Germinal center entry not selection of B cells is controlled by peptide-MHCII complex density

B cells expressing high affinity antigen receptors are advantaged in germinal centers (GC), perhaps by increased acquisition of antigen for presentation to follicular helper T cells and improved T-cell help. In this model for affinity-dependent selection, the density of peptide/MHCII (pMHCII) complexes on GC B cells is the primary determinant of selection. Here we show in chimeric mice populated by B cells differing only in their capacity to express MHCII (MHCII+/+ and MHCII+/−) that GC selection is insensitive to halving pMHCII density. Alone, both B cell types generate identical humoral responses; in competition, MHCII+/+ B cells are preferentially recruited to early GCs but this advantage does not persist once GCs are established. During GC responses, competing MHCII+/+ and MHCII+/− GC B cells comparably accumulate mutations and have indistinguishable rates of affinity maturation. We conclude that B-cell selection by pMHCII density is stringent in the establishment of GCs, but relaxed during GC responses.

T he primary repertoire of B-cell antigen receptors (BCR) is generated by the combinatorial association of V, D, and J gene segments during B-cell development. This primary BCR repertoire is expanded and refined by somatic hypermutation and affinity-driven selection in germinal centers (GC), resulting in a secondary BCR repertoire capable of high affinity binding to virtually any antigen. Selection for entry into nascent GCs seems to be controlled by interclonal competition for T-cell help based on the different levels of peptide/MHC class II (pMHCII) displayed by antigen-activated B cells 1 . Concordantly, even B cells expressing BCRs with very low affinity for antigen can form GCs in the absence of competition from higher-affinity clones 2,3 . In organized GCs, B cells participate in iterative rounds of interzonal migration, switching between the centroblast state in the GC dark zone (DZ) and the centrocyte state in the light zone (LZ) 4 . Rapid proliferation and fixation of V(D)J mutations characterize the GC DZ, whereas antigen presentation and affinity-dependent selection occur among the T FH and follicular dendritic cells (FDC) in the LZ 5,6 . Selection in the LZ is thought to represent intraclonal and interclonal competition; the successful B-cell competitors return to the DZ for additional rounds of proliferation and mutation and by this cyclic process maximize the somatic evolution of BCR affinity [7][8][9][10] . How FDC and T FH cells function to select higher affinity BCRs from newly mutated B-cell populations, however, is unclear.
Affinity-driven selection in GCs has been proposed to be controlled by the density of pMHCII displayed by B cells during cognate interaction with helper T cells 4 . This "T-cell help" model is supported by mathematical modeling 11,12 , the finding that BCRs retrieve antigen for processing in an affinity-dependent manner 13 , and the critical function of T FH cells in GC responses 14 . Direct evidence for the role of pMHCII density in controlling GC B-cell competition comes from experiments that deliver antigen to GC B cells by a BCRindependent mechanism that bypasses FDCs 5,9,15,16 . In this experimental model, targeted LZ B cells with increased pMHCII densities have prolonged interaction with T FH cells and preferentially re-enter the DZ for further rounds of proliferation and mutation 5 . These studies also indicate that prolonged, cognate T:B-cell interaction increases the proliferative capacity of GC B cells in the DZ and speeds transit through the cell cycle 9,15,16 .
To quantify the role of pMHCII in controlling B-cell selection into and during the GC reaction, we use an alternative strategy to map the limits of T-cell help in the selection of antigen-specific B cells for humoral responses. By short-and long-term B-cell reconstitutions, we place congenic MHCII +/+ and haploinsufficient MHCII +/− B cells in direct competition for GC entry and affinity-dependent selection. Even though MHCII expression by B cells is modulated during the course of humoral responses, these competing B-cell populations consistently express twofold differences in MHCII and pMHCII surface density. Our competition experiments confirm that MHCII +/+ B cells are preferentially seeded to nascent GCs even though wild type (WT) and haploinsufficient B cells are comparably activated by antigen in vivo. Once GCs are formed, however, MHCII +/+ GC B cells have no competitive advantage over haploinsufficient B cells with regard to their persistence, proliferation, acquisition of V(D)J mutations, and affinity maturation. We conclude that pMHCIIdriven selection is more stringent for B cells entering GCs than for B cells in established GCs. In this relaxed environment of pMHCII selection, GC B cells with a broad range of BCR affinities can co-exist, increasing the potential for rare evolutionary trajectories to contribute to protective, humoral immunity.
To ensure that MHCII haploinsufficiency equated to a comparable reduction in pMHCIIs, we made a chimeric antigen of NP-streptavidin bound to biotinylated I-Eα52-73 peptide (NP-SA-Eα) 1 . Quantification of Eα52-68 pMHCII complexes can be determined by the Y-Ae monoclonal Ab, which is specific for pEα:I-A b complexes 20 . B1-8.MHCII +/+ and B1-8.MHCII +/− mice were immunized in the footpad with NP-SA-Eα in alum; MHCII expression on MF B cells in unimmunized controls was halved in haploinsufficient animals, for both NP-binding and non-binding cells ( Supplementary Fig. 1A). In immunized mice, by 16 h post immunization populations of NP-binding B cells with lower membrane IgD levels appeared in both WT and haploinsufficient mice (Fig. 1b). Within these activated IgD low NP + B-cell subsets, both MHCII and Eα52-68 pMHCII were halved in MHCII +/− B cells compared to WT controls. This quantitative difference is maintained when B cells are exposed to TLR ligands in vitro (Supplementary Figs. 1B-D).
We observed no significant effects of MHCII haploinsufficiency on T-cell-dependent humoral responses. MHCII +/+ and MHCII +/− B6 mice immunized with NP-OVA exhibited comparable IgG Ab levels and GC responses on days 8, 12, 16, and 24 post immunization. Indeed, the kinetics and magnitude of GC responses in MHCII +/+ and MHCII +/− mice were indistinguishable ( Fig. 1c) and serum IgG for NP and NIP (4-hydroxy-3-iodo-5-nitrophenyl acetyl) were similar as well (Fig. 1d). As expected 21 , heteroclitic (NIP-binding) IgG levels rose faster than NP-specific IgG, but NP-and NIP-specific serum IgG levels converged by day 24 (Fig. 1d). These data demonstrate that MHCII +/+ and MHCII +/− B cells have similar intrinsic capacities to produce GCs and serum IgG Ab in response to NP-OVA. In the absence of MHCII +/+ competitors, reduced MHCII and pMHCII expression on haploinsufficient B cells does not impact GC responses or affinity maturation of serum IgG Ab.
To determine whether MHCII haploinsufficiency might affect the average or distribution of BCR avidities within GCs, we sorted single MF and GC B cells from the spleens of MHCII +/+ and MHCII +/− mice immunized with NP-OVA for single-cell Nojima cultures (Supplementary Fig. 2A To compare BCR affinity distributions among GC B cells from MHCII +/+ and MHCII +/− mice, we determined the avidity indices (AvIns) for every clonal IgG Nojima culture to NP and NIP 22 . The AvIn represents the ratio of specific (NP-or NIP) binding by individual clonal IgGs to a standard, heteroclitic NP/ NIP IgG mAb, H33Lγ1 (K a = 2.0 × 10 7 M −1 ) 21 . We determined AvIn values for both high density (permissive) and low density (stringent), NP-and NIP-binding. The expected, canonical GC response is both heteroclitic (NIP > NP binding) and stringent. From MHCII +/+ and MHCII +/− GC, respectively, 44.7% (155/ 349) and 60.3% (186/308) of clonal IgGs bound to the permissive (low and high avidity), high-density NIP 25 conjugated Luminex beads; both cohorts exhibited similar AvIn distributions and geometric means that were not significantly different (Fig. 1e). Stringent, heteroclitic binding to NIP 2 conjugated beads was also identical between the MHCII +/+ and MHCII +/− cohorts with similar distributions and geometric means of AvIn values that were comparably higher than those determined for NP 2 beads (Fig. 1e). In no case did the mean AvIn values for MHCII +/+ GC and MHCII +/− GC B cells differ significantly for the same antigen ligand and both WT and haploinsufficient GC B-cell clones exhibited comparable heteroclicity (≅twofold relative to H33Lγ1) for NIP 2 over NP 2 (Fig. 1e). We conclude that even at the level of individual GC B-cell clones, MHCII haploinsufficiency has little or no effect on primary GC B-cell responses to NP-OVA.
These short-term reconstitution experiments support the notion that pMHCII density controls B-cell entry and/or proliferation in nascent GCs 1 . Nonetheless, whereas MHCII haploinsufficient B cells are strongly disadvantaged in the earliest stages of the GC response, in organized GCs, MHCII +/+ and MHCII +/− B cells expressing identical BCR appear to be equally fit.
In our short-term transfer studies, diversity in BCR affinity for NP or NIP was minimized by using B1-8 VDJ knock-in donors 23 .
To determine whether MHCII haploinsufficient B cells exhibit reduced fitness in GCs when BCR affinity is not constrained, we followed the dynamics of MHCII +/+ and MHCII +/− B cells responding to immunization with NP-OVA and asked whether MHCII differences have longer term effects in an environment of direct competition. In the large group of chimeric mice studied As for short term transfers (Fig. 2), CD45.1 + MHCII +/+ B cells in chimeric mice exhibited a significant advantage over CD45.2 + MHCII +/− B cells for acquisition of the GC phenotype (Fig. 4). By day 4 post immunization, T-cell-dependent B-cell proliferation and acquisition of the GC phenotype 24,25 was biased in favor of MHCII +/+ B cells (MHCII +/+ :MHCII +/− = 1.9:1) (Fig. 4a). This bias for MHCII +/+ GC-phenotype B cells grew on days 5 and 6, as GCs form and become organized; by day 8, ratios of MHCII +/+ :MHCII +/− GC B cells stabilized at ≅1.7:1, remaining stable on days 16 and −24 even as GC responses waned to ≅0.5% of the MF B-cell compartment (Fig. 4a). Immunization of BM chimeric mice revealed a common pattern of B-cell competition: B cells with higher MHCII densities are strongly advantaged in the earliest stages of the GC response but exhibit no increase in fitness once the GCs become organized. This ratio (≅2:1) of MHCII +/+ :MHCII +/− B cells is conserved even in the chronically activated, Peyer's patch GCs of BM chimeric mice (Fig. 4a).

Discussion
BCR affinity maturation in GCs has been proposed to be driven by the quality of T FH and GC B-cell interactions, which in turn are determined by pMHCII density on the B-cell surface 16 . To explore this hypothesis quantitatively, we established by B-cell transfer, a venue for competition between congenic B cells that differed by their capacity to express MHCII. B cells hemizygous for MHCII I-A b19 were haploinsufficient for MHCII expression 26 regardless of physiologic state (Figs. 1a, 2j and Supplementary  Fig. 5). The haploinsufficiency extended to pMHCII as well; NPbinding B cells from MHCII +/− mice immunized with NP-SA-Eα expressed half the level of Eα peptide/MHCII complex of MHCII +/+ controls ( Fig. 1b and Supplementary Fig. 3). This quantitative model allows us to explore the role of pMHCII density in clonal selection in humoral immune responses.
MHCII +/− mice mounted serum IgG and GC responses to NP-OVA that were indistinguishable from those of B6 controls (Fig. 1d). This identity shows that pMHCII density per se does not determine the magnitude of humoral responses and recalls the observation that GC responses in B1-8 hi and B1-8 lo mice exhibit comparable kinetics despite great differences (40-fold) in BCR affinity for NP 23 . However, when MHCII +/+ and MHCII +/− B cells directly compete after immunization, MHCII +/+ B cells exhibit a significant advantage over MHCII +/− competitors in populating nascent GCs (Fig. 2d-f). This early advantage is not associated with differences in antigen binding, as it persists even when BCR affinities between the competing B cells are homogenized by a shared VDJ knock-in (Fig. 2d-f). This early advantage of MHCII +/+ B cells follows B-cell activation, as the appearance and numbers of IgD low NP + B cells following immunization with NP immunogens is identical for MHCII +/+ and MHCII +/− B cells (Supplementary Fig. 3). We conclude that the early advantage of MHCII +/+ B cells reflects pMHCIIdependent selection at the initial, T H -dependent entry of antigenactivated B cells into humoral responses 1,27 .
To our surprise, the early competitive advantage for MHCII +/+ B cells was lost once GCs became organized (Figs. 2f  and 4a). The skewed ratios of MHCII +/+ and MHCII +/− B cells entering GCs become stable and persist without significant change once GC organization is established (Figs. 2f and 4a). Stability in the ratios of MHCII +/+ :MHCII +/− GC B cells did not reflect anatomical segregation as histologic studies confirmed that all GCs were populated by both MHCII +/+ and MHCII +/− B cells (Fig. 3c-f). In GCs, B cells with twofold differences in MHCII expression exhibit the same capacity for persistence within LZ and DZ GC pools. If affinity-dependent competition among GC B cells essentially reflects the "mapping" of BCR affinity onto pMHCII density, our experiments indicate that while MHCII density and T-cell help are limiting factors in pre-GC selection, in established GCs this selection is less stringent. This relaxation of pMHCII selection intensity may support the "permissive" affinity-dependent selection noted for complex protein antigens in established GCs 22 . Selection on complex phenotypes often results in compensation. For example, mice deficient in CD21/CD35 exhibit reduced serum IgG responses but enhanced affinity maturation 28 and mixed chimera mice with normal and ICOSL −/− B cells show affinity compensation in ICOSL −/− GC B cells and plasmacytes 29 . To exclude the possibility that MHCII +/− GC B cells compensate for lower pMHCII densities by increased BCR affinity, we determined the BCR AvIns for ≈800 individual MHCII +/+ and MHCII +/− GC B cells to NP and NIP (Fig. 5).
Heteroclitic, affinity maturation was highly similar between MHCII +/+ and MHCII +/− GC B cells, with no compensatory affinity increases observed in MHCII +/− GC B cells (Fig. 5 and Supplementary Fig. 6). Mean AvIn values did not differ significantly between MHCII +/+ and MHCII +/− GC B cells at any sample time and AvIn distributions were comparable in each cohort (Fig. 5). Direct measurement of BCR avidity obviates the limitations of predicting affinity maturation by enumerating affinity-enhancing mutations 29 .
Immunofluorescence studies of immunized chimeric mice showed that all GCs comprised B cells from both MHCII +/+ and MHCII +/− donors, demonstrating that these B cells competed in common GC niches (Fig. 3c-h). That both groups exhibited similar AvIn distributions, including comparable lower avidity "tail" populations ( Fig. 5) 22,30 , ensures that spatial segregation cannot account for persistence and continuing affinity maturation of MHCII +/− GC B cells.
We find it highly unlikely that pMHCII on MHCII +/+ and MHCII +/− GC B cells are ever equalized. Bannard et al. have reported that MHCII molecules turn over rapidly in DZ B cells by ubiquitin-mediated degradation 31 , perhaps to ensure that pMHCII density accurately represents BCR affinity. If so,  (Fig. 3a). a NIP 25 -, b NIP 2 -, and c NP 2 -specific AvIn values (relative to the H33Lγ1 standard) are shown. Each symbol (MHCII +/+ , black circles; MHCII +/− , triangles) represents the AvIn value of one IgG + clonal culture sample (n = 12-220, Table 1). Boxes represent the 25th, 75th percentiles and median. Bars (blue) indicate the geometric mean and ±S.D. Statistical significance (P < 0.05) was measured using two-way ANOVA with Friedman test followed by Dunn's multiple comparison post tests. N.S. not significant pMHCII densities on WT and haploinsufficient B cells would reflect the abundance of the immediate precursor, the MHCII-CLIP complex, which is halved also in MHCII +/− GC B cells ( Supplementary Fig. 5).
The notion that GC B cells with higher pMHCII densities are advantaged in GCs by enhanced T FH help came from experiments in which antigen linked to DEC-205 mAb (αDEC-205-Ag) was delivered to DEC-205 + GC B cells 5,9 . This targeted loading of antigen is BCR and FDC independent but correlated with increased B:T FH interaction, DZ proliferation, and V H mutation frequencies 9,15 . We see no evidence for these effects when competing B cells acquire antigen via their BCR. What is surprising is that B-cell entry into GC responses is strongly affected by a twofold reduction in MHCII and pMHCII (Figs. 2 and 4). If the GC represents the paradigm for affinity-dependent, B-cell selection, it seems counter-intuitive that T:B interactions that initiate humoral responses are more stringent than those within the GC.
Results from other experimental models that establish competition between B cells with higher-or lower levels of MHCII expression are similar to our own. For example, in the absence of H2-O expression, B cells with elevated pMHCII densities exhibit a pronounced advantage over WT B cells on entry into GCs, but once GCs become established ratios of H2-O-deficient and -sufficient GC B cells remain constant for 21 days 32 . Likewise, Bannard et al. infected mixed BM chimeric mice containing B cells that expressed normal MHCII or MHCII resistant to ubiquitination mediated turnover with influenza 31 . One week after infection, ubiquitin resistant GC B cells represented 10−50% of GC B cells while at 5 weeks these cells remained almost as abundant at 0-40% 31 . Finally, immunization of mixed BM chimeric mice containing WT and CD83 −/− B cells (with lower MHCII expression) with SRBC resulted in an early preference for CD83-sufficient GC B cells, but with little or no change in CD83 +/+ :CD83 −/− ratios between days 6 and 12 post immunization 33 .
The permissive nature of selection in GCs 22 was recently underscored when Turner et al. showed that B cells responding to a dissimilar antigen could enter into and persist within an ongoing GC response 34,35 . These authors show that HyHEL-10 B cells specific for hen-(HEL) and duck egg lysozyme (DEL) exposed ex vivo to DEL-OVA enter and proliferate in GCs elicited by DEL-OVA or OVA immunization with equal efficiency 35 . One surprising interpretation of this result is that the B cell's initial exposure to antigen determines its fitness for the GC response rather than the ability to recover antigen repeatedly from GC FDCs. If these results 34,35 are generalizable, current models for affinity-dependent selection in GCs require substantial revision. At least one possibility is that in the LZ, low affinity GC B cells have a substantial chance of receiving "bystander help" from local T FH activated by fitter cells. Such unspecific help would be consistent with recent results from the Nussenzweig laboratory 36 showing that apoptosis in the GC LZ is essentially independent of BCR affinity.
The GC is a dynamic microenvironment where antigenactivated B cells iteratively undergo proliferation, hypermutation, and affinity-driven selection. By direct observation, selection for higher BCR affinities in GCs is rapid and relies on the unequal success of mutant B cells in generating progeny. Nonetheless, GC B-cell populations also comprise substantial subsets of mutated B cells with very low BCR avidities and GC selection may be functionally permissive for these less fit populations 22,30 . Given that pathogens evolve under selection by host immunity, permissive selection may be a strategy to optimize memory B-cell compartments against pathogen variants that have escaped immune control. The development of broadly neutralizing antibodies to influenza and HIV from B-cell clonal lineages characterized by extraordinary frequencies of V(D)J mutation are consistent with this tortuous pathway to protective efficacy 37-40 . In summary, by allowing congenic MHCII +/+ and MHCII +/− B cells to compete directly, we show that WT and haploinsufficient GC B cells exhibit similar fitness in GCs as determined by their proliferation, persistence, mutation frequencies, and affinity maturation. In contrast, haploinsufficient B cells are significantly disadvantaged during the initiation of humoral responses, most likely during the initial T:B collaboration that marks the start of humoral responses. If affinity-driven selection is determined by pMHCII density on B cells, that mechanism appears to be significantly more stringent at the initiation of the humoral response than during the GC reaction itself. Short-term transfer and mixed bone marrow chimeric mice. For short-term cell transfers, single-cell suspensions were harvested and processed from spleens of B1-8 +/+ MHCII +/+ (CD45.1 + ) and B1-8 +/+ MHCII +/− (CD45.2 + ) or B1-8 +/+ MHCII +/+ (CD45.2 + ) mice. Splenocytes were stained with biotinylated-Abs (αCD4, αCD11c, αCD43, αCD90.2, αF4/80, and αGL-1) and subsequently labeled with Streptavidin MicroBeads (Miltenyi Biotec). B cells were then negatively purified using magnetic activated cell sorting with CS column on a VarioMACS separator (Miltenyi Biotec). After sorting, purified B cell samples were stained and examined using flow cytometry to determine the purity and percentage of NP + populations. 100 μL of cell mixtures containing 2 × 10 6 cells with 1:1 ratios of NP + CD45.1 +/+ B1-8 +/+ MHCII +/+ and NP + CD45.2 +/+ B1-8 +/+ MHCII +/− B cells were transferred i.v. to individual recipient (B6.SJL × B6) F 1 (CD45.1 + /CD45.2 + ) mice. To generate mixed BM chimeric mice, (B6.SJL × B6) F 1 (CD45.1 + /CD45.2 + ) mice were sub-lethally irradiated with a single dose of 6.5 Gy X-ray and then injected i.v. with equal numbers (2.5 × 10 6 cells) of BM cells harvested from B6.SJL (CD45.1 + ) and MHCII +/− (CD45.2 + ) or B6 (CD45.2 + ) mice. Reconstituted mice were rested for 6-8 weeks before use in experiments.
Single B-cell culture. Single MF and GC B cells from control mice and immunized mice were expanded in the presence of NB-21.2D9 feeder cells (Nojima cultures) as described 22 . Briefly, NB-21.2D9 feeder cells (2000 cells/well) were pre-seeded into 96 well culture plate with RPMI 1640 media (Life Technologies) supplemented with 10% FBS (Thermo Fisher Scientific), 10 mM HEPES, 1 mM sodium pyruvate, 1× MEM nonessential amino acid, 100 U/mL penicillin-streptomycin and 5.5 × 10 −5 M 2-mercaptoethanol (All Life Technologies). Next day, single MF and GC B cells were directly sorted into NB-21.2D9 culture plates and cultured in the presence of exogenous recombinant IL-4 (2 ng/ml, Peprotech) and CD154/BAFF/IL-21-expressing NB-21.2D9 cells. Two-thirds of culture media were replaced with fresh media daily from days 2 to −8. On day 10, culture supernatants and cell pellets were harvested for subsequent ELISA IgG determinations and V(D)J amplifications, respectively.
ELISA and luminex multiplex immunoassays. Concentrations of IgG in harvested culture supernatants were determined by standard sandwich ELISA as described 22 . Briefly, 384-well ELISA plates (Corning) were coated with anti-mouse Igκ Ab and anti-mouse Igλ Ab (2 μg/ml each; Southern Biotech) in carbonate buffer overnight at 4°C. After blocking with PBS containing 0.5% BSA for 1 h at room temperature, serially diluted samples (1:100, 1:1000, and 1:2000 dilutions) or the H33Lγ1 mAb standard were applied to the plate and incubated overnight at 4°C. After washing, HRP-conjugated anti-mouse IgG (1:5000 dilution; Southern Biotech) was added and incubated at room temperature for 1 h. After washing, bound HRP activity was visualized using a TMB peroxidase kit (BioLegend) and the optical densities (O.D.) were determined at 450 nm with a SpectraMax M2 Microplate Reader (Molecular Devices). IgG + supernatants screened by ELISA and serum samples were subjected to Luminex Multiplex Immunoassay for the detection of avidity against different antigen as reported 41 . Briefly, 1 × 10 7 coded MicroPlex microspheres (Luminex) were covalently linked to 50 µg of NP 2 -BSA, NP 30 -BSA, NIP 2 -BSA, NIP 25 -BSA (Biosearch Technologies), OVA (Sigma-Aldrich), Igκ, Igλ, or IgG (Southern Biotech), respectively. Conjugated microsphere mixture were incubated with diluted samples for 2 h at room temperature. After washing, any IgG Ab bound to the microspheres was detected with 2 µg/ml PE goat anti-mouse IgG (Southern Biotech). Fluorescence was measured on a Bio-Plex 3D suspension array system (Bio-Rad Laboratories).