequencing analyses of heavy and light chain variable domains (VH and VL) indicated that the three panels of hit mAbs from FBC panning were each comprised of multiple unique mAbs (Suppl. Tables 1 and 2). Further, this collection of mAbs possessed heavy chain CDR3 lengths and IGHV germline deviation frequencies (data not shown) consistent with previous findings.^{21, 27} To further assess the diversity of unique hits enriched by FBC, a panel of H929-reactive mAbs were expressed and purified as soluble Fabs, and then subjected to biotinylation by Sortase A ligation. SDS-PAGE analysis indicated that the Fabs retained proper interchain disulfide bond formation after biotinylation (Suppl. Fig. 7). The specificity of these biotinylated Fabs was then assessed against various cell lines of human origin by flow cytometry using streptavidin detection (Fig. 6). All Fabs generated positive signals against H929 cells, confirming that they are biotinylated and retain antigen reactivity. Four unique reactivity patterns were observed in this expression profiling experiment. Further epitope binning analysis indicated that this panel recognized five unique epitopes (Suppl. Fig. 8). These findings support the assertion that diverse panels of target-reactive mAbs can be enriched using the FBC approach. Collectively, this work highlights the utility of the FBC approach in the selection of phage antibody libraries.

Figure 6. Expression Profiling of Fab panel isolated by FBC panning.

The binding of biotinylated H929-reactive Fabs against various cell lines of human origin was examined by flow cytometry, using streptavidin-PE for detection. Dashed lines indicate peaks of histograms in which no Fab was added.

Discussion

Nonspecific adherence to target substrates is an intrinsic property of phage particles, which can complicate the selection efficiency of phage libraries via the propagation of irrelevant clones. This phenomenon is exacerbated in antibody selections using phagemid display systems, as the majority of produced phage lack proper display of the selectable fusion protein. In this study, we present the FBC panning approach to address this impediment. In FBC phagemid pC3Csort, the Sortase A motif LPETGG is fused to the C-terminus of the Fab light chain. This enables efficient incorporation of biotin onto Fab fragments displayed on phage by Sortase A ligation, which facilitates the depletion of bald phage during the panning process. We have shown that both the biotinylation and capture components of FBC can be robustly applied to a naive Fab-phage library. We also showed this approach enables a more efficient enrichment of cell-surface reactive mAbs from this library compared to conventional panning.

Several approaches have been previously developed to address the aberrant propagation of bald phage during panning. For example, modified helper phage constructs can be used to render bald phage non-infectious by compromising the functionality of helper phage- supplied pIII.^{12, 13} Additional engineered helper phage constructs have been developed with reduced or ablated pIII supply, which enforces oligovalent antibody fragment display.^14–17 Phage generated from the above approaches depend on phagemid-derived antibody fragment-pIII fusion protein for bacterial host infectivity. In contrast, FBC relies on the capture of Fab- displaying phage in the panning eluent for bald phage depletion, which enables the use of N- terminally truncated pIII as a fusion partner for Fab display and helper phage-supplied wild type pIII for bacterial host infectivity. This distinction offers multiple benefits. First, pC3Csort-bearing F+ bacterial hosts are not susceptible to superinfection resistance mediated by wild type pIII expression.²⁸ pC3Csort-derived phage particles are engendered with redundant copies of wild type plll, which is likely beneficial for phage packaging and infectivity. Moreover, it allows the use of robust helper phage constructs such as VCSM13, thus obviating the need of plll complementing components for engineered helper phage production. The FBC approach also retains monovalent display of Fab fragments, which is advantageous for identifying high affinity mAbs via panning.^29,30 Although it may be possible to utilize alternative affinity-based reagents for Fab-phage capture, such as anti-Fab antibodies, we reasoned that the virtually irreversible streptavidin-biotin interaction (K_d ≈ 10⁻¹⁵ M) would enable a superb distinction between Fab- displaying and bald phage. This assertion is supported by the finding that Fab-biotinylated phage have a ~1,000-fold recovery advantage compared to non-biotinylated phage using streptavidin-coated magnetic beads. After examining several suboptimal approaches for Fab- phage biotinylation at the C-terminus of the Fab light chain, including BirA-mediated biotin ligation, the Sortase A-catalyzed approach presented here provided sufficient biotin incorporation. In the context of a diverse Fab-phage library, ~1% of all phage and ~20% of Fab- displaying phage were recoverable via the FBC approach, which together suggest library Fab display levels of ~5%. Given typical panning inputs of ~10¹³ phage, the FBC approach thus affords a selectable diversity of ~10¹¹, which is suitable for panning most large combinatorial antibody libraries. It is possible that the ~20% Fab-displaying phage recovery can be increased by further optimization of this approach. Nonetheless, this selection inefficiency is far outweighed by the selection advantage conferred by a ~ 10⁵-fold depletion of bald phage. In support of this notion, the FBC process enabled the isolation of diverse panels of cell surface- reactive mAbs after its implementation in the 2^nd panning round. In contrast, conventional antibody selections from similar libraries typically require 3–4 selection rounds to yield hit mAbs against pre-defined, purified antigens,²¹ and whole cells.^31,32 By reducing the number of panning rounds required to obtain hit mAbs, the FBC approach should minimize the outgrowth of contaminant clones.^33,34 It may also aid in the identification of low-abundance hits which are out competed over multiple cycles of phage selection and re-amplification.³⁵ Moreover, compression of library diversity will likely be advantageous for antibody discovery efforts involving next- generation sequencing of Fab-phage libraries.³⁶ Given the ubiquitous nature of antibody- independent phage adherence, the FBC approach will likely afford similar enhancements in antibody selection performance against a variety of target substrates, including pre-defined purified antigens, and perhaps even more challenging targets such as tissue sections.^{37, 38} Further optimization of the FBC approach may lead to enhanced recovery of Fab-displaying phage to greater than 20%, which would make its implementation in the first round of selection more attractive.

We also found that soluble, purified Fabs isolated from FBC panning could be biotinylated using Sortase A. As Sortase A ligation is compatible with a wide range of molecules,³⁹ this approach may support facile payload incorporation for streamlined assessments of Fabs in downstream diagnostic and therapeutic applications.^40–43

In this study, the FBC approach was examined using a naive chimeric rabbit/human Fab library. Given that the LPETGG peptide is fused to human light chain variable domains in pC3Csort, this approach should at least be compatible with fully human Fab libraries, as well as chimeric Fab libraries using human constant domains. As presented here, the FBC approach has no precisely analogous application to the panning of single chain libraries, including scFv antibody fragments or non-immunoglobulin scaffolds,⁴⁴ as such proteins lack free C-termini for LPETGG fusion when displayed on phage. However, N-terminal labeling of phage-displayed peptides at the plll site using Sortase A has been established,⁴⁵ which could be potentially utilized for Sortase A-catalyzed biotinylation of single chain protein phage libraries.

Collectively, the findings presented in this study suggest the FBC approach enhances the enrichment of target-reactive mAbs from Fab-phage antibody libraries. As such selections are staples in a variety of applications, including de novo generation and in vitro evolution of mAbs as well as cell surface antigen discovery, this approach is anticipated to have broad utility in the field of antibody engineering.

Materials and Methods

Cell lines

H929, BxPC-3, MDA-MB-468, U266, RPMI-8226, Jurkat, HBL-2, and Kasumi-1 cells were purchased from ATCC. Epithelial cell lines BxPC-3 and MDA-MB-468 were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) containing 10% (v/v) fetal bovine serum (FBS; Atlas Biologicals), 1× penicillin-streptomycin (PS), and 1× GlutaMAX. H929, U266, RPMI-8226, HBL-2, and Jurkat cell lines were maintained in Roswell Park Memorial Institute (RPMI) 1640 Medium supplemented with 10% (v/v) FBS and 1× PS. Kasumi-1 cells were maintained in RPMI 1640 Medium supplemented with 20% (v/v) FBS and 1× PS. The mouse RAG2^−/− pre-B-cell line 63–12 ectopically expressing human ROR1 was described previously.²¹ Parental 63–12⁴⁶ and 63–12(ROR1) cells were cultured in Iscove’s Modified Dulbecco’s Medium (IMDM) containing 10% (v/v) FBS, 0.1% (v/v) β-mercaptoethanol (Sigma-Aldrich), and 1× PS. Adherent BxPC-3 and MDA-MB-468 cells were passaged using TrypLE Express and harvested under subconfluent growth conditions using Stem Pro Accutase Cell Dissociation Reagent. All cells were washed twice with Dulbecco’s Phosphate-Buffered Saline (DPBS) prior to panning and flow cytometry experiments. Unless noted otherwise, all above reagents were from Thermo Fisher Scientific.

Phagemid cloning

pC3Csort phagemids were generated from ERR1–324 (k) and ERR1-TOP54 (λ) encoding pC3C phagemids²¹ by two fragment overlap extension PCR using Phusion High-Fidelity DNA Polymerase (New England Biolabs). Rabbit VL/human CL/sort fragments were amplified using universal sense primer C-5’SFIVL and antisense primers sortase-kappa (5’- GCCT CCT GT CTCT GGCAAGCT ACCCCCTCCACCACACTCT CCCCT GTTGAAGC-3’) for ERR1–324, or sortase-lambda (5’- GCCT CCT GT CTCT GGCAAGCT ACCCCCTCCACCT GAACATT CT GT AGGGGCCA-3’) for ERR1-TOP54. Sort/pelB/rabbit VH fragments were generated from both ERR1–324 and ERR1- TOP54 encoding phagemids using sense primer SORT VH (5’-GCCAGAGACAGGAGGCT AATT CT AGAT AATT AATT AGGAGGAATTT AAAAT G-3’) and antisense primer C-3’sfivh. The sequences of C-5’SFIVL and C-3’sfivh have been reported.⁴⁷ Appropriate ERR1–324 and ERR1-TOP54 encoding rabbit VL/human CL/sort and sort/pel B/rabbit VH fragments were then fused in a second PCR step using flanking C-5’SFIVL and C-3’sfivh primers. These rabbit VL/human CK/rabbit VH cassettes were then digested with Sfil (New England Biolabs), ligated into Sfil-digested pC3C using T4 DNA Ligase (New England Biolabs), and transformed into ER2738 cells (New England Biolabs) using a classical calcium chloride method.⁴⁸ Cognate pC3Csort-Cam phagemids were generated by transferring pC3Csort VL/human CK/rabbit VH cassettes into pC3C-Cam⁴⁹ by asymmetric Sfil cloning. ∆Fab phagemid was generated by introducing an ochre stop codon (TAA) at the first residue of the VH domain of ERR1–324 in pC3C by site-directed mutagenesis. The sequences of phagemids were confirmed by DNA sequencing of both heavy and light chain encoding fragments.

Monoclonal phage preparation

Phagemid transformed ER2738 bacteria were expanded from glycerol stocks in 100 mL super broth supplemented with 10 µg/mL tetracycline (Sigma-Aldrich) (SBT) supplemented with either 100 µg/mL carbenicillin (Sigma-Aldrich) or 25 µg/mL chloramphenicol (Thermo Fisher Scientific) at 37° C with shaking at 250 rpm until OD_6oo=0.5, and then ~10¹² plaque forming-units VCSM13 helper phage (Agilent) was added. After 1 h of growth at 37° C with shaking, 140 pL of 50 pg/pL kanamycin sulfate (Sigma-Aldrich) was added, and cultures proceeded with phage production overnight at 30° C with shaking at 250 rpm. Phage harvesting proceeded as previously described,⁴⁷ with modifications. Precipitated phage pellets from 100 ml_ overnight cultures were first suspended in 4 mL TBS buffer (50 mM Tris-HCI, 0.15 M NaCI, pH 7.6) and bacterial debris cleared by centrifugation at 15,000 x g for 5 min. Phage containing supernatants were passed through 0.45-pm filters, and then to each was added a 5× PEG/NaCI solution containing 20% (w/v) PEG-8000 (Sigma-Aldrich) and 15% (w/v) NaCI (Sigma-Aldrich) to a concentration of 1×. Phage were then precipitated on ice for 15 min and centrifuged 15,000 x g for 5 min to collect phage. After decanting supernatant, phage pellets were washed once with 1 mL TBS and then resuspended in 200 µL TBS.

Quantification of phage titer

For panning experiments, input and output phage titer was determined using colony-forming assays essentially as previously described.⁵⁰ Phage virion input for ELISA and immunoblot analyses was determined spectrophotometrically using an established procedure.⁵¹

Recombinant human ROR1 protein

Recombinant human ROR1 containing an AviTag (GLNDIFEAQKIEWHE) for biotinylation and His₆ tag for purification was expressed, purified, and biotinylated as previously described.²¹

Panning of monoclonal phage and pC3C library

For analysis of antibody-independent phage adherence, panning against immobilized human ROR1 antigen and 63–12(ROR1) cells using a Fab-phage library in pC3C²¹ and ΔFab monoclonal phage was conducted using a previously described protocol.⁵⁰ This procedure was also implemented against 63–12(ROR1) cells using BSA-containing selection buffer and a milk- based selection buffer consisting of 3% (w/v) skim milk, 0.025% (w/v) NaN₃ and 1 mM EDTA in PBS. For examination of FBC method using orthogonally resistant reporter system, monoclonal CamR-resistant 324 phage prepared from pC3Csort-Cam or pC3C-Cam were subjected to Fab- phage biotinylation, then diluted in a Fab-phage library in pC3C at a ratio of 1:10⁴. These mixtures were then subjected to a single round of whole-cell panning essentially as previously described.⁵⁰ The primary modification was elution of phage from cell surfaces, which involved low pH treatment, followed by neutralization as described in the Whole-cell panning using pC3Csort library methods section. Half of each phage eluent were then subjected to colony-forming assays on LB agar plates containing either 100 µg/mL carbenicillin or 25 µg/mL chloramphenicol. The other half was subjected to Fab-phage capture and then analyzed by colony-forming assays on LB agar plates containing carbenicillin or chloramphenicol.

Production of a naive chimeric rabbit/human Fab-phage library in pC3Csort

The pC3Csort phage antibody library was generated using established procedures.²¹ The primary modification was the incorporation of the G₄SLPETGG peptide fused to the C-terminus of the Fab light chain. This was accomplished by PCR amplifying human CKSort-pelB and human Cλsort-pelB fragments from pC3Csort phagemids encoding ERR1–324 and ERR1- TOP54, respectively, as described previously.⁵² A modified pC3C-derived phagemid was used as a destination for library ligation by asymmetric Sfil cloning. The G₄SLPETGG peptide sequence was absent in the Fab expression cassette of this phagemid to prevent selection of the encoded Fab during FBC panning. This phagemid also possessed the 34 nucleotide BBa_B1002 transcriptional terminator (iGEM Parts Collection) positioned upstream of the Lac promoter to minimize Fab expression in the presence of glucose.⁵³ The κ and λ sub-libraries were comprised of 1.7 × 10⁸ and 2.7 × 10⁸ independent transformants, respectively.

pC3Csort library re-amplification

The starting pC3Csort library was re-amplified from phage stock solutions using an established procedure.⁴⁷ Re-amplification of 1^st round output phage and 1^st and 2^nd round output phagemid harvesting were conducted using a modified procedure. Phagemid-bearing ER2738 colonies grown overnight on 1^st and 2^nd round panning output agar plates (see Whole-cell panning using pC3Csort library) were harvested using 10 mL SBT media and a cell spreader (Thermo Fisher Scientific). For 1^st round output phage re-amplification, a portion of these bacterial stocks (~2 mL) was diluted in 100 mL SBT media supplemented with 100 µg/mL carbenicillin at an OD₆₀₀=0.25. Re-amplification cultures grew at 37° C for 30 min with shaking, and then ~10¹² pfu VCSM13 helper phage (Agilent) was added. After 1 h growth at 37° C with shaking, 140 µL of 50 µg/mL kanamycin sulfate was added to re-amplification cultures. Cultures proceeded with phage production overnight at 30° C with shaking. Remaining bacterial stocks from panning outputs were supplemented with 10% glycerol (v/v) (Sigma-Aldrich) and stored at −80° C. The re-amplified pC3Csort starting library was harvested by standard procedures⁴⁷ and resuspended in selection buffer containing 3% (w/v) Bovine Serum Albumin (BSA), 0.025% (w/v) NaN₃, and 1 mM EDTA in PBS. Re-amplified 1^st round output phage were harvested and then subjected to tandem PEG/NaCI precipitation and resuspension in TBS as described in Monoclonal phage preparation.

Whole-cell panning using pC3Csort library

10–30 million target mammalian cells and input phage samples were blocked separately in 0.5–1 mL selection buffer for 1 h on a tumbler at 4° C in Low-Retention Microcentrifuge Tubes (Thermo Fisher Scientific). Mammalian cells were pelleted, supernatant decanted, and then cells resuspended in phage containing solution. Selections proceeded on a tumbler at 4° C for 2 h, and then cells were washed 2× with selection buffer. For each washing step, cells were first resuspended in 0.5 mL selection buffer, and then incubated on tumbler at 4° C for 10 min. 0.5 mL selection buffer was added to tubes, cells were pelleted by centrifugation, and then the supernatant was decanted. Prior to pelleting cells in the second wash, resuspended cells were transferred to fresh Low-Retention Microcentrifuge Tubes. Cell-adherent phage were eluted by adding 0.2 mL of 100 mM glycine-HCI (pH 2.2), followed by a 10 min incubation at room temperature on a tumbler. Cells were pelleted at 15,000 x g for 1 min, and then the phage containing supernatants transferred to fresh tube containing 100 µL of 1 M Tris-HCL (pH 8.0). The cell pellet was then subjected to a second elution following the same procedure as above, and the supernatants from both elutions were pooled. These neutralized eluents were then again centrifuged at 15,000 x g for 1 min to remove remaining cellular debris. Phage containing supernatants were then either propagated in 0.5 mL mid-log phase ER2738 bacteria or subjected to Fab-phage capture. For conventional (non-FBC) selections, phage-infected ER2738 bacteria were added to 100 mm x 15 mm Square Petri Dishes (Simport Scientific) containing Lysogeny Broth (LB) with 1.5% (w/v) agar supplemented with 100 µg/mL carbenicillin and 1% (w/v) D-glucose (Sigma-Aldrich). Plates were then grown overnight at 37° C.

Recombinant Sortase A expression and purification

An expression cassette encoding residues 60–206 of a previously described variant (clone 8.3)²⁶ of Sortase A from Staphylococcus aureus was synthesized (Integrated DNA Technologies) and transferred into the pET28a vector using the Ndel and BamHI restriction sites. N-terminally His6- tagged Sortase A was then cytoplasmically expressed from this plasmid by induction with isopropyl β-D-l-thiogalactopyranoside (Sigma-Aldrich) in BL21 bacterial cells. Bacteria were lysed using a Nano DeBee Homogenizer (BEE International), and then cleared lysates were purified by IMAC using a 1-mL HisTrap HP column (GE Healthcare) per the manufacturer’s protocol. Eluted Sortase A was then concentrated and buffer exchanged to TBS using 15-mL Amicon Ultra Centrifugal Filter Devices equipped with 10-kDa MWCO membranes (EMD Millipore). Purified Sortase A was analyzed by SDS-PAGE, and yield calculated using absorbance at 280 nm.

Fab-phage biotinylation

Concentrated phage preparations (~10¹¹ cfu/µL) in TBS were subjected to Sortase A-mediated biotinylation using 10 μΜ Sortase A, 250 pM Gly₃-biotin, and 10 mM CaCI_2. Custom synthesis of the Gly₃-biotin peptide was carried out by LifeTein. After incubating for 15 min at 37° C, biotinylated phage were diluted to 800 µL and 200 µL 5× PEG/NaCI was added. Tubes were placed on ice for 15 min to precipitate the phage. Phage were then pelleted by centrifugation at 15,000 x g for 5 minutes. Phage pellets were then washed once with 1 mL TBS. Phage for panning were then resuspended in 0.5 mL selection buffer and blocked for 1 hour at 4° C prior to addition to whole cells. Phage pellets for ELISA, immunoblot, or direct Fab-phage capture analyses were resuspended in 800 µL TBS, subjected to another cycle of precipitation/pelleting/washing, and then finally resuspended in TBS.

Fab-phage capture

To neutralized whole-cell panning eluents containing phage previously subjected to Fab-phage biotinylation (10⁶−10⁸ cfu) or unselected pC3Csort κ or λ sub-libraries diluted in 500 µL selection buffer was added 10 µL Dynabeads MyOne Streptavidin C1 (Invitrogen). Phage/bead mixtures were incubated at room temperature on a tumbler for 15 min. They were then placed on a Magnetic Particle Concentrator (Thermo Fisher Scientific) for at least 1 min to isolate beads and adherent phage. Supernatants were removed using a pipette or vacuum line, and beads were then washed 3× with 1 mL PBST. Each wash involved vigorous vortexing, bead isolation on the Magnetic Particle Concentrator, and supernatant removal. After a final wash with 1 mL PBS, phage were eluted from beads using 50 µL 10 mg/mL trypsin (Sigma-Aldrich) in PBS for 30 min at 37°C. Trypsinized phage eluents were then used to infect 0.95 mL of mid-log phase ER2738 bacterial hosts. After a 30 min incubation at room temperature, colony-forming assays were carried out using LB agar plates containing 100 pg/mL carbenicillin or 25 pg/µL chloramphenicol, as appropriate. Propagation of phagemids was carried out by plating remaining phagemid-infected bacteria as described in the Whole-cell panning using pC3Csort library methods section.

Analysis of FBC efficiency for Fab-displaying phage

Fab-displaying phage were pre-selected by panning 10⁹ cfu phage from the naïve chimeric rabbit/human Fab-phage library in pC3Csort (+/− Fab-phage biotinylation) against a 1:1 mixture of goat anti-human κ light chain and goat anti-human λ light chain antibodies (Southern Biotech) using an established procedure for immobilized panning on ELISA plates.⁵⁰ Phage elution was modified to a low pH treatment, followed by neutralization. Two equal parts (1/5^th) of the phage eluent were either directly analyzed by colony-forming assays or subjected to Fab-phage capture and then colony-forming assays.

Identification of mAb hits and unique mAb hits from whole-cell panning outputs

Phagemid bearing ER2738 colonies from panning outputs were transferred to individual wells in a 2-mL deep 96-well plate (USA Scientific) each containing 400 µL ZYM-5052 autoinduction media⁵⁴ supplemented with 10 µg/mL tetracycline and 100 pg/mL carbenicillin. After overnight growth at 30° C and 250 rpm, periplasmic extracts were prepared for each well following a previously described method⁵⁵ using 50 µL TSE buffer (200 mM Tris-HCI, pH 8.0, 500 mM sucrose, 1 mM EDTA). 50 µL FACS buffer (0.5% (w/v) BSA + 0.1% (w/v) NaN₃ in PBS) was added to each well after TSE extraction. Bacterial debris was then cleared by centrifugation, yielding periplasmic extracts containing soluble Fab-ΔρΙll fragments. 25 µL from each well was then used to stain 25,000 target cells for 30 min on ice. After two washes with 100 μL FACS buffer, cells were then stained with 25 μL of a 100-fold dilution of PE-conjugated goat antihuman IgG, F(ab’)₂ fragment specific antibody (Jackson ImmunoResearch #109–116-097) in FACS buffer. After 15 min incubation on ice, cells were washed twice with 100 μL FACS buffer and then resuspended in 25 μL FACS buffer and analyzed on an Accuri C6 flow cytometer (BD Biosciences) equipped with a Hypercyt autosampler (Intellicyt). Identification of hits was determined with FlowJo10 software (Tree Star) by a population comparison analysis between screened periplasmic extracts and a control well lacking soluble Fab-ΔρΙll. A cutoff of χ² = 5 was used as the lower bound for determining positive clones. Hit clones from flow cytometry-based screening were analyzed by DNA fingerprinting using Alul restriction enzyme and their VH and VL-encoding sequences were determined using established procedures.⁴⁷

Immunoblot

pC3Csort phage sub-libraries previously subjected to Fab-phage biotinylation were precipitated again using 800 μL and 200 μL 5× PEG/NaCI. After pelleting phage, supernatant was decanted and phage pellet resuspended in TBS. 10¹¹ phage virions per lane were mixed with 4× NuPage LDS Sample Loading Buffer (Thermo Fisher Scientific) and loaded onto a 4–12% NuPAGE Bis- Tris Gel (Thermo Fisher Scientific). Prior to this, reduced phage samples were mixed with 4% (v/v) β-mercaptoethanol and heated at 95°C for 5 min. After electrophoresis, resolved phage proteins were transferred to a PVDF membrane (EMD Millipore). The membrane was then blocked with a solution of 1× Western Blocking Reagent (Roche) in TBS, and then incubated with an HRP-conjugated goat anti-human IgG, F(ab’)₂ fragment specific antibody (Jackson ImmunoResearch #109–036-097) diluted 10,000-fold in 0.5× Western Blocking Reagent in TBS. After washing with TBST, immunoreactive bands were developed with Amersham ECL Prime Western Blotting Detection Reagent (GE Healthcare) and then visualized using CL-XPosure Film (Thermo Fisher Scientific). The membrane was stripped with Restore Western Blot Stripping Buffer (Thermo Fisher Scientific) per the manufacturer’s protocol, and then probed using a solution of ExtrAvidin-Peroxidase (Sigma-Aldrich) diluted 10,000-fold in 0.5× Western Blocking Reagent in TBS. After washing with TBST, immunoreactive bands were developed and visualized as above.

SDS-PAGE

2 µg of protein was mixed with 4× NuPage LDS Sample Loading Buffer and loaded onto a 4–12% NuPAGE Bis-Tris Gel. Reduced protein samples were treated with 4% (v/v) β- mercaptoethanol and heated at 95°C for 5 min prior to gel loading. After electrophoresis, gels were stained with PageBlue Protein Staining Solution (Thermo Fisher Scientific) and then de- stained per the manufacturer’s procedure.

ELISA

25 or 50 ng of goat anti-human light chain antibodies, Streptavidin (Sigma-Aldrich), or Rat-anti- HA mAb 3F10 (Roche) were coated on a 96-well half-area ELISA plate (Costar) in 25 pL PBS. For coating of human ROR1, wells coated with streptavidin were washed once with 150 pL H₂0 and then 25 ng of recombinant biotinylated human ROR1 protein was added in 25 µL PBS. All coating steps proceeded for 1 h incubation at 37° C or overnight at 4° C. Wells were then blocked with 3% (w/v) BSA in PBS. For analysis of Fab-ΔρΙll protein expression, 25 µL clonal Fab-ΔρΙII TSE extracts diluted three-fold in PBS were then added to wells. For analysis of pC3Csort sub-library phage samples, serial dilutions were prepared with 1% (w/v) BSA in PBS, and 25 pL added to ELISA plate wells. After 1 h incubation 37° C, samples were removed and wells washed 3× with either 150 pL H₂0 (for Fab-ΔρllΙ TSE extracts) or 150 µL PBST (for phage sub-library samples). For analysis of Fab-ΔρΙll protein expression, 25 µL of a 1,000-fold dilution of Rat-anti-HA mAb 3F10 conjugated to HRP (Roche) was then added to wells. For analysis of phage samples, 25 µL of a 1,000-fold dilution of a mouse anti-M13 conjugated to HRP (GE Healthcare) was then added to wells. After 1 h incubation at 37°C, wells were then washed 3× with 150 μL H20, and then 50 μL ABTS One Component HRP Microwell Substrate (SurModics) added to each well. After appropriate development, the absorbance at 405 nm was read for each well using a microplate reader (Molecular Devices).

Expression, purification, and Sortase A-catalyzed biotinylation of chimeric rabbit/human Fabs

Unique Fab hits from FBC panning, as assessed by VH and VL sequencing, were sub-cloned into a modified pET11a expression vector by asymmetric Sfil cloning.⁵⁶ This vector introduces a His₆ tag onto the C-terminus of the Fab heavy chain. These plasmids were transformed into E. coli strain Rosetta(DE3) (EMD Millipore) using a classical calcium chloride method. Fabs expression was conducted by growing transformed bacteria in autoinduction cultures containing 100 µg/mL carbenicillin and 25 µg/mL chloramphenicol overnight at 30°C with shaking at 250 rpm. Periplasmic extracts from these Fab expression cultures were generated⁵⁵ and then IMAC- purified using a 1-mL HisTrap HP column (GE Healthcare) as previously described.⁵⁷ A portion of the I MAC purified Fabs were biotinylated via Sortase A ligation using the following reaction conditions: 20 µM Fab (1 mg/ml_), 2 μΜ Sortase A, 250 μΜ Gly₃-biotin, and 10 mM CaCI₂ in TBS. After incubating for 3 h at 42°C, biotinylated Fabs were purified using CaptureSelect Kappa-XL or CaptureSelect LC-lambda (Hu) affinity matrices (Thermo Fisher Scientific) per the manufacturer’s protocol. Fabs were then concentrated and buffer exchanged to TBS using Amicon Ultra-0.5 mL 10 kDa Centrifugal Filter Units (EMD Millipore). The quality of Fabs was analyzed by SDS-PAGE. Absorbance at 280 nm was used to quantify purified Fab concentration.

Expression profiling and epitope binning of biotinylated chimeric rabbit/human Fabs

Mammalian cells were incubated with purified chimeric rabbit/human Fabs at a concentration of 1 million cells/ml in FACS buffer. For expression profiling analysis, biotinylated Fabs were added to cells at a concentration of 1 ng/µL. For epitope binning analysis, H929 cells were first incubated with 10 ng/µL non-biotinylated competitor Fabs to mask epitopes. After pelleting cells and decanting the supernatant, cells were then co-incubated with biotinylated probe and non- biotinylated competitor Fabs at concentrations of 1 ng/µL and 10 ng/µL, respectively. For all analyses, cells were then washed twice with 100 µL FACS buffer, and then stained with a 1,000-fold dilution of PE-conjugated Streptavidin (BD Biosciences #554061) in FACS buffer. After 15 mins incubation on ice, cells were washed twice with 100 µL FACS buffer and then analyzed on an Accuri C6 flow cytometer equipped with a Hypercyt autosampler. Histograms were generated using FlowJo10 software.

Highlights.

• A phagemid display system for site-specific biotinylation of Fabs displayed on phage was developed.

• Based on this phagemid display system, a Fab-phage biotinylation & capture (FBC) approach, which depletes phage that do not display Fabs, was established.

• A naive rabbit antibody library that leverages FBC was generated and selected against whole cells.

• FBC improves target-reactive mAb enrichment.

Abbreviations:

mAb

monoclonal antibody

ELISA

enzyme-linked immunosorbent assay

CDR

complementary-determining region

IMAC

immobilized metal ion affinity chromatography

HRP

horseradish peroxidase

SDS-PAGE

sodium dodecyl sulfate polyacrylamide gel electrophoresis

cfu

colony-forming units

Fab

fragment of antigen binding

scFv

single-chain variable fragment