Opposing effects of pre-existing antibody and memory T cell help on the dynamics of recall germinal centers

Secondary GC B cells show impaired binding to antigen

We hypothesized that the failure of recall GCs to contribute to antigen-specific titers upon homologous boosting¹² might be related to differences in their antigen-binding properties. To investigate this possibility, we followed a strategy we used previously to separate primary and secondary responses by anatomical location.²⁰ We first immunized wild-type (WT) C57BL/6 mice or S1pr2-CreERT2.Rosa26^{Lox-Stop-Lox-tdTomato} mice (S1pr2-Tomato, which allow us to fate-map primary GC B cells in a tamoxifen-dependent manner²⁵) with recombinant hemagglutinin (HA_PR8) in the right hind footpad (FP) to create a primary GC in the draining popliteal lymph node (pLN). One month later, we boosted these mice in the contralateral (left) FP with the same immunogen to generate a recall response (Fig. 1A). We then analyzed the left (boost-draining) pLN by flow cytometry 9 days after HA_PR8 immunization, when GCs are fully formed in these mice as well as in “primary” control mice primed with the unrelated antigen ovalbumin (OVA) (Fig. 1B,C, Supplementary Fig. 1A). Whereas a large proportion of B cells in primary GCs bound HA_PR8 tetramers, binding was near undetectable in secondary GCs (Fig. 1D,E). Fate-mapping of GC B cells in the S1pr2-Tomato mice by tamoxifen treatment during primary GC formation allowed us to measure HA binding separately in secondary GC B cells derived from naïve and memory precursors. Confirming our previous findings,²⁰ secondary GCs consisted almost entirely of tdTomato⁻ B cells, with only a minor (median 2.1%) contribution from memory-derived tdTomato⁺ cells (Fig. 1F). Tetramer binding was in most cases negligible also among this small tdTomato⁺ population (Fig. 1F, Supplementary Fig. 1A). Antigen binding among secondary GC B cells remained suppressed 18 days after homologous boosting (Supplementary Fig. 1B), indicating that affinity maturation is unable to rescue the poor antigen binding of recall GC B cells.

To rule out that these findings represent an artifact of tetramer staining, we sorted single primary and secondary GC B cells onto tissue culture wells containing feeder cells expressing CD40L, BAFF, and IL-21 (NB-21) that allow single B cells to proliferate and produce assayable amounts of antibody²⁶ (Supplementary Fig. 1C). We then assayed wells producing detectable amounts of Ig for their ability to bind HA_PR8 by ELISA (Fig. 1G,H). To estimate binding affinity, we calculated the ratio of HA_PR8 binding to Ig concentration, or “avidity index,”²⁶ categorizing samples into 3 strata (below detection, low, and high binding). Although the majority of supernatants from all experimental groups did not react detectably with the immunizing antigen in this assay, reactivity was markedly less frequent among GC B cell supernatants from secondary compared to primary GCs, which was true for both total GC and tdTomato⁺ sorted B cells (Fig. 1G,H). Thus, primary exposure to an antigen skews the binding properties of B cells recruited to secondary GCs, leading to loss of positivity in both tetramer binding and ELISA assays.

Poor antigen binding among recall GC B cells is not due to depletion of the naïve repertoire

A potential explanation for the skewed binding pattern of B cells in secondary GCs is that primary immunization may have depleted the naïve repertoire of B cells specific for the immunizing antigen. To rule this out, we allowed the naïve B cell pool time to re-form prior to boosting, using CD20-CreERT2.Rosa26^{Lox-Stop-Lox-tdTomato} (CD20-Tomato) mice²⁷ to measure naïve B cell turnover. We first fate-mapped the entire mature B cell compartment in these mice with two consecutive doses of tamoxifen (Fig. 1I). Three days later, when virtually all circulating IgM⁺IgD⁺ naïve B cells expressed tdTomato (Fig. 1J, Supplementary Fig. 2A), mice were immunized with HA_H3 both intraperitoneally and in one footpad (Fig. 1I). The fraction of fate-mapped naïve B cells declined progressively thereafter, with a median of 51.1% and 30.9% cells labeled after 25 and 41 days respectively, yielding a calculated half-life of 21.7 days (Fig. 1J). Turnover slowed down after 6 weeks—possibly reflecting the existence of a long-lived IgM⁺IgD⁺ population—stabilizing at approximately 16.0% at 14 weeks post-tamoxifen, including a minor contribution of spontaneous Rosa26^{Lox-Stop-Lox-tdTomato} recombination in the absence of tamoxifen estimated at 1.6% (Supplementary Fig. 2B). Boosting fate-mapped mice on the contralateral footpad at 3 months post-priming still generated GCs in which very few B cells (median 0.4%) were able to bind the HA_H3 tetramer (in contrast to occasional residual primary GCs with high binding) (Fig. 1K,L, Supplementary Fig. 2C). Therefore, the decreased ability of secondary GC B cells to detectably bind the immunizing antigen is not due to depletion of antigen-binding precursors from the naïve pool by primary immunization.

Of note, these experiments also strengthen our previous finding that secondary GCs are composed almost exclusively of naïve- rather than memory-derived B cells:²⁰ whereas recall ASCs were markedly enriched for tdTomato⁺ cells compared to follicular B cells in the same LN, indicative of their memory origin, the same was not true for secondary GC B cells, which were fate-mapped to the same extent as naïve B cells in the same mice (Fig. 1M). Thus, recall GCs are predominantly composed of B cells that had not yet been generated at the time of priming and are therefore unequivocally naïve- rather than memory-derived.

Poor antigen binding among recall GC B cells is caused by pre-existing antibody

Serum antibodies compete at the epitope level with naïve B cells of the same fine specificity, preventing their entry into GCs.^7–10,23,28 We hypothesized that, if loss of antigen binding in recall GCs is related to this phenomenon, mice boosted heterologously with antigenically drifted HAs containing novel epitopes not targeted by primary antibody would generate secondary GCs with improved binding to antigen. To test this, we primed mice with HA_FM1, an HA variant with 10% amino acid divergence from HA_PR8, and boosted these mice with HA_PR8 as above (Fig. 2A). In contrast to secondary GCs of homologously boosted mice, in which antigen-binding was again suppressed, HA_PR8-binding was partially rescued in GCs of heterologously boosted mice (Fig. 2B–D). These results were corroborated by ELISA measurements on the supernatants of single-GC B cell cultures (Fig. 2E,F). Rescue of HA_PR8-binders was also observed when mice were primed with HA_CA’09 and boosted with HA_PR8 (representing a 20% amino acid-level divergence) (Fig. 2G–J).

To formally test whether these differences were due to antibody-mediated feedback, we used Aicda^CreERT2/+.Rosa26^{Confetti/Confetti}.Prdm1^flox/flox mice (Prdm1ΔGC), in which the gene required for ASC differentiation (Prdm1, encoding for the transcription factor Blimp-1) can be deleted in a tamoxifen-inducible manner from the cohort of B cells that responded to priming, rendering these cells and their descendants incapable of differentiating into antibody-secreting cells. In mice treated with tamoxifen between days 3–12 post immunization, ASCs were efficiently depleted, whereas GC formation and export of HA⁺Confetti⁺ MBCs were largely unaffected compared to Aicda^CreERT2/+.Rosa26^{Confetti/Confetti}.Prdm1^+/+ (AID-Confetti²⁹) control mice (Supplementary Fig. 3A). To induce secondary GCs, we boosted Prdm1ΔGC or AID-Confetti mice approximately 50 days after priming with HA_PR8 in alhydrogel (Fig. 3A). Anti-HA_PR8 IgG was reduced by 9.0 and 5.4-fold in Prdm1^flox/flox compared to Prdm1^+/+ mice prior to and after boosting, respectively (Fig. 3B). This partial depletion of antigen-specific antibodies resulted in a partial rescue of the secondary GC phenotype, with secondary GCs in Prdm1ΔGC mice containing a median of 14.3% HA-binding B cells compared 1.5% in control mice (Fig. 3C,D). HA-binding among secondary GC B cells showed a moderate but significant inverse correlation with anti-HA IgG titers in homologously-boosted Prdm1ΔGC mice (r²=0.40, p=0.021, Supplementary Fig. 3C) which became much stronger when Prdm1^+/+ control mice were include in the calculation (r²=0.63, p<0.0001, Supplementary Fig. 3D). Thus, primary-derived antibodies directly inhibit the participation of high-affinity B cells in recall GCs.

The presence of a fate-mapping Rosa26^{Confetti/Confetti} allele in these mice also allowed us to assess the re-entry of MBCs into secondary GCs. The fraction of memory-derived B cells was significantly higher in Prdm1ΔGC mice (median 9.8%) compared to controls (median 0.93%) (Fig. 3D). Because Blimp-1-deficient B cells are precluded from differentiating into ASCs, loss of this factor may affect both MBC export from the primary GC and the decision of MBCs to re-enter GCs. These experiments therefore do not allow us to determine the extent to which antibody feedback is responsible for the increase in MBC-derived cells in secondary GCs. Nevertheless, most (>90%) of secondary GC B cells remained non-fate mapped—and are therefore of naïve origin—even when circulating antibody is depleted (Fig. 3E). Importantly, the large majority of HA⁺ GC B cells was not fate mapped in Prdm1ΔGC mice, indicating that recovery of HA-binding in these GCs was mostly due to enhanced recruitment of HA⁺ naïve B cells (Fig. 3F,G). We conclude that, while antibody-mediated feedback cannot explain the dominance of naïve-derived B cells in secondary GCs, it does account at least partially for the loss of antigen-binding among this population.

Memory T cells are required for recall GCs to form in the presence pre-existing antibody

A potential contributor to the ability of B cells with poor antigen binding capacity to enter recall GCs is the presence of a pre-expanded population of memory T cells, which would act by lowering the affinity threshold required for GC entry.^30,31 To test this possibility, we employed a hapten-carrier model where T cell help can be de-coupled from B cell specificity. We induced GCs by footpad immunization with 4-hydroxy-3-nitrophenylacetyl (NP)-CGG in alhydrogel of WT mice previously primed in the opposite leg with OVA (primary group), NP-CGG (homologous recall group), or NP-OVA (carrier-switch group) (Fig. 4A). In both recall groups, antibodies and MBCs specific for the hapten NP will be present at the time of boosting, whereas memory T cells specific for the carrier protein CGG will only be present in the homologous recall group. As with protein immunization, NP-CGG induced abundant GCs in the homologous recall group, which displayed moderately but significantly reduced tetramer binding compared to primary GCs (Fig. 4B,C). Usage of Igλ⁺ B cells, characteristic of an anti-NP response, was also significantly reduced in secondary GCs, but remained enriched compared to the available B cell repertoire (which is ~5% Igλ⁺ in C57BL/6 mice³²) (Supplementary Fig. 4A). More dramatically, secondary GC formation was severely suppressed in the carrier-switch group (Fig. 4B,C), indicating that help from naïve CGG-specific T cells is insufficient to overcome the negative effect of anti-NP antibody on GC formation. Moreover, ASC formation was markedly inhibited in carrier-switch mice compared to the homologous group (Fig. 4B,C), indicating that T cell help is required for the MBC to ASC transition in this setting. Notably, a CGG-specific GC response was not induced even in the carrier-switch group, in which CGG was not subject to epitope masking, possibly because NP-lation blocked access to immunodominant CGG antigens. Consistently with this, unhaptenated CGG was able to elicit a specific GC response when administered in conjunction with NP-CGG at the boost (Supplementary Fig. 4B,C). GCs remained suppressed at day 10 post immunization, although low numbers of non-NP and non-Igλ GC B cells began to be detectable at this point (Supplementary Fig. 4D). To test whether the inability of carrier-switch mice to generate recall GC and ASC responses could be directly attributed to the presence of anti-NP antibodies, we immunized Prdm1ΔGC and control AID-Confetti mice with OVA (primary group) or NP-OVA (carrier-switch group) and administered tamoxifen during the primary response to prevent ASC formation (Fig. 4D). NP-specific IgG titers were strongly if not fully depleted in Prdm1ΔGC mice compared to controls at both pre- and post-boost timepoints (6.3- and 7.3-fold reduction, respectively; Fig. 4E). The lack of boosting of serum anti-NP IgG titers from pre- to post-boost (Fig. 4E) is consistent with the impaired induction of ASCs in carrier-switch mice (Fig. 4C). In carrier-switch settings, boosting with NP-CGG resulted in significantly increased GC B cell and ASC populations in Prdm1ΔGC compared to Prdm1-sufficient mice (Fig. 4F), with anti-NP IgG titers inversely correlating with GC size (r²=0.43, P=0.0032, Supplementary Fig. 4E), indicating that a partial reduction in anti-NP antibodies partially rescues GC formation also in the absence of T cell memory (Fig. 4F). In this setting, depletion of antibodies led to no increase in the participation of MBCs in secondary GCs (Fig. 4G).

Finally, we sought to confirm that the ability of secondary GCs to form in the face of antibody feedback is a consequence of the increased availability of helper T cells. To this end, we used adoptive transfer of monoclonal T cells to increase helper T cell availability prior to boosting in carrier-switch settings, where GCs fail to form (Fig. 4H). We primed mice with NP-CGG (carrier-switch) or NP-OVA (homologous recall), which we boosted one month later in the contralateral footpad with NP-OVA. Carrier-switch mice received either no adoptive transfer or a low (5 × 10⁴) or high (5 × 10⁵) number of congenically-marked (CD45.1) OVA-specific OT-II T cells one day prior to boosting (Fig. 4H, Supplementary Fig. 3F,G). Whereas formation of recall GCs and ASCs was again strongly suppressed in carrier-switch mice that did not receive OT-II cells, this was overcome by adoptive transfer of high, but not low, numbers of OT-II T cells (Fig. 4I), with the abundance of transfer-derived T cells correlating positively with GC size (r²=0.83, P<0.0001, Supplementary Fig. 3H). NP-binding remained low in all recall settings (Supplementary Fig. 3I). Thus, increasing T cell help overcomes the suppressive effect of antibody feedback on GC formation, but not on high-affinity B cell recruitment. We conclude that, in homologous boosting regimens, the increased availability of help from memory T cells allows GCs to rapidly form even in the presence of antibody-mediated feedback, but at the cost of a loss of effective antigen binding among recruited B cells.

RESOURCE AVAILABILITY

Experimental Model and Subject Details

Method Details