Blimp-1 Controls Plasma Cell Function Through the Regulation of Immunoglobulin Secretion and the Unfolded Protein Response

Home , ATF4, ATF6, BCL6, IRF4, PAX5, PRDM1

A rt i c l e s

Blimp-1 controls plasma cell function through the regulation of immunoglobulin secretion and the unfolded protein response

Julie Tellier1,2, Wei Shi1,3, Martina Minnich4, Yang Liao1,2, Simon Crawford5, Gordon K Smyth1,6, Axel Kallies1,2, Meinrad Busslinger4 & Stephen L Nutt1,2

Plasma cell differentiation requires silencing of B cell transcription, while it establishes antibody-secretory function and long-term survival. The transcription factors Blimp-1 and IRF4 are essential for the generation of plasma cells; however, their function in mature plasma cells has remained elusive. We found that while IRF4 was essential for the survival of plasma cells, Blimp-1 was dispensable for this. Blimp-1-deficient plasma cells retained their transcriptional identity but lost the ability to secrete antibody. Blimp-1 regulated many components of the unfolded protein response (UPR), including XBP-1 and ATF6. The overlap in the functions of Blimp-1 and XBP-1 was restricted to that response, with Blimp-1 uniquely regulating activity of the kinase mTOR and the size of plasma cells. Thus, Blimp-1 was required for the unique physiological ability of plasma cells that enables the secretion of protective antibody.

The production of antibodies is an essential arm of the immune by the following triad of transcription factors: IRF4 (encoded by Irf4), response that provides both immediate protection against a current Blimp-1 (encoded by Prdm1) and XBP-1 (encoded by Xbp1). IRF4 not infection and long-term immunity to re-exposure to the same pathogen. only has high expression in PCs but also is essential for PC develop- The antibody-secreting cell compartment consists of short-lived ment, at least in part due to its regulation of Prdm1 (refs. 7–9). Blimp-1 proliferating plasmablasts (PBs), which are generated early in an is expressed in all antibody-secreting cells and is also required for immune response, and long-lived post-mitotic plasma cells (PCs), their differentiation beyond an early point10–12. Blimp-1 has so far which reside in specialized niches in the bone marrow (BM)1,2. These been thought of as a transcriptional repressor that silences several Nature America, Inc. All rights reserved. America, Inc. © 201 6 Nature long-lived PCs have been shown to maintain high titers of protective important B cell genes, including Pax5 (ref. 13), Myc14, Ciita15, Bcl6, antibody for decades after exposure to pathogen or immunization3. Spib and Id3 (ref. 16), although there is only limited understanding Thus, understanding the factors that control the production, func- of its targets in PCs17. XBP-1, an important component of the UPR, 18 npg tion and long-term survival of PCs is critical for both improved was initially proposed to be essential for PC formation ; however, vaccine design and the development of novel approaches for targeting subsequent evidence has suggested that XBP-1 is required more spe- pathogenic PCs in diseases such as multiple myeloma and systemic cifically for immunoglobulin production19–22. As a consequence of lupus erythematosus. the important functions of IRF4 and Blimp-1 early in the differentia- To achieve the dual goals of maintaining an extremely high rate tion process, there is little current knowledge about the function of of immunoglobulin secretion while ensuring their own long-term these factors in long-lived PCs23,24. Here we used a genetic approach survival, PCs show a highly specialized morphology with an enlarged to investigate the functional consequences of the loss of IRF4, Blimp-1 cytoplasm and a tightly arranged endoplasmic reticulum (ER). PCs or XBP-1 in mature post-mitotic BM PCs. also constitutively activate the unfolded protein response (UPR), a specialized sensing mechanism for detecting and dealing with large RESULTS amounts of protein that pass through the ER4. Inactivation of IRF4 and Blimp-1 in PCs The differentiation of activated B cells into PCs requires coordi- To assess the importance of IRF4 and Blimp-1 in mature BM PCs, we nated changes in the expression of many hundreds of genes, including crossed mice carrying loxP-flanked alleles encoding either transcrip- the silencing of B cell–associated transcripts, such as those encoding tion factor (Irf4fl/fl or Prdm1fl/fl)25,26 with Rosa26-CreERT2 mice (which the transcription factors Pax5, Bach2 and Bcl-6, and the activation of have tamoxifen-inducible expression of Cre recombinase from the a suite of PC-specific genes5,6. This developmental program is guided ubiquitous Rosa26 locus)27. This system allowed tamoxifen-inducible

1The Walter and Eliza Hall Institute of Medical Research, Parkville, Australia. 2Department of Medical Biology, The University of Melbourne, Parkville, Australia. 3Department of Computing and Information Systems, The University of Melbourne, Parkville, Australia. 4Research Institute of Molecular Pathology, Vienna Biocenter, Vienna, Austria. 5School of BioSciences, The University of Melbourne, Parkville, Australia. 6Department of Mathematics and Statistics, The University of Melbourne, Parkville, Australia. Correspondence should be addressed to S.L.N. ([email protected]).

Received 21 July 2015; accepted 2 November 2015; published online 18 January 2016; doi:10.1038/ni.3348

nature immunology VOLUME 17 NUMBER 3 MARCH 2016 323 A rt i c l e s

Figure 1 Inactivation of Irf4 and Prdm1 a fl/+ +/gfp b +/gfp in PCs. (a) Frequency of Irf4fl/+CreERT2 Irf4 Prdm1 Prdm1 Irf4fl/– Prdm1fl/gfp fl/gfp (Irf4fl/+) or Irf4fl/−CreERT2 (Irf4fl/−) PCs, 0.0005 0.4 ** 0.4 Prdm1 + lo + 0.0004 identified as CD138 B220 GFP (left), * ** 0.3 0.3 ** +/gfp ERT2 +/gfp or Prdm1 Cre (Prdm1 ) or 0.0003 * fl/gfp ERT2 fl/gfp 0.2 0.2 Prdm1 Cre (Prdm1 ) PCs, PCs (%) + 0.0002 PCs (%) + + PCs (%) identified as CD138 Blimp-1–GFP (right), 0.0001 0.1 0.1 among total BM cells from Rag1−/− host GFP 0 0 0 mice given B cells of the those genotypes, 72 14–21 7–14 21–28 35–49 7–14 21–28–35 56–63 assessed at various times (horizontal Time after treatment (d) Time after treatment (d) Time after treatment (d) axes) after treatment of host mice with tamoxifen to induce inactivation of Irf4 (reported by GFP expression) or Prdm1 (experimental plan, Supplementary Fig. 1a). (b) Frequency of BM PCs from Prdm1+/gfpCreERT2 or Prdm1fl/gfpCreERT2 mice various days (horizontal axis) after treatment with tamoxifen (experimental plan, Supplementary Fig. 1b). Each symbol (a,b) represents an individual mouse; small horizontal lines indicate the mean. *P < 0.05 and **P < 0.005 (paired t-test). Data are pooled from three experiments.

inactivation of Irf4 or Prdm1 in pre-existing PCs. To facilitate the the mice with tamoxifen. Analysis of the frequency of Prdm1 tran- tracking of PCs, Prdm1fl/fl and Prdm1+/+ control mice also carried scripts spanning exons 5–6, which are removed by Cre-mediated a Prdm1gfp reporter allele that expresses green fluorescent protein excision of the loxP-flanked exon 5, revealed an 87% reduction in fre- (GFP) but no functional Blimp-1 (Prdm1fl/gfpCreERT2 or Prdm1+/gfp quency of full-length transcripts in Prdm1fl/gfpCreERT2 cells, compared CreERT2)11, while Irf4fl/fl mice carried an internal GFP-encoding cas- with their abundance in Prdm1+/gfpCreERT2 cells (Supplementary sette that reports gene inactivation26. We transferred B cells from Fig. 2c). 465 genes were expressed differentially in Blimp-1-deficient those mice into B cell– and T cell–deficient Rag1−/− mice to generate (Prdm1fl/gfpCreERT2) PCs relative to their expression in control a large population of PCs and induced Cre activity by administration (Prdm1+/gfpCreERT2) PCs (Fig. 2a), with 170 genes activated by Blimp-1 of tamoxifen to the host mice 14 d later. Although inactivation of and 295 genes repressed by Blimp-1 (normalized average expression Irf4 occurred equivalently in Irf4fl/+CreERT2 and Irf4fl/−CreERT2 B cells of ≥4 RPKM (reads per kilobase of exon model per million mapped (approximately 15% GFP+ cells for each genotype; data not shown), reads) in at least one sample; false-discovery rate (FDR), ≤0.05; GFP+ PCs were lost from the BM as early as 2 d after treatment with Supplementary Table 1). Cross-referencing of those differentially tamoxifen (Fig. 1a and Supplementary Fig. 1a), which demonstrated expressed genes with their binding of Blimp-1, as assessed by chroma- that IRF4 was indispensable for PC survival. In contrast, inactivation tin immunoprecipitation followed by deep sequencing with a biotin- of Prdm1 by an identical strategy resulted in a PC population was that tagged Prdm1 allele29 in PBs generated in vitro, revealed that 28% of stable for many weeks after treatment with tamoxifen (Fig. 1a and genes activated by Blimp-1 (47 of 170) and 41% of genes repressed by Supplementary Fig. 1a). To confirm that result, we induced inactiva- Blimp-1 (120 of 295) were bound by Blimp-1 and were thus potential tion of Prdm1 in naive Prdm1fl/gfpCreERT2 mice and again found that BM direct targets of Blimp-1 (Supplementary Table 1). Partitioning of the PCs persisted without Blimp-1 (Fig. 1b and Supplementary Fig. 1b). proteins encoded by the differentially expressed genes into functional We observed a similar difference in the dependence of PC survival categories showed that the largest group of genes activated by Blimp-1 on IRF4 and Blimp-1 in the spleen (Supplementary Fig. 1a–d). encoded proteins involved in metabolism and nutrient transporters, Although both models showed a significant reduction in PC numbers whereas many genes repressed by Blimp-1 encoded receptors and Nature America, Inc. All rights reserved. America, Inc. © 201 6 Nature at the latest time points after inactivation of Prdm1, the data derived signaling molecules (Fig. 2b). This analysis also suggested a role for from the two genotypes were not strictly comparable, as the Prdm1+/gfp Blimp-1 in inhibiting B cell function, including antigen presentation B cells continued to produce new PCs throughout the course of the (through its effect on genes encoding major histocompatibility com- fl/gfp npg experiment, while the Prdm1 cells lacked this capacity (Fig. plex class II molecules), pathogen recognition (through its effect on 1a,b and Supplementary Fig. 1c). Consistent with that conclusion, Tlr9) and signaling through the activating receptors (through its effect we observed no significant change in the proportion of spleen or on Cd79b and Bank1). Blimp-1 also appeared to be important for the BM cells in tamoxifen-treated Prdm1fl/gfp mice that displayed a PC transcription of genes encoding products related to metabolic activity phenotype, either at steady state or after immunization with a pro- of the PCs, most probably related to their immunoglobulin-secreting tein antigen in alum, over the time frame examined (Fig. 1a,b and function (through its effect on Tpst2, Syvn1 and Fkbp11). Notably, Supplementary Fig. 1c–e). We assessed the efficiency of the ablation very few genes encoding products involved in the cell cycle or sur- of Prdm1 by transferring Prdm1fl/gfpCreERT2 or Prdm1+/gfpCreERT2 vival were affected by the loss of Blimp-1. For example, Myc, a known B cells into Rag1−/− host mice and treating the recipients with target of Blimp-1-mediated repression that encodes a central regula- tamoxifen 2 d after transfer, before PCs could be formed. This tor of proliferation in B cells that is normally repressed in PCs14, was approach completely blocked PC differentiation (Supplementary not re-expressed in Blimp-1-deficient PCs. The expression of genes Figs. 1f and 2a), which mimicked the conditional removal of Prdm1 encoding products essential to PC survival, such as MCL1 (encoded by from activated B cells in vitro12,28. PCR genotyping of purified B cells Mcl1)30, Bim (encoded by Bcl2l11) or BCMA (encoded by Tnfrsf17)31, and PCs confirmed efficient inactivation of the Prdm1 locus in both remained similarly unchanged (Fig. 2c and data not shown). cell types (Supplementary Fig. 2b). These data demonstrated that As Blimp-1 has been linked to silencing of the B cell transcrip- Blimp-1 was not essential for the long-term survival of BM PCs. tional program after PC differentiation16, we investigated whether the BM PCs reverted to the B cell stage without Blimp-1. Several Blimp-1-regulated genes in PCs B cell–associated genes, including Cd22, Spib, Cd79b and Ciita, which The long-term persistence of Blimp-1-deficient PCs enabled us to are usually silenced in PCs, were re-expressed in PCs in the absence of investigate the effect of the loss of Blimp-1 on the transcriptome of Blimp-1 (Fig. 2c). Most notably, Pax5 and Bcl6, which encode two of long-lived PCs. We sequenced RNA from BM PCs purified from the main regulators of the B cell transcriptome and are known targets Prdm1fl/gfpCreERT2 and Prdm1+/gfpCreERT2 mice 21 d after treating of Blimp-1 repression13,16 were not re-expressed in Blimp-1-deficient

324 VOLUME 17 NUMBER 3 MARCH 2016 nature immunology A rt i c l e s

Figure 2 Transcriptional analysis of Blimp-1- a b c deficient PCs. (a–c) Whole-genome RNA- 60 56 49 50 sequencing analysis of PCs sorted from the 50 Fo B +/gfp PC fl/gfp PC +/gfp ERT2 fl/gfp BM of Cre (+/gfp) or 40 Prdm1 Prdm1 Repressed CreERT2 (fl/gfp) mice 21 d after treatment 40 5 30 with tamoxifen. (a) Scatterplot of differential 23 Genes

fold) 22 expression, showing genes with significantly 2 20 15 14 increased expression (blue) or decreased 11 Tnfrsf17 0 10 8 expression (red) in the absence of Blimp-1 4 3

Expression 0 (FDR <0.05; normalized average expression of 0 4 3 2 ≥4 RPKM in at least one sample). (b) Functional 10 7 fl/gfp vs +/gfp (log –5 PC

11 Activated Hspa5 classification and quantification of proteins signature 20 17 Xbp1 0 5 10 18 encoded by target genes repressed (blue) or 2223 23 activated (red) by Blimp-1 in PCs. Numbers Mean expression Genes 30 (log2 RPKM) Ern1 40 Ell2 above or below bars indicate total number 40 of genes in each category. (c) Expression of 50

published gene signatures for wild type follicular Other d Survival B cell Receptor Secreted Cd93 Signaling Trafficking Cell cycle Unknown 6 6 +/gfp Metabolism Transporter B cells (FoB) and BM PCs from Prdm1 +/gfp PC Transcription Cytoskeleton CreERT2 mice (+/gfp PC) or Prdm1fl/gfpCreERT2 fl/gfp PC Edem3 mice (fl/gfp PC); right margin, positions of some 100 genes of interest. (d) Flow cytometry analyzing 80 expression of the Blimp-1-regulated surface 60 40 molecules major histocompatibility class II Cd22 +/gfp 20 Ciita

(MHCII), CD22 and CD93 on Prdm1 Events (% of max) 0 Cd19 ERT2 hi hi B cell Pax5 Cre mature B cells (B220 CD19 ) and 0 103 104 105 signature +/gfp ERT2 fl/gfp ERT2 Prdm1 Cre or Prdm1 Cre PCs MHCII CD22 CD93 Bcl11a (CD138+Blimp-1-GFP+) obtained from the BM Id3 e H2–Aa of mice 21 d after treatment with tamoxifen. Blimp-1 PC signature Spib activated Cd79b (e) Overlap and differences in Blimp-1-activated target genes (Supplementary Table 1) and 6 the PC gene signature . Data are from two 37133 264 Row z-score experiments with samples pooled from multiple mice (a–c,e) or are representative of at least –1–0.5 0 0.5 1 three experiments (d).

PCs (Fig. 2c and data not shown), which demonstrated that the PCs independent measure of secretory activity of the PCs, we measured were not reverting to the more developmentally immature B cell fate cell-surface exposure of the lysosome-associated protein CD107a. in the absence of Blimp-1. The absence of the expression of Pax5 and Blimp-1-deficient BM PCs had much less staining of CD107a than Bcl6 might also explain the only partial reactivation of B cell genes, did their Blimp-1-sufficient counterparts (Fig. 3c), again indicative such as those encoding the lectin CD22 and major histocompatibility of impaired fusion of lysosomes with the plasma membrane. Thus, Nature America, Inc. All rights reserved. America, Inc. © 201 6 Nature complex class II molecules, in Blimp-1-deficient PCs, compared Blimp-1 was needed to maintain the characteristic PC morphology with their expression in wild-type B cells (Fig. 2c,d). Blimp-1 was and cytoplasmic organization. also required for the normal expression of some PC genes, including

npg the gene encoding the C-type lectin transmembrane receptor CD93 Control of immunoglobulin secretion by Blimp-1 (Fig. 2c,d). To extend our analysis, we compared the genes activated As the fundamental function of PCs is to produce immunoglobulin, by Blimp-1 with a published signature of 301 PC genes6 and found we assessed the secretory capacity of Blimp-1-sufficient and Blimp- that 88% of the signature genes (264 of 301) were expressed independ- 1-deficient PCs by enzyme-linked immunospot assay. We treated ently of Blimp-1 (Fig. 2e). Combined with the results of published Prdm1fl/gfpCreERT2 and Prdm1+/gfpCreERT2 mice with tamoxifen, then studies10–12,16,20,29, these results demonstrated that although Blimp-1 purified splenic and BM PCs from the mice 14–28 d later and assessed was essential for the establishment of the full PC gene-expression their secretion of immunoglobulin M (IgM) and all IgG. The secretion program, once formed, PCs maintained their unique transcriptome of both immunoglobulin isotypes was reduced in the spleen and, most largely independently of Blimp-1. strikingly, in the BM in the absence of Blimp-1 (Fig. 4a). To determine if this defect resulted from a decrease in the transcription of Control of PC size and ultrastructure by Blimp-1 immunoglobulin-encoding genes, we analyzed our RNA-sequencing During our analysis of survival kinetics, we noticed that the PCs in data and found that with the notable exception of those encoding IgM which Blimp-1 was deleted were smaller than and had less granularity and IgG3, immunoglobulin-encoding transcripts were not affected than that of Blimp-1-sufficient PCs and had a continuum of reduced by the loss of Blimp-1 (Fig. 4b). expression of the PC marker CD138 and Blimp-1–GFP relative to that The immunoglobulin heavy-chain isotypes are produced in two dif- of Blimp-1-sufficient PCs (Fig. 3a). Transmission electron microscopy ferent isoforms that differ in their use of polyadenylation sites32. The of cellular ultrastructure also revealed that Blimp-1-deficient BM PCs longer membrane-bound receptor form has high expression in B cells, had severe disruption of their distinctive dense ER (Fig. 3b), a finding while the shorter secreted form predominates in PCs (Supplementary substantiated by staining with a fluorescent dye specific for the ER Fig. 3a). Closer inspection of the expression of exons encoding the (Fig. 3c). In contrast, staining for secretory granules was increased in immunoglobulin heavy-chain constant regions revealed an increase Blimp-1-deficient PCs relative to this staining in Blimp-1-sufficient in the expression of transcripts encoding the transmembrane form in PCs (Fig. 3c), suggestive of impaired lysosomal trafficking. As an Blimp-1-deficient PCs compared with that in Blimp-1-sufficient PCs,

nature immunology VOLUME 17 NUMBER 3 MARCH 2016 325 A rt i c l e s

a b +/gfp fl/gfp c PC gated

8 5 3 +/gfp fl/gfp

fl/gfp ) ) 3 * 2

0.24 4 ) 6 3 2 3 5 250K ** 10 4 200K 104 2 *** 150K 1 +/gfp 3 10 0.25 100K 2 1 CD107a MFI ( × 10 LysoTracker MFI ( × 10 0 50K MitoTracker MFI ( × 10 CD138 SSC 0 3 4 5 0 10 10 10 0 0 0 0 50K GFP 100K150K200K250K FSC

Figure 3 Blimp-1 controls the size and morphology of PCs. (a) Flow cytometry of BM cells (left) and gated CD138+Blimp-1–GFP+ PCs (middle and right) obtained from Prdm1+/gfpCreERT2 and Prdm1fl/gfpCreERT2 mice 35 d after treatment with tamoxifen. SSC, side scatter; FSC, forward scatter. Numbers adjacent to outlined areas (left) indicate percent CD138+Blimp-1–GFP+ PCs. (b) Electron microscopy of PCs isolated from the BM of Prdm1+/gfpCreERT2 and Prdm1fl/gfpCreERT2 mice 21 d after treatment with tamoxifen. Scale bars, 2 µm (top row) or 5 µm (bottom row). (c) Mean fluorescence intensity (MFI) of the organelle-specific dye MitoTracker (left) or LysoTracker (middle) and of surface CD107a (right) of PCs obtained from the BM of Prdm1+/gfpCreERT2 and Prdm1fl/gfpCreERT2 mice 35 d after treatment with tamoxifen. *P < 0.05, **P < 0.01 and ***P < 0.005 (paired t-test). Data are representative of five experiments (a) or two experiments (b) or are from three experiments (c; mean and s.d.).

which thus lowered the ratio of transcripts encoding the secreted form activating Atf6, which encodes a known inducer of Xbp1 (ref. 36), and to those encoding the membrane-specific form in Blimp-1-deficient also promoting formation of the active spliced form of XBP-1 through PCs compared with that observed in Blimp-1-sufficient PCs (Fig. 4c,d). the direct regulation of Ern1 (which encodes IRE1, another compo- We then assessed protein expression by intracellular staining and nent of the UPR)37 (Fig. 5b). The requirement for Blimp-1 for the full flow cytometry. The proportion of PCs expressing IgG or IgA or the expression of the key regulators of the UPR ‘translated’ into significant immunoglobulin κ- or λ-chain was similar in Blimp-1-deficient PCs down-modulation of the expression of the majority of genes encod- and Blimp-1-sufficient PCs (Fig. 4e and data not shown). In line with ing the downstream components of the pathway (P = 0.001; Fig. 5c, the reduced expression of transcripts encoding IgM in the absence Supplementary Table 2 and Supplementary Fig. 4a). Notably, the of Blimp-1, the frequency of IgM+ PCs was substantially diminished function of Blimp-1 in the UPR was much broader than the regula- (Fig. 4e). Intra- and extracellular flow cytometry staining of IgA and tion of genes encoding the key transcription factors of the UPR, as IgM confirmed that whereas expression of total protein was decreased Blimp-1 directly bound to 38% of UPR genes expressed in PCs (45 of in the absence of Blimp-1, expression of the membrane-specific form 119 genes), encoding products that represented all branches of the was maintained (Fig. 4f). The elongation factor Ell2 has been shown pathway (Supplementary Table 2). Together these data demonstrated to have a major role in this alternative polyadenylation process33,34. that Blimp-1 as a central transcription factor of the UPR in PCs. Ell2 was a direct target of Blimp-1 (Supplementary Table 1), and its To investigate the overlap between Blimp-1 and XBP-1 in controlling Nature America, Inc. All rights reserved. America, Inc. © 201 6 Nature expression was reduced in Blimp-1-deficient PCs relative to that in the UPR and the PC transcriptome more generally, we generated mice Blimp-1-sufficient PCs (Fig. 4g). These data suggested that Blimp-1 with inducible deletion of Xbp1 through use of the CreERT2 system contributed to the efficient transition from a membrane-specific form (Supplementary Fig. 2d). As reported19,21,22, inactivation of Xbp1 had

npg of the immunoglobulin heavy chain to a secreted form in PCs, at least no effect on the size of the PC population (Fig. 6a). RNA-sequencing in part through direct control of Ell2 expression. analysis revealed that 632 genes were expressed differentially in XBP-1-deficient PCs versus XBP-1-sufficient PCs (Fig. 6b and Regulation of the UPR by Blimp-1 Supplementary Table 3) and, notably, Prdm1 was not down-modulated The partial re-expression of the membrane-specific immunoglobulin but instead was slightly upregulated in the absence of XBP-1 (with a heavy-chain at the expense of the secreted isoform in Blimp-1-deficient change in expression of 1.680-fold; FDR = 0.047). The comparison PCs could explain only in part the secretion defect observed of genes regulated by Blimp-1 or XBP-1 showed only a modest over- after loss of Blimp-1. Therefore, we assessed expression of the main lap of 56 genes (31 activated and 25 inhibited; approximately 10% of regulators of the UPR that control protein synthesis, folding and total genes regulated by either factor) (Supplementary Fig. 4b), which post-translational modifications, and expansion of the secretory meant it was unlikely that the Blimp-1-mediated control of PC function apparatus (Supplementary Fig. 3b). Notably, expression of XBP-1, was achieved predominantly through regulation of XBP-1, a conclu- ATF4 and ATF6, the three main transcription factors known to imple- sion in agreement with published studies of PCs derived in vitro20. Of ment this pathway4,35, was significantly downregulated at the level of note, XBP-1 deficiency in BM PCs resulted in a global decrease in all mRNA and protein in BM PCs after inactivation of Prdm1 (Fig. 5a and transcripts encoding the immunoglobulin heavy-chain complex (Igh) Supplementary Table 2). Wes generated PBs in vitro from the prog- (Fig. 6c). Reduced immunoglobulin expression was confirmed at the eny of mice carrying sequence encoding a biotin acceptor motif (Bio) protein level (Fig. 6d), and this correlated with reduced secretory inserted into the endogenous Prdm1 locus crossed with mice express- function, as measured by expression of CD107a (Fig. 6e) and the size ing the Escherichia coli biotin ligase BirA (Prdm1Bio/BioRosa26BirA/BirA of enzyme-linked immunospots for IgA and IgM (Fig. 6f and data mice)29 and analyzed the cells by chromatin immunoprecipita- not shown). Overall analysis of the UPR pathway revealed significant tion followed by deep sequencing; this identified Blimp-1-binding down-modulation (P = 0.0005; Fig. 6g) similar to that observed without sites in Atf6 but not in Atf4 or Xbp1 (Fig. 5b and data not shown). Blimp-1 (Fig. 5c); however, in contrast to the activity of Blimp-1, the The regulation of Xbp1 by Blimp-1 is probably two-pronged, with Blimp-1 activity of XBP-1 was focused on particular processes of the pathway,

326 VOLUME 17 NUMBER 3 MARCH 2016 nature immunology A rt i c l e s

a b c ) 3 +/gfp fl/gfp +/gfp 600 +/gfp 500 +/gfp ×4 300 200 fl/gfp fl/gfp 400 fl/gfp Spleen 400 ×6 150 300 ×17 200 * 100 200 ×6 ×3 200 Sec/MB ASCs per ASCs per + 100 + 400 cells 400 cells 100 50 ×3 ×10 IgG BM IgM *** *** 0 0

0 0 Expression (RPKM × 10 IgA IgE IgD IgM IgA IgE IgM Spleen BM Spleen BM IgG1 IgG3 IgG1 IgG3 IgG2c IgG2b IgG2cIgG2b λ-chainκ-chain

100 d e +/gfp fI/gfp fI/gfp 20 250 43 10.3 80 +/gfp GC B 60 5,500 4,000 *** 40 +/gfp PC cells among PCs (%) + 5,500 1,000 20 IgM

fI/gfp PC Ighm (RPM)

IgM 0 Ighg2b (RPM) MB Spleen BM sec 105 4.6 10.6 25 104 20 Figure 4 Blimp-1 controls immunoglobulin production. (a) Enzyme-linked 3 10 15 immunospot assay of the secretion of IgM and total IgG from PCs isolated 0 10 +/gfp ERT2 fl/gfp ERT2 cells among PCs (%) IgG

from Prdm1 Cre and Prdm1 Cre mice 14 d after treatment + 5 with tamoxifen, presented as images of wells from an assay with antibody to 3 4 5 0 10 10 10 IgG + + GFP 0 IgM (left) and as quantification of IgM ASCs (middle) or IgG ASCs (right). Spleen BM (b–d) RNA-sequencing analysis of Igh transcripts in Prdm1+/gfpCreERT2 and Prdm1fl/gfpCreERT2 BM PCs. (b) Normalized expression (in RPKM) of constant f +/gfp g +/gfp 120 fl/gfp 25 15 fl/gfp region–encoding transcripts from the locus encoding the immunoglobulin 250 ) ) ) 3 heavy-chain complex and the loci encoding the - or -chain complex 90 3 20 3 κ λ 10 200 (horizontal axis: products encoded). (c) Ratio of mRNA encoding the ** 15 60 ** 150

10 Ell2 secreted form to mRNA encoding the membrane-specific form (Sec/MB) 5 100

for each immunoglobulin heavy-chain isotype (horizontal axis). Numbers 30 5 Intracellular

IgA MFI ( × 10 50 IgM MFI ( × 10 +/gfp ERT2 IgM MFI ( × 10

above bars indicate the difference in this ratio in Cre PCs Plasma memberane Prdm1 0 0 0 expression (RPKM) 0 versus Prdm1fl/gfpCreERT2 PCs. (d) RNA-sequencing analysis of Ighg2b MB Tot (encoding IgG2b) and Ighm (encoding IgM) in Prdm1+/gfp GC B cells and Prdm1+/gfpCreERT2 and Prdm1fl/gfpCreERT2 BM PCs; dotted outlines delineate secretion-specific sequences; red outlines indicate use of Ighg2b exons encoding the membrane-specific form of IgG2b (enlarged 85×); results are presented as reads per million reads (RPM), and numbers at top right in each plot indicate the highest expression value along axis (throughout). Bottom, exons (blue boxes) and introns (black lines) of Ighm and Ighg2b encoding the membrane-specific form (MB) or secreted form (sec). (e) Flow cytometry of PCs obtained from the BM of Prdm1+/gfpCreERT2 and Prdm1fl/gfpCreERT2 mice 28 d after treatment with tamoxifen, showing intracellular staining of IgM and IgG (left), and frequency of IgM+ or IgG+ cells among the Blimp-1–GFP+ PC population in spleen and BM (right). Numbers adjacent to outlined areas (left) indicate percent IgM+Blimp-1–GFP+ PCs (top row) or IgG+Blimp-1–GFP+ PCs (bottom row). Each symbol (right) represents an individual mouse; small horizontal lines indicate the mean. (f) Mean fluorescence intensity of IgA at the plasma membrane (MB) and total IgA (membrane and intracellular), detected by sequential staining Nature America, Inc. All rights reserved. America, Inc. © 201 6 Nature with the same antibody (left), and of IgM at the plasma membrane (middle) or intracellular IgM (right), detected with separate antibodies to IgM, for PCs obtained from the BM of Prdm1+/gfpCreERT2 and Prdm1fl/gfpCreERT2 mice 35 d after treatment with tamoxifen. (g) RNA-sequencing analysis of the expression of Ell2 in PCs obtained from the BM of Prdm1+/gfpCreERT2 and Prdm1fl/gfpCreERT2 mice 21 d after treatment with tamoxifen. *P < 0.05,

npg **P < 0.01 and ***P < 0.005 (paired t-test). Data are from one experiment representative of two experiments (a; mean and s.d.), two experiments (b–d,g; mean of two biological replicates pooled from multiple individual mice) or are from two or more experiments (e,f; mean and s.d. in f).

notably protein folding and targeting of proteins to the ER, with the mTORC1 complex through analysis of the phosphorylation almost no effect on other effectors, such as the transcription factors of mTOR itself (at Ser2448) and one of its downstream targets S6 (Supplementary Table 2 and Supplementary Fig. 4a). Loss of XBP-1 (at Ser235 and Ser236) revealed considerable upregulation of the also had very little effect on the expression of B cell and PC signature pathway activity in PCs, compared with its activity B cells, and that genes, beyond those encoding components of the UPR (Supplementary this was dependent upon Blimp-1 (Fig. 7a). As a control, we assessed Fig. 4c). This analysis revealed that while the functions of Blimp-1 and phosphorylation of the mTORC2 target Akt (at Ser473) and found XBP-1 overlapped in the control of Igh expression and some aspects of this was unaffected (Fig. 7a). The supply of amino acids, in par- the UPR, most of their functions in PCs were unique. ticular of leucine, is a crucial regulator of mTORC1 activity39, and we determined that the expression of several carriers of amino Requirement for Blimp-1 for full activity of the kinase mTOR acids was lower in Blimp-1-deficient BM PCs than in Blimp-1- A characteristic phenotype of the Blimp-1-deficient PCs was their sufficient BM PCs (Fig. 7a,b). The expression of one of those carriers, smaller cell size, a feature that is under the control of the mTOR CD98, was particularly high on PCs and was very much dependent (‘mammalian target of rapamycin’) pathway38. The kinase mTOR is on Blimp-1 (Fig. 7a). CD98 is encoded by Slc3a2 and Slc7a5, both part of two complexes, mTORC1 and mTORC2; the former regulates Blimp-1-regulated genes, and the latter was also directly bound by cell size, organelle biogenesis and protein synthesis (Supplementary Blimp-1 (Fig. 7b,c). Expression of the transferrin receptor CD71 Fig. 3c). Analysis of the RNA-sequencing data revealed that the (encoded by Tfrc), another carrier known to modulate mTORC1 expression of genes encoding the core components of either complex, activity, was also decreased in absence of Blimp-1 (Fig. 7b). such as mTOR, Raptor and Rictor, was not affected by loss of Blimp-1 Furthermore, genes encoding two members of the sestrin family40, in PCs (data not shown). In contrast, assessment of the activity of Sesn1 and Sesn3, were targets of Blimp-1-mediated repression; they

nature immunology VOLUME 17 NUMBER 3 MARCH 2016 327 A rt i c l e s

Figure 5 Blimp-1 controls the UPR. a 100 B cell (a) Flow cytometry analyzing intracellular +/gfp PC 12 +/gfp staining of the active spliced form of XBP-1 80 fl/gfp PC fl/gfp ) (XBP-1s), ATF4 and ATF6 in +/gfpCreERT2 3 8 Prdm1 60 ** B cells and in PCs obtained from the BM of 40 +/gfp ERT2 fl/gfp ERT2 4 Prdm1 Cre and Prdm1 Cre MFI ( × 10 * 20 ** mice 35 d after treatment with tamoxifen Events (% of max) (left), and mean fluorescence intensity of 0 0 3 4 5 XBP-1s ATF4 ATF6 each (right). *P < 0.05 and **P < 0.01 (paired 0 10 10 10 XBP-1s ATF4 ATF6 t-test). (b) Binding of Blimp-1 to Atf6 (top) or Ern1 (bottom) in lipopolysaccharide-stimulated b c Bio/Bio BirA/BirA 5 Prdm1 Rosa26 PBs, showing 5.2 (RPM) binding Blimp-1-binding regions (small horizontal Blimp-1 lines below plots) identified by peak calling 10 kb UPR gene

with model-based analysis of ChIP-seq data enrichment Atf6 0 algorithm. Below, exon-intron structure 3 Up in Down (vertical lines, exons), with arrow indicating fl/gfp in fl/gfp (RPM) binding direction of transcription, and a scale bar (left) Blimp-1 10 kb in kilobases (kb). (c) Gene-set–enrichment 1.95 1.08 0.63 0.28 17.59 –0.05–0.34–0.70–1.17–1.95 –12.85 analysis (bottom) of genes (shaded rectangles; Ern1 Statistics horizontally ranked by moderated t-statistic) upregulated (pink; t > 1), downregulated (blue (t < −1) or not altered (gray) in Prdm1fl/gfpCreERT2 BM PCs relative to their expression in Prdm1+/gfpCreERT2 BM PCs, obtained from mice 21 d after treatment with tamoxifen; vertical black lines indicate 119 genes encoding components of the UPR (‘UPR gene’), identified by gene annotation and filtered for expression in PCs. Top, enrichment for the position of genes encoding components of the UPR (‘UPR gene’). Data are from two or more experiments (a; mean and s.d.) or two experiments (b,c).

were bound by Blimp-1, and their expression was augmented in (at Ser79) and Raptor (at Ser792), was significantly enhanced follow- Blimp-1-deficient PCs (Fig. 7b,c). Members of this family are acti- ing loss of Blimp-1 (Fig. 7a). These data suggested that Blimp-1 acted vators of the AMP-activated protein kinase (AMPK), which inhibits at multiple points to positively regulate mTORC1 activity, through mTORC1 through the phosphorylation of one of its components, activation of amino acid carriers, including CD98, and repression Raptor41. In keeping with the possibility of increased AMPK activity, of the expression of the negative regulatory members of the sestrin phosphorylation of two targets of AMPK, acetyl-CoA carboxylase family, and thus prevented AMPK activity.

a b c d e

+/+ ) +/+ +/+ +/+ Xbp1 2 5 Xbp1 Xbp1 Xbp1 fl/fl ) fl/fl fl/fl fl/fl 0.20 Xbp1 3 600 Xbp1 30 Xbp1 Xbp1 (log 3 ) 3 +/+

fl/fl 200 10 MFI ( × 10 1 0.05 –5

Plasma cells (%) * *** Xbp1 0 0 0 CD107a MFI ( × 10 0 0 5 10 Expression (RPKM × 10 lgl lgM lgG lgA lgA lgE lgD lgM lgk npg Mean expression lgG1 lgG3 lgG2c lgG2b (log2 RPKM)

f Xbp1+/+ g Xbp1+/+ Xbp1fl/fl 4.4 Xbp1fl/fl

150 800 )

2 Spleen UPR gene 600 enrichment 0 100 Up in Down in 400 fl/fl fl/fl Xbp1 Xbp1

50 spots (mm + 200 ASC per 500 cells Surface area of + BM lgM ** ** 2.5 1.5 1.0 0.5 0.1 16.4 –0.3 –0.8 –1.6 –2.9

lgM 0 0 –31.6 Spleen BM Spleen BM Statistics

Figure 6 Loss of XBP-1 leads to diminished Igh expression and UPR activity. (a) Frequency of BM PCs (CD138+Blimp-1–GFP+ throughout) in Xbp1+/+Prdm1+/gfpCreERT2 mice (Xbp1+/+) or Xbp1fl/fl Prdm1+/gfpCreERT2 mice (Xbp1fl/fl) 35 d after treatment with tamoxifen (to induce inactivation of Xbp1). (b) Scatterplot of differential gene expression in Xbp1+/+ Prdm1+/gfpCreERT2 PCs versus Xbp1fl/fl Prdm1+/gfpCreERT2 PCs obtained from mice 21 d after treatment with tamoxifen (as in a), showing genes with significantly increased expression (blue) or decreased expression (red) in the absence of XBP-1 (FDR <0.05, normalized average expression of ≥4 RPKM in at least one sample). (c) Normalized expression of the constant region–encoding exons of genes in immunoglobulin-encoding loci (as in Fig. 4b) in cells as in b. (d) Intracellular staining of IgM in IgA−IgG− cells, IgA in IgM−IgG− cells, and IgG in IgA−IgM− cells, from mice treated as in b. (e) CD107a expression at the surface of BM PCs in cells as in a. (f) Enzyme- linked immunospot assay of the secretion of IgM by isolated PCs as in b, presented as IgM+ cells per 500 sorted Blimp-1-GFP+CD138+ PCs (left), surface area of the spots identified (middle), and wells of an assay with antibody to IgM (right). (g) Gene-set–enrichment analysis of genes encoding components of the UPR in cells as in b (ranked and presented as in Fig. 5c), upregulated (pink; t > 1), downregulated (blue (t < −1) or not altered (gray) after loss of XBP-1 (vertical black lines and plot above, as in Fig. 5c). *P < 0.05, **P < 0.01 and ***P < 0.005 (paired t-test.). Data are from one experiment representative of two experiments (a; mean and s.d.) or are from two experiments (b–g; mean and s.d. in c–f).

328 VOLUME 17 NUMBER 3 MARCH 2016 nature immunology A rt i c l e s

Figure 7 Blimp-1 regulates the mTOR pathway. B cell 12 +/gfp * (a) Flow cytometry analyzing the expression a +/gfp PC fl/gfp

fl/gfp PC ) of CD98 and the phosphorylation of S6 at 3 8 ** Ser235 and Ser236 (p-S6(S235,S236)), 4 mTOR at Ser2448 (p-mTOR(S2448)), MFI ( × 10 ** * Akt at Ser473 (p-Akt(S473)), Raptor at *** Ser792 (p-Raptor(S792) or acetyl-CoA 0

carboxylase at Ser79 (p-ACC(S79)) in p-S6 p-Akt CD98 p-ACC mature B cells (B220hiCD19hi) from the BM p-mTOR p-Raptor of Prdm1+/gfpCreERT2 and in PCs CD98 p-S6(S235,S236) p-mTOR(S2448) b (CD138+Blimp-1-GFP+) from the BM of 100 2 +/gfp ERT2 fl/gfp ERT2 80 Prdm1 Cre and Prdm1 Cre fold)

2 1 mice 35 d after treatment with tamoxifen 60 0 (left), and mean fluorescence intensity of 40 those results (right). *P < 0.05, **P < 0.01 20 –1 and ***P < 0.005 (paired t-test). (b) RNA- Expression (log sequencing analysis of key genes (horizontal Events (% of max) 0 –2 3 4 5 Tfrc 0 10 10 10 * Slc1a5 Slc3a2Slc7a5Slc7a1 Sesn1Sesn3 axis) upstream of the mTOR pathway in Slc38a2 * * Slc36a1 * * p-Akt(S473) p-raptor(S792) p-ACC(S79) * Prdm1fl/gfpCreERT2 BM PCs, presented relative to their expression in Prdm1+/gfpCreERT2 BM c PCs. *, direct target of Blimp-1. (c) Binding 3 3 (RPM) (RPM) binding binding Blimp-1 of Blimp-1 at Slc7a5 (which encodes a chain Blimp-1 of CD98), Slc38a2, Sesn1 and Sesn3 in 10 kb Bio/Bio 5 kb lipopolysaccharide-stimulated Prdm1 Slc38a2 BirA/BirA Slc7a5 Rosa26 tagged PBs (presented 5 5 as in Fig. 5b). Data are from two or more (RPM) (RPM) binding binding experiments per condition (a; mean and s.d), Blimp-1 Blimp-1 10 kb 10 kb or two experiments (b,c). Sesn1 Sesn3

DISCUSSION of others was maintained indirectly through other mechanisms in IRF4, Blimp-1 and XBP-1 are the best characterized transcription fac- fully mature PCs. Our current findings, together with published tors active in mature PCs. By removing each one in fully differentiated studies10–12,16,20,29, suggest a model in which Blimp-1 is essential for PCs in vivo, we have demonstrated specific roles for these factors in the establishment of the full PC transcriptome, but once it is established, PC biology. PC survival critically relied on IRF4, potentially through PC identity is maintained independently of Blimp-1. the regulation of key survival molecules such as Mcl-1 (ref. 30). Immunoglobulin-encoding transcripts account for approximately Ablation of the PC lineage at even early points after inactivation of 70% of the total mRNA in long-lived PCs6, which allows each cell Irf4 was in agreement with published studies of various myeloma cell to produce close to 2 ng of antibodies per day44. To reach and stably lines23 and precluded any further investigation of the downstream maintain this secretory capacity, PCs require specialized machinery targets of IRF4 with these in vivo models. In contrast, we determined and metabolic activity. Our data demonstrated that Blimp-1 had a Nature America, Inc. All rights reserved. America, Inc. © 201 6 Nature that mature PCs survived acute loss of either Blimp-1 or XBP-1. These central role in this process through the enforcement of the UPR and findings were contrast to published studies concluding that both fac- mTOR pathways. In PCs, the UPR has been proposed to function tors are essential for PC survival18,28. The main point of difference not as a stress response but as a preemptive physiological pathway

npg among these studies was that the others tracked mainly the abundance implemented during the early differentiation that precedes the high and specificity of antibody as a surrogate for PC frequency18,28 and expression of immunoglobulins45. In support of that conclusion, we thus the loss of antibody production in both mutant mouse strains demonstrated that XBP-1, not Blimp-1, maintained the expression of gave the misleading impression of the disappearance of PCs. The use immunoglobulin-encoding transcripts in mature PCs. Blimp-1 was of a Blimp-1–GFP reporter provided a superior tool for us to track the intimately involved in the UPR, directly regulating Atf6 and Ern1, PCs at high resolution, compared with monitoring CD138 or immu- both of which encode products required for full XBP-1 expression and noglobulin secretion11. The survival of Blimp-1- or XBP-1 deficient UPR function, as well as directly regulating 38% of the downstream PCs was also compatible with published reports showing that both genes of the UPR. Blimp-1 was also needed for maintenance of the proteins might act as tumor suppressors in multiple myeloma42,43. characteristic cytoplasmic morphology of PCs, a function also linked Blimp-1 has long been known to be essential for the developmental to the inability to fully activate XBP-1 (refs. 21,46). transition from activated B cell to PC12. Blimp-1 is thought to act Beyond the UPR, the targets of Blimp-1 and XBP-1 were largely by repressing genes encoding the main regulators of the B cell pro- independent, suggestive of distinct functions in PC biology. The regu- gram, including Pax5, Bcl6, Spib and Ciita, as well as silencing Myc lation of activity of the mTOR pathway by Blimp-1 represented one to facilitate the post-mitotic state of mature PCs24. By comparison such unique role. mTOR has a function complementary to that of of our transcriptomic data with a published PC signature6, it was the UPR47, as it modulates protein synthesis or organelle biogenesis apparent that Blimp-1 was not needed to maintain the core identity, in response to environmental stimuli and, in particular, in response as defined by their transcriptional program, of already mature PCs. to nutrient availability. Moreover, enhanced mTOR activity can pro- Blimp-1 does serve some role in repressing B cell transcription, as mote immunoglobulin production independently of XBP-1 (ref. 46). various B cell genes, including those encoding CD23, CD22, Spi-B, Our data showed that Blimp-1 directly fostered this pathway through Id3 and TLR9, but not those encoding Pax-5, Bcl-6 or Myc, were re- two mechanisms. Blimp-1 supported the amino acid supply of the expressed in Blimp-1-deficient PCs; this suggested that while some PCs by directly promoting the expression of specific carriers, such as targets required active Blimp-1-mediated repression, the repression CD98, and prevented the activity of the inhibitory sestrin-AMPK axis.

nature immunology VOLUME 17 NUMBER 3 MARCH 2016 329 A rt i c l e s

Together our studies have demonstrated that while the survival and 16. Shaffer, A.L. et al. Blimp-1 orchestrates plasma cell differentiation by extinguishing the mature B cell gene expression program. Immunity 17, 51–62 (2002). identity of long-lived PCs is largely independent of Blimp-1, this mul- 17. Doody, G.M. et al. An extended set of PRDM1/BLIMP1 target genes links binding tifunctional transcriptional regulator is essential for the molecular and motif type to dynamic repression. Nucleic Acids Res. 38, 5336–5350 (2010). cellular changes that support the extremely high and sustained rates of 18. Reimold, A.M. et al. Plasma cell differentiation requires the transcription factor XBP-1. Nature 412, 300–307 (2001). antibody secretion that are essential for protective immunity. 19. Hu, C.C., Dougan, S.K., McGehee, A.M., Love, J.C. & Ploegh, H.L. XBP-1 regulates signal transduction, transcription factors and bone marrow colonization in B cells. Methods EMBO J. 28, 1624–1636 (2009). 20. Shaffer, A.L. et al. XBP1, downstream of Blimp-1, expands the secretory apparatus Methods and any associated references are available in the online and other organelles, and increases protein synthesis in plasma cell differentiation. version of the paper. Immunity 21, 81–93 (2004). 21. Taubenheim, N. et al. High rate of antibody secretion is not integral to plasma cell differentiation as revealed by XBP-1 deficiency. J. Immunol. 189, 3328–3338 Accession codes. GEO: raw sequence reads, read counts and (2012). normalized expression, GSE70981. 22. Todd, D.J. et al. XBP1 governs late events in plasma cell differentiation and is not required for antigen-specific memory B cell development. J. Exp. Med. 206, Note: Any Supplementary Information and Source Data files are available in the 2151–2159 (2009). online version of the paper. 23. Shaffer, A.L. et al. IRF4 addiction in multiple myeloma. Nature 454, 226–231 (2008). Acknowledgments 24. Shapiro-Shelef, M. & Calame, K. Regulation of plasma-cell development. Nat. Rev. We thank U. Klein (Columbia University) for Irf4fl/fl mice; L. Glimcher (Weill Immunol. 5, 230–242 (2005). 25. Kallies, A., Xin, A., Belz, G.T. & Nutt, S.L. Blimp-1 transcription factor is required Cornell Medical College) for Xbp1fl/fl mice; S. Wilcox, M. Chopin and C. Seillet for for the differentiation of effector CD8+ T cells and memory responses. Immunity technical assistance; and J. Leahy for animal care. Supported by the National Health 31, 283–295 (2009). and Medical Research Council of Australia (G.K.S. and S.L.N.; 361646, 575500 and 26. Klein, U. et al. Transcription factor IRF4 controls plasma cell differentiation and 1054925 to S.L.N.; 1054618 to G.K.S.; 1023454 to G.K.S. and W.S.; and 1049416 class-switch recombination. Nat. Immunol. 7, 773–782 (2006). to A.K.), the Sylvia and Charles Viertel Foundation (A.K.), the Multiple Myeloma 27. Seibler, J. et al. Rapid generation of inducible mouse mutants. Nucleic Acids Res. Research Foundation (S.L.N.), Boehringer Ingelheim (Busslinger laboratory) 31, e12 (2003). and the European Research Council (291740-LymphoControl for the Busslinger 28. Shapiro-Shelef, M., Lin, K.I., Savitsky, D., Liao, J. & Calame, K. Blimp-1 is required laboratory), and made possible through Victorian State Government Operational for maintenance of long-lived plasma cells in the bone marrow. J. Exp. Med. 202, 1471–1476 (2005). Infrastructure Support. 29. Minnich, M. et al. Multifunctional role of the transcription factor Blimp1 in coordinating plasma cell differentiation. doi:10.1038/ni.3349 (18 January 2016). AUTHOR CONTRIBUTIONS Nat. Immunol. 30. Peperzak, V. et al. Mcl-1 is essential for the survival of plasma cells. Nat. Immunol. J.T. performed most experiments; W.S., Y.L. and G.K.S. performed the 14, 290–297 (2013). bioinformatics and statistical analyses; M.M. and M.B. provided data for chromatin 31. O’Connor, B.P. et al. BCMA is essential for the survival of long-lived bone marrow immunoprecipitation followed by deep sequencing; S.C. performed electron plasma cells. J. Exp. Med. 199, 91–98 (2004). microscopy; A.K. provided mouse models; S.L.N. supervised the study; and J.T. and 32. Bayles, I. & Milcarek, C. Plasma cell formation, secretion, and persistence: S.L.N. wrote the manuscript, for which all authors provided editorial input. the short and the long of it. Crit. Rev. Immunol. 34, 481–499 (2014). 33. Martincic, K., Alkan, S.A., Cheatle, A., Borghesi, L. & Milcarek, C. Transcription COMPETING FINANCIAL INTERESTS elongation factor ELL2 directs immunoglobulin secretion in plasma cells by The authors declare no competing financial interests. stimulating altered RNA processing. Nat. Immunol. 10, 1102–1109 (2009). 34. Park, K.S. et al. Transcription elongation factor ELL2 drives Ig secretory-specific Reprints and permissions information is available online at http://www.nature.com/ mRNA production and the unfolded protein response. J. Immunol. 193, 4663–4674 reprints/index.html. (2014). 35. Brewer, J.W. Regulatory crosstalk within the mammalian unfolded protein response. 1. Tarlinton, D., Radbruch, A., Hiepe, F. & Dorner, T. Plasma cell differentiation and Cell. Mol. Life Sci. 71, 1067–1079 (2014). survival. Curr. Opin. Immunol. 20, 162–169 (2008). 36. Yoshida, H., Matsui, T., Yamamoto, A., Okada, T. & Mori, K. XBP1 mRNA is induced 2. MacLennan, I.C. Extrafollicular antibody responses. 194, 8–18 by ATF6 and spliced by IRE1 in response to ER stress to produce a highly active Nature America, Inc. All rights reserved. America, Inc. © 201 6 Nature et al. Immunol. Rev. (2003). transcription factor. Cell 107, 881–891 (2001). 3. Hammarlund, E. et al. Duration of antiviral immunity after smallpox vaccination. 37. Calfon, M. et al. IRE1 couples endoplasmic reticulum load to secretory capacity Nat. Med. 9, 1131–1137 (2003). by processing the XBP-1 mRNA. Nature 415, 92–96 (2002). 4. Bettigole, S.E. & Glimcher, L.H. Endoplasmic reticulum stress in immunity. Annu. 38. Powell, J.D., Pollizzi, K.N., Heikamp, E.B. & Horton, M.R. Regulation of immune npg Rev. Immunol. 33, 107–138 (2015). responses by mTOR. Annu. Rev. Immunol. 30, 39–68 (2012). 5. Nutt, S.L., Hodgkin, P.D., Tarlinton, D.M. & Corcoran, L.M. The generation of 39. Edinger, A.L. & Thompson, C.B. Akt maintains cell size and survival by increasing antibody-secreting plasma cells. Nat. Rev. Immunol. 15, 160–171 (2015). mTOR-dependent nutrient uptake. Mol. Biol. Cell 13, 2276–2288 (2002). 6. Shi, W. et al. Transcriptional profiling of mouse B cell terminal differentiation defines 40. Lee, J.H., Budanov, A.V. & Karin, M. Sestrins orchestrate cellular metabolism to a signature for antibody-secreting plasma cells. Nat. Immunol. 16, 663–673 (2015). attenuate aging. Cell Metab. 18, 792–801 (2013). 7. Kwon, H. et al. Analysis of IL-21-induced Prdm1 gene regulation reveals functional 41. Chen, C.C. et al. FoxOs inhibit mTORC1 and activate Akt by inducing the expression cooperation of STAT3 and IRF4 transcription factors. Immunity 31, 941–952 (2009). of Sestrin3 and Rictor. Dev. Cell 18, 592–604 (2010). 8. Ochiai, K. et al. Transcriptional regulation of germinal center B and plasma cell 42. Leung-Hagesteijn, C. et al. Xbp1s-negative tumor B cells and pre-plasmablasts fates by dynamical control of IRF4. Immunity 38, 918–929 (2013). mediate therapeutic proteasome inhibitor resistance in multiple myeloma. Cancer 9. Sciammas, R. et al. Graded expression of interferon regulatory factor-4 coordinates Cell 24, 289–304 (2013). isotype switching with plasma cell differentiation. Immunity 25, 225–236 (2006). 43. Lohr, J.G. et al. Widespread genetic heterogeneity in multiple myeloma: implications 10. Kallies, A. et al. Initiation of plasma-cell differentiation is independent of the for targeted therapy. Cancer Cell 25, 91–101 (2014). transcription factor Blimp-1. Immunity 26, 555–566 (2007). 44. Brinkmann, V. & Heusser, C.H. T cell-dependent differentiation of human B cells 11. Kallies, A. et al. Plasma cell ontogeny defined by quantitative changes in blimp-1 into IgM, IgG, IgA, or IgE plasma cells: high rate of antibody production by IgE expression. J. Exp. Med. 200, 967–977 (2004). plasma cells, but limited clonal expansion of IgE precursors. Cell. Immunol. 152, 12. Shapiro-Shelef, M. et al. Blimp-1 is required for the formation of immunoglobulin 323–332 (1993). secreting plasma cells and pre-plasma memory B cells. Immunity 19, 607–620 45. Gass, J.N., Gifford, N.M. & Brewer, J.W. Activation of an unfolded protein (2003). response during differentiation of antibody-secreting B cells. J. Biol. Chem. 277, 13. Lin, K.I., Angelin-Duclos, C., Kuo, T.C. & Calame, K. Blimp-1-dependent repression 49047–49054 (2002). of Pax-5 is required for differentiation of B cells to immunoglobulin M-secreting 46. Benhamron, S., Pattanayak, S.P., Berger, M. & Tirosh, B. mTOR activation promotes plasma cells. Mol. Cell. Biol. 22, 4771–4780 (2002). plasma cell differentiation and bypasses XBP-1 for immunoglobulin secretion. 14. Lin, Y., Wong, K. & Calame, K. Repression of c-myc transcription by Blimp-1, Mol. Cell. Biol. 35, 153–166 (2015). an inducer of terminal B cell differentiation. Science 276, 596–599 (1997). 47. Goldfinger, M., Shmuel, M., Benhamron, S. & Tirosh, B. Protein synthesis in plasma 15. Piskurich, J.F. et al. BLIMP-I mediates extinction of major histocompatibility class II cells is regulated by crosstalk between endoplasmic reticulum stress and mTOR transactivator expression in plasma cells. Nat. Immunol. 1, 526–532 (2000). signaling. Eur. J. Immunol. 41, 491–502 (2011).

330 VOLUME 17 NUMBER 3 MARCH 2016 nature immunology ONLINE METHODS 75- to 100-bp single-end sequencing on an Illumina HiSeq 2500 instrument at Mice. Prdm1gfp/+ mice11, Prdm1fl mice25, Irf4fl mice26, Xbp1fl mice22 and the Australian Genome Research Facility. Raw RNA-seq data previously gener- Rosa26-CreERT2 mice27 were bred and maintained at the animal facilities of ated for follicular B cells (GSE60927) were also included in this analysis. the Walter and Eliza Hall Institute. Prdm1fl mice carried a loxP-flanked exon 5, Sequence reads were aligned to the GRCm38/mm10 build of the while Prdm1gfp mice expressed an allele truncated after exon 5 (of a total of Mus musculus genome using the Subread aligner49. Only uniquely mapped 7 exons). The PCR analysis for exon 5 loss has been described previously25. reads were retained. Genewise counts were obtained using featureCounts50. All mice were on C57BL/6 background. Animal experiments were conducted Reads overlapping exons in annotation build 38.1 of NCBI RefSeq database according to the protocols approved by the Walter and Eliza Hall Institute were included. Immunoglobulin-encoding genes were excluded from the gene- ethics committee. Immunization was a single intraperitoneal injection of level expression analysis and were analyzed separately. Genes were filtered 100 µg NP-KLH precipitated onto alum (ThermoScientific). For cell transfer from downstream analysis if they failed to achieve a CPM (counts per million experiments, total splenocytes (5 × 106) or isolated B cells (2 × 106) were mapped reads) value of at least 1 in at least one library. Counts were converted −/− injected intravenously into Rag1 recipients. Cre-mediated deletion of to log2 counts per million, quantile normalized and precision weighted with loxP-flanked alleles was triggered by the administration of tamoxifen (Sigma- the ‘voom’ function of the limma package51,52. A linear model was fitted to Aldrich, 0.2 mg/g) by oral gavage on 2 consecutive days. each gene, and empirical Bayes moderated t-statistics were used to assess differences in expression53. Genes were called ‘differentially expressed’ if they PB generation in vitro. B cells were isolated from splenocytes by positive achieved a FDR of 0.05 or less. The gene set enrichment plots were generated selection using the B220- or CD19-coated beads (Miltenyi Biotec) and cultured with the ‘barcodeplot’ function in limma. Gene-set–enrichment analysis was as previously described48. Cultures were seeded at 1 × 106/ml with optimal carried out using the ‘roast’ method in limma with 999 rotations54. One-sided concentrations of CD40L (100 ng/ml; R&D Systems), interleukin 4 (10 ng/ml; P values are reported. R&D Systems) and interleukin 5 (5 ng/ml; R&D Systems). Cultures were For the analysis of immunoglobulin-encoding genes, sequence reads were treated with 4-hydroxytamoxifen (Sigma-Aldrich) at 100 nM. re-aligned to the genome using the Subjunc aligner with the ‘–allJunctions’ option to disable the requirement for the presence of donor and receptor sites Flow cytometry and cell sorting. Single-cell suspensions were stained with when mapping exon-spanning reads49. Mapped reads were then assigned antibodies to the following surface molecules: CD138 (281.2; BD Biosciences), to all exons using featureCounts. Exon-spanning reads were assigned to all CD19 (1D3; in-house), B220 (RA3-6B2; in-house), IgM (II41; eBiosciences), their overlapping exons. RPKM (reads per kilobase of exon length per mil- CD98 (RL388; BioLegend), CD93 (AA4.1; BD Biosciences), major histocom- lion mapped fragments) values for exons encoding immunoglobulin constant patibility complex class II (M5; eBiosciences), CD22 (OX-97; BioLegend), regions were computed on the basis of exon lengths and the total exon counts CD107a (1D4B; BD Biosciences), CD71 (B2; BD Biosciences). Cells were in each library. The Subjunc mapping data were also used to identify the fre- stained with lysotracker deep red or ERtracker red (Molecular Probes) accord- quency of splicing between Prdm1 exons 5–6 and 4–6 and thus the efficiency ing to manufacturer’s instructions. For intracellular transcription factor and of Cre-mediated deletion of the loxP-flanked exon 5 in PCs. immunoglobulin measurement, cells were fixed and permeabilized using the eBiosciences transcription factor staining buffer set and BD Cytofix/ Transmission electron microscopy. BM Blimp-1–GFP+CD138+ PCs were Cytoperm kit, respectively, then were stained with antibodies specific for the purified by flow cytometry and prepared for transmission electron microscopy following: XBP-1s (Q3-695; BD Biosciences), ATF4 (D4B8; Cell Signaling), as previously described21. ATF6 (ab37149; Abcam), IgM (715-475-140; Jackson ImmunoResearch), IgG (715-475-151; Jackson ImmunoResearch), IgA (mA-6E1; eBiosciences), immu- Analysis of Blimp-1 binding by chromatin immunoprecipitation followed noglobulin κ-chain (187.1; BD Biosciences) and immunoglobulin λ-chain by deep sequencing. CD138+ PBs were generated by lipopolysaccharide- (TB28-2; BD biosciences). For detection of phosphorylated proteins, cells induced differentiation of mature B cells from Prdm1Bio/BioRosa26BirA/BirA were prepared using BD Phosphoflow lyse/fix buffer and perm buffer III, and mice as described in detail elsewhere29. Chromatin from ~1 × 108 PBs was labeled with specific antibodies that recognized the following: S6 phospho- prepared using a lysis buffer containing 0.25% SDS before chromatin precipita- Nature America, Inc. All rights reserved. America, Inc. © 201 6 Nature rylated at Ser235/236 (D57.2.25; Cell Signaling), mTOR phosphorylated at tion by streptavidin precipitation (Bio-ChIP), as described55. The precipitated Ser2448 (D9C2; Cell Signaling), ACC phosphorylated at Ser79 (D7D11; Cell genomic DNA was quantified by real-time PCR, and about 1–5 ng of DNA pre- Signaling), Raptor phosphorylated at Ser792 (2083; Cell Signaling) or Akt cipitated via chromatin immunoprecipitation was used for library preparation

npg phosphorylated at Ser473 (M89-61; BD Biosciences). Stained cells were ana- and subsequent Illumina deep sequencing. Peak calling and gene assignment lyzed on a FACSCantoII or Fortessa X20 (BD Biosciences). PCs were enriched is described in detail elsewhere29. from BM cells with biotinylated antibody to CD138 (8B12; in-house) and anti- biotin microbeads (Miltenyi Biotec) and sorted with FACSDiva or FACSAria cytometers (BD Biosciences). 48. Hasbold, J., Corcoran, L.M., Tarlinton, D.M., Tangye, S.G. & Hodgkin, P.D. Evidence from the generation of immunoglobulin G-secreting cells that stochastic mechanisms regulate lymphocyte differentiation. Nat. Immunol. 5, 55–63 (2004). Enzyme-linked immunospot assay. Multiscreen-HA plates (Millipore) were 49. Liao, Y., Smyth, G.K. & Shi, W. The Subread aligner: fast, accurate and scalable coated with antibody to mouse IgM (1021-01) or IgG (1030-01). Cells were read mapping by seed-and-vote. Nucleic Acids Res. 41, e108 (2013). incubated on the plates overnight at 37 °C. After incubation with biotinylated 50. Liao, Y., Smyth, G.K. & Shi, W. featureCounts: an efficient general purpose program for assigning sequence reads to genomic features. Bioinformatics 30, 923–930 goat anti-mouse IgM (1021-08) and IgG (1030-08), followed by streptavidin– (2014). alkaline phosphatase, spots were developed using the substrate VectorBlue 51. Law, C.W., Chen, Y., Shi, W. & Smyth, G.K. voom: Precision weights unlock linear (Vector Labs). All antibodies used for coating and labeling were obtained model analysis tools for RNA-seq read counts. Genome Biol. 15, R29 (2014). from Southern Biotech. 52. Ritchie, M.E. et al. limma powers differential expression analyses for RNA- sequencing and microarray studies. Nucleic Acids Res. 43, e47 (2015). 53. Smyth, G.K. Linear models and empirical bayes methods for assessing differential Whole-transcriptome analysis. RNA was isolated from ex vivo–sorted Blimp- expression in microarray experiments. Stat. Appl. Genet. Mol. Biol. 3, Article3 1-GFP+CD138+ PCs from previously tamoxifen-treated Prdm1+/gfpCreERT2, (2004). Prdm1fl/gfpCreERT2, Xbp1fl/fl Prdm1+/gfpCreERT2 and Xbp1+/+ Prdm1+/gfpCreERT2 54. Wu, D. et al. ROAST: rotation gene set tests for complex microarray experiments. Bioinformatics 26, 2176–2182 (2010). BM using the Qiagen RNeasy Micro kit. Libraries were generated using the 55. Ebert, A. et al. The distal VH gene cluster of the Igh locus contains distinct regulatory Illumina Truseq RNA sample preparation kit following the manufacturer’s elements with Pax5 transcription factor-dependent activity in pro-B cells. Immunity instructions. For all samples, 100 ng RNA were subjected to a transcriptome 34, 175–187 (2011).

doi:10.1038/ni.3348 nature immunology