Mir-17–92 regulates bone marrow homing of plasma cells and production of immunoglobulin G2c

The polycistronic mir-17–92 cluster, also known as oncomir-1, was previously shown to be essential for early B lymphopoiesis. However, its role in late-stage B-cell differentiation and function remains unexplored. Here we ablate mir-17–92 in mature B cells and demonstrate that mir-17–92 is dispensable for conventional B-cell development in the periphery. Interestingly, mir-17–92-deficiency in B cells leads to enhanced homing of plasma cells to the bone marrow during T-cell-dependent immune response and selectively impairs IgG2c production. Mechanistically, mir-17–92 directly represses the expression of Sphingosine 1-phosphate receptor 1 and transcription factor IKAROS, which are, respectively, important for plasma cell homing and IgG2c production. We further show that deletion of mir-17–92 could reduce IgG2c anti-DNA autoantibody production and hence mitigate immune complex glomerulonephritis in Shp1-deficient mice prone to autoimmunity. Our results identify important roles for mir-17–92 in the regulation of peripheral B-cell function. After activation and selection, plasma cells home to the bone marrow where they persist and continue to make antibodies. Here the authors show that the mir-17–92cluster coordinates the process by regulating the homing receptor S1PR1 and the transcription factor IKAROS that controls IgG2c production.

B cells and antibodies are central elements of adaptive humoral immunity and critical for host protection. B lymphopoiesis begins with committed haematopoietic stem cells differentiating into pro-B, pre-B and immature B cells in the bone marrow (BM) and cumulates with mature B cells seeding the peripheral lymphoid organs. On activation by foreign antigens, naive B cells undergo terminal differentiation into memory B and antibodysecreting plasma cells (PCs) that provide life-long protection 1,2 . B-cell differentiation is governed by a network of transcription factors 3,4 . Recently, microRNAs (miRNAs) are shown to exert additional control over B-cell differentiation and function 5,6 .
MiRNAs are endogenously expressed short non-coding RNAs that fine-tune gene expression at the post-transcriptional level by either promoting the degradation or impeding the translation of target mRNAs [7][8][9] . A 'seed region' of six to eight nucleotides at the 5 0 end of a miRNA determines its target specificity 10 . The importance of miRNAs in B-cell biology is highlighted by the drastic defects in mice with global deletion of miRNAs caused by ablating Dicer, a RNAse III endonuclease critical for miRNA processing, in precursor 11 , peripheral mature 12 and activated 13 B cells. When Dicer was deleted at the earliest stage of B-cell maturation using mb-1 Cre, B lymphopoiesis was severely blocked at the pro-B to pre-B-cell transition 11 . This defect was mainly attributed to the absence of mir-17-92 whose seed sequences were enriched in the 3 0 -untranslated regions (3 0 UTRs) of mRNAs upregulated in Dicer-deficient pro-B cells 11 .
MiR-17-92, or oncomir-1, is a polycistronic miRNA cluster that yields six mature miRNAs, including mir-17, mir-18a, mir-19a, mir-20a, mir-19b and miR-92a. It was first identified as a potent miRNA oncogene on human chromosome 13q31, a region that was amplified in a number of B-cell malignancies 14,15 . Overexpression of mir-17-92 in haematopoietic cells significantly accelerated c-Myc induced B-lymphomagenesis in mice 16 and B-cell-specific mir-17-92 transgenic mice developed lymphoma with high penetrance 17 . On the other hand, germline deletion of mir-17-92 compromised early B lymphopoiesis suggesting that mir-17-92 plays essential roles in early B-cell development 18 . Pro-apoptotic BIM was identified as a direct target of mir-17-92 and its upregulation in mutant pro-B cells accounted for the impairment in B-lymphopoiesis 18 .
In this study, we deleted mir-17-92 in mature B cells using CD19-Cre to examine its role in late-stage B-cell differentiation and function. In contrast to its essential role in early B lymphopoiesis in the BM, mir-17-92-deficency does not perturb conventional B-cell populations in the periphery. Interestingly, during T-cell-dependent (TD) antibody response, there is increased PCs in the BM due to elevated expression of Sphingosine 1-phosphate receptor 1 (S1PR1), which mediates BM homing of PCs 19 . We show that five out of six members of mir-17-92 cluster directly target the 3 0 UTR of S1pr1 and removal of one S1pr1 allele could rebalance S1pr1 expression and normalize the enhanced BM homing phenotype of mir-17-92deficient PCs. We also reveal that immunoglobulin G2c (IgG2c) production is impaired in the absence of mir-17-92 and this defect is due to increased expression of the transcription factor IKAROS, which is important for IgG2c production 20 and directly repressed by mir-92a (ref. 21). Furthermore, we demonstrate that mir-17-92 deletion in B cells could reduce IgG2c autoantibody production and attenuate immune complex glomerulonephritis in autoimmune Shp1-deficient mice. Taken together, our findings identify novel roles for mir-17-92 in regulating peripheral B-cell function by fine-tuning different target genes.

Results
B-cell development in mice lacking mir-17-92 in B cells. To address the role of mir-17-92 in peripheral B cells, we deleted loxP-flanked (floxed) mir-17-92 alleles in mice using CD19-Cre that is inefficient with floxed gene deletion in early B cells but mediates efficient gene deletion in mature B lymphocytes 22 . This allows us to bypass the pro-B to pre-B-cell block in mir-17-92 germline-deleted mice 18 to investigate its role in the later stages of B-cell maturation and function.
The frequency of B220 þ IgM þ conventional B cells in the spleen and lymph nodes were also indistinguishable between mir-17-92 bKO and control mice (Fig. 1b). We further observed comparable percentage of follicular B (B220 þ CD93 À CD23 þ CD21 À ), marginal zone B (B220 þ CD93 À CD23 À CD21 þ ) and fraction I, II and III transitional B cells in the spleens of mir-17-92 bKO and control mice (Fig. 1c). B-cell follicles were maintained normally in mir-17-92 bKO mice (Fig. 1d). The mir-17-92 fl/fl alleles were indeed almost completely deleted ( Supplementary Fig. 1e) and the expression of individual miRNAs significantly reduced (480%) in various mutant peripheral B-cell populations ( Supplementary Fig. 1f-h). However, mir-17-92 bKO mice had slightly reduced (B20%) total B-cell numbers in their spleens and lymph nodes compared with control mice (Fig. 1e). As the enhanced expression of pro-apoptotic protein BIM, a known target of mir-17-92, was thought to account for the drastic reduction of mir-17-92-deficient pre-B cells 18 , we examined if the mild reduction of total peripheral B cells in mir-17-92 bKO mice could be ascribed to dysregulated BIM expression. Interestingly, Bim mRNA level was comparable between mir-17-92 bKO and control B cells (Fig. 1f), but BIM protein was enhanced in mir-17-92 bKO B cells (Fig. 1g,h), suggesting that mir-17-92 could modulate BIM expression at the protein level and the subtle decrease in total numbers of mutant peripheral B cells could be due to mildly increased BIM expression. Together, these results suggest that mir-17-92 cluster is not essential for conventional B-cell development in the periphery as it is for early B lymphopoiesis.
By contrast, both the frequency and total B-cell numbers were found to be reduced in the peritoneal cavities of mir-17-92 bKO mice compared with control mice (Fig. 2a,b). The reduction was ascribed to decreased B1-a (CD19 þ CD43 þ CD5 þ ) and B1-b (CD19 þ CD43 þ CD5 À ) cells as their numbers were reduced by approximately seven-and twofold, respectively, whereas conventional B (B2) (CD19 þ CD43 À CD5 À ) cells were not significantly affected. These data indicate that mir-17-92 is required for B1 cell development or maintenance.
Antibody responses in mir-17-92 bKO mice. We determined if mir-17-92 is required for B-cell function by examining antibody responses in mutant mice. When immunized with T-cellindependent (TI) antigen 4-hydroxy-3-nitrophenylacetyl (NP) coupled to Ficoll, which elicits IgM and IgG3 antibody production, mir-17-92 bKO mice exhibited significantly decreased NPspecific IgM and IgG3 antibody production compared with control mice at day 7 and 14 post immunization (Fig. 2c). This defect is consistent with the drastic reduction of B1 cells in these mice, which are the major responders to TI antigens 23 . Thus, mir-17-92 is required for TI antibody response.
We next examined TD antibody response in mir-17-92 bKO mice by challenging them with NP coupled to chicken globulin (NP-CGG). We observed that mir-17-92 bKO and control mice produced similar amounts of NP-specific IgM and IgG1 antibodies, the two major antibody isotypes elicited by NP-CGG immunization, at various time points post immunization (Fig. 3a). The CD19 þ Fas þ CD38 À germinal centre (GC) B cells were comparably present in mir-17-92 bKO and control mice at day 10 post immunization, and class-switched antigen-specific NIP þ IgG1 þ (GC) B cells also appeared to be unperturbed in the mutant mice (Fig. 3b,c). We further stimulated B cells in vitro with anti-CD40 antibody and interleukin (IL)-4 and found IgG1 class switching to be indistinguishable between mutant and control B cells ( Supplementary Fig. 2a). These data indicate that mir-17-92 is largely not required for GC response and the production of TD class-switched antigen-specific IgG1 antibodies.
Increased homing of PCs to BM in mir-17-92 bKO mice. We also examined in mutant mice the generation of memory B cells and long-lived PCs, which are differentiated through the GC response 24,25 . The frequency of NP-specific CD38 þ IgG1 high memory B cells was found to be similar in the spleen of mir-17-92 bKO and control mice at day 56 post immunization, suggesting that mir-17-92 is dispensable for memory B-cell generation and maintenance (Fig. 3d,e). However, when longlived PCs in the BM were examined by enzyme-linked immunospot (ELISPOT) assays, we consistently observed an increase (20-30%) in the frequency of NP-specific IgG1 PCs in mir-17-92 bKO compared with control mice (Fig. 3f).
The increased long-lived PCs in the BM of mir-17-92 bKO mice prompted us to examine PCs in detail at early time points     after immunization. Strikingly, at day 10 post immunization, the frequency of PCs was B2.5-fold higher in the BM of mutant mice compared with controls ( Fig. 4a,b). However, the frequency of splenic PCs from the same mice was not higher but rather reduced (Fig. 4a,b). The increased presence of PCs in the BM and decreased frequency of PCs in the spleen of mir-17-92 bKO mice was evident as early as day 7 post immunization ( Fig. 4c). These results suggest that the increased PCs in the BM of immunized mir-17-92 bKO mice is not due to exaggerated generation of PCs in the spleen but could be due to the enhanced homing of splenic PCs to the BM during immune response. We further assessed if PC differentiation was perturbed in the absence of mir-17-92 by culturing B cells with lipopolysaccharide (LPS) in vitro. After 3 days of LPS stimulation, the percentage of B220 low CD138 þ plasmablasts from mir-17-92 bKO and control B-cell cultures was comparable ( Supplementary Fig. 2b), implying that mir-17-92 deficiency does not affect PC differentiation per se. Taken together, our data suggest that during an acute antibody response, PC differentiation is normal but homing to the BM is enhanced in the absence of mir-17-92.
Mir-17-92 directly targets S1pr1 expression in B cells. The homing of PCs from secondary lymphoid organs to BM is mainly mediated by sphingosine 1-phosphate (S1P) and chemokine CXCL12 through their engagements of S1PR1 (ref. 19) and chemokine-receptor CXCR4 (ref. 26), respectively. To understand how mir-17-92 modulates BM homing of PCs, we searched for possible targets of mir-17-92 using two miRNA target prediction software-TargetScan and MicroCosm, focusing on S1P-S1PR1 and CXCL12-CXCR4 signalling axes, and revealed that S1pr1 is a predicted target gene of mir-17-92 with five out of six members of the cluster (namely mir-17, mir-19a, mir-19b, mir-20a and mir-92a) targeting it.
We next determined if S1pr1 was indeed targeted by mir-17-92. The 3 0 UTR of S1pr1 contains four predicted binding sites with mir-19a and mir-19b targeting two sites and mir-17, mir-20a and mir-92a each targeting one site (Fig. 5a). A luciferase reporter construct with full-length S1pr1 3 0 UTR DNA fragment was cloned downstream of firefly luciferase coding sequence and transfected into HEK293 cells together with either the whole mir-17-92 cluster or individual relevant miRNAs. We observed significantly reduced luciferase activity with the overexpression of either the whole cluster or individual miRNAs (Fig. 5b). We also generated reporter constructs containing individual wild-type (A to D) or mutant (Am to Dm)-binding sites where the DNA sequences complementary to the miRNA seed regions were mutated ( Supplementary Fig. 3). When transfected into HEK293 cells with individual miRNAs, the constructs containing wild-type 3 0 UTR yielded lower luciferase activities compared with the various mutant counterparts (Fig. 5c). Thus, our results suggest that S1pr1 is a bona fide target of mir-17-92, with five out of six members directly targeting its 3 0 UTR at four different sites.
Modulation of S1PR1 expression in mir-17-92-deficient B cells. We proceeded to examine the effect of mir-17-92 deficiency on S1PR1 expression in B cells. First, splenic B220 þ IgM þ B cells were sorted and S1pr1 mRNA was examined by qRT-PCR. We found increased (B30%) S1pr1 mRNA level in mir-17-92 bKO B cells compared with controls ( Fig. 5d), which was correlated with a slight increase in cell surface S1PR1 expression as determined by FACS analyses (Fig. 5e). We also examined S1pr1 mRNA in sorted CD138 þ B220 low plasmablasts differentiated in vitro using LPS and B220 À CD138 þ NIP þ plasmablasts from the spleens of immunized mice. We found that mir-17-92 deficiency led to increase in S1pr1 mRNA (Fig. 5f,h), which was corroborated by increased S1PR1 surface levels in the mutant cells (Fig. 5g,i). Hence, our data suggest that mir-17-92 modulates S1PR1 expression in PCs.
Next we investigated if modulating S1PR1 activity or expression in mir-17-92-deficient PCs could influence their homing to BM. It is difficult to test S1P1 function in vitro due to receptor desensitization caused by non-physiological ligand exposure and the inability to model S1P concentration differential that exists across endothelial cells in vivo 27 . Thus, we examined BM homing   of PCs in immunized mir-17-92 bKO mice in response to different dosages of FTY720 (Fig. 6a), a potent S1P1 inhibitor 28,29 that impedes the BM homing of PCs from secondary lymphoid tissues 19 . We reasoned that higher S1PR1 expression level in mir-17-92-deficient PCs would render them less sensitive to FTY720 inhibition. Consistent with previous report 19 , when mice were given high dose of FTY720 (3.0 mg kg À 1 ), there was hardly any PCs detected in the BM of both mir-17-92 control and bKO mice at day 10 post immunization ( Fig. 6b). Interestingly, when FTY720 dosage was reduced to 1.0 mg kg À 1 , NP-specific PCs were found in the BM of mir-17-92 bKO but not control mice. When the dosage was further lowered to 0.3 mg kg À 1 , we could ARTICLE detect NP-specific PCs in the BM of both mir-17-92 control and bKO mice but with the frequency of PCs significantly higher (approximately threefold) in the latter (Fig. 6b). These data are consistent with our hypothesis that a higher level of S1PR1 expression in mir-17-92-deficient PCs renders them less sensitive to FTY720 inhibition.
Impaired IgG2c antibody production in mir-17-92 bKO mice. During our initial characterization of mir-17B92 bKO mice, we found normal levels of IgM, IgG1, IgG3 and IgG2b, but significantly reduced IgG2c antibodies, in the sera of unchallenged mutant mice compared with controls (Fig. 7a). It is known that inbred mouse strains such as C57BL/6 and SJL with the Igh-1b allele possess only the gene encoding the IgG2c isotype, whereas BALB/c and 129/Sv strains with the Igh-1a allele express only IgG2a (ref. 30). The allele-specific PCR analysis confirmed the presence of IgG2c but not IgG2a gene alleles in mice used in our study ( Supplementary Fig. 4), which have been backcrossed to C57BL/6 background for more than 10 generations, and therefore ruled out the possibility that the reduced serum IgG2c in mutant mice was due to different genetic backgrounds.
We next examined if the production of antigen-specific IgG2c antibodies would be affected during acute antibody response. Interestingly, mir-17-92 bKO mice had markedly reduced NP-specific IgG2c antibodies at various time points post immunization compared with control mice (Fig. 7b). This is in contrast to the normal production of NP-specific IgG1 antibodies in mir-17-92 bKO mice (Fig. 3a). To exclude the possibility that defective IgG2c production in mir-17-92 bKO mice could be due to the modest reduction of peripheral B-cell numbers in these mice, we generated and analysed mir-17-92 fl/fl Aicda Cre/ þ mice in which the floxed mir-17-92 allele was deleted only in antigenactivated but not naïve B cells ( Supplementary Fig. 5a). Unchallenged mir-17-92 fl/fl Aicda Cre/ þ mice had normal B-cell populations in their spleens and lymph nodes compared with mir-17-92 þ / þ Aicda Cre/ þ mice ( Supplementary Fig. 5b,c). NP-specific IgG1 antibodies were also not drastically affected in mir-17-92 fl/fl Aicda Cre/ þ mice, especially at later time points post immunization ( Supplementary Fig. 5d). In contrast, the production of NP-specific IgG2c antibodies was significantly reduced in mir-17-92 fl/fl Aicda Cre/ þ mice (Fig. 7c), suggesting that mir-17-92 is indeed important for IgG2c antibody production.
To further determine if mir-17-92-deficient B cells were intrinsically defective in IgG2c production, we purified B cells from mir-17-92 bKO and control mice and stimulated them in vitro with LPS þ IFN-g, which induces preferential Ig class switching to IgG2c/2a, and assessed the presence of IgG2c þ cells by FACS analysis. After 3 days, mir-17-92 bKO B-cell culture generated significantly reduced IgG2c þ cells (B30% of control; Fig. 7d,e). However, the generation of IgG3 þ and IgG1 þ B cells after LPS and LPS þ IL-4 stimulations, respectively, was comparable between mir-17-92 bKO and control cultures. The switching to IgG2b in response to LPS þ TGFb þ IL-5 stimulation appeared to be reduced but it was not statistically significant (Fig. 7d,e). These data suggest that mir-17-92 is intrinsically required for IgG2c production in B cells.
We explored how mir-17-92 could regulate IgG2c production. As AID is absolutely required for Ig class switch recombination (CSR), we first assessed if the gene expression of Aicda was affected in mir-17-92-deficent B cells using qRT-PCR. We found comparable Aicda expression between stimulated mir-17-92 bKO and control B cells ( Supplementary Fig. 6), thus excluding the possibility that the defective IgG2c production was due to dysregulated AID expression. Next we examined germline transcription (GLT) in IgH switch (S) regions, which is important for AID to access ssDNA targets 31 . When germline transcripts for g3, g1, g2c and g2b in various B-cell cultures were determined by qRT-PCR, we detected marked reduction in g2c GLT in mir-17-92 bKO cells stimulated with LPS þ IFNg compared with controls, whereas g3, g1 and g2b GLT levels were largely comparable between mutant and control cells (Fig. 7f).
Chromatin remodelling is required for many DNA-directed processes, such as transcription, recombination and replication.   (b,c) Luciferase reporter assay validating mir-17-92 targeting of S1pr1. HEK293 cells were transfected with full-length S1pr1 3 0 UTR construct and vectors containing mir-17-92 cluster or individual mir-17-92 members or control vector (b), or with various WT (a-d) or mutant (Am to Dm) S1p1 3 0 UTR constructs and individual miRNAs as depicted (c). Firefly luciferase activity is normalized to renilla luciferase activity (Fluc/Rluc). Results are presented relative to control, set as 1. The Fluc/Rluc for WT 3 0 UTR constructs (a-d) was presented relative to that for corresponding mutant constructs (Am to Dm).
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7764 ARTICLE Acetylation (Ac) on amino-terminal tails of histone H3 and H4 at S regions has been shown to be important for Ig GLT and CSR 32,33 . We therefore performed chromatin immunoprecipitation (ChIP)-qPCR to assess the acetylation of histone H3 (H3Ac) and H4 (H4Ac) at the S region and Cg2c I exon of LPS þ IFNg-treated B cells using specific antibodies. When measured by ChIP-qPCR, mutant cells had lower H3Ac levels were lower at the S region and Cg2c I exon compared with control cells at 48 h after LPS þ IFNg stimulation (Fig. 8a), correlating with decreased IgG2c GLT in mutant B cells (Fig. 7f). Consistent with the normal IgG1 GLT (Fig. 7f), H3Ac levels at Sg1/Ig1 loci were comparable between mutant and control B cells stimulated with LPS þ IL-4 (Fig. 8b). We also examined H4Ac at Cg2c S region, which is known to reflect AID accessibility and activity at S regions and important for CSR 32 . We showed that H4Ac level at Sg2c was lower in LPS þ IFNg-stimulated mir-17-92 bKO cells compared with similarly treated control cells (Fig. 8c), indicating reduced AID accessibility at Sg2c region of mutant cells. However, H4Ac level at the Sg1 region was comparable in LPS þ IL-4-treated mir-17-92 bKO and control cells (Fig. 8d). Taken together, these results suggest that mir-17-92 deficiency affects chromatin remodelling at Cg2c/2a locus and consequently, GLT, AID accessibility and CSR.
As miRNAs negatively regulate gene expression, we hypothesized that mir-17-92 could target negative regulator(s) of IgG2c CSR that likely also regulates chromatin remodelling. Hence, the absence of mir-17-92 would lead to the derepression of negative regulator(s) and result in reduced IgG2c production. It was previously shown that IKAROS deficiency led to increased and ectopic switching to IgG2c/2a (ref. 20). IKAROS was also showed to maintain repressive chromatin at Cg2c/2a and suppress GLT, AID accessibility and CSR of IgG2a/2c (ref. 20). More recently, Ikzf1, the gene encoding IKAROS, was demonstrated to be directly targeted by mir-92a in T cells 21 . Given that mir-17-92deficient B cells exhibited reduced GLT and hypoacetylation of histone at Sg2c region, we postulated that mir-17-92 likely regulated IgG2a/2c switching by modulating IKAROS expression in activated B cells. We found that Ikzf1 was indeed targeted by mir-17-92 or mir-92a but not by other members of the cluster using luciferase reporter assays ( Supplementary Fig. 7). We demonstrated that mir-17-92-deficient B cells had higher (B60%) Ikzf1 mRNA levels than control cells when treated with LPS þ IFNg (Fig. 8d). Consistent with increased Ikzf1 mRNA, the protein level of IKAROS was also enhanced in stimulated mutant cells as determined by western blotting (Fig. 8e,f). These results together indicate that mir-17-92 directly targets Ikzf1 and modulates IKAROS expression in B cells. We next asked if elevated IKAROS expression could account for the defective IgG2c production in mir-17-92-deficient B cells. We used short interfering RNA (siRNA) to knock-down IKAROS in the mir-17-92-deficient cells with nucleofection, which can efficiently transfect activated B cells in vitro ( Supplementary  Fig. 8). Western blotting analyses showed that IKAROS protein level was noticeably reduced in mutant B cells transfected with Ikzf1-targeting siRNA compared with control siRNA (Fig. 9a) Figure 6 | Effects of FTY720 inhibition and S1pr1 haploinsufficiency on mir-17-92 bKO PCs. (a) Timeline of FTY720 treatment and immunization of mice. Mice were immunized with NP 38 -CGG at day 0 and given FTY720 at different dosages at day 4 and day 7 before analyses at day 10. (b) Frequency of NP-specific IgG1 ASCs in the BM of FTY720-treated mice at day 10 after immunization. (c) Reduced S1pr1 expression by S1pr1 haploinsufficiency. S1pr1 mRNA in mir-17-92 fl/fl S1pr1 fl/ þ Cd19 Cre/ þ B cells was determined by qRT-PCR and presented relative to that in mir-17-92 þ / þ S1pr1 þ / þ Cd19 Cre/ þ B cells. (d,e) NP-specific IgG1 ASCs in the BM of mir-17-92 þ / þ S1pr1 þ / þ Cd19 Cre/ þ , mir-17-92 fl/fl S1pr1 þ / þ Cd19 Cre/ þ and mir-17-92 fl/fl S1pr1 fl/ þ Cd19 Cre/ þ mice at day 10 post immunization. NP-specific IgG1 ASCs were determined by ELISPOT assays, shown as images of assays (d) and frequency (e). Unpaired two-tailed Student's t-test was used for statistical analyses. *Po0.05. **Po0.01. Each symbol represents one mouse and small horizontal lines indicate the mean (b,e). Bar graphs represent mean±s.d. (n ¼ 3). Data are representative of three (c) or five (d) experiments. (Fig. 9b,c). We further demonstrated that introduction of mir-92a mimics but not others could enhance IgG2c production in mir-17-92-deficient and control B cells (Fig. 9d). Taken together, our data suggest that mir-92a from mir-17-92 cluster directly represses IKAROS expression and regulates IgG2c production.
Mir-17-92 deficiency attenuates murine autoimmunity. IgG2c/ 2a is the most prevalent autoantibody isotype found in murine lupus 34,35 . These autoantibodies bind self DNA and ribonucleoproteins to form immune complexes that cause tissue damage. The systemic autoimmunity in mir-17-92-transgenic mice 36,37 and selective IgG2c defect in mir-17-9 fl/fl Cd19 Cre/ þ mice led us to explore if mir-17-92 ablation could attenuate B-cell-mediated autoimmunity. Hence, we introduced mir-17-92 deficiency into Ptpn6 fl/fl Cd19 Cre/ þ mice that had Shp1 tyrosine phosphatase deleted in B cells and developed autoimmunity resembling that of human lupus 38 .
These results suggest that the ablation of mir-17-92 could potentially suppress IgG2c autoantibody production and therefore alleviate symptoms of systemic autoimmunity in mice.

Discussion
We examined the role of mir-17-92 in late-stage B-cell differentiation and function by ablating it using CD19-cre. We show that mir-17-92 is not absolutely required for conventional B-cell maturation in the periphery in contrast to its essential role in early B lymphopoiesis. However, we unveil two novel roles for mir-17-92 in peripheral B cells, namely the modulation of PC homing to the BM and selective regulation of IgG2c production. We further identify S1PR1 and IKAROS as direct physiological targets of mir-17-92 responsible for these two outcomes.
Previous studies have postulated that mir-17-92 could largely account for Dicer's function in early B-lymphoiesis by targeting BIM 11,18 . Here we demonstrate that mir-17-92 is largely dispensable for the maturation of conventional B cells in the periphery, suggesting that it plays a more important role in early B-lymphoiesis than in peripheral B-cell maturation. However, we demonstrate that mir-17-92 is required for B1 cell development and TI immune response. We also show that BIM protein but not mRNA level is increased in mir-17-92-deficient B cells and could account for the modestly decreased total peripheral B cells in mir-17-92 bKO mice. Previous studies of mir-17-92-deficient oroverexpressing B cells 11,39 have demonstrated either enhanced or decreased BIM protein expression in mutant B cells, but Bim mRNA level was not examined in those studies. Our results suggest that mir-17-92 regulation of BIM mainly occurs at the protein level.
In contrast to the absolute requirement for Dicer in GC response 13 , mir-17-92 bKO mice have normal NP-specific IgM and IgG1 antibody production and unperturbed GC during TD antibody response, suggesting that miRNAs other than mir-17-92 are important for these processes. In support of this view, a previous miRNA deep-sequencing study has shown that mir-17-92 family members are not amongst the most highly expressed miRNAs in GC B cells 40 .
Interestingly, we show that mir-17-92 is important for modulating PC homing to BM where survival niches are provided for their long-term persistence [41][42][43] . We show that five out of six members of mir-17-92 cluster target the 3 0 UTR of S1pr1 at four different sites and the absence of mir-17-92 leads to higher S1pr1 expression in PCs. It is possible that S1pr1 could be synergistically and simultaneously targeted by these five miRNAs in PCs. The enhanced BM homing of PCs in mir-17-92 bKO mice can be normalized by removing one copy of S1pr1 gene. Thus, our data suggest that mir-17-92-mediated fine-tuning of S1PR1 expression is another means to control BM homing of PCs, and this insight could be useful for the development of miRNA-based drugs to interfere with PC homing to BM survival niches in PC-related diseases such as multiple myeloma and autoimmunity.
The mir-17-92-mediated modulation of S1PR1 expression and enhanced homing of PCs in mutant mice raise the possibility that the slightly decreased peripheral B cells in mutant mice might be due to increased egress of B cells from secondary lymphoid organs into circulation. However, our analyses showed that the distribution of naïve B cells in blood and secondary lymphoid organs was comparable between mir-17-92 bKO and control mice ( Supplementary Fig. 9), suggesting that the mildly increased S1PR1 level in mutant naïve B cells does not significantly shift the equilibrium distribution of B cells in favour of blood. The noticeable homing phenotype of mutant PCs but not naïve B cells could be due to the different expression levels of mir-17-92, which is upregulated in PCs ( Supplementary Fig. 10), or the differential expression of other S1PR1-targeting miRNAs in these cells.
We also show that mir-17-92 deficiency in B cells impairs IgG2c antibody production, which is accompanied by abnormal GLT and chromatin remodelling at Cg2a/2c locus. We further demonstrate that mir-17-92 can inhibit the expression of IKAROS, the transcription factor known to maintain repressive chromatin at Cg2C/2a, and therefore suppress transcription, AID accessibility and CSR of IgG2a/2c 20 . It was recently demonstrated that IKAROS expression in B cells could be modulated by the transcription factor IRF5 (ref. 44). Thus, our current study uncovers a post-transcriptional layer of IKAROS regulation by miRNAs. Interestingly, in IKAROS hypomorphic mice that have 10-fold less IKAROS protein than wild-type mice 20 , the production of IgG2b was also enhanced. However, in our study, we did not observe a significant change in IgG2b production in mir-17-92-deficient B cells. This could be due to the mild (approximately twofold) increase in IKAROS expression in these cells. In agreement with this viewpoint, IgG2b was also not affected in IRF5-deficent B cells, which manifested modest changes in IKAROS expression 44,45 . It is possible that IgG2c/2a CSR is more sensitive to subtle changes in IKAROS levels than that of other Ig isotypes.
Another interesting finding from our study is that mir-17-92 deletion in B cells can significantly reduce the amount of pathogenic IgG2c anti-DNA antibodies in autoimmune Shp1-deficient mice. In addition, IgG immune complex glomerulonephritis in Shp1-deficient mice could be attenuated by mir-17-92 deficiency. Murine IgG2c/IgG2a and its analogue isotype IgG1 in humans are the most prevalent and potent Ig isotypes found in autoimmune diseases [46][47][48][49] . Therefore, our results suggest that mir-17-92 could be a possible target for therapeutic intervention by providing an avenue to control pathogenic IgG2c autoantibody production in lupus.

Methods
Mice. Mir-17-92 fl/fl , S1pr1 fl/fl , Ptpn6 fl/fl and Cd19 Cre/ þ mice were obtained from The Jackson Laboratory. Aicda Cre/ þ mouse was generated and described previously 13 and maintained in C57BL/6 background. Mir-17-92 fl/fl mice were backcrossed 410 generations into C57BL/B background before intercrossing with Cd19 Cre/ þ and Aicda Cre/ þ mice to generate mir-17-92 fl/fl Cd19 Cre/ þ and mir-17-92 fl/fl Aicda Cre/ þ mice. Genotyping of IgG2a and IgG2c alleles was performed by PCR using allele-specific primers as described 30 previously. Mice of 10 to 16 weeks old of both genders were used according to the protocols approved by the Institutional Animal Care and Use Committees (IACUCs) based on guidelines from the National Advisory Committee on Laboratory Animal Research (NACLAR) of Singapore.
Immunofluorescence histology. Splenic and renal sections of 8 mm thickness were prepared from tissues frozen in Tissue-Tek (Sakura) as described previously 13 . After blocking, sections were stained at room temperature for 30 min with a cocktail of antibodies and mounted with CYTOSEAL 60 (Electron Microscopy Sciences) and analysed on a FV1000 confocal microscope (Olympus).
ELISA and ELISPOT assay. Sera were collected from unchallenged or immunized mice on various days. Total or NP-specific antibodies of various isotypes in sera were measured via enzyme-linked immunosorbent assay (ELISA). The relative titres of NP-specific serum antibodies are shown as the relative of dilutions that give absorbance reading within the linear range of the assay 51 . The frequency of NP-specific IgG1 antibody-secreting cells was determined by ELISPOT assay 13 .
Plasmid construction and luciferase reporter assay. Mir-17, mir-18a, mir-19a, mir-19b, mir-20a, mir-92a and mir-17-92 cluster were cloned into MSCV retroviral vector. The 3 0 UTR fragments of mouse S1pr1 and Ikzf1 with predicted binding sites for individual miRNAs and their mutants were amplified from genomic DNA and inserted into pmirGLO plasmid (Promega) and confirmed by sequencing. For luciferase reporter assay, HEK 293 cells were plated into 48-well plates at 1 Â 10 5 cells per well 16 h before transfection. pmirGLO (0.2 mg) reporter plasmid containing the S1pr1 and Ikzf1 3 0 UTR fragments or their mutants and 0.2 mg different miRNA-expressing plasmids were transfected into each well using Lipofectamine 2000 transfection reagent (Invitrogen) in triplicate. Luciferase assays were performed 48 h after transfection using the Dual-Luciferase Reporter Assay System (Promega) on a SpectraMax Luminescence Microplate Reader (Molecular Devices). Luciferase activity was expressed as ratio between firefly and renilla luciferase activities.
Cell culture and nucleofection of B cells. Splenic B cells were purified with anti-CD43 mAb-coupled magnetic beads using magnetic-activated cell sorting negative selection protocol (Miltenyi Biotech, Bergisch Gladbach, Germany). Cells were stimulated for 3 days with 10 mg ml À 1 LPS (Sigma) in DMEM supplemented with the necessary additives. IL-4 (10 ng ml À 1 ), IL-5 (5 ng ml À 1 ), IFNg (10 ng ml À 1 ) and TGFb (2 ng ml À 1 ) from R&D Systems were used to induce preferential Ig class switching to various isotypes. For nucleofection, purified B cells were stimulated with 10 mg ml À 1 LPS and 10 ng ml À 1 IFNg overnight and transfected with siRNA or miRNA mimics from Life Technologies using P4 Primary Cell 4D-Nucleofector X Kit on an Amaxa 4D-nucleofector (Lonza). Transfected cells were recultured and analysed 48 h later.
Western blotting. Cells were lysed and protein concentrations were determined as described previously 52 . Equal amount of proteins was loaded and subjected to SDS-polyacrylamide gel electrophoresis, transferred and blotted with specific antibodies. BIM-specific antibody (C34C5) was from Cell Signalling. The IKAROSspecific antibody (M20) was from Santa Cruz. Western blot images shown in Figs 1, 8 and 9 have been cropped for presentation. Full-sized images are presented in Supplementary Fig. 11.
ChIP-qPCR. ChIP was performed using EZ-ChIP kit (Millipore) according to the manufacturer's protocol. In brief, 1 Â 10 7 stimulated B cells were crosslinked in 1% formaldehyde. Chromatin was sonicated to 500-1,000 bp using Misonix Sonicator 3000 Ultrasonic Cell Disruptor (QSonica). Sonicated chromatin was precleared with Protein G Agarose and subsequently subjected to immunoprecipitation with anti-H3Ac, anti-H4Ac or control antibodies (Millipore). One per cent of the precleared chromatin was saved as input. Protein/DNA complexes were eluted and reverse crosslinked to free DNA, which was further purified using spin columns (Millipore). The input DNA was diluted 10 times. Two ml of DNA samples was analysed by qPCR using Ig1, Sg1, Ig2c, Sg2c and Gapdh primers 20,53 . The acetylation index was calculated described previously 20 .
Statistics. Unpaired two-tailed Student's t-test was used for statistical analyses.