Engineering an Antibody V Gene-Selective Vaccine

The ligand-binding surface of the B cell receptor (BCR) is formed by encoded and non-encoded antigen complementarity determining regions (CDRs). Genetically reproducible or ‘public’ antibodies can arise when the encoded CDRs play deterministic roles in antigen recognition, notably within human broadly neutralizing antibodies against HIV and influenza virus. We sought to exploit this by engineering virus-like-particle (VLP) vaccines that harbor multivalent affinity against gene-encoded moieties of the BCR antigen binding site. As proof of concept, we deployed a library of RNA bacteriophage VLPs displaying random peptides to identify a multivalent antigen that selectively triggered germline BCRs using the human VH gene IGVH1-2*02. This VLP selectively primed IGHV1-2*02 BCRs that were present within a highly diversified germline antibody repertoire within humanized mice. Our approach thus provides methodology to generate antigens that engage specific BCR configurations of interest, in the absence of structure-based information.


INTRODUCTION
Antibody target specificity originates from the interaction between germline B cell receptor (BCR) and cognate antigen. Each germline BCR displays an antigen binding site formed by six antigen binding loops or complementarity determining regions (CDRs) in which antibody gene-encoded CDRs surround the sequence-variable CDRH3 loops (1)(2)(3). CDRH3 hypervariability is generated through stochastic N-junctional diversification, where it accounts for the majority of germline BCR diversity and thus forms the principal source of antigen contact affinity (4)(5)(6). However, CDRH3 loops do not always explore the antigenic space equally, which is reflected by immunodominance hierarchies in which low frequency BCR target solutions are unable to compete for selection during subsequent antibody affinity maturation within B cell germinal centers (7,8). Such immunological subdominance is a hallmark of broadly neutralizing antibody (bnAb) responses against pathogens that defy conventional vaccine approaches, including HIV and influenza virus (8)(9)(10)(11)(12). It follows that if bnAb targeting solutions are both rare and reliant on randomly emerging CDRH3 configurations, then vaccine-expansion of the corresponding bnAb response will likely prove difficult.
Given the wide and overlapping utilities of antibody V Hendowed or public B cell responses, we sought to develop a vaccine immunogen that selectively primes human B cell lineages via preferential contact to the gene-encoded features of the antigen-binding surface. We hypothesized that affinity for a targeted gene-encoded BCR motif would enable selective vaccine-expansion of human germline B cells bearing this feature. To test this central hypothesis, we applied directed evolution on a virus-like particle (VLP) vaccine platform to identify VLPs with multivalent specificity for germline BCRs displaying the CDRs encoded by IGHV1-2*02. This strategy employed a highly immunogenic RNA bacteriophage VLP platform that has been engineered so that it can display diverse random peptide sequences. Starting with a large library of VLPs, we deployed a series of positive and negative selector antibodies with chimeric CDR displays to identify a VLP that engages the genetically conserved features of IGVH1-2*02 BCRs. We then demonstrated that this reagent selectively expands IGHV1-2*02 BCRs in vivo, within a purpose-built humanized mouse vaccine model.

Construction of VLP Selection Libraries
The expression plasmid pDSP62 was described previously (54,55). Briefly, this plasmid expresses a single-chain dimer version of the MS2 bacteriophage coat protein. The upstream copy has been 'codon-juggled' to allow discrimination of annealing sites by primers during reverse transcription (RT) and PCR steps. The plasmid contains the phage T7 promoter and terminator regions from the pET3d vector, a kanamycin resistance gene, and an M13 origin of replication. Unique SalI and BamHI restriction sites have been engineered upstream and downstream of the insert-containing copy of coat protein, for use in cloning during affinity selection. VLPs produced using pDSP62 contain 90 copies of the displayed peptide per VLP.
We previously constructed random peptide plasmid libraries for use in our VLP affinity selection protocol (55). Briefly, oligonucleotides were synthesized with 6,7,8,9,10,11,12,13,or 15 NNS codons, where N represents an equimolar mixture of all four nucleotides and S is an equal mixture of C and G. The thirty-two possible NNS codons encode all 20 amino acids and only a single stop codon. Using the Kunkel site-directed mutagenesis method and a ssDNA phagemid template, we produced plasmid libraries using the pDSP62 vector backbone. Each of the nine plasmid libraries (encoding inserts of different sizes) was generated independently and consisted of at least 10 10 individual transformants. Plasmid libraries were purified using Qiafilter maxi kits (Qiagen, Venice CA).

Production of Recombinant VLP Libraries
Individual plasmid libraries were electroporated into the E. coli T7 expression strain C41(DE3) (Lucigen) and grown to mid-log phase in LB media with Kanamycin (50 µg/mL). In order to maintain the high diversity of the plasmid library, the efficiency of transformation was monitored, and only highly efficient transformations (with >10 10 individual transformants) were used to produce VLP libraries. Coat protein expression was induced by the addition of IPTG (1 mM) for three to five hours and bacteria were collected by centrifugation and the pellet was stored at -20°C overnight. Bacteria were lysed in SCB buffer (50 mM Tris, pH 7.5, 100 mM NaCl) with 10 mg/ml of hen egg lysozyme for 1 hour at 4°C, treated with deoxycholate (at a final concentration of 0.05%) for 30 minutes at 4°C, sonicated, and then treated with 10 units/mL of DNase I for 1 hour at 4°C. Soluble protein (which includes VLPs) was separated from insoluble bacterial debris by centrifugation. Soluble protein was concentrated by precipitation with ammonium sulfate at 70% saturation followed by centrifugation, and then resuspended in SCB. VLPs were purified away from contaminating bacterial proteins by size exclusion chromatography using Sepharose CL-4B resin (Sigma-Aldrich) as previously described (56). Fractions that contained VLPs were identified by agarose gel electrophoresis, pooled, and then re-precipitated by the addition of ammonium sulfate at 70% saturation. Precipitated VLPs were collected by centrifugation, solubilized in SCB buffer and dialyzed in SCB overnight using SnakeSkin Dialysis Tubing with a 10 kDa pore size (Thermo Fisher Scientific). VLP libraries were stored at -20°C.
The HC and LC sequences of the LK mAbs were cloned into pVRC8400 (a plasmid containing the CMV IE Enhancer/ Promoter, HTLV-1 R Region and Splice Donor site, and the CMV IE Splice Acceptor site upstream of the open reading frame) and co-transfected in 293F cells (59). After six days of expression, the cultures were centrifuged (2,000 × g, 10 min) and after filtration of the supernatant (VacuCap 8/0.2mm filters, Pall Corporation), the sample mixed for one hour with Protein G-Agarose (Pierce, Cat #20398). The resin was then washed (6 column volumes of PBS) and the bound IgG was eluted into 50 mM Tris, pH 8, using 1.5 column volumes of low pH IgG elution buffer (Pierce, Cat#21004). The LK mAbs were then concentrated using Amicon Ultra concentrators (30 kDa cut off) and were further resolved by size exclusion FPLC using a Superdex 200 10/300 column (GE Healthcare) followed by SDS PAGE and staining with GelCode Blue (Thermo Scientific, MA).

VLP Selections
The selection scheme is shown schematically in Figure 3A. Selections were performed by coating ELISA plates (Immulon 2HB; Thermo Fisher Scientific) with a total of 500 ng of antibody (LK1, selection rounds 1 and 2, or LK4, rounds 3 and 4) in 50 µL of PBS (phosphate-buffered saline) overnight at 4°C. The wells were then blocked using PBS/2% bovine serum albumin (BSA) in a total volume of 100 µL for 2 hours at room temperature (~25°C). A mixed library of VLPs was generated by mixing of equal amounts of the nine different VLP libraries displaying random sequence inserts ranging from 6 to 15 amino acids in length. 50 µg of the pooled VLPs were suspended in a total of volume of 50 µl of PBS/1% BSA added to wells and then incubated at room temperature for 2 hours at room temperature. After extensive washing with PBS, bound VLPs were eluted by incubating wells with 50 µl of 0.1M glycine (pH 2.7) for 5 minutes. Eluted VLPs were then brought to neutral pH by addition of 5 µl 1M Tris (pH 9.0).
Reverse transcription (RT) was performed using 8 µl (~15%) of the eluent as template, with 1.25 µM of a primer annealing downstream of the MS2 coat protein sequence (5′-TCAGCGGTGGCAGCAGCCAA-3′ ) and MMLV-RT (Invitrogen) following the manufacturer's instructions. The product of this reaction was amplified by PCR using High Fidelity Platinum Taq (Invitrogen) and primers that annealed upstream (5'-CTATGCAGGGGTTGTTGAAG-3') and downstream 5'-CGGGCTTTGTTAGCAGCCGG-3') of the portion of coat protein that contained the inserted sequence. The PCR product was purified using the QIAquick PCR Purification Kit (Qiagen), digested using BamHI-HF and SalI-HF (New England Biolabs), and then re-ligated into the pDSP62 expression vector. Ligation reactions were ethanol precipitated and resuspended in 10 µl of nuclease-free water. The entire volume was used to transform electrocompetent 10G cells (Lucigen); pooled transformants were grown overnight in a 100 ml culture LB media with kanamycin (50 µg/ml). Plasmid DNA was recovered from these pooled cultures by midiprep (Qiagen) and then used to generate VLPs for subsequent rounds of selection, as described above. As is shown in Figure 2A, we performed four rounds of selection, alternating between LK1 (rounds 1 and 2) and LK4 (rounds 3 and 4). After the fourth round of selection, individual transformants were isolated and sequenced. A more detailed protocol describing the generation of VLPs displaying random peptides and affinity selections is also available (60).

ELISAs
Selected VLPs were tested by ELISA to measure binding to the positive selectant antibodies (LK 1 and LK 4) and negative selectors (LK 5 and LK 6). Briefly, 500 ng of VLPs in a total volume of 50 µl were adsorbed to Immulon II HB ELISA plates overnight at 4°C. The wells were blocked from nonspecific binding using 0.5% dry milk in PBS in a 100 µl volume for 2 hours at room temperature. Wells were then incubated with serial dilutions (diluted in PBS/0.5% milk) of the antibodies of interest (LK1, LK4, LK5, or LK6) for 2.5 hours at room temperature. Following extensive washing with PBS, 50 µL of a 1:4000 dilution of horse radish peroxidase (HRP)-conjugated goat anti-human IgG (Jackson Immunoresearch), diluted in PBS/0.5% milk, was added to each well and incubated for 1 hour at room temperature. ELISAs were developed using TMB substrate (50 µL, EMD Millipore), stopped using 50 µL of 1% HCl, and then absorbance was measured at 450 nm using an accuSkan FC plate reader (Fisher Scientific).

BCR Triggering In Vitro
To evaluate V H -dependent signaling by VLP selectants, the LK1, LK4, LK5 and LK6 antibodies were expressed as IgM BCRs within a BCR-surface negative Ramos B cell line that enables display of mono-specific IgM BCRs of interest (17,59). Ectopic BCRs are stably expressed by through lenti-viral mediated delivery of membrane anchored HC and LC sequences (59). This BCR reporter system has been now widely described and deployed as a tool to rank-order candidate immunogens in antigen receptor triggering studies (15,31,59,(61)(62)(63)(64)(65)(66)(67) and its display of user-defined BCR sequences has also been extensively characterized and validated by deep sequencing (68).
In this study, reporter B cells displaying LK1, LK4, LK5 or LK6 IgM BCRs were evaluated for antigen receptor signaling following incubation with VLP selectants, as per our standard method (59). In these experiments, 1×10 6 cells displaying monoclonal BCR were resuspended in RPMI media and exposed to 10µg/ml VLP selectant or 0.5 mg/ml anti-IgM F(ab') 2 (Southern Biotech). BCR stimulation was measured kinetically by flow cytometry (LSR II, BD) as the ratio of the Ca 2+ bound/unbound states of the membrane permeable and ratiometric dye Fura Red. For each LK cell line, the ratiometric measurements were made before and after antigen exposure, wherein data was acquired for 300 seconds after stimulation. All the values were normalized to total Ca 2+ flux capacity, as defined by exposure of the cells to 10 mg/ml ionomycin (59). Downstream analyses of the data were performed using FlowJo software version 9.5.2 (TreeStar).

Transgenic Mice
The transgenic mice used in this study were of a previously established model in which human antibody V H usage is constrained to user-defined gene segments, while allowing for normal and random recombination with diverse human D and J segments, which generates an antibody CDRH3 repertoire that is similar to humans, both in relation to length distribution and amino acid usage (31,53). In this study mice constrained to the human V H gene IGHV1-2*02 (31,53) were used, and were a gift to D.L. from Bristol-Myers Squibb (Redwood City, CA). The animals were maintained within Ragon Institute's HPPF barrier facility and the experiments were conducted with IACUC approval (MGH protocol 2014N000252). In this study, both male and female animals, aged 6-10 weeks, were used.

Statistics
Statistical analyses was performed using Prism Graphpad software. The expansion of CD45.2 +/+ B cells in response to VLP-F2 or VLP displaying irrelevant peptide was compared using Students' T-test. The sample sizes were n=5 animals per treatment and an alpha level of 0.05 was deployed throughout.

Systematic Overview
We present a pipeline for generating multivalent VLP antigens with biochemical specificity to user-defined features of the germline antibody binding site (Figure 1). We first apply a highly immunogenic RNA bacteriophage VLP platform that has been engineered so that it can display diverse random peptide sequences (54,55,69). Each library contains more than 10 10 individual transformants, and each individual VLP displays a different guest peptide on its surface and encapsidates its own mRNA, meaning that the VLPs can be deployed for affinity selection. Recombinant antibodies displaying chimeric antigen binding sites are then applied positive and negative selectors to identify VLP with affinity for a particular antibody region of interest, in this study the human IGHV1-2*02 domain ( Figure 1). The selected VLPs are then evaluated for antibody binding and capacity to elicit the corresponding BCR signaling within a B cell reporter system (59

Affinity Selections to Identify VLPs That
Bind the IGHV1-2*-02 Domain Recombinant selector antibodies displaying chimeric antigen binding sites were referred to as 'LK' mAbs ( Figure 2). Each LK was assembled using IgG1 and displayed the following chimeric paratopes: 1) IGHV1-2*02 + VRC01 CDRH3 and identical to LK1 (LK5) or LK4 (LK6). The LKs were generated recombinantly and proper IgG assembly was confirmed using size exclusion chromatography and SDS PAGE (Figures 2B, C). LKs were then applied as positive and negative VLP selectors, where substrate specificity for the IGHV1-2*02 domain could be identified by reactivity to LK1 and LK4 but not to LK5 and LK6. We deployed the LK mAbs to select and screen for VLPs that bound specifically to the IGHV1-2*-02 domain ( Figure 3A). The VLP library was subjected to four rounds of positive selection, alternating between the LK1 (rounds 1 and 2) and LK4 (rounds 3 and 4) mAbs. Following the final selection, a small number of individual selectants were cloned, sequenced, used to express VLPs, and then tested for binding to the panel of LK antibodies by ELISA. This analysis identified classes of VLPs with different LK binding specificities ( Figure S1). One VLP (HMRGGAYAYTD; designated F2) bound strongly to both LK1 and LK4, but not LK5 and LK6 ( Figure 3B), indicating that it was selected for binding to the IGHV1-2*-02 domain. Conceivably, LK5-and LK6-mAbs could have been co-applied as negative selectors to further direct the evolution of VLP selectants toward IGHV1-2*02 domain specificity, however this was not required.

Selected VLPs Engage and Activate B Cells Expressing IGHV1-2*-02
Multivalent antigen display is a principle that enables BCR crosslinking and receptor signaling (75,76) and has long been applied to enhance humoral output by vaccines (77)(78)(79)(80). As this principle was built into our VLP platform (54,55), we could test whether VLP-F2 triggered IGHV1-2*02 BCRs. Accordingly, we expressed LK1, LK4, LK5 and LK6 as IgM BCRs in a B cell reporter system (59). In this system, IgM BCRs of interest are stably expressed in Ramos B cells lacking surface display of their endogenous BCRs, enabling evaluation of specific antigen receptor triggering by vaccine candidates in vitro (15,31,(61)(62)(63)(64)(65)(66)(67)(68). LK IgM BCR activation was evaluated kinetically, by Ca 2+ flux, where we found that VLP-F2 triggered signaling from LK1 and LK2 BCRs but not from LK4 or LK5 BCRs ( Figure 3C). Control VLP failed to stimulate any of the LK BCRs, despite the fact that BCR signaling in response to crosslinking by anti-IgM was comparable across the LK sequences. These data indicated that the VLP-F2 specificity for the IGHV1-2*02-encoded domain translated into IGHV1-2*02-dependent BCR signaling in vitro. The VLP-F2 displaying the peptide sequence HMRGGAYATD was isolated which showed affinity for LK1 and LK4 mAbs (mean ± SD, n = 3 replicates), but not LK5 or LK6 mAbs (see other VLP reactivity profiles in Figure S1). (C) IGHV1-2*02-depedent BCR triggering activity by VLP-F2. To confirm that selective affinity for the IGHV1-2*02 germline domain translated into V H -gene-selective BCR activation, the LK antibodies were stably expressed as IgM BCRs in an engineered B cell reporter line that enables monospecific display of user defined BCRs (59). BCR activation was measured kinetically by Ca 2+ flux using the ratiometric dye Fura Red. Following baseline acquisition (20s), LK BCRs were exposed to anti-IgM, VLP-F2 or VLP-control and the data was acquired for 300s. The baseline reading was then subtracted and the Ca 2+ flux values were standardized to total cell Ca 2+ flux capacity as defined by exposure to the ionophore ionomycin. Each run represents the average of two independent fluxes per BCR.

DISCUSSION
Deployment of shared features underscores reproducible broadly neutralizing antibody responses against microbial pathogens. We present methodology for generating virus-like-particle (VLP) vaccines that expand targeted human B cell lineages via engineering antigen specificity to the gene-encoded features of the germline antibody binding site. Antibody V gene-selective B cell priming has the potential to broadly augment antibody output from any public/V gene-selective response, such as against influenza virus, HCV, HBV, HIV, SARS-CoV-2, yellow fever virus, or the malaria parasite Plasmodium falciparum (13-22, 24-28, 88-92). While these antibody-target solutions can be immunologically recessive (and thus resistant to expansion by traditional vaccines), we have previously shown that germlineencoded affinity for cognate antigen can provide natural reproducible substrate for pathway-amplifying normally subdominant human influenza bnAbs (31,32). Compared to VLP displaying irrelevant peptide, we found that VLP-F2 enriched for the differentiation of naïve IgM IGHV1-2*02 B cells into plasmablasts, a stage of B cell expansion that is supported by both extrafollicular and follicular input (84)(85)(86). This provides experimental proof of concept for V H -biased B cell priming within an otherwise highly diverse germline antibody repertoire. Similar approaches have deployed adoptive transfer of monoclonal BCRs which are then activated by structure-based germline stimulating vaccines (37-39, 46, 70). By contrast, the IGHV1-2*02 B cells in our system were polyclonal, bearing unconstrained human-like diversity in their hypervariable CDRH3 loops (31,53), which are the principal source of BCR diversity (4-6). Thus a key difference was to broadly prime human B cell lineages independent of their hypervariable features. While we cannot not rule out additional selection of CDRH3 structures in the response, a critical aspect of our approach was engineering VLP-affinity against the geneencoded features of the BCR. This was enabled by the chimeric LK constructs, which facilitated the identification of VLP specificity to the IGHV1-2*02 domain. This specificity then manifested as a capacity to selectively trigger IGHV1-2*02 BCRs in vitro and selectively prime and expand polyclonal IGHV1-2*02 B cell lineages in vivo. Given the functional utility of IGHV1-2*02 in both follicular anti-viral responses and in extrafollicular antibacterial responses (15,48,53,83), we suggest that V H -dependent B cell expansion could serve to intensify such activities, and could be conceivably serve as a directed activation step for any V H -gene dependent antibody response.
Another difference from current germline stimulating vaccine concepts is that our method does not rely on structure-based information for immunogen design, which can constrain experimental work on a few well-described B cell development pathways (37-39, 46, 70). Notably, we isolated VLPs bearing selective affinity for each LK selector antibody, suggesting that directed evolution on the VLP library harbors the potential to engage whichever BCR. Thus while we focused on specificity for the IGHV1-2*02 domain as proof of concept, our method could provide broad utility in engaging and expanding any BCR target of interest.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding authors.

ETHICS STATEMENT
The animal study was reviewed and approved by MGH Animal Committee, IACUC protocol 2014N000252.

AUTHOR CONTRIBUTIONS
LR, AY, JP, DP, BC, and DL designed the research studies. LR, AY, JP, VO, PD, and AM performed the research. RB, DR, and NL provided the transgenic mice. LR, AY, JP, DP, BC, and DL analyzed the data and wrote the paper. All authors contributed to the article and approved the submitted version.

FUNDING
This work was supported by NIH funding to D.L. and B.C. (R01AI124378, R01AI137057, DP2DA042422, R01AI153098, R01AI155447), the Harvard University Milton Award, and the Gilead Research Scholars Program. The funder was not involved in the study design, collection, analysis, interpretation of data, the writing of this article or the decision to submit it for publication. JP was supported by NIH grant T32-AI007538. We thank the members of the Lingwood and Chackerian labs for helpful discussion and technical assistance.