Main

Purinergic receptors of the P2RX family are ligand-gated cation channels that interpret the microenvironmental release of ATP2. In the immune system, P2RX receptors can augment innate and adaptive immunity by sensing the uncontrolled release of ATP by infected or dying cells3,4. By contrast, gap-junction proteins such as pannexin 3 (PANX3) control the regulated release of ATP by developing and mature bone marrow (BM) osteoblasts5. We analysed haematopoietic and immune cell populations in Panx3−/− mice, which have reduced BM cellularity because they have shorter bones with narrower medullary cavities but otherwise have relatively intact BM architecture6. Consistently, numbers of developing BM myeloid and B cell precursors, including immature and mature B cells, were reduced by 50–70% (Extended Data Fig. 1a,b). We observed minimal impact on developing T cells in the thymus, B cells and T cells in the spleen or peritoneal cavity B1 B cells. (Extended Data Fig. 1c–e; see Extended Data Figs. 2 and 3 for flow cytometric gating strategies for BM plasma cells (PCs) and precursor populations and activated splenic B cells). However, serum IgG, IgM and IgA titres were reduced substantially in Panx3−/− mice (Fig. 1a) and ELISPOT analyses revealed far fewer IgG- and IgM-secreting PCs among BM but not spleen cells from Panx3−/− mice (Fig. 1b). Furthermore, BM PCs identified by surface co-expression of CD138 and SCA-1 (ref. 7) were reduced profoundly in Panx3−/− mice (Fig. 1c). CD138highSCA-1+ BM PCs include newly generated antibody-secreting cells identified by surface B220 expression, as well as a B220− population enriched for mature long-lived PCs (LLPCs)8 and both BM PC subsets were compromised by Panx3 mutation (Extended Data Fig. 1f). By sharp contrast, frequencies and numbers of splenic PCs were unchanged (Extended Data Fig. 1g). Therefore, PANX3 promotes the generation and/or maintenance of mature BM PCs.

Fig. 1: Panx3 mutation ablates BM PCs.
figure 1

a, ELISA data for total IgG, IgM and IgA in adult Panx3+/+ and Panx3−/− sera. b, Representative images and quantification of ELISPOT assays for IgG- and IgM-secreting BM and spleen cells from Panx3+/+ and Panx3−/− mice. c, Viable BM PCs from Panx3+/+ and Panx3−/− adults. Left-most plot shows representative Dump/IgD gate in which the Dump channel is antibodies to CD4, CD8, TER-119 and F4/80. Shown on the right are means for BM CD138highSCA-1+ cells in the indicated mice using the graph legend shown for panel a. d, Representative ELISPOT images and numbers of IgG-secreting BM cells from Panx3+/+ or Panx3−/− mice on day zero or 4 days after culture with osteoblastic calvarial cells from Panx3+/+ or Panx3−/− mice. Data shown were combined from two experiments. e, eATP in supernatants from sextuplicate cultures of induced osteoblastic cells from Panx3+/+ or Panx3−/− mice. f, Representative ELISPOT images and numbers of IgG-secreting BM cells on day zero or after addition of 100 μM ATP, ADP or AMP for 48 h before transfer to ELISPOT wells. For ELISPOT assays 106 total cells per well were added except for assays for IgM-secreting cells in b for which 5 × 104 cells per well were evaluated. n = 6 per group for a,c–f; *P = 0.0022. n = 9 per group for b; *P < 0.0001. All P values were derived from two-tailed Mann–Whitney tests, without any adjustment for multiple comparisons. All experiments were performed at least thrice except ELISPOT assays in b which were performed once for splenic IgG and IgM. All bar graphs are means.

Source Data

Full size image

Because PANX3 is expressed chiefly by osteoblasts5, we induced osteoblast differentiation from calvarial cells from wild-type (WT) and Panx3−/− mice and established cocultures with these osteoblasts and BM cells from WT or Panx3−/− mice (Extended Data Fig. 4a). Four days later we performed ELISPOT assays to quantify PCs and noted far fewer IgG-secreting cells when WT BM cells were cultured on Panx3−/− osteoblast feeder cells compared to all other combinations (Fig. 1d). Of note, numbers of PCs recovered from BM cells cultured with WT osteoblasts mirrored PC numbers on day zero (Fig. 1d). Whereas CXCL12, SCF, interleukin-6 (IL-6) and APRIL can each support PC survival9, we found no differences in mRNA abundance for each in induced WT and Panx3−/− osteoblasts (Extended Data Fig. 4b). However, supernatants of Panx3−/− osteoblasts contained reduced extracellular ATP (eATP) concentrations (Fig. 1e), raising the possibility that osteoblasts support PCs by using PANX3 to release eATP. Consistent with this possibility, adding ATP but not ADP or AMP without osteoblastic feeders drove recovery of PCs from WT BM cells to amounts in line with PC numbers observed on day zero (Fig. 1f). Similarly, eATP addition also promoted recovery of the few PCs available in Panx3−/− BM cells, again to amounts identified on day zero (Fig. 1f). Together, these data suggest that PANX3-regulated eATP release modulates BM PC function and/or survival with a lesser role for PCs outside the BM.

P2RX receptors include seven different family members with different expression patterns. To test for a role for P2RX receptors in BM PCs, we first treated mice with the global P2RX inhibitors suramin or PPADS10,11. Whereas suramin is less effective at inhibiting P2RX4 function12, both suramin and PPADS reduced frequencies of CD138highB220− BM cells with the phenotype of PCs in vivo (Extended Data Fig. 5a). Available RNA-sequencing data indicate that, among P2RX family members, P2RX4 is highly expressed by mouse BM PCs and is also expressed by human BM PCs13,14 (Extended Data Fig. 5b,c). Thus, we mutated P2rx4 in B-lineage cells by crossing P2rx4fl/fl mice with mice harbouring either an mb-1-cre allele (mb-1 is also known as Cd79a) for mutation in early B cell precursors15 or a CD20-regulated tamoxifen-induced Cre-ERT2 BAC transgene (CD20-TAM-Cre;  human CD20 is also known as MS4A1) for induced mutagenesis in mature B cells16.

Steady-state serum immunoglobulin titre and both B220+ and B220− BM PCs were significantly reduced in P2rx4fl/fl;mb-1-cre mice compared to P2rx4+/+;mb-1-cre controls, whereas numbers of splenic PCs were unchanged (Fig. 2a,b and Extended Data Fig. 6a,b). ELISPOT assays revealed that IgG- and IgM-secreting cells were reduced substantially in BM but not spleen in P2rx4fl/fl;mb-1-cre adults (Fig. 2c). By contrast, frequencies of naive B cells in the spleen, developing B-lineage cells in the BM and peritoneal cavity B1 B cells were minimally changed (Extended Data Fig. 6c,d). Consistent with the notion that BM PCs use P2RX4 to sense eATP, adding eATP augmented numbers of IgG-secreting cells in short-term cultures of P2rx4+/+;mb-1-cre but not P2rx4fl/fl;mb-1-cre BM cells (Fig. 2d). In parallel we fed P2rx4fl/fl;CD20-TAM-Cre and P2rx4+/+;CD20-TAM-Cre mice tamoxifen-laced chow for 4 weeks beginning at 5 weeks of age. This strategy also caused a twofold to threefold decrease in steady-state serum immunoglobulin titre (Fig. 2e). Also, delivery of tamoxifen in this fashion to P2rx4fl/fl;CD20-TAM-Cre and P2rx4+/+;CD20-TAM-Cre mice beginning at 12 weeks of age caused substantial loss of the bulk of steady-state BM PCs (Fig. 2f) and equally compromised both B220+ and B220− PCs in the BM without affecting numbers of steady-state PCs in the spleen and with minimal effects on naive and developing B-lineage cells (Extended Data Fig. 6e–g). ELISPOT analyses of cultured BM cells confirmed that tamoxifen-pulsed P2rx4fl/fl;CD20-TAM-Cre mice possessed reduced numbers of PCs, even with addition of eATP (Extended Data Fig. 6h). We conclude that BM PCs critically depend on P2RX4.

Fig. 2: P2RX4 is required for BM PCs.
figure 2

a, ELISA data for total IgG, IgM and IgA in sera from WT;mb-1-cre and P2rx4fl/fl;mb-1-cre adults. b, BM PCs in WT;mb-1-cre and P2rx4fl/fl;mb-1-cre adults gated on Dump−IgD− cells as shown in Fig. 1. c, Representative images and quantification of ELISPOT wells for IgG- and IgM-secreting BM and spleen cells from WT;mb-1-cre and P2rx4fl/fl;mb-1-cre mice. d, Representative images and quantification of ELISPOT wells for IgG-secreting BM cells from the indicated mice cultured for 2 days with or without ATP addition before transfer to ELISPOT wells with more ATP where indicated. e–g, WT;CD20-TAM-Cre and P2rx4fl/fl;CD20-TAM-Cre mice were given tamoxifen-laced chow for 4 weeks before direct analysis (e,f) or immunization followed by analysis (g). e, ELISA data for total IgG, IgM and IgA in sera from WT;CD20-TAM-Cre and P2rx4fl/fl;CD20-TAM-Cre adults immediately after tamoxifen delivery. f, BM PCs gated as in Fig. 1c from WT;CD20-TAM-Cre and P2rx4fl/fl;CD20-TAM-Cre adults fed tamoxifen-laced chow for 4 weeks beginning at 12 weeks of age. g, WT;CD20-TAM-Cre and P2rx4fl/fl;CD20-TAM-Cre mice were fed tamoxifen for 4 weeks beginning at 5 weeks of age, then immunized with NP-KLH in alum. Shown are ELISPOT analyses for NP-specific PCs in spleen and BM on days 7 and 30 after immunization, respectively. For ELISPOT assays 106 total cells per well were added except for assays for IgM-secreting cells in c for which 5 × 104 cells per well were evaluated. n = 6 per group for a,b,d–g; *P = 0.0022. n = 9 per group for c; *P < 0.0001. All P values were derived from two-tailed Mann–Whitney tests, without any adjustment for many comparisons. The experiments in a,b,d–f were performed at least thrice. The experiments in c were performed at least twice and the experiment in g was performed twice. All bar graphs are means.

Source Data

Full size image

The impact on steady-state numbers of BM PCs in P2rx4fl/fl;CD20-TAM-Cre mice was perhaps surprising given that PCs are not reduced dramatically in mice given anti-CD20 (ref. 17). Because BAC transgene expression can extend beyond prescribed cell- and stage-restricted expression patterns and amounts18, we probed for Cre protein expression and evidence of P2rx4 mutation in mature BM PCs. As shown, the Cre protein was readily detected by flow cytometry in mature B220− BM PCs from CD20-TAM-Cre adults (Extended Data Fig. 7a). Likewise, using a genomic PCR assay we observed evidence for P2rx4 mutation among BM B220− PCs that remained in P2rx4fl/fl;CD20-TAM-Cre mice after pulsing with tamoxifen for only 7 days (Extended Data Fig. 7b). We conclude that CD20-TAM-Cre mice are a model for inducible mutagenesis in BM PCs, in addition to their known utility for studies of splenic naive B cells and germinal centre B cells16. Further, these results are consistent with the idea that P2RX4 is required by established BM PCs to persist in the BM.

To examine antigen-specific PCs, we fed P2rx4fl/fl;CD20-TAM-Cre and control P2rx4+/+;CD20-TAM-Cre mice with tamoxifen-laced chow for 4 weeks, then immunized these mice with the T cell-dependent hapten-carrier (4-hydroxy-3-nitrophenyl)acetyl-keyhole limpet haemocyanin (NP-KLH). We quantified NP-specific PCs in the spleen on days 7 and 30 and the BM at day 30 after immunization by ELISPOT. For P2rx4fl/fl;CD20-TAM-Cre mice numbers of NP-specific PCs were reduced substantially in the BM at day 30 but not the spleen at either day 7 (Fig. 2g) or day 30 (Extended Data Fig. 7c). Also, P2rx4fl/fl;CD20-TAM-Cre mice showed no reduction of NP-specific splenic germinal centre B cells (Extended Data Fig. 7d). We conclude that P2RX4 is required to establish BM PC populations to an exogenous antigen.

We next evaluated the impact of the P2RX4-specific inhibitor 5-BDBD. In short-term BM cultures, baseline numbers of antibody-secreting cells and the capacity of eATP to sustain PC numbers were both abolished by 5-BDBD but not by the P2RX7 inhibitor A438079 (Fig. 3a). Further, treatment of WT mice with 5-BDBD significantly reduced numbers of BM PCs including B220− CD138+ LLPCs (Fig. 3b and Extended Data Fig. 8a), without affecting numbers of steady-state splenic PCs or naive B cells or CD4+ and CD8+ splenic T cells in unimmunized mice (Extended Data Fig. 8b,c) or NP-specific germinal centre B cells in NP-KLH immunized mice (Extended Data Fig. 8d), or peritoneal cavity B cells in unimmunized mice (Extended Data Fig. 8e).

Fig. 3: P2RX4 inhibition causes loss of serum antibodies by depleting BM PCs.
figure 3

a, ELISPOT analyses for total IgG-secreting cells among B6 BM cells (106 cells per well) first cultured with or without 100 μM ATP and with or without 50 μM 5-BDBD or 50 μM A438079 versus dimethylsulfoxide alone for 2 days. b, B6 adult females were given four injections of 5-BDBD (4.25 mg kg−1) or dimethylsulfoxide intravenously (n = 6 per group) every 2 days and then analysed on day 8. Dump− IgD− gate shown in Fig. 1. c, 5-BDBD depletes pre-established serum antibodies. B6 and Panx3−/− females were immunized with NP-KLH/alum. Sixty days later (day 0) mice were given dimethylsulfoxide (dotted line) or 5-BDBD (solid line) five times over 10 days (n = 3 per group). Shown are optical density values for NP-binding serum IgG and IgM at each time point. d–g, NZB/W females were given dimethylsulfoxide (dotted line) or 5-BDBD (solid line) at 29–33 weeks of age (grey rectangles). Shown are anti-dsDNA titre (d), overall proteinuria (e) (see Methods for scoring), urine albumin (f), kidney pigmentation and histological images revealed with PAS staining (g). n = 6 per group (d,e); n = 3 (f,g). h,i, B6 adults previously given B6.H2-Ab1bm12 splenocytes (day 0) were given 5-BDBD (solid line) or dimethylsulfoxide (dotted line) (n = 3 per group) at 5–9 weeks after cell transfer (grey rectangles) and monitored for dsDNA-specific serum IgG (h) and proteinuria (i). Means are shown for all graphs. For a,b, *P = 0.0022, with P values derived from two-tailed Mann–Whitney tests, without any adjustment for multiple comparisons. For c–f,h,i, *P < 0.05, **P < 0.01, ***P < 0.001, derived from a repeated measures analysis of variance (ANOVA) (Methods). The experiments in a,b, were performed at least thrice and the experiments in c–e and g–i were performed twice. Scale bar, 100 μm (g).

Source Data

Full size image

We also examined the impact of 5-BDBD on serum antibody titre in WT mice that were previously immunized with NP-KLH. We delivered 5-BDBD to WT adults immunized with NP-KLH 60 days previously, using Panx3−/− mice as further controls. Serum NP-specific IgM and IgG titres were each severely reduced by 5-BDBD and remained reduced for at least 80 days after ending 5-BDBD treatments (Fig. 3c). As expected, hapten-specific antibodies were highly compromised across all time points in Panx3−/− mice, with no further effect of P2RX4 inhibition by 5-BDBD (Fig. 3c).

We evaluated two models of systemic lupus erythematosus characterized by serum antibodies against double-stranded DNA (dsDNA) and progressive proteinuria. NZB/W mice experience a spontaneous and progressive age-associated disease beginning at 20 weeks of age19. Treatment of NZB/W mice with 5-BDBD beginning at 29 weeks of age caused robust arrest in IgG anti-dsDNA titre and a precipitous drop in overall proteinuria (Fig. 3d,e). In NZB/W mice, 5-BDBD also arrested age-associated increases in urine albumin concentrations (Fig. 3f) and other disease indicators including kidney hypopigmentation and renal tubular hypertrophy with large hyaline casts (Fig. 3g). Also, we induced anti-dsDNA antibodies and proteinuria in B6 mice by transferring B6.H2-Ab1bm12 splenocytes20. Similar to our observations in NZB/W mice, 5-BDBD administration reduced pre-established proteinuria and arrested anti-dsDNA IgG titre when administered beginning 5 weeks after transfer of B6.H2-Ab1bm12 splenocytes (Fig. 3h,i). Consistent with the conclusion that P2RX4 inhibition severely depletes mature BM PCs, rebound kinetics of anti-dsDNA antibody serum titre and proteinuria were relatively slow compared to the rapid rebound observed in NZB/W mice given the mTOR inhibitor rapamycin which arrests antibody synthesis in mature BM PCs without causing their demise8 (Fig. 3d,e,h,i and Extended Data Fig. 9). We conclude that short-term P2RX4 inhibition causes loss of pathogenic antibody titre by depleting BM PCs.

Because P2RX receptors are non-selective channels for Ca2+ and other cations needed for optimal endoplasmic reticulum function21, we considered the possibility that eATP/P2RX4 signalling promotes endoplasmic reticulum homeostasis in BM PCs. Robust sustained antibody synthesis requires the endoplasmic reticulum sensor IRE1α, which activates the unfolded protein response (UPR) by splicing nascent RNA transcripts encoded by the Xbp1 gene to encode a functional transcription factor termed XPB1s (ref. 22). We witnessed increased concentrations of XPB1s and the additional UPR-affiliated transcriptional regulator ATF4 (ref. 22) in BM LLPCs 24 h after delivering a single dose of 5-BDBD to WT adults (Fig. 4a,b). Hyperacute endoplasmic reticulum stress can induce apoptosis by causing ATF4 to promote expression of the transcription factor CCAAT/enhancer-binding protein-homologous protein (CHOP)23 which then drives transcription of the pro-apoptotic BCL-2 family member BIM24. We therefore questioned whether P2RX4 inhibition causes BM PC depletion by unleashing the ATF4–CHOP pathway. To test this, we evaluated the impact of 5-BDBD on PCs in Chopfl/fl;CD20-TAM-Cre and control Chop+/+;CD20-TAM-Cre adults which were fed tamoxifen for 4 weeks before immunizing with NP-KLH. ELISPOT analysis (Fig. 4c,d) revealed that Chop (also known as Ddit3) mutation prevented 5-BDBD-mediated depletion of IgG-secreting BM LLPCs, including NP-specific BM PCs in mice that were immunized 4 weeks previously and 14 days after ending tamoxifen delivery. We confirmed that 4 weeks of tamoxifen exposure was sufficient to induce Chop mutation in B cells and BM PCs but not T cells of Chopfl/fl;CD20-TAM-Cre young adults (Extended Data Fig. 10). We conclude that P2RX4 inhibition causes an acute endoplasmic reticulum stress response that initiates a CHOP-dependent death pathway in BM PCs.

Fig. 4: P2RX4 inhibits endoplasmic reticulum-affiliated apoptosis in BM PCs.
figure 4

a,b, Intracellular XBP1s (a) and ATF4 (b) expression for BM PCs 24 h after treating mice once with 5-BDBD or dimethylsulfoxide alone (control) (n = 6 per group). c,d, WT;CD20-TAM-Cre and Chopfl/fl;CD20-TAM-Cre adults were fed tamoxifen-laced chow for 4 weeks before immunization with NP-KLH/alum. Four weeks after immunization, cohorts were given 5-BDBD (4.25 mg kg−1) or dimethylsulfoxide thrice intravenously for 7 days before analysis (n = 6 per group). c,d, Representative images (c) and quantification (d) of ELISPOT analyses of 106 total cells per well for total (left) and NP-specific BM PCs (right) in indicated mice. All bar graphs are means. Data in d are pooled from two experiments with a total of 6 per group. For a,b,d, *P = 0.0022, with P values derived from two-tailed Mann–Whitney tests without any adjustment for multiple comparisons. All experiments were performed at least twice.

Source Data

Full size image

Discussion

Our experiments show that establishing and maintaining optimal serum antibody titre requires activation of the purinergic receptor and cation channel P2RX4 on PCs. Our results further suggest that BM PCs including LLPCs use P2RX4 to avoid endoplasmic reticulum-regulated apoptosis by recognizing the homeostatic release of eATP by PANX3-expressing osteoblastic niche cells. Mutation of P2rx4 in developing B-lineage cells using mb-1-cre resulted in reduced serum antibody titre and greatly reduced numbers of BM PCs, indicating that P2RX4 is required to establish normal BM PC populations. Moreover, induced mutation of P2rx4 in B cells and PCs using CD20-TAM-Cre reduced serum antibody titre to a comparable degree and greatly depleted BM PCs, including the B220− fraction that is enriched for mature LLPCs. It is therefore likely that most BM-resident PCs, including LLPCs, depend on P2RX4. In each setting, other aspects of B cell development including induction of germinal centre B cells and splenic PCs were unaffected. Further, eATP release by osteoblasts and their capacity to support PC function in vitro was compromised by Panx3 mutation and the ability of eATP to support PCs in vitro was abrogated by arresting P2RX4 function with 5-BDBD. Thus, BM PCs sense eATP through P2RX4. Our results support the conclusion that PANX3-dependent release of eATP by osteoblastic BM cells and its detection by P2RX4 on PCs are essential events defining BM niches for PCs.

How mature PCs avoid apoptosis to sustain serum antibody titre remains poorly understood. A widely accepted model holds that mature PCs must localize to specialized BM survival niches orchestrated by the chemokine CXCL12 (ref. 25). However, the identity of all relevant niche-associated survival factors remains uncertain and PC lifespan may also be affected by more cell-intrinsic pathways26. One key cell-extrinsic factor is the cytokine APRIL, which delivers anti-apoptotic signals to PCs by binding to the surface receptor BCMA27,28. Indeed, BCMA-mutant mice possess only residual numbers of BM PCs in a manner similar to PANX3 mutants as well as mice harbouring P2RX4-mutant B-lineage cells29. However, BM PC numbers are unaffected by April mutation without concomitant depletion of the cytokine BAFF, an alternative BCMA ligand27. Because BAFF is available outside the BM including in serum30,31, PC access to BAFF and perhaps APRIL is unlikely to be restricted to BM PC niches. By contrast, our results point to PANX3-regulated eATP release by osteoblastic lineage cells and its interpretation by P2RX4 on PCs as a defining and unique feature of PC survival niches in the BM.

In the immune system, eATP is often viewed as a pro-inflammatory signal that opens P2RX channels such as P2RX7 to trigger or augment macrophage and memory T cell function in response to infection and tissue injury3,32. By contrast, our results reveal that PCs use P2RX4 to sense the regulated homeostatic release of eATP in non-inflammatory microenvironments in the BM. This idea is consistent with the observation that P2RX4 possesses a uniquely high binding affinity for eATP among P2RX receptors33 and other work proposing that pro-inflammatory events are actively suppressed in the BM as a strategy to protect specialized niches, including those harbouring mature PCs34,35. Thus, our findings suggest that P2RX receptors have evolved to interrogate eATP concentrations to regulate a diverse array of cell fates, including in the immune system.

Our work also identifies P2RX4 as a key upstream regulator of endoplasmic reticulum function, as short-term P2RX4 blockade caused rapid activation of the IRE1–XBP1 pathway and increased abundance of the extra UPR regulator ATF4 in BM PCs. An intimate relationship between P2RX4 and the endoplasmic reticulum is further indicated by the capacity of Chop mutation to rescue BM PCs experiencing P2RX4 inhibition, as CHOP is known to mediate apoptosis due to acute or unresolved endoplasmic reticulum stress36. In this regard, the lack of impact of P2RX4 blockade on antibody-secreting PCs in the spleen was perhaps surprising. One possible explanation for these data centres on previous results indicating that newborn PCs experience progressive increases in immunoglobulin synthesis which in turn causes stepwise increases in UPR activity37. Therefore, relatively immature PCs such as those in the spleen may be intrinsically less sensitive to P2RX4 inhibition. Given that small numbers of PCs can persist in the spleen for at least several months after induction without input from activated B cells38, it may be informative to learn if PCs that persist outside the BM for extended periods eventually become susceptible to P2RX4 blockade. Alternatively, PCs in spleen and BM may depend on different signals to regulate the UPR.

PCs play roles in a broad range of afflictions ranging from antibody-mediated autoimmunity and transplant rejection to multiple myeloma, an incurable malignancy of BM PCs. Our results demonstrate that the P2RX4 inhibitor 5-BDBD can be used to deplete BM PCs and substantially reduce pre-existing serum antibody titre, including for autoantibodies generated in two different mouse models of autoimmune disease. Treatment of previously immunized mice with 5-BDBD caused rapid and sustained loss in pre-established serum antibodies and delivery of 5-BDBD to disease-prone NZB/W mice reduced lupus-affiliated symptoms including urine albumin and overall proteinuria and prevented several more metrics of kidney disease. Current strategies for eliminating problematic PCs mainly target processes unique to PC biology survival and function. These include bispecific antibodies or engineered patient T cells targeting BCMA39,40 and proteasome inhibitors intended to induce a terminal endoplasmic reticulum stress response due to perturbed proteostasis in normal and malignant LLPCs41,42,43. Given that 5-BDBD increased intracellular abundance of XBP1s and ATF4 and caused CHOP-dependent LLPC depletion, we suggest that the eATP/P2RX4 pathway may prove useful as an alternative strategy to induce hyperacute endoplasmic reticulum stress and apoptosis in problematic LLPCs in autoimmunity and other settings. Ultimately, for humoral autoimmunity and allo-antibody producing cells such strategies will dovetail with approaches that target antigen-specific PCs44 to avoid depleting PCs producing beneficial antibodies. In addition, given the capacity of eATP to sustain numbers of antibody-secreting cells in culture, we speculate that the insights from the present study will illuminate pathways leading to new methods to increase protective antibody titre.

Methods

Mice

C57BL/6 (B6), B6.mb-1-cre (B6.C(Cg)-Cd79atm1(cre)Reth/EhobJ), B6.H2-Ab1bm12, NZB/W and Chop+/fl mice were obtained from the Jackson Laboratory (USA) and bred and/or maintained in our colonies at the University of Pennsylvania and/or the National Cancer Institute (NCI), USA. Panx3 KO mice were provided by Y. Yamada6 (National Institutes of Health (NIH), USA) and housed in our NCI colony. P2rx4fl/fl (P2rx4tm1c(EUCOMM)Wtsi) mice were obtained from the MRC Harwell Institute (UK) and rederived at the NCI. Human CD20 (hCD20) transgenic mice containing an internal ribosomal entry site (IRES)-ERT2Cre in the 3′-untranslated region (UTR) of the gene encoding hCD20 (hCD20-TAM-Cre)16 were provided by M. J. Shlomchik (University of Pittsburgh). All mice were used at 8–12 weeks of age unless otherwise indicated.

Immunizations and treatments

Mice were immunized intraperitoneally with 50 μg of NP18-KLH in alum. To induce Cre activity, adults were fed tamoxifen-laced chow (ENVIGO) for 4 weeks before initiating experiments. Mice were treated with 5-BDBD (4.25 mg kg−1 intravenously (i.v.); R&D), suramin (2.0 mg kg−1 i.v.; Abcam) or PPADS (2.5 mg kg−1 i.v.; Santacruz)45 in a volume of 30 μl of 5-BDBD and the P2RX7 inhibitor A438079 (R&D Systems) were each used in vitro at a final concentration of 50 μM. All P2RX inhibitors were dissolved initially in dimethylsulfoxide.

Flow cytometry

BM cells were flushed from femurs, tibiae and ilia using a 27-gauge needle and spleens were mechanically disrupted using frosted microscope slides. BM cells and splenocytes were stained and analysed in FACS buffer (PBS containing 0.1% sodium azide, 1 mM EDTA and 0.5% BSA). Cells were counted using a Cellaca MX cell counter (Nexcelom Biosciences). Antibodies specific for CD4 (RM4-4), CD8α (53-6.7), TER-119 (TER-119), F4/80 (BM8), B220 (RA3-6B2), CD19 (1D3), CD138 (300506), CD38 (90), IgD (11-26c), GL7 (GL-7), AA4.1 (R3), IgM (II/41), CD43 (84-3C1), TCRβ (H57-597), CD25 (PC61.5), CD44 (IM7), CD62L (MEL-14), CD11b (M1/70) and Streptavidin APC-eFluor 780 were from eBioscience. ATF4 (CL594-60035) was from Proteintech. XBP1s (Q3-695), CD5 (53-7.3) and CD43 (S7) were from Becton Dickinson. All Dump channels consisted of antibodies to CD4, CD8, TER-119 and F4/80. Non-viable cells were identified using the violet-excited Zombie aqua fixable dye (BioLegend). To evaluate intracellular Cre expression, surface-stained cells were fixed for 5 min with 10% formalin, washed, then incubated for 30 min in BD Phosflow Perm Buffer III (BD Biosciences). Cells were washed and then stained for 30 min with a PE-conjugated rabbit anti-Cre (Cell Signaling Technology). Samples were acquired using a LSR Fortessa or Symphony A3 lite flow cytometer (Becton Dickinson) using FACSDiva v.9.1 and analysed using FlowJo software (Tree Star). All analyses were first gated on viable singlet cells.

Osteoblast cultures

Calvaria were digested six times for 10 min by 0.1% collagenase type 1 (Worthington Biochemical) and 0.2% dispase II (Roche) in PBS. The last two fractions were collected in culture medium consisting of α-minimum essential medium (Invitrogen) with 10% FBS, 100 U ml−1 of penicillin and 100 µg ml−1 of streptomycin. For the coculture assay, primary calvarial cells were induced by the addition of 50 µg ml−1 of ascorbic acid (Sigma-Aldrich) and 5 mM β-glycerophosphate (Sigma-Aldrich) for 14 days before coculture experiments.

ELISPOT and ELISA

ELISPOT assay was determined by mouse IgG ELISpot kit (Mabtech). In brief, plates (Millipore) were coated with NP4-BSA, anti–H + L Ig (all 15 μg ml−1) or dsDNA (10 μg ml−1) in sodium bicarbonate buffer. Unless indicated otherwise, for ELISPOT analyses we plated 106 cells per well unless cells were taken from earlier cultures that were also initiated with 106 cells per well. Plates were blocked overnight with RPMI with 10% FBS. Plates were washed using PBS with 0.1% Tween 20, followed by incubation with biotinylated anti-IgG (Mabtech). Biotinylated antibodies were revealed using streptavidin-HRP (Mabtech) and spots visualized with TMB (Mabtech). Wells were imaged using ImmunoSpot software (Cellular Technology) and manually counted using Fiji (ImageJ) software. Background wells used no cells and contained 0 or 1 spots. ELISA for total serum IgG, IgM and IgA was performed with ELISA kit (Thermo) according to the manufacturer’s instructions. In brief, diluted sera were applied to antigen-coated 96-well plates. Primary antibodies were detected with horseradish peroxidase-conjugated anti-IgG, anti-IgM or anti-IgA secondary antibodies (Thermo), followed by incubation with TMB substrate. ELISA analyses for NP-specific serum antibodies were performed with NP26-BSA (Biosearch Technologies) coated plates and ELISAs for dsDNA antibody titre were performed with a mouse anti-DNA ELISA Kit (Abcam).

Urine protein and albumin analyses

Proteinuria in NZB/W and B6.H2-Ab1bm12 mice was measured using Albustix Urinalysis Test Strips (Siemens) according to the manufacturer’s instructions. Briefly, fresh urine was applied to reagent strips and the colorimetric change determined using the following scale: 1, ≥30 mg per dl protein; 2, ≥100 mg dl−1; 3, ≥300 mg dl−1; and 4, ≥2,000 mg dl−1. Urine albumin concentrations were determined with a mouse albumin ELISA Kit (Abcam) according to the manufacturer’s instructions.

Histological analyses

Kidney samples were fixed in 4% paraformaldehyde and embedded in paraffin wax. After deparaffinization and rehydration, sections were stained with haematoxylin and eosin or periodic acid-Schiff (PAS) reagent with the assistance of the Penn Vet Comparative Pathology Core.

ATP flux

SATP flux was examined by luminometry, as previously described5,46. Primary calvarial cells isolated from Panx3 KO mice were seeded at 104 cells per well in a 96-well plate and induced by osteogenic culture media for 14 days. The cells were then washed with PBS, followed by incubation in PBS for 2 min. The supernatant was collected and assayed with luciferase and luciferin (Promega). The luminescence was measured using an Infinite 200 PRO (TECAN).

Statistics

Statistical differences between two groups were evaluated using the two-tailed non-parametric Mann–Whitney U-test. For analyses involving several time points per group, a repeated measures ANOVA was used with correction for many comparisons using the original method of Benjamini and Hochberg, with a desired false discovery rate of 0.05. All statistics were performed in Prism software (v.9.51, GraphPad).

Bioinformatics

Publicly available BM single-cell RNA-sequencing data14 were downloaded in h5ad format from the Chan–Zuckerberg CellxGene website and loaded into python 3.9 using SCANPY (v.1.9.3)47. Batch correction was performed by selecting 2,000 highly variable genes48, taking the different sequencing assays into account and fed into scVI (single-cell variational inference)49, with n_layers set to 2, n_latent set to 30 and gene likelihood set to negative binomial. Latent dimensions were used to identify clusters using the Leiden algorithm50. B cell lineage cells were isolated on the basis of their cluster identity and published annotation. This subset was reanalysed in the same fashion, followed by gene expression data normalization, log transformation. A UMAP manifold was computed on the latent dimensions obtained from scVI.

Study approval

All animal studies described herein were reviewed and approved by the National Cancer Institute ACUC or the University of Pennsylvania IACUC. Animals were bred and maintained in accordance with institutional guidelines for animal welfare.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.