Anti-idiotypic antibodies elicit anti-HIV-1–specific B cell responses

We developed an anti-idiotypic antibody that specifically engages target B cells while mitigating off-target responses. This approach can be used as a vaccine strategy for variable pathogens where specific precursor B cells have been identified to correlate with protection.


Introduction
Despite nearly four decades of investigation, there is still no vaccine against HIV-1. However, broadly neutralizing antibodies (bNAbs) targeting conserved epitopes on the HIV-1 envelope (Env) are protective against infection by diverse viral strains in animal models (Burton and Hangartner, 2016;Haynes et al., 2016;Escolano et al., 2017;Kwong and Mascola, 2018). Thus, it is widely believed that a vaccine that elicits such antibodies would also be protective in humans.
VRC01-class antibodies are among the broadest and most potent bNAbs isolated to date. These antibodies share the same Ig heavy chain (IgH) V gene, VH1-2*02, which encodes a CDRH2 (complementary determining region 2, heavy chain) region that makes critical contacts with the CD4 binding site (CD4bs) on Env (Zhou et al., 2010(Zhou et al., , 2015Diskin et al., 2011;Huang et al., 2016;Sajadi et al., 2018). In addition, these antibodies express Ig light chains (IgLs) with 5-aa-long CDRL3s (complementary determining region 3, light chain), a rare feature of human IgLs that is nonetheless required to accommodate the CDRH2restricted mode of binding (West et al., 2012;Zhou et al., 2013).
An important impediment to vaccine development against HIV-1 is that immunization with soluble or multimerized recombinant Env proteins elicits antibodies that are type specific but show little or no neutralization breadth against heterologous viral variants (McCoy and Weiss, 2013;Sliepen and Sanders, 2016;Karlsson Hedestam et al., 2017;van Schooten and van Gils, 2018). This problem is not specific to HIV-1 but is also shared to a lesser extent with other variable pathogens such as influenza, Zika, and dengue virus (Brien et al., 2010;Fajardo et al., 2016;Wu and Wilson, 2017). Several different strategies have been devised to try to resolve this issue by focusing humoral responses on conserved epitopes. Examples include the development of scaffolds that carry the conserved epitopes (Ofek et al., 2010;Azoitei et al., 2011;Correia et al., 2014;Zhou et al., 2014), the use of short linear peptides that encompass target epitopes (Wang et al., 2010;Alam et al., 2017;Xu et al., 2018), shielding or deletion of off-target antigenic surfaces that are immunodominant Cherpelis et al., 2001;Mörner et al., 2009;Eggink et al., 2014;Yassine et al., 2015;Duan et al., 2018), and sequential immunization with different antigens that share the target epitope (Mörner et al., 2009;Guenaga et al., 2011).
To expand bNAb-producing B cells and their precursors and thereby give these cells a competitive advantage, we explored an alternative, non-Env-derived immunogen approach using a monoclonal anti-idiotypic antibody, iv8, that binds VRC01-class precursor B cell receptors (BCRs) with high affinity. When used as a prime, iv8 activated and expanded B cells expressing VRC01-class antibodies in a murine adoptive transfer system. Our results demonstrate that anti-idiotypic antibodies can expand VRC01-class B cells in vivo without activating B cell clones that produce off-target responses to Env.

Results
An anti-idiotypic antibody specific for VRC01-class bNAbs We produced an anti-idiotypic monoclonal antibody against germline VRC01, iv8, using hybridoma technology (see Materials and methods). In addition to germline VRC01, iv8 binds other germline VRC01-class monoclonal antibodies, including NIH45-46, 3BNC60, 12A21, and VRC-CH31 Bonsignori et al., 2012). It does not bind to mature VRC01-class or all tested non-VRC01-class anti-HIV-1 bNAbs ( Fig. 1 A). iv8 also bound a synthetic intermediate (SI) antibody comprising the mature 3BNC60 IgH and its germline IgL (3BNC60 SI ), but not to a 3BNC60 chimera composed of the germline IgH and mature IgL. Similarly, iv8 bound to a chimeric antibody comprising the germline 8ANC131 IgH (anti-CD4bs specific but derived from VH1-46) paired with germline VRC01 IgL, but not to germline NIH45-46 IgH paired with germline 8ANC131 IgL that contains a 9-aa CDRL3 ( Fig. 1 A), indicating that binding mostly depends on VRC01-class germline light chains. Another idiotypic monoclonal antibody, iv1, bound to glVRC01 and glNIH45-46 and also bound weakly to mature VRC01. The more restricted binding of iv1 indicates that its binding depends at least partially on the CDRH3 (Fig. 1 A).
To determine the molecular basis for iv8 specificity, we solved the crystal structure of iv8 single-chain variable fragment (scFv) in complex with germline VRC01 scFv at 2.4Å ( Fig. 1 B and Table S1). Indeed, iv8 primarily interacts with the IgL of germline VRC01 (constituting 80% of the total buried surface area [BSA]) including CDRL3 residues (Tyr91, Glu96, and Phe97). Of the remaining germline VRC01 BSA; 60% of the contacts are with CDRH3 (Asp99, Tyr100, and Trp100B) and 40% with the germline-encoded VH1-2*02, including residues Trp47, Trp50, and Asn58 that make critical contacts with Env ( Fig. 1, B and C; and Table S2; Zhou et al., 2010;West et al., 2012;Scharf et al., 2013). Together, the crystal structure and binding data establish the dominant role of the germline VRC01 light chain for recognition by the iv8 paratope.

VRC01-class B cells respond to iv8 in a polyclonal immune system
To determine whether iv8 binds to mouse B cells expressing VRC01-class antibodies, we isolated naive B cells from wild-type and 3BNC60 SI knock-in mice (Dosenovic et al., 2018). Whereas nearly all 3BNC60 SI knock-in B cells were positive for iv8 binding, the wild-type B cells were negative (Fig. 2 A).
To determine whether iv8 can activate VRC01-class B cells in the context of a polyclonal immune repertoire, we performed adoptive transfer experiments using 3BNC60 SI B cells (Dosenovic et al., 2018). To assess cell division of target B cells in vivo at early time points, an excess of naive 3BNC60 SI B cells (>2 × 10 6 ) were labeled with CellTrace Violet (CTV) and transferred into wild-type congenic recipient mice. The mice were injected with a control anti-idiotypic antibody (ib5; Bancroft et al., 2019) in Ribi adjuvant, or with iv8 in Ribi adjuvant, or with a multimeric nanoparticle in which the iv8 fragment antigen binding (Fab) was fused to the N-terminus of a multimerization domain derived from the Gallus gallus complement component 4B (C4b) binding protein with or without Ribi adjuvant (Hofmeyer et al., 2013). 5 d after injection, there was no detectable cell division of 3BNC60 SI B cells in mice that received the negative control antibody ib5. In contrast, both intact iv8 (mAb iv8) and iv8 Fab nanoparticles induced strong proliferation of 3BNC60 SI B cells in vivo as measured by dilution of CTV fluorescence (Fig. 2 B). However, iv8 Fab nanoparticles were more efficient at inducing cell proliferation than mAb iv8, and the addition of Ribi adjuvant enhanced the effect (Fig. 2 B). Control PGT121 knock-in B cells, which recognize a glycopeptide epitope on the third variable region of Env, showed no measurable cell proliferation in response to iv8 Fab nanoparticles (Fig. 2 B; Escolano et al., 2016). We conclude that iv8 can induce proliferation of VRC01-class B cells in vivo in the context of a polyclonal repertoire, and that iv8 Fab nanoparticles are more efficient in doing so than the intact monoclonal antibody.
To investigate 3BNC60 SI B cell responses to iv8 at later time points, we injected iv8 into mice harboring ∼25,000 naive 3BNC60 SI B cells. We assayed the mice at 14 d after injection. Although the transferred B cells were readily detected at this time point in noninjected mice, they were significantly decreased and barely detectable in mice injected with intact iv8 (Fig. 2, C and D). To evaluate whether the observed loss of activated B cells was due to insufficient T cell help, recipient mice were preprimed with 2W1S peptide to elicit CD4 + helper T cells and injected with iv8 fused to the 2W1S peptide (mAb iv8-2W1S; Fig. 2, C and D) in Ribi (Rees et al., 1999;Moon et al., 2007). The frequency of 3BNC60 SI B cells in 2W1S peptide primed mice injected with mAb iv8-2W1S was comparable to that of unimmunized mice at day 14. In contrast, mice immunized with iv8 Fab nanoparticles with no 2W1S peptide showed no decrease in 3BNC60 SI B cells over controls. Conjugation of the 2W1S helper peptide to the C-terminus of the iv8 Fab nanoparticle increased the frequency of 3BNC60 SI B cells up to ∼65-fold over the controls, an improvement that we ascribe to the combination of multimerization and recruitment of preexisting T cell help (Fig. 2,C and D).
Next, we compared the fate of 3BNC60 SI B cells in response to injection with iv8 or an authentic HIV-1 Env-derived germlinetargeting immunogen, 426c-N276D, which has an affinity of ∼40 µM compared with ∼2 nM for iv8 for 3BNC60 SI (Table S3; McGuire et al., 2013;Dosenovic et al., 2018). 3BNC60 SI B cells were labeled with CTV and transferred into 2W1S-primed wildtype mice. Recipients were injected with 426c-N276D-2W1S nanoparticles or iv8 Fab-2W1S nanoparticles and assayed for plasmablast and early B cell memory formation 5 d later by flow cytometry. We observed an increased population of 3BNC60 SI B cells after iv8 Fab-2W1S nanoparticle injection, with an arithmetic mean of 1.8 × 10 4 plasmablasts and 5.4 × 10 4 early memory B cells compared with 1.0 × 10 3 plasmablasts and 2.2 × 10 4 early memory B cells in mice injected with 426c-N276D-2W1S nanoparticles. In contrast, these cells were barely detectable in unimmunized control mice (Fig. 3, A and B, left panels; and Fig. S1, A and B). Similar results were obtained when analyzing the relative frequency of 3BNC60 SI -derived plasmablasts and early memory B cells (Fig. 3, A and B, right panels).
Germinal center (GC) development was also assayed 14 d after injection using the same experimental setup. Draining lymph nodes contained an arithmetic mean of 1.4 × 10 4 and 8 × 10 3 3BNC60 SI -derived GC B cells after 426c-N276D-2W1S or iv8 Fab-2W1S nanoparticle injection, respectively ( In comparison, 3BNC60 SI GC cells were difficult to detect in unimmunized controls (Fig. 3 C, left). We conclude that injection with 426c-N276D-2W1S or iv8 Fab-2W1S nanoparticles induces similar levels of GC responses by 3BNC60 SI B cells. However, iv8 Fab-2W1S nanoparticles induce more plasmablast and early memory B cell responses than 426c-N276D-2W1S nanoparticles.

Antibody responses in serum after injection with iv8 and Env
To evaluate serum antibody responses, we injected 2W1Sprimed wild-type recipient mice harboring ∼25,000 naive 3BNC60 SI B cells with 426c-N276D-2W1S or iv8 Fab-2W1S nanoparticles. Serum antibody responses were evaluated by ELISA using eOD-GT8, which is a distinct HIV-1 Env-derived immunogen Dosenovic et al., 2015).
Mice receiving 3BNC60 SI B cell transfers developed eOD-GT8-binding serum antibodies in response to both 426c-N276D-2W1S and iv8 Fab-2W1S nanoparticles, but control mice did not (Fig. 4 A). The eOD-GT8 responses were specific for the VRC01 epitope (the CD4bs) because there was minimal reactivity to an epitope specific knockout (eOD-GT8-KO; Fig. 4 A).
Adoptively transferred mice injected with iv8 Fab-2W1S or 426c-N276D-2W1S nanoparticles showed eOD-GT8-specific antibody responses for 2-3 mo (Fig. 4 B). However, 426c-N276D-2W1S nanoparticles also elicited significant off-target responses Figure 1. iv8 binding to inferred germline VRC01-class BCRs. (A) iv8 (black circles) and iv1 (purple triangles) binding to the indicated recombinant inferred-germline (gl), mutated (mt), or chimeric human monoclonal antibodies using BLI. Data points depict the maximum binding response (R Max) of the mAb following a 500-s association step. Each point represents the R Max of an independent replicate (n = 3). Lines indicate arithmetic means; error bars indicate SEM; HC and LC indicate IgHs and IgLs, respectively. (B) Structure of iv8 scFv bound to germline VRC01 scFv. The iv8 IgH and IgLs are shown in dark and light gray, respectively. Germline VRC01 IgH and IgLs are shown in beige and pink, respectively. (C) Distribution of BSA on glVRC01 upon iv8 binding. BSA on light-chain regions is shown in pink; BSA on heavy-chain regions is shown in beige.
to HIV-1 Env that were not observed after iv8 Fab-2W1S nanoparticle injection (Fig. 4 C). We conclude that iv8 injection engages target B cells and elicits CD4bs-specific antibody responses with minimal off-target responses to Env.

iv8 selects B cells with VRC01-class IgLs
To determine whether iv8 can select VRC01-class B cells in vivo, we used knock-in mice expressing only the germline or mature IgHs of the human 3BNC60 antibody (3BNC60 HCgl [heavychain germline] or 3BNC60 HCmt [heavy-chain mutated], respectively). Naive B cells in these mice express the 3BNC60 knock-in IgH paired with mouse IgLs (Dosenovic et al., 2015). To ensure that stimulated B cells would receive adequate T cell help, the mice were preprimed with the 2W1S peptide.
Single IgG1 + GC B cells were purified from draining lymph nodes of iv8 Fab-2W1S nanoparticle-injected and controlinjected (iv1 Fab-2W1S nanoparticles) 3BNC60 HCgl mice, and their IgLs were sequenced (Fig. S2 A). These cells were not selected based on Env binding, because the 3BNC60 HCgl antibodies do not bind to Env with high affinity. When all IgLs were considered, there was a significant preference for shorter CDRL3s in iv8 Fab-2W1S nanoparticle-injected mice (P = 0.0014; Tables S4 and S5). This result was also seen in expanded B cell clones (Fig. 5 B and Tables S4 and S5). Four of five recombinant, soluble IgGs corresponding to BCRs from five representative expanded 3BNC60 HCgl clones with 6-aa CDRL3s (Table S7) bound strongly to iv8 (Fig. 5 C). None of these mAbs bound to the 426cTM 4 ΔV1-3 or eOD-GT8 germline-targeting immunogens (Fig. 5 C).

Fig. 5 A and
IgL sequences were also obtained from single Env-binding 3BNC60 HCmt B cells in naive or iv8 Fab-2W1S nanoparticleinjected mice. When comparing all sequences, we found that iv8 selected for expansion of 3BNC60 HCmt memory B cells with significantly shorter CDRL3s compared with B cells of naive mice (P < 0.0001; Fig. 5 D and Tables S6 and S7). Moreover, all of the IgG1 + memory B cell clones obtained from two 3BNC60 One representative experiment of three is shown. (B) Representative flow cytometry plots showing proliferation of adoptively transferred 3BNC60 SI cells as indicated by CTV dilution at 5 d after injection of (from left to right) control mAb (ib5, binds glb12 but not gl3BNC60) with Ribi, mAb iv8 with Ribi, iv8 Fab nanoparticles (np) without Ribi, iv8 Fab nanoparticles with Ribi, and CTV-labeled control PGT121 knock-in B cells in response to iv8 Fab nanoparticles with Ribi. One representative animal of four per group is shown from one experiment of three. (C) Flow cytometry plots pregated on total B cells (lymphocytes, singlets, dump − , live, B220 + ) indicating frequency of 3BNC60 SI knock-in cells of total B cells at 14 d after injection of the indicated constructs in Ribi adjuvant conjugated to 2W1S peptide or not. One representative animal of two to five per group in one experiment of three. (D) Frequency of 3BNC60 SI knock-in cells in mice injected as in C. Lines indicate arithmetic mean values of individual mice per group, with error bars indicating SEM of two replicate experiments with two to five mice per group per experiment. Statistics were calculated using the Mann-Whitney U test. *, P ≤ 0.05; **, P ≤ 0.01.

Dosenovic et al. Journal of Experimental Medicine
HCmt mice injected with iv8 Fab-2W1S nanoparticles expressed IgLs with 5-aa-long CDRL3 sequences (Fig. 5 E and Tables S6 and  S7). Four of five recombinant, soluble IgGs corresponding to BCRs from five representative expanded 3BNC60 HCmt clones (Table S7) bound strongly to iv8, and all bound to the Figure 3. B cell fates in response to iv8 Fab-2W1S or N276D-2W1S nanoparticles. (A) Number of 3BNC60 SI knock-in (KI) plasmablasts (PB; CD138 + B220 low Env high ; left) or frequency of 3BNC60 SI plasmablasts of total 3BNC60 SI cells (right) per draining LN in 2W1S-primed adoptively transferred wild-type mice 5 d after injection (inj.) with iv8 Fab-2W1S or N276D-2W1S nanoparticles (np) compared with noninjected controls. Lines indicate arithmetic means, and error bars indicate SEM. Each data point represents the average from two draining lymph nodes collected from a single animal (n = 4 uninjected, n = 11 N276D-2W1S np injected, n = 11 iv8 Fab-2W1S np injected). Merged data from two independent experiments are shown. (B) As in A, but measuring number of 3BNC60 SI early memory B cells (mB; B220 + CTV low GL7 − CD38 + ; left) or frequency of 3BNC60 SI early memory B cells of total 3BNC60 SI (right) in spleen at 5 d after injection. Lines indicate arithmetic means, with error bars indicating SEM. Each data point represents the analysis in spleen of one animal (n = 4 uninjected, n = 11 N276D-2W1S np injected, n = 11 iv8 Fab-2W1S np injected). Merged data from two independent experiments. (C) As in A, but measuring 3BNC60 SI GC B cells (B220 + CD95 + CD38 − ; left) and frequency of 3BNC60 SI cells of total GCs (right) 14 d after injection. Lines indicate arithmetic means, and error bars indicate SEM. Merged data from three independent experiments are shown. Each data point represents the average number from two draining lymph nodes collected from a single animal (n = 7 uninjected, n = 17 N276D-2W1S np injected, n = 19 iv8 Fab-2W1S np injected). Statistics in A-C were calculated using the Mann-Whitney U test. **, P ≤ 0.01; ***, P ≤ 0.001; ns, nonsignificant. 426cTM 4 ΔV1-3 germline-targeting immunogen in a CD4bsspecific manner (Fig. 4 F). None of the cloned mAbs bound to eOD-GT8 (not depicted). We conclude that iv8 selects for B cells with shorter-than-average CDRL3s in 3BNC60 HC knock-in mice, and that iv8-responding B cells from 3BNC60 HCmt mice with VRC01, like 5-aa CDRL3, cross-react with a germlinetargeting Env.
iv8 recognizes B cells expressing antibody light chains similar in length to VRC01-class antibodies in human peripheral blood To determine if iv8 binds to naive human B cells expressing VRC01-class antibodies, we purified iv8 + peripheral blood B cells from seven HIV-1 negative donors and analyzed their IgHs and/ or IgLs by RT-PCR (Tiller et al., 2008) or high-throughput paired sequencing on the chromium platform from 10X Genomics. More than 80% of the iv8 + B cells expressed the VK3-11 light chain found in VRC01 (Fig. 6 A). This is consistent with the extensive interactions found between iv8 and the glVRC01 light chain in the crystal structure (Fig. 1, B and C). In contrast, the overall frequency of VK3-11 in naive B cells was ∼8% (P < 0.0001; Fig. 6 A). In addition, an arithmetic mean of 2.8% of the iv8 + B cells expressed a light chain containing a 5-aa CDRL3, corresponding to a significant enrichment over the unsorted naive cells (P = 0.0056; Fig. 6 B). Although iv8 strongly selected B cells that express a VK3-11 light chain, only an arithmetic average of 29% of the light chains that expressed a 5-aa CDRL3 were derived from VK3-11 (Fig. 6 C). As might be expected from the structural information, the IgHs were not enriched for VH1-2. Thus, iv8 enriches for human B cells expressing antibodies with a VK3-11 light chains and 5-aa CDRL3s.

Discussion
The relatively poor reactivity of inferred germline precursors of VRC01-class antibodies with conventional recombinant HIV-1 Env proteins led to the development of novel Env-derived immunogens that engage B cells expressing these antibodies Jardine et al., 2013Jardine et al., , 2016Klein et al., 2013;McGuire et al., 2013McGuire et al., , 2016Medina-Ramírez et al., 2017). However, germline targeting alone is not sufficient to elicit mature VRC01-class antibodies.
To overcome the glycan barrier to the CD4bs at position N276, the precursors of VRC01-class antibodies undergo prolonged affinity maturation and extensive somatic hypermutation in GCs (Liao et al., 2013;Doria-Rose et al., 2014). Under physiological conditions, this process involves an arms race between the virus and the humoral immune system in which antibody-mediated selection of viral mutants over a period of 1-3 yr leads to emergence of one or more bNAbs (Mascola and Haynes, 2013;Burton and Hangartner, 2016;Escolano et al., 2017;Landais and Moore, 2018). This process has been reproduced in knock-in mice expressing the inferred germline precursors of bNAbs targeting the base of the third variable region of Env and, in part, for VRC01-class antibodies, by sequential immunization (Dosenovic et al., 2015;Briney et al., 2016;Escolano et al., 2016;Tian et al., 2016).
However, the B cell compartment in knock-in mice is dominated by cells expressing inferred germline bNAbs, and there is little if any competition with other B cells. Effective competition between bNAb precursors and off-target B cells depends on their relative affinity to the immunogen and their frequency in the population (Dal Porto et al., 2002;Shih et al., 2002;Abbott et al., 2018;Dosenovic et al., 2018). The inferred germline precursor B cells of VRC01-class antibodies are rare in the normal human repertoire, and they would have to compete with a diverse set of more common B cells that recognize potentially irrelevant parts of the immunogen. Consistent with this idea, the 426c-derived germline-targeting immunogen elicited significant off-target anti-Env responses that compete with and may ultimately interfere with bNAb development (Dosenovic et al., 2015;Briney et al., 2016;McGuire et al., 2016;Tian et al., 2016;Duan et al., 2018).
We have tested the notion that anti-idiotypic antibodies can selectively expand rare B cells to increase the probability that such cells would contribute to the immune response. Several modifications were required to enhance the activity of iv8 in vivo. Whereas injection of the intact antibody led to targeted B cell depletion, a multimerized Fab fragment was a potent inducer of cell division, GC, early memory, and plasma cell development. These responses were further improved by Figure 6. iv8-binding to human peripheral blood cells. Single naive human B cells were purified by cell sorting with fluorescently labeled iv8, and the IgL kappa transcripts were recovered by RT-PCR and sequenced or subjected to paired BCR sequencing using the chromium platform from 10X Genomics. As a control, naive human B cells were enriched using magnetic beads, and paired BCR sequencing was performed using the chromium platform. (A) Frequency of IgL kappa transcripts derived from VK3-11 in iv8 + or total B cells. Each dot represents one iv8 + sort (n = 13) or paired chromium sequencing run of magnetic bead-enriched (total naive) B cells (n = 7). (B) Frequency sequenced IgL kappa transcripts with a 5-aa CDLR3 in iv8 + or total B cells. Each dot represents one iv8 + sort (n = 13) or chromium sequencing run of magnetic bead-enriched (total naive) B cells (n = 7). Lines in A and B indicate arithmetic means, and error bars indicate SEM. (C) IgVK gene usage for of iv8 + B cells with 5-aa CDRL3s identified in B. Number in the middle of the pie chart indicates the total number of IgL kappa sequences analyzed. Statistics were calculated using two-tailed unpaired Student's t tests in A and B. **, P ≤ 0.01; ***, P ≤ 0.001. LC, light chain. preexisting T cell help, which was provided by priming with, and then adding, a T cell epitope to iv8. There are three potential advantages of this approach. First, any B cells expanded by anti-idiotype that express antibodies that do not react with Env will not be further expanded upon subsequent immunization with HIV-1 Env antigens (Fig. 7 A). In contrast, priming with a germline-targeting Env would, in addition to activating the target B cells, activate off-target B cell responses that would continue to compete upon Env boosting (Fig. 7 B). Second, the anti-idiotype approach does not require a priori structural understanding or specific engineering to produce a reagent that targets the inferred germline antibody-producing B cell. Third, this method could be applied to any vaccine for which there are shared antibody signatures that are associated with protection. For example, potent anti-Zika virus VH3-23/ VK1-5 antibodies are found in many high-responder individuals, and bNAbs to influenza are typically VH1-69 (Sui et al., 2009;Corti et al., 2010;Ekiert et al., 2011;Robbiani et al., 2017).
In summary, the data indicate that anti-idiotypic antibodies that target specific germline BCRs can be used to activate and expand antigen-specific naive B cells that are present in a polyclonal repertoire. This strategy may complement HIV-1 Envbased immunization approaches and is potentially applicable to other vaccines where immunodominant off-target responses interfere with effective vaccine development.

Hybridoma generation
We injected mice three times with germline VRC01 (VH1-2*02 IgH/VK3-11 IgL). 3 d after the final injection spleens were harvested and used to generate hybridomas at the Fred Hutchinson Antibody Technology Center. Hybridoma supernatants were initially screened against glVRC01 to identify antigen-specific hybridomas. Supernatants from positive wells were then screened against a small panel of monoclonal antibodies that included glVRC01, as well as other germline non-VRC01-class antibodies that served as isotype controls by ELISA. Wells containing hybridomas that displayed positive binding for glVRC01, but negative binding to isotype controls, were sub-cloned by limiting dilution and screened for binding to a larger panel of antibodies using Bio-Layer Interferometry (BLI) on the Octet Red96 (ForteBio). Anti-mouse IgG Fc capture sensors (ForteBio) were immersed in hybridoma supernatant for 200 s. After loading, the baseline signal was then recorded for 60 s in kinetics buffer (KB; 1× PBS, 0.01% BSA, 0.02% Tween 20, and 0.005% NaN 3, pH 7.4). The sensors were then immersed in KB containing 20 µg/ml of purified human antibody for a 500 s association step. The maximum response was determined by averaging the nm shift over the last 5 s of the association step after subtracting the background signal from each analyte-containing well using empty anti-mouse IgG Fc capture ensors sensors at each time point. The binding screen included germline and mature VRC01-class mAbs, non-VRC01-class mAbs, and chimeric mAbs comprised of either a VRC01-class IgH paired with a non-VRC01-class IgL, or a non-VRC01-class IgH paired with a VRC01-class IgL. Most hybridomas including iv1 bound glVRC01 or glNIH45-46 which differ only in the CDRH3 region, but showed no cross reactivity to other antibodies in the panel.

Plasmids
The variable heavy (VH) and variable light (VL) sequences of the iv8 and iv1 hybridomas (GenBank accession nos. MK611925-MK611928) were recovered using the mouse iG primer set (Millipore) using the protocol outlined in Siegel (2009), and Sanger sequenced (Genewiz). The VH/VL sequences were codon-optimized and cloned into full-length pTT3 derived IgG1 and IgL kappa expression vectors containing human constant regions using Gibson assembly . The iv8 and iv1 VH regions were also cloned in frame with the human IgG-C1 domain fused to the modified multimerization domain of the C4b binding protein (C4b) from Gallus gallus (Ogun et al., 2008) followed by a cathepsin cleavage site (Schneider et al., 2000) and the 2W1S peptide (Rees et al., 1999). Another variant of iv8-C4b was made without the C-terminal cathepsin-2W1S fusion. pTT3-426c-N276D-gp140-KO was generated by introducing the D368R and E370A mutations (McGuire et al., 2014a) into pTT3-426c-N276D gp140 (McGuire et al., 2013) plasmid using site directed mutagenesis according to the manufacturer's recommendations (Agilent Quickchange XLII). pTT3-426c-N276D-C4b-2W1S was created by cloning cDNA corresponding to amino acids 1-479 of the 426c N276D from pTT3-426c-N276D gp140 in-frame with the C4b multimerization domain and 2W1S peptide described above. The coding sequence of eOD-GT8  preceded by a leader peptide (MDAMKRGLCCVLLLCGAVFVSPSAS) was synthesized by IDT DNA technologies and cloned into pTT3 followed by a gly-ser linker and a his-avi tag (McGuire et al., 2014a) to create pTT3 eOD-GT8his-avi. pTT3 eOD-GT8-KO was generated by introducing the D276N, W277F, R278T, D279A, and D368R mutations into pTT3-eOD-GT8-his-avi by site directed mutagenesis.

Antibody purification
Plasmids encoding mAb heavy and light chains were transfected into 293F cells at a density of 10 6 cells ml −1 in Freestyle 293 media using the 293Free transfection reagent according to the manufacturer's instructions or PEI. Expression was performed for 6 d, after which cells and cellular debris were removed by centrifugation at 4,000 ×g followed by filtration through a 0.22-µm filter. Clarified cell supernatant containing recombinant antibodies was passed over Protein A or G Agarose, followed by extensive washing with PBS, and then eluted with 1 ml of Pierce IgG Elution Buffer, pH 2.0, into 0.1 ml of Tris HCl, pH 8.0. Purified antibodies were then dialyzed overnight into PBS, and stored at −20°C before use.

Binding screen with recombinant anti-idiotypic mAbs
To verify the specificity of recombinant iv8 and iv1, they were biotinylated at a theoretical 1:1 mAb to biotin ratio using the EZ-Link NHS-PEG4-Biotin kit (Thermo Fisher Scientific) and then subjected to the BLI screen described in the "hybridoma generation" section with the following change to the first capture step: iv8 or iv1 was immobilized on a streptavidin biosensor (ForteBio) by immersing the biosensor into a 10 µg/ml solution of biotinylated mAb.
scFv expression and complex formation scFvconstructs were cloned into the pTT3 plasmid vector. ScFvs contain only the Fv antibody domains of VH/VL connected by an amino acid linker GGGGSGGGGSGGGGS from the C-terminus of the light chain to the N-terminus of the heavy chain. Iv8 scFv and glVRC01 scFv were expressed using HEK293ENBA cells (Tom et al., 2008). Cells were cultured in suspension and transfected with 500 µg/liter plasmid DNA using 293 Free Transfection Reagent (Novagen). After 6 d, cells were centrifuged at 4,500 rpm for 20 min. Supernatant was filter sterilized and incubated with His60 Ni-Superflow Resin (Novagen) overnight at 4°C, capturing scFv's with a 6his-tag on the C-terminus of the proteins. The Ni resin was separated from the supernatant and washed with a solution of 150 mM NaCl, 20 mM Tris (pH 8.0), 20 mM imidazole (pH 7.0) and eluted with a solution of 300 mM NaCl, 50 mM Tris pH 8.0, 250 mM imidazole (pH 7.0). To separate the scFv monomers from diabodies and triabodies and to further purify the proteins, the sample was run on size-exclusion chromatography using a HiLoad 16/600 Superdex 200 pg (GE) column. Complexes of iv8 scFv and glVRC01 scFv were formed by mixing proteins at a 1:1 molar ratio. ScFv complexes were removed from monomeric scFv using sizeexclusion chromatography. Complexes were concentrated to ∼10 mg/ml for crystallization trials.

Crystallization
Crystallization conditions for iv8scFv and glVRC01scFv complexes were screened and monitored with a Formulatrix NT8 drop setting and Rock Imager. Screening was done with Hampton Research Crystal Screen HT, Molecular Dimensions Proplex screen HT-96, and Rigaku Wizard Precipitant Synergy block no. 2 using the vapor diffusion method. Final crystals were grown in a solution of 0.13 M NaCl, 0.13 M Tris (pH 8.0), 10.4% PEG 20,000. Crystals were cryoprotected in solutions containing 30% molar excess of their original reagents and 20% ethylene glycol. Crystals diffracted to 2.42Å. Data were collected at Advanced Light Source 5.0.2 and processed using HKL2000 (Otwinowski and Minor, 1997).

Structure solution and refinement
The structure of iv8scFv-glVRC01scFv was solved by molecular replacement with Phaser in the CCP4 suite (Collaborative Computational Project, Number 4, 1994) using the glVRC01 Fv portion of PDB ID: 6MFT (Borst et al., 2018) as the search model for both iv8 and glVRC01. Model building and refinement were performed using COOT (Emsley and Cowtan, 2004) and Phenix (Adams et al., 2010). Statistics are summarized in Table S1.
Measurement of iv8 binding to 3BNC60 SI Kinetic measurements were obtained on the Octet Red instrument at 30°C with shaking at 1,000 rpm. Recombinant humanized iv8 was immobilized on Anti-Human IgG Fc capture sensors (ForteBio). After loading, the baseline signal was then recorded for 60 s in KB. The sensors were then immersed into wells containing serial dilutions of purified recombinant 3BNC60 SI Fab for a 300 s association step. Sensors were then immersed in KB for 600 s to determine the dissociation step. The background signal from each analyte-containing well was measured using reference sensors loaded with a negative control antibody, and subtracted from the signal obtained with each corresponding ligand-coupled sensor at each time-point. Curve fitting was performed using a 1:1 binding model and the ForteBio data analysis software. Mean association rate (k on ) and dissociation rate (k off ) values were determined by averaging all binding curves that matched the theoretical fit with an R 2 ≥0.98.
Mice 3BNC60 HCmt and HCgl knock-in mice (CD45.2 + ) carry the prerearranged Ig V(D)J genes encoding the mature (somatically mutated) or the predicted germline IgH respectively, of human bNAb 3BNC60. 3BNC60 SI knock-in mice (CD45.2 + ) carry the prerearranged Ig V(D)J genes encoding the mature IgH and predicted germline IgL of human bNAb 3BNC60 (Dosenovic et al., 2015;McGuire et al., 2016) (Dosenovic et al., 2018) where ∼5% of the transferred B cells were shown to survive. These high numbers of B cells are needed in order to be accurately detect at early times after injection. In experiments where adoptively transferred 3BNC60 SI cells or serum responses were analyzed at later time points (>10 d), 500,000 enriched naive 3BNC60 SI B cells were transferred i.v., resulting in 25,000 naive 3BNC60 SI B cells/mouse. When total 3BNC60 SI GC cells and plasmablasts per lymph node were determined, the average number of cells found in two lymph nodes collected from the same animal was used in the plots. , and nuclease-free water (3.1 µl). The sorted plates were stored at −80°C until further processing (Tiller et al., 2009). IgL sequences were retrieved using primers described in (Dosenovic et al., 2015). A clone was defined as nucleotide sequences with the same IgVL and IgJL germline gene and similar CDRL3 nucleotide sequences (similarity of CDRL3 was defined using hamming distance threshold of 0.15). Samples were acquired on a BD FACSAria.

Fluorescent iv8 probes
To generate iv8-APC, 1 µg of iv8 was incubated with 5 µl Zenon APC Human IgG labeling reagent, and incubated at RT for 10 min, then incubated with 5 µl of Zeonon blocking reagent and stored at RT until use. To generate iv8-PE, 1 µg of iv8 was incubated with 5 µl Zenon PE Human IgG labeling reagent, and incubated at RT for 10 min, then incubated with 5 µl of Zenon blocking reagent and stored RT until use.

Human subjects
Peripheral blood mononuclear cells (PBMCs) and serum were collected from HIV-uninfected adults recruited at the Seattle HIV Vaccine Trials Unit (Seattle, Washington) as part of the study "Establishing Immunologic Assays for Determining HIV-1 Prevention and Control," also referred to as Seattle Assay Control. All participants signed informed consent, and the following institutional human subjects review committee approved the protocol before study initiation: Fred Hutchinson Cancer Research Center institutional review board (Seattle, Washington). PBMC samples from donors were blindly selected at random with no considerations made for age or sex.

Statistical analysis
Data were analyzed with Prism 6 (GraphPad) for normal distribution using the D'Agostino-Pearson normality test. Data points that displayed normal Gaussian distribution were further analyzed for significant differences with Prism 6 (GraphPad) using two-tailed unpaired Student's t tests. Data points that did not display a normal Gaussian distribution were analyzed for statistical differences using the Mann-Whitney U test. Data were considered significant at *, P ≤ 0.05; **, P ≤ 0.01, ***, P ≤ 0.001.
Online supplemental material Fig. S1 depicts flow cytometric gating strategies for enumeration of B cells after injection of iv8 Fab-2W1S and N276D-2W1S nanoparticles. Fig. S2 depicts flow cytometry gating strategy for single cell sorting of heavy chain knock-in mice. Table S1 describes the data collection and refinement statistics for iv8scFv bound to glVRC01scFv. Table S2 describes the molecular interactions between glVRC01 and iv8 in detail. Table S3 contains the kinetic parameters of iv8-binding to 3BNC60 SI determined by BLI. Table S4 contains BCR sequence data obtained from controlinjected 3BNC60 HCgl knock-in mice. Table S5 contains BCR sequence data obtained from iv8 Fab-2W1S nanoparticleinjected 3BNC60 HCgl knock-in mice. Table S6 contains BCR sequence data information from HIV-1 Env-binding naive B cells in a 3BNC60 HCmt knock-in mouse. Table S7 contains BCR sequence data from HIV-1 Env binding B cells in a 3BNC60 HCmt knock-in mouse after injection with iv8.