|Home | About | Journals | Submit | Contact Us | Français|
The HIV-1 broad neutralizing antibody (bnAb) 2F5 has been shown to be poly/self-reactive in vitro, and we previously demonstrated that targeted expression of its VDJ rearrangement alone was sufficient to trigger a profound B cell developmental blockade in 2F5 VH knockin (KI) mice, consistent with central deletion of 2F5 H chain-expressing B cells. Here, we generate a strain expressing the entire 2F5 bnAb specificity, 2F5 VHxVL KI mice, and find an even higher degree of tolerance control than observed in the 2F5 VH KI strain. Although B-cell development was severely impaired in 2F5 VHxVL KI animals, we demonstrate rescue of their B-cells when cultured in IL-7/BAFF. Intriguingly, even under these conditions, most rescued B-cell hybridomas produced mAbs that lacked HIV-1 Envelope (Env) reactivity due to editing of the 2F5 L chain, and the majority of rescued B-cells retained an anergic phenotype. Thus, when clonal deletion is circumvented, κ editing and anergy are additional safeguards preventing 2F5 VH/VL expression by immature/transitional B-cells. Importantly, 7% of rescued B-cells retained 2F5 VH/VL-expression and secreted Env-specific mAbs with HIV-1 neutralizing activity. This “partial” rescue was further corroborated in vivo, as reflected by the anergic phenotype of most rescued B-cells in 2F5 VHxVL KI × Eμ-bcl2 tg mice, and significant (yet modest) enrichment of Env-specific B-cells and serum Igs. The rescued 2F5 mAb-producing B-cell clones in this study are the first examples of in vivo-derived bone marrow precursors specifying HIV-1 bnAbs, and provide a starting point for design of strategies aimed at rescuing such B-cells.
The HIV-1 envelope (Env) has several conserved regions to which broadly neutralizing antibodies (bnAbs) bind, yet only rarely do infected persons make antibodies to these conserved neutralizing epitopes (1). When bnAbs do develop, they do so in only a minority of patients, and only after 2–3 years after infection (2–4). A primary goal of HIV-1 vaccine development is to develop strategies to induce bnAbs following immunization with Env immunogens, but so far, these strategies have been unsuccessful. Several hypotheses concerning the unusual features of Env have been proposed to account for the inability to routinely elicit bnAbs, either in the setting of natural infection or in vaccinated individuals. These include the genetic plasticity, complex nature, and masking of Env epitopes (5–10) and competitive suppression/induction of non-neutralizing Ab responses by highly immunogenic antigens on non-native Env structures (1, 3).
More recently, the role of host tolerance mechanisms in limiting the bnAb response has been proposed as an additional and/or alternative explanation (11, 12). This explanation has been put forward based on the fact that many rare HIV-1 bnAbs isolated thus far share unusual traits, including extensive polyreactivity, extensive somatic hypermutation, and/or H chain third complementarity determining regions (HCDR3s) expressed by B cells that are normally counterselected early in developmental maturation (reviewed in (1, 13)). In this regard, bnAbs 2F5 and 4E10, two human mAbs specific for the highly-conserved Env gp41 membrane proximal external region (MPER), have unusually long, charged, and hydrophobic HCDR3s and exhibit a high degree of polyreactivity to a variety of host molecules in vitro (11, 14, 15). These characteristics led us to hypothesize that their bnAb dearth was due to the routine tolerization of B cells from which these specificities could originate (11, 12).
Using gene targeting, we recently demonstrated in 2F5 VH knock-in (KI) mice that expression of the 2F5 H chain led to a profound blockade in development at the pre-B/immature B cell transition, consistent with 2F5 H chain-expressing immature BM B cells being subjected to clonal deletion due to their physiologically significant self-reactivity in vivo (16). While the 2F5 VH KI strain allowed us to establish the role of the 2F5 H chain in triggering B cell tolerance mechanisms, the B cells in this model are forced to pair their 2F5 H chains with endogenous mouse L chains, and thus do not allow us to monitor regulation of B cells bearing the original 2F5 VH/VL pair in vivo. Because we previously showed that most endogenous L chains in vitro do not complement MPER binding (16), the 2F5 VH model is therefore of limited value for inferring strategies aimed at eliciting MPER-specific bnAbs, i.e. whether B cells expressing the original 2F5 VDJ+VJ rearrangements can be rescued from tolerance controls, while still retaining specificities that are functionally comparable to the 2F5 mAb.
To determine how B cells expressing the original 2F5 mAb are limited by tolerance mechanisms in vivo and if they can be rescued from such controls while retaining functional specificity (i.e. neutralization potential), we generated a novel mouse strain whose B cells have the potential to express the original 2F5 VH/VL pair: the 2F5 “complete” KI mouse. We found that whereas essentially no arrest in B cell development was observed in the 2F5 VL KI strain, the BM B cell developmental arrest observed in the 2F5 VH strain was dramatically accentuated in 2F5 complete KI mice. These results are consistent with the hypothesis that BM B cells expressing the original 2F5 VH/VL pair, relative to those expressing 2F5 VH in combination with endogenous L chains, are subject to an even more stringent degree of tolerance controls and rules out the notion that lack of pairing with the original 2F5 L chain partner imparts the profound developmental blockade observed in 2F5 VH KI mice. Importantly, we also show that sIg+ BM B cells bearing 2F5 VH/VL pairs can be rescued from tolerance control in vitro, with the majority being developmentally arrested at the immature B cell stage, and express non-neutralizing Igs, due to loss of MPER specificity via replacement of their 2F5 L chains. Seven percent of rescued B cells retain 2F5 L chains, MPER binding and neutralizing activity, providing the first evidence of isolation of in vivo-derived BM B cell precursors specifying a bnAb. This partial in vitro rescue, limited by receptor editing and anergy, was further corroborated in vivo, as reflected by a modest, yet significant enrichment of MPER-specific B cells and serum Igs in 2F5 complete KI mice that over-express the anti-apoptotic gene bcl-2 specifically in the B-cell lineage.
A targeting vector containing the rearranged 2F5 VL gene inserted within the joining (Jκ) region of the murine Ig κ L chain locus was used to selectively disrupt the endogenous Jκ1, Jκ2, and Jκ3 segments. To generate the 3′ and 5′ homology arms of the vector, the Jκ region and its flanking upstream/downstream regions were isolated from a mouse C57BL/6 genomic library-derived BAC clone. The targeting backbone contained CAG-DTA and loxP-flanked Neo selection cassettes. Homologous recombination of ES cells was confirmed by Southern blotting using Bam HI, targeted ES clones were subjected to in vitro Cre recombinase-mediated deletion of the neo selection cassette, and four correctly targeted, neo− clones were injected into C57BL/6J Tyrc-2J blastocysts, one of which produced chimeric mice that transmitted the 2F5 VL insertion. 2F5 VL+/− and 2F5 VL+/+ genotypes were determined in the offspring by PCR primers specific for WT or targeted alleles and a primer common to both alleles (see Fig. 1 for vector targeting scheme and screening strategy). To detect Igκ transcripts in 2F5 VH+/− or control C57BL/6 mice, a murine Cκ-specific primer was used in combination with either a 2F5 VL-specific or a forward degenerate Vκ primer (that can detect most leader sequences including the 2F5 targeting construct’s VκOx1 leader sequence) in PCR amplifications of cDNA from purified splenic B-cells.
Eμ-bcl2- transgenic mice, under the control of the Eμ promoter (line C57BL/6-tgN (BCL2) 22 Wehi), have B-lineage specific over-expression of the human bcl-2 gene (17) were obtained from Jackson Labs. 2F5 VH KI mice (16) were either used alone, or crossbred with 2F5 VL KI mice to generate 2F5 complete KI mice. These strains and all other derivatives used in this study were housed in the Duke MSRB2 vivarium in a pathogen-free environment with 12h light/dark cycles at 20–25°C under AAALAC guidelines and in accordance with all Institutional Animal Care and Use Committee and Duke University Institutional Biosafety Committee-approved animal protocols.
For flow cytometric analysis, single cell suspensions from spleen, BM, LNs, peritoneal lavage, or PBLs were isolated from 6–12 week old naïve mice of various genotypes and phenotypically assessed using standard staining methods. Briefly, 106 cells were suspended in FACS Buffer containing 1X×PBS (pH7.2), 3% FBS (Sigma) and 0.01% Sodium Azide, and B cells were stained with pre-mixed combinations of fluorochrome-labeled mAbs at empirically-determined optimal concentrations, and total B cells were gated as singlet, live, lin−, CD19+ and/or B220+. All Abs were from BD unless otherwise stated. Primary labeled mAbs used were: Pacific Blue, APC, or Texas Red-conjugated α-B220 (clone RA3-6B2), PE-Cy7 a-CD19, FITC-conjugated α-IgD (clone 11–26), FITC, APC or PE-Cy7-conjugated α-IgM (clone 15F9), PE-conjugated α-CD21, PE-Cy7-labeled α-CD23 (eBiosciences), APC-conjugated α-CD93 (eBiosciences), FITC conjugated α-CD43, PE-conjugated α-BP-1, APC-labeled α-HSA, PE-conjugated α-kappa, and FITC-conjugated α-lambda1–3. Depending on the experiment, either Propidium Iodide (PI) or v-amine live/dead violet dye (Molecular Probes) was used to exclude dead cells, and B cell lineage excluding markers (lin−) included biotinylated mAbs against Thy1 (Abcam), F4/80 (Abcam), CD11c, Gr-1, TER-119, NK-1.1, CD4 and CD8. Other reagents used included APC-labeled MPER tetramers, used as previously described (18), Fc block (α-CD16/32), and for secondary staining, Texas-Red-conjugated Streptavadin. All FACS analysis was performed using a BD LSRII flow cytometer and data was acquired and analyzed using Cell Quest (BD) and FloJo (Tree Star) software, respectively.
Serum samples were collected from 6–12 wk old naïve mice, and serum Ab concentrations of all Ig subclasses were determined by Luminex analysis, using a Milliplex mouse Ig isotyping immunoassay kit and a BioRad Luminex Bead Array Reader, with baseline Ig levels set by subtracting values obtained in RAG-deficient animals. Quantitative measurements of serum IgM and IgG-specific binding to the 2F5 nominal MPER epitope peptide SP62 was determined by ELISA, as described previously (11, 14, 16), with one modification (for more sensitive detection): using AP-conjugated goat-anti-mouse μ or γ H chain-specific reagents (both from Southern Biotech) and attophos substrate (Promega) according to the manufacturer’s instructions. Endpoint titers were calculated as the reciprocal of the highest serum dilution used where >3 background binding ODs were still observed.
Culture-derived BM B cells from 2F5 complete (VH+/+x VL+/+ ) KI or wild-type (WT) littermate control mice were generated based on methodologies outlined in (19). Briefly, 8 wk old naïve mice were euthanized, BM was collected by repeated flushing of hind leg long bones with cold IMDM media, single cell suspensions were prepared by repeated pipetting, and viability was assessed by trypan blue exclusion staining. BM suspensions were then incubated briefly (15 min at 37 degrees) in 10 cm culture dishes to allow for cells to adhere. Non-adherent cells were then recovered by centrifugation, depleted of erythrocytes by ACK lysis, washed, transferred into T-75 flasks, and incubated at 7.5 × 105 for 4 days in IMDM media supplemented with recombinant mouse IL-7 (10 ng/ml), followed by washing, and re-plating in IMDM media supplemented with BAFF (20–100 ng/ml) for an additional 3–4 days.
To generate primary cultures from rescued 2F5 complete (VH+/+x VL+/+) KI B cells grown in the B cell culture system, electrofusions were performed using rescued culture-derived 2F5 complete KI bone marrow B cells described in previous methods section, and as phenotypically confirmed by flow cytometry (as shown in Fig. 3). The electrofusions were done based on previously-described techniques (20, 21). Briefly, NS0-Bcl2 myeloma fusion partner cells and culture-derived 2F5 VH+/+x VL+/+ KI B cells were washed twice with an isosmolar electrofusion buffer (Eppendorf), and fused at a 1:2 B cell: myeloma ratio using a PA-4000/PA-101 electrofusion apparatus with FE-20/800 electrode fusion chamber (Cyto Pulse Sciences, Inc.). Pre-fusion dielectrophoresis was performed with an alternating current voltage of 40V–60V at 1.4 MHz for 20 s. Cells were fused with a single square-wave direct current voltage of 525 V for 0.04 ms. Post-fusion dielectrophoresis was performed with an alternating current voltages of 50-55 V at 1.4 MHz for 30s. After fusion, cells were harvested and distributed into 96 well plates (flat-bottom) at 1,000 B cells per well and incubated in culture medium supplemented with 100 μM hypoxanthine, 0.4 μM aminopterin, and 16 μM thymidine.
For primary screens, culture supernatants from all wells derived from single colonies were assayed 14 days post-fusion, for Ab production as well as the ability to neutralize HIV-1. Ab production was determined using a sandwich ELISA to measure total Ig levels using purified goat a-mouse Ig (H+L) and AP-conjugated goat α-mouse κ+λ reagents (both from Southern Biotech) for capture and detection, respectively. Raw OD was converted to mg/ml by using defined concentrations of IgM or IgG mAbs to construct a standard curve. HIV-1 neutralization was determined using the TZM-bl pseudovirus infectivity assay as previously described (22), but specifically using the MN HIV-1 strain, which we have previously shown to be sensitive to bnAbs of both IgG and IgM isotypes. All Ab-secreting lines, (including all neutralizing clones), as well as a statistically relevant cohort of the non-secreting lines (determined by paired t test analysis), for comparison of Ig sequences with secreting line, were subcloned one more time.
Secondary screens of all subcloned lines included confirming isotypes/quantities of Abs produced in supernatants by Luminex analysis as described above for serum assays, and Ab reactivities to the 2F5 MPER epitope, cardiolipin, histones, and NIH-3T3 cytoplasmic/nuclear antigens in normalized culture supernatants over a full concentration curve were determined by ELISA. MPER-specific ELISA assays were as described above for serum assays, cardiolipin-specific ELISA reactivity assays were performed as previously described (11, 16), and histone-specific ELISAs were done using purified calf thymus histones (Worthington Biochemical Corp.) based on techniques described in (23). For ELISAs measuring reactivity to NIH-3T3 antigens, NIH-3T3 cells (2.5×104 cells/ml) were plated in 96-well plates in 200μl media (DMEM, 10% FCS, 50μM 2-mercaptoethanol, penicillin/streptomycin) and cultured for 24 hrs. Cells were fixed in cold methanol:acetone (1:1) at −20 °C for 10 min. Plates were then dried and stored at 4°C until use. NIH-3T3 cells were rehydrated and blocked overnight with PBS containing 0.5% BSA, 0.1% Tween-20 and 1.0% normal goat serum at 4°C. Hybridoma culture supernatants were screened against NIH-3T3 wells for 2 hrs at room temperature, and all washing and detection steps were done as for other ELISA assays. Finally, for certain subclones, including those with neutralizing activity, mAbs were purified from supernatants, and assayed either as pentamers or reduced to monomeric form and subjected to the assays as described above to determine various reactivities and neutralization, and FPLC analysis to confirm IgM form (data not shown).
For analysis of variable regions from all 2F5 KI hybridoma lines described in this study, cDNA was synthesized as described above, and their VH regions were amplified using a 2F5 VH-specific forward primer in combination with the above Cμ-specific reverse primer, whereas VL regions were amplified using a Cκ-specific reverse primer in combination with either a forward 2F5 VL-specific primer or a common degenerate forward murine Vκ forward primer ((24); which we determined also can detect the human 2F5 VL. PCR products obtained from amplifications of hybridoma VH and VL regions were cloned into pGEM-T Easy vectors (Promega), transformed in GC5 competent cells (CLP), and transformant DNA was isolated. In order to determine the closest germline gene of origin and to identify potentially somatically mutated residues, V regions were sequenced in both orientations and analyzed by aligning them with the 2F5 VH and VL regions of the original mutated 2F5 bnAb using L-ALIGN software, and “blasting” them against the public IgBLAST database of mouse Ig sequences at National Institutes of Health/National Center of Biotechnology Information (NCBI).
As the most stringent test of the hypothesis that HIV-1 bnAbs like 2F5 are under immunological tolerance control, we constructed 2F5 complete (VH+/+ × VL+/+) KI mice that have B cells expressing Igs containing the original, somatically mutated 2F5 VH/VL pair. We first constructed a novel KI line, 2F5 VL, generated by directed targeting of the original, somatically mutated 2F5 VκJκ into the mouse IgL kappa locus, as outlined in Figure 1. Four independent ES clones were confirmed to harbor the expected homologous replacement event at the Ig kappa locus (Fig. S1A), and were then used to generate hemizygous or homozygous progeny bearing germ-line transmission of the 2F5 VκJκ rearrangement, that could be distinguished using a PCR-based strategy (Fig. S1B). Appropriate splicing of the 2F5 VκJκ rearrangement to the mouse Cκ region in RNA from germ-line transmitted animals was determined by RT-PCR, (Fig. S1C). The 2F5 complete KI line was then generated by crossbreeding 2F5 VL KI mice with our previously described 2F5 VH KI line (16). Analogous to the targeting strategy used for generating the 2F5 VH KI line, in which all physiological diversification/modification processes at the H chain locus were retained, the strategy for generating 2F5 VL mice retained all flanking genetic elements at the kappa locus, including upstream Vκ genes, downstream Jκ4, Jκ5 mini-gene segments, and the Recombination Signal (RS). Thus, the 2F5 complete KI strain provides a physiologically-relevant in vivo system for determining how Ig diversification processes at either H or L chain loci, including potential receptor editing events, may impact tolerization of B cells expressing original, mutated 2F5 VH/VL pairs.
To determine what potential contribution the targeted 2F5 VκJκ rearrangement may have on 2F5’s ability to trigger tolerance mechanisms in vivo, we first compared BM and splenic B cell development of homozygous 2F5 complete KI mice with that of 2F5 VL, 2F5 VH, and C57BL/6 littermates (Fig. 2). In contrast to the B cell developmental block we reported in the 2F5 VH KI strain (whose B cells express 2F5 H chains paired with endogenous L chains), 2F5 VL KImice (whose B cells initially express 2F5 L chains paired with endogenous H chains) had no alterations in B cell development. Importantly, the developmental arrest observed in the 2F5 VH KI strain was further accentuated in 2F5 complete KI mice, including a virtually near-complete reduction of bone marrow B cells (>97%) at the pre-B to immature B stage (Fig. 2A, S2), further reductions in frequencies of splenic B cells, with accompanying lowered sIg, CD19, and B220 densities (Fig 2B), and significantly lower total B cells numbers (both in terms of frequency and absolute numbers) and lowered Ig densities in all other tissues assessed (Fig 2C and data not shown). These results are consistent with B cells expressing the original VH/VL pair, relative to those in 2F5 VH KI mice (expressing 2F5 VH in combination with endogenous L chains), being subjected to an even more profound level of clonal deletion in the BM and a higher degree of developmental arrest and/or functional silencing of residual peripheral B cells.
The profound development block at the BM pre-B to immature B cell transition in 2F5 complete KI mice is consistent with the highly efficient removal of immature B cells by clonal deletion upon expression of 2F5 VH/VL pairs on their surface. To examine if and to what extent normally deleted 2F5VH/VL-expressing BM B cells could be rescued and/or allowed to progress in B cell ontogeny, we used a stromal cell-independent culture system previously validated for efficiently rescuing differentiation of autoreactive Ig+ B cell clones, such as those from 3H9 KI mice (19). Using this system, BAFF/IL-7-cultured BM B cells from 2F5 complete KI mice differentiated from the pre-B cell stage to sIg+ immature and mature BM B cell compartments, but had lowered sIg and B220 densities relative to corresponding B cell subsets from littermates (Fig. 3; middle panels), a phenotype consistent with a high degree of developmental arrest/B cell anergy. A small fraction of B cells exhibited sIg densities comparable to those in littermates (Fig 3; right panels), and thus represented sIg+ B cells potentially completing BM differentiation and/or breaking anergy. We interpret these results to mean that B cells expressing 2F5 VH/VL pairs on their surfaces can be partially rescued from negative selection at the immature BM B cell stage, but full rescue of BM differentiation using the current in vitro culture conditions occurs in only a minor 2F5VH/VL-expressing B cell subset. These results also reinforce the notion that 2F5 bnAb-expressing B cell precursors are under highly stringent central tolerance controls, given that similar in vitro culture in IL-7/BAFF (and in the absence of tolerizing BM stroma) can efficiently rescue BM B cell subsets from KI mice expressing less autoreactive specificities relative to those expressing the original 2F5 VH/VL pair, including those from the low affinity 3H9 VH KI (19) and 2F5 VH KI strains (25).
Critical to the question of what controls bnAb expression is the ability to find the B cells that can make these specificities and to show that they exist in early B cell development. Thus, we screened 800 primary hybridoma cultures from rescued 2F5 complete KI BM B cells for cell growth and generated 77 cloned hybridoma lines. First, we screened all cloned lines for Ab production and identified 58 secreting lines, all of the IgM isotype. Supernatants from secreting B cell lines were then screened for neutralization and seven percent (4 hybridomas) were found to neutralize HIV-1 isolate B.MN, which has previously been shown to be sensitive to neutralization by human recombinant IgM 2F5 (26). IC50 neutralization titers of purified mAbs from these four hybridoma lines confirmed neutralization of nearly all isolates tested, including potent neutralization of B.MN (Table 1A). Finally, supernatants from the four neutralizing lines were then assayed for binding to the 2F5-specific MPER epitope peptide SP62, cardiolipin, histones, and NIH-3T3 nuclear/cytoplasmic antigens (Fig. 4), reactivities previously defined for the original human 2F5 mAb (11) as well as for m2F5, the “mouserized” 2F5 chimeric, recombinant Ab (i.e. bearing human V/mouse Cγ; (16)). All Abs retained complete binding of these 2F5-specific reactivities, including full reactivity with lipids and the MPER epitope (Tables II, III, Fig. 4), which recently have both been shown to be critical for 2F5’s neutralization ability (27). Also consistent with the neutralization ability of these hybridomas, cloning of their Ig VH and VL regions by RT-PCR using Ig-specific primers and sequencing analysis revealed that they all contained the original, knocked-in 2F5 VH and VL regions, free of any replacement somatic mutations (Tables II, III).
To determine the oligomeric nature of the IgM produced by the four neutralizing hybridoma lines, mAbs were purified and run on FPLC size exclusion chromatography and found to contain predominantly IgM pentamer (data not shown). As a representative experiment, we used one of these neutralizing mAbs, V3–1.4, to produce IgM monomer from pentamer by limited IgM reduction with dithiothreotal (DTT) and tested it as both IgM pentamer and monomer in neutralization assays, compared to the original IgG 2F5 mAb (Table IB). We found both the V3-1.4 IgM pentamer and monomer had identical neutralizing specificity as the original human IgG1 2F5, with the pentamer having slightly more potency. Overall, these data therefore definitely demonstrate that in vivo, H and L chains from 2F5 complete KI mice (containing the original knocked-in 2F5 VH and VL regions, respectively) are capable of functionally pairing with each other and reinforces the notion that immature B cells expressing functional 2F5 VH/VL pairs are absent due to their clonal deletion in the BM, but can be rescued by in vitro culture in IL-7 and BAFF.
The majority of secreting hybridoma lines (93%) derived from rescued naïve BM B cells of 2F5 complete KI mice produced Abs that were non-neutralizing (Table II), suggesting that they had lost one or more of the defined 2F5-specific reactivities, including lipid and/or MPER reactivity. To determine which reactivities were lost in these lines, we assayed all supernatants for binding to the 2F5-specific MPER epitope, cardiolipin, histones, and NIH-3T3 nuclear/cytoplasmic antigens (Fig. 4 and Tables II, III). Strikingly, we found all non-neutralizing secreting lines lost MPER reactivity. In addition, 48% of these lines also lost all reactivity to NIH-3T3 cytoplasmic/nuclear antigens, cardiolipin, and histones, although the remainder retained either partial binding across all specificities or at least partial reactivity to some of these specificities, suggesting that the stringent loss of MPER specificity observed in these lines was the main factor in limiting their neutralizing potential.
The increased survival window and/or removal from deletional controls afforded by in vitro culture in IL-7+BAFF may have allowed receptor editing of either the H chain (VH replacement) or L chain to occur, and such modifications may have impacted MPER specificity and neutralization potential of 2F5 complete KI B cells. To examine if and what Ig V region modifications were potentially associated with loss of MPER reactivity in these partially rescued, non-neutralizing 2F5 complete KI hybridoma lines, we cloned and sequenced their Ig VH and VL regions, as described above for the neutralizing clones (Fig. 5 and Table III). As with the neutralizing lines, all non-neutralizing− lines had H chains that retained their 2F5 VH regions, with no observable VH replacement events and either no mutations or silent mutations; thus H chain modification events likely didn’t account for lack of MPER binding. In striking contrast, all non-neutralizing hybridomas replaced their 2F5 L chains with endogenous L chains, and all replacements utilized Jκ4 or Jκ5 gene segments, indicative of secondary L chain rearrangements that were intriguingly restricted to the κ locus.
Interestingly, secondary L chain rearrangement events observed in the non-neutralizing clones involved restricted V segment usage, with the amount of V gene restriction correlating with the degree that defined reactivities measured in our assays other than MPER reactivity could be eliminated. In particular, two Vκ gene segments, the Jκ-proximal members 8–19 and 21–4 were the most utilized L chain genes (used in 22% and 30% of all secreting clones, respectively; Fig. 5 and Table III), and correlated with partial and full loss of 3T3/cardiolipin/histone binding, respectively. It is also intriguing that the Vκ21 family member 21–4 (21D) is the most effective and utilized overall L chain editor of the 2F5 H chain, as seen in the 3H9/56R H chain KI model (28). Conversely, it is also interesting that many different V gene segments within the Vκ4 family used were largely ineffective L chain “editors” of 2F5 H chain self-reactivity, since this family has been observed to infrequently pair with the 3H9 H chain and is largely ineffective at removing deletion of α-DNA reactive B cells in vivo conferred by 3H9, relative to the Vκ8 family (29–31).
We also cloned and sequenced Ig VH and VL regions from non-secreting lines, and interestingly, found that several retained the knocked-in 2F5 L chain (Table S1). Additionally, we found several others had edited to Vκ11–27, Vκ12–42, and Vκ12–41, three close mouse VL gene homologues of the 2F5 VL germline gene Vκ1–13 (Table S1), the latter two segments which we have cloned and shown can complement all defined 2F5 reactivities, including at partial MPER binding (data not shown). Finally, several other non-secreting clones also used increased frequencies of edited L chains with V segments associated with retention of binding to or more reactivities of 3T3, cardiolipin, or histone binding (for example, overrepresented usage of L chains containing 3T3/cardiolipin/histone-complementing Vκ4 family members; Table III, S1). Finally, non-secreting lines lacked usage of the editor L chain Vκ21–4, the family most frequently used by secreting clones. Taken together, these suggest that Vκ usage between secreting and non-secreting lines differs significantly and raises the intriguing possibility that non-secreting lines potentially represent rescued B cell clones that were under a higher degree of clonal anergy/functional unresponsiveness than those from which secreting lines were derived.
Overall, our reactivity/Ig usage analysis of rescued 2F5 complete KI hybridomas clearly demonstrates that when removed from deletional controls in vitro, most autoreactive 2F5VH/VL-expressing BM B cells replace their original 2F5 L chains with endogenous L chains, and this event most strongly correlates with loss of MPER reactivity and neutralization capability.
To further examine the impact that replacing the 2F5 L chain from the original 2F5 H/L chain pair has on B-cell clonal deletion and/or anergy in vivo, we compared B cell development in 2F5 complete (VH+/+ × VL+/+ ) completeKI mice (homozygous for the targeted 2F5 VJ insertion), with that in 2F5 VH+/+ × VL+/− KI mice (hemizygous for this insertion), which have the potential to use a fully intact alternate, endogenous kappa L chain allele. Interestingly, we found that 2F5 VH+/+ × VL+/− KI mice, relative to 2F5 VH+/+ × VL+/+ completeKI mice, had higher frequencies of total BM and splenic B cells (Fig. 6A and B), thus partially rescued B cells from deletion, yet still exhibited comparable lowered sIgM densities (Fig 6C), a result similar to that we observed in our in vitro rescue of B cells from 2F5 complete KI mice.
Because partial rescue of B cell differentiation and replacement of the 2F5 L chain in vitro correlates with complete loss of 2F5 nominal MPER epitope specificity (and to a lesser extent, loss of cardiolipin, histone, and NIH-3T3 binding; Fig. 4 and Tables II, III), this suggests that these measured reactivities (combined) are not the sole contributors in the deletion of 2F5-expressing B cells. However, a partial role for L chain-dependent MPER reactivity in 2F5-expressing B cell deletion is suggested by the observation that 2F5 complete KI mice (engineered to express 2F5 H chain/L chain pairs shown to confer lipid and MPER reactivity), have an accentuated B cell developmental arrest (Fig. 2B) relative to 2F5 VH KI mice (engineered to express 2F5 H chains paired with endogenous L chain pairs, and previously demonstrated by random in vitro pairing to be capable of complementing lipid, but not MPER specificity (16)).
To establish the contribution of MPER reactivity in the tolerization of 2F5-expressing B cells in vivo, we measured normalized α-MPER endpoint 2F5 complete (VH+/+ X VL+/+), 2F5 VH+/+ X VL+/−, and 2F5 VH KI models (Fig. 6D), which should be revealed as a gradient, with the greatest enrichment expected in serum IgM from 2F5 complete 2F5 VH+/+ × VL+/+ KI mice. Indeed, as measured by normalized α-MPER endpoint titers, IgM-specific sera from 2F5 VH KI and 2F5 VH+/+ × VL+/− KI mice had low, detectable MPER reactivities, but significantly higher reactivities were observed in IgM-specific sera from 2F5 VH+/+ × VL+/+ KI mice (Fig. 6D). These results were further corroborated by staining residual total splenic B cell populations from 2F5 VH, 2F5 VH+/+ × VL+/− and 2F5 complete KI strains with MPER-specific tetramer reagents (18), where again we saw a significant relative increase in the frequencies of MPER+ B cells in 2F5 complete KI mice, relative to 2F5 VH KI and strains (Fig. 6E). Interestingly, despite the relative enrichment of MPER-specific splenic B cells in 2F5 complete KI mice, this population only represented <10% of the total B220+CD19+ splenic population, suggesting strong counterselection for MPER+ VH/VL pairs in the periphery, and selection/rescue of MPER-, anergic peripheral B cells which likely have been subjected to additional tolerance mechanisms modifying H chain and/or L chain specificity.
Taken together, these results provide additional evidence that 2F5 H+L chain-dependent interactions with self-antigen(s) mimicking the nominal 2F5 MPER epitope are at least partly responsible for tolerizing 2F5-expressing B cells in vivo.
Our studies of cultured 2F5 complete KI BM B cells demonstrate that extensive L chain editing occurs when the survival window is extended, yet only partially reduce 2F5’s self-reactivity, based on the fact that most cells appear developmentally arrested/anergic (Fig. 3). To confirm this, and as a first step in devising potential in vivo strategies aimed at rescuing 2F5-specific B cells, we crossed the 2F5 complete KI strain with Eμ-bcl2 tg mice (i.e. expressing the anti-apoptotic gene bcl2 only in the B cell lineage), and examined B cell development or enumerated MPER-specific B cells and serum Igs (Fig. 7). Analogous to the phenotype of complete KI BM B cells that were cultured in vitro with BAFF+IL-7, we find that a subset of BM and splenic B cells in 2F5 complete KI × Eμ-bcl2 tg mice cells are rescued from deletion (based on B cell frequencies; Fig. 7A), but not from their anergic/unresponsive phenotype (based on comparable reduced MFIs in 2F5 complete KI mice sufficient or deficient for Eμ-bcl2 tg mice). Further confirming the partial rescue phenotype in 2F5 complete KI × Eμ-bcl2 tg mice were significant increases in MPER-2F5 epitope-specific splenic B cells (Fig. 7B) and serum Igs (Fig. 7C).
In this study, we make two important observations regarding the regulation of B cells expressing the HIV-1 Env gp41 2F5 bnAb. Under normal selection conditions, we show in a novel murine line, 2F5 complete (VH/VL) KI mice, that pairing of the 2F5 H chain with its cognate L chain in vivo results in stringent tolerance controls. Secondly, under conditions where apoptosis of BM B cells is circumvented, we show that sIg+ B cells bearing self-reactive and functional 2F5 VH/VL pairs can be rescued. We also find that extensive L chain editing and anergy prevent efficient 2F5 VH/VL expression, thus revealing that additional tolerance mechanisms in 2F5 complete (VH/VL) KI mice control 2F5 mAb expression.
The profound reduction of Ig+ 2F5 bnAb-expressing-bone marrow B cells in 2F5 complete (VH+/+ × VL+/+) KI mice confirms that the predominant tolerizing mechanism/checkpoint of 2F5-expressing B cells is clonal deletion in the BM, and is similar to that seen in other KI models made with autoantibody Igs exhibiting high degrees of self-/polyreactivity in vivo (reviewed in (32, 33)). Additionally, since this strain was engineered to produce a pre-antigenic B cell repertoire initially comprised of H/L chain pairs containing the original (mutated) 2F5 mAb’s VH and VL regions, this consolidates the notion that this Ig’s inherent self-reactivity triggered the profound developmental blockade, regardless of when in B cell ontogeny the original 2F5 H chain acquired somatic mutations, and irrespective of if/what extent ineffective pairing of mutated 2F5 H chains with surrogate or conventional L chains potentially occurred in the 2F5 VH KI pre-antigenic repertoire. This study also defines the relative impact of 2F5 H and L chains in the development of B cells expressing them: the dominance of the 2F5 H chain is demonstrated in that its in vivo expression, and not the L chain’s, results in profound loss of BM B cells. On the other hand, a definitive, albeit additive role for the 2F5 L chain is supported by the demonstration that 2F5 H+L chain co-expression in vivo enhances the extent of clonal deletion/anergy in immature B cells. In this context, given the extensive amount of L chain editing in residual/rescued B cells from 2F5 complete KI mice, it will be of interest to explore the possibility that normal B cell development in 2F5 VL KI mice is due to extensive editing of its 2F5 L chain, and if so, to determine if this occurs either in response to 2F5 L chain associations with self-reactive, endogenous H chains, or alternatively, results from general difficulties the 2F5 L chain has pairing with endogenous H chains.
Our study’s demonstration that B cells producing the HIV-1 bnAb 2F5 from the BM of 2F5 complete (VH xVL) KI mice, can be recovered in vitro in the absence of the stromal tolerizing environment, is critical for three reasons. First, it provides evidence, under normal selection conditions in vivo, that the absence of B cell precursors capable of expressing 2F5 VH/VL pairs with broadly neutralizing activity is due to their de novo synthesis prior to their clonal deletion, rather than not having been initially made. Secondly, it provides the first example of isolated in vivo-derived BM B cell precursors specifying a HIV-1 bnAb, and provides a starting basis to manipulate the immune system such that these desirable clones may be elicited. Thirdly, it reveals that under conditions where deletional controls are removed, two additional tolerance mechanisms remain that limit the efficient rescue of these bnAb-specific B cell precursors: extensive L chain editing, resulting in loss of MPER specificity and neutralization (Table III, Fig. 6), and anergy (Figs. 3, 6B,C).
With respect to the hurdle of overcoming anergy, our in vivo studies of BM B cells from 2F5 VH complete × Eμ-bcl2 mice confirm the above in vitro studies, demonstrating the rescue of many 2F5-expressing B cells from central deletion, but not from anergy (Fig. 7A), and is also consistent with previous studies of enforced B cell-specific overexpression of bcl2 in vivo in certain autoreactive Ig transgenic systems such as the membrane-bound HEL tg model, where central deletion can be at least partially circumvented, but a large fraction of rescued autoreactive B cell clones remain phenotypically anergic (34). However, the fact that we can rescue a significant MPER-specific IgG endpoint titer in naïve 2F5 complete KI × Eμ-bcl2 tg mice, relative to 2F5 complete KI mice (Fig. 7C) does also suggest that a subset of B cells can break anergy in vivo, and in this context, it will be of particular interest to correlate rescued B cell subsets with serum MPER reactivity (and potentially neutralization titers) in immunized 2F5 complete KI × Eμ-bcl2 tg animals.
Previously, it has been shown that the Eμ-bcl2 transgene not only increases the survival window of self-reactive bone marrow B cells but also allows them to concomitantly increase their frequency of κ and λ L chain editing events (35). The anergic phenotype observed in our in vitro cultured B cells was also associated with extensive (but largely ineffective) κ-specific L chain editing events under these rescue conditions. As expected, over-expression of the Eμ-bcl2 transgene in our studies increased L chain editing in WT mice (as demonstrated by decreased κ/λ ratios and increased κ+λ ratios), but interestingly had little effect on λ L chain editing in the context of 2F5 complete KI B cells (based on their unaltered κ/λ ratios), suggesting that either LC editing doesn’t occur, or, as seen in hybridomas derived from in vitro- rescued 2F5 complete KI B cells, only κ-specific secondary rearrangement events can occur and/or are preferentially selected for. In support of this latter possibility, studies in 2F5 VH KI mice suggest that the 2F5 H chain also imparts initial pairing constraints to restricted Vκ families and disfavors λ partners (manuscript in preparation). Regardless of if/what types of L chain editing events occur in bcl2-overexpressing 2F5 complete KI B-cells and the causal link these events have with anergy, these results reinforce our in vitro findings that rescued 2F5 VH/VL-expressing B cells cannot reduce their self-reactivity sufficiently to overcome their anergic (functionally unresponsive) phenotype.
The unusually high extent of L chain editing observed in 2F5 complete KI mice could be influenced by two potential factors. One is compensation for the lack of observed VH replacement events in hybridomas derived from our 2F5 KI models (Tables II, III). Consistent with this possibility is the fact that the original (mutated) 2F5 VH bears “atypical” embedded/nonamer motifs (i.e. cryptic Recombination Signal Sequences (cRSS)) in the well-known VH 3′ FRW3 site, where cRSS have been experimentally linked with VH replacement of the 3H9 H chain (36, 37). In particular, even though the 2F5 and 3H9 VH 3′ FRW3 regions share identical embedded heptamers, the 2F5 VH 3′ FRW3 lacks the consensus embedded nonamer found in the 3H9 3′ FRW3 region, as well as many other VH genes examined (Table S2). Another, likely possibility for the high rate of L chain editing in 2F5 complete KI mice relates to the largely unsuccessful nature of sequential κ L chain editing events in overcoming the threshold of sensitivity to anergy, for example in situations where increased attempts are permitted due to an increased survival window (Table III, Fig. 3, ,7)7) and/or there is increased availability of endogenous κ L chain elements (Fig. 6B, C).
The unsuccessful κ L chain editing pattern observed in rescued 2F5 complete KI B cells (in which only a highly-restricted set of κ L chain partners i.e. such as Vκ21-4 are used, and only partially mitigate 2F5 H chain reactivity), is analogous to the situation in KI mice bearing the dominant, high affinity anti-DNA H chain 3H9-76R, which, like 2F5, contains multiple HCDR3 positively-charged residues, and where almost all L chains are ineffective at vetoing H chain-encoded DNA reactivity. That the pattern of L chain editing in 2F5 complete KI mice appears to be κ-restricted however, is distinct from other situations where editing to the λ locus effectively vetoes H chain self-reactivity, and may reflect an inherent inability of λLCs to either mitigate 2F5 H chain self-reactivity, or alternatively, properly pair with 2F5 H chains). Further investigation using 2F5 VH and VL KI models into the contribution of the individual 2F5 H and L chains in inducing self-reactivity and H/L chain pairing constraints and the potential role of L chain editing in these two processes should be useful in this regard.
A final important finding of this study is that several specificities within the spectrum of 2F5’s overall polyreactivity likely contribute in its in vivo tolerogenicity. The contribution of the original 2F5 L chain in specifying self-reactivity for a self-antigenic component mimicking the nominal 2F5 MPER epitope, is corroborated by several lines of evidence, including: a) 2F5 L+H chain co-expression conferring co-ordinate MPER specificity and enhanced degree of clonal deletion in immature BM B cells in vivo, 2) many rescued secreting hybridoma lines derived from 2F5 complete KI BM B cells use endogenous L chain editors that completely eliminate MPER binding in vitro, and 3) 2F5 complete KI BM B cell lines follow a hierarchy wherein higher frequencies of L chain editing events using putative MPER-complementing endogenous L chains, occur in non-secreting hybridomas, relative to secreting (potentially less anergic) ones (Table S1). However, our study also implies an important role for 2F5 H chain-specific interactions with one or more self-antigens, since most 2F5 H chain-expressing-B cells in this study are only partially rescued from negative selection, even when “editor” L chains remove all measurable reactivities. Because such 2F5 H chain-specific interactions cannot be measured in our in vitro reactivity assay, but may either be with the same MPER-mimicking self-antigen(s) recognized by the original 2F5 HC/LC, and/or to distinct self-ag(s) altogether, further studies aimed at comprehensively identifying the physiologically-relevant in vivo self-antigen target(s) of 2F5 will be critical.
Flow cytometry was performed by the Duke Center for AIDS Research, and the Center for HIV/AIDS Vaccine Immunology. We thank John F. Whitesides, Patrice McDermott, and Letealia M. Oliver for expert technical assistance in flow cytometry, Greg Sempowski and Jeff Hale for expert advice and technical assistance with Luminex assays, Brad Lockwood for affinity purification of mAbs from hybridoma supernatants, and Shi-Mao Xia for performing neutralization assays.
1This work was conducted as part of the Collaboration for AIDS Vaccine Discovery (CAVD) with support from a Bill and Melinda Gates Foundation grant to BFH (38643) and NIH, NIAID grants to BFH (AI067854), GK (AI081579) and LV (AI087202).