Detection and activation of HIV broadly neutralizing antibody precursor B cells using anti-idiotypes

There are many pathogens, such as HIV-1, respiratory syncytial virus, and influenza virus, for which the development of a protective vaccine has been elusive despite decades of effort. In many cases, the failure does not appear to be an absolute inability of the immune system to produce protective antibodies, since they have been characterized in some infected individuals (Beeler and van Wyke Coelingh, 1989; Burton et al., 1991; Okuno et al., 1993; Johnson et al., 1997; Karron et al., 1997; Scanlan et al., 2002; Kashyap et al., 2008; Throsby et al., 2008; Ekiert et al., 2009, 2011; Scheid et al., 2009, 2011; Sui et al., 2009; Walker et al., 2009, 2011; Wu et al., 2010; Corti et al., 2011; Dreyfus et al., 2012, 2013; Huang et al., 2012, 2014; Magro et al., 2012; Mouquet et al., 2012; Liao et al., 2013; Ngwuta et al., 2015). It is not entirely clear why traditional vaccine development approaches have failed to induce similar protective antibodies, but mounting evidence suggests that protective vaccines for these pathogens will likely need to stimulate specific lineages of antibodies targeting defined epitopes (Pantaleo and Koup, 2004; Rappuoli et al., 2016; Lang et al., 2017; Robbiani et al., 2017; Goodwin et al., 2018; Kwong and Mascola, 2018). In contrast, traditional vaccines generally stimulate polyclonal antibody responses against whole pathogens, or large subunits from pathogens, rather than focus the immune response toward known protective epitopes.

The observation that a subset of HIV-1–infected individuals develop broadly neutralizing antibodies (bNAbs) targeting the HIV-1 envelope protein (Env) that are capable of neutralizing diverse HIV-1 viral isolates over the course of infection demonstrates that the human immune system is capable of generating bNAb responses provided it receives the appropriate antigenic stimulation. Indeed, numerous bNAbs have been isolated from infected individuals that define the sites of vulnerability and mechanisms of neutralization of HIV-1 (West et al., 2014; Burton and Hangartner, 2016; Kwong and Mascola, 2018). Importantly, bNAbs have been shown to protect from experimental infection in animal models, suggesting that they will be an important component of an effective vaccine against HIV-1 (Mascola et al., 1999; Shibata et al., 1999; Parren et al., 2001; Hessell et al., 2009; Moldt et al., 2012; Pietzsch et al., 2012; Gruell et al., 2013; Balazs et al., 2014; Shingai et al., 2014; Gautam et al., 2016; Liu et al., 2016).

Unfortunately, bNAbs have not been elicited by vaccination in humans, despite the use of diverse recombinant Env-derived immunogens and immunization schemes (Burton et al., 2004; Cohen and Dolin, 2013; Schiffner et al., 2013). Recent advances in the amplification of antibody genes from individual B cells have revealed that bNAbs often require extensive somatic mutation to achieve broad and potent neutralizing activity (Scheid et al., 2009; Dimitrov, 2010; Klein et al., 2013a; Kwong et al., 2013; Mascola and Haynes, 2013; West et al., 2014). Although mutated bNAbs are capable of broad Env recognition and virus neutralization, most inferred germline bNAb variants do not display reactivity with any Env variant tested, which led us and others to propose that one reason for the lack of bNAb elicitation through immunization is the lack of stimulation of B cells expressing bNAb progenitor B cell receptors (BCRs) by Env-based immunogens (Xiao et al., 2009; Hoot et al., 2013; Jardine et al., 2013; McGuire et al., 2013; Andrabi et al., 2015; Zhou et al., 2015; Gorman et al., 2016).

Env-based immunogens capable of binding to certain germline bNAbs have been recently developed that effectively activate B cells in knock-in mice expressing the desired BCRs in vivo (Dosenovic et al., 2015; Escolano et al., 2016; Jardine et al., 2016; McGuire et al., 2016; Steichen et al., 2016; Abbott et al., 2018). For many bNAbs, however, antigens capable of being bound by the corresponding germline forms have not yet been identified or designed. Here, we demonstrate the use of anti-idiotypes as an alternative approach to structure-based immunogen design to target the inferred germline version of the HIV-1 bNAb b12. We focused upon b12 because despite being one of the first HIV-1 bNAbs isolated, Env-based immunogens have not been identified or developed to recognize its ancestral germline form (Xiao et al., 2009; Hoot et al., 2013). The mature form of b12 was isolated through phage display (Barbas et al., 1992), and it used a heavy chain derived from V_H1-3*02/D_H2-21/J_H6*03 and light chain derived from V_K3-20*01/J_K2 (Xiao et al., 2009; Hoot et al., 2013). Using anti-idiotypes specific for the inferred germline version of b12 (iglb12) as baits for single B cell sorting, we identified a subset of human germline BCRs using V_H1-3 with some heavy chain CDRH3 similarity to iglb12. While crystal structures indicated that one of these anti-iglb12 idiotypes made extensive contacts with the iglb12 CDRH3 region encoded by D_H2-21, this gene segment was not enriched in sorted B cells. Moreover, we could not identify any CDRH3 regions with >50% amino acid sequence identity to iglb12 within a dataset including hundreds of BCRs. In light of this, we hypothesized that B cells expressing BCRs containing CDRH3s with high similarity to iglb12 may be deleted from the repertoire as a result of autoreactivity with self-antigens. In line with this notion, we demonstrate that the iglb12 antibody stains HEp-2 cells, a feature of other autoreactive HIV-1–specific antibodies (Haynes et al., 2005; Verkoczy et al., 2010; Ota et al., 2013). To evaluate the consequence of this autoreactivity in vivo, we generated mice with the iglb12 heavy chain knocked into the endogenous murine heavy chain locus. The iglb12 heavy chain knock-in mice exhibited B cell deletion during development, with the surviving B cells exhibiting BCR down-regulation and anergy. Despite the presence of anergy, immunization with a multimerized version of an anti-iglb12 idiotype stimulated the proliferation and differentiation of transgenic B cells expressing the iglb12 heavy chain. In summary, our results establish a proof of concept that anti-idiotypes can be used as immunogens to identify and stimulate the ancestral germline version of protective antibodies.

To generate anti-idiotype antibodies to iglb12 (Hoot et al., 2013), we immunized and boosted mice intraperitoneally with iglb12 in adjuvant. 3 d after the final immunization, splenocytes were isolated and fused to myeloma cells, resulting in the generation of hundreds of hybridomas producing antibodies able to bind iglb12. Nearly all of the clones were excluded because they bound to a panel of control germline antibodies as well as iglb12, suggesting specificity for the heavy or light chain constant regions of human Ig (data not shown). Importantly, four clones (IB1, IB2, IB3, and IB5) exhibited binding to iglb12 with similar apparent affinities (dissociation constants [K_D] of ∼10⁻⁹) but did not bind a panel of control antibodies derived from germline BCR sequences of anti–HIV-1 antibodies (Fig. 1, A–C; and Table S1). The anti-iglb12 idiotypes also failed to bind the fully mature version of b12, indicating a preference for the unmutated sequence (Fig. 1 C). One clone, IB2, bound a chimera expressing the iglb12 heavy chain and a noncognate control light chain, while the other three clones, IB1, IB3, and IB5, did not bind this chimera (Fig. 1 C), suggesting different modes of recognition between IB2 and IB1, IB3, and IB5. We next assessed whether binding of these clones was specific for the iglb12 heavy chain CDRH3 region as opposed to a dependence on V_H1-3, the variable gene segment used by iglb12. Consistent with our clones specifically recognizing features unique to iglb12, all four clones failed to bind naturally paired control antibodies using V_H1-3 (DeKosky et al., 2016) whether or not they were paired with a V_k3-20 light chain, from which the b12 light chain is derived (Fig. 1 C). Together, these data suggest that these four clones preferentially recognized the iglb12 idiotype.

Among the four anti-iglb12 idiotypes identified by our screen, we focused upon IB2 and IB3 because we solved their crystal structures in complex with iglb12 at a resolution of 2.6 Å and 3.0 Å, respectively. The crystal structures revealed that IB2 and IB3 displayed distinct modes of interaction with iglb12 (Fig. 2). IB3 interacted with both the heavy and light chains of iglb12 (Fig. 2, A–C), consistent with biolayer interferometry (BLI) data (Fig. 1 C). Most interactions occurred within the heavy chain of iglb12 (Fig. 2, B and C), which contributed 76% of the buried surface area (BSA) and six of the eight hydrogen bonds. One hydrogen bond occurred in each of the CDRH3 and CDRL3 regions of iglb12, with the remaining six hydrogen bonds and BSA occurring in the CDR1, CDR2, and framework regions of iglb12, encoded by V_H1-3 or V_K3-20 (Fig. 2, D and E). In contrast, IB2 approached iglb12 at a 45° angle relative to IB3 and bound exclusively to the iglb12 heavy chain (Fig. 2, A and F). The CDRH3 region of iglb12 contributed nearly half of the BSA and two of the six hydrogen bonds, with the remaining BSA and hydrogen bonds largely focused upon CDRH1 and CDRH2, which are encoded by V_H1-3 (Fig. 2, G–J). The extensive interactions with the iglb12 heavy chain CDRH3 likely explain why IB2 and IB3 do not bind other antibodies using V_H1-3 (Fig. 1 C).

We next assessed whether the anti-iglb12 idiotypes could be used to identify B cells expressing iglb12-like BCRs from within the polyclonal human B cell repertoire of HIV-uninfected individuals using single B cell sorting followed by RT-PCR amplification and sequencing of heavy and light chain genes. In an initial set of experiments, IB1, IB2, and IB3 were biotinylated and tetramerized using streptavidin conjugated to allophycocyanin (APC) and incubated together with 100 million human peripheral blood mononuclear cells (PBMCs) before enrichment using anti-APC microbeads. Using this approach ∼2% of B cells in the APC-enriched fraction bound to IB1/2/3-APC, but not APC-Dylight755 (APC755)–conjugated isotype control tetramers (Fig. 3 A). Highlighting the efficiency of this enrichment approach, few IB1/2/3-APC⁺ B cells could be detected in the APC-depleted fraction (Fig. 3 A). Using this approach, single B cells binding IB1/2/3 tetramers were sorted into individual wells of a 96-well plate, and the heavy and light chains expressed by these cells were sequenced using nested RT-PCR (Tiller et al., 2008). In total, we FACS-purified 91 single IB1/2/3-binding B cells from an individual. This yielded 66 heavy chain and 38 light chain sequences, with 31 heavy and light chain pairs (Table S2). To confirm the specificity of our flow cytometry approach, antibodies were produced from paired heavy and light chains from 10 cells. 80% of these antibodies bound IB2, but not IB1, IB3, or IB5, when assessed by BLI (Fig. 3 B). Interestingly, the eight antibodies that bound IB2 all used the same heavy chain variable region as igbl12, V_H1-3, while the two antibodies that failed to bind any of the anti-idiotypes used other V_H alleles (Table S2). In two of the cloned antibodies, the V_H1-3⁺ heavy chains were paired with V_K3-20⁺ light chains, suggesting that the lack of binding to IB1, IB3, and IB5 was not due to the absence of this segment (Fig. 1 C). These results highlight the high specificity of our flow cytometry approach and suggest that IB2 is bound by a higher number of cells compared with other anti-idiotypes.

To gain a more in-depth understanding of the BCRs recognized by individual anti-idiotypic antibodies, we next conducted sorts with IB2 and IB3 labeled with different fluorophores. We focused our analysis on these anti-idiotypes because the high-resolution crystal structures of IB2 and IB3 complexed with iglb12 allowed us to make predictions about the BCRs they may preferentially recognize. From the crystal structures, we predicted that IB2 would be highly selective for B cells expressing BCRs using V_H1-3 with CDRH3 regions similar to iglb12. In contrast, we predicted that IB3 would be selective for B cells expressing BCRs using V_H1-3 with more diverse CDRH3s paired with V_k3-20 light chains with some CDRL3 similarity to iglb12. Due to the random nature of V(D)J recombination that gives rise to CDR3 regions, we did not expect to find B cells expressing CDRH3s or CDRL3s identical to those of iglb12.

To test these predictions, IB2 and IB3 were biotinylated and tetramerized using streptavidin conjugated to APC or R-phycoerythrin (PE). Staining PBMCs with anti-idiotype tetramers revealed that <0.05% and 0.008% of the live B cell population from HIV-uninfected individuals bound IB2-APC or IB3-PE, respectively, but not APC755- or PE-Dylight594 (PE594)–conjugated isotype control tetramers (Fig. 4 A). To more easily identify rare B cells binding IB2 and/or IB3, tetramer-binding cells were enriched using APC-specific and PE-specific magnetic microbeads from samples of 200 million PBMCs before flow cytometry (Fig. 4 A). Using this approach, single B cells binding IB2 and/or IB3 tetramers were sorted into individual wells of a 96-well plate, and the heavy and kappa light chains expressed by these cells were sequenced using nested RT-PCR (Tiller et al., 2008). In total, we FACS-purified 576 single IB2-binding B cells and 150 single IB3-binding B cells from three individuals. From IB2-binding B cells, we obtained 302 heavy chain and 174 kappa light chain sequences, with 136 heavy and light chain pairs (Table S3). IB3-binding B cells yielded 84 heavy chain and 44 kappa light chain sequences, with 27 heavy and light chain pairs (Table S4). We were unable to recover heavy or light chain sequences from the small number of sorted cells binding both IB2 and IB3 tetramers.

The BCR sequences from B cells binding to IB2 or IB3 were next compared with iglb12. On average, over 71% of the IB2-binding B cells from all three individuals used the same V_H gene as iglb12, V_H1-3 (Fig. 4 B). In contrast, only 4.6% of individually sorted B cells expressed V_H1-3 when neither anti-idiotype was used to select cells for sorting (Fig. 4 B), consistent with previous studies (DeKosky et al., 2016). In terms of the entire B cell repertoire, an average of 0.04% of the B cells bound IB2 and used V_H1-3 (Fig. 4 C). In contrast, only 11% of the IB3-binding B cells used V_H1-3, which amounted to ∼0.001% of the B cell repertoire (Fig. 4, B and C). Consistent with the crystal structure revealing that IB3 interacts with both the heavy and light chains of iglb12 (Fig. 2, B and C), four out of the five IB3⁺ V_H1-3⁺ BCRs in which light chain sequences were recovered used the same light chain as igbl12, V_K3-20, or the closely related V_K3D-20 (Fig. 4 D). Surprisingly, IB2⁺ V_H1-3⁺ BCRs exhibited a slight preference for V_K3-20 (Fig. 4 D), which was not predicted by the heavy chain–only mode of binding indicated by the crystal structure (Fig. 2, F–J).

We next examined the unique heavy chain CDRH3 regions used by IB2⁺ V_H1-3⁺ BCRs in search of further similarity to iglb12. Since we were only able to detect 10 IB3⁺ V_H1-3⁺ BCRs in this dataset, these BCRs were not included in this analysis. The crystal structure of IB2 bound to iglb12 revealed that 44% of the BSA was contributed by the CDRH3 of iglb12 (Fig. 2 J). Nearly all of the CDRH3 BSA, 82% (i.e., 36% of the total BSA), is focused on the portion encoded by D_H2-21 (Fig. 4 E). However, only 3.6% of IB2⁺ V_H1-3⁺ BCRs used D_H2-21, which was not different from control V_H1-3⁺ BCRs (Fig. 4 F). Of the CDRH3 regions in IB2⁺ V_H1-3⁺ BCRs, the two most similar to iglb12 CDRH3 only displayed 50% amino acid identity (Fig. 4 G).

The moderate similarity of IB2⁺ BCR sequences to iglb12 led us to consider whether there were biological mechanisms selecting against CDRH3s that more closely matched iglb12. Specifically, we hypothesized that the iglb12 heavy chain was autoreactive, resulting in deletion of similar sequences from the B cell repertoire. Autoreactivity of iglb12 has not been reported, and conflicting reports of autoreactivity have been reported for the mature form of b12. One study reported that mature b12 exhibited autoreactivity using an in vitro assay (Haynes et al., 2005; Ota et al., 2013), however B cell development in transgenic mice expressing this antibody was normal (Ota et al., 2013). As a preliminary assessment of potential autoreactivity, we measured iglb12 binding to human HEp-2 cells using immunofluorescence. This approach revealed that iglb12 bound moderately to HEp-2 cells compared with 4E10 (Fig. 5, A and B), an HIV-specific antibody previously shown to exhibit binding in this assay (Haynes et al., 2005; Verkoczy et al., 2010). The iglb12 staining was similar to the mature form of b12, both of which were brighter than that of 10E8 (Fig. 5, A and B), an antibody that was previously shown to be nonreactive in this assay (Haynes et al., 2005; Verkoczy et al., 2010). Together, these data suggest that iglb12 is autoreactive.

To assess whether the autoreactivity detected in the HEp-2 assay resulted in functional consequences in vivo, we generated a transgenic mouse in which all B cells expressed the iglb12 heavy chain. Previous work has demonstrated that B cells from BCR knock-in mice expressing 4E10, but not mature b12, are deleted from the repertoire as a consequence of autoreactivity (Doyle-Cooper et al., 2013; Ota et al., 2013). We found that iglb12 heavy chain transgenic mice displayed a three- to fourfold reduction in the number of CD19⁺ B220⁺ CD93⁻ CD3⁻ F4/80⁻ Gr-1⁻ mature B cells in the spleen compared with WT mice or mice expressing a control inferred germline heavy chain generated using the same knock-in strategy (Jardine et al., 2015; Fig. 6, A and B). Deletion could be traced back to the immature stage in the bone marrow, where iglb12 heavy chain transgenic mice contained approximately twofold fewer CD19⁺ B220⁺ CD93⁺ IgM⁺ CD3⁻ F4/80⁻ Gr-1⁻ mature B cells compared with both WT and control transgenic animals (Fig. 6, C and D). Together, the data indicate that a significant portion of murine B cells expressing an iglb12 heavy chain are deleted in the bone marrow.

Previous studies have shown that autoreactive B cells that escape deletion are often rendered functionally unresponsive to antigenic stimulation, a state referred to as anergy (Goodnow et al., 1988, 2010; Cyster et al., 1994; Fulcher and Basten, 1994; Klinman, 1996; Nemazee, 2006, 2017). Given this, we next considered whether the B cells remaining in iglb12 heavy chain transgenic mice were anergic. Previously, several phenotypes of anergic B cells have been described in mice, including CD93⁺ CD23⁺ IgM^LOW “T3” transitional phenotype, as well as mature B cells expressing low levels of IgM (Allman et al., 2001; Merrell et al., 2006; Taylor et al., 2012; Agrawal et al., 2013; Sabouri et al., 2016). In agreement with the hypothesis that circulating iglb12 heavy chain transgenic B cells were anergic, the number of T3 cells was increased 29-fold in these animals compared with control transgenic animals (Fig. 6, C and E). In addition, IgM expression was 6.4-fold lower on mature B cells from iglb12 heavy chain transgenic mice compared with control heavy chain transgenic animals (Fig. 7 A). IgD expression was also reduced 1.9-fold on mature B cells from iglb12 heavy chain transgenic mice compared with control transgenic animals (Fig. 7 B), which suggested a global down-regulation of surface BCR on these cells. Down-regulated surface BCR expression by iglb12 heavy chain transgenic mature B cells was confirmed through the assessment of CD79β (Igβ; Fig. 7 C), a component of the BCR signaling complex required for surface BCR expression.

The increased frequency of iglb12 heavy chain transgenic B cells exhibiting an anergic phenotype in combination with decreased BCR expression suggested that these cells would respond poorly to BCR stimulation. To test this, we labeled purified mature CD93⁻ B cells from WT and iglb12 heavy chain transgenic mice with CellTrace Violet (CTV) and incubated them with varying concentrations of anti-Ig for 72 h in vitro before analysis. At all concentrations of anti-Ig, fewer iglb12 heavy chain transgenic B cells had diluted CTV compared with B cells from WT mice (Fig. 8, A and B). These data suggest that the iglb12 heavy chain transgenic B cells that escape deletion are functionally anergic. Combined with the HEp-2 staining data, our murine data suggest that the iglb12 heavy chain is autoreactive, supporting the hypothesis that deletion and anergy of iglb12-like BCRs occurs in humans.

The presence of anergy within iglb12 heavy chain transgenic B cells suggests that immunogens targeting this population would need to overcome these functional defects. Before performing immunization experiments, we modified IB2 to ensure optimal murine responses. Since IB2 was generated in a mouse, it would likely generate a poor response if injected unmodified, since tolerance mechanisms would be expected to eliminate most sources of T cell help targeting it. To increase possible CD4⁺ T cell help, the heavy and light chain constant regions used by this anti-idiotype were humanized (Tiller et al., 2009). Since multimerization can overcome anergic responses (Cooke et al., 1994; Gautam et al., 2016; McGuire et al., 2016), humanized IB2 antigen-binding fragment (Fab) was fused to a modified multimerization domain of the chicken C4b-binding protein (huIB2-C4b), which self-assembles into a multimer containing up to seven humanized IB2 Fab domains (Ogun et al., 2008; Hofmeyer et al., 2013). To further garner T cell help, the 2W epitope (Rees et al., 1999) was added to the C-terminus of C4b separated by a cathepsin consensus sequence in order to promote endosomal processing and MHC presentation (Schneider et al., 2000). For immunization experiments, 4 × 10⁵ B cells from CD45.2⁺ iglb12 heavy chain transgenic mice were labeled with CTV and transferred into CD45.1⁺ WT recipient mice. 1 d after transfer, the mice were immunized with huIB2-C4b or an irrelevant humanized isotype control-C4b fusion (control-C4b). 5 or 14 d following the injection of the control-C4b, CD45.2⁺ donor B cells accounted for an average of ∼0.0006% of total B cells in control recipient animals, which we detected by enriching for CD45.2⁺ cells before analysis (Fig. 9, A and B). As expected of resting cells that had not encountered antigen and are therefore not proliferating, these cells largely retained high levels of CTV (Fig. 9, C and D). In contrast, 5 d following injection of huIB2-C4b, donor B cells increased twofold to an average of ∼0.001% of total B cells (Fig. 9, B–D). Approximately half of the donor B cells had low levels of CTV, indicating that proliferation accounted for the increased frequency of donor cells (Fig. 9, B–D). In addition to expansion, ∼33% of transgenic B cells in huIB2-C4b–immunized mice bound GL7 and down-regulated CD38 (Fig. 9, E and F), a phenotype of germinal center B cells (Allen et al., 2007; Kurosaki et al., 2010; Victora et al., 2010; Vinuesa et al., 2010; Kitano et al., 2011). In these experiments, IB2-binding germinal center B cells derived from both the transgenic donor and recipient animals could be detected using IB2 tetramers (Fig. 10, A–C). This analysis revealed that transgenic CD45.2⁺ iglb12 B cells accounted for 5.8% of the IB2-binding germinal center B cells (Fig. 10 D). This frequency represented an increase from the 1–2% of transgenic cells found in cells outside the germinal center in both huIB2-C4b and control-C4b injected animals (Fig. 10 D). Together, these results indicate that anti-iglb12 idiotype antibodies can be used as immunogens that can overcome anergy and induce B cell activation and entry into germinal centers.

Our results demonstrate that anti-idiotype antibodies can identify and activate bNAb precursor B cells in vivo and represent a potential alternative or complement to structure-based immunogen design. Anti-idiotypes present several advantages. First, antigens derived from pathogens are generally difficult to produce and are often highly unstable or difficult to maintain in the desired conformation, especially after injection. From a manufacturing standpoint, anti-idiotypes are easily produced once identified and would be expected to be highly stable in vivo, as evidenced by the licensure of several therapeutic antibodies. Another advantage to this approach is that one could simultaneously use numerous anti-idiotypes to stimulate multiple potentially protective lineages. However, it must be recognized that it is unlikely that the injection of a single anti-idiotype containing immunogen would induce the production of a protective antibody in situations where high levels of somatic mutation are required (Scheid et al., 2009; Dimitrov, 2010; Pancera et al., 2010; Zhou et al., 2010; Haynes et al., 2012; Mouquet et al., 2012; Jardine et al., 2013; Klein et al., 2013a,b; McGuire et al., 2013; Escolano et al., 2016). In these situations, successive injections of immunogens containing anti-idiotypes of increasingly mature antibodies may guide antibody maturation toward highly protective antibodies. Alternatively, anti-idiotypes may be better suited for activation of protective antibodies that require less somatic mutation, such as those that can neutralize Middle East respiratory syndrome (Agrawal et al., 2016) and respiratory syncytial virus (Goodwin et al., 2018), or as the first step in a maturation pathway that includes pathogen-derived antigen in subsequent boosts.

To identify four clones that produced anti-iglb12 idiotypes, we used multiplexed screening approaches to evaluate thousands of hybridomas. In future studies the screening burden can be further reduced by enrichment or sorting of B cells expressing BCRs with desirable binding properties before and/or after hybridoma fusion (Taylor et al., 2012; Spanier et al., 2016).

To our knowledge, this study represents the first time anti-idiotypes have been used to identify and analyze naive B cells expressing BCRs with similarities to the inferred germline sequence of a broadly neutralizing HIV-1–specific antibody. An important result within these analyses is the finding that different anti-idiotypic antibodies can have different modes of interaction with BCRs similar to the one of interest. The binding profile analysis (Fig. 1 C) and crystal structure data (Fig. 2) suggested that IB2 would select only for BCRs with iglb12-like V_H1-3⁺ heavy chains, while IB3 would select for BCRs with iglb12-like V_H1-3⁺ heavy and V_K3-20⁺ light chains, perhaps as a subset of the IB2-binding population. Sorting and sequencing BCRs from human B cells was therefore important and revealed that IB2 and IB3 selected for completely different populations of cells, with IB2 selecting far more stringently for V_H1-3⁺ BCRs compared with IB3. Given these findings, we are exploring ways to better predict which anti-idiotypes will be better at identifying BCRs of interest. Nevertheless, our results indicate that probing the preimmune repertoire with anti-idiotypes or other potential immunogens is an informative step in vaccine design.

In the course of our sequencing experiments, we screened through the equivalent of 30 million B cells and only identified a few CDRH3 sequences recognized by the anti-idiotypes that showed >50% amino acid identity to iglb12. It is conceivable that the anti-idiotypes were able to recognize rare CDRH3 sequences with higher identity to iglb12 but that we were unable to identify them due to limitations in the efficiency of cell sorting and BCR sequencing. The ability to screen a larger number of cells currently exceeds our capacity but is obtainable using enrichment and cell sorting in combination with droplet-based BCR sequencing approaches.

Nevertheless, the absence of CDRH3 sequences with high similarity to iglb12 led us to hypothesize that these sequences may be purged from the repertoire due to autoreactivity. Support for this hypothesis came from iglb12 staining of HEp-2 cells and experiments in which B cell deletion and anergy were observed in transgenic mice expressing the iglb12 heavy chain. These tolerance effects may explain why b12-like bNAbs have not been commonly identified from HIV-infected individuals in several studies. In contrast, bNAbs with other specificities are present and are identified in a large fraction of HIV-1 subjects who develop bNAb responses, such as those targeting the N332 supersite (Pancera et al., 2014; Landais et al., 2016; Rusert et al., 2016; Ditse et al., 2018).

The existence of deletion and anergy in mice suggested that immunization strategies able to overcome tolerance may be required to stimulate the production of b12-like bNAbs in humans. The need to overcome tolerance is not unique to iglb12 knock-in mice, since deletion and anergy have also been detected in mice expressing the inferred germline forms of the HIV-1 bNAbs 4E10 and 3BNC60 (McGuire et al., 2016; Zhang et al., 2016). Autoreactivity is not a unique feature of some HIV-specific lineages and has been detected in assessments of influenza bNAbs (Andrews et al., 2015; Bajic et al., 2019). Notably, it remains possible that the tolerance exhibited in transgenic mice expressing the iglb12 heavy chain is an artifact of the inability to accurately infer the true unmutated heavy chain CDRH3 from the mature b12 antibody. For example, the CDRH3 in this sequence contains two N regions that cannot be inferred and are instead carried over from the mature sequence. It is possible that the true germline heavy chain CDRH3 had different sequences in these locations that would not have resulted in deletion and anergy.

Importantly, activation and germinal center entry of iglb12-expressing B cells is just the first step of the maturation process. This lineage would need to be guided toward broad Env binding and neutralization, which will likely require many rounds of mutation and selection. Future work will explore whether prime boost experiments with IB2/IB3 and Env can be used to guide the maturation of bNAb responses.

Overall, our study demonstrates the utility of using anti-idiotypes to identify and stimulate B cells expressing desirable BCRs. Future work is aimed at including anti-idiotypes in the vaccine design process, both as tools to identify cells of interest and as immunogens themselves.

Blood was obtained by venipuncture from healthy, HIV-seronegative adult volunteers enrolled in the General Quality Control study in Seattle, WA, which was approved by the Fred Hutchinson Cancer Research Center institutional review board. Informed consent was obtained from all study participants before enrollment into the parent protocols. PBMCs were isolated from whole blood using Accuspin System Histopaque-1077 (Sigma-Aldrich) resuspended in 10% dimethylsulfoxide in heat-inactivated fetal bovine serum and cryopreserved in liquid nitrogen before use.

All experiments were performed in accordance with relevant institutional and national guidelines and were approved by Fred Hutchinson Cancer Research Center Institutional Animal Care and Use Committee. 6–14-wk-old male and female WT C57BL/6J and B6.SJL-Ptprca Pepcb/BoyJ (CD45.1⁺) mice were purchased from the Jackson Laboratory or bred in house. Generation of iglb12 heavy chain transgenic mice and control heavy chain transgenic mice was performed at The Scripps Research Institute Mouse Genetics Core Facility similarly to what was described previously for the mature b12 heavy chain (Ota et al., 2013). Briefly, the D_Q52-J_H cluster was replaced in a C57BL/6-derived C2 embryonic stem cell line with the iglb12 heavy chain VDJ exon or inferred germline VRC01 heavy chain VDJ exon. Mice derived from these modified embryonic cells carrying the transgenic heavy chain were interbred and homozygous transgenic heavy chain^+/+ male or female mice were used for experiments. Control transgenic animals expressing the iglVRC01 heavy chain were generated in an identical manner (Jardine et al., 2015).

The sequence of iglb12 has been previously reported (Hoot et al., 2013, and the CDRH3 and CDRL3 sequences are listed in Table S2. Anti-idiotypic antibodies were generated by the Fred Hutchinson Cancer Research Center Antibody Technology Core. Briefly, 8–12-wk-old female BALB/c mice (Jackson Laboratory) were immunized and boosted intraperitoneally with iglb12 in adjuvant five times over a 5-mo period. 3 d following the final boost, splenocytes were electrofused to myeloma cells, and thousands of hybridoma colonies were screened for binding to iglb12 and a panel of control germline antibodies. Hybridomas producing antibodies specific for iglb12 were subcloned from single cells and expanded before anti-idiotype purification. Supernatant from these expanded hybridomas was harvested and purified over protein G resin (Pierce) following the manufacturer’s recommendations. To produce recombinant anti-idiotypes, RNA was extracted from 1 × 10⁶ cells using the RNeasy kit (Qiagen) and the heavy and light chain sequences of the murine hybridomas were by obtained using the mouse Ig-primer set (69831; EMD Millipore) using the protocol developed by Seigel et al. (Siegel, 2009). Sequences were codon optimized and cloned into pTT3-based IgG expression vectors with human constant regions (Snijder et al., 2018) using In-Fusion cloning (Clontech).

Purified anti-idiotype antibodies were biotinylated using an EZ-link Sulfo-NHS-LC-Biotinylation kit (Thermo Fisher Scientific) using a 1:1 molar ratio of biotin to anti-idiotype. Unconjugated biotin was removed by centrifugation using a 50-kD Amicon Ultra size exclusion column (EMD Millipore). To determine the average number of biotin molecules bound to each anti-idiotype molecule, streptavidin-PE (ProZyme) was titrated into a fixed amount of biotinylated anti-idiotype in increasing concentrations and incubated at room temperature for 30 min. Samples were run on an SDS-PAGE gel (Bio-Rad Laboratories) and transferred to nitrocellulose and incubated with streptavidin–Alexa Fluor 680 (1:10,000; Thermo Fisher Scientific) to determine the point at which there was excess biotin available for the streptavidin–Alexa Fluor 680 reagent to bind. Biotinylated anti-idiotypes were mixed with streptavidin-PE or streptavidin-APC at the ratio determined above to fully saturate streptavidin and incubated for 30 min at room temperature. Unconjugated anti-idiotypes were removed by several rounds of dilution and concentration using a 300K Nanosep centrifugal device (Pall Corporation). Control PE594 and APC755 tetramers were created by mixing isotype control antibodies with streptavidin-PE preconjugated with Dylight 594 (PE594) or streptavidin-APC preconjugated with Dylight 755 (APC755; both from Thermo Fisher Scientific) following the manufacturer’s instructions. On average, PE594 and APC755 contained 4–8 Dylight molecules per PE/APC. The concentration of each anti-idiotype tetramer was calculated by measuring the absorbance of PE (565 nm, extinction coefficient = 1.96 µM⁻¹ cm⁻¹) or APC (650 nm, extinction coefficient = 0.7 µM⁻¹ cm⁻¹) and the lot-specific ratio of streptavidin-PE or streptavidin-APC.

For human PBMC analysis, 200 × 10⁶ frozen cells were thawed into RPMI with 10% fetal bovine serum (Thermo Fisher Scientific), 100 U/ml penicillin (Thermo Fisher Scientific), 100 µg/ml streptomycin (Thermo Fisher Scientific), and 0.0275 mM 2-mercaptoethanol (Sigma-Aldrich). Cells were centrifuged and resuspended to 0.2 ml in ice-cold FACS buffer composed of 1× Dulbecco's PBS (DPBS) with 1% newborn calf serum (Thermo Fisher Scientific) containing 2% rat and mouse serums. PE594- and APC755-conjugated control anti-idiotype tetramers were added at a final concentration of 5–15 nM and incubated on ice for 5 min to allow binding to B cells of unwanted specificities. IB2-PE and IB3-APC or IB1-APC, IB2-APC, and IB3-APC tetramers were added at a final concentration of 5 nM and incubated on ice for 25 min, followed by a 10-ml wash with ice-cold FACS buffer. 25 μl of both anti-PE- and/or anti-APC-conjugated microbeads (Miltenyi Biotec) were added and incubated on ice for 30 min. 3 ml of FACS buffer was then added to the cell mixture and passed over a magnetized LS column (Miltenyi Biotec). The column was washed once with 5 ml ice-cold FACS buffer and then removed from the magnetic field. 5 ml ice-cold FACS buffer was pushed through the unmagnetized column twice using a plunger to elute the bound cell fraction.

Purified naive mature B cells were prepared from spleens of iglb12 heavy chain transgenic mice using a B cell negative selection kit (Miltenyi Biotec) supplemented with 1.25 µg/ml biotinylated anti-CD93 (AA4.1; eBioscience) to remove transitional B cells. Purified B cells were washed in 1× DPBS, adjusted to a concentration of 5 × 10⁷ cells/ml in warm 1× DPBS and incubated with 5 µg CTV (Thermo Fisher Scientific) for 12 min at 37°C before being washed with media containing 10% fetal bovine serum (Thermo Fisher Scientific). Cells were resuspended in warm 1× DPBS and 4 × 10⁵ cells per mouse were injected retro-orbitally into CD45.1⁺ recipients. 1 d following the transfer, recipient mice were injected intraperitoneally with 0.2 ml of the following solution: 0.1 ml (25 µg) diluted Sigma Adjuvant System (Sigma-Aldrich) and 20 µg of a humanized version of IB2 containing Fabs fused to the 2W epitope (EAWGALANWAVDSA) heptamerized on the C4B nanoparticle as described previously (Hofmeyer et al., 2013) in 0.1 ml 1× DPBS. A humanized isotype control Fab-C4b in diluted Sigma Adjuvant System was used as a control. 5 d following injection, the spleen, inguinal, brachial, cervical, and mesenteric lymph nodes were harvested into 1× DPBS, minced, and then digested as described previously (Taylor et al., 2015) in dispase (0.8 mg/ml; Thermo Fisher Scientific), collagenase (0.2 mg/ml; Roche), and DNase (0.1 mg/ml; Roche) for 20 min at 37°C. Enzymes were inactivated with 5 mM EDTA, and single-cell suspensions were obtained. Samples were centrifuged and cell pellets resuspended to 0.2 ml in 1× DPBS containing 1% newborn calf serum, 5 µg/ml of anti-CD16/CD3 (2.4G2, BioXCell), and 1.25 µg/ml anti-CD45.2 biotin (104; eBioscience) and enriched using 25 µl anti-biotin microbeads (Miltenyi Biotec) as described for tetramer enrichment.

After centrifugation, supernatant was discarded and cell pellets were resuspended in FACS buffer. The entire enriched fraction and 1/40th of the flow-through or unenriched fractions and were incubated with antibodies against surface markers in a total volume of 50 µl for 25 min on ice and washed with FACS buffer. Human PBMCs were labeled with 6.24 µg/ml anti-IgM FITC (G20-127; BD Biosciences), 4 µg/ml anti-IgD PerCP-Cy5.5 (IA6-2; BD Biosciences), 2 µg/ml anti-CD20 eFluor 450 (2H7; eBioscience), 1 µl anti-CD27 BV650 or BV480 (L128; BD Biosciences), 1 µl anti-CD3 BV711 (UCT1; BD Biosciences), 1 µl anti-CD14 BV711 (MϕP9; BD Biosciences), 2.8 µg/ml anti-CD16 BV711 (3G8; BioLegend), and 1 µl anti-CD19 PE-Cy7 (HIB19; BD Biosciences).

In adoptive transfer experiments, mouse spleen and lymph node cells were labeled with 2.5 µg/ml GL7 FITC (BD Biosciences), 2 µg/ml anti-CD45.1 PerCP-Cy5.5 (A20; BioLegend), 2 µg/ml anti-CD38 Alexa Fluor 700 (90; eBioscience), 2 µg/ml anti-CD45.2 PE-Cy7 (104; BioLegend), 2 µg/ml anti-CD3 BV510 (145-2C11; BD Biosciences), 2 µg/ml anti-F4/80 BV510 (BM8; BioLegend), 2 µg/ml anti–Gr-1 BV510 (RB6-8C5; BD Biosciences), 2 µg/ml anti-B220 BV786 (RA3-6B2; BD Biosciences), and 2 µg/ml anti-CD19 BUV395 (1D3; BD Biosciences). 1 µg/ml streptavidin-BV650 (BD Biosciences) was also included to detect donor cells that were enriched using anti-CD45.2 biotin.

For B cell subset analysis, cells were labeled with 2.5 µg/ml anti-CD79β FITC (HM79-12; BD Biosciences), 2 µg/ml anti-IgD PerCP-Cy5.5 (11-26c.2a; BD Biosciences), 2 µg/ml anti-CD21 PE-Cy7 (eBio8D9; eBioscience), 2 µg/ml anti-CD93 BV421 (AA4.1; BD Biosciences), 2 µg/ml anti-CD3 BV510 (145-2C11; BD Biosciences), 2 µg/ml anti-F4/80 BV510 (BM8; BioLegend), 2 µg/ml anti-Gr-1 BV510 (RB6-8C5; BD Biosciences), 2 µg/ml anti-IgM^b BV650 (AF6-78; BD Biosciences), 2 µg/ml anti-CD23 BV786 (B3B4; BD Biosciences), 2 µg/ml anti-B220 BUV395 (RA3-6B2; BD Biosciences), and 2 µg/ml anti-CD19 BUV737 (1D3; BD Biosciences).

For in vitro proliferation experiments, cells were labeled with 2 µg/ml anti-CD3 BV510 (145-2C11; BD Biosciences), 2 µg/ml anti-F4/80 BV510 (BM8; BioLegend), 2 µg/ml anti-Gr-1 BV510 (RB6-8C5; BD Biosciences), and 2 µg/ml anti-CD19 BUV395 (1D3; BD Biosciences).

For all experiments, 0.25 µl Fixable Viability Dye eFluor 506 or eFluor 780 (eBioscience) or Fixable Viability Stain 700 (BD Biosciences) was included in the surface antibody cocktail to distinguish live from dead cells. Flow cytometry was performed on a five-laser (355 nm, 405 nm, 488 nm, 561 nm, and 640 nm) LSR II, LSRFortessa, FACSARIA II, or FACSymphony device (BD Biosciences) and analyzed with FlowJo 10 software (Tree Star). 2 × 10⁴ fluorescent AccuCheck counting beads (Thermo Fisher Scientific) were added to the samples before flow cytometry and used to calculate total numbers of cell subtypes in the column-bound and flow-through suspensions, as previously described (Taylor et al., 2015).

Naive mature B cells were purified and labeled with CTV as described for adoptive transfer experiments. Cells were adjusted to a concentration of 2 × 10⁶ cells/ml in RPMI (Thermo Fisher Scientific) containing 10% fetal bovine serum (Thermo Fisher Scientific), 100 U/ml penicillin (Thermo Fisher Scientific), 0.1 mg/ml streptomycin, 2 mM L-glutamine (Thermo Fisher Scientific), 0.02 mg/ml gentamicin (Thermo Fisher Scientific), and 0.027.5 mM 2-mercaptoethanol (Sigma-Aldrich). 2 × 10⁵ cells from each sample were added per well of a 96-well flat bottomed plate, and cells were cultured in the presence or absence of 2, 5, 10, or 25 µg/ml F(ab′)₂ goat anti-mouse Ig (Jackson ImmunoResearch) in a final volume of 0.2 ml and incubated for 72 h at 37°C before analysis by flow cytometry.

Single human B cells were isolated using a FACSARIA II cell sorter with Diva configuration (BD Biosciences) following tetramer enrichment and cell surface marker staining using human PBMCs as described above. Single B cells from defined subpopulations were sorted according to surface marker expression patterns. Specifically, IB2- or IB3-specific B cells were sorted using side scatter–based doublet discrimination, followed by gating on FVD⁻ CD19⁺ CD20⁺ CD3⁻ CD14⁻ CD16⁻ B cells that bound IB2-PE or IB3-APC tetramers but not PE650 or APC755 tetramers containing isotype control antibodies. We also sorted 576 control FVD⁻ CD19⁺ CD20⁺ CD3⁻ CD14⁻ CD16⁻ CD27⁻ IgM/D⁺ B cells from two individuals that were not stained with an anti-idiotype. Single cells were sorted into individual wells of 96-well PCR plates (Eppendorf) containing 10 µl/well ice-cold lysis buffer containing 0.25 µl (12.5 U) RNaseOUT (Thermo Fisher Scientific), 2.5 µl 5× SuperScript IV First Strand Buffer (Thermo Fisher Scientific), 0.625 µl 0.1 M dithiothreitol (Thermo Fisher Scientific), 0.3125 µl 10% Igepal detergent (Sigma-Aldrich), and 6.625 µl diethyl pyrocarbonate–treated water. Plates were sealed with adhesive PCR plate seals (Thermo Fisher Scientific), centrifuged briefly, and immediately frozen on dry ice before storage at −80°C.

RT was performed using SuperScript IV (Thermo Fisher Scientific) as previously described (Tiller et al., 2009; Wu et al., 2010). Briefly, 3 µl RT reaction mix consisting of 1.5 µl 50 µM random hexamers (Thermo Fisher Scientific), 0.4 µl 25 mM deoxyribonucleotide triphosphates (dNTPs; Thermo Fisher Scientific), 0.5 µl (10 U) SuperScript IV RT, and 0.6 µl water was added to each well containing a single sorted B cell in 10 µl lysis buffer and incubated at 50°C for 1 h. Following RT, 2 µl cDNA was added to 19 µl PCR reaction mix so that the final reaction contained 0.2 µl (0.5 U) HotStarTaq Polymerase (Qiagen), 0.075 µl 50 µM 3′ reverse primers, 0.115 µl 50 µM 5′ forward primers, 0.24 µl 25 mM dNTPs, 1.9 µl 10× buffer (Qiagen), and 16.5 µl water. The PCR program was 94°C 30 s, 57°C 30 s, 72°C 55 s, 50 cycles, 72°C 10 min for IgM/IgG/kappa light chains and 94°C 30 s, 60°C 30 s, 72°C 55 s, 50 cycles, 72°C 10 min for lambda light chains. After the first round of PCR, 2 µl of the PCR product was added to 19 µl of the second-round PCR reaction so that the final reaction contained 0.2 µl (0.5 U) HotStarTaq Polymerase, 0.075 µl 50 µM 3′ reverse primers, 0.075 µl 50 µM 5′ forward primers, 0.24 µl 25 mM dNTPs, 1.9 µl 10× buffer, and 16.5 µl water. PCR programs were the same as the first round of PCR. 4 μl of the PCR product was run on an agarose gel to confirm the presence of a ∼500-bp heavy chain band or 450-bp light chain band. 5 μl from PCR reactions showing the presence of heavy or light chain amplicons was mixed with 2 µl of ExoSAP-IT (Thermo Fisher Scientific) and incubated at 37°C 15 min, followed by 80°C for 15 min to hydrolyze excess primers and nucleotides. Hydrolyzed second-round PCR products were sequenced by Genewiz with the respective reverse primer, and sequences were analyzed using IMGT/V-Quest to identify V, D, and J gene segments.

Non-V_H1-3–derived control antibodies were previous described (Hoot et al., 2013; Liao et al., 2013; McGuire et al., 2014a,b). Paired heavy chain VDJ and light chain VJ sequences from BCR sequencing experiments or naive V_H1-3⁺ B cells identified previously (DeKosky et al., 2016) were codon optimized and cloned into pTT3-derived expression vectors containing the human IgG1, IgK, or IgL constant regions using In-Fusion cloning (Clontech).

BLI assays were performed on the Octet.Red instrument (ForteBio). For anti-idiotype binding screens, anti-mouse IgG capture sensors (ForteBio) were immersed in kinetics buffer (1× PBS, 0.01% BSA, 0.02% Tween 20, and 0.005% NaN₃, pH 7.4) containing 5 µg/ml purified anti-idiotypic antibody for 100 s. After loading, the baseline signal was then recorded for 60 s in kinetics buffer. The sensors were then immersed in kinetics buffer containing 20 µg/ml purified human antibody for a 100 s association step. The maximum response was determined by averaging the nanometer 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 sensors at each time point.

Kinetic analyses were performed at room temperature with shaking at 500 rpm. Anti-mouse Ig capture sensors were immersed in kinetics buffer containing 20 µg/ml purified anti-idiotypic antibody for 240 s. After loading, the baseline signal was then recorded for 60 s in kinetics buffer. The sensors were next immersed into wells containing serial dilutions of purified iglb12 Fab in kinetics buffer for a 300-s association phase, followed by immersion in kinetics buffer for an additional 600-s dissociation phase. The background was signal measured from each analyte-containing well using sensors loaded with a negative control antibody and subtracted from the signal obtained with each corresponding ligand-coupled sensor at each time point. Kinetic analyses were performed at least twice with an independently prepared analyte dilution series. Curve fitting was performed using a 1:1 binding model and ForteBio data analysis software. The mean on rate and off rate values were determined by averaging all binding curves in each dilution series that matched the theoretical fit with an R² value of ≥0.95.

The ZEUS IFA ANA HEp-2 Test System was used as recommended by the manufacturer (Zeus Scientific) with minor changes. Briefly, HEp-2–coated slides were incubated with 25 µl antibody at 0.1 mg/ml for 30 min at room temperature. Slides were then washed three times in 1× DPBS followed by the incubation of 25 µl of 0.01 mg/ml goat anti-human IgG Alexa Fluor 594 (Thermo Fisher Scientific) in 1× DPBS for 30 min at room temperature in the dark. After washing three times with 1× DPBS, slides incubated with ZEUS IFA ANA HEp-2 Test System fixative, and coverslips were applied. Images were acquired using the EVOS Cell Imaging System (Thermo Fisher Scientific) and analyzed using ImageJ to determine the average Alexa Fluor 594 fluorescence per HEp-2 cell for each image.

IB2 and IB3 IgG purified from hybridomas were digested over activated Ficin on Agarose Resin (Thermo Fisher Scientific) for 48 h to obtain Fabs. Fabs for iglb12 were generated by digesting with 1 µg Lys C (Roche) for 10 mg of IgG overnight at 37°C. The Fabs were purified from Fc regions by collecting the fraction that did not bind to a Protein A column (Thermo Fisher Scientific) and by size exclusion chromatography using Superdex 200 (GE Healthcare Life Sciences) in 5 mM Hepes and 150 mM NaCl.

A c/scFv iglb12 was cloned and expressed using lentiviral-based stable transduction of human embryonic kidney suspension adapted freestyle 293F cells (Thermo Fisher Scientific) as described previously (Bandaranayake et al., 2011). Briefly, plasmids containing appropriate constructs were incorporated into replication-incompetent lentiviral particles and transduced into human embryonic kidney suspension adapted freestyle 293F cells plated in fresh media at 10⁶/ml in 10 ml and incubated at 37°C with 8% CO₂ and 80% humidity. Cells were supplemented with 25 ml fresh freestyle media 8–16 h after transduction. 3 d following transduction, cells were replated at a cell density of 0.5 × 10⁶/ml in cultures as large as 2 liters. Culture supernatants were harvested by centrifugation at 8,000 rpm to remove cells and clarified by passing through 0.22-μm filter (Thermo Fisher Scientific). Clarified supernatants were supplemented with 150 mM NaCl and concentrated using a 10,000 molecular weight cutoff Amicon Stirred Cell concentrator (EMD Millipore). Recombinant proteins were further purified by size exclusion chromatography using a Superdex 75 and/or Superdex 200 column (GE Healthcare) running in 25 mM piperazine-N,N′-bis(2-ethanesylfonic acid) pH 7.0, 150 mM NaCl, 1 mM EDTA, and 0.02% NaN₃. Eluted fractions were analyzed using SDS-PAGE to assess purity. Protein concentrations were determined using molar extinction coefficients and absorbance at 280 nm.

Complexes of IB2 Fab + iglb12 c/scFv and IB3 Fab + iglb12 Fab were formed by mixing both components at a 1:1 molar ratio and incubating overnight at 4°C. Complexes were purified by size exclusion chromatography using Superdex 200 in 5 mM Hepes and 150 mM NaCl at pH 7.4. Crystals of IB2 Fab + iglb12 c/scFv and IB3 Fab + iglb12 Fab were obtained using a mosquito and an NT8 dispensing robots, and screening was done using commercially available screens (Rigaku Wizard Precipitant Synergy block #2, Molecular Dimensions Proplex screen HT-96, Hampton Research Crystal Screen HT) by mixing 0.1 µl/0.1 µl (protein/reservoir) by the vapor diffusion method. Crystals used for diffraction data were grown in the following conditions: IB3 Fab + iglb12 Fab crystals were grown in solution containing 0.2 M ammonium phosphate monobasic, 0.1 M Tris, pH 8.5, and 50%(+/−) 2-methyl-2,4-pentanediol. IB2 Fab + iglb12 c/scFv crystals were grown in solution containing 13.4% polyethylene glycol 8000, 13.4 PEG 400, 0.1 M MgCl₂, and 0.1 M Tris, pH 8.5, using the vapor diffusion method. Crystals were cryoprotected in solutions containing 30% molar excess of their original reagents and 20% glycerol for IB3 Fab + iglb12 Fab and 20% 2R-3R Butanediol for IB2 Fab + iglb12 c/scFv. Crystal diffracted to 3 Å for IB3 Fab + iglb12 Fab and 2.6Å for IB2 Fab + iglb12 c/scFv. Data were collected with advanced light source 5.1 and 5.2 and processed using HKL2000 (Otwinowski and Minor, 1997).

The structures were solved by molecular replacement using Phaser in CCP4 (Collaborative Computational Project, Number 4, 1994) and PDB accession number 2NY7 (Zhou et al., 2007) for the mature b12 portion as a search model. Iterating rounds of structure building and refinement was performed in COOT (Emsley and Cowtan, 2004) and Phenix (Adams et al., 2010). The refinement statistics are summarized in Table S5, and the structural figures were made with Pymol (DeLano, 2002).

Atomic coordinates and structure factors of the IB2/iglb12 (6OL6) and IB3/iglb12 (6OL5) complex x-ray crystal structures have been deposited in the Protein Data Bank. The heavy and light chain sequences of IB1 (MK775682 and MK775686), IB2 (MK775683 and MK775687), IB3 (MK775684 and MK775688), and IB5 (MK775685 and MK775689) have been deposited in GenBank.

Table S1 contains sequence information for the control antibodies used in Fig. 1 C. Table S2 contains sequence information for the heavy and light chain sequences derived from B cells binding the IB1/2/3 tetramer cocktail. Table S3 contains sequence information for the heavy and light chain sequences derived from human B cells binding the IB2 tetramers. Table S4 contains sequence information for the heavy and light chain sequences derived from human B cells binding the IB3 tetramers. Table S5 contains data collection and refinement statistics for the crystal structures shown in Fig. 2.

We thank A. Macy, M. Baker, D. Alwan, and the Fred Hutchinson Cancer Research Center Flow Cytometry Shared Resource for technical assistance and S. Voght for careful reading of the manuscript. We thank the James B. Pendleton Charitable Trust for generous support of Formulatrix robotic instruments. We also thank J. McElrath for PBMC (Fred Hutchinson Cancer Research Center, Seattle, Washington).

X-ray diffraction data were collected at the Berkeley Center for Structural Biology beamlines 5.0.1 and 5.0.2, which are supported in part by the National Institutes of Health, National Institute of General Medical Sciences. The Advanced Light Source is supported by the Director, Office of Science, Office of Basic Energy Sciences of the US Department of Energy under contract DE-AC02-05CH11231. Research reported in this publication was supported by the National Institute of Allergy and Infectious Diseases of the National Institutes of Health under award numbers R01AI122912 (J.J. Taylor), R01AI081625 (L. Stamatatos), and R21AI127249 (A.T. McGuire). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

The authors declare no competing financial interests.

Author contributions: L. Stamatatos, A.T. McGuire, and J.J. Taylor designed the study, analyzed results, and wrote the manuscript; T. Bancroft conducted flow cytometry, cell sorting, adoptive transfer, and BCR sequencing experiments, analyzed results, and wrote the manuscript; B.L. DeBuysscher conducted and analyzed HEp-2 binding experiments; H.R. Steach assisted with cell sorting experiments; K.S. Fitzpatrick conducted and analyzed flow cytometry experiments; C.E. Weidle and M. Pancera determined crystal structures; A. Schwartz conducted adoptive transfer and in vitro stimulation experiments; A. Wall, M.D. Gray, and J. Feng, performed binding screens, cloned and purified recombinant antibodies or antigens, and conducted BLI experiments; M.M. Gewe and R.K. Strong provided the iglb12 c/scFv; and P.D. Skog, C. Doyle-Cooper, T. Ota, and D. Nemazee generated iglb12 IgH transgenics.