B cell and plasma cell fates are controlled by different transcriptional networks, as exemplified by the mutually exclusive expression and cross-antagonism of the B cell identity factor Pax5 and the plasma cell regulator Blimp1. It has been postulated that repression of Pax5 by Blimp1 is essential for plasma cell development. Here, we challenged this hypothesis by analyzing the IghPax5/+ mouse, which expressed a Pax5 minigene from the immunoglobulin heavy-chain locus. Despite high Pax5 expression, plasma cells efficiently developed in young IghPax5/+ mice at steady state and upon immunization, while their number moderately declined in older mice. Although Pax5 significantly deregulated the plasma cell expression program, key plasma cell regulators were normally expressed in IghPax5/+ plasma cells. While IgM and IgA secretion by IghPax5/+ plasma cells was normal, IgG secretion was modestly decreased. Hence, Pax5 repression is not essential for robust plasma cell development and antibody secretion, although it is required for optimal IgG production and accumulation of long-lived plasma cells.
Plasma cells provide acute and long-term protection of the host against infection through secretion of high-affinity antibodies that recognize an almost unlimited number of pathogens (Nutt et al., 2015). The highly diverse B cell antigen receptor (BCR) repertoire is generated in early B cell development by V(D)J recombination of the immunoglobulin heavy-chain (Igh) and light-chain (Igk or Igl) gene loci (Alt et al., 2013). Somatic hypermutation of the rearranged Ig genes subsequently creates pathogen-specific high-affinity BCRs in germinal center (GC) B cells (Victora and Nussenzweig, 2012). The terminal differentiation of B cells to antibody-secreting plasma cells is characterized by a massive reprogramming of gene expression (Minnich et al., 2016; Shi et al., 2015). While the transcription factors Irf4, Blimp1, E2A, and E2-2 are required for the development of plasma cells (Gloury et al., 2016; Minnich et al., 2016; Nutt et al., 2015; Tellier et al., 2016; Wöhner et al., 2016), the unfolded protein response regulator Xbp1 is essential for controlling the enormous expansion of the ER and thus the high capacity of antibody secretion in plasma cells (Reimold et al., 2001; Shaffer et al., 2004). In contrast, the development, function, and identity of B cells is regulated by a different transcriptional network, including the B cell–specific regulators Pax5 and Ebf1 (Boller and Grosschedl, 2014; Medvedovic et al., 2011).
The transcription factor Pax5 controls the commitment of lymphoid progenitors to the B cell pathway at the onset of B-lymphopoiesis (Medvedovic et al., 2011; Nutt et al., 1999). Pax5 fulfills its B cell commitment function by repressing lineage-inappropriate genes to suppress alternative lineage fates and by activating B-lymphoid–specific genes to promote B cell development (Delogu et al., 2006; Nutt et al., 1999; Revilla-I-Domingo et al., 2012; Schebesta et al., 2007). Pax5 functions as a B cell identity factor throughout B cell development, as mature splenic B cells upon conditional Pax5 loss are able to dedifferentiate to uncommitted progenitors in the bone marrow, which can further develop into functional T cells (Cobaleda et al., 2007). In late B lymphopoiesis, Pax5 is required for the generation or survival of all mature B cell types, as it controls signaling from both the BCR and Toll-like receptors (unpublished data). In addition, PAX5 functions as a haploinsufficient tumor suppressor in one third of all human B-progenitor acute lymphoblastic leukemias (Gu et al., 2019), while it acts as an oncoprotein in a subset of B cell non-Hodgkin’s lymphomas carrying an IGH-PAX5 translocation (Medvedovic et al., 2013).
The transcriptional regulators Pax5 and Blimp1 (encoded by the Prdm1 gene) are expressed in a mutually exclusive manner in the B cell lineages. Pax5 is expressed at all stages of B cell development from the pro–B cell to the activated B cell stage and, upon terminal differentiation, is repressed in plasmablasts and plasma cells (Fuxa and Busslinger, 2007). In contrast, Blimp1 expression is not detected in B cells, but is initiated in pre-plasmablasts and maintained in all antibody-secreting cells (ASCs) in peripheral lymphoid organs and in long-lived plasma cells in the bone marrow (Kallies et al., 2004). Consistent with their mutually exclusive expression patterns, Pax5 and Blimp1 cross-antagonize each other and their respective gene expression programs, as Pax5 directly represses the Prdm1 (Blimp1) gene in mature B cells (Delogu et al., 2006; Revilla-I-Domingo et al., 2012), while Blimp1 directly represses Pax5 in plasmablasts and plasma cells (Minnich et al., 2016).
Deletion of the endogenous Pax5 gene in the chicken DT40 B cell line leads to plasmacytic differentiation at the expense of the B cell phenotype, as Blimp1 and Xbp1 expression is induced together with increased IgM secretion, while Bcl6 and Ebf1 expression is repressed (Nera et al., 2006). Ectopic expression of Pax5 was furthermore shown to suppress the spontaneous differentiation of a mature B cell line to ASCs (Usui et al., 1997). Moreover, retroviral expression of Pax5 in splenic B cells was shown to prevent lipopolysaccharide-induced differentiation to IgM-secreting plasmablasts (Lin et al., 2002). Together, this evidence led to the commonly held view that the repression of Pax5 is essential for plasma cell differentiation to occur (Nutt et al., 2015). However, this hypothesis has never been experimentally challenged in an in vivo mouse model.
We have previously generated a mouse model for a human IGH-PAX5 translocation (Busslinger et al., 1996) by inserting a PAX5 minigene into the endogenous Igh locus adjacent to the Eμ enhancer (Souabni et al., 2007; Fig. S1 A). Heterozygous IghP5ki/+ mice in the absence of a cooperating oncogenic mutation do not give rise to B cell lymphoma and only rarely develop T-lymphoblastic lymphoma after a long latency (Souabni et al., 2007). In wild-type mice, the expression of the rearranged Igh mRNA is known to increase 30-fold during the transition from activated B cells to plasmablasts, while the Pax5 gene is simultaneously repressed (Minnich et al., 2016). In contrast, the plasmablasts and plasma cells of IghP5ki/+ mice should continue to express Pax5 from the Pax5 minigene due to the increased transcriptional activity of the Igh locus, while the endogenous Pax5 gene is repressed by Blimp1. Using this mouse model, we have now demonstrated that, despite extensive gene expression changes, plasmablasts and plasma cells efficiently developed in the presence of continued Pax5 expression. The Pax5-expressing plasma cells were functional as they secreted antigen-specific antibodies in response to immunization, although the longevity of plasma cells in the bone marrow was moderately decreased with time. Contrary to the current dogma, these data therefore demonstrate that the repression of Pax5 is not required for plasma cell development to occur.
B and T cells are present at normal frequencies in IghPax5/+ mice
As plasmablasts and plasma cells are defined by their expression of Blimp1 (Prdm1), we used the Prdm1Gfp reporter allele (Kallies et al., 2004) for identifying these rare cells by flow cytometry. For this purpose, we first deleted the frt-flanked IRES-Gfp reporter gene located downstream of the PAX5 minigene in the IghP5ki allele (Souabni et al., 2007) by Flpe-mediated recombination to generate the IghPax5 allele (Fig. S1 A). We subsequently performed all analyses with experimental IghPax5/+Prdm1Gfp/+ and control Igh+/+Prdm1Gfp/+ mice. For simplicity, we will abbreviate the two genotypes to IghPax5/+ and Igh+/+, except for experiments where GFP expression from the Prdm1Gfp allele was essential for plasma cell identification. Since the Igh locus is known to be already active in lymphoid progenitors of IghP5ki/+ mice (Souabni et al., 2007), we examined B cell development in the bone marrow as well as B and T cells in the spleen and lymph nodes of experimental and control mice by flow cytometry. B cell development was normal in the bone marrow of IghPax5/+ mice compared with control mice, except for a twofold increase in pro-B cells (Fig. 1 A), which is likely caused by the precocious expression of Pax5 in the preceding lymphoid progenitors (Souabni et al., 2007). Similarly, the follicular and marginal zone B cells were present at comparable frequencies in the spleen of IghPax5/+ and control littermates (Fig. 1 B). The CD8+ T cells were also unaffected in the lymph nodes of IghPax5/+ mice, while a small reduction in CD4+ T cells was observed in these mice compared with control littermates (Fig. 1 C).
Analysis of Pax5 expression by intracellular flow cytometry revealed that the Pax5 protein levels were similar or only slightly increased in the different B cell populations of IghPax5/+ mice compared with control littermates (Fig. 1, D and E), consistent with the absence of a B cell phenotype. Both CD4+ and CD8+ T cells in the lymph nodes of IghPax5/+ mice showed ectopic and variegated Pax5 expression (Fig. 1 F), in agreement with the fact that the Igh locus is transcriptionally active in T cells (Kemp et al., 1980). Hence, Pax5 expression levels were minimally increased in B cells and moderately elevated in T cells, consistent with normal B and T cell development in IghPax5/+ mice.
Efficient plasma cell development in the presence of Pax5 expression
To determine whether ectopic Pax5 expression can inhibit plasma cell differentiation, naive B cells from IghPax5/+Prdm1Gfp/+ mice were stimulated in vitro with bacterial LPS. After 4 d of stimulation, IghPax5/+Prdm1Gfp/+ plasmablasts could be readily detected as GFP+CD138+ cells and were only minimally decreased in frequency compared with control Igh+/+Prdm1Gfp/+ plasmablasts (Fig. 2 A). As expected, the IghPax5/+ plasmablasts expressed strongly elevated levels of Pax5 protein, as shown by intracellular Pax5 staining (Fig. 2 B). Detailed analysis revealed a rapid and progressive loss of Pax5 protein expression during the differentiation of activated Igh+/+ B cells to plasmablasts (Fig. S1, B and C). In contrast, Pax5 protein expression was only moderately reduced during the transition from activated IghPax5/+ B cells to plasmablasts, which was likely caused by repression of the endogenous Pax5 gene (Fig. S1 C). Ectopic transcription from the IghPax5 allele instead gave rise to the high Pax5 protein expression observed in IghPax5/+ plasmablasts (Fig. S1 C). Together, these data indicate that, contrary to expectation, ectopic Pax5 expression from the Igh locus did not prevent in vitro plasmablast differentiation.
We next interrogated the ability of IghPax5/+Prdm1Gfp/+ mice to generate plasma cells in response to an antigen challenge in vivo. 14 d after immunization with the T cell–dependent antigen 4-hydroxy-3-nitrophenylacetyl (NP)–conjugated keyhole limpet hemocyanin (NP-KLH; in alum), GFP+CD138+ plasma cells were detected in the spleens of IghPax5/+Prdm1Gfp/+ mice at similar frequencies as in control Igh+/+Prdm1Gfp/+ mice (Fig. 2 C). As the Prdm1Gfp reporter allele can distinguish between less mature GFPint and mature GFPhi plasma cells (Kallies et al., 2004), closer analysis of the GFP expression level within the bulk GFP+CD138+ plasma cell population revealed a moderate skewing of the plasma cells in IghPax5/+Prdm1Gfp/+ mice toward the less mature GFPint phenotype in contrast to the control Igh+/+Prdm1Gfp/+ mice (Fig. 2 C).
In addition to investigating plasma cells generated in secondary lymphoid organs following an antigen challenge, we also examined the long-lived plasma cells that accumulate in the bone marrow at steady state. In young mice at the age of 2–3 mo, plasma cells (GFP+CD138+) were present in IghPax5/+Prdm1Gfp/+ mice at a slightly elevated frequency compared with control Igh+/+Prdm1Gfp/+ mice (Fig. 2 D, top row). This increase in IghPax5/+Prdm1Gfp/+ mice was due to an increase of less mature GFPint plasma cells, while the mature GFPhi plasma cells were present at a similar frequency in experimental and control mice (Fig. 2 D, top row). In contrast, fewer total plasma cells were detected in the bone marrow of older IghPax5/+Prdm1Gfp/+ mice relative to control mice at the age of 6–8 mo (Fig. 2 D, bottom row). While the less mature GFPint plasma cells were present at similar frequency, older IghPax5/+Prdm1Gfp/+ mice failed to accumulate mature GFPhi plasma cells to the same extent as their control Igh+/+Prdm1Gfp/+ littermates (Fig. 2 D, bottom row). In summary, ectopic Pax5 expression is compatible with robust plasma cell development, but subsequently interferes with long-term accumulation of mature plasma cells in the bone marrow.
Using intracellular flow cytometry, we confirmed that IghPax5/+Prdm1Gfp/+ plasma cells expressed greatly elevated levels of Pax5 protein (Fig. 2 E). Flow-cytometric analyses furthermore revealed increased expression of the cell surface proteins CD19 and CD22 as well as decreased expression of CD93 and Sca1 on plasma cells of IghPax5/+Prdm1Gfp/+ mice compared with control Igh+/+Prdm1Gfp/+ littermates (Fig. 2 E). Importantly, Cd19 and Cd22 are known activated Pax5 target genes in pro-B cells, while Sca1 and Cd93 are repressed Pax5 target gene identified in pro-B and mature B cells, respectively (Revilla-I-Domingo et al., 2012). Together, these observations demonstrated that the ectopically expressed Pax5 protein was transcriptionally active in IghPax5/+Prdm1Gfp/+ plasma cells.
Pax5-dependent deregulation of the plasma cell gene expression program
To systematically analyze the effect of Pax5 on gene expression in plasma cells, we first investigated the genome-wide Pax5-binding pattern in IghPax5/+ and control Igh+/+ plasmablasts, which were isolated at day 4 of LPS stimulation to a purity of 92–95% by immunomagnetic enrichment with CD138-MicroBeads (Fig. S2 A). In parallel, we analyzed the CD138-depleted cell fractions, which were strongly enriched for activated B cells (73–75%; Fig. S2 A) and were therefore referred to as “activated B cells.” As we expected minimal Pax5 binding in Igh+/+ plasmablasts, we performed calibrated chromatin immunoprecipitation coupled with deep sequencing (ChIP-seq) experiments (Hu et al., 2015) to verify the quality of the ChIP-seq data obtained with these cells. For this, we mixed chromatin prepared from the purified mouse plasmablasts or activated B cells with chromatin from chicken DT40 B cells at a 9:1 ratio before ChIP-seq analysis with an antibody that recognized the conserved paired domain of mouse and chicken Pax5 (see Materials and methods). Peak calling with a stringent P value of <10−10 identified 17,907 and 9,934 Pax5 peaks in activated IghPax5/+ and Igh+/+ B cells, respectively, which were characterized by a similar average peak height (read density; Fig. 3, A and B; and Fig. S2, B and C). Notably, 20,481 Pax5 peaks of a high read density were identified in IghPax5/+ plasmablasts. In contrast, only 1,618 Pax5 peaks, characterized by a low read density, could be determined in Igh+/+ plasmablasts (Fig. 3, A and B; and Fig. S2, B and C), indicating that only residual Pax5 was still bound to chromatin upon loss of Pax5 expression in wild-type plasmablasts (Fig. 2 B and,Fig. S1 C). Pax5 binding at the genes Cd19, Cd22, Cd79a, and Id3, all of which have previously been identified as activated Pax5 target genes in pro-B cells (Revilla-I-Domingo et al., 2012; Schebesta et al., 2007), exemplified how Pax5 peaks are maintained in IghPax5/+ plasmablasts and are lost in control Igh+/+ plasmablasts (Fig. 3 C). Hence, these data demonstrate efficient binding of Pax5 across the genome in plasmablasts that normally do not express this transcription factor.
We next investigated the degree to which the ectopically expressed Pax5 protein could interfere with the transcriptional program of plasmablasts and plasma cells in IghPax5/+Prdm1Gfp/+ mice compared with control Igh+/+Prdm1Gfp/+ mice. For this, we performed RNA-sequencing (RNA-seq) analysis with sorted in vitro LPS-differentiated plasmablasts (CD22–CD138+GFP+) or plasma cells (CD138+GFP+), which were purified ex vivo from the bone marrow of 6–8-mo-old mice by immunomagnetic enrichment with CD138-MicroBeads before flow-cytometric isolation (Fig. S3 A). As the mature GFPhi plasma cells expressed higher levels of CD138 compared with the less mature GFPint plasma cells in the bone marrow of IghPax5/+Prdm1Gfp/+ mice (Fig. 2 D), the immunomagnetic enrichment of CD138+ cells efficiently selected for mature GFPhi plasma cells (Fig. S3 A). By comparing the gene expression changes between IghPax5/+ and control Igh+/+ plasmablasts, we identified 122 Pax5-activated and 78 Pax5-repressed genes, based on an expression difference of more than twofold, an adjusted P value of <0.05, and an expression value of >5 TPM (transcripts per million) in one of the two plasmablast cell types (Fig. S3 B and Table S1). By using a higher threefold cutoff, we identified 246 Pax5-activated and 66 Pax5-repressed genes in IghPax5/+ plasma cells compared with control Igh+/+ plasma cells (Fig. 4 A and Table S2), which resulted in an overlap of 62 activated and 14 repressed genes between in vitro differentiated plasmablasts and long-term plasma cells (Fig. S3 C). Annotation of the Pax5-regulated genes identified in IghPax5/+ plasma cells revealed that a large fraction of the activated and repressed genes code for cell surface proteins, signal transducers, and metabolic enzymes, which could potentially affect the function of Pax5-expressing plasma cells (Fig. 4 B). Pax5 binding, determined in IghPax5/+ plasmablasts (Fig. 3), was observed at 190 (77%) of the 246 Pax5-activated genes and only at 28 (42%) of the 66 Pax5-repressed genes (Table S2), suggesting that Pax5 may be more effective in directly activating genes in IghPax5/+ plasma cells. Consistent with this finding, 67 of the 246 Pax5-activated genes in IghPax5/+ plasma cells were previously identified as activated Pax5 target genes, while only 8 of the 66 Pax5-repressed genes were earlier described as repressed Pax5 target genes in pro-B and mature B cells (Delogu et al., 2006; Revilla-I-Domingo et al., 2012; Schebesta et al., 2007; Table S2).
Gene set enrichment analysis (GSEA) revealed that genes up-regulated during normal plasmablast or plasma cell development (Minnich et al., 2016) were significantly enriched as Pax5-repressed genes, while conversely genes down-regulated during terminal differentiation were enriched as Pax5-activated genes in IghPax5/+ plasmablasts and IghPax5/+ plasma cells, respectively (Fig. 4 C and,Fig. S3 D). The expression of selected genes, which qualified as activated Pax5 target genes both in IghPax5/+ plasma cells and pro-B or mature B cells, are shown in Fig. 4 D, while the expression of genes that were identified as activated Pax5 target genes only in IghPax5/+ plasma cells is shown in Fig. 4 E. Likewise, the expression of common and plasma cell–specific repressed Pax5 target genes is indicated in Fig. S3, E and F. Despite the substantial gene expression changes, the key plasma cell regulators Blimp1, Irf4, E2A, E2-2, Ikaros, RelA, and Zbtb20 (Chevrier et al., 2014; Nutt et al., 2015; Roy et al., 2019) were expressed at similar or slightly elevated levels in IghPax5/+ plasma cells relative to Igh+/+ plasma cells (Fig. 4 F), which explains why plasma cells could efficiently develop in IghPax5/+ mice. The twofold decreased expression of Xbp1 in IghPax5/+ plasma cells mimics the heterozygous Xbp1 condition, which was reported not to affect antibody secretion in Xbp1+/− plasmablasts (Taubenheim et al., 2012). Notably, Blimp1 was normally expressed in IghPax5/+ plasma cells despite the fact that Pax5 activated the expression of Irf8, SpiB (Spib), Bach2, and c-Rel (Rel; Fig. 4 G), which have also been implicated as repressors of the Prdm1 gene in mature B cells (Carotta et al., 2014; Ochiai et al., 2006; Roy et al., 2019; Willis et al., 2017). Importantly, the expression of the pro-survival genes Mcl1 and Bcl2 was similar or twofold increased in IghPax5/+ plasma cells compared with Igh+/+ plasma cells (Fig. 4 F). Finally, the expression of genes, which code for cell adhesion proteins and surface receptors involved in the control of plasma cell homing and survival (Nutt et al., 2015), was 1.6-fold increased (Cxcr4) or 1.3-fold to 2.4-fold decreased (Tnfrsf17 [BCMA], Cd44, Il6ra, Cd28, and Itga4 [VLA4]) in IghPax5/+ plasma cells relative to control cells (Fig. S3 G). More importantly, the mRNA expression of the cell surface lectin CD93, implicated in plasma cell persistence (Chevrier et al., 2009), was 5.3-fold repressed by ectopic Pax5 expression in plasma cells (Fig. 2 E and,Fig. S3 E), suggesting that the loss of CD93, possibly in combination with the minimally reduced expression of other cell surface receptors, may contribute to the impaired longevity of IghPax5/+ plasma cells in older mice.
Normal IgM secretion by Pax5-expressing plasma cells
Given the Pax5-dependent deregulation of gene expression in plasma cells, we next investigated whether the plasma cells of IghPax5/+ mice were functional, particularly with regard to antibody secretion. To this end, we performed ELISPOT assays to determine the number and secretory capacity of IgM and IgG ASCs. The secretory capacity was determined by the size of the ELISPOT, which correlates with the amount of antibody secretion by individual plasma cells. Under steady-state conditions, the number and ELISPOT size of IgM-producing ASCs were similar in the bone marrow of young IghPax5/+ and control Igh+/+ mice at the age of 2–3 mo (Fig. 5 A), indicating that the Pax5-expressing plasma cells retained the ability to undergo the transcriptional and morphological changes required for immunoglobulin secretion. However, the number and ELISPOT size of IgG-producing ASCs were reduced in IghPax5/+ mice (Fig. 5 A). Similar results were obtained with mice at the age of 6–8 mo, except for the observed 2.2-fold decrease in the number of IgM-producing ASCs in the bone marrow of IghPax5/+ mice compared with control littermates (Fig. 5 B). This reduction likely reflects the twofold decrease in total plasma cell numbers that we observed by flow-cytometric analysis in old IghPax5/+ mice (Fig. 2 D, bottom row). Moreover, ELISA measurements of the serum titers of the different Ig isotypes in young and old mice revealed a moderate decrease only for IgG1 in IghPax5/+ mice (Fig. S4, A and B), in agreement with the corresponding reduction of the ELISPOT size of IgG ASCs in these mice (Fig. 5, A and B).
To interrogate the capacity of ectopic Pax5-expressing plasma cells to mount a specific antibody response to a defined T cell–dependent antigen, we immunized IghPax5/+ and control mice with NP-KLH (in alum) and performed ELISPOT assays at day 14 after immunization to detect NP-specific plasma cells. NP-specific IgM ASCs were present at similar numbers and had a similar ELISPOT size in both the spleen and bone marrow of IghPax5/+ and control mice (Fig. 5, C and D). However, as observed already under steady-state conditions, both the number and ELISPOT size (secretion) of the NP-specific IgG1 ASCs were reduced in the spleens and bone marrow of immunized IghPax5/+ mice compared with control mice (Fig. 5, C and D). Together, these data indicate that IgM-specific plasma cells were generated in normal numbers and secreted similar amounts of IgM in IghPax5/+ mice in contrast to the IgG-specific plasma cells, which were present at lower numbers and showed a moderate defect in IgG secretion.
To gain further insight into the IgG-specific plasma cell phenotype, we compared the expression of Igh mRNAs of different Ig isotypes by interrogating the RNA-seq data of IghPax5/+ and control Igh+/+ plasma cells (Fig. 4 A). Notably, the Ighm and Igha mRNAs were present at similar abundance in plasma cells of both genotypes, whereas the expression of Ighg1, Ighg2c, and Ighg3 mRNA was reduced in IghPax5/+ plasma cells compared with control plasma cells (Fig. S4 C). These data strongly suggest that the IgG-specific plasma cells were generated in lower numbers due to a defect in class switch recombination (CSR) to IgG isotypes in IghPax5/+ B cells.
Finally, we used transition electron microscopy to analyze the morphology of the plasma cells isolated from the bone marrow of 6–8-mo-old IghPax5/+ and control Igh+/+ mice. The ultrastructure of the Pax5-expressing plasma cells was mostly normal compared with that of control plasma cells, as the cytoplasm of both plasma cell types was strongly increased by an extended network of the ER (Fig. 5 E and,Fig. S4 D). While the stacking of the ER appeared slightly less tight in Pax5-expressing plasma cells, the cytosolic surfaces of the ER were similarly densely occupied with ribosomes, as shown at higher magnification (Fig. 5 E and,Fig. S4 D). To measure the increase of the ER network by a second independent method, we stained plasma cells from the bone marrow of 6–8-mo-old mice with the ER-Tracker Red dye. Surprisingly, flow-cytometric analysis revealed that the intensity of the ER-Tracker staining was moderately increased in IghPax5/+ plasma cells compared with control Igh+/+ plasma cells (Fig. S4 E). In summary, we conclude that Pax5 expression did not interfere with the formation of a functional secretory apparatus in plasma cells.
Impaired CSR to IgG1 in IghPax5/+ mice
We next investigated the potential of IghPax5/+ B cells to undergo CSR to IgG under in vitro and in vivo conditions. For this, we stimulated naive splenic B cells for 4 d with LPS or anti-CD40, IL-4, and IL-5 to induce CSR to IgG3 or IgG1, respectively (Fig. 6, A and B). Under both conditions, IghPax5/+ B cells underwent CSR to IgG3 and IgG1 at similar frequency as control Igh+/+ B cells (Fig. 6, A and B), indicating that ectopic Pax5 expression in B cells did not interfere with CSR under in vitro stimulation conditions. However, the analysis of GC B cells (CD19+GL7+Fas+) at day 14 after immunization with NP-KLH (in alum) revealed a 1.8-fold decrease of GC B cells in the spleen of IghPax5/+ mice compared with control Igh+/+ littermates (Fig. 6 C). While the antigen-specific (NP+) cell fraction within the GC B cell population was similar, the fraction of IgG1-switched cells within the NP + GC B cell population was 2.9-fold reduced in IghPax5/+ mice compared with control littermates (Fig. 6 C). A similar result was obtained upon immunization with sheep RBCs (SRBCs), which elicits a strong polyclonal B cell response (Fig. S5 A). At day 14 after immunization with this antigen, GC B cells were present at a similar frequency in the spleen of IghPax5/+ and Igh+/+ mice, while the frequency of IgG1+ GC B cells was still twofold reduced in IghPax5/+ mice compared with control littermates (Fig. S5 A). Together, these data indicate that ectopic Pax5 expression interfered with CSR to IgG1 in vivo, but not under in vitro conditions.
We hypothesize that the in vivo versus in vitro discrepancy is likely explained by the complex GC microenvironment consisting of several cell types that support the activation and differentiation of B cells upon cognate antigen interaction. In particular, CSR and GC B cell differentiation require the interaction with T follicular helper (TFH) cells (Crotty, 2019). Upon recruitment to the GC, the TFH cells provide signals to B cells through their costimulatory surface molecules CD40L and ICOS (CD278) as well as through secretion of the cytokines IL-4 and IL-21 (Crotty, 2019). At 14 d after NP-KLH immunization, TFH cells were present at normal numbers in splenic GCs of IghPax5/+ mice (Fig. 6 D). As observed for all T cells, the TFH cells of IghPax5/+ mice also exhibited ectopic and variegated expression of Pax5 (Fig. 6 D). It is therefore conceivable that ectopic Pax5 expression may impair the function of the TFH cells, thus leading to impaired CSR in IghPax5/+ mice.
The presence of normal T cells rescues CSR in IghPax5/+ B cells
As the TFH cells ectopically expressed Pax5 in IghPax5/+ mice, we next examined whether the CSR defect of these mice was B cell–intrinsic or dependent on the Pax5-expressing TFH cells. To this end, we generated mixed bone marrow chimeras by reconstituting lethally irradiated C57BL/6 mice with a 1:1 mixture of IghPax5/+ and IghΔJh/ΔJh (JHT) bone marrow (Fig. S5 B). All B cells in these chimeric mice originated from the IghPax5/+ stem cells, as the homozygous JHT mutation, eliminating the DHQ52, JH and Eμ elements of the Igh locus (Gu et al., 1993), prevented B cell development of the JHT progenitors. In contrast, all T cells should be derived from the JHT stem cells, as the IghPax5/+ progenitors cannot contribute to T lymphopoiesis in a competitive setting due to a partial block of thymic T cell development in IghPax5/+ mice (Souabni et al., 2007). For comparison, we generated bone marrow chimeras with a mixture of Igh+/+ and JHT bone marrow and, 3 mo after transplantation, immunized all chimeric mice with SRBCs, followed by flow-cytometric analysis after 14 d (Fig. S5 B). The TFH cells in the experimental IghPax5/+ chimeric mice did not express Pax5 and were thus of JHT origin (Fig. S5 C). Notably, the IgG1+ GC B cells were present at similar frequencies in the spleen of experimental IghPax5/+ and control Igh+/+ chimeric mice (Fig. S5 D), indicating that the decrease of IgG1+ GC B cells in IghPax5/+ mice (Fig. S5 A) was not a B cell–intrinsic defect.
In addition, we performed an adoptive cell transfer experiment by coinjecting splenic IghB1-8hi/Pax5Prdm1Gfp/+ B cells (CD45.1+CD45.2+) and OT-II TCR-transgenic Rag2−/− CD4+ T cells (CD45.2+) into C57BL/6 mice (CD45.1+; Fig. 7 A). The IghB1-8hi allele expresses the NP-specific high-affinity BCR B1-8hi (Shih et al., 2002), and the OT-II transgene expresses an OVA-specific TCR (Barnden et al., 1998). In the context of this experiment, it is important to note that Pax5 was not expressed in TFH cells derived from OT-II TCR-transgenic Rag2−/− CD4+ T cells. Control cell transfer experiments were performed with IghB1-8hi/+Prdm1Gfp/+ B cells (CD45.1+CD45.2+; Fig. 7 A). 1 d after injection, the mice were immunized with the NP-OVA antigen (in alum) and analyzed by flow cytometry and ELISPOT assay 6 d after immunization. Analyses of the transferred CD45.1+CD45.2+ B cells and their GFP+ plasma cell progeny revealed similar frequencies of total and IgG1+ GC B cells (Fig. 7 B) as well as GFP+ plasma cells (Fig. 7 C) in the spleen of mice injected with experimental (IghB1-8hi/Pax5) or control (IghB1-8hi/+) B cells. The numbers of NP-IgM and NP-IgG1 ASCs was minimally reduced in the experimental compared with the control setting (Fig. 7 D). The ELISPOT size of experimental NP-IgM and NP-IgG1 ASCs was 1.5- and 1.9-fold reduced compared with control ASCs, respectively, thus further indicating a minor defect of antibody secretion in Pax5-expressing plasma cells. In summary, these experiments indicate that CSR to IgG was rescued in IghPax5/+ B cells in the presence of functional T cells, which resulted in normalized numbers of IgG-secreting plasma cells.
Normal B cell immune responses in Peyer’s patches of IghPax5/+ mice
Since we observed that the B cell immune responses in IghPax5/+ mice are restored in the presence of functional T cells, we predicted that immunization with the T cell–independent TNP-Ficoll antigen should give rise to normal CSR and plasma cell development in IghPax5/+ mice. Indeed, total and IgG2b+ plasma cells (TACI+CD138+) were present at similar frequencies in the spleen of IghPax5/+ and Igh+/+ mice at day 14 after TNP-Ficoll immunization (Fig. 8 A). Consistent with this result, the level of Ighg2b transcripts was also similar in bone marrow plasma cells of both genotypes (Fig. S4 C). We next analyzed Peyer’s patches, as CSR to IgA in these gut-associated lymphoid organs is also known to be T cell–independent (Bergqvist et al., 2010). Total and IgA-switched GC B cells were present in similar number in the Peyer’s patches of IghPax5/+ and Igh+/+ mice (Fig. 8 B). Moreover, total and IgA+ plasma cells were equally abundant in Peyer’s patches of both genotypes (Fig. 8 C). Hence, these data provide further evidence that ectopic Pax5 expression in the B cell lineage does not interfere with CSR, GC B cell formation, and plasma cell development.
The distinct gene expression programs of B cells and plasma cells are controlled by different transcription factors as exemplified by the mutually exclusive expression of Pax5 and Blimp1 and their cross-antagonism in B cell and plasma cell development, respectively (Nutt et al., 2015). In addition, ectopic expression of Pax5 in in vitro cultured B cells was shown to prevent plasmablast differentiation (Lin et al., 2002; Usui et al., 1997). Together, these observations led to the commonly held view that plasma cell differentiation strictly depends on the repression of Pax5 (Nutt et al., 2015). Here, we challenged this hypothesis by investigating the development of plasma cells in the IghPax5/+ mouse model, which demonstrated, contrary to expectation, that ectopic Pax5 expression from the Igh locus is compatible with efficient development of plasmablasts and plasma cells, as discussed below.
While Pax5 is similarly expressed during B cell development in IghPax5/+ mice as in control littermates, its expression and genome-wide binding are maintained at a high level only in plasmablasts and plasma cells of the IghPax5/+ mice. As a consequence, the ectopically expressed Pax5 protein activates or represses many genes that are normally down- or up-regulated in plasma cells, respectively. Notably, one third of the deregulated and Pax5-bound genes in IghPax5/+ plasma cells were previously identified as activated or repressed Pax5 target genes in pro-B or mature B cells (Delogu et al., 2006; Revilla-I-Domingo et al., 2012; Schebesta et al., 2007; Table S2). Despite the large gene expression changes, the key plasma cell regulators Blimp1, Irf4, E2A, E2-2, and Ikaros are similarly expressed in IghPax5/+ and control plasma cells. Consistent with this finding, the IghPax5/+ mice undergo efficient plasma cell development under steady-state conditions and in response to immunization. These data indicate first and foremost that plasma cell development can occur in the absence of Pax5 repression, contrary to the currently held dogma. Second, Pax5-expressing plasma cells exhibit normal trafficking and homing from secondary lymphoid organs to their survival niches in the bone marrow. The discrepancy between earlier published data and our current results may be explained by the fact that the effect of ectopic Pax5 expression was previously analyzed by transfection in in vitro cultured B cells in contrast to our in vivo analysis of the more physiological IghPax5/+ mouse model.
The initiation of plasmablast differentiation in the presence of continuous Pax5 expression was an unexpected result, as Pax5 is known to transcriptionally repress the expression of the essential plasma cell regulator Blimp1 in mature B cells (Delogu et al., 2006; Revilla-I-Domingo et al., 2012). In addition to Pax5, other transcription factors such as Bach2, SpiB (Spib), Irf8, and c-Rel (Rel) also contribute to the repression of the Prdm1 gene in B cells (Carotta et al., 2014; Ochiai et al., 2006; Roy et al., 2019; Willis et al., 2017). Although Pax5 activated the expression of these repressors also in IghPax5/+ plasma cells, Blimp1 expression remained normal in these cells. On the other hand, multiple transcription factors including Irf4, Stat3, RelA (Rela), E2A (Tcf3), and E2-2 (Tcf4) are required to activate the Prdm1 gene at the onset on plasmablast differentiation (Klein et al., 2006; Kwon et al., 2009; Roy et al., 2019; Sciammas et al., 2006; Wöhner et al., 2016). Notably, B cells that undergo strong BCR and CD40 signaling in the light zone of the GC first activate the Irf4 gene to become plasmablast precursors before the initiation of Prdm1 expression in plasmablasts (Ise et al., 2018). It is therefore conceivable that the increased expression of the activator Irf4 can override the Pax5-mediated repression of Prdm1 during the B cell–to–plasmablast transition. Once expressed, Irf4 and Blimp1 subsequently control the plasma cell fate regardless of ectopic Pax5 expression, indicating that these plasma cell regulators are dominant over Pax5 in plasma cells.
One complication of the IghPax5/+ mouse model is the fact that ectopic and variegated expression of Pax5 is also observed in the T cell lineage of this mouse strain, including the TFH cells (Souabni et al., 2007; this study). It is therefore not surprising that the IghPax5/+ mice exhibit impaired CSR to IgG and GC B cell differentiation, as both pathways strictly depend on functional TFH cells (Crotty, 2019). Replacing the dysfunctional IghPax5/+ TFH cells with wild-type Igh+/+ TFH cells in competitive cell transfer experiments resulted in normal numbers of IgG1+IghPax5/+ GC B cells in response to immunization, indicating that the observed CSR and GC B cell defects are indeed not B cell–autonomous. In contrast, the moderate defect of IgG1 secretion by IghPax5/+ plasma cells is B cell–intrinsic, while the secretion of IgM is normal, consistent with the finding that Pax5-expressing plasma cells exhibit a morphology characterized by an extended ER network similar to wild-type plasma cells. The Fc regions of the different IgG isotypes are known to specifically undergo fucosylation, which is catalyzed by the fucosyltransferase Fut8 in the ER (Li et al., 2015). Fut8 is known to be required for efficient secretion of the different IgG isotypes (Li et al., 2015). In this context, it is interesting to note that Pax5 directly represses the Fut8 gene fourfold in IghPax5/+ plasma cells (Fig. S3 F). These data therefore suggest that the Pax5-dependent repression of Fut8 may be responsible for the impaired IgG secretion of IghPax5/+ plasma cells.
While young IghPax5/+ mice have modestly elevated levels of long-lived plasma cells in the bone marrow compared with control mice, the Pax5-expressing plasma cells are skewed toward the less mature Blimp1-GFPint phenotype (Kallies et al., 2004). Moreover, the number of total as well as mature Blimp1-GFPhi plasma cells decreases with progressing age in IghPax5/+ mice, demonstrating that ectopic Pax5 expression interferes with the longevity of plasma cells in the bone marrow. Notably, the Bcl-2 family member Mcl-1, which is essential for the survival of plasma cells in the bone marrow (Peperzak et al., 2013), is normally expressed in Pax5-expressing plasma cells. As the cell surface lectin CD93 is also required for the persistence of plasma cells in the bone marrow niche (Chevrier et al., 2009), it is conceivable that the fivefold repression of Cd93 expression by Pax5 (Fig. 2 E and,Fig. S3 E) may contribute to the impaired longevity of IghPax5/+ plasma cells.
In summary, our study has unequivocally demonstrated that antibody-secreting plasma cells can robustly develop even in the absence of Pax5 repression. However, Pax5 repression is required to ensure efficient IgG secretion and normal accumulation of long-lived plasma cells with progressing age.
Materials and methods
The following mice were maintained on the C57BL/6 genetic background: IghP5ki/+ (Souabni et al., 2007), IghPax5/+ (this study), Prdm1Gfp/+ (Kallies et al., 2004), JHT (Gu et al., 1993), IghB1-8hi/B1-8hi (Shih et al., 2002), Rag2−/− (Shinkai et al., 1992), OT-II TCR-tg (Barnden et al., 1998), and transgenic Flpe mice (Rodríguez et al., 2000). All animal experiments were performed according to valid project licenses, which were approved and regularly controlled by the Austrian Veterinary Authorities.
The following monoclonal antibodies were used for flow cytometry: B220/CD45R (RA3-6B2), CD2 (RM2-5), CD4 (GK1.5), CD8α (53–6.7), CD19 (1D3), CD21 (7G6), CD22 (Cy34.1), CD23 (B3B4), CD45.1 (A20), CD45.2 (104), CD49b/Itga2 (DX5), CD93 (AA4.1), CD95/Fas (Jo2), CD117/Kit (2B8), CD138 (281–2), GL7 (GL7), CD267/TACI (8F10), IgA (mA-6E1), IgD (11-26c.2a), IgM (II/41), IgG1 (A85-1), IgG2b (R12-3), IgG3 (R40-82), Sca1/Ly6a (D7), TCRβ (H57-597), CD185/Cxcr5 (REA215), and PD1 (REA802). The following monoclonal antibodies were used for intracellular flow cytometry: Pax5 (1H9) and Bcl6 (K112-91). NP-BSA-biotin was obtained from Biosearch Technologies. The ER content of B lymphocytes was determined by staining with the ER-Tracker Red dye (BODIPY TR Glibenclamide; Invitrogen).
The following antibody was used for ChIP-seq analysis: anti-Pax5 (rabbit polyclonal antibody, detecting the paired domain; amino acids 17–145; Adams et al., 1992; Busslinger laboratory). The following antibodies were used for ELISPOT and ELISA: goat anti-mouse IgM, IgG, IgG1, IgG2b, and IgA (human adsorbed; SouthernBiotech), HRP-coupled goat anti-mouse IgM, IgG, and IgG1 (human adsorbed; SouthernBiotech), and AP-coupled goat anti-mouse IgM, IgG1, IgA, and IgE (human adsorbed; SouthernBiotech).
Definition of cell types by flow cytometry
Pro-B cells (B220+CD19+Kit+CD2–IgM–), pre-B cells (B220+CD19+Kit–CD2+IgM–), immature B cells (B220+CD19+IgM+IgD–), recirculating B cells (B220+CD19+ IgMloIgDhi), follicular B cells (B220+CD19+CD23hiCD21lo), marginal zone B cells (B220+CD19+CD21hiCD23lo), GC B cells (CD19+B220+GL7+Fas+), in vitro activated B cells (B220+CD22+CD138–Blimp1-GFP–), in vitro cultured plasmablasts (CD22–CD138+Blimp1-GFP+), plasma cells (Lin–[CD4–CD8–CD21–CD49b–]B220intCD138+Blimp1-GFP+), TFH cells (CD4+CXCR5+PD1+Bcl6+). Flow cytometry experiments and FACS sorting were performed on LSR Fortessa (BD Biosciences) and FACSAria III (BD Biosciences) machines, respectively. FlowJo software (Treestar) was used for data analysis.
Generation of bone marrow chimeras
Bone marrow cells from JHT, IghPax5/+Prdm1Gfp/+, and Igh+/+Prdm1Gfp/+ mice were enriched for hematopoietic progenitors by immunomagnetic depletion of cells expressing the cell surface markers CD4, CD8, CD21, and CD49b. Progenitor cells from IghPax5/+Prdm1Gfp/+ and Igh+/+Prdm1Gfp/+ mice were mixed in a 1:1 ratio with progenitor cells from JHT mice before intravenous injection into lethally irradiated (2 × 5.5 Gy) JHT recipients. The mice were immunized 3 mo after bone marrow reconstitution.
Adoptive transfer of B and T cells
CD43– B cells from the spleen of IghB1-8hi/Pax5Prdm1Gfp/+ and IghB1-8hi/+Prdm1Gfp/+ mice (CD45.1+ CD45.2+) were isolated by immunomagnetic depletion with CD43-MicroBeads (Miltenyi Biotec). CD4+ T cells from the spleen of OT-II TCR-tg Rag2−/−mice (CD45.2+) were selected by immunomagnetic depletion of CD11b+, CD11c+, and CD115+ myeloid cells. Approximately 3 × 106 B cells and 5 × 105 T cells from the respective donors were transferred intravenously into C57BL/6 recipients (CD45.1+).
Intracellular Pax5 staining
Pax5 staining was performed with a monoclonal anti-Pax5 antibody (1H9; BioLegend) after fixation and permeabilization with the Foxp3/transcription factor staining buffer set (eBioscience).
In vitro B cell stimulation experiments
Mature B cells from spleen and lymph nodes were isolated as CD43– B cells by immunomagnetic sorting with CD43-MicroBeads (Miltenyi Biotec). For LPS stimulation, mature B cells were seeded at a density of 5 × 105 cells/ml in IMDM medium containing 10% FCS (A15-101; GE Healthcare), 1 mM glutamine, 50 µM β-mercaptoethanol, and 25 µg/ml LPS (L4130; Sigma-Aldrich) for 4 d. For anti-CD40 plus IL-4 and IL-5 stimulation experiments, mature B cells were plated at 5 × 105 cells/ml in IMDM medium supplemented with 10% FCS, 1 mM glutamine, and 50 µM β-mercaptoethanol, and containing anti-CD40 antibody (2 µg/ml; HM40-3; eBioscience), IL-4 (20 ng/ml), and IL-5 (10 ng/ml) for 4 d.
The immune response to a T cell–dependent antigen was studied by intraperitoneal injection of 100 µg NP-KLH (Biosearch Technologies) in alum, or 100 µg of NP conjugated to OVA (NP-OVA; Biosearch Technologies) in alum. To investigate the response to a polyclonal antigen, SRBCs were washed in PBS and resuspended at 109 cells/ml followed by intraperitoneal injection of 100 µl into an adult mouse. GC B cells and plasma cells were analyzed by flow cytometry 6–14 d after immunization. The immune response to a T cell–independent antigen was studied by intraperitoneal injection of 10 µg of TNP-Ficoll (Biosearch Technologies) in PBS and by flow-cytometric analysis 14 d after immunization.
The frequencies of ASCs were determined by ELISPOT assay as described (Smith et al., 1997). ELISPOT plates (Merck) coated with goat anti-mouse IgM or IgG antibody (SouthernBiotech) were used for capturing total IgM or IgG antibodies, respectively, which were secreted by ASCs in the bone marrow of nonimmunized mice at steady state. Alternatively, NP-BSA–coated ELISPOT plates were used for capturing NP-specific antibodies secreted by splenic ASCs from immunized mice. Cells from bone marrow or spleen were treated with the ACK lysing buffer (Thermo Fisher Scientific) before incubation for 6 h at 37°C in ELISPOT wells at 2 × 105 cells/well (steady state) or 2 × 106 cells/well (after immunization). 6 d after NP-OVA immunization (Fig. 7 D), CD138+GFP+ plasma cells were directly sorted into ELISPOT wells at 500 cells/well before incubation. ELISPOTs were visualized with goat anti-mouse IgM, IgG, or IgG1 antibodies conjugated to HRP (SouthernBiotech), and color was developed by addition of the AEC staining reagents (Sigma-Aldrich). After extensive washing, the spots were counted with an ELISPOT Reader System (Autoimmun Diagnostika).
Measurement of ELISPOT size
The size of all ELISPOTs in each well was analyzed using a macro developed in Fiji macro language to perform automated segmentation and measurement in batch mode. Borders of the well area were defined to exclude outside reflections from the analysis, and then color deconvolution was used to separate ELISPOTs from background. Laplace filtering and background subtraction were used to enhance contrast and create binary images of the objects. Segmentation was performed in three rounds to cover different sizes, and the resulting images were then merged. The segmented images were then refined, and objects with low contrast were excluded. The resulting areas were measured.
Serum titers of total IgM, IgG1, IgA, and IgE in unimmunized mice were determined by ELISA (Smith et al., 1997). ELISA plates (Sigma-Aldrich) were coated with goat anti-mouse IgM, IgG1, IgA, and IgE antibody (human adsorbed; SouthernBiotech) before sera was incubated at 4°C for 16 h. Sera antibody concentrations were determined by addition of goat anti-mouse IgM, IgG1, IgA, and IgE antibodies conjugated to alkaline phosphatase (human adsorbed, SouthernBiotech), and color was developed by addition of pNPP One Component Microwell Substrate Solution (SouthernBiotech). The color intensity was quantified using an Epoch Microplate Spectrophotometer (BioTek), using a standard consisting of a mixture of all the sera analyzed.
Plasma cells from the bone marrow were isolated by enrichment of CD138+ cells using immunomagnetic beads (STEMCELL Technologies), followed by flow cytometry sorting of CD138+GFP+ plasma cells. Plasma cells were fixed using 2.5% glutaraldehyde in 0.1 M sodium phosphate buffer, pH 7.4, for 1 h at room temperature. Due to the low numbers of plasma cells, the pelleted cells were mixed with 3% agarose and cut into small pieces. Samples were post-fixed in 2% osmium tetroxide in 0.1 M sodium phosphate buffer on ice and dehydrated in a graded series of acetone on ice before embedding in Agar 100 resin. Subsequently, 70-nm sections were cut parallel to the substrate and post-stained with 2% uranyl acetate and Reynolds lead citrate. Sections were examined with a FEI Morgagni 268D (FEI) operated at 80 kV. Images were acquired using an 11-megapixel Morada charge-coupled device camera (Olympus-SIS).
ChIP analysis of Pax5
In vitro activated B cells and plasmablasts derived by LPS stimulation from IghPax5/+ and Igh+/+ B cells, and DT40 chicken cells were separately cross-linked with 1% formaldehyde (Sigma-Aldrich) for 10 min. Nuclei were prepared and lysed in the presence of 0.25% SDS, and chromatin was sheared by sonication with the Bioruptor Standard (Diagenode). Murine chromatin samples were mixed with 10% of chicken DT40 chromatin, followed by immunoprecipitation with a polyclonal antibody recognizing the conserved paired domain of Pax5 (Adams et al., 1992). Specific enrichment was measured and calculated as the precipitated DNA amount relative to input DNA. The ChIP-precipitated DNA (0.5–1 ng) was used for library preparation and subsequent Illumina deep sequencing.
cDNA preparation for RNA-seq
Plasma cells (CD138+GFP+) were sorted from the bone marrow as described in Fig. S3 A, and plasmablasts were isolated by flow-cytometric sorting as CD22–CD138+GFP+ cells after 4 d of LPS stimulation of B cells from the spleen and lymph nodes. Total RNA from in vitro stimulated plasmablasts or ex vivo sorted plasma cells was isolated with the RNeasy Plus Mini Kit (Qiagen), and mRNA was purified by two rounds of poly(A) selection with the Dynabeads mRNA purification kit (Invitrogen). The mRNA was fragmented by heating at 94°C for 3 min in fragmentation buffer. The fragmented mRNA was used as template for first-strand cDNA synthesis with random hexamers and the Superscript Vilo First-Strand Synthesis System (Invitrogen). The second-strand cDNA synthesis was performed with 100 mM deoxy ATP (dATP), dCTP, dGTP, and dUTP in the presence of RNase H, Escherichia coli DNA polymerase I, and DNA ligase (Invitrogen). The incorporation of dUTP allowed for specific elimination of the second DNA strand during library preparation, thereby preserving strand specificity (Parkhomchuk et al., 2009).
Library preparation and Illumina deep sequencing
0.5 ng to 1 µg of cDNA or ChIP-precipitated DNA was used as starting material for the generation of sequencing libraries with the NEBNext Ultra II DNA library prep kit for Illumina (NEB). Cluster generation and sequencing were performed using the Illumina HiSeq 2500 system with a read length of 50 nucleotides, according to the manufacturer’s guidelines.
Database of RefSeq-annotated genes
The database generation of RefSeq-annotated genes was performed as previously described (Wöhner et al., 2016). To refine the annotation of immunoglobulin genes, the immunoglobulin λ light-chain segments were replaced with their corresponding converted GRCm38.p3 annotations (Ensembl version 79; Yates et al., 2016). The resulting number of genes was 24,732.
Processing of Pax5 ChIP-seq data
All reads that passed a basic quality control filter (Illumina chastity) were further processed. Adapters were removed with Cutadapt, and the reads were filtered against the rDNA with Bowtie 2 (Langmead and Salzberg, 2012). The remaining reads were aligned with Bowtie 1.0 (Langmead et al., 2009), first to the mouse genome (mm9), and remaining unaligned reads to the chicken genome (gal5). Peaks were called with the MACS program version 2.1.0 (Zhang et al., 2008), using a Rag2−/− pro-B cell input (GSM1145867, GSM1296537) for mouse and no input for chicken. Chicken reads were counted over union chicken peaks using featureCount version 1.5.0 (Liao et al., 2014). Normalization factors for the scaling of the mouse Pax5 ChIP-seq track based on the chicken spike-in reads were calculated using the estimateSizeFactors function of DESeq2 (Love et al., 2014) with R version 3.4.1 (https://www.r-project.org). This resulted in the following factors: 0.83 (GSM4473547), 2.18 (GSM4473548), 0.78 (GSM4473549), and 0.78 (GSM4473550). Aligned mouse reads were converted to read coverage tracks using a pipeline that involves samtools version 1.9 (Li et al., 2009), bedtools version 2.7.21 (Quinlan and Hall, 2010), and University of California Santa Cruz kent tools (Kuhn et al., 2013), considering the appropriate normalization factors. The aligned chicken reads were deduplicated using the rmdup command (samtools version 1.9) before coverage conversion. Genome-wide read coverages were normalized to the number of aligned reads per million. Peak overlaps were calculated with the Multovl program (Aszódi, 2012). Peak-to-gene assignment was performed as described (Revilla-I-Domingo et al., 2012). Read density profiles were calculated with customized R scripts based on genome-wide peak-centered read coverages.
RNA-seq data processing
Mouse reads that passed the basic quality control (Illumina chastity) were further processed. The first six nucleotides at the 5′ end were trimmed, adapters were removed with Cutadapt, and the reads were filtered against the rDNA with Bowtie 2 (Langmead and Salzberg, 2012). The remaining reads were aligned with TopHat version 1.4.1 (Trapnell et al., 2009) to the transcriptome and full genome of Mus musculus, version mm9. Reads were counted over merged gene models using featureCounts version 1.5.0 (Liao et al., 2014). Differential expression between IghPax5/+ (KI) and Igh+/+ (WT) samples were analyzed for each cell type by DESeq2 (Love et al., 2014). In brief, the model design “~ group” and Negative Binomial GLM fitting and Wald statistic (“DESeq” function) were used, whereby “group” specifies the combinations of the conditions (plasma_cell, plasma_blast, wt, ki). Genes with an adjusted P value of <0.05, a mean TPM of >5, and an absolute fold-change of >3 in plasma cells or >2 in plasmablasts, respectively, were called as significantly differentially expressed. Immunoglobulin and TCR genes were filtered from the list of significantly expressed genes. Regularized log transformations were computed with the blind option set to “FALSE” and were transformed from the log2 to log10 scale for the scatterplots shown in Fig. 4 A and,Fig. S3 B. TPM values were calculated as described (Wagner et al., 2012).
GSEA was performed using the GSEA software from the Broad Institute (Subramanian et al., 2005). Genes were ranked based on log2 fold change from the DESeq2 package and compared with our own defined gene sets.
Statistical significance was determined with the GraphPad Prism 7 software, using the tests described in each figure legend.
Online supplemental material
Fig. S1 shows a schematic diagram of the IghPax5 allele and the Pax5 protein expression during LPS-induced differentiation of IghPax5/+ and Igh+/+ plasmablasts. Fig. S2 describes the flow-cytometric sorting of LPS-induced IghPax5/+ and Igh+/+ plasmablasts and the Pax5-binding patterns in sorted activated B cells and plasmablasts of both genotypes. Fig. S3 further examines the Pax5-dependent gene expression in IghPax5/+ plasmablasts and plasma cells. Fig. S4 shows the antibody titers and ER morphology of plasma cells in IghPax5/+ and Igh+/+ mice. Fig. S5 describes the immune responses to SRBC immunization in IghPax5/+ mice. Table S1 and Table S2 contain the RNA-seq data of all Pax5-regulated genes identified in LPS-induced plasmablasts and ex vivo sorted bone marrow plasma cells of the IghPax5/+ genotype, respectively. Table S3 describes all Illumina sequencing experiments generated for this study.
We thank S. Jurado for help with injection experiments, M. Fischer for initial analysis of RNA-seq data, B. Vilagos for initial immunization experiments, T. Lendl for writing the program script for ELISPOT size analysis, K. Aumayr’s team for flow-cytometric sorting, A. Sommer’s team at the Vienna BioCenter Core Facilities (VBCF) for Illumina sequencing, and N. Drexler at VBCF for electron microscopy.
This research was supported by Boehringer Ingelheim, the European Research Council under the European Union’s Horizon 2020 research and innovation program (grant agreement no. 740349-PlasmaCellControl), the Austrian Industrial Research Promotion Agency (Headquarter Grant FFG-852936), and the Austrian Science Fund (FWF) DK W1212 (to B. Agerer and A. Bergthaler).
Author contributions: G.J. Liu did most experiments; M. Wöhner helped with cell injections and advice on plasma cell experiments; S.G. Malin performed initial plasma cell analyses; M. Jaritz bioinformatically analyzed the RNA-seq and ChIP-seq data; B. Agerer and A. Bergthaler performed immunization experiments; G.J. Liu and M. Busslinger planned the project, designed the experiments, and wrote the manuscript.
Disclosures: The authors declare no competing interests exist.
S.G. Malin’s present address is Center for Molecular Medicine, Department of Medicine, Karolinska Institutet, and Karolinska University Hospital, Stockholm, Sweden.