Conditional antibody expression to avoid central B cell deletion in humanized HIV-1 vaccine mouse models

Significance

Mouse models can provide fast and cost-effective systems to test HIV-1 vaccine candidates at the preclinical stage. To serve this purpose, mouse models are engineered to express the precursors of human broadly neutralizing antibodies (bnAbs) against diverse HIV-1 strains. Immunization of such mouse models can evaluate the ability of vaccines to mature the precursor antibodies into bnAbs. However, due to the unusual properties of bnAbs, mouse models expressing their precursors often have B cell developmental defects. Herein, we devised and validated a strategy to address this problem. This approach could also facilitate the expression of other clinically relevant antibodies in mature B cells in transgenic mice; immunization of such mice could be used to generate novel antibodies with desirable properties.

Abstract

HIV-1 vaccine development aims to elicit broadly neutralizing antibodies (bnAbs) against diverse viral strains. In some HIV-1–infected individuals, bnAbs evolved from precursor antibodies through affinity maturation. To induce bnAbs, a vaccine must mediate a similar antibody maturation process. One way to test a vaccine is to immunize mouse models that express human bnAb precursors and assess whether the vaccine can convert precursor antibodies into bnAbs. A major problem with such mouse models is that bnAb expression often hinders B cell development. Such developmental blocks may be attributed to the unusual properties of bnAb variable regions, such as poly-reactivity and long antigen-binding loops, which are usually under negative selection during primary B cell development. To address this problem, we devised a method to circumvent such B cell developmental blocks by expressing bnAbs conditionally in mature B cells. We validated this method by expressing the unmutated common ancestor (UCA) of the human VRC26 bnAb in transgenic mice. Constitutive expression of the VRC26UCA led to developmental arrest of B cell progenitors in bone marrow; poly-reactivity of the VRC26UCA and poor pairing of the VRC26UCA heavy chain with the mouse surrogate light chain may contribute to this phenotype. The conditional expression strategy bypassed the impediment to VRC26UCA B cell development, enabling the expression of VRC26UCA in mature B cells. This approach should be generally applicable for expressing other bnAbs that are under negative selection during B cell development.

Some HIV-1–infected individuals develop broadly neutralizing antibodies (bnAbs) against diverse viral strains (1, 2). A goal of HIV-1 vaccine development is to elicit comparable bnAbs. The bnAb developmental pathway during natural infection can serve as a blueprint for immunogen design. In bnAb donors, each bnAb lineage evolved from a germline antibody precursor or unmutated common ancestor (UCA). The maturation of a UCA into bnAb entails extensive somatic hypermutation. Similarly, conversion of a UCA into bnAb via vaccination may require a series of immunogens to guide the maturation process. These immunogens can be tested in transgenic mice that express bnAb precursors (3).

To express bnAb precursors in transgenic mice, a standard method is to integrate, or knock in (KI), the preassembled variable region (V) exons of the immunoglobulin (Ig) heavy chain (HC) and light chain (LC) of the bnAb precursor into the mouse J_H and Jκ loci, respectively (3). By feedback regulation, preassembled V exons inhibit rearrangement of the endogenous Ig loci (4); in theory, all of the B cells in such transgenic mice should express the KI antibody. However, a common problem with such KI mice is that B cells expressing certain bnAbs are deleted or rendered anergic (5–10). Such phenotypes are characteristic of B cells expressing autoreactive antibodies. Some bnAbs are indeed poly-reactive or autoreactive (11–16). To eliminate the autoreactivity, V_H replacement and LC receptor editing can allow for selection of B cells that have deleted the HC and LC V exons of autoreactive antibodies and replaced them with endogenously assembled V exons (17–22). Clonal deletion or anergy can eliminate or inactivate autoreactive B cells (23–27). Exclusion from B cell follicles can preclude autoreactive B cells from the recirculating mature B cell compartment (28, 29). These tolerance control mechanisms hinder the expression of certain bnAbs in KI mice (30).

We have developed a method to overcome the restriction on KI bnAb expression in KI mice. Thus, we expressed a human bnAb UCA conditionally in mature B cells, thereby circumventing the tolerance control checkpoints during B cell maturation. As a proof-of-principle experiment, we used this method to express the UCA of human VRC26 bnAb (31, 32). The VRC26 bnAb interacts with the V1V2 region at the apex of the HIV-1 envelop. The HC of VRC26 features a long complementarity determining region 3 (CDR H3), consisting of 37 amino acids. The extended CDR H3 penetrates through the thick glycan layer on the HIV-1 envelop to contact a conserved peptide epitope underneath (33). Antibodies with such a long CDR H3 are extremely rare in both human and mouse (34, 35), likely due at least in part to negative selection during B cell development (36). We anticipated similar obstacles for VRC26 expression in mice. In this study, we confirmed strong negative selection against VRC26 expression in transgenic mice, studied the underlying cause, and tested our conditional expression strategy to address the problem.

Results

We constructed two mouse models for this study, constitutive and conditional expression models. In the constitutive expression model, the V exons of the VRC26UCA HC and LC were integrated into the mouse J_H and Jκ locus, respectively (Fig. 1A and SI Appendix, Fig. S1). This conventional KI mouse should reveal potential negative selection against the VRC26UCA and serve as a reference point for the conditional expression model. The design of the conditional expression model was adapted from a method to switch antibody V regions in memory B cells (37). In this model, two V exons were integrated in tandem into the mouse J_H locus (Fig. 1B and SI Appendix, Fig. S1). The V exon proximal to the μ constant region (Cμ) was derived from the HC of a mouse antibody that supports normal B cell development (38). The intronic enhancer (iEμ), immediately downstream of the mouse V exon, should preferentially activate the expression of the mouse V exon versus that of the upstream VRC26UCA V exon. In addition, two polyadenylation sites (pAx2) can truncate transcripts from the upstream VRC26UCA V exon (Fig. 1C). These two measures were designed to ensure that the mouse HC is expressed in B cell progenitors and drives B cell maturation, without interference from the VRC26UCA. For this reason, we refer to the mouse antibody as the “driver.” Two loxP sites flank the driver V exon and polyadenylation sites (Fig. 1C). This mouse model also harbors a CD21-cre transgene (39), which becomes active in mature B cells and excises the driver V exon and the polyadenylation sites. Consequently, VRC26UCA replaces the driver for expression in mature B cells. This strategy should enable VRC26UCA expression to bypass the B cell maturation block. The conditional expression method is applicable for both HC and LC. However, for the present study on VRC26UCA, we hypothesized that the long CDR H3 would be the primary target for negative selection. To test this idea, we expressed the HC of VRC26UCA conditionally, but the LC constitutively (Fig. 1B). With this approach, phenotypic differences between the conditional and constitutive models should reflect differential HC expression.

Fig. 1.

Diagram of the constitutive and the conditional expression model of VRC26UCA. (A) Illustration of the constitutive expression model of VRC26UCA. The diagram is for illustration purpose only and is not drawn to scale. The genetic organization of the IgH and Igκ locus of the B1-8/3-83 KI mice is similar to that of the constitutive expression model VRC26UCA (43). For details of the B1-8/3-83 KI mice, see the original report of the mouse line (43) and Materials and Methods in this study. (B) Illustration of the conditional expression model of VRC26UCA. (C) Illustration of the conditional expression cassette.

Both the constitutive and the conditional expression models were derived from one integration event in embryonic stem (ES) cells (SI Appendix, Fig. S1). Through homologous recombination, a conditional expression cassette containing the VRC26UCA V exon and driver V exon replaced the mouse J_H region. The integrated cassette included three loxP sites. Under limiting cre recombinase conditions, partial recombination between different pairs of loxP sites gave rise to ES clones for either the constitutive or the conditional expression model (SI Appendix, Fig. S1). These two types of derivative ES clones were separately injected into Rag2-deficient blastocysts to generate chimeric mice (40). Because Rag2 is essential to V(D)J recombination, all of the B and T cells in the chimeric mice originate from the injected Rag2-sufficient ES cells. We refer to this technique as Rag2-deficient blastocyst complementation (RDBC) (40). For the constitutive expression model, we further bred the chimeric mice for germline transmission, allowing all experiments with the constitutive expression model to be carried out with germline KI mice. However, germline breeding is not suitable for generating the conditional expression model, because the “floxed” driver V exon was frequently deleted among the progeny, presumably due to leaky expression of the CD21-cre transgene (39) in germ cells. For this reason, all of the experiments with the conditional expression model were performed with RDBC chimeric mice. The RDBC approach has the added advantage of obviating the inefficient breeding step of bringing together three genetic components on separate chromosomes: HC KI, LC KI, and CD21-cre transgene.

B Cell Deletion in the Constitutive Expression Model of VRC26UCA.

We used allotypic markers to monitor VRC26UCA expression in the constitutive expression model. The two IgH alleles of the mouse model were derived from the 129/Sv and C57BL/6 mouse strains, respectively. The KI VRC26UCA HC resides on the IgH^a allele from 129/Sv strain; the IgH^b allele from C57BL/6 strain is unmodified (Fig. 1A). In naïve B cells, the VRC26UCA HC was expressed as IgM^a, whereas mouse HC was expressed as IgM^b. In the wild-type F1 mouse, equal proportions of splenic B cells expressed IgM^a or IgM^b (Fig. 2A). Expression of a prerearranged V exon from the IgH^a allele should allelically exclude the IgH^b allele; consequently, most B cells in such KI mice should express IgM^a. The B1-8/3-83 KI mouse exemplified this phenomenon (Fig. 2B). This mouse line harbors prerearranged V exons for the B1-8HC and 3-83LC at the J_H and Jκ loci, respectively (41–43). B1-8 is a mouse antibody against 4-hydroxy-3-nitrophenylacetyl hapten (44, 45); 3-83 is a mouse antibody against H-2K^k class I MHC (46). The combination of B1-8HC and 3-83LC forms an “innocent” antibody that supports B cell development (43). As in the VRC26UCA KI mouse (Fig. 1A), the V exon of the B1-8HC is on the IgH^a allele, and the other IgH^b allele is unmodified. Splenic B cells from the B1-8/3-83 KI mouse expressed exclusively IgM^a (Fig. 2B). In contrast, splenic B cells in the constitutive expression model of VRC26UCA expressed predominantly IgM^b. In 7-wk-old mice, some IgM^a B cells were detectable, but their surface IgM^a levels were abnormally low (Fig. 2C). By 37 wk of age, even fewer B cells were IgM^a+ (Fig. 2D). Another sign of a B cell anomaly was that the VRC26UCA KI mouse had reduced numbers of B cells in the spleen relative to F1 mouse (compare SI Appendix, Fig. S2 E and F with SI Appendix, Fig. S2C). Using T cell number as an internal reference, the ratio of B cells versus T cells in the F1 mouse was 1.1; the ratio was reduced to 0.28 in the 7-wk-old VRC26UCA KI mouse. A KI mouse with VRC26UCA HC alone had a similar, but less severe, phenotype (Fig. 2 I and J and SI Appendix, Fig. S2 G and H). On the other hand, a KI mouse with only the VRC26UCA LC had nearly normal B cell numbers (SI Appendix, Fig. S2 I and J).

Fig. 2.

FACS analysis of splenic B cells from the constitutive expression model of VRC26UCA. All of the plots are gated on B cells. Gating strategy is illustrated in SI Appendix, Fig. S2. The staining antibodies are indicated next to the axis. The surface phenotype and percentage of the gated population relative to total events in the plot are indicated next to the gates. (A–D and I–L) FACS staining patterns of IgM^a versus IgM^b. (E–H and M–P) FACS staining patterns of Igκ versus Igλ. Mouse genotype is shown above each plot: F1, 129/Sv × C57BL/6 wild-type F1 mouse; B1-8/3-83, KI mice with B1-8HC and 3-83LC; VRC26H/L, the constitutive expression model of VRC26UCA HC and LC; VRC26H, KI mice for constitutive expression of the VRC26UCA HC; VRC26L, KI mice for constitutive expression of the VRC26UCA LC. The ages of the VRC26UCA KI mice are indicated above the plot.

The pattern of LC expression in the VRC26UCA KI mouse was also unusual. Unlike IgH, Igκ alleles in this model were not distinguishable by allotypic markers. Igκ⁺ B cells could express either the VRC26UCA LC or a mouse LC, but all Igλ⁺ B cells expressed mouse LC. In wild-type mice, about 3 to 5% of B cells expressed Igλ, as shown in F1 mouse (Fig. 2E). Because of feed-back regulation, expression of a prerearranged V exon from the κ locus, as in the B1-8/3-83 KI mouse, further diminished Igλ⁺ B cells (Fig. 2F). In contrast, the VRC26UCA KI mouse contained higher fractions of Igλ⁺ B cells than the F1 mouse (Fig. 2 G and H). Within the Igλ⁺ population in a 7-wk-old VRC26UCA KI mouse, some B cells expressed abnormally low levels of Igλ (Fig. 2G). These Igλ^lo B cells likely corresponded to those with low surface IgM^a expression (Fig. 2C). In concert with the loss of IgM^a B cells in older mice, Igλ^lo B cells also disappeared (Fig. 2H). Again, the KI mouse with VRC26UCA HC alone had a similar phenotype as the complete VRC26UCA KI mouse, but with a more prominent Igλ^lo population (Fig. 2M), correlating with larger numbers of IgM^a,lo B cells (Fig. 2I). The Igκ and Igλ loci were unmodified in the VRC26UCA HC mouse. The increase in Igλ⁺ B cells, and the distinct IgM^a,loIgλ^lo population, may reflect preferential pairing of the VRC26UCA HC with mouse λ LC, or a paucity of mouse κ LCs able to pair with the VRC26UCA HC. Expression of the VRC26UCA LC also caused a noticeable increase in Igλ⁺ B cells, presumably due to deletion of the KI LC via receptor editing in some B cells (Fig. 2 O and P).

These analyses showed strong negative selection against the VRC26UCA, primarily due to its HC. V_H replacement can delete rearranged V exons in the IgH locus (17, 47). Most V_H segments contain a cryptic recombination signal sequence (RSS), which can undergo V(D)J recombination with upstream canonical RSSs to delete the KI V_H portion of the V(D)J exon and replace it with upstream V_H sequence (17, 47). To test whether V_H replacement accounted for the loss of VRC26UCA HC expression, we generated hybridomas from splenic B cells of the constitutive mouse model. Consistent with the FACS analysis, the majority of hybridomas secreted IgM^b allotype antibodies (SI Appendix, Fig. S3 A and B). Based on PCR analysis, all of the examined hybridomas, including those expressing IgM^a allotype antibody, lost the region upstream of the cryptic RSS in the VRC26UCA HC (SI Appendix, Fig. S3 C and D). Using a digestion circularization-PCR (DC-PCR) method, we identified the sequences that substituted for the VRC26UCA HC in some hybridomas (SI Appendix, Fig. S3 E–H and Table S1). These recombination products were typical V_H replacement events, in which the cryptic RSS in the VRC26UCA HC mediated V(D)J recombination with upstream mouse V_H or mouse D segments. In hybridomas producing IgM^b allotype antibodies, the V_H replacement products were nonproductive, including D to VRC26UCA HC rearrangements, out-of-frame V_H-D-VRC26UCA HC rearrangements, or in-frame V_H-D-VRC26UCA HC rearrangements with stop codons (SI Appendix, Table S1). One IgG1^b hybridoma contained an in-frame V_H-VRC26UCA HC recombination event. It was unclear why this HC was not expressed as an IgM^a allotype antibody; one possibility was that the V_H replacement product did not fold properly. In hybridomas producing IgM^a allotype antibodies, V_H-D-VRC26UCA HC recombination events were in-frame (SI Appendix, Table S1). Thus, in all examined cases, V_H replacement eliminated the VRC26UCA HC.

Receptor editing can delete rearranged V exons in the Igκ locus, leading to the rearrangement of the other Igκ allele or the Igλ locus (19–22). This process likely contributed to the increase in Igλ⁺ B cells in the VRC26UCA KI mouse. Deletion of the VRC26UCA LC via receptor editing could also result in the expression of the other Igκ allele. Without an allotypic marker to distinguish the two Igκ alleles, we could not assess the frequency of such events with FACS analysis. Thus, the increase in Igλ⁺ B cells may not reflect the full extent of VRC26UCA LC deletion. As judged by B cell number (SI Appendix, Fig. S2 I and J), expression of the VRC26UCA LC had less of an impact on B cell development than expression of the VRC26UCA HC. Nonetheless, combined expression of both VRC26UCA HC and LC exacerbated B cell deletion.

B Cell Developmental Block in the Constitutive Expression Model of VRC26UCA.

The analyses above showed that VRC26UCA expression essentially blocked B cell maturation. To define the stage of developmental arrest, we analyzed B cell progenitors in the bone marrow of the VRC26UCA KI mouse, in comparison with B cell progenitors in wild-type mouse and in the B1-8/3-83 KI mouse. Based on B220, CD19, and IgM surface markers, we broadly separated bone marrow B cells from the wild-type F1 mouse into three major populations: prepro-B cells (B220⁺CD19⁻IgM⁻; population I), pro-B cells and pre-B cells (B220⁺CD19⁺IgM⁻; population II), and immature and mature B cells (B220⁺CD19⁺IgM⁺; population III) (Fig. 3A and SI Appendix, Fig. S4) (48). As reported previously (43), the bone marrow from the B1-8/3-83 KI mouse had minimal numbers of pro-B cells and pre-B cells (population II) (Fig. 3B). This phenotype was likely attributable to the preassembled V exons in the KI mouse, which obviated the need for V(D)J recombination to assemble HC and LC V exons in pro-B cells and pre-B cells (43). The VRC26UCA KI mouse had a different bone marrow B cell pattern from either the F1 mouse or the B1-8/3-83 KI mouse (compare Fig. 3C with Fig. 3 A and B). Relative to the B1-8/3-83 KI mouse, the VRC26UCA KI mouse had more CD19⁺IgM⁻ pro-B and pre-B cells (compare population II in Fig. 3 B and C). In addition, the bone marrow of the VRC26UCA KI mouse contained a prominent B cell population with intermediate levels of CD19 (CD19^int) and low levels of surface IgM (IgM^lo) (population IV in Fig. 3C). These B cells were IgM^a+ (Fig. 3I), the allotype of VRC26UCA HC, and CD93^hi, a marker for immature B cells (Fig. 3K). A minor population of CD19^hiIgM⁺ immature and mature B cells was discernable (population III in Fig. 3C), but these B cells expressed IgM^b mouse antibodies (Fig. 3 J and K). The accumulation, and apparent developmental arrest, of the IgM^a+CD93^hi B cells suggests that the maturation of VRC26UCA B cells was impeded at the immature B cell stage. Like splenic B cells, the bone marrow B cells from the VRC26UCA HC-only mouse exhibited a similar but less severe, phenotype as those expressing both the HC and LC KI alleles (compare Fig. 3C with Fig. 3D and Fig. 3 I–K with Fig. 3 L–N). The bone marrow B cell profile of the KI mouse with VRC26UCA LC alone was largely normal (Fig. 3E).

Fig. 3.

FACS analysis of bone marrow B cells from the constitutive expression model of VRC26UCA. All of the plots are gated on B cells. The gating strategy is illustrated in SI Appendix, Fig. S4. The plots are labeled in the same manner as those in Fig. 2; the bone marrow in this figure and the spleen in Fig. 2 for each genotype were isolated from the same mouse. (A–E) FACS staining patterns of CD19 versus IgM. (F, I, and L) FACS staining patterns of CD19 versus IgM^a. (G, J, and M) FACS staining patterns of CD19 versus IgM^b. (H, K, and N) Overlays of histograms of CD93 expression on IgM^a and IgM^b B cells. These two populations differ in size in VRC26UCA KI mice; to compensate for this disparity, the y axis is normalized to mode, and the highest peak in each distribution is set to 100%. Due to smoothening of the curve, the highest peaks are not visible in some plots.

Some bnAbs are poly-reactive, and expression of such antibodies can trigger tolerance control mechanisms (11–16). This mechanism could underlie the developmental arrest of immature B cells in VRC26UCA KI mouse. To test this possibility, we assessed the cross-reactivity of VRC26UCA to mouse bone marrow and splenocytes. We preferred this assay to the conventional method of antibody staining of HEp2 cells (13), which originated from human epithelial cells and do not represent relevant antigens for B cells in KI mice. For comparison, we used the driver mouse antibody, which was used to support B cell maturation in the conditional expression system. Relative to this control, VRC26UCA exhibited stronger cross-reactivity with B cells from both spleen and bone marrow (Fig. 4 A and C), intermediate binding to non-B cells in bone marrow and non-B/T cells in spleen (Fig. 4 B and E), and no detectable interaction with splenic T cells and red blood cells (Fig. 4 D and F) (see SI Appendix, Fig. S5 for definition of these cell populations). This assay did not sample all antigens, such as soluble factors, in bone marrow and in spleen, and could not define the cross-reactive antigens. Despite these limitations, this experiment did reveal the cross-reactive nature of VRC26UCA, providing a potential basis for the developmental arrest of immature B cells in VRC26UCA KI mice.

Fig. 4.

In vitro studies on potential causes of the B cell developmental block in the constitutive expression model of VRC26UCA. (A–F) FACS analysis of cross-reactivity of VRC26UCA antibody. In this experiment, splenocytes or bone marrow cells from the constitutive expression model of VRC26UCA were stained with VRC26UCA antibody or the driver antibody. As indicated above each plot, binding activity was assessed on B cells and non-B cells from bone marrow (BM) or B cells, T cells, non-B/T cells and red blood cells (RBCs) from spleen. The gating strategy is illustrated in SI Appendix, Fig. S5. The binding activity is displayed in histograms, and the plots for the VRC26UCA and the driver antibodies are overlaid for comparison. The y axis of the histogram is normalized to mode, as in Fig. 3 H, K, and N. (G–K) ELISA analysis of pairing efficiencies of antibody HC with SLC. (G) ELISA measurement of the levels of μHC secreted in association with mouse SLC composed of mouse VpreB (mVpreB) and mouse λ5 (mλ5). Each line represents the titration curve of supernatant from one transfection experiment. Each transfection was done in triplicate. Titration curves for different HCs are labeled by color: driver HC, red; B1-8HC, brown; VRC26UCA HC, blue. Each plot includes a titration curve of IgM standard (black) as a reference; green curve represents background. The x axis of the plots represents relative concentration of the culture supernatant, and the highest concentration corresponds to undiluted supernatant. The y axis displays OD405, which correlates with antibody concentration, as measured for μ HC. Plots H–K are labeled in the same manner as plot G, except that B1-8HC was not included in the analyses in I–K. (H) Levels of μHC secreted in association with LC. (I) Levels of μHC secreted in association with human SLC composed of human VpreB (hVpreB) and human λ5 (hλ5). (J) Levels of μHC secreted in association with hVpreB and mλ5. (K) Levels of μHC secreted in association with mVpreB and hλ5.

In light of the largely empty pro-B and pre-B compartments in the B1-8/3-83 KI mouse (population II in Fig. 3B), the preassembled V exons of B1-8/3-83 antibody appeared to have accelerated B cell development through these stages (43). In comparison, the presence of a sizable population of pro-B cells and pre-B cells in the VRC26UCA KI mouse (population II in Fig. 3C) suggests that, relative to the B1-8/3-83 antibody, the VRC26UCA was less effective in this respect. This observation was reminiscent of the phenotype of a KI mouse that expresses an autoreactive antibody against class I MHC, where the accumulation of pre-B cells reflects receptor editing activity to delete the LC of the autoreactive antibody (43). Such a mechanism could explain the increase in pre-B cells in the VRC26UCA KI mouse, which also showed obvious signs of receptor editing, as revealed by elevated frequencies of Igλ⁺ B cells (Fig. 2G). In addition, V_H replacement on the VRC26UCA KI IgH^a allele and rearrangement of the other IgH^b allele, as evidenced by the emergence of IgM^b+ B cells from the pro-B/pre-B cell compartment (Fig. 3J), could contribute to the accumulation of pro-B cells in the bone marrow of the VRC26UCA KI mouse. Expression of a functional HC in pro-B cells normally suppresses V(D)J recombination activity on the IgH locus (4). One potential explanation for the failure of the VRC26UCA to do so could be poor pairing of this HC with the surrogate light chain (SLC) to form a pre-B cell receptor (49, 50), which promotes the transition of pro-B cells to pre-B cells and terminates IgH rearrangement (51–56).

Although the pre-B cell receptor is normally expressed on cell surface, a soluble form of the HC/SLC complex has been produced for structural and biochemical studies (57). Thus, to test the potential role of HC/SLC pairing in the failure of the VRC26UCA to promote early B cell development, we cotransfected the secretory form of VRC26UCA HC with SLC components, VpreB (58) and λ5 (59), into 293T cells. If the HC does not pair with the SLC, the HC should not fold properly for secretion. Thus, the level of secreted HC, presumably in association with the SLC, should correlate with their pairing efficiency. For comparison, we performed the same experiment with two mouse HCs: The driver HC and the B1-8 HC (Fig. 4G and SI Appendix, Fig. S6). In support of our prediction, cotransfection of the VRC26UCA HC with mouse SLC yielded no detectable HC secretion, in contrast to the results with the driver HC and the B1-8 HC, which were readily detectable in culture supernatants (Fig. 4G). The pairing defect was specific for the mouse SLC, as cotransfection of the VRC26UCA HC and LC led to robust antibody secretion (Fig. 4H). Furthermore, cotransfection of the VRC26UCA HC with the human SLC also improved antibody secretion (Fig. 4I), and the human λ5 accounted primarily for the stimulatory effect (Fig. 4 J and K).

Conditional Expression Strategy to Bypass the Developmental Block of VRC26UCA B Cell Progenitors.

The second goal of this study was to test whether a conditional expression strategy could overcome the impediment to B cell development associated with constitutive expression of VRC26UCA. Toward this end, we generated a conditional expression model of VRC26UCA (Fig. 1B). In contrast to the constitutive expression model, where the majority of splenic B cells expressed mouse IgM^b, most of the splenic B cells in the conditional expression model expressed IgM^a (Fig. 5B). This result indicates that the driver HC, which was expressed in B cell progenitors, effectively excluded the other IgH^b allele and supported B cell maturation. Nonetheless, relative to the control IgH^a 129/Sv mouse, both IgM^a and IgD expression on B cells from the conditional expression model were below normal (Fig. 5 A–F). With respect to LC expression, the frequency of Igλ⁺ B cells, including a distinct Igλ^lo population, was increased (Fig. 5 G–I). On the other hand, B cells from the conditional expression model of VRC26UCA displayed normal levels of major B cell surface markers, except for a minor increase in class II MHC (I-A^b) (SI Appendix, Fig. S7). Low surface Ig is characteristic of anergic B cells, likely because persistent stimulation by self-antigens down-modulates the B cell receptor (24, 25). The poly-reactivity of VRC26UCA (Fig. 4 A and C) may be a contributing factor to the low surface Ig phenotype. Expression of VRC26UCA followed the activation of CD21-cre in mature B cells (Fig. 1C). CD21 expression is up-regulated during the transition of immature B cells to mature B cells (60). Premature expression of VRC26UCA in immature B cells, as a consequence of elevating CD21-cre expression in transitional B cells (39), may also contribute to the anergic phenotype of B cells in the VRC26UCA mouse model, since immature B cells are susceptible to tolerance control.

Fig. 5.

FACS analysis of splenic B cells from the conditional expression model of VRC26UCA. All of the plots are gated on B cells, as illustrated in SI Appendix, Fig. S2. The plots are labeled in the same manner as those in Fig. 2. The genotypes of the mice are indicated above the plots: 129/Sv, wild-type 129/Sv mouse; VRC26H/L conditional, the conditional expression model of VRC26UCA HC and LC. (A and B) FACS plots showing the distribution of IgM^a versus IgM^b expressing B cells. (C) The histogram is an overlay of IgM^a expression between 129/Sv (A) and the conditional expression model of VRC26UCA (B). The y axis of the histogram represents relative cell numbers, and the peak of each histogram is set to 100% (modal scale). (D–I) These plots compare the expression patterns of IgM/IgD and Igκ/Igλ between control 129/Sv mouse and the conditional expression model of VRC26UCA. The plots are labeled in the same manner as in A–C.

Due to partial deletion of the driver V exon by CD21-cre, the IgM^a B cell population consisted of a mixture of VRC26UCA and driver-expressing B cells. To determine the ratio of these two types of B cells, we used the same DC-PCR method as for V_H replacement analysis (SI Appendix, Fig. S8). Because of its long CDR H3, the DC-PCR product for VRC26UCA HC is longer than that of the driver HC (Fig. 6 and SI Appendix, Fig. S8B). Control experiments confirmed unbiased amplification of the VRC26UCA HC and driver HC with this DC-PCR method (SI Appendix, Fig. S8B). Thus, the ratio of the two DC-PCR products should reflect the fraction of B cells that express the two antibodies. Strictly speaking, the method detects V exons that are positioned for expression at the DNA level, but not the actual expression of the V region at the RNA or protein level. Since B cells are dependent on functional B cell receptors for survival (61), it is reasonable to assume that the V exon, which is proximal to the constant regions, is expressed. Since CD21-cre acts primarily in mature B cells, VRC26UCA expression should happen predominantly in mature B cells. We sorted splenic B cells into IgM^loIgD^hi and IgM⁺IgD^lo populations (Fig. 6A), which were enriched for mature and immature B cells, respectively. Based on the DC-PCR assay, about 30% of mature splenic B cells expressed the VRC26UCA, and consistent with the expression pattern of CD21-cre, few immature B cells expressed VRC26UCA (Fig. 6B). VRC26UCA expression was further enriched among B cells with lower IgM expression (Fig. 6 C and D), presumably because of poly-reactivity of this antibody.

Fig. 6.

Measurement of the fraction of B cells expressing the VRC26UCA in the conditional expression model. (A–D) DC-PCR analysis of the ratio of splenic B cells expressing the driver or VRC26UCA. (A) The FACS plot shows the B cell populations that were sorted for DC-PCR analysis: P6, IgM^loIgD^hi; P7, IgM⁺IgD^lo. The plot is gated on B220⁺ B cells. (B) DC-PCR analysis was performed on sorted B cells from A. The PCR products were run on agarose gel, detected with Southern hybridization and quantified with PhorsphorImager. The identity of the PCR products is labeled to the right of the autoradiogram; an unknown PCR product is indicated with an asterisk (*). At the top of the autoradiogram, the B cell population analyzed in this assay is indicated; VRC26% = VRC26/(VRC26+driver). C and D show a similar experiment as A and B, except that the P6 gate was shifted toward lower IgM levels. (E–G) Single-cell RT-PCR assay to determine the frequency of activated IgG1⁺ splenic B cells expressing VRC26UCA in the conditional expression model. (E) Splenic B cells from the conditional expression model were activated in vitro with anti-CD40 antibody plus IL4, and B220⁺IgG1⁺ B cells (P5) were sorted as single-cells into 96-well plate. (F) VRC26UCA HC or (G) LC transcripts were amplified from the sorted B cells. Each lane corresponds to PCR of one cell. Only part of the gel image is shown. The percentage of VRC26UCA HC- or LC-positive B cells out of the total sorted B cells, 95 in this experiment, is indicated at the top of the gel image.

The low surface Ig phenotype of VRC26UCA B cells raised concerns about their function. As a preliminary assessment of their functional status, we stimulated splenic B cells from the VRC26UCA mice with anti-CD40 antibody plus IL-4, which activate B cells to undergo IgH class switching to IgG1 and IgE. We sorted IgG1⁺ B cells and performed single-cell RT-PCR (Fig. 6E). About 38% of the sorted IgG1⁺ B cells expressed the VRC26UCA HC (Fig. 6F), and 93% of these B cells expressed the VRC26UCA LC (Fig. 6G). This result showed that VRC26UCA-expressing B cells could at least be activated in vitro. Furthermore, the analysis confirmed the expression of VRC26UCA at the RNA level in a substantial fraction of splenic B cells, in corroboration with the DC-PCR analysis. As a validation of the conditional expression strategy, we also confirmed that VRC26UCA expression was dependent on driver deletion (SI Appendix, Fig. S9).

Discussion

We show that constitutive expression of VRC26UCA in KI mice caused a B cell developmental arrest, primarily at the immature B cell stage. These immature B cells displayed atypical patterns of surface markers: Intermediate CD19 and low IgM (Fig. 3C). Bone marrow from wild-type mouse did not contain substantial numbers of B cells with this surface phenotype (Fig. 3A). Such B cells have also not been reported in other bnAb KI mice, but bone marrow B cells were not examined with the combination of CD19 and IgM markers in those studies (5–10). Since a similar CD19^int B cell population was also detectable in the B1-8/3-83 KI mouse (Fig. 3B), this type of B cell may be present, in varying numbers, in other KI mice too. The origin of this B cell population is unclear. One possibility is that these B cells developed conventionally through pro-B cells and pre-B cells, but down-modulated both CD19 and IgM at the immature B cell stage, potentially due to poly-reactivity of the nascent B cell receptor. However, the presence of this type of B cell in both the VRC26UCA and the B1-8/3-83 KI mice led us to favor the possibility that the phenomenon may relate more generally to the preassembled V exons. Since expression of preassembled V exons does not depend on V(D)J recombination in pro-B cells and pre-B cells, HC and LC expression could potentially begin earlier than usual, during the transition from CD19⁻ prepro-B cells to CD19⁺ pro-B cells (62), hence their CD19^int phenotype. In this scenario, these B cells essentially skipped the normal pro-B cell and pre-B cell stages. For some antibodies, such as B1-8/3-83, B cells could apparently reach maturity through this pathway (43). However, in the case of VRC26UCA and perhaps other antibodies, the nonphysiological maturation pathway might exacerbate the negative impacts of other factors, such as poly-reactivity. Receptor editing can rescue autoreactive B cells from developmental arrest or clonal deletion (19–22, 43). However, the mechanism may be ineffective for antibodies such as VRC26UCA, whose HC, with its long CDR H3, is the dominant cause of negative selection. Deletion of the autoreactive HC via V_H replacement cannot take place in immature B cells, where the IgH locus is no longer active for V(D)J recombination. As a result, the aberrant immature B cells could not be rescued from developmental arrest and accumulated in the bone marrow of VRC26UCA KI mouse.

One way to address this B cell developmental problem is to express bnAbs through de novo assembly of their V exons via V(D)J recombination, thereby enforcing the physiological maturation pathway for B cell progenitors. We previously took this approach to express the precursors for the VRC01 class antibodies in mice (63). The VRC01 class antibodies target the CD4-binding site of the HIV-1 envelop (64–66). Unlike VRC26, the VRC01 antibodies do not have long CDR H3s. The VRC01 antibodies contact the CD4-binding site primarily through CDR H2 and framework 3 region of the human V_H1–2 segment, whereas CDR H3 plays a relatively minor role in epitope contact (64, 65, 67). For this reason, the CDR H3 sequences are variable among VRC01 family members (68). To recapitulate this CDR H3 diversity in the VRC01 mouse model, we allowed the human V_H1–2 segment to recombine with the whole set of mouse D and J_H segments (63). B cells in this VRC01 model developed normally and expressed diverse HCs with normal CDR H3 length distribution; immunization of this mouse model elicited VRC01-like antibodies (63, 69). The physiological process of V(D)J exon assembly and the usual CDR H3 length may have contributed to the normal B cell phenotype in the VRC01 mouse model.

Not all B cells expressed surface IgM prematurely, prior to the pro-B cell stage in the VRC26UCA KI mouse; the CD19⁺IgM⁻ population (population II in Fig. 3C) in the VRC26UCA KI mouse may consist of such B cells. This population was more numerous in the VRC26UCA KI mouse than in the B1-8/3-83 KI mouse. We interpreted this difference as a reflection of the inefficiency of VRC26UCA, even when expressed as a preassembled V exon, in driving B cell maturation through the pro-B cell and pre-B cell stages. The abundant V_H replacement events on the HC KI IgH^a allele and rearrangement of the other mouse IgH^b allele in the VRC26UCA KI mouse also suggest that the VRC26UCA HC failed to effectively suppress V(D)J recombination and to expedite B cell maturation through the pro-B cell stage. These observations led us to propose that poor pairing of the VRC26UCA HC with the mouse SLC may underlie these phenomena, and our in vitro experiments supported this hypothesis. The basis for the pairing defect is unclear, but the extraordinarily long CDR H3 of the VRC26UCA HC could be relevant. It has been shown that some antibodies with long CDR H3s are under negative selection during B cell maturation (36), and SLC could play a role in this selection. The non-Ig domains of VpreB and λ5 form the equivalent of CDR L3 and interact with CDR H3 of the HC (57). It is possible that long CDR H3, at least in some cases, structurally clashes with the non-Ig domain of the SLC. The 37-amino acid CDR H3 of VRC26UCA is far above the average CDR H3 length of mouse antibodies, and the mouse SLC may not be adapted to accommodating such extraordinarily long CDR H3. In comparison, antibodies with long CDR H3s are more common in human than in mouse repertoires (34, 35, 70), and the human SLC may have evolved to accommodate long CDR H3s, thus explaining the more efficient pairing of VRC26UCA HC with the human SLC than with the mouse SLC (Fig. 4 G and I). To test the physiological relevance of these in vitro observations, an obvious experiment is to swap the mouse SLC with the human counterpart and determine whether the human SLC facilitates B cell maturation in this and potentially other bnAb mouse models. If validated, incorporation of the human SLC into KI mice could be a general strategy to alleviate B cell developmental defects in bnAb mouse models (5–10).

One implication of the present study is that multiple factors may contribute to B cell defects in bnAb KI mice. Some of these factors, such as a nonphysiological B cell developmental pathway and mouse SLC pairing defect, are specific to the KI mouse system and may not be relevant to human vaccination. Eliminating these nonphysiological factors would not only increase target B cells for immunogens, but also reveal physiologically relevant hurdles for bnAb induction, and the information could facilitate the development of intervention strategies. The conditional expression approach is one step in this direction. The conditional expression method bypasses obstacles at the B cell progenitor stage, including some potentially nonphysiological hindrances, and generates mature B cells expressing the target antibody for testing immunogens. In the case of the conditional expression model of VRC26UCA, the mature B cells still expressed abnormally low levels of surface Ig and may be functionally anergic. Activating such B cells may be challenging. On the other hand, if the anergic state reflects bona fide peripheral tolerance, the model could be used to develop strategies to revitalize the anergic B cells, for example with multimeric immunogens or potent adjuvants. Given the prevalence of poly-reactivity among bnAbs, such intervention may be necessary for bnAb induction (30, 71). The conditional expression method is obviously not a feasible way to overcome deletion of bnAb precursors in humans. The value of the strategy lies in generating mouse models with sufficient target B cells for testing and optimizing immunization strategies at the preclinical stage. In human vaccination, an effective vaccine most likely will need to activate rare target B cells that have survived tolerance control checkpoints or other restrictions. Vaccine optimization in the model system that we describe here, or in related mouse model systems, may increase the chance of success in overcoming such potential hurdles.

Materials and Methods

Generation of the Constitutive and Conditional Expression Model of VRC26UCA.

The conditional expression system requires a CD21-cre transgene. To expedite the set-up of the conditional expression system, we derived an ES cell line from CD21-cre transgenic mice (39). The genetic background of the original CD21-cre transgenic mouse line was C57BL/6. As described in the text, we planned to utilize IgH allotypic markers to differentiate B cells expressing the KI VRC26UCA antibody versus mouse antibodies. Toward this end, we crossed the CD21-cre mice with 129/Sv mice and derived an ES cell line from 3.5-d embryos. In this CD21-cre ES cell line, one IgH allele was an allotype from the 129/Sv mouse strain, whereas the other IgH allele was a “b” allotype from the C57BL/6 mouse strain. We used this CD21-cre ES cell line for incorporating the VRC26UCA HC and LC expression constructs.

The organization of the integration construct for VRC26UCA HC is illustrated in SI Appendix, Fig. S1A; the diagram also applies to the LC KI construct. The sequences of the V(D)J exons for VRC26UCA HC, VRC26UCA LC, and driver HC in these KI constructs are shown in SI Appendix, Supplementary Materials and Methods. The homology arms of the integration constructs were amplified with high-fidelity PCR from the genomic DNA of a 129/Sv ES cell line. Linearized construct was electroporated into the CD21-cre ES cell line. Clones with stable integration of the construct were selected with G418. These clones were screened via Southern blotting. Correct clones were transduced with Adeno-cre virus. As illustrated in SI Appendix, Fig. S1A, partial recombination among the three loxP sites in the construct gave rise to ES clones for either the constitutive expression or the conditional expression of VRC26UCA. LoxP recombination pattern was determined by Southern blotting. From this screening, we selected ES clones for constitutive or conditional expression of VRC26UCA. All clones for generating mouse models were verified for genotype (SI Appendix, Fig. S1 B and C) and karyotype (96% ES cells with 40 chromosomes for the conditional expression model; 90% ES cells with 40 chromosomes for the constitutive expression model). To generate the KI mice, the ES cells were injected into Rag2-deficient blastocysts to generate chimeric mice. For long-term studies, we also bred the chimeric mice for germline transmission. In the present study, all analyses of the constitutive expression model were performed with germline KI mice that resulted from cross between the VRC26UCA KI mice and C57BL/6 mice. These mice were of mixed 129/Sv and C57BL/6 genetic background. The conditional expression model also gave germline transmission. As explained in Results, the driver V gene flanked with loxP sites tended to be deleted in a substantial fraction of progeny during breeding. For this reason, we used RDBC chimeric mice for the analysis of the conditional expression model. All mouse work in this study has been approved by the Institutional Animal Care and Use Committee at Boston Children’s Hospital.

B1-8/3-83 KI Mice.

The B1-8/3-83 KI mice were published previously (41–43). These mice were originally generated with the E14 ES cell line from the 129/Ola mouse strain (IgH^a allotype). The B1-8 HC was integrated into the J_H locus of one IgH^a allele. To distinguish the B1-8 HC from endogenous mouse HC with allotypic marker, as in the VRC26UCA KI mice, we bred the B1-83/3-83 KI mice with C57BL/6 mice so that the B1-8 HC was expressed as IgH^a, whereas mouse HC was expressed as IgH^b.

Characterization of Splenic and Bone Marrow B Cells in Mouse Models.

For splenic B cell analysis, spleen was dissociated mechanically into cell suspension. Red blood cells were eliminated with Red Blood Cell lysing buffer (Sigma R7757). To generate the data in Figs. 2 and 5 and SI Appendix, Fig. S2, the splenocytes were stained with the following antibodies: PE-Cy5 anti- B220 (eBioscience 15-0452-83), PE anti-Thy1.2 (PharMingen 553006), PE anti-IgM (eBioscience 12-5790-83), PE anti-IgM^a (PharMingen 553517), PE anti-Igλ (Biolegend 407308), FITC anti-IgD (BD PharMingen 553439), FITC anti-IgM^b (PharMingen 553520), FITC anti-Igκ (SouthernBiotech 1050-02). The stained cells were analyzed on BD FACS Calibur, and FACS plots were generated with FlowJo10 software. For analysis of B cell marker expression in SI Appendix, Fig. S7, splenic B cells were stained with the following antibodies: PerCP/Cy5.5 anti-B220 (eBioscience 45-0452-82), APC anti-IgM (eBioscience 17-5790-82), PE anti-IgM (eBioscience 12-5790-83), APC anti-IgD (Biolegend 405713), FITC anti-IgD (BD PharMingen 553439), PE anti-CD93 (eBiosceince 12-5892-82), PE anti-CD22 (SouthernBiotech 1580-09), PE anti-I-A^b (SouthernBiotech 1896-09), PE anti-CD86 (SouthernBiotech 1735-09), PE anti-CD40 (eBioscience 12-0401-81), PE anti-CD19 (BD Pharmingen 557399), PE anti-CD23 (BD Pharmingen 553139), FITC anti-CD21 (BD Pharmingen 553818). The stained cells were analyzed on Attune NxT from Invitrogen, and the data were analyzed with FlowJo10 software.

For bone marrow B cell analysis, bone marrow cells were flushed out from femur and tibia with a syringe. Red blood cells were eliminated with Red Blood Cell lysing buffer (Sigma R7757). In the experiments shown in Fig. 3 A–E and SI Appendix, Fig. S4, bone marrow cells were stained with the following antibodies: FITC anti-B220 (eBioscience 11-0452-85), APC anti-CD19 (SouthernBiotech 2018-01), PE anti-IgM (eBioscience 12-5790-83), PE-Cy7 anti-CD43 (BD Pharmingen 562866), Sytox blue (S34857). Since CD43 expression was not discussed in this study, CD43 staining was not shown in the figures. For the experiments shown in Fig. 3 F–N, bone marrow cells were stained with FITC-anti-B220 (eBioscience 11-0452-85), APC-anti-CD19 (SouthernBiotech 2018-01), PerCP/Cy5.5 anti-CD93 (eBioscience 45-5892-82), PE-anti-IgM^a (PharMingen 553517); PE anti-B220 (eBioscience 12-0452-82), APC anti-CD19 (SouthernBiotech 2018-01), PerCP/Cy5.5 anti-CD93 (eBioscience 45-5892-82), FITC anti-IgM^b (PharMingen 553520). Flow cytometry was performed on Attune NxT from Invitrogen, and the data were analyzed with FlowJo10 software.

Analysis of V_H Replacement.

Splenocytes were isolated from the constitutive expression model of VRC26UCA and stimulated with anti-CD40 antibody (Invitrogen 16-0402-86) plus IL-4 (72). After 3 d of stimulation, activated B cells were fused with NS1 plasmacytoma cells with the polyethylene glycol method (73). Hybridoma clones were selected with hypoxanthine–aminopterin–thymidine. Supernatants from the hybridoma clones were screened with enzyme-linked immunosorbent assay (ELISA) to determine the isotype and allotype of secreted antibodies. Simulation with anti-CD40 antibody plus IL-4 induced class switching to IgG1 and IgE (72). Unswitched B cells continued to express IgM. As a result, most of the hybridoma clones secreted IgM, IgG1 or IgE. Some supernatants contained more than one antibody isotype; these samples likely were derived from mixed hybridoma clones and were not analyzed further. Hybridoma clones producing only IgM or IgG1 were used for the experiments in SI Appendix, Fig. S3. The supernatants from these hybridomas were further screened with allotype-specific antibodies. For IgM⁺ hybridomas, their supernatants were screened with anti-IgM^a and anti-IgM^b antibodies. Hybridoma clones expressing only IgM^a or IgM^b were chosen for further analysis. For IgG1⁺ hybridomas, their supernatants were screened with anti-IgG1^a antibody only; the anti-IgG1^b antibody worked poorly in ELISAs. IgG1^a-/IgG1⁺ hybridomas were considered as IgG1^b+. Without direct identification of IgG1^b hybridomas, some IgG1^a+ hybridomas could be mixed with IgG1^b+ hybridomas, causing over-estimation of the frequency of IgG1^a hybridomas.

In SI Appendix, Fig. S3C, genomic DNAs from IgM⁺ hybridomas were amplified with primers corresponding to sequences downstream of the cryptic RSS. In SI Appendix, Fig. S3D, the same DNA samples from SI Appendix, Fig. S3C were amplified with primers for sequences upstream and downstream of the cryptic RSS, respectively. To amplify V_H replacement products with DC-PCR, genomic DNA from hybridomas were digested with NcoI and XbaI, or PstI and XbaI. The digested DNA was self-ligated into circles with adaptor primers that had cohesive ends for NcoI and XbaI, or PstI and XbaI. The ligated product was subject to PCR amplification with primers a and b (SI Appendix, Fig. S3E). The PCR product was sequenced. The identities of the mouse D or V_H segments joined to VRC26UCA HC, as shown in SI Appendix, Fig. S3 F–H and Table S1, were determined with Blast or Ig Blast.

Assessment of Cross-Reactivity of VRC26UCA Antibody.

To produce recombinant VRC26UCA antibody and the driver antibody, the cDNAs for the secretory form of HCs and LCs of these antibodies were cloned into the pcDNA3 expression vector. The constant regions for the HCs and LCs were mouse μ and κ isotypes, respectively. A 6xHis tag was appended to the C terminus of the HC to facilitate purification. The cDNA sequences for the VRC26UCA and the driver antibodies in these expression constructs are shown in SI Appendix, Supplementary Materials and Methods. The expression constructs were transfected into Expi293 cells, using ExpiFectamine (Gibco A14524). The supernatant of the culture was harvested 5 to 6 d after transfection. Antibodies were purified from the supernatant with HisTrap column on fast protein liquid chromatography (FPLC). The purified antibody was conjugated to AlexaFuro647 with microscale protein labeling kit (Invitrogen A30009). For the cross-reactivity assay, bone marrow cells or splenocytes from the constitutive expression model of VRC26UCA were stained with the following antibodies: FITC anti-B220, PE anti-Thy1.2, Biotin Ter119 (Biolegend 116203), BV605 Streptavidin (Biolegend 405229), PE-Cy7 anti-TNP (Biolegend 401627), AF647 VRC26UCA or AF647 Driver. The stained splenocytes or bone marrow cells were analyzed on Attune NxT flow cytometer from Invitrogen, and FACS plots were generated with FlowJo10 software.

Analysis of HC Pairing with SLC.

The cDNAs for the secretory form of the HC and for the LC, VpreB, or λ5 chains were cloned into the pcDNA3 expression vector; the cDNA sequences in these expression vectors are available in SI Appendix, Supplementary Materials and Methods. The expression constructs were transfected into 293T cells, using the PEI method (74). To control for transfection efficiency, a pMaxGFP expression construct (Lonza) was cotransfected with the antibody expression constructs. The ratio of HC, LC, and GFP expression constructs was: 50:50:5; the ratio of HC, VpreB, λ5, and GFP expression constructs was: 33:33:33:5. Two days after transfection, the supernatant of the culture was collected. Antibody concentration in the supernatant was measured with ELISA, using anti-mouse IgM antibody (SouthernBiotech 1021-01) for capture and anti-mouse IgM-alkaline phosphatase antibody (SouthernBiotech 1021-04) for detection. Purified mouse IgM (SouthernBiotech 0101-01) served as standard.

DC-PCR Assay to Determine the Frequency of Splenic B Cells Expressing VRC26UCA in the Conditional Expression Model of VRC26UCA.

Splenic B cells were sorted into B220⁺IgM⁺IgD^lo and B220⁺IgM^loIgD^hi fractions on BD FACS Aria. Genomic DNA was isolated from the sorted B cells. The genomic DNA was digested with NcoI and EcoRI. The digested DNA was ligated with an adaptor with compatible cohesive ends with NcoI and EcoRI. In the ligation reaction, low DNA concentration favored intramolecular recircularization. Circularized DNA containing the V regions of the VRC26UCA HC or the driver HC shared a common region downstream of J_H. PCR amplification with primers a and b, which annealed to the common sequence, yielded products containing the two V regions with comparable efficiencies (SI Appendix, Fig. S8). The PCR product was run on agarose gel. The DNA was transferred to Zeta-Probe GT membrane (Bio-Rad 162-0197). The DNA was hybridized to an oligonucleotide probe that was common to the PCR products of both V regions. The hybridization signal was quantified with PhorsphorImager (GE Storm 865).

Single-Cell RT-PCR Assay to Determine the Fraction of B Cells Expressing the VRC26UCA in Activated Splenic B Cells.

Splenic B cells were isolated from the conditional expression model of VRC26UCA. The B cells were stimulated with anti-CD40 antibody plus IL-4. After 3 d of stimulation, single B220⁺IgG1⁺ B cells were sorted into 96-well plate with BD FACS Aria. Reverse transcription of the RNA from the sorted single cells yielded cDNA, which served as template for PCR amplification for the VRC26UCA HC and LC. The VRC26UCA HC cDNA was amplified with forward primer for the HC V region and reverse primer for Cγ1. Similarly, the VRC26UCA LC cDNA was amplified with forward primer for the LC V region and reverse primer for Cκ.

PCR Amplification of the VRC26UCA HC or the Driver HC from Hybridomas of the Conditional Expression Model of VRC26UCA.

Hybridomas were generated and screened with the same method as that for the constitutive model, as described above. SI Appendix, Fig. S9 shows the analysis of 8 IgG1⁺ hybridomas. As illustrated in the diagram above each gel image, RT-PCR amplification of the driver or the VRC26UCA HC transcripts were achieved with forward primers specific for each V region and reverse primer in Cγ1. PCR amplification of the driver HC or the VRC26UCA HC directly from hybridoma DNA was achieved with forward primers specific for each V region and a reverse primer downstream of J_H.

Data Availability.

All of the data supporting this study are shown in the main text figures and SI Appendix. The materials involved in this study are available upon request to the corresponding authors (M.T. and F.W.A.).