Local antibody feedback limits affinity maturation in the germinal center and promotes epitope spreading

SUMMARY

Circulating antibodies from previous immune encounters impact initiating humoral responses. Here we investigated how local epitope-specific competition shapes ongoing germinal center (GC) responses by delivering an mRNA-LNP encoded membrane-bound immunogen displaying three conserved HIV-1 Envelope (Env) epitopes to mouse models bearing B cell receptors (BCRs) of defined affinities. High-affinity B cells exhibited shorter GC residency than lower affinity counterparts. B cells engaged GC reactions at equivalent rates in the presence or absence of clonal lineages binding the same epitope with similar affinities; however, higher-affinity clones suppressed lower-affinity counterparts targeting the same epitope. Spatial transcriptomics revealed plasma-like cells within and adjacent to the GC, and early IgG was detectable in draining lymph nodes. Our findings suggest that a self-modulated local antibody feedback loop limits epitope-specific recognition—dampening selection for higher-affinity B cells and facilitating epitope spreading by redirecting the response toward alternative epitopes.

eTOC.

Feedback from circulating antibodies can shape initiating immune responses. Yan et al examine ongoing germinal center (GC) responses upon immunization with an mRNA-LNP encoded membrane-bound immunogen displaying three conserved HIV-1 Envelope epitopes, and find antibody-mediated intra-epitope competitive effects and a contribution of local antibodies to determining B cell residency in GCs and affinity maturation.

Graphical Abstract

INTRODUCTION

Affinity maturation is the fundamental process underpinning the humoral response to infection or vaccination^1,2. Exposure to antigen activates a polyclonal B cell response derived from those cells with B cell receptors (BCRs) meeting a minimum affinity threshold³. Those with very high initial affinities will differentiate into extrafollicular plasma cells (PC)⁴, while others will form the germinal center (GC). GCs generate and diversify B cells: therein, somatic hypermutation (SHM) introduces mutations into the variable region of BCRs, producing further variation in affinity for the antigen⁵.

This expanded diversity is shaped by the pressure of competition in GC selection. Expansion in the GC is predicated upon survival and proliferation signals from T follicular helper (Tfh) cells^6,7. B cells engage Tfh by presenting internalized antigen complexed with MHC II^8,9; the efficiency of this presentation is affinity-dependent^3,10. Though low-affinity clones can enter and participate in GCs, they may be excluded or disfavored in the presence of higher affinity competitors^11–13. Lower affinity B cells surviving in the GC may be selected for differentiation into memory B cells (MBC)^14,15, but whether fate differentiation into MBC and long-lived plasma cells (LLPC) is determined by BCR affinity is disputed¹⁶, and there is substantial variation in the affinity of antibodies expressed by PC emerging from the GC¹⁷. Ultimately, slower antigen-BCR dissociation reaches a point of diminishing returns in terms of receptor activation enhancement, suggesting that, even if high-affinity cells appear early in the response, maturation in the GC reaches a ceiling above which affinity improvements are no longer visible to selection^3,18,19. The antibodies of varying affinities produced by terminally differentiated B cells arising throughout this response may also shape the interactions of B cells in GCs.

Feedback from existing antibodies influences the downstream humoral response^20–23, but its influence on the diversity of B cells entering and exiting GCs over the course of the response remains underexplored. Pre-existing antibody allows immune complex formation²⁴, and the BCR affinity activation floor lowers for soluble antigens complexed with antibody³. Conversely, circulating antibody may be competitive, intensifying selective pressure in GCs by limiting antigen access²⁵ or blocking cognate B cell entry²⁶. Prior work from our group suggests a circulating Ab affinity tipping point, with lower affinity Abs enhancing GC responses and higher affinity Abs blocking²⁷. Antibodies thus affect B cell responses in dose, affinity, and epitope-specific manners²⁸. However, when high-affinity mAbs are present prior to SARS-CoV-2 immunization, a preponderance of low-affinity clones are observed in post-vaccination GCs²⁹; this booster response is also characterized by a shift to subdominant epitopes. Breadth, and not potency alone, is an important determinant of protection—in malaria, vaccine boosting Abs from prior vaccination limit overall recall responses, but those later responses are characterized by a shift towards subdominant epitopes³⁰.

Feedback may be particularly important for vaccines requiring a longer immune response, multiple boosts, or multiple epitopes. Those are characteristics of HIV-1 retrovaccinology based on broadly neutralizing antibodies (bnAbs) to conserved epitopes on the HIV Envelope (Env). In “germline-targeting” (GT) strategies, priming immunogens are designed for high affinity to the unmutated B cells inferred to give rise to bnAbs; these precursors are often rare and lack affinity to the mature HIV trimer^31–42. After priming, several boosts with sequentially more native-Env-like immunogens may be required to guide precursors towards functional anti-HIV bnAbs^32,39,41,43. An effective vaccine may need to elicit bnAbs to two or more distinct epitopes on Env to prevent viral escape^44–50. Individuals undergoing these proposed vaccination schedules would thus have a variety of B cell lineages at varying affinities for conserved epitopes on Env and fluctuating antibody titers, converging on the same native trimer. A complicating factor is the capacity to introduce membrane-anchored antigen presentations via mRNA-LNPs; increasing avidity by presenting antigen on a surface may lower the threshold for initial activation⁵¹, or otherwise unpredictably alter the competitive environment.

Here we established a series of late-stage HIV preclinical mouse models carrying human BCRs of known affinities to one of three epitopes on the HIV-1 Env: the CD4 binding site (CD4bs), V2-Apex, and V3-glycan. These BCRs are “intermediate”—falling between germline and mature bnAbs in affinity and sequence evolution. By vaccinating variable-affinity preclinical B cell models with an mRNA-LNP encoding a single, membrane-anchored native trimer immunogen presenting all three epitopes, we found that lower affinity cells reacted robustly to the trimer in isolation, but that competition from higher affinity lineages to the same epitope was inhibitory. Conversely, lower affinity B cell lineages to the same epitope, and lineages with different epitope specificities, did not affect each other’s GC kinetics. High-affinity antibodies alone reproduced inhibition, suggesting an antibody feedback mechanism; we furthermore identified local PCs within GCs as the likely sources of blocking antibody. Our findings provide insights into the mechanisms governing B cells within GCs and also offer guiding principles for the rational design of vaccines to elicit potent and broad antibody responses.

RESULTS

High-affinity BCR clones are efficiently recruited to GCs but are not preferentially maintained over lower affinity counterparts

To investigate how inherent affinity affects GC kinetics following immunization with HIV-1 Env glycoproteins, we developed mouse models with BCRs derived from the phylogenies of human bnAbs (Fig. S1A–C). We selected clinically-relevant bnAbs targeting three non-overlapping neutralizing epitopes on the highly glycosylated HIV-1 Env: N6-I3 (CD4bs)^52,53, PCT64–18D (V2-apex)⁵⁴, and minBG18.6 (V3-glycan) (Fig. S1D–F)⁴⁰. Using our CRISPR/Cas9 homologous directed recombination method^55,56, we generated transgenic mouse models with B cells expressing the heavy chains (HCs) and/or light chains (LCs) of these three Abs. Lines were bred to homozygosity before use; follicular cell development was normal, and though lines showed some variation in peritoneal B cells all expected populations were present (Fig. S1G&H). B cells isolated from these lines survived, differentiated and proliferated at normal rates in vitro (Fig. S1I). Peripheral blood mononuclear cells (PBMCs) in the resulting IgH^N6-I3/N6-I3IgK^N6-I3/N6-I3 (referred to as CD4bs-Int¹ below), IgH^{PCT64-18D/PCT64-18D}IgK^{PCT64-18D/PCT64-18D} (V2-Int¹), and IgH^{minBG18.6/ minBG18.6}IgK^WT/WT (V3-Int¹) mouse lines were examined by 10x single-cell RNA sequencing (scRNAseq). N6-I3 HC and N6-I3 LC sequences were co-expressed by 93.1% of cells from CD4bs-Int¹ mice, and PCT64–18D HC and PCT64–18D LCs by 99.7% of cells from V2-Int¹ mice, confirming intended BCR assembly. For V3-Int¹, HCs paired with diverse native murine LCs, most frequently IGLV1, IGKV4–68, IGKV4–61, IGKV12–98, and IGKV4–91 (Fig. 1A).

Figure 1. An mRNA-encoded membrane-anchored BG505 native trimer recruits intermediate bnAb-precursor B cells targeting different epitopes on Env.

(A) (Top) Nested pie chart of CD4bs-Int¹ heavy chain (HC) (red), murine HC (gray), human CD4bs-Int¹ light chain (LC) (shaded red) and murine LC (shaded gray) sequences amplified from single-cell sorted epitope-specific (BG505⁺KO⁻) naive B cells in CD4bs-Int¹ mice. (Middle) As top for V2-Int¹ HC (blue), murine HC (gray), human V2-Int¹ LC (shaded blue) and murine LC (shaded gray). (Bottom) As prior for human V3-Int¹ HC (purple), murine HC (gray) and murine LC sequences (multi-colored). Outer rings represent HCs and inner rings represent LCs. Data are shown from one representative of two experiments (n=1–2 mice). Total denotes single cells amplified. For V3-Int¹, the top 15 of 20 LCs amplified are listed.

(B) Representative FACS plots of BG505-double-positive and BG505-epitope-specific-KO-negative peripheral B cells in (descending): naïve CD4bs-Int¹, V2-Int¹, V3-Int¹ mice, or WT controls. Events were pre-gated on lymphocytes/singlets/CD4⁻CD8⁻F4/80⁻Gr1⁻/B220⁺ B cells.

(C) Quantification of (B). BG505-specific binders in blood peripheral B cells. n=8–10 for CD4bs-Int¹, V2-Int¹, or V3-Int¹ mice, n=5 for WT, pooled from 2–4 individual experiments. Bars are mean + SD.

(D) Affinity of CD4bs-Int¹ or V2-Int¹ or V3-Int¹ Ab against BG505 trimer measured by SPR dissociation constant. For V3-Int¹, Abs were expressed with human HC and representative murine LCs (see Key Resources Table). Each dot represents the mean of 4 technical replicates. Dotted line marks LOD.

(E) Schematic of GC recruitment evaluation experiments for CD4bs-Int¹, V2-Int¹, or V3-Int¹ CD45.2 B cells. Immunization output panels below display representative data from one of at least two experiments with 3–5 mice per condition.

(F) Representative FACS plots of B cells obtained from dLNs after immunization with mRNA-LNP encoding BG505 Env glycoprotein trimer. Days post-immunization at top. Events pre-gated on lymphocytes/singlets/live/CD4⁻CD8⁻F4/80⁻Gr1⁻/B220⁺ B cells and represent GC in B cells, or CD45.2 cells in GC.

(G) Time course plots of GC cells as a percentage of total B cells (left) and CD45.2⁺ cells as a percentage of GC B cells (right). Values of zero were plotted as UD (undetected) on the log₁₀ scale. Each dot represents one mouse. Bars are mean ± SD.

(H) Dotted violin plots of HC amino acid mutations across all sites at D21. Each dot represents an HC sequence from one B cell.

(I) Pie charts of class-switch profiles at D21. Total=sequences amplified.

(J) Nested pie chart showing V3-Int¹ HC and LC usage from single-cell sorted epitope-specific (BG505⁺KO⁻) B cells at D28; no murine HC (grey) detected. Outer layer, human V3-Int¹ IGHV; inner layer, murine IGKV. Nested pie chart shows average of all mice in a group (Total=sequences amplified, n=mice).

See also Figure S1.

We required an immunogen capable of activating B cells targeting diverse epitopes across a broad affinity range to evaluate how inherent BCR affinity impacts GC kinetics. BG505 MD39.3 gp151 (BG505 hereafter), which has performed well in NHPs and a human clinical trial^57,58, met that criterion. PBMCs from all three lines showed competent specific binding to BG505, and lack of binding to corresponding epitope knockout (KO) probes (Fig. 1B and Fig.1C). All three BCR variants showed detectable but varied affinities to BG505: the dissociation constant [K_Ds] of CD4bs-Int¹ was 3 nM, V2-Int¹ 770 nM, and V3-Int¹, where the minBG18.6 heavy chain paired with representative mouse light chains, ranged from 820 nM to the limit of detection (LOD) at 10 μM (Fig. 1D).

CD4bs-Int¹, V2-Int¹, or V3-Int¹ CD45.2^+/+ cells were adoptively transferred into WT CD45.1^+/+ hosts individually to reach frequencies of 20 in 10⁶, and recipient mice were then immunized with an mRNA-LNP-encoded membrane-anchored BG505 (Fig. 1E). The GC response and the proportion of CD45.2⁺ B cells in GCs was evaluated in dLNs at D7, D14, and D28. Across all transfer recipients, GCs constituted ~2–5% of B cells on average (Fig. 1F and Fig. 1G). CD45.2⁺ cells from each line were effectively activated and present in GCs at D7, though their average fractions in GCs ranged from 5% for V3-Int¹ to 67% for CD4bs-Int¹ and 78% for V2-Int¹. At D14 and D28, V2-Int¹ and V3-Int¹ CD45.2 B cells retained a share of the GC (43% and 47% for V2-Int¹, 53% and 27% for V3-Int¹); V3-Int¹ responses displayed higher variability at D28, consistent with their polyclonality. CD4bs-Int¹ cells, which have the highest affinity for their epitope (K_D, 3nM), dropped to 2.88% by D14 and 0.16% at D28 (Fig. 1F and Fig. 1G). Sequencing by single-cell BCRseq of sorted antigen-specific CD45.2⁺ B cells at D21 revealed that immunization stimulated class-switching and SHM in all three cell lines (Fig. 1H, Fig. 1I, Fig. S1J). Immunization selected for particular V3-Int¹ BCRs, with a shift towards IGKV4–91 (79.88%) and IGKV4–61 (10.93%) (Fig. 1A, and Fig. 1J). Thus, independent of initial affinity, all three cell lines could be stimulated by an mRNA-LNP-delivered membrane-anchored immunogen, enter the early GC, and undergo class switching and SHM, but each line displayed distinct GC kinetics, with a shorter GC half-life for the line with highest initial affinity.

High-affinity clones do not influence GC dynamics of low-affinity B cells targeting other epitopes

We next sought to determine whether the different B cell lines would compete despite targeting different epitopes on the same trimer. CD4bs-Int¹, V2-Int¹, and V3-Int¹ CD45.2^+/+ B cells were adoptively transferred into WT CD45.1^+/+ mice individually to reach a frequency of 20 cells per 10⁶ total B cells, or in equal-proportion combinations to reach 20 (1x Mix) or 60 (3x Mix) cells per 10⁶ total B cell. Total CD45.2^+/+ B cells in 1x Mix matched individual transfer recipients; in the 3x Mix, B cells from each individual CD45.2^+/+ line were at frequencies equivalent to individual transfers, but there were more total CD45.2^+/+ B cells. Recipient mice were immunized intramuscularly with 2 μg of BG505 mRNA-LNP one day later. Local, draining LNs were analyzed by flow cytometry at D7 and D21 post-immunization (Fig. 2A). GC sizes were similar across groups at both timepoints; V2-Int¹ alone produced slightly larger GCs at D7 than 3x Mix, but this difference did not persist to D21 (Fig. 2B–2C). Mice transferred with V3-Int¹ alone averaged fewer CD45.2⁺ B cells in GCs at D7 relative to recipients of other individual transfers or either Mix, but no significant differences were observed at D21 (Fig. 2D). To determine the composition of the CD45.2 B cells in the mixed transfer recipients, all CD45.2 cells in the GCs were sorted for 10x scRNA-seq. CD4bs-Int¹, V2-Int¹, and V3-Int¹ cells were recovered from both the 1x and 3xMix recipients at variable frequencies (Fig. 2E). The ratio for each CD4bs-Int¹, V2-Int¹, or V3-Int¹ cell identified by scRNAseq was multiplexed back to the CD45.2 cell percentage obtained by flow cytometry for each individual mouse to estimate specific single CD4bs-Int¹, V2-Int¹, or V3-Int¹ B cell rates in dLN B cells. The percentage of CD4bs-Int¹, V2-Int¹, and V3-Int¹ B cells among total B cells at D21 was indistinguishable between single transfers and either 1x or 3x Mix (Fig. 2F, Fig. S2A). Slightly higher mutation rates were observed in all three B cell lines isolated from 1x Mix recipients compared to individual transfer recipients; in 3x Mix recipients, SHM was only increased for V3-Int¹ HCs (Fig. S2B). No differences were observed in class-switching profiles, or the identities of the three most common LC found with V3-Int¹, though LC frequencies varied by treatment (Fig. S2C–D). The GC kinetics of each epitope-specific cell line was thus unaffected by the presence of lines targeting non-overlapping epitopes on the same immunogen.

Figure 2. A single trimer simultaneously activates intermediate B cells targeting V2-apex, V3-glycan, and the CD4bs in the same host.

(A) Schematic of individual- or co-adoptive-transfer and immunization experiments for CD4bs-Int¹, V2-Int¹, and V3-Int¹ B cells.

(B) Representative FACS plots of B cells obtained from dLNs after individual- or co-adoptive-transfer. Events pre-gated on lymphocytes/singlets/live/CD4⁻CD8⁻F4/80⁻Gr1⁻/B220⁺ B cells and represent (upper) GC and (lower) CD45.2 cells in GC.

(C–D) Quantification of (C) GC cells as the percentage of total B cells and (D) CD45.2⁺ cells as the percentage of GC B cells at D7 and D21. q-values were calculated by Kruskal-Wallis test followed by pairwise comparisons with Benjamini–Krieger–Yekutieli (BKY) correction.

(E) CD4bs-Int¹ (red), V2-Int¹ (blue), and V3-Int¹ (purple) B cell lineage frequency analysis from 10x scRNAseq of GC CD45.2 B cells sorted at D21 from the 3x Mix (top) and 1x Mix (bottom) groups. Pie charts are averages of all mice in a group (Total=sequences amplified, n=mice).

(F) CD4bs-Int¹, V2-Int¹, and V3-Int¹ BCR composition of GC CD45.2 B cell sequences as percentages of total B cells at D21. Percentage was calculated from FACS data for individual groups, and from 10x scRNAseq plus FACS data for Mix groups. Values of zero on the log₁₀ scale were plotted as UD. q-values were calculated by 2-way ANOVA test followed by pairwise comparisons with BKY correction.

Figures represent data from one of at least two experiments with 3–5 mice per condition. Bars are mean + SD. Each dot represents one mouse. q-values: *p < 0.05, ns = not significant. See also Figure S2.

B cell lines with low, overlapping affinities to the same epitope can be co-stimulated

Although we did not observe inhibition among B cell lines targeting different epitopes on the same trimer, to explore possible crosstalk between lineages targeting the same epitope, we generated an additional mouse line using minBG18.11 (V3-Int²), which targets the same V3-glycan epitope as minBG18.6 (V3-Int¹) but represent a more affinity-matured variant with higher SHM from germline⁴⁰. Homozygous V3-Int² mice (IgH^{minBG18.11/ minBG18.11}IgK^WT/WT) displayed normally developing B cells which could survive, proliferate, and differentiate in vitro (Fig. S3A–C). BG505-specific PBMC frequency was 27.9%, similar to that observed in V3-Int¹ (Fig. 3A, Fig. S3D, Fig. 1B–C). As in the V3-Int¹ mouse line, the V3-Int² HC sequences also paired with a variety of endogenous murine LCs, most commonly IGKV12–46, IGKV12–41, IGKV10–94, IGKV4–50, and IGKV12–98 (Fig. 3B). V3-Int¹ and V3-Int² displayed similar GC kinetics post-immunization by BG505 mRNA-LNP, with V3-Int² showing a slightly decreased GC size from 3.2% to 1.5% between D14 and D28; CD45.2⁺ cells in GC were 35% at D14 and 45% at D28 (Fig. 3C–D, Fig. S3E–F). While V3-epitope recognition is strongly HC dependent⁴⁰, the recombination with murine LCs adds further diversity and the affinity of monoclonal antibodies (mAbs) representative of V3-Int² BCRs to BG505 ranged from 1 μM to 10 μM, comparable to V3-Int¹ representative mAbs (Fig. 3E).

Figure 3. Two B cell lineages targeting V3-glycan can be simultaneously activated without competitive interference.

(A) Representative FACS plots of BG505-double-positive and BG505-epitope specific KO-negative peripheral B cells in naïve V3-Int² mice. Events were pre-gated on lymphocytes/singlets/CD4⁻CD8⁻F4/80⁻Gr1⁻/B220⁺ B cells.

(B) Nested pie chart of human V3-Int² HC (purple), murine HC (gray) and murine LC sequences (multi-colored) amplified from single-cell sorted epitope-specific (BG505⁺KO⁻) naive B cells in V3-Int² mice. Legend shows names of most frequent 11 of 43 LCs amplified. Outer rings represent HCs and inner rings represent LCs. Total = single cells amplified.

(C) Representative FACS plots of B cells obtained from dLNs in V3-Int² adoptively transferred mice at D14 and D28 after immunization with mRNA-LNP-encoded BG505. Events pre-gated on lymphocytes/singlets/live/CD4⁻CD8⁻F4/80⁻Gr1⁻/B220⁺ B cells and represent GC, CD45.2 cells in GC.

(D) Quantification of (C). Each dot represents one mouse. Bars are mean ± SD.

(E) Affinity measurement of V3-Int¹ and V3-Int² Ab against BG505 trimer measured by SPR dissociation constant. V3-Int¹ and V3-Int² Abs were expressed with human HC and representative murine LCs, detailed sequence can be found in the key resource table. Each dot represents the mean of four technical replicates. Dotted line marks LOD. Data for V3-Int¹ reproduced from Fig. 1D.

(F) Schematic of individual or co-adoptive-transfer and immunization experiments for V3-Int¹ and V3-Int² B cells.

(G) Representative FACS plots of B cells obtained from dLNs of mice V3-Int¹ and V3-Int² B cells at D14 and D28 after BG505. Events were pre-gated on lymphocytes/singlets/live/CD4-CD8-F4/80-Gr1⁻/B220⁺ B cells and represent GC and CD45.2 cells in GC.

(H) (Upper) GC cells as a percentage of total B cells and (lower) CD45.2⁺ cells as a percentage of total GC B cells at D14 and D28. Each dot represents one mouse. q-values calculated by 2-way ANOVA with pairwise post-hoc comparisons adjusted using BYK correction.

(I) The V3-Int¹ (light purple) and V3-Int² (dark purple) B cell lineage frequency analysis from 10x scRNAseq of GC CD45.2 B cells sorted from the Mix group. Pie charts are averages of all mice in a group (Total=sequences amplified, n=mice).

(J) V3-Int¹ and V3-Int² BCR composition of GC CD45.2 B cell sequences as a percentage of total B cells. Percentage was calculated from FACS data for individual groups, and from 10x scRNAseq plus FACS data for Mix groups. Values of zero were plotted as UD on the log₁₀ scale. q-values calculated by 2-way ANOVA Kruskal-Wallis with pairwise post-hoc comparisons adjusted using BYK correction.

(K) Dotted violin plot of HC amino acid mutations across all sites at D28. Each dot represents an HC sequence from one B cell. Analyzed using Mann-Whitney’s test.

(L) (Upper) Nested pie chart showing V3-Int¹ HC and LC usage from single-cell sorted CD45.2 B cells at D28 from individual transfer and mix recipients. Outer ring, human V3-Int¹ IGHV; inner ring, murine IGKV. (Lower) Nested pie chart showing V3-Int² HC and LC usage as upper. Nested pie charts are averages of all mice in a group (Total=sequences amplified, n=mice).

Panels in C, D, G, H display representative data from one of at least two experiments with 3–5 mice per condition. Bars are mean ± SD in (H) and (J). Significance is indicated as: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; ns = not significant. See also Figure S1.

To establish a repertoire of B cells targeting the same epitope, V3-Int² and V3-Int¹ CD45.2^+/+ B cells were adoptively transferred into CD45.1^+/+ host mice either individually, to reach a frequency of 20 cells per 10⁶ total B cells, or in equal-proportion combinations (mix) to reach a combined frequency of 40 per 10⁶. Recipients were i.m. immunized one day later (D0) with BG505 mRNA-LNP, and local dLNs were analyzed at D14 and D28 (Fig. 3F). Mix-transfer recipients had larger GCs than either single-transfer group at D14, but no significant differences were observed at D28 (Fig. 3G–H). The frequency of antigen-specific CD45.2⁺ B cells in GCs increased over time. At D28, CD45.2⁺ B cells were slightly more frequent in GCs in the mix-recipients compared to V3-Int²-recipients, but frequencies were otherwise similar across groups at both time points (Fig. 3G–H). CD45.2⁺ cells from mix-transfer recipients were sorted for 10x scRNAseq; both V3-Int¹ and V3-Int² B cells were recovered at D14 and D28 (Fig. 3I, Fig. S3G). Upon multiplexing to GC CD45.2⁺ percentages per mouse, no significant differences in relative B cell frequencies were observed between the single-transfer and the mix-transfer groups (Fig. 3J, Fig.S3H). While V3-Int¹ HC isolated from mix recipients had higher SHM at D14 than those from individual transfer groups, these differences equilibrated by D28 and no differences were apparent in V3-Int² HC SHM across treatments (Fig. 3K, Fig. S3I). V3-Int¹ HC maintained their preference for IGKV4–91, IGKV4–61, and IGKV4–50 LCs whether developing alone or accompanying V3-Int²; similarly, V3-Int² HC prefers IGKV12–46, IGKV4–50, and IGKV4–61 in both single and mixed transfer scenarios, though the frequencies of these most common pairings varied (Fig. 3L). We observed no differences in class-switching profiles for either V3-Int¹ or V3-Int² B cells between individual and mix-transfer recipients (Fig. S3J). Thus, despite engaging precisely the same epitope on the mRNA-LNP-delivered immunogen with overlapping affinity ranges, V3-Int¹- and V3-Int²-derived B cells do not alter one another’s GC participation or SHM accumulation.

High-affinity B cells filter out lower affinity cells targeting the same epitope

As V3-Int¹ and V3-Int² have variable, and overlapping, affinities to BG505, to clarify the effects of affinity on intra-epitope competition, we generated an additional HC+LC line targeting the CD4bs. We used the sequences of the Ab min12A21⁵³ to generate IgH^{min12A21/min12A21}IgK^{min12A21/min12A21} (referred to as CD4bs-Int² below), which has lower levels of SHM than CD4bs-Int¹. The affinity of CD4bs-Int² to the BG505 trimer (46 nM) is lower than that of CD4bs-Int¹ (3 nM) (Fig. 4A), though both are quite high compared to V3-Int¹ and V3-Int², which ranged between 1–10 μM. (Fig. 3E). An average of 31.2% of CD4bs-Int² PBMCs bound the BG505 probe (Fig. 4B, Fig. S4A), and 10x scRNAseq demonstrated that the BCRs of binders almost exclusively expressed the full length CD4bs-Int² HC and LC sequences (95% HC and 97% LC), similar to CD4bs-Int¹ (Fig. 4C, Fig.1A). CD4bs-Int² development to follicular B cells, as well as survival, proliferation and differentiation in vitro, were comparable to CD4bs-Int¹ (Fig. S4B–D). Following stimulation with BG505 mRNA-LNP, CD4bs-Int² B cells could be recruited to GCs at D7 and maintained, averaging 43% at D7 and 29% at D28 (Fig. 4D). This contrasts sharply with CD4bs-Int¹ B cells, which were nearly undetectable in GCs at D21 (Fig. 1G).

Figure 4. Affinity-dependent inhibition of GC recruitment occurs between B cell lines competing for the CD4bs.

(A) Affinity of CD4bs-Int¹ and CD4bs-Int² human Ab against BG505 trimer measured by SPR dissociation constant. Each dot represents the mean of four technical replicates. Dotted line represents LOD.

(B) Representative FACS plots of BG505-double-positive and BG505-epitope specific KO-negative peripheral B cells in naïve CD4bs-Int² mice. Events were pre-gated on lymphocytes/singlets/CD4⁻CD8⁻F4/80⁻Gr1⁻/B220⁺ B cells. Data is representative of ten total mice from five individual experiments.

(C) Nested pie chart of human CD4bs-Int² HC (shaded pink), murine HC (shaded gray), CD4bs-Int² LC (pink), and murine LC sequences (gray) amplified from single-cell sorted epitope-specific (BG505⁺KO⁻) naive B cells in CD4bs-Int² mice. Total = single cells amplified. Data produced from one representative of three mice.

(D) CD45.2⁺ cells as a percentage of total GC B cells in dLNs of CD4bs-Int² adoptive transfer recipients at D7 and D21 post-immunization by BG505. Values of zero were plotted as UD on the log₁₀ scale. Each dot represents one mouse. Bars are mean ± SD.

(E) Schematic of individual or co-adoptive-transfer and immunization experiments for CD4bs-Int¹ and CD4bs-Int² B cells. Recipient mice were adoptively transferred with either CD4bs-Int¹ or CD4bs-Int² B cells to reach a frequency of 20 in 10⁶ B cells before immunization with 2 μg of BG505 mRNA. Mice were sacrificed at D7 and D21.

(F) Representative FACS plots of B cells obtained from dLNs at D7 and D21 post-immunization by BG505 in individual or mix transfer recipients. Events were pre-gated on lymphocytes/singlets/live/CD4⁻CD8⁻F4/80⁻Gr1⁻/B220⁺ B cells and represent GC, CD45.2 cells in GC.

(G) (upper) GC cells as a percentage of total B cells and (lower) CD45.2⁺ cells as a percentage of total GC B cells at D7 and D21.

(H) The CD4bs-Int² (pink) and CD4bs-Int¹ (red) B cell lineage frequency analysis from 10x scRNAseq of GC CD45.2 B cells sorted from the Mix groups at D21. Pie charts are averages of all mice in a group (Total=sequences amplified, n=mice).

(I) BCR composition of GC CD45.2 B cell sequences as a percentage of total B cells at D21. Percentage was calculated from FACS data for individual groups, and from 10x scRNAseq plus FACS data for Mix groups. Values of zero were plotted as UD on the log₁₀ scale.

Panel G–I present data pooled from 2 experiments of 3–5 mice per condition. Bars are mean ± SD in (G) and (I). Each dot represents one mouse. q-values calculated by nonparametric Kruskal-Wallis test with pairwise post-hoc comparisons adjusted using BYK correction: ***p < 0.001; ns = not significant. See also Figure S4.

After demonstrating lack of inhibitory competition between V3 cells with low, overlapping-affinity ranges, the CD4bs-Int² and CD4bs-Int¹ KIs allow for the exploration of same-epitope B cells with high affinities differing by approximately 15-fold. CD4bs-Int¹ and CD4bs-Int² CD45.2⁺ B cells were adoptively transferred into WT mice either individually (to achieve a frequency of 20 per 10⁶ total B cells) or in equal combination (combined frequency: 40 per 10⁶), and recipients immunized with BG505 mRNA-LNP (Fig. 4E). All groups displayed similar GC sizes (Fig. 4F&G, upper). At D7, CD45.2⁺ B cell frequencies in GCs were comparable across single and mixed transfer recipients. By D21, however, CD4bs-Int² B cells maintained high GC presence only when transferred individually, with significantly reduced frequencies observed when co-transferred with high-affinity CD4bs-Int¹ B cells (Fig. 4F&G, lower). In contrast, the frequency of high-affinity CD4bs-Int¹ B cells in GCs decayed rapidly to nearly undetectable levels at D21 in both single and mixed transfer recipients (Fig. 4F&G, lower). Doubling the antigen dose did not rescue the CD4bs-Int² response (Fig. 4E–G).

In mix-transfer recipients, scRNAseq of CD45.2⁺ GC B cells revealed that the higher affinity CD4bs-Int¹ B cells dominated GCs at both D7 and D21 independent of administered antigen dose, with only ~5–10% of CD45.2⁺ GC B cells belonging to the CD4bs-Int² lineage (Fig. 4H, Fig. S4E&F). As above, we multiplexed the ratio to make the comparison per mouse and found a 1000-fold decrease of CD4bs-Int² cells in GCs after the addition of CD4bs-Int¹; CD4bs-Int¹, in contrast, was unaffected by CD4bs-Int² (Fig. 4I), demonstrating inhibition of CD4bs-Int² recruitment to or retention in GCs in the presence of CD4bs-Int¹. While CD4bs-Int² HC underwent lower rates of mutation in mixture recipients than when transferred alone, the CD4bs-Int¹ mutation rate was not affected by the presence of the low-affinity CD4bs-Int² line (46 nM) (Fig. S4G). No differences in class-switch profiles were observed in either line (Fig. S4H).

To determine whether competition for GC residence was mediated by inherent BCR affinity to the antigen, we generated a third CD4bs-targeting-bnAb-derived mouse line, IgH^N6-I2/N6-I2IgK^{N6-I2/ N6-I2}, referred to as CD4bs-Int³, from another CD4bs targeting Ab: N6-I2⁵², with affinity for the BG505 MD39 antigen (K_D, 5.7 nM) that was only slightly lower than the CD4bs-Int¹ affinity (K_D, 3 nM) (Fig. S5A). Though CD4bs-Int³, like CD4bs-Int¹, was less prevalent than CD4bs-Int² cells when transferred alone (Fig. S5B–D), frequency within CD45.2⁺ GC B cells was in order of descending affinity to BG505 in the triple mix (Fig. S5E). The addition of CD4bs-Int³ to the transfer-mix did not affect the kinetics of CD4bs-Int¹ (Fig. S5F), though greater SHM was observed in CD4bs-Int¹ (3 nM) after the addition of similarly high-affinity CD4bs-Int³ (5.7 nM) cells in the triple mixture (Fig. S5G). Thus, while no inhibition was observed among B cell lineages with overlapping micromolar affinities targeting the V3-glycan supersite, same-epitope B cells with higher, nanomolar-range affinities targeting the CD4bs suppressed their lower affinity counterparts.

Lowering antigen affinity extends GC residence

The absence of CD4bs-Int² in later GCs in hosts also containing higher affinity CD4bs-Int¹ cells (Fig. 4F–I), as well as the limited time in GCs for CD4bs-Int¹ after immunization (Fig. 1F), led us to examine the effects of BCR affinity to the antigen on GC kinetics using protein immunization. We immunized mice transferred individually with CD4bs-Int¹ or CD4bs-Int² with BG505 protein adjuvanted with saponin/MPLA nanoparticles (SMNP)⁵⁹ (Fig. 5A&B). D8 GCs were similar after either transfer (2–3%); GCs in lower affinity CD4bs-Int² recipients peaked at D16 at 6% and returned to 2% at D21, while GCs in recipients of the higher affinity CD4bs-Int¹ remained at 2.5% on D16 before dropping to 0.7% on D21 (Fig. 5C&D). Within the GC, both CD4bs-Int¹ and CD4bs-Int² comprised approximately 6–7% of GC B cells at D8, demonstrating successful recruitment. At D16, however, GC occupancy rates of the lower affinity CD4bs-Int² increased to approximately 12% and then to 16% on D21; in contrast, by D16 the higher affinity CD4bs-Int¹ cells underwent a massive contraction to 0.4%, from which they did not recover on D21 (Fig. 5C&E).

Figure 5. Affinity-related GC kinetics of competing CD4bs-targeting precursor lines.

(A) Schematic of BG505 immunization experiments for CD4bs-Int¹ and CD4bs-Int² B cells. Mice were immunized with BG505 Env glycoprotein trimer protein with SMNP adjuvant.

(B) Affinity of CD4bs-Int¹ and CD4bs-Int² human Ab against BG505 trimer measured by SPR dissociation constant. Each dot represents the mean of 4 technical replicates. Dotted line represents LOD; dashed line proposes a potential affinity selection ceiling in our system in a low nanomolar range. Affinity data reproduced from Fig. 4D for reference.

(C) Representative FACS plots of GC B cells obtained from dLNs after BG505 immunization at D8, D16, and D21. Events were pre-gated on lymphocytes/singlets/live/CD4⁻CD8⁻F4/80⁻Gr1⁻/B220⁺/GC B cells and represent CD45.2 cells in GC.

(D) Kinetics of GCs over time after BG505 immunization.

(E) Kinetics of CD45.2 cells in GCs over time after BG505 immunization. Values of zero were plotted as UD on the log₁₀ scale.

(F) Schematic of DU172 immunization experiments for CD4bs-Int² and CD4bs-Int¹ B cells. Mice were immunized with DU172 Env glycoprotein trimer protein with SMNP adjuvant.

(G) Affinity of CD4bs-Int¹ and CD4bs-Int² human Ab against DU172 trimer measured by SPR dissociation constant. Each dot represents the mean of 4 technical replicates. Dotted line represents LOD, dashed line proposes the potential affinity selection ceiling in a low nanomolar range.

(H) Representative FACS plots of GC B cells obtained from dLNs after DU172 immunization at D8, D16, and D21. Events were pre-gated on lymphocytes/singlets/live/CD4⁻CD8⁻F4/80-Gr1⁻/B220⁺/GC B cells and represent CD45.2 cells in GC.

(I) Kinetics of GC B cells after DU172 immunization.

(J) Kinetics of CD45.2 cells as the percentage of GC B cells after DU172 immunization. Values of zero were plotted as UD on the log₁₀ scale.

Panels show data from one representative experiment of at least two performed; 3–5 mice per condition. Dots represent single mice. Bars are mean ± SD. q-values calculated by nonparametric Kruskal-Wallis test followed by pairwise post-hoc comparisons adjusted using BKY correction: *p < 0.05, **p < 0.01, ns = not significant. See also Figure S6.

To determine the effect of BCR affinity for different antigens on these kinetics, we then deployed another HIV protein trimer adjuvanted with SMNP, DU172 (Fig. S6A–D), which displays lower affinity to both CD4bs-Int¹ (K_D, 13 nM for DU172 vs 3 nM for BG505) and CD4bs-Int² (K_D, 375 nM for DU172 vs 46 nM for BG505) (Fig. 5F&G). Total GC sizes were similar in both lines throughout (~2–5%) (Fig. 5H&I). Strikingly, no significant difference in CD45.2 cell presence in GCs was observed between these lines (Fig. 5H&J), despite the fact that the ratio of CD4bs-Int¹:CD4bs-Int² BCR affinities for DU172, or the “affinity gap,” is 29.4 folds, greater than their 15.6-fold affinity gap for BG505 (Fig. 5B&G). Both CD4bs-Int¹ and CD4bs-Int² were present in GCs until D21 (Fig. 5H&J). Thus, with BG505 immunization, CD4bs-Int¹ and CD4bs-Int² were recruited to GCs at D8 at equivalent rates, but the high-affinity CD4bs-Int¹ (K_D, 3nM for BG505) B cells diminished by D16 and were absent at D21. In contrast, after immunization with the lower affinity antigen DU172, both CD4bs-Int¹ and CD4bs-Int² remained in GCs out to D21. Thus, lowering antigen affinity below the ~3–12 nM ceiling prolongs GC residence.

Antibody-mediated epitope masking suppresses competing B cells in an affinity-dependent manner

Since B cells targeting the same epitope did not directly compete for antigen, while lower affinity CD4bs-specific B cells were inhibited by CD4bs-Int¹ (Fig. 4G–I), we investigated whether epitope masking from secreted antibody after early fate determination could inhibit B cells from the same epitope. First, we established a model with higher affinity passively transferred Abs and lower affinity B cells targeting the same epitope (Fig. 6A): CD45.2^+/+ B cells from CD4bs-Int² donor mice (46 nM affinity to BG505) were adoptively transferred into CD45.1^+/+ host mice through the retro orbital sinus (day −1); approximately 8 hours later (timepoint referred to as day −0.5), CD4bs-Int¹ Abs (3 nM affinity to BG505) were injected into host mice through the tail vein; approximately 16 hours later (D0), recipient mice were immunized with BG505 mRNA. At D10 post-immunization, dLNs were sampled for flow cytometry (Fig. 6A). Although GCs formed and were a larger fraction of B cells at the higher mAb doses (3 μg and 30 μg), all CD4bs-Int¹ mAb doses higher than 0.05 μg diminished the percentage of CD4bs-Int² B cells in GCs substantially relative to a control mAb; 0.3 μg decreased CD4bs-Int² B cell participation from 14% to 3%; and at 3 μg and 30 μg, CD4bs-Int² B cells were almost entirely blocked from GCs (1% for 3 μg, 0.3% for 30 μg) (Fig. 6B–C). Next, we repeated the passive transfer of CD4bs-Int¹ Ab followed by immunization in mice instead adoptively transferred with equivalently high-affinity CD4bs-Int¹ B cells (Fig. 6D). GC size increases were not significant after mAb delivery, and no significant change in the proportion of CD4bs-Int¹ B cells in GCs was observed at 0.3 μg compared to the control group; however, near-total blocking was observed when the Ab dose was increased to 3 μg or 30 μg (Fig. 6E–F). Thus, low concentrations (0.3 μg) of high-affinity CD4bs-Int¹ mAb were sufficient to inhibit lower affinity CD4bs-Int² cells, but higher concentrations (3 μg and 30 μg) also inhibited cells of equivalent affinity for the antigen.

Figure 6. Affinities and concentrations of antibodies competing for the same epitope determine GC response.

(A) Schematic of CD4bs-Int¹ Ab blocking experiments for CD4bs-Int² B cells.

(B) (upper) Representative FACS plots of GC CD4bs-Int² CD45.2 B cells obtained from dLNs after CD4bs-Int¹ Ab pre-injection and BG505 mRNA immunization. Events were pre-gated on lymphocytes/singlets/live/CD4⁻CD8⁻F4/80⁻Gr1⁻/B220⁺ B cells and represent GC in B cells. (lower) Representative FACS plots of CD45.2 B cells. Events were pre-gated on lymphocytes/singlets/live/CD4⁻CD8⁻F4/80⁻Gr1⁻/B220⁺/CD38⁻CD95⁺ B cells and represent CD4bs-Int² CD45.2 cells in GC.

(C) (left) GC B cells as a percentage of total B cells and (right) CD45.2⁺ CD4bs-Int² cells as a percentage of GC B cells at D10.

(D) Schematic of CD4bs-Int¹ Ab blocking experiments for CD4bs-Int¹ B cells.

(E) Representative FACS plots of GC CD4bs-Int¹ CD45.2 B cells obtained from dLNs at D10. Gated as in (B).

(F) (left) GC B cells as a percentage of total B cells and (right) CD45.2⁺ CD4bs-Int¹ B cells as a percentage of GC B cells at D10.

(G) Schematic of CD4bs-Int² Ab blocking experiments for CD4bs-Int² B cells.

(H) Representative FACS plots of GC CD4bs-Int² B cells obtained from dLNs after CD4bs-Int² Ab pre-injection. Gated as in (B).

(I) (left) GC B cells as a percentage of total B cells and (right) CD45.2⁺ CD4bs-Int² B cells as a percentage of total GC B cells at D10. Control group reproduced from C.

(J) Schematic of CD4bs-Int² Ab blocking experiments for CD4bs-Int¹ B cells.

(K) Representative FACS plots of GC CD4bs-Int¹ B cells obtained from dLNs after CD4bs-Int² Ab pre-injection. Gated as in (B).

(L) (left) GC B cells as a percentage of total B cells and (right) CD45.2⁺ CD4bs-Int¹ B cells as a percentage of total GC B cells at D10. Control group reproduced from F.

Thirty μg of irrelevant flu MEDI8852 Ab was pre-injected to each mouse in all Ctrl Ab groups. Figures represent data pooled from two experiments with 3–5 mice per condition. Each dot represents one mouse. Bars are mean + SD. q-values calculated using Kruskal-Wallis test followed by pairwise comparisons with BKY correction: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; ns = not significant.

We then repeated both adoptive transfers with the lower affinity CD4bs-Int² mAb (K_D, 46 nM for BG505) (Fig. 6G&J). As before, high mAb doses increased GC size. While 0.05 μg had no effect, CD4bs-Int² mAb inhibited CD4bs-Int² B cells in GCs beginning at a dose of 0.3 μg and reached near-total inhibition at 3 μg and 30 μg. (Fig. 6H–I). In contrast, while high mAb doses also increased GC size, 0.3 μg of CD4bs-Int² mAb did not produce any significant decrease in CD4bs-Int¹ B cells in GCs. Higher doses of CD4bs-Int² Ab did reduce CD4bs-Int¹ B cell numbers in GCs, however (Fig. 6K–L). Thus, antibody-mediated epitope masking suppresses GC participation in an affinity- and concentration-dependent manner, with feedback sensitivity governed by BCR affinity.

Local PCs provide antibody feedback to GCs

To determine whether emerging PCs and subsequent antibody production could contribute to GC kinetics, we next interrogated plasma cell populations in and around GCs. GC formation occurs 5–7 days post-immunization and short-lived PCs arise early⁶⁰. As the frequency of CD4bs-Int¹ B cells in GCs declined dramatically between D7 and D16 post immunization, we focused on days 6 to 9 to capture this transition. Mice were adoptively transferred with either CD4bs-Int¹ or CD4bs-Int² cells and then immunized with 8 μg BG505 mRNA, after which flow cytometry was used to evaluate both CD45.2 recruitment to GCs and differentiation into plasma-like (CD138⁺) cells in and out of GCs in dLNs (Fig. 7A). At D6 to D8, significantly more CD45.2⁺ cells were recruited to GCs after CD4bs-Int¹ transfer compared to CD4bs-Int²-transfer, but the difference was no longer apparent by D9 (Fig. 7B–C). Within GCs (inGC) at D6, numbers of plasma-like (CD138⁺) cells were similar for CD4bs-Int¹ (25%) and CD4bs-Int² (18%). At D7 and D8, CD138⁺ CD45.2 B cells were significantly higher in CD4bs-Int¹ recipients (D7=24%; D8=15%) than CD4bs-Int² recipients (D7=10%; D8=1%); but by D9, few or no cells remained in GCs in either model (CD4bs-Int¹=3.6%; CD4bs-Int²=0%) (Fig.7B–C). In unimmunized control mice, dLN from mice adoptively transferred with CD4bs-Int¹ or CD4bs-Int² showed no detectable CD138⁺CD45.2 cells in GCs from D6–D9 (Fig. S7B). In immunized mice, for local draining lymph node B cells not in GCs (exGC), the two lines were similar at D6 and highly divergent after that: CD4bs-Int¹ maintained ~40% of plasma-like CD138⁺ cells exGC throughout; CD4bs-Int² CD138⁺ cells exGC decreased from 25% at D6 to 6.5% at D9 (Fig.7B–C). The exGC environment thus contained detectable plasma-like cells from D6 to D9 for either line, but CD4bs-Int¹-derived PCs remained at far higher percentages. This group of plasma-like CD138⁺ cells were also observed at D8 inGC (9%) and exGC (30%) in CD4bs-Int¹ recipient mice but not CD4bs-Int² recipients after lower affinity DU172 protein immunization; the population decreased below 1% between D8 and D16 (Fig. S6E&F). The lower numbers of early CD138⁺ cells in dLN, especially inGC, could be related to longer GC residence in lower affinity DU172 immunization.

Figure 7. PCs in and adjacent to GCs produce a local Ab pool that adjusts the GC response.

(A) Schematic of plasma cell experiments. Recipient mice were adoptively transferred with either CD4bs-Int¹ or CD4bs-Int² B cells to reach a frequency of 50 in 10⁶ B cells before immunization with 8 μg of BG505 mRNA. Mice were sacrificed at D6, D7, D8, and D9.

(B) (upper) Representative FACS plots of CD45.2 B cells. Events were pre-gated on lymphocytes/singlets/live/CD4⁻CD8⁻F4/80⁻Gr1⁻/B220⁺/CD38⁻CD95⁺ B cells and represent CD45.2 B cells in GCs. (lower) Representative FACS plots of plasma like cells. Events were pre-gated on lymphocytes/singlets/live/CD4⁻CD8⁻F4/80⁻Gr1⁻/B220⁺/CD38⁻CD95⁺/CD45.1⁻CD45.2⁺ B cells and represent plasma-like cells among GC CD45.2 B cells.

(C) (upper) CD45.2⁺ B cells as a percentage of GC B cells, (middle) CD138⁺ cells as a percentage of GC CD45.2 B cells, and (bottom) CD138⁺ cells as a percentage of exGC CD45.2 B cells at D6, D7, D8, D9. Dots are individual mice, bars are mean + SD. q-values calculated by 2-way ANOVA with pairwise post-hoc comparisons adjusted using BYK correction.

(D) Representative immunofluorescent staining of dLN isolated from mice transferred with CD4bs-Int¹ B cells at D6 post-immunization by BG505, staining with B220 (blue, surface), CD45.2 (green, surface), and IRF4 (red, nucleus). Large tile presents overview of architecture of the whole LN, inset boxes present individual GCs and subsets of PCs in GC. Fo, Follicle; MC, Medullar cords; TZ, T cell zone. Scale bar: 200 μm.

(E) Slide-seq spatial map of dLN isolated from mice with CD4bs-Int¹ B cells D6 post-immunization by BG505. Mapping of dLN is (upper) colored by cell type annotations from gene expression profiles, or (lower) colored by anatomical location inGC or exGC. Plasma cells (PC); GC B cells (GCBC); naïve B cells (NBC); memory B cells (MBC); cytotoxic T cells (Cytotoxic); epithelial cell (Epithelial); CD4 T cells (CD4); follicular dendritic cells (FDC); follicular regulatory T cells (Tfr); follicular helper T cells (Tfh); plasmacytoid dendritic cells (PDC). n = subset cells. Non-serial sections of the same dLN were used as in (D). Scale bar: 500 μm.

(F) Violin plots of expression level of genes related to Ab secretion in GC B cells and PCs in and out of GC by anatomical location in the dLN at D6. Pairwise group comparisons were performed using two-sided Mann–Whitney U tests, followed by Holm–Bonferroni correction.

(G) ELISA quantification of BG505-binding (circle) and BG505-CD4bs-KO-binding (diamond) IgG from 100 μl of dLN homogenates of CD4bs-In¹(red) or CD4bs-Int² (pink) recipient mice immunized with BG505 mRNA or left unimmunized. A 5x dilution multiplier was applied to obtain the final estimated (est) endpoint titer of dLNs. Dots represent mean values of technical triplicate from homogenates generated from one popliteal LN from each of the four mice pooled at D7, D8, and D9. Unimmunized dots include all data collected in groups left unimmunized after corresponding adoptive transfers. Dotted lines represent LOD.

(H) Schematic of single-side immunization experiment for CD4bs-Int¹ and CD4bs-Int² B cells. Recipient mice were adoptively transferred with CD4bs-Int¹ or CD4bs-Int² B cells respectively to reach a frequency of 50 in 10⁶ B cells before immunization with 8 μg of BG505 mRNA. Mice were sacrificed at D8 for ipsilateral dLNs and contralateral LNs as controls.

(I) ELISA quantification of BG505-binding IgG from 100 μl of dLN homogenates ipsilateral dLNs (circle) and contralateral LNs (diamond) of CD4bs-In¹(red) or CD4bs-Int² (pink) recipient mice immunized with BG505 mRNA or left unimmunized. A 5x dilution multiplier was applied to obtain the final est endpoint titer of dLNs. For each group, homogenates were generated from one popliteal LN from each of the four mice pooled into 100 μl of buffer. Technical triplicates were performed for each group. Dots represent individual values pooled from two experiments. Bars are mean ± SD. Dotted lines represent LOD.

Significance is indicated as: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; ns = not significant. See also Figure S7.

The observation of CD138⁺ cells both inGC and exGC by flow cytometry pushed us to detail the spatial distribution of plasma-like cells in the dLN. To determine localization, we first deployed immunofluorescence (IF) staining in immunized mice adoptively transferred with either CD4bs-Int¹ (Fig. 7D) or CD4bs-Int² (Fig. S7C); dLN from mice adoptively transferred with CD4bs-Int² but left unimmunized were also examined as a control (Fig. S7D). Nucleus staining for IRF4 was utilized for better visualization of plasma-like cell colocalization. Multiple GCs formed by D6 after BG505 immunization in the B cell follicles at the cortex of the dLN in adoptively transferred mice (Fig. 7D & S7C). Some GCs were comprised almost entirely of CD45.2 cells. Of those CD45.2 cells within GCs, some were also positive for IRF4 nucleus staining, indicating that they were plasma-like; CD45.2 plasma-like (IRF4⁺) cells were primarily observed in small groups at GC borders (Fig. 7D, Fig. S7C). Using Slide-seq^61,62, we then developed a whole-distribution map of plasma-like cells in dLN. Plasma-like cells, assigned by Robust Cell Type Decomposition (RCTD)⁶³ using an immune scRNA-seq reference dataset⁶⁴, were observed both within and around GCs, accumulating in the medullary cords, paracortex, or T-B border (Fig. 7D–E; Fig. S7C, S7E–F). Quantification of sufficiently sized GCs (>=100 beads) showed an average of 1.5 PCs per 100 GC cells in the inner GC core, 6.6 per 100 in the outer GC ring, and 20.3 per 100 in the 30 μm periphery of the GC (Fig. S7G–H). Gene expression analysis found upregulation of Ab secretion-related genes, including secretion components (Jchain, Ighg1, Ighg2b, and Ighg2c), as well as protein folding machinery (Calr, Pdia4, and Hspa5), in PCs within the anatomical range of the GC relative to non-PC GC B cells; expression by inGC PCs reached similar or slightly lower levels compared to PCs exterior to the GC (Fig. 7F, Fig. S7I), suggesting potential Ab secretion functions in those cells.

To determine whether these plasma cell populations in and around the GC produced Ab, dLNs from different time points were mechanically disrupted in buffer and ELISA against BG505 was performed on the resulting supernatant. Based on prior measurements of murine LN volumes^65–70, we estimated that a minimum of a 5x dilution occurs at the mechanical disruption stage; we therefore applied a 5x correction to estimate endpoint titer in the dLNs. Compared to unimmunized mice adoptively transferred with the same CD45.2 cells, high titers of IgG were detected in dLNs at D6 to D9 after immunization. For CD4bs-Int², the estimated endpoint titer for dLN IgG increased from 3×10⁴ at D6, to 1×10⁵ at D7, to a peak of 2×10⁵ at D8, and then dropped to 1×10⁴ at D9. High-affinity CD4bs-Int¹ presented a higher and more stable dLN IgG titer curve above 8×10⁴ from D6—9, with a slightly earlier peak of 2×10⁵ at D7. At D7, a small BG505-CD4bs-KO probe peak was observed for CD4bs-Int² (at 5×10³) and CD4bs-Int¹ (8×10²), potentially indicating a small amount of IgG produced by host mouse B cells targeting other epitopes (Fig. 7G). Compared to IgG, IgM titers peaked at a lower level (5×10³) and early (by D6) in dLNs of both types of transfer recipients, with low-affinity CD4bs-Int² persisting until D9 while, in contrast, high-affinity CD4bs-Int¹ dropped at D7. No IgM were detected above the detection limit of 250 for either the CD4bs-Int¹- or CD4bs-Int²-transferred unimmunized group, or by BG505-CD4bs-KO probes (Fig. S7J). The observation that neither IgG nor IgM were detected in the dLN of unimmunized adoptive transfer recipients excludes the possibility of a base from self-reactivity (Fig. 7G, Fig. S7J). Local dLN titers may be higher than those in circulation at early timepoints (Fig. S7K). To more precisely establish the localization of Ab production to the dLN, we performed single-side immunization on recipient mice and compared LN homogenate from ipsilateral dLNs and the contralateral LNs (Fig. 7H, Fig. 7I, Fig. S7L). At D8, the ipsilateral dLNs showed markedly higher average IgG titers in both CD4bs-Int² recipient (5.1 × 10⁴) and CD4bs-Int¹ recipient (6.4 × 10⁴), while the contralateral LNs showed much lower IgG titers just above LOD (6 × 10² for CD4bs-Int²; 1 × 10³ for CD4bs-Int¹) (Fig. 7I, Fig. S7L). This result suggests that early Ab feedback could be more local than systematic.

To determine the affinities of the early antibody pool, mutated BCR sequences from D7 GC B cells were expressed as IgG mAbs in vitro and tested for affinity against the BG505 trimer. In D7 GC, mAbs isolated from CD4bs-Int¹-transfer recipients had relatively high affinities, with K_D values ranging from 3 nM to 19 nM and a median K_D of 5 nM, while mAbs from CD4bs-Int² recipients had relatively lower affinities, with K_D values ranging from from 34 nM to 2.5 μM and a median K_D of 85 nM (Fig. S7M). This indicated a maintenance of an at least 10-fold overall affinity gap between the Ab reservoirs generated by these two lineages. In sum, high early plasma-cell and antibody abundance in dLNs demonstrate the existence of a local antibody feedback loop that rapidly tunes GC competition.

DISCUSSION

The classical model of GC selection is Darwinian, with higher affinity B cells progressively outcompeting lower affinity clones for antigen and survival signals⁷¹. Our data reveal an additional layer of regulation in which antibody feedback shapes GC dynamics. Using mouse models with BCRs of defined affinities and epitope specificities to the same immunogen, we found substantial initial recruitment of both high- and low-affinity B cells, but only lower affinity cells persisted in GCs. Persistence was not a fixed characteristic of particular B cell lines but depended on affinity for the presented immunogen. Importantly, B cells targeting the same epitope with similar affinities could be costimulated without changing their individual GC residency patterns, whereas high-affinity clones suppressed lower affinity counterparts. Spatial transcriptomics revealed the presence of PCs near GCs as a probable source of local secreted antibody and feedback mediating this suppression. Together, these findings suggest that antibody feedback may provide a critical regulatory “brake,” enforcing affinity thresholds that shape B cell selection and exit from GCs, while also promoting epitope spreading as lower affinity clones to other epitopes are not suppressed. This dual mechanism suggests vaccine strategies aiming to balance breadth and potency may need to modulate affinity and feedback dynamics.

B cells must meet a minimum affinity-to-antigen “floor” to enter GCs for affinity maturation^{3,5,11,12,72–74}. As affinity increases, selective pressure on antigen–BCR binding may decrease^3,18,19; though high-affinity variants may still emerge, an affinity ceiling exists above which comparative selective advantage is lost. In the classic hapten-based system, B cells with higher initial BCR affinities to antigen accumulate fewer V_H mutations than lower affinity counterparts despite similar mutation rates, due to either relaxed positive selection, or to negative selection as mutations in high-affinity BCRs may be likelier to diminish affinity¹³. In our real-world HIV-trimer-based system, similarly, only high-affinity competitors but not low-affinity competitors increased SHM rates in counterpart high-affinity cells, potentially due to positive selection. Antibodies produced by those first activated B cells can modulate this selective process by directly competing for antigen or inducing apoptosis in lower affinity B cells²⁵; high-affinity antibodies may be induced quite early, without affinity maturation¹⁹. Recent studies have highlighted the blocking effect injected high-affinity Abs can exert on immune responses^27,28. In agreement with these studies, our findings suggest an Ab-driven self-modulation loop acts as a “brake”, reinforcing the upper affinity maturation ceiling in GCs. Notably, this brake is an intra-epitope phenomenon reproducible by passive antibody administration, suggesting that epitope masking by early antibody production may drive this system. The brake’s epitope-specificity is congruent with recent findings in an influenza infection model, where affinities from interclonal PC differed by factors of a thousand or more, while intraclonal PCs only differed by factors of 10 to 30¹⁷. An intraclonal ceiling effect produced by Ab braking may drive GC B cell diversity and subdominant responses observed by multiple studies^{28–30,75–77}. Interestingly, feedback from serum Abs has been observed not only hindering but also enhancing responses^27–29,78. The increase in total GC sizes after high doses of Ab in our passive transfer experiments is suggestive of enhancement, and is notable in light of recent work suggesting that Abs can essentially adjuvant mRNA-LNP immunization²⁸. In contrast to the “brake”, Ab-driven self-modulation may also serve as the intra-epitope “filter” to exclude low-affinity lines from GC entry or residence—though avidity effects may rescue some low-affinity B cell GC participation⁷⁹. Thus, initial BCR affinity to antigen and the timing of high-affinity Ab secretion may establish a self-modulating loop in which the “brake” drives diversity while the “filter” drives potency.

Much of our understanding of Ab feedback relates to circulating Ab titers and affinities^28,80,76,27. In contrast, our observations of early Ab titer in dLN and PC in and around the GC suggests that a local, real-time PC-driven feedback loop determines B cell composition in primary GCs. Plasma fate commitment occurs in the light zone (LZ) with the upregulation of Irf4^16,81–83. Immunofluorescent staining using single markers, such as IRF4, BLIMP1, CD138, or cytosol IgG1, had placed PC precursors predominantly in the T-B border, with rare subsets in the GC^81,82,84,85. Gene expression of plasma-like cells (IRF4⁺) defined as GC-resident on the basis of CD38^-Fas⁺ are similar to that of total plasma-like population (CD138⁺TACI⁻)⁸⁶. Our Slide-seq data confirms the anatomical localization of inGC PCs designated by a comprehensive gene signature. Similar levels of upregulation of secretion components and protein folding machinery in inGC and exGC PCs implies inGC PC Ab secretion; recent observations of unusually low extracellular protease levels inside B cell follicles in mouse LN⁸⁷ further support the possibility of a local Ab pool. Like the GC response itself, post-GC PC proliferation is affinity-dependent⁸⁸, and our own observations suggest more inGC PCs from higher affinity lines were present early. Thus, alongside precursor frequency, Ag affinity, and Ag avidity⁸⁹, localized PC populations may determine GC kinetics by producing a local high-affinity secreted Ab pool.

Affinity- and kinetics-driven competition and Ab feedback have substantial implications for the vaccine-mediated development of breadth, particularly in contexts where the maturation of multiple B cell lineages are driven by the same immunogens to achieve broad neutralization, such as HIV sequential immunization^{32,39,41,90–92}. While progress has been made in simultaneously priming diverse bnAb precursors^93,94, little is known about how crosstalk between minimally matured antibodies and B cells will impact late-stage boosting strategies. The lack of inter-epitope B cell competition in our assays indicates that the brake on intra-epitope B cell development may be essential to a multi-epitope response. Where long GC reactions are preferable, the use of lower affinity immunogens may release the self-braking mechanism. Furthermore, antigen affinity to the broader Ab and B cell repertoire must be considered. Though not ascribed to antibody production, the presence of higher affinity non-precursor same-epitope responses has been observed as a limiting factor in B cell GC residence¹². BnAb development strategies involving sequential immunization should identify functional affinity ranges for each immunization stage. A related concept, that the “affinity drop” between sequential antigens should not be too high or too low, was previously described, though also not ascribed to antibody production³⁹.

In sum, our study elucidated a multi-faceted Ab feedback loop in B cell competition for the same immunogen, which served as a self-modulating affinity-dependent “brake” for development of B cell lineages to the same epitope, as well as a “filter” excluding lower affinity B cells, while preserving a parallel evolutionary path for lineages to other epitopes. The use of vaccine models applicable to pre-clinical development provides direct applications to fine-tuning vaccine strategies to avoid inhibitory antibody feedback.

Limitations of the Study

Our findings demonstrate that inherent affinity affects the durability and magnitude of GC B cell responses. While we found antibody feedback compelling based on multiple lines of evidence, more direct tests using antibody-production-deficit B cells, such as Blimp-1 knock-outs, remain for future investigations. Additionally, Slide-seq, while a powerful approach, has a lower capture rate than traditional RNA sequencing. As to our model system, where cross-epitope comparisons are made, we should note that the bnAbs used in our BCR transgenic mice differ in binding stoichiometry (monovalent vs trivalent), binding modality (HCDR3 vs V_H-gene dominant), approach geometry, and epitope composition. Binding stoichiometry and approach geometry may influence BCR crosslinking on antigen-decorated surfaces, while binding modality and epitope composition could affect BCR clustering. In particular, signaling thresholds may differ for BCRs recognizing predominantly proteinaceous epitopes to those engaging carbohydrate-rich epitopes, as glycan-mediated interactions may exhibit distinct binding kinetics or activation dampening via CD22 interactions⁹⁵. Finally, the membrane-anchored presentation of the immunogen used here may be subject to quite different modulating forces than a soluble immunogen^57,58,96, though no obvious differences in individual kinetics were observed in our soluble protein experiments. Overall, however, consistent affinity-dependent effects were observed both when varying BCR affinity through the generation of transgenic mouse lines expressing sequences from distinct bnAb precursors and when varying antigen affinity through different immunogens.

RESOURCE AVAILABILITY

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Facundo D. Batista (fbatista1@mgh.harvard.edu), except where noted otherwise below.

Materials availability

Model animals and antibodies minBG18.6–1, minBG18.6–2, minBG18.6–3, minBG18.6–4, minBG18.11–1, minBG18.11–2, minBG18.11–3, minBG18.11–4, N6-I3-D7–1, N6-I3-D7–2, N6-I3-D7–3, N6-I3-D7–4, N6-I3-D7–5, N6-I3-D7–6, N6-I3-D7–7, N6-I3-D7–8, N6-I3-D7–9, Min12A21-D7–1, Min12A21-D7–2, Min12A21-D7–3, Min12A21-D7–4, Min12A21-D7–5, Min12A21-D7–6, Min12A21-D7–7, Min12A21-D7–8, Min12A21-D7–9 are available from corresponding author (FDB) on request, under a standard material transfer agreement (MTA) with Massachusetts General Hospital. Plasmids or recombinant proteins for immunogens and sort reagents related to BG505 MD39.3 and DU172 MD39.2; or antibodies 12A21, min12A21, N6-I2, N6-I3, N6, PCT64–18D; or SPR reagents in this study, are available from WRS under a material transfer agreement with the Scripps Research Institute. mRNA-LNP vaccine construct for BG505 MD39.3 can be made available from SH if the recipient and Moderna are able to agree upon the terms of an MTA.

Data and code availability

Slide-seq data has been deposited on Broad Institute Single cell portal, and the repository URL is listed in the Key Resources Table (https://singlecell.broadinstitute.org/single_cell/). Custom code used for analysis has been deposited on GitHub, and the repository URL is listed in the Key Resources Table (https://github.com/). BCR sequences have been deposited to Zenodo and the identifier DOI is listed in the Key Resource Table (https://zenodo.org/).

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
PE/Cy7 anti-mouse IgD Antibody	Biolegend	Cat#: 405720
APC/Cy7 anti-mouse/human CD45R/B220 Antibody	Biolegend	Cat#: 103224
BV711 Rat Anti-Mouse Ig, λ1, λ2 & λ3 Light Chain	BD Biosciences	Cat#: 744527
BUV395 Rat Anti-Mouse Ig, K light chain	BD Biosciences	Cat#: 742839
BV421 Rat Anti-Mouse IgM	BD Biosciences	Cat#: 743323
PE anti-mouse CD45.2 Antibody	Biolegend	Cat#: 109808
BV786 Rat Anti-Mouse IgD	BD Biosciences	Cat#: 563618
BV421 Rat Anti-Mouse IgG1	BD Biosciences	Cat#: 562580
BUV395 Rat Anti-Mouse IgM	BD Biosciences	Cat#: 743329
Alexa Fluor® 594 anti-mouse IgD Antibody	Biolegend	Cat#: 405740
BV786 Rat Anti-Mouse IgM	BD Biosciences	Cat#: 564028
BV785 anti-mouse CD138 (Syndecan-1) Antibody	Biolegend	Cat#: 142534
anti-mouse CD38, BV510	BD Biosciences	Cat#: 740129
PE anti-mouse CD138 (Syndecan-1) Antibody	Biolegend	Cat#: 142503
Alexa Fluor® 700 anti-mouse IgD Antibody	Biolegend	Cat#: 405729
anti-mouse CD38, BV510	BD Biosciences	Cat#: 740129
PE Hamster Anti-Mouse CD95	BD Biosciences	Cat#: 561985
PE/Cyanine7 anti-mouse CD138 (Syndecan-1) Antibody	Biolegend	Cat#: 142513
BUV395 Rat Anti-Mouse IgG2b	BD Biosciences	Cat#: 743180
Alexa Fluor® 594 AffiniPure® Goat Anti-Mouse IgG2c	Jackson ImmunoResearch	Cat#: 115-587-188
CD4 Monoclonal Antibody (GK1.5), APC-eFluor 780	Thermo Fisher Scientific	Cat#: 47-0041-82
CD8a Monoclonal Antibody (53-6.7), APC-eFluor 780	Thermo Fisher Scientific	Cat#: 47-0081-82
F4/80 Monoclonal Antibody (BM8), APC-eFluor 780	Thermo Fisher Scientific	Cat#: 47-4801-82
Ly-6G Monoclonal Antibody (1A8-Ly6g), APC-eFluor 780	Thermo Fisher Scientific	Cat#: 47-9668-82
Alexa Fluor® 700 anti-mouse CD4 Antibody	Biolegend	Cat#: 100429
CD8a Monoclonal Antibody (53-6.7), Alexa Fluor™ 700	Thermo Fisher Scientific	Cat#: 56-0081-82
Alexa Fluor® 700 anti-mouse F4/80 Antibody	Biolegend	Cat#: 123129
Alexa Fluor® 700 anti-mouse Ly-6G Antibody	Biolegend	Cat#: 127621
APC/Cyanine7 anti-mouse CD23 Antibody	Biolegend	Cat#: 101629
BUV395 Rat Anti-Mouse CD21/CD35	BD Biosciences	Cat#: 740249
PE/Cyanine7 anti-mouse CD24 Antibody	Biolegend	Cat#: 101821
PE anti-mouse CD43 Antibody	Biolegend	Cat#: 143205
PE/Cy7 anti-mouse IgD Antibody	BD Biosciences	Cat#: 743328
BV421 anti-mouse IgD Antibody	BD Biosciences	Cat#: 405725
PerCP/Cyanine5.5 anti-mouse CD93 (AA4.1, early B lineage) Antibody	Biolegend	Cat#: 136511
BUV395 Rat Anti-Mouse CD45R/B220	BD Biosciences	Cat#: 563793
BUV805 Rat Anti-CD11b	BD Biosciences	Cat#: 568345
BV605 Rat Anti-Mouse CD5	BD Biosciences	Cat#: 563194
BV650 Rat Anti-Mouse CD23	BD Biosciences	Cat#: 740456
PE anti-mouse CD23 Antibody	Biolegend	Cat#: 101608
PE/Cyanine7 anti-mouse IgM Antibody	Biolegend	Cat#: 406514
PE/Cyanine7 anti-mouse CD19 Antibody	Biolegend	Cat#: 152418
PerCP/Cyanine5.5 anti-mouse/human CD45R/B220 Antibody	Biolegend	Cat#: 103236
Alexa Fluor® 488 anti-mouse CD45.2 Antibody	Biolegend	Cat#: 109816
Alexa Fluor® 594 anti-mouse/human CD45R/B220 Antibody	Biolegend	Cat#: 103254
Alexa Fluor® 647 anti-IRF4 Antibody	Biolegend	Cat#: 646408
Rat monoclonal anti-mouse/human IRF4 Alexa Fluor 647 (clone: 3E4)	Biolegend	Cat #646408, RRID: AB_2564048
Rat monoclonal anti-mouse/human CD45R/B220 Alexa Fluor 594 (clone: RA3-6B2)	Biolegend	Cat #103254, RRID: AB_2563229
Mouse monoclonal anti-mouse CD45.2 Alexa-Fluor 488 (clone: 104)	Biolegend	Cat #109815, RRID: AB_492869
MEDI8852 Ab	Kallewaard et al., 2016
19b Ab	Moore et al., 1995	N/A
F105 Ab	Burton et al., 1991	N/A
B6 Ab	Pantophlet et al., 2003	N/A
PGT145 Ab	Walker et al., 2011	N/A
PGT151 Ab	Falkowska et al., 2014	N/A
PGT121 Ab	Walker et al., 2011	RRID: AB_2491041
N6 Ab	Huang et al., 2016	N/A
12A21 Ab	Scheid et al., 2011	N/A
Min12A21 Ab	Jardine et al., 2016	N/A
N6-I2 Ab	Huang et al., 2016	N/A
N6-I3 Ab	Huang et al., 2016	N/A
PCT64-18D Ab	Landais et al., 2017	N/A
minBG18.6-1 Ab	This paper.	N/A
minBG18.6-2 Ab	This paper.	N/A
minBG18.6-3 Ab	This paper.	N/A
minBG18.6-4 Ab	This paper.	N/A
minBG18.11-1 Ab	This paper.	N/A
minBG18.11-2 Ab	This paper.	N/A
minBG18.11-3 Ab	This paper.	N/A
minBG18.11-4 Ab	This paper.	N/A
N6-I3-D7-1 Ab	This paper.	N/A
N6-I3-D7-2 Ab	This paper.	N/A
N6-I3-D7-3 Ab	This paper.	N/A
N6-I3-D7-4 Ab	This paper.	N/A
N6-I3-D7-5 Ab	This paper.	N/A
N6-I3-D7-6 Ab	This paper.	N/A
N6-I3-D7-7 Ab	This paper.	N/A
N6-I3-D7-8 Ab	This paper.	N/A
N6-I3-D7-9 Ab	This paper.	N/A
Min12A21-D7-1 Ab	This paper.	N/A
Min12A21-D7-2 Ab	This paper.	N/A
Min12A21-D7-3 Ab	This paper.	N/A
Min12A21-D7-4 Ab	This paper.	N/A
Min12A21-D7-5 Ab	This paper.	N/A
Min12A21-D7-6 Ab	This paper.	N/A
Min12A21-D7-7 Ab	This paper.	N/A
Min12A21-D7-8 Ab	This paper.	N/A
Min12A21-D7-9 Ab	This paper.	N/A
Bacterial and virus strains
DH5α Competent Cells	Thermo Fisher Scientific	Cat # GACC-96
Chemicals, Peptides, and Recombinant Proteins
Saponin/MPLA nanoparticles (SMNP)	Silva et al., 2021	N/A
Invitrogen™ Molecular Probes™ DAPI (4’,6-Diamidino-2-Phenylindole, Dihydrochloride)	Thermo Fisher Scientific	Cat#: D1306
Alexa Fluor 488 Streptavidin	Biolegend	Cat#: 405235
Alexa Fluor 647 Streptavidin	Biolegend	Cat#: 405237
Alexa Fluor 594 Streptavidin	Biolegend	Cat#: 405240
Superscript™ III Reverse Transcriptase	Thermo Fisher	Cat#: 18080085
HotStarTaq DNA Polymerase	Qiagen	Cat#: 203205
cOmplete™, EDTA-free Protease Inhibitor Cocktail	MilliporeSigma	Cat#: 12352204
RNasin® Ribonuclease Inhibitors (Recombinant)	Promega	Cat#: N2515
CountBright™ Absolute Counting Beads, for flow cytometry	Thermo Fisher Scientific	Cat#: C36950
SIGMAFAST™ p-Nitrophenyl phosphate Tablets	MilliporeSigma	Cat#: N2770-50SET
NP40	MilliporeSigma	Cat#: 492016-100ML
UltraComp eBeads™ Compensation Beads	Thermo Fisher Scientific	Cat#: 01-2222-42
Recombinant Mouse IL-4 Protein	R and D systems	Cat#: 404-ML-025
Recombinant Mouse IL-5 Protein	R and D systems	Cat#: 405-ML-025
Recombinant Mouse CD40 Ligand/TNFSF5 (HA-tag) Protein	R and D systems	Cat#: 8230-CL-050
RPMI 1640 Medium	Thermo Fisher Scientific	Cat#: 11875119
Fetal Bovine Serum	MilliporeSigma	Cat#: F4135-500ML
HEPES	Thermo Fisher Scientific	Cat#: 15630080
GlutaMAX™ Supplement	Thermo Fisher Scientific	Cat#: 35050079
MEM Non-Essential Amino Acids Solution	Thermo Fisher Scientific	Cat#: 11140076
Penicillin-Streptomycin	Thermo Fisher Scientific	Cat#: 15140122
β-mercaptoethanol	MilliporeSigma	Cat#: M6250-100ML
DU172-17 MD39.2 SOSIP	This paper.	N/A
DU172-17 MD39 SOSIP CD4bs-KO (His-Avi-tagged)	This paper.	N/A
DU172-17 MD39 SOSIP (His-Avi-tagged)	This paper.	N/A
DU172-17 MD39 SOSIP (His-tagged)	This paper.	N/A
BG505 MD39.3 SOSIP	Ramezani-Rad et al., 2025	N/A
BG505 MD39.3 gp151 mRNA	Ramezani-Rad et al., 2025	N/A
BG505 MD39.3 SOSIP (His-Avi-tagged)	Ramezani-Rad et al., 2025	N/A
BG505 MD39.3 SOSIP CD4bs-KO (His-Avi-tagged)	Ramezani-Rad et al., 2025	N/A
BG505 MD39.3 SOSIP V2-KO (His-Avi-tagged)	This paper.	N/A
BG505 MD39.3 SOSIP V3-KO (His-Avi-tagged)	This paper.	N/A
BG505 MD39.3 SOSIP (His-tagged)	Ramezani-Rad et al., 2025	N/A
BG505 MD39.3 SOSIP CD4bs-KO (His-tagged)	Ramezani-Rad et al., 2025	N/A
Saponin/MPLA nanoparticles (SMNP)	Silva et al., 2021	N/A
Critical commercial assays
LIVE/DEAD™ Fixable Blue Dead Cell Stain Kit, for UV excitation	Thermo Fisher Scientific	Cat#: L34962
LIVE/DEAD™ Fixable Violet Dead Cell Stain Kit, for 405 nm excitation	Thermo Fisher Scientific	Cat#: L34964
CellTrace™ CFSE Cell Proliferation Kit, for flow cytometry	Thermo Fisher Scientific	Cat#: C34554
Pan B Cell Isolation Kit II, mouse	Miltenyi Biotec	Cat#: 130-104-443
Chromium Next GEM Single Cell 5′ Kit v2	10x Genomics	PN-1000263
Library Construction Kit	10x Genomics	PN-1000190
Chromium Single Cell Mouse BCR Amplification Kit	10x Genomics	PN-1000255
Chromium Next GEM Chip K Single Cell Kit	10x Genomics	PN-1000286
Dual Index Kit TT Set A	10x Genomics	PN-1000215
Dual Index Kit TN Set A	10x Genomics	PN-1000250
Deposited data
Raw and analyzed data	This paper	https://singlecell.broadinstitute.org/single_cell/study/SCP3331
BCR sequences	This paper	10.5281/zenodo.17603995
Cell lines
HEK293F	Thermo Fisher	Cat# R790-07; RRID: CVCL_6642
Experimental Models: Organisms/Strains
Mouse: CD4bs-Int1 BCR mouse model	This paper	N/A
Mouse: CD4bs-Int2 BCR mouse model	This paper	N/A
Mouse: CD4bs-Int3 BCR mouse model	This paper	N/A
Mouse: V2-Int1 BCR mouse model	This paper	N/A
Mouse: V3-Int1 BCR mouse model	This paper	N/A
Mouse: V3-Int2 BCR mouse model	This paper	N/A
Mouse: B6.SJL-Ptprcapepcb/BoyJ	The Jackson Laboratory	JAX: 002014
Mouse: C57BL/6J	The Jackson Laboratory	JAX: 000664
Recombinant DNA
pHL-sec		Addgene #99845
pCW-sec		N/A
pCW-CHIg-hG1		N/A
pCW-CLIg-hk		N/A
Software and Algorithms
BioRender	BioRender.com	https://biorender.com/
Byos™ (Version 5)	Protein Metrics Inc.	https://www.proteinmetrics.com/products/byonic/
C-SIDE	Cable et al., 2022	https://doi.org/10.1101/2021.12.26.474183
Carterra Software	Carterra Inc.	N/A
Flowjo X	Treestar	https://www.flowjo.com/
Geneious Prime	Biomatters	https://www.geneious.com/
Fiji (ImageJ)	Schindelin et al., 2012	https://fiji.sc/
Illustrator	Adobe	N/A
IMGT/V-quest	IMGT®, the international ImMunoGeneTics information system® (Université de Montpellier, CNRS, France)	http://www.imgt.org/IMGTindex/V-QUEST.php/
Microsoft Office	Microsoft	https://www.office.com/
Orbitrap Fusion Tune application v3.1	Thermo Fisher Scientific	N/A
Prism 8	GraphPad	https://www.graphpad.com/
Python version 3.9.19	The Python Software Foundation (PSF)	https://www.python.org/; RRID: SCR_008394
R version 4.5.0	R Foundation for Statistical Computing	https://www.r-project.org; RRID: SCR_001905
Robust decomposition of cell type mixtures	Cable et al., 2022a	https://doi.org/10.1038/s41587-021-00830-w
Rstudio version 2025.05.0+496	Rstudio, Inc. (2019)	https://posit.co; RRID: SCR_000432
Scanpy	Wolf et al., 2018	https://github.com/theislab/scanpy; RRID: SCR_018139
TissueFAXS SL 7.1.135 Confocal	TissueGnostics
TissueFAXS SL Viewer 7.1.6245.135	TissueGnostics
UCSF ChimeraX v1.10.1	Goddard et al., 2021	N/A
XCalibur Version v4.2	Thermo Fisher Scientific	N/A
Custom analysis scripts	This paper	https://github.com/immunoliugy/affinity_brake
Other
Amicon® Ultra Centrifugal Filter, 100 kDa MWCO (15 mL)	Millipore Sigma	Cat# UFC9100
Amicon® Ultra Centrifugal Filter, 100 kDa MWCO (4 mL)	Millipore Sigma	Cat# UFC8100
Amicon® Ultra Centrifugal Filter, 30 kDa MWCO (15 mL)	Millipore Sigma	Cat# UFC9030
Amicon® Ultra Centrifugal Filter, 30 kDa MWCO (4 mL)	Millipore Sigma	Cat# UFC8030
Dawn HELEOS II	Wyatt	N/A
EasySpray PepMap RSLC C18 column (75 μm × 75 cm)	Thermo Fisher Scientific	Cat# ES805
Endosafe nexgen-PTS Instrument	Charles River	N/A
HisTrap HP column (5 mL)	Cytiva	Cat# 17524801
Microdialysis plate 48-wells 1 mL	Thermo Fisher Scientific	Cat# A50466
NanoDrop 2000c Spectrophotometer	Thermo Fisher Scientific	ND-2000
Oasis MCX 96-well μElution Plate	Waters	186001830BA
Octet RED96e Instrument	FortéBio	RED96E
Optilab T-REX	Wyatt	N/A
Orbitrap Eclipse mass spectrometer	Thermo Fisher Scientific	N/A
Superdex 200 10/300 GL	Cytiva/GE	Cat# 17517501
Ultimate 3000 HPLC	Thermo Fisher Scientific	N/A
Vivaspin 500, 3 kDa MWCO, Polyethersulfone	Sigma-Aldrich	Cat# GE28-9322-18

STAR Methods

Experimental model and subject details

H^N6-I3/N6-I3K^N6-I3/N6-I3 (CD4bs-Int¹), H^{min12A21/min12A21}K^{min12A21/min12A21} (CD4bs-Int²), H^N6-I2/N6-I2K^{N6-I2/ N6-I2} (CD4bs-Int³), HPCT64^{-18D/PCT64-18D}K^{PCT64-18D/PCT64-18D} (V2-Int¹), H^{minBG18.6/minBG18.6}K^WT/WT (V3-Int¹), and H^{minBG18.11/ minBG18.11}K^WT/WT (V3-Int²) BCR transgenic mouse lines were generated on the background of C57BL/6J (CD45.2^+/+) mice, as described previously^55,56. Note, the HC sequences for V3-Int¹ and V3-Int² differ from the previously published minBG18.6 and minBG18.11 HC sequences by one amino acid in the J_H gene (ARNAIRIYGVVALGEWFHYGMDVWGQGTAVTVSS for V3-Int^1/2; ARNAIRIYGVVALGEWFHYGMDVWGQGTTVTVSS for minBG18.6/11)⁴⁰. All transgenic mouse lines were generated in the animal facility of the Gene Modification Facility (Harvard University). Subsequent breeding, colony maintenance, and experimental procedures were performed at the animal facility of the Ragon Institute of Mass General Brigham, MIT, and Harvard. For experiments, wild-type (WT) adult male B6.SJL-Ptprca Pepcb/BoyJ (CD45.1^+/+) mice between the age of 8–12 weeks were purchased from The Jackson Laboratory (Bar Harbor, ME). Experimental mice were housed at the animal facility of the Ragon Institute with free access to food and water, controlled temperature, and a 12:12 hour light-dark cycle. All animal experiments were conducted in accordance with the Institutional Animal Care and Use Committee (IACUC) of Massachusetts General Hospital (MGH)’s approved Animal Study Protocols 2016N000286 and 2016N000022. The MGH Center for Comparative Medicine (CCM) is an Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC) International-approved program.

Immunogen and probe design

In this work we sought to employ native-like immunogens derived from wild-type HIV-1 isolates to present unmodified epitopes capable of engaging moderately evolved broadly neutralizing antibodies (bnAbs) targeting diverse neutralizing supersites on Env: CD4-binding site (CD4bs), V2-apex (V2), and V3-glycan (V3). We selected Env from two strains whose soluble antigens fulfill two independent criteria to evaluate how affinity differences on both the antigen and BCR sides influence antibody feedback mechanisms: (1) Env-1 must show detectable binding affinity to the diverse minimally mutated bnAbs of interest (PCT64–18, minBG18.6, minBG18.11, min12A21 and N6-I3), and (2) Env-2 must display significantly lower affinity than Env-1 for CD4bs antibodies. BG505 (clade A) and DU172–17 (clade C) served as Env-1 and Env-2, respectively, fulfill these criteria (Fig. S6B–C). BG505 MD39.3 construct design and characterization was previously described⁵⁸. DU127–17 was selected because it is resistant to neutralization by several CD4bs-targeting bnAbs and its recombinant antigens display low affinity against VRC01-class bnAbs^52,97. DU172 MD39.2 was designed using established HIV-1 Env stabilization strategies: (1) SOSIP mutations for improved stability^98–100, (2) C-terminal truncation at residue D664 for enhanced homogeneity of trimeric pre-fusion gp140^100,101, (3) MD39 mutations for improved antigenic profile, expression yield, and thermostability³⁹, and (4) replacement of the furin cleavage site with a non-cleavable linker termed link14 between gp120 and gp41 (linker sequence: SHSGSGGSGSGGHA)⁴⁰, where we use the terminology MD39.2 to denote a cleavage-independent MD39 stabilized trimer (Fig. S6A)⁵⁸. Mass spectrometry analysis and antigenic profiling against a panel of neutralizing and non-neutralizing antibodies performed by BLI confirmed DU172 MD39.2 constructs exhibit the expected N-linked glycan profile and adopt a trimeric pre-fusion conformation (Fig. S6B–D).

Probes for BG505 and DU172 were generated by addition of His-Avi-tag (HHHHHHGGSGGSGLNDIFEAQKIEWHE) and His-tag (HHHHHH) sequences at the C terminus of gp41 after residue D664. while His-tagged trimers served as ELISA probes, and His-Avi-tag SOSIP served as FACS probes.

His- and His-Avi-tagged epitope-specific KO reagents were generated by introducing four VRC01-class-specific KO mutations (280R, 365L, 368R, and 371R) onto BG505 MD39.3 and DU172 MD39.2 SOSIP trimers as previously described^90,102. His-Avi-tagged V2-apex epitope-specific KO mutants were created by introducing R169E and K171E mutations that abrogate binding by long-HCDR3 Apex bnAbs and related precursors⁹⁶. His-Avi-tagged V3-glycan epitope-specific KO variants were produced by introducing R327D, H330K and N332T mutations⁴⁰. All HIV-1 Env residues are denoted using HxB2 numbering.

Immunogen, probe, and Ab production

For protein immunogen production, untagged trimeric immunogens were produced by transient transfection of HEK-293F cells (ThermoFisher) and purified by gravity-flow affinity chromatography using Galanthus nivalis lectin resin (Vectorlabs) followed by SEC using Superdex 200 Increase 10/300 GL columns (Cytiva). The homogeneity and molecular weight of antigens was confirmed by size-exclusion chromatography-multi-angle light scattering (SEC-MALS) in PBS using Superdex 200 Increase 10/300 GL columns (Cytiva) with an isocratic flow of 0.5 mL/min followed by DAWN HELEOS II and Optilab T-rEX detectors (Wyatt Technology). Endotoxin levels in immunogen preparations were confirmed to be <5 EU/mg of endotoxin using an Endosafe nexgen-PTS instrument (Charles River).

For mRNA immunogen production, the amino acid sequence encoding BG505 MD39.3 (gp151) was provided to Moderna for production and formulation of mRNA-LNP immunogens.

For probe production, His-tagged and His-Avi-tagged trimeric antigens were produced by transient transfection of HEK 293F cells (ThermoFisher) and purified by immobilized metal ion affinity chromatography (IMAC) using HisTrap excel columns 5 mL (Cytiva) followed by size-exclusion chromatography (SEC) using Superdex 200 Increase 10/300 GL columns (Cytiva). The homogeneity and molecular weight of antigens was confirmed by size-exclusion chromatography-multi-angle light scattering (SEC-MALS) in PBS using Superdex 200 Increase 10/300 GL columns (Cytiva) with an isocratic flow of 0.5 mL/min followed by DAWN HELEOS II and Optilab T-rEX detectors (Wyatt Technology). Biotinylation of His-Avi-tagged trimers was performed using BirA (Avidity) and purified by SEC with Superdex 200 Increase 10/300 GL columns (Cytiva) to remove unconjugated biotin molecules.

For Ab production, sequences encoding the antibody Fv regions were synthesized by GenScript and cloned into antibody expression vectors pCW-CHIg-hG1 and pCW-CLIg-hk for heavy and light chain genes, respectively. Monoclonal antibodies (mAbs) were produced using transient transfection of HEK 293F cells (ThermoFisher) and purified by gravity-flow affinity chromatography using rProteinA Sepharose Fast Flow resin (Cytiva). Min12A21, N6-I3, minBG18.6, and minBG18.11 antibody mutants were produced by GenScript using the TurboCHO expression service. All antibodies were produced as human IgG1s.

SMNP was provided by the Irvine lab of Scripps Research Institute⁵⁹.

Bio-layer interferometry (BLI)

Native-like conformation of soluble DU172 gp140 antigens was confirmed by Bio-Layer Interferometry (BLI) using a panel of broadly neutralizing (quaternary-specific PGT151 and PGT145; V3-specific PGT121; CD4bs-specific N6 and min12A21) and non-neutralizing antibodies (CD4bs-specific F105 and B6; V3-specific 19b) and conducted on an Octet RED Instrument (FortéBio). Non-neutralizing antibodies bind non-native trimers and monomeric gp120; BG505 gp120 foldon trimer served as a negative control representing poorly assembled Envs (binding to non-nAbs and lack of binding to quaternary-specific bnAbs). Monoclonal antibodies were captured on anti-hIgG Fc capture (AHC) biosensors (Sartorius) at a concentration of 10 μg/mL in kinetics buffer (PBS, pH 7.4, 0.01% [w/v] BSA, and 0.002% [v/v] Tween 20) for 120 seconds after baseline determination. Association was measured for 120 seconds by dipping IgG-loaded biosensors into wells containing 1 μM recombinant SOSIP trimers. Dissociation was monitored for 120 seconds in kinetics buffer. Relative binding was determined by subtracting baseline absorbance and calculating the maximum response (nm) at endpoint of the association phase.

Biotinylated sort reagents of BG505 MD39.3, BG505 MD39.3 CD4bs-KO, DU172 MD39.2, and DU172 MD39.2 CD4bs-KO were loaded onto streptavidin (SA) biosensors (Sartorius) at 25 μg/mL in kinetics buffer for 120 seconds after baseline determination. After reacquiring a kinetics buffer baseline, biosensors were transferred to wells containing 1 μM monoclonal antibodies (neutralizing and non-neutralizing IgGs) and allowed to associate for 120 seconds. The biosensors were dipped into kinetics buffer alone to monitor dissociation for additional 120 seconds. Relative binding was determined by subtracting baseline absorbance values from end-point measurements.

Site-specific glycan profiling

N-linked site-specific glycan profiling was conducted as previously described to determine the degree of glycan occupancy and the extent of glycan heterogeneity (proportion of complex vs oligomannose/hybrid glycan types)¹⁰³.

Surface Plasma Resonance (SPR)

We measured kinetics and affinity of antibody-antigen interactions on Carterra LSA using CMDP Sensor Chip (Carterra) and 1x HBS-EP⁺ pH 7.4 running buffer (20x stock from Teknova, Cat. No H8022) supplemented with BSA at 1 mg/mL. The chip surface for ligand capture was prepared following Carterra software instructions. About 800–1000 RU of capture antibody (SouthernBiotech Cat.no 2047–01) in 10 mM Sodium Acetate pH 4.5 was amine coupled, and regeneration buffer Phosphoric Acid 1.7% was injected three times per cycle for 60 seconds. Ligand concentrations of 1 μg/mL were used with a contact time of 5 minutes. Analyte samples (SOSIP trimers) were buffer exchanged into the running buffer using dialysis and analyte concentrations were quantified on NanoDrop 2000c Spectrophotometer (Thermo Fisher Scientific) using absorption signal at 280 nm. Raw sensograms were analyzed using Kinetics software (Carterra), interspot and blank double referencing, Langmuir model. We typically cover a broad range of affinities in our runs and the best referencing practices differ depending on how fast the dissociation rate is for a particular ligand. For fast dissociation rates (faster than 9e⁻³ 1/s) we use automated batch referencing that includes overlay y-align and higher analyte concentrations. For slow dissociation rates (9e⁻³ 1/s or less) we use manual process referencing that includes serial y-align and lower analyte concentrations. After automated data analysis by Kinetics software, we also performed additional filtering to remove datasets with highest response signals smaller than signals from negative controls. This additional filtering was performed automatically using an R-script (available upon request).

B cell in vitro stimulation assay

Naive live B cells were purified from spleens of BCR transgenic and WT mice to reach more than 90% purity using negative B-cell isolation (Miltenyi Biotec). 10⁷ per mL B cells were labeled with 2 μM CFSE (Thermo Fisher) for 5 min at 37°C then were washed with complete B cell medium [RPMI supplemented with 10% FCS, 25 mM Hepes, Glutamax, Non-Essential Amino Acids, penicillin streptomycin (Thermo Fisher), and 50 μM β-mercaptoethanol (MilliporeSigma)]. CFSE Labelled cells were then stimulated in complete B cell medium supplemented with combinations of 10 ng/mL IL4 (R and D systems), 10 ng/mL IL5 (R and D systems), and 50 ng/mL CD40L (R and D systems). After 3 days of culture, proliferation status was measure by percentages of CFSE^low cells in flow cytometry based on CFSE levels at day 0 for each well. Survival status was measure by percentages of DAPI⁻ cells after staining with DAPI (Thermo Fisher Scientific). The differentiation of plasmablasts was measured by percentages of CD138⁺ cells after CD138 (281–2) staining.

Adoptive transfer and immunization

For adoptive transfer, spleens were collected from donor BCR transgenic mice with a C57BL/6J (CD45.2^+/+) background. Spleens were crushed through a 70 μm cell strainer and subjected to pan B cell isolation kits (Miltenyi Biotec). Isolated B cells were then qualified for live cells on a LUNA-FX7 automated cell counter (Logos Biosystems) and adjusted to the desired number and volume (100 μl/mouse) in phosphate-buffered saline (PBS) before transfer to CD45.1^+/+ recipient mice by intravenous (i.v.) injection through the orbital sinus, establishing frequencies of 20 in 10⁶ B cells except where otherwise specified. For mRNA immunization, mRNA-LNP immunogens were diluted to desired quantity and volume in PBS. Unless otherwise stated, diluted mRNA was injected at 2 μg per mouse in 100 μl per mouse, intramuscularly (i.m.) through gastrocnemius, 50 μl each leg. For protein immunization, 10 μg immunogens mixed with 5 μg of SMNP adjuvant were immunized to each mouse in 100 μl PBS subcutaneously (s.c.), 50 μl at each side of tail base.

Flow cytometry and cell sorting

For single cell suspensions, inguinal, popliteal, and iliac lymph nodes were crushed through a 70 μm cell strainer, centrifuged and re-suspended in FACS buffer (2% fetal bovine serum/PBS). Single-cell suspensions were kept on ice after. Probes were conjugated with streptavidin-Alexa Fluor 488, streptavidin-Alexa Fluor 647, or streptavidin-Alexa Fluor 594 for at least 30 min. Cells were blocked by Fc block (clone 2.4G2, BD Biosciences) for 15 min and then pre-incubated with freshly conjugated specific probes for 15 min. Probe-stained cells were then co-incubated with surface antibodies for another 15 min. If sorting for 10x sequencing, cell barcodes (BioLegend) and anti-mouse CD45 hashtags (BioLegend) were incubated with cells for 15 min during the coincubation step. Cells were washed 3 times with FACS buffer and re-suspended with DAPI 1:5000 diluted in FACS buffer, after which they were loaded into a BD LSR Fortessa analyzer or a BD FACS Aria Fusion sorter. Sorted cells were kept on ice for subsequent 10x sequencing procedures or at −80° for Sanger sequencing procedures. Data analysis was performed using FlowJo software (TreeStar).

Single-cell BCR sequencing

For 10x Genomics single-cell BCR sequencing, bulk sorted cells were loaded onto the 10x Genomics Chromium Controller at 5,000–20,000 cells per reaction and encapsulated in gel beads in emulsion. Single-cell gene expression, V(D)J, and hash-tag oligo libraries were constructed using the Next GEM Single-cell 5′ Reagent Kits v2 (10x Genomics, Pleasanton, CA) following the manufacturer’s protocol. Libraries were then quantified by Tapestation 4200 (Agilent, Santa Clara, CA) and the Qubit double-stranded DNA High Sensitivity assay (AAT Bioquest, Sunnyvale, CA). Qualified libraries were pooled and sequenced on the Nextseq2000 sequencer (Illumina, San Diego, CA). Finally, sequence data were analyzed using Cell Ranger pipelines and by customized analysis for specific BCR sequences.

For experiments with fewer cell numbers, single cell 96-well plate PCR was performed, as described previously¹².

LN homogenization and Enzyme-linked immunosorbent assays (ELISA)

To test IgG or IgM specific end point titers in LN, a homogenizing preparation step is needed prior to ELISA: one popliteal LN from each of the four mice in each group was carefully isolated and washed in PBS three times before all four were pooled into 100 μl of PBS with 1x Complete EDTA-free protease inhibitor cocktail (Roche). Pooled dLNs were homogenized on ice and centrifuged at 300 g, 4°C for 5 min. Supernatants were centrifuged again at 14000 g, 4°C for 10 min. Supernatants were transferred to a clean tube as prepared dLN homogenate and stored at −80°C for the next step.

Homogenate of dLN or mouse serum was then used for ELISA. Anti-His Ab (1 mg/ml, 50 μl per well) were incubated in 96-well high-absorption ELISA plates (NUNC/Corning) pre-coated with BG505 or epitope-specific-KO probes with His-tag (1 mg/ml, 50 μl per well) for 2 h at room temperature (RT). Plates were then washed 5 times with 0.05% Tween 20 in PBS (tPBS) and blocked with tPBS with 3% BSA for 2 h at RT. After washing, serial-diluted (2–5 folds depending on preliminary estimates of initial Ab titer) dLN homogenate or serum were incubated at 4°C for 4 h with requisite starting dilutions. Plates were washed again and incubated with 50 μl per well of Alkaline Phosphatase AffiniPure Goat Anti-Mouse IgG or IgM (Jackson Immuno Research) at 1:5000 in tPBS + 0.5 BSA FOR 1 h at RT. 50 μl per well of p-Nitrophenyl phosphate dissolved in ddH2O was added to each well for incubation of 20 min at RT for the final chromogenic reaction. Absorbance at 405 nm was determined with a Synergy Neo2 plate reader (BioTek). Endpoint titer was determined as dilution of the last serial-diluted well with an OD405 read over the threshold value, which was set as (mean + 3 × SD) of the OD405 read in all negative wells. A 5x dilution multiplier was applied to obtain the final est endpoint titer of dLNs.

Immunofluorescence and histology

Frozen murine lymph node tissue sections were immersed in Harris hematoxylin for 15 seconds before rinsing in deionized (DI) water and washing in 1X PBS for 30 seconds. Sections were dipped in Scott’s Tap Water Substitute for 45 seconds and rinsed in DI water before they were dipped in 70% EtOH and 90% EtOH for 30 seconds each. Sections were immersed in alcoholic-eosin for 2 minutes and rinsed in DI water. Sections were dipped in 90% EtOH and 100% EtOH for 15 seconds each and dipped in 100% EtOH for 30 seconds. Sections were immersed in xylene for one minute, mounted, and coverslipped.

To verify the presence of adoptively transferred plasma cells in murine lymph node tissue sections, mounted 10 μm frozen tissue sections were fixed in 10% formalin for 10 minutes and washed three times in 1X PBS. Sections were permeabilized in 0.1% Triton X-100 for 15 minutes and washed three times in 1X PBS. Sections were stained with antibodies for adoptively transferred cells (Alexa Fluor 488 anti-mouse CD45.2, 104, Biolegend, 1:200), B cells (Alexa Fluor 594 anti-mouse/human CD45R/B220, RA3–6B2, Biolegend, 1:200), and plasma cells (Alexa Fluor 647 anti-IRF4, 3E4, Biolegend, 1:200) for one hour at room temperature. Sections were washed three times in 1X PBS, mounted, and coverslipped. Confocal imaging was performed on the TissueFAXS SL Q (TissueGnostics).

Spatial transcriptomics: sample processing

Fresh frozen murine lymph node tissues were embedded in Tissue-Tek OCT compound and frozen on dry ice before transferring to −80°C for long-term storage. Frozen blocks were warmed to −20°C in a cryostat (Leica CM3050S) for 30 min prior to handling. Tissues were sliced at a 10 μm thickness and Slide-seq was performed on the fresh frozen sections as previously described¹⁰⁴. Serial sections were taken for tissue staining at a later time. Spatial libraries were sequenced on an Illumina NextSeq 2000 P3 flow cell with the following read structure: 42 bases Read 1, 8 bases Index 1, 41 bases Read 2, 0 bases Index 2. Each library received approximately 250 million reads.

Spatial transcriptomics: data analysis

As a quality control measure, we first removed beads with a unique molecular identifier (UMI) count of less than 100. Next, to delineate tissue boundaries and remove off-tissue beads, we performed Density Based Spatial Clustering of Applications with Noise (DBSCAN)¹⁰⁵. To ensure tissue regions were continuous, we subsequently performed dilation, whereby beads initially removed by DBSCAN were added back to the tissue if they were located within a small radius of on-tissue beads.

Robust Cell Type Decomposition (RCTD)⁶³ was used to assign cell types (plasma cell (PC), germinal center B cell (GCBC), cytotoxic T cell (Cytotoxic), follicular dendritic cell (FDC), myeloid cell (myeloid), CD4 T cell (CD4), naive B cell (NBC), memory B cell (MBC), epithelial cell (Epithelial), T follicular regulatory cell (Tfr), T follicular helper cell (Tfh), and plasmacytoid dendritic cell (PDC)) as singlets or doublets to each bead on Slide-seq data from lymph node samples. Mouse orthologs of human genes were identified using Ensembl BioMart¹⁰⁶ based on reference scRNA-seq datasets⁶⁴.

Germinal center regions were identified through marker gene expression (Bcl6, Aicda, Rgs13, and Stnb1m) and RCTD cell type assignments, including beads classified by RCTD as GCBC singlets or doublets containing GCBCs. These masks were merged, followed by DBSCAN clustering and spatial dilation.

To perform differential gene analysis, we used Cell-type Specific Inference of Differential Expression (C-SIDE)¹⁰⁷ to compare plasma cells inside versus outside germinal centers.

Statistical analysis

For Slide-seq data, analyses were conducted in Python 3.9.19 (The Python Software Foundation). For each gene/feature, pairwise group comparisons were performed using two-sided Mann–Whitney U tests. P-values were then adjusted for multiple testing using the Holm–Bonferroni correction.

All other statistical analyses were performed using Prism 10 (GraphPad). Normal distribution was not assumed. For comparisons involving two groups, Mann-Whitney’s test was utilized. For comparisons involving three or more groups, q-values were calculated by either 2-way Analysis of Variance (ANOVA) or nonparametric ANOVA (Kruskal-Wallis tests), as indicated in the legends; this was followed by post-hoc pairwise comparisons with Benjamini-Krieger-Yekutieli (BKY) correction for head-to-head comparisons.

Adjusted p-values (q-values) are indicated as follows: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; ns = not significant.

Supplementary Material

SUPPLEMENTAL INFORMATION

Document S1. Figures S1–S7

HIGHLIGHTS.

• One mRNA immunogen activated B cells to three different epitopes

• Higher affinity B cells displayed shorter germinal center (GC) half-lives

• Local antibody feedback established intra-epitope GC affinity floors and ceilings

• Antibodies from local plasma cells determined GC kinetics