ZBTB32 Restricts the Duration of Memory B Cell Recall Responses

This article is featured in In This Issue, p.1007

After clearance of infection or vaccination, Ag-specific long-lived plasma cells and memory B cells persist to mediate distinct aspects of long-term humoral immunity (1). Long-lived plasma cells constitutively secrete enormous quantities of Abs irrespective of the presence of Ag (2, 3). In contrast, memory B cells secrete Abs only when they are re-exposed to cognate Ags, after which they generate more rapid and robust responses than do their naive precursors (4). Differences between primary and secondary responses are mediated by several factors. First, the precursor frequency of Ag-specific memory B cells is greater than that of their naive counterparts (5). By expanding a larger number of clones, recall responses generate more plasma cells and Ab production than in primary responses. Second, unique cell-intrinsic properties mediate the rapid expansion and differentiation of memory B cells into plasma cells. For example, Ag engagement of isotype-switched IgG, expressed by many memory B cells, leads to more robust plasma cell differentiation in vivo than does IgM signaling (6–10). Consistent with these findings, upon reactivation, IgG-expressing memory B cells robustly generate plasma cells but yield comparatively fewer germinal center B cells (5, 11, 12). Additional transcriptional mechanisms mediate rapid plasma cell differentiation by memory B cells irrespective of Ab isotype (13). As one example, mouse CD80⁺ memory B cells express low levels of the transcription factor BACH2, which otherwise inhibits plasma cell differentiation (14). Although the rapid production of Abs by memory B cells upon re-exposure to pathogens such as influenza viruses is advantageous (15), mechanisms must exist to attenuate this response once the immunogen is cleared. Given the intrinsic gene expression differences between naive and memory B cells (16–18), it is possible that unique transcriptional programs curtail secondary Ab responses.

We and others (19, 20) recently demonstrated that ZBTB20, a member of the BTB/POZ transcription factor family, promotes durable primary Ab responses when alum is used as the adjuvant. Members of this family contain an N-terminal BTB/POZ domain that mediates dimerization and recruitment of transcriptional repressors and a C-terminal domain with a variable number of zinc fingers that mediate DNA binding (21). Hallmark members of this family that regulate aspects of the immune system include BCL6, which controls germinal center and T follicular helper cell development (22–27), ThPOK, which promotes CD4 versus CD8 thymocyte fate decisions (28, 29), and PLZF, which controls NKT cell development and function (30, 31). Another member of this family, ZBTB32, was initially identified through its ability to interact with testes-specific kinases, FANCC, and GATA3 (32–34), the latter of which leads to the in vitro suppression of cytokine production by CD4 T cells. ZBTB32 is essential for the proliferative burst of NK cells in vivo (35), but other reported immunological phenotypes of Zbtb32^−/− mice have been relatively subtle (36, 37). Subsequent work revealed that ZBTB32 is highly induced in B cells by LPS stimulation, partially represses Ciita transcripts, and is preferentially expressed by the CD80⁺ subset of memory B cells (13, 38). Yet the functional consequences of ZBTB32 expression in the B cell lineage are uncertain. In this study, we demonstrate that ZBTB32 specifically limits the rapidity and duration of memory B cell–mediated recall responses.

All animal procedures were approved by the Animal Studies Committee at Washington University in St. Louis (approval number 20140030). C57BL/6N, B6.SJLPtprc^aPepc^b (B6.SJL) and B6.CgIgh^aThy1^aGpi1^a (IgH^a) mice were purchased from Charles River Laboratories. Zbtb32^−/− mice have been described previously (36). All mice were bred in the animal facilities of the Washington University School of Medicine under pathogen-free conditions and experiments were performed in compliance with Washington University Animal Studies guidelines.

Total RNA was extracted with TRIzol (Life technologies) and first strand cDNA synthesis was performed with Superscript III Reverse transcription kit using oligo (dT) primers or random hexamers (Life Technologies) according to the manufacturer’s instructions. Quantitative RT-PCR was performed using SYBR Green PCR master mix (Applied Biosystems) on a Prism 7000 Sequence Detection System (Applied Biosystems). The primer sequences are as follows: Zbtb32, 5′-GGTGCTCCCTTCTCCCATAGT-3′ (forward) and 5′-GGAGTGGTTCAAGGTCAGTG-3′ (reverse); β-actin, 5′-CCTGAACCCTAAGGCCAAC-3′ (forward) and 5′-ACAGCCTGGATGGCTACG-3′ (reverse).

Zbtb32^+/+ and Zbtb32^−/− mice 8–10 wk of age were immunized i.p. with a single dose of 100 μg of 4-hydroxy-3-nitrophenyl-acetyl–chicken gammaglobulin (NP-CGG; hapten protein ratio: 15–22; Biosearch Technologies) precipitated in 5% aluminum potassium sulfate (Thermo Fisher Scientific) in PBS. Spleens were harvested 8–10 wk post immunization, and single-cell suspensions of splenocytes were subjected to gradient centrifugation using Histopaque 119 (Sigma-Aldrich) for 10 min at 2000 × g to remove the noncellular debris. Interface cells were then collected, and RBCs were lysed by resuspending in buffer containing 0.15 M NH₄Cl, 10 mM KHCO₃, and 0.1 mM EDTA (pH 7.2). Cells were washed twice with PBS, and 10% of the cells were retained for flow cytometric analysis. The remaining splenocytes were adoptively transferred into one, two, or three nonirradiated IgH^a- or B6.SJL-recipient mice, as described in each figure, by i.v. injection. A recall response was then elicited in the recipient mice 24 h later by i.v. administration of 50 μg of soluble NP-CGG.

ELISA plates were coated overnight at 4°C with 5 μg/ml NP₁₆BSA, NP₄BSA, or CGG (Biosearch Technologies) in bicarbonate coating buffer (0.1 M sodium bicarbonate and 0.02% sodium azide at pH 9.6). Plates were washed with wash buffer (PBS containing 0.05% Tween 20), and after blocking 1 h with blocking buffer (PBS supplemented with 2% BSA and 0.05% Tween 20) at 37°C, serially diluted serum samples were added and incubated for 1 h at room temperature. Technical duplicates were performed for every serum sample. Plates were washed with PBS with 0.05% Tween 20 and incubated with 1 μg/ml biotinylated anti-IgG1^b (B682; BD Biosciences) for 1 h, followed by streptavidin-conjugated HRP for 45 min. Peroxidase activity was detected by tetramethylbenzidine substrate (DakoCytomation), and the reaction was quenched with 2 N H₂SO₄, and ODs were quantified at 450 nm. The end-point titer of each sample was determined using Prism software (GraphPad Software) from a one phase exponential decay curve defined as the dilution that generates an OD₄₅₀ value of the background plus 3 SDs.

For the detection of NP-specific Ab-secreting cells, MultiScreen filter plates (EMD Millipore) were coated with 50 μg/ml NP₁₆BSA in PBS overnight. Total bone marrow cells were harvested 9 wk after recall from IgH^a-recipient mice and were seeded at 5–10 × 10⁶ cells/well in technical triplicates for every sample. Cells were cultured in 100 μl of RPMI 1640 medium/5% FBS overnight at 37°C. Wells were washed with PBS containing 0.05% Tween 20 and stained with biotinylated anti-IgG1^b and streptavidin-conjugated HRP (BD Biosciences). Spots were developed with 3-amino 8-ethyl carbazole (Sigma-Aldrich), and the reaction was stopped by rinsing the plates with water and the spots were counted in ImmunoSpot S6 Analyzer (CTL Laboratories).

The following mAbs were purified from hybridoma supernatants by BioXCell: 2C11 (anti-CD3), GK1.5 (anti-CD4), 536.7 (anti-CD8), 1D3 (anti-CD19), 8C5 (anti-Gr1), M1/70 (CD11b), TER119 (anti-Ter119), A20.1.7 (anti-CD45.1). Purified Abs were conjugated to Pacific Blue or Alexa Fluor 680 (Life Technologies), according to the manufacturer’s instructions. The following Abs were purchased from eBioscience: II/41 (anti-IgM)-PerCP–eFluor 710 or FITC, 1126 (anti-IgD)-FITC or Qdot 605, and GL-7 conjugated to biotin. The following Abs were purchased from BioLegend: A-20 (CD45.1)-allophycocyanin-Cy7, 1610A1 (anti-CD80)-PE, 29-2L17 (anti-CCR6)-PeCy7, GL-7-Pacific Blue, RA3-6B2 (B220)-Alexa 488 or -PerCp-Cy5.5, 281-2 (CD138)-PE or -allophycocyanin, ICRF44 (CD11b)-PerCp-Cy5.5, 104D2 (CD117)-PerCp-Cy5.5, RB68C5 (Ly6G)-BV605, anti-human CD20 (2H7)-allophycocyanin-Cy7, anti-human CD27 (O323)-Alexa 647, anti-human CD38 (HIT2)-PE-Cy7, anti-human CD2 (RPA-2.10)-PE, anti-human CD19 (HIB2)-BV421, anti-human IgM (MHM-88)-FITC, anti-human CD138 (MI15)-FITC, and streptavidin-Qdot 605. Unlabeled Mouse anti rat IgG (H+L) was bought from Southern Biotechnology Associates. NP–allophycocyanin used for surface staining and in intracellular staining was made by conjugating allophycocyanin (Sigma-Aldrich) with NP-O–succinimide ester (Biosearch Technologies) as described previously (39, 40).

Bone marrow plasma cells were sorted following CD138 enrichment in which total bone marrow cells were stained with 1 μl of anti–CD138-PE (BioLegend) for every 10⁷ cells. CD138⁺ cells were selectively enriched using anti-PE magnetic beads (4 μl of beads /10⁷ cells) and MACS LS columns (Miltenyi Biotec) prior to flow cytometric analysis and/or purification. Total splenocytes were depleted of the non-B cell lineage cells and IgM⁺, IgD⁺, and GL-7⁺ cells by cellular panning. Briefly, cells were surface stained with rat anti-mouse Abs against non-B cell lineage cells (CD3, CD4, CD8, Gr1, CD11b, and Ter119), non–class-switched (IgM⁺, IgD⁺), and germinal center B (GL-7⁺) B cells for 30 min on ice. Cells were then washed with and incubated in plates precoated with 1 μg/ml mouse anti rat IgG (Southern Biotechnology Associates) for 30 min at 4°C. The nonadherent cell fraction was then collected, washed, and stained to identify the resting and NP⁺ memory B cells (CD19⁺CCR6⁺CD80⁺ class-switched IgM⁻IgD⁻ cells). For microarrays, resting memory B cells and donor (CD45.2)-derived NP-specific activated memory B cells and bone marrow plasma cells were sorted into PBS with 5% FBS or directly into RNA lysis buffer (Macherey–Nagel). All cell analysis and sorting was done on a FACS Aria II (BD Biosciences).

Following adoptive transfer and NP-CGG rechallenge, mice were fed with 2 mg/ml BrdU in the drinking water for the durations indicated. Splenic plasma cells, memory B cells and CD138-enriched bone marrow plasma cells were first stained for surface expression of respective Abs as described in Supplemental Fig. 1. Cells were then processed and stained for incorporated BrdU with the FITC BrdU Flow kit (BD Biosciences), according to the manufacturer’s instructions.

Total RNA was prepared from purified resting memory B cells (5,000–10,000 cells), donor-derived NP-specific day 7 memory B cells (100–500 cells), and donor-derived NP-specific bone marrow plasma cells (100–500 cells) using a NucleoSpin RNA isolation kit (Macherey–Nagel), which includes a DNase digestion step. cDNA amplification using the GeneChip WT Pico Kit and hybridization of labeled cDNA to GeneChip Mouse Gene 1.0 ST Arrays (Affymetrix) were performed by the Genome Technology Access Center core facility at Washington University in St. Louis. Analysis of microarray data was performed using Arraystar software (DNASTAR). Quantiles-based normalization was performed, and Student two-tailed t tests were used to calculate differentially expressed genes with p < 0.05 without multiple test correction. All such differentially expressed genes were entered into the Consensus Pathway Database (http://cpdb.molgen.mpg.de/), and enriched pathways with q < 0.05 were considered significant. Data can be accessed through the National Center for Biotechnology Information Gene Expression Omnibus database under accession number GSE83194 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE83194).

All procedures in this study were approved by the Human Research Protection Office at Washington University. Bone marrow was obtained from total hip arthroplasty samples from patients undergoing elective surgery (Barnes Jewish Hospital). For bone marrow aspirates, 20–30 ml of IMDM was added, samples were dissociated using vigorous agitation, and filtered through 70-μM nylon mesh. RBCs were removed using 0.15 M NH₄Cl, 10 mM KHCO₃, and 0.1 mM EDTA (pH 7.2) buffer, samples were stained with 1 μl/10⁷ cells anti-CD138 microbeads (Miltenyi Biotec), and enriched on an AutoMacs machine (Miltenyi Biotec) over two columns. Peripheral blood was obtained from the Barnes Jewish Hospital Pheresis center from waste Trima fillers. Samples were spun through a 1.077 g/ml Histopaque gradient (Sigma-Aldrich), interface cells were collected, lysed with 0.15 M NH₄Cl, 10 mM KHCO₃, and 0.1 mM EDTA (pH 7.2), and stained for FACS as in Supplemental Fig. 1.

Human samples gated as shown in Supplemental Fig. 1 were double-sorted into RLT buffer (Qiagen), and RNA was prepared using an RNeasy Micro kit (Qiagen). Sequencing libraries were generated using a Clontech Smart-Seq kit. Single-end 50-bp reads were acquired using an Illumina HiSEquation 2500. Reads were mapped to the Ensembl_R72 reference human genome using Tophat (https://usegalaxy.org/). Resultant .bam files were processed, and reads per kb of transcript per million mapped read values were computed using Partek software. Data can be accessed through the National Center for Biotechnology Information Gene Expression Omnibus database under accession number GSE81443 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE81443).

Means, geometric means, SEM, 95% confidence interval, unpaired Student 2-tailed t tests, Mann–Whitney tests, and one-phase exponential decay curve fitting for end-point dilution estimation were calculated with Prism software.

ZBTB32 is highly expressed by CD80⁺PD-L2⁺ memory B cells (13), independent of whether they express IgM or IgG. These cells preferentially differentiate into plasma cells instead of germinal center B cells upon antigenic rechallenge (13). To expand the analysis to other B cell subsets (Supplemental Fig. 1), we performed quantitative RT-PCR analysis. Mouse polyclonal CD80⁺ memory B cells express 20- to 30-fold higher levels of ZBTB32 transcripts than do naive and germinal center B cells and polyclonal bone marrow plasma cells (Fig. 1A). RNA-seq analysis of human lineages also revealed expression of ZBTB32 in both IgM⁺ and IgG⁺ memory B cells but undetectable expression in naive B cells and polyclonal bone marrow plasma cells (Fig. 1B). Thus, ZBTB32 is specifically expressed by both mouse and human memory B cells.

We next performed experiments to define the functional role of ZBTB32 in primary and secondary Ab responses. Zbtb32^−/− mice were immunized with the T-dependent Ag NP-CGG, and Ag-specific serum Ab titers were quantified at 1 and 8 wk postimmunization. No differences were observed in NP-specific serum Ab titers between Zbtb32^+/+ and Zbtb32^−/− mice at either time point (Fig. 2A). At 10 wk postimmunization, Zbtb32^+/+ and Zbtb32^−/− mice possessed similar numbers of NP-specific splenic memory B cells (Fig. 2B). Moreover, the ratios of Ag-specific CD80⁺PD-L2⁺, CD80⁻PD-L2⁺, and CD80⁻PD-L2⁻ memory B cell subsets were unchanged in Zbtb32^−/− mice (dat not depicted). Thus, ZBTB32 is dispensable for primary Ab responses.

To test secondary responses, splenocytes from Zbtb32^+/+ and Zbtb32^−/− mice 10 wk after NP-CGG immunization were adoptively transferred into allotype-distinct naive IgH^a mice. One day later, these recipients were challenged i.v. with soluble NP-CGG to elicit a recall response. Donor IgG1^b NP-specific serum Ab titers derived from Zbtb32^−/− cells were elevated relative to controls at 1 wk postimmunization (Fig. 3A). In subsequent weeks, wild-type Ag-specific IgG1^b titers gradually declined, but Abs derived from Zbtb32^−/− cells remained elevated (Fig. 3A). The affinities of these Abs were similar between donor genotypes, as measured by NP₄/NP₁₆ binding ratios (Fig. 3B). At 9 wk postimmunization, ELISPOT assays were performed to quantify the numbers of NP-specific Ab-secreting cells. Far greater numbers of bone marrow Zbtb32^−/− IgG1^b NP-specific Ab-secreting cells were observed relative to Zbtb32^+/+ controls (Fig. 3C). We also examined Ab titers specific for CGG because these responses qualitatively differ from those against haptens (41, 42). Although end-point titers were difficult to calculate in this response, CGG-specific Abs derived from Zbtb32^−/− cells were clearly elevated relative to controls at both early and late time points (Fig. 3D). Thus, ZBTB32 restricts the rapidity and duration of Ab production following memory B cell recall responses to both haptens and proteins.

The rapid and prolonged Zbtb32^−/− memory B cell recall responses could be driven by either cell-extrinsic or -intrinsic changes. These changes could manifest in memory B cells and/or their downstream plasma cells. To distinguish between these possibilities, CD45.2⁺ Zbtb32^+/+ and Zbtb32^−/− and CD45.1⁺ wild-type mice were immunized with NP-CGG. At 10 wk postimmunization, CD45.1⁺ and CD45.2⁺ splenocytes were mixed at equal ratios and transferred into naive CD45.1⁺ recipients (Fig. 4A). Mice were challenged with i.v. NP-CGG, and Ag-specific B cell subset chimerism was assessed 7 and 21 d later. Under these conditions, we observed almost no CD45.2⁺ donor germinal center B cells (data not depicted). CD45.2⁺ donor chimerism was similar between Zbtb32^+/+ and Zbtb32^−/− bone marrow plasma cells at day 7 (Fig. 4B). However, Zbtb32^−/− bone marrow plasma cell chimerism was elevated at day 21 relative to controls (Fig. 4B, Supplemental Fig. 2). A slight decrease in CD45.2⁺ chimerism was observed between days 7 and 21 in both genotypic groups (Fig. 4B), likely because of delayed host CD45.1⁺ responses to NP-CGG. No differences in chimerism were observed between Zbtb32^+/+ and Zbtb32^−/− memory B cells at either day 7 or 21 (Fig. 4C, Supplemental Fig. 2). Similarly, splenic plasma cell chimerism was similar between Zbtb32^+/+ and Zbtb32^−/− genotypes at day 7 (Fig. 4D, Supplemental Fig. 2). At day 21, we were unable to identify splenic Ag-specific plasma cells above background staining (data not depicted). These data demonstrate that ZBTB32 deficiency prolongs Ab responses by elevating the numbers of secondary bone marrow plasma cells at late time points through a cell-intrinsic mechanism.

Though these data provide a cellular basis for durable Zbtb32^−/− Ab recall responses, they do not explain the elevated Ab levels at early timepoints (Fig. 3A). We hypothesized that ZBTB32 deficiency leads to more rapid Ab production relative to controls at timepoints before day 7 in the recall response. Because Ag-specific cells are difficult to quantify by flow cytometry at these early stages (data not depicted), we turned to a more sensitive system. Zbtb32^+/+ and Zbtb32^−/− mice were immunized, and 8 wk later, T cell–depleted splenocytes were transferred along with wild-type T cells from NP-CGG–primed mice into Rag1^−/− recipients. At 3.5 and 4.5 d posttransfer and immunization, ELISPOT assays were performed on splenocytes. At both time points, Zbtb32^−/− memory B cells yielded more splenic Ab-secreting cells than did control memory cells (Fig. 4E). Taken together, these data demonstrate that ZBTB32 deficiency leads to more rapid production of secondary plasma cells in the spleen and more prolonged maintenance of secondary plasma cells in the bone marrow relative to ZBTB32-sufficient controls.

The prolonged Ab production phenotype could be caused by ZBTB32-mediated regulation of differentiation programs in memory B cells and proliferative and/or survival programs in bone marrow plasma cells. We performed microarray analyses of Zbtb32^+/+ and Zbtb32^−/− resting memory B cells, NP-specific activated memory B cells isolated 7 d after adoptive transfer and immunization, and NP-specific secondary bone marrow plasma cells at this same time point (Supplemental Fig. 3). The magnitude of gene expression differences between Zbtb32^+/+ and Zbtb32^−/− resting memory B cells was relatively modest, with only 34 genes displaying statistically significant >2-fold expression changes (Fig. 5A). Even when all 1386 statistically significant transcripts were analyzed, no clear signatures could be derived from Consensus Pathway Database analysis (http://cpdb.molgen.mpg.de/). Upon activation in vivo, 147 genes became differentially expressed >2-fold between Zbtb32^+/+ and Zbtb32^−/− CD80⁺ memory B cells (Fig. 5A). Pathway analysis of all statistically significant changes in Zbtb32^−/− activated memory B cells revealed elevated expression of genes such as cathepsins A, B, and D, which are involved in lysosomal function and Ag processing (43), and the dual-specificity phosphatases 1, 7, and 16, which regulate MAPK signaling (44) (Fig. 5B). Yet, we observed no transcriptional evidence of changes in signatures related to proliferation or survival, consistent with the data in Fig. 4C.

Among secondary bone marrow plasma cells, ZBTB32 expression itself was extinguished (Fig. 5A). When all 1177 statistically significant differences were considered, Zbtb32^−/− NP-specific secondary bone marrow plasma cells displayed elevated expression of numerous genes known to promote cell cycle progression (Fig. 5C), such as E2f3, which transcriptionally promotes entry into S-phase (45), and Pcna, which promotes DNA replication (46, 47). In addition, we observed elevated expression of numerous mitochondrial ribosomal genes (Mrpl18, Mrps22, Mrpl51, Mrps25, Mrps30, Mrpl30, and Mprs27), which promote translation (Fig. 5C). Mitochondrial translation is essential for production of proteins involved in the electron transport chain (48), and we have shown that maximal respiratory capacity is essential for long-lived plasma cell survival (49). Thus, the transcriptional data of bone marrow Zbtb32^−/− secondary plasma cells suggest increased proliferation and survival, either or both of which could contribute to prolonged Ab production in vivo.

To distinguish between these possibilities, we performed BrdU labeling experiments. Immunizations and adoptive transfers were established, and mice were given BrdU in the drinking water for 3 d prior to analysis (Fig. 6A, Supplemental Fig. 4) (50). The majority of both Zbtb32^+/+ and Zbtb32^−/− bone marrow plasma cells had incorporated BrdU at day 7, suggesting either recent proliferation or derivation from upstream proliferating precursors (Fig. 6B, Supplemental Fig. 4). There was a trend toward increased BrdU incorporation by Zbtb32^−/− secondary plasma cells (Fig. 6B), perhaps reflective of the transcriptional data, elevated serum Abs at this time point, and increases in early plasma cell numbers (Figs. 3A, 4E, 5C). However, this difference did not reach statistical significance, and by day 21, we observed little BrdU incorporation in either genotypic group (Fig. 6B). The majority of splenic memory B and plasma cells also incorporated BrdU at day 7 (Fig. 6C, 6D, Supplemental Fig. 4). Splenic plasma cells were difficult to detect at day 21, but memory B cells showed little evidence of BrdU incorporation at day 21 (Fig. 6C). Thus, by day 21, proliferation had subsided in all Ag-specific B cell subsets. To further demonstrate this point, we performed BrdU pulse-chase experiments. Adoptive recipients of Zbtb32^+/+ and Zbtb32^−/− immune splenocytes were challenged with NP-CGG and concomitantly fed BrdU in the drinking water for 7 d. Animals were then given normal drinking water for 14 d, and bone marrow plasma cells were analyzed. BrdU label retention in donor bone marrow plasma cells was similar between Zbtb32^+/+ and Zbtb32^−/− genotypes (Fig. 6E). These data demonstrate that ZBTB32 restricts the duration of secondary Ab responses by attenuating survival, rather than the proliferation of plasma cells.

In this study, we have demonstrated that the transcription factor ZBTB32 is important for attenuating Ab production after a memory B cell recall response. Given that ZBTB32 expression is turned off by the plasma cell stage, the data imply that lifespan is imprinted during the upstream activation phase, as has been previously proposed (51, 52). Yet, no molecular mechanism has yet been determined that would explain this process. In our study, we observed that activated Zbtb32^−/− memory B cells elevate expression of genes involved in lysosomal and MAPK pathways. The elevated expression of lysosomal genes such as cathepsins is of particular interest, given their importance in MHC class II Ag processing (43). It is possible that upon reactivation, Zbtb32^−/− memory B cells are better at presenting peptide:MHC class II and receiving T cell help than are their wild-type counterparts. Although the extreme paucity of Ag-specific memory B cells in vivo has limited our ability to test this hypothesis, this model is consistent with other studies showing that the ability to interact with T cells is a major determinant of B cell fate (53, 54). Exactly how this interaction would promote longevity in plasma cells is not clear but may involve mitochondrial reprogramming.

Durable maintenance of Ag-specific Abs following secondary infection or booster immunization is desirable, so it is not clear why a ZBTB32-dependent genetic program exists to restrict such responses. One possibility is that only a finite number of long-lived plasma cells can be maintained because of limiting numbers of survival niches (55). An inability to properly attenuate secondary plasma cell responses could thus competitively antagonize pre-existing immunity over the course of a lifetime (56). A second possibility is that in autoimmunity-susceptible backgrounds, ZBTB32 deficiency could exacerbate disease by prolonging autoantibody production. We did not observe any obvious signs of spontaneous autoimmunity in Zbtb32^−/− mice, but they have been maintained in the largely resistant C57BL/6 strain. Finally, it is possible that ZBTB32 deficiency predisposes to B and plasma cell transformation and malignancy. Although not frequent, in the Multiple Myeloma Genomics Portal (https://www.broadinstitute.org/mmgp/home), we have identified at least one sample carrying a Zbtb32 deletion. Moreover, the levels of ZBTB32 expression distinguish different types of diffuse large B cell lymphomas (57). Appropriate genetic models will help distinguish between these etiological possibilities.

We thank I-Cheng Ho for permission to use Zbtb32^−/− mice. We also thank Krista Kennerly for laboratory management, You Zhou, Adrianus Boon, and Chyi Hsieh for technical assistance with experiments, Erica Lantelme and Dorian Brinja for assistance with flow cytometry, and Rachel Wong, Derrick Callahan, Wing Lam, and Hannah Pizzato for helpful comments on the manuscript.

The authors have no financial conflicts of interest.