Hypoxia-inducible factor-1α is a critical transcription factor for IL-10-producing B cells in autoimmune disease

Hypoxia-inducible factors (HIFs) are key elements for controlling immune cell metabolism and functions. While HIFs are known to be involved in T cells and macrophages activation, their functions in B lymphocytes are poorly defined. Here, we show that hypoxia-inducible factor-1α (HIF-1α) contributes to IL-10 production by B cells. HIF-1α regulates IL-10 expression, and HIF-1α-dependent glycolysis facilitates CD1dhiCD5+ B cells expansion. Mice with B cell-specific deletion of Hif1a have reduced number of IL-10-producing B cells, which result in exacerbated collagen-induced arthritis and experimental autoimmune encephalomyelitis. Wild-type CD1dhiCD5+ B cells, but not Hif1a-deficient CD1dhiCD5+ B cells, protect recipient mice from autoimmune disease, while the protective function of Hif1a-deficient CD1dhiCD5+ B cells is restored when their defective IL-10 expression is genetically corrected. Taken together, this study demonstrates the key function of the hypoxia-associated transcription factor HIF-1α in driving IL-10 expression in CD1dhiCD5+ B cells, and in controlling their protective activity in autoimmune disease.

B cells are traditionally known for their effector function involved in antigen presentation and antibody secretion upon their differentiation into plasmablasts and plasma cells conferring humoral immunity 1 . However, increasing attention has been directed to the immune regulatory function of B cells 2,3 . This regulatory function is associated with their production of anti-inflammatory cytokines such as IL-35, TGF-β, and in particular IL-10 [4][5][6] . Previous studies have shown that CD1d hi CD5 + B cells, transitional 2-marginal zone precursors (T2-MZP; CD23 hi CD21 hi IgM + ), antibody-secreting cells (CD44 hi CD138 + plasmablasts), and peritoneal CD5 + B1a cells can, through the production of IL-10, suppress pathogenic T cells and inhibit autoimmune inflammatory diseases such as experimental autoimmune encephalomyelitis (EAE), arthritis, and colitis, as well as contact hypersensitivity [7][8][9][10] .
Hypoxia-inducible factors (HIFs) are heterodimeric transcription factors, consisting of an oxygen-labile alpha subunit (HIF-α) and a constitutively stable beta subunit (HIF-β), that exert pivotal roles in inducing cellular responses to hypoxia 11 . While hypoxia causes alpha subunits stabilization and induction of respective target genes, HIF-1α and HIF-2α are hydroxylated by prolyl hydroxylases (PHD) and degraded after binding protein von Hippel Lindau (pVHL) under normoxic conditions 12,13 . HIFs were shown to be involved in innate and adaptive immune activation. In macrophages, HIF-1α increases cell motility and the expression of pro-inflammatory cytokines 14,15 . In adaptive immunity, HIF-1α has been shown to promote Th17 cell development and to enhance the expression of cytolytic molecules such as granzyme B and perforin in CD8 T cells 16,17 . The function of HIFs in B cells, however, is incompletely determined. Interestingly, abnormalities of peritoneal B1 cells and high levels of IgG and IgM antibodies directed against dsDNA have been described in Hif1a-deficient chimeric mice 18 , suggesting a possible regulation of B cell functions by HIF-1α.
In this study, we delineate the function of HIFs in B cells during autoimmune disease with a particular interest in IL-10producing CD1d hi CD5 + B cells. We generated B cell-specific Hif1a or Hif2a mutant mice to test the influence of HIFs on B cell cytokine production and on the course of autoimmune disease. B cell activation through B cell antigen receptor (BCR) induces upregulation of HIF-1α expression, and B cell-specific ablation of Hif1a, but not Hif2a, impairs IL-10 production by B cells. HIF-1α transcriptionally regulates Il10 gene expression in cooperation with phosphorylated-STAT3, and is required to establish the glycolytic metabolism driving CD1d hi CD5 + B cells expansion. Furthermore, compared with wild-type (WT), mice lacking HIF-1α in B cells have exacerbated collagen-induced arthritis (CIA) and EAE, which can be rescued by ectopic expression of IL-10 in Hif1a-deficient CD1d hi CD5 + B cells and their adoptive transfer in vivo. Our findings reveal HIF-1α as a critical transcription factor for IL-10 production by B cells. HIF-1α expression controls CD1d hi CD5 + B cells expansion and may be considered as a potential target in autoimmune disease.

HIF-1α expression increases in activated B cells.
To investigate the role of HIFs in B cells, the expression of HIF-1α and HIF-2α was determined in C57BL/6 WT splenic B cells stimulated with lipopolysaccharide (LPS) (10 μg/ml) or anti-IgM (10 μg/ml). Even under normoxic conditions, Hif1a mRNA expression is induced in B cells stimulated with LPS or anti-IgM (Fig. 1a), whereas the expression of Hif2a is almost undetectable and remains unchanged when analyzed in fold change (Fig. 1a). Accordingly, HIF-2α protein is hardly detectable, whereas HIF-1α protein increases at 4, 8, and 12 h after LPS or anti-IgM stimulation in B cells (Fig. 1b). Since HIF-1α induction by LPS has been already reported to be dependent on NF-κB signaling 19 , we also checked whether this pathway is effective in B cells. Indeed, knockdown of RelA not only decreases p65 phosphorylation but also HIF-1α protein level in B cells stimulated by LPS for 4 h (Supplementary Fig. 1a).
Since B cells stimulation by anti-IgM also induces HIF-1α (Fig. 1b), we delineated the pathways of HIF-1α induction in BCR-stimulated B cells. Therefore, ERK and STAT3 proteins levels were analyzed. As shown in Fig. 1c, phosphorylated-ERK (pERK) and phosphorylated-STAT3 Ser727 (pSTAT3 727 ) are increased in splenic B cells after anti-IgM stimulation, whereas phosphorylated-STAT3 Tyr705 (pSTAT3 705 ) is virtually undetectable. Using specific inhibitor of ERK, STAT3, and AKT pathways, which are not affecting B cell viability ( Supplementary  Fig. 1b), we analyzed the pathway essential for HIF-1α protein expression. Indeed, HIF-1α protein induction is suppressed in a dose-dependent manner when BCR-stimulated B cells are treated with ERK or STAT3 inhibitors, but not AKT inhibitor treatment (Fig. 1d). Similarly, decrease of HIF-1α protein is observed when STAT3 or ERK are knocked down in B cells using siRNA approach ( Supplementary Fig. 1c). Interestingly, STAT3 727 phosphorylation is decreased after B cell treatment with ERK inhibitor (Fig. 1d), suggesting that phosphorylation of ERK is essential for STAT3 727 phosphorylation.
Next, we determined whether pSTAT3 727 could also transcriptionally regulate Hif1a gene expression. To do so, chromatin immunoprecipitation (ChIP) analysis was performed on a putative STAT3 binding site on Hif1a promoter at −309 bp/ −319 bp from the transcription starting site (TSS) (Fig. 1e). Indeed, low level of pSTAT3 727 can bind to Hif1a promoter in splenic B cells in homeostasis (Fig. 1e). Interestingly, pSTAT3 727 binding on Hif1a promoter is strikingly enhanced in BCRmediated activated B cells (Fig. 1e). Our results demonstrate that HIF-1α is increased at mRNA and protein levels in LPS-treated B cells via the NF-κB pathway and in BCR-stimulated B cells via ERK-STAT3 activation.
B1a population is reduced in Mb1 cre Hif1a f/f mice. To determine the roles of HIF-1α and HIF-2α during B cell development in vivo, we bred mice carrying a loxP-flanked Hif1a or Hif2a allele with mice expressing cre recombinase from the Mb1 promoter to delete Hif1a or Hif2a specifically in B lymphocytes (referred to herein as Mb1 cre Hif1a f/f or Mb1 cre Hif2a f/f mice). As expected, HIF-1α or HIF-2α protein is completely abolished in splenic B cells but not in T cells isolated from Mb1 cre Hif1a f/f or Mb1 cre Hif2a f/f mice compared with WT control mice (Supplementary Fig. 1d). Next, flow cytometric analysis of the B cell subpopulations in Mb1 cre Hif1a f/f , Mb1 cre Hif2a f/f , and WT control mice were performed (Fig. 2a). No difference can be detected in the populations of pre-pro-B, pro-B, pre-B, immature, and recirculating B cells (Hardy fractions A-F) in WT and mutant mice (Fig. 2b). The splenic B cell subpopulations, transitional type 1 and 2 as well as follicular cells, are also similar in Mb1 cre Hif1a f/f , Mb1 cre Hif2a f/f , and control mice (Fig. 2c, d). However, percentage and absolute numbers of marginal zone B cells are moderately decreased in Mb1 cre Hif1a f/f mice compared to WT mice (Fig. 2d). Next, we analyzed peripheral B cell subsets in inguinal lymph nodes and blood from Mb1 cre Hif1a f/f , Mb1 cre Hif2a f/f , and WT mice. HIF-1α or HIF-2α deletion does not alter the immature and mature B cell populations in the periphery (Fig. 2e, f). Interestingly, only B1a cell number is drastically decreased in the peritoneum of Mb1 cre Hif1a f/f mice when compared to WT or Mb1 cre Hif2a f/f mice, whereas no difference is observed for B1b cells (Fig. 2g).
To further determine the effects of HIF-1α and HIF-2α on B cell functions in vitro, proliferation and apoptosis rates were examined in splenic B cell stimulated with LPS, anti-CD40, or anti-IgM. Of note, Hif1a-or Hif2a-deficient splenic B cells have a normal proliferative or survival ratio after stimulation (Supplementary Fig. 2a, b). We also examined T cell independent (TI) antibody responses and T cell dependent (TD) antibody responses in Mb1 cre Hif1a f/f and Mb1 cre Hif2a f/f mice. Antigenspecific antibody production is similar in Mb1 cre Hif1a f/f or Mb1 cre Hif2a f/f mice and WT controls, indicating that HIFs are not essential for TI or TD antibody responses ( Supplementary  Fig. 2c-h). Altogether, these data show that HIF-2α has no essential role during B cell development, whereas HIF-1α is important for the B1a population in the peritoneum.  increased level of HIF-1α protein in IL-10 + B cells (Supplementary Fig. 3a). Because IL-10-producing B cells have been described in different B cell subpopulations such as CD1d hi CD5 + CD19 + B cells 22 and CD23 + CD21 hi IgM + (T2-MZP) B cells 23 , we speculated that these two subsets are modified in Mb1 cre Hif1a f/f mice in vivo. It is noteworthy that percentage and absolute numbers of CD1d hi CD5 + or T2-MZP B cells are reduced in Mb1 cre Hif1a f/f mice compared to WT littermates ( Fig. 3b and Supplementary  Fig. 3b). Less BrdU-positive cells are observed in CD1d hi CD5 + and T2-MZP populations of Mb1 cre Hif1a f/f mice compared to WT control mice ( Fig. 3c and Supplementary Fig. 3c), implying a defect of regulatory B cell proliferation in Hif1adeficient mice. In accordance, IL-10 intracellular staining in CD1d hi CD5 + B cells confirms the decreased frequency of IL-10 + CD1d hi CD5 + B cells in Mb1 cre Hif1a f/f mice (Fig. 3d). Moreover, analysis of anti-inflammatory cytokines expression in sorted CD1d hi CD5 + B cells from Mb1 cre Hif1a f/f mice reveals a significant decrease in Il10 mRNA expression as well as a decreased IL-10 production after stimulation (*P < 0.05 and **P < 0.01, by t-test; Fig. 3e and Supplementary Fig. 3d), whereas Tgfb, P35, and Ebi3 mRNA levels are not altered ( Supplementary  Fig. 3d). Altogether, these data suggest that HIF-1α is an important factor for the expansion of CD1d hi CD5 + B cells, and their IL-10 production.
HIF-1α regulates glycolysis in CD1d hi CD5 + B cells. Since CD1d hi CD5 + B cell number is normal in Il10-deficient mice 24 , we hypothesized that the reduced IL-10 level is likely not responsible for the reduced CD1d hi CD5 + B cell number in Mb1 cre Hif1a f/f mice. HIF-1α was previously identified as a key factor for glycolytic activity and glucose metabolism in immune cell function and proliferation [25][26][27] . To further dissect the expansion of CD1d hi CD5 + B cells, we examined the level of HIF-1α in this population. Indeed, HIF-1α protein level is higher in CD1d hi CD5 + B cells than in CD1d lo CD5 − B cells ( Supplementary  Fig. 4a). Next, glucose uptake was examined in FACS-sorted CD1d lo CD5 − B and CD1d hi CD5 + B cells from WT mice. CD1d hi CD5 + B cells display a two-fold increase in glucose transport activity compared to CD1d lo CD5 − B cells (Fig. 4a), suggesting that CD1d hi CD5 + B cells preferentially use glucose metabolism. Moreover, CD1d hi CD5 + B cells from Mb1 cre Hif1a f/f mice exhibit a lower level of glucose uptake and lactate secretion (Fig. 4b, c) compared to CD1d hi CD5 + B cells from WT mice, whereas there is no difference in glucose uptake between Hif1adeficient and WT CD1d lo CD5 − B cells ( Supplementary Fig. 4b).
Accordingly, mRNAs expression of HIF-1α-targeted glycolytic genes, glucose transporter 1 (Glut1), pyruvate kinase M2 (Pkm2), hexokinase 2 (Hk2), lactate dehydrogenase A (Ldha), phosphoinositide-dependent kinase 1 (Pdk1), and glucose-6phosphate isomerase 1 (Gpi1), are markedly decreased in Hif1adeficient CD1d hi CD5 + B cells compared to WT CD1d hi CD5 + B cells (Fig. 4d). Next, we delineated whether the high glycolytic activity of CD1d hi CD5 + B cells was critical for the expansion of CD1d hi CD5 + B cells. As shown in Fig. 4e, partial inhibition of glycolysis by treatment with competitive glycolytic inhibitor 2deoxyglucose is sufficient to inhibit WT CD1d hi CD5 + B cells proliferation to a similar level as found in untreated Hif1a-deficient CD1d hi CD5 + B cells. Taken together, these data suggest that HIF-1α expression controls the expansion of CD1d hi CD5 + B cells by orchestrating their high glycolytic activity.
HIF-1α and STAT3 cooperatively regulate Il10 transcription. To delineate how HIF-1α can regulate IL-10 expression in B cells, splenic B cells from Mb1 cre Hif1a f/f and WT mice were cultured under normoxic or hypoxic condition. Interestingly, Il10 mRNA expression is strongly increased in B cells cultured under hypoxia compared to normoxia (Fig. 5a). Consistent with the reduced IL-10 production in Hif1a-deficient B cells (Fig. 3), Il10 mRNA expression is also lower in Hif1a-deficient B cells than WT B cells under hypoxic condition (Fig. 5a). We next examined whether Il10 gene expression could be transcriptionally regulated by HIFs in B cells. Bio-informatics promoter analysis, using JASPA with the consensus core (A/GCGTG), reveals several putative hypoxiaresponsive element (HRE) regions (I-V) on Il10 promoter ( Fig. 5b and Supplementary Fig. 5a). By ChIP assay, we show that HIF-1α can bind to HRE I and HRE II regions under hypoxic condition (Fig. 5c). Interestingly, the pattern of HIF-1α binding is similar to that of histone H3 (trimethylK4) antibodies, whereas no specific binding is detected when using control IgG antibodies ( Fig. 5d and Supplementary Fig. 5b), suggesting that these regions are transcriptionally active under hypoxia. Next, luciferase reporter assays with putative HRE constructs were performed in 293T cells after hypoxic or normoxic culture. As expected, the luciferase activity of the HRE I and HRE II constructs are increased under hypoxic condition, suggesting that HIF-1α activates Il10 transcription through HRE I and HRE II regions (Fig. 5e).
Since STAT3 and HIF-1α were previously shown to cooperate on HIF target genes such as CA9 and PGK1 28 , we hypothesized that the highly expressed pSTAT3 727 in BCR-activated B cells ( Fig. 1c) might form a complex with HIF-1α to activate Il10 transcription. To test this hypothesis, we confirmed the binding of HIF-1α in B cells after anti-IgM stimulation (Fig. 5f). In addition, HIF-1β can also bind to the HRE I and HRE II regions on the Il10 promoter in B cell after anti-IgM stimulation, whereas no binding of HIF-2α or control IgG is detected (Supplementary Next, co-immunoprecipitation of pSTAT3 727 and HIF-1α was performed. Indeed, pSTAT3 727 protein binds to HIF-1α protein in BCR-stimulated B cells (Fig. 5g). Furthermore, twostep ChIP assays pulling-down HIF-1α and pSTAT3 727 sequentially show that pSTAT3 727 could also bind to the HRE I and HRE II regions on Il10 promoter (Fig. 5h), implying that a complex comprising HIF-1α and pSTAT3 727 might be involved in IL-10 production by BCR-stimulated B cells.
HIF-1α deficiency in B cells exacerbates autoimmune diseases. IL-10 production by B cells was previously shown to influence the course of inflammatory autoimmune diseases 6 . Therefore, we hypothesized that HIF-1α in B cells represented a critical node for the modulation of autoimmune diseases. To test this hypothesis, Mb1 cre Hif1a f/f mice were subjected to CIA, a standard murine model of arthritis resembling human rheumatoid arthritis. As shown in Fig. 6a, Mb1 cre Hif1a f/f mice show a significantly increased incidence of arthritis after immunization with collagen II (CII) compared to littermate controls (*P < 0.05, by Kaplan-Meier analysis with log-rank test). The induction of arthritis is dependent on CII immunization, since no clinical symptom is observed in Mb1 cre Hif1a f/f mice without immunization ( Fig. 6a, b). Moreover, Mb1 cre Hif1a f/f mice exhibit an earlier disease onset and develop higher clinical arthritis scores than WT mice after CII immunization (Fig. 6b). Accordingly, Mb1 cre Hif1a f/f mice have an increased paw thickness, synovial inflammation, bone erosions, and number of osteoclasts, confirming an exacerbation of arthritis symptoms in mutant mice (Fig. 6c, d). Similar to TD or TI antibody responses ( Supplementary Fig. 2), levels of IgG, IgG1, IgG2a, and IgG2b are not changed in Mb1 cre Hif1a f/f and WT arthritic mice ( Supplementary Fig. 6a). Next, cytokines mRNA expression pattern was analyzed in synovial tissues. As   expected, an increased level of pro-inflammatory cytokines such as Tnf, Ifng, Il17, Il1b mRNA and a reduced level of Il10 mRNA are detected in synovial tissue of Mb1 cre Hif1a f/f mice compared to WT mice after CIA induction (Fig. 6e). Furthermore, the levels of pro-inflammatory cytokines like IL-17 and IFN-γ are higher in the supernatant of CII-stimulated splenocytes from Mb1 cre Hif1a f/ f mice than WT mice (Fig. 6f). Whereas no change in TGF-β is detected, the level of IL-10 is reduced in CII-stimulated splenocytes and splenic B cells from Mb1 cre Hif1a f/f mice (Fig. 6f, g).
To further confirm the physiological role of HIF-1α in B cells mediated by IL-10 production, the myelin oligodendrocyte glycoprotein peptide (MOG  induced EAE was applied to Mb1 cre Hif1a f/f mice and WT littermates. After immunization with MOG  , Mb1 cre Hif1a f/f mice develop higher clinical scores than WT mice (Fig. 7a). Histopathological analyses reveal an increased number of inflammatory foci and demyelination areas in the spinal cords of Mb1 cre Hif1a f/f mice (Fig. 7b). In addition, increased numbers of infiltrated CD4 + , CD8 + T cells, and F4/80 + macrophages are observed in the central nervous system (CNS) of Mb1 cre Hif1a f/f mice compared to WT littermates (Fig. 7c). Like for the CIA model, EAE pathogenesis is associated to Th17, Th1 cells, and production of pro-inflammatory cytokines like IL-17 and IFN-γ. At the peak of EAE disease, increased levels of IL-17 and IFN-γ and a reduced level of IL-10 are found in serum from Mb1 cre Hif1a f/f mice compared to WT littermates ( Supplementary  Fig. 7a). Furthermore, after in vitro re-stimulation with MOG 35-55 for 48 h, splenocytes and splenic B cells from Mb1 cre Hif1a f/f mice produce a reduced level of IL-10 when compared to WT cells (Fig. 7d, e). However, there is no difference in TGF-β and IL-35 production by splenic B cells from Mb1 cre Hif1a f/f mice after restimulated by MOG   (Fig. 7e). Next, Th1 (IFN-γ + CD4 + ) cells, Th17 (IL-17 + CD4 + , IL-23R + IL-17 + , or GM-CSF + IL-17 + ) cells, Treg (CD25 + Foxp3 + CD4 + ) cells, type 1 regulatory T cells (Tr1) (IL-10 + CD4 + ) cells, IL-10 + Foxp3 + T cells as well as IL-10 + B cells, ICAM + B cells, CD73 + B cells, GITRL + B cells, FasL + B cells, and PD-L1 + B cells were quantified in the spleen, dLNs, and CNS by FACS. No difference is detected in the ICAM + , CD73 + , GITRL + , FasL + , and PD-L1 + B cell populations in spleen and dLNs ( Supplementary Fig. 7b, c). Th1 and Th17 populations, including pathogenic IL-23R + IL-17 + and GM-CSF + IL-17 + Th17 cells are increased, whereas Tr1, IL-10 + Foxp3 + Treg cells, and IL-10producing B cells are decreased in Mb1 cre Hif1a f/f mice compared to WT mice (Fig. 7f-i and Supplementary Fig. 7d    our results demonstrate that HIF-1α expression in B cells has a crucial protective function in autoimmune diseases.

Impaired suppressive function of Hif1a-deficient B cells.
To determine whether the phenotypes of Mb1 cre Hif1a f/f mice in autoimmune diseases, including the observed exacerbated proinflammatory T cell response, were secondary to a defect in IL-10 production by B cells, naïve CD4 T cells were co-cultured with CD1d hi CD5 + B cells from WT mice in presence or absence of anti-IL-10 antibody. Indeed, CD4 T cells co-cultured with CD1d hi CD5 + B cells in presence of anti-IL-10 antibody shows higher Th1, Th17 and lower Tr1 polarization than the ones cultured in the absence of anti-IL-10 antibody (Fig. 8a). Next, naïve CD4 T cells were cocultured with CD1d hi CD5 + B cells sorted from Mb1 cre Hif1a f/f mice and WT littermates under T cell-polarizing conditions. Naïve CD4 T cells co-cultured with CD1d hi CD5 + B cells from Mb1 cre Hif1a f/f mice (CD1d hi CD5 + (ΔHif1a)) show higher polarization into Th1 (IFN-γ + CD4 + ) and Th17 (IL-17 + CD4 + ) cells than their counterparts co-cultured with sorted CD1d hi CD5 + B cells from WT mice (CD1d hi CD5 + (WT)) (Fig. 8b). These data suggest that the suppressive function of Hif1a-deficient CD1d hi CD5 + B cells on Th1 and Th17 cells differentiation is impaired (Fig. 8b). Interestingly, less Tr1 (IL-10 + CD4 + ) cells are also found when CD4 T cells where co-cultured with Hif1a-deficient CD1d hi CD5 + B cells, whereas no difference is detected in Foxp3 + Treg cells (Fig. 8b, c), suggesting that CD1d hi CD5 + B cells regulate the differentiation of Tr1 cells in a HIF-1α-dependent manner in this culture system, as observed in vivo (Figs. 6 and 7). This culture system was then used to define whether the impaired regulatory function of Hif1a-deficient CD1d hi CD5 + B cells was due to their defect in IL-10 production. To this end, Hif1a-deficient CD1d hi CD5 + B cells were transduced with an IL-10-overexpression lentivirus ( Supplementary Fig. 8a), before coculturing them with naïve CD4 T cells. Remarkably, the regulatory effects on Th1, Th17, and Tr1 cells differentiation are rescued for IL-10-transduced Hif1a-deficient CD1d hi CD5 + B cells (IL-10-CD1d hi CD5 + (ΔHif1a)) compared to the mocktransduced Hif1a-deficient CD1d hi CD5 + B cells (mock-CD1d hi CD5 + (ΔHif1a)) (Fig. 8b). These data indicate that HIF-1α-dependent IL-10 production is required for the suppressive function of CD1d hi CD5 + B cells on T cells in vitro.

Discussion
Herein, we describe a novel molecular mechanism of immune modulation, which determines the function of IL-10-producing B cells and thereby influences the course of autoimmune disease. We identified HIF-1α as a critical transcriptional factor involved in IL-10 production by B cells, thereby influencing the course of T cell-mediated autoimmune diseases such as EAE and arthritis.
IL-10-producing B cells have been identified as potent players in the inhibition of inflammation in autoimmune disease 6 . Hence, it has been shown that adoptive transfer of B cells taken from DBA/1 mice in the remission phase of arthritis prevents the onset of CIA via the secretion of IL-10 7 . In accordance, transfer of B-cell activating factor of TNF family (BAFF) expanded CD1d hi CD5 + B cells decreased Th17 activation and reduced disease severity of arthritis 29 . In agreement with these data from arthritis models, adoptive transfer of MOG-sensitized CD1d hi CD5 + B cells into WT mice also mitigate the severity of EAE 8 . The immune regulatory function of CD1d hi CD5 + B cells appears to be tightly bound to the production of anti-inflammatory cytokines like IL-10, which limits the immune response to pathogens and thereby prevents damage to the host 30 . Studies have shown that CD1d hi CD5 + B cells have the property to differentiate into plasmablasts after stimulation 31 . Accordingly, we found that CD44 hi CD138 + plasmablasts are also reduced in Hif1a-deficient Fig. 6 HIF-1α deficiency in B cells exacerbates collagen-induced arthritis. a Arthritis incidence in Mb1 cre Hif1a f/f (n = 9) and WT mice (n = 9) after collagen immunization. NI non-immunized mice; CIA collagen immunized mice. b Clinical score of arthritis in mice as described in a. c Picture and quantification of paw thickness at day 45 after the first immunization in mice as described in a. Scale bar, 2 mm. d Histopathology sections and quantifications of erosion area (H&E), inflammation area (H&E), and osteoclast number (TRAP) in paw from mice as in a. Arrows indicate erosion or inflammation area. Scale bars, 500 μm. e Quantitative RT-PCR analysis of Tnf, Ifng, Il17a, Il1b, and Il10 mRNA expression in knee synovial tissue from mice as described in a. f IL-17, IFN-γ, IL-10, and TGF-β expression by splenocytes isolated from Mb1 cre Hif1a f/f (n = 9) and WT mice (n = 9) after collagen immunization followed by in vitro restimulation with collagen (20 μg/ml) for 48 h. g IL-10, TGF-β, and IL-35 production by enriched splenic B cells isolated from Mb1 cre Hif1a f/f (n = 6) and WT mice (n = 6) after collagen immunization followed by in vitro re-stimulation with collagen (20 μg/ml) for 48 h. h, i Representative plots and quantification of IL-17A + CD4 + (Th17), IFN-γ + CD4 + (Th1), CD25 + Foxp3 + CD4 + (Treg), IL-10 + CD4 + (Tr1), and IL-10 + CD19 + (IL-10 + B) cells in spleen (h) and draining lymph nodes (dLNs) (i) from Mb1 cre Hif1a f/f (n = 7) and WT mice (n = 7) after collagen immunization. j Representative plots and percentage of IL-23R + IL-17A + CD4 + and GM-CSF + IL-17A + CD4 + T cells in spleen and dLNs from Mb1 cre Hif1a f/f (n = 6) and WT mice (n = 6) after collagen immunization. Data are shown as mean ± s.e.m. Pictures are representative of three independent experiments. NS not significant; *P < 0.05, **P < 0.01, and ***P < 0.001 (Kaplan-Meier analysis with log-rank test (a) or two-tailed unpaired Student's t-test (b-h)) (see also Supplementary Figure 6) mice after EAE induction, suggesting that loss of HIF-1α causes impaired CD1d hi CD5 + B cells and increases likelihood to develop autoimmune disease.
Published studies have shown that calcium sensor stromal interaction molecules (STIM) and IL-21-dependent cognate interactions are required for the function of IL-10-producing B cells 32,33 . However, the molecular regulation of IL-10 production in B cells was incompletely defined to date. Our data now show that HIF-1α is crucial in inducing IL-10 production by B cells. Lack of HIF-1α in B cells causes reduced IL-10 production followed by enhanced Th17 cells. In addition, increased IL-17 and IFN-γ production in Hif1a-deficient mice is associated with a   NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-02683-x ARTICLE strong exacerbation of EAE and inflammatory arthritis. HIFs have been previously suggested to influence adaptive and innate immunity 34 . Differential effects of HIF-1α and HIF-2α have previously been shown in immune cells 14,16,17 . However, the roles of HIF-1α and HIF-2α in B cells have not been shown. Our data suggest that despite the description of hypoxic niches in the bone marrow and regions within the spleen 35 , HIF-1α and HIF-2α appear non-essential for the development of B cell subsets in the bone marrow as well as spleens and lymph nodes in homeostasis. However, LPS or BCR-mediated activation of B cells causes massive induction of HIF-1α, but not HIF-2α, in an oxygenindependent way. While in macrophages, HIF-1α accumulation requires NF-κB dependent transcriptional event 19 , HIF-1α in BCR-stimulated B cells is induced via the ERK-STAT3 signaling pathway. After B cell activation, STAT3 is phosphorylated at position Ser727, but not at the Thr705 site 36 . Our data show that phosphorylated-STAT3 727 then effectively induces Hif1a gene transcription in activated B cells.
Our findings indicate that HIF-1α contributes to CD1d hi CD5 + B cell proliferation and IL-10 production. Emerging studies indicate that metabolism is important for B cell function and proliferation 25 . In accordance with other publications 29,37,38 , we found that CD1d hi CD5 + B cells have a higher glycolytic activity compared to CD1d lo CD5 − B cells, which can control the normal expansion of CD1d hi CD5 + B cells. Our study shows for the first time that this glycolytic metabolism is dependent on HIF-1α expression. In addition, HIF-1α effectively binds to the Il10 promoter at two HRE elements, which also correlates with the actively transcribing regions. Since previous studies have shown that proximal broad H3K4me3 domains are highly dynamic in different cell types and conditions [39][40][41] , the high enrichment of H3K4me3 in HRE I or HRE II regions is probably related to the hypoxic condition. We show that HIF-1α transcriptionally enhances Il10 mRNA expression. In line with our results, it has been shown that both IL-21 and IL-27 induce ERK and STAT3 activation and upregulate IL-10 in CD4 T cells 42,43 . Notably, HIF-1α appears to specifically induce IL-10 in B cells, as other immune regulatory factors expressed in B cells, such as TGF-β and IL-35, are not altered in Hif1a-deficient B cells. In accordance with our findings, the HIF-1α-dependent regulation of IL-10 has been previously reported in macrophages, Tr1 cells, and myocytes using different approaches of HIF-1α knockdown [44][45][46] . Our data also show that manipulation of HIF-1α expression in B cells influences Tr1 cells. Tr1 cells have been intimately associated with IL-10-producing B cells as they drive the differentiation of cognate B cells into IL-10-producing B cells 47 , suggesting a mutual regulation between IL-10-producing B cells and Tr1 cells. In accordance with that, our data show that impaired IL-10 production by HIF-1α-dependent B cells is associated with decreased Tr1 cells, indicating a regulatory network between IL-10producing B cells and Tr1 cells.
Therapeutically, fostering of HIF-1α expression may provide a tool to increase IL-10-producing B cells and limit autoimmune diseases such as EAE and arthritis. Inhibitors of prolylhydroxylases (PHD), for instance, are agents that can induce HIF-1α. Treatment with PHD inhibitors has been shown to ameliorate endotoxic shock as well as inflammatory bowel disease in mice [48][49][50] . In these studies, PHD inhibitors enhanced the numbers of IL-10-producing B cells and reduced expression of inflammatory cytokines 49 . Therefore, activating the HIF-1α axis through pharmacologic agents may indeed provide a tool to augment the immune regulatory potential of IL-10-producing B cells with the potential to prevent and/or treat systemic autoimmune inflammatory diseases 51 .
In summary, we provide a novel molecular mechanism for the regulation of autoimmune disease by CD1d hi CD5 + B cells. By modulating glycolytic metabolism, HIF-1α regulates CD1d hi CD5 + B cell expansion. Moreover, we identified HIF-1α as a critical node involved in IL-10 production by B cells. HIF-1α effectively binds to hypoxia response elements on the Il10 promoter, resulting in expression of IL-10 in B cells. In consequence, HIF-1α expression in B cells regulates autoimmune diseases such as EAE and arthritis.

Methods
Mice. C57BL/6 WT mice (027) were purchased from Charles River Laboratories (Sulzfeld, Germany). To generate B cell-specific Hif1a or Hif2a-deficient mice, Hif1a flox/flox mice or Hif2a flox/flox mice were crossed with Mb1-cre mice. Hif1a flox/flox mice, Hif2a flox/flox mice, and Mb1-cre mice were previously described [52][53][54] . The mice were bred and maintained on a C57BL/6 background and Hifs flox/flox cre-negative or Hifs +/+ cre-positive littermates were used as WT controls. Sex-and age-matched (8-10 weeks) mice were killed using CO 2  Flow cytometry and cell sorting. Single-cell suspensions were prepared from bone marrow (femurs), spleen, inguinal lymph nodes, peritoneal cavity, and peripheral blood. Red cells were lysed with ammonium-chloride-potassium (ACK) buffer. Cells were Fc-blocked (CD16/CD32) and stained with antibodies (Supplementary Table 1). Analyses of the expression of cell surface molecules on a single cell level were performed by flow cytometry with Calibur (BD) or Cytoflex (Beckman Coulter) flow cytometer. Dead cells were detected using a LIVE/DEAD Fixable Violet Dead Cell Stain Kit (L34955, Life Technologies) before cell surface staining. For analysis of intracellular IL-10 expression by B cells, LPS, phorbol-12myristate-13-acetate (PMA), ionomycin, and monensin (L+PIM) were added to the cultures 5 h before fixing and permeabilizing with the Foxp3/Transcription Factor Staining Buffer Set (00/5523/00, ebioscience) according to the manufacturer's instruction. All flow cytometry experiments were gated on viable, single lymphocytes and data were analyzed with FlowJo software (Treestar).
Full size images are presented in Supplementary Fig. 11.
Luciferase reporter assay. HRE regions (I-V) of Il10 promoter (Supplementary  table 2) were amplified by PCR from genomic DNA extracted from splenocytes in C57BL/6 WT mice and cloned into the pGL3 firefly reporter vector (Promega). 293T cells were co-transfected with luciferase reporter construct and β-gal plasmid using Lipofectamine 2000 (invitrogen). Transfected cells were cultured under normoxic (21% O 2 ) or hypoxic (1% O 2 ) conditions for 24 h. Cells were then lysed and luciferase activity was quantified and normalized to the activity of the cotransfected β-gal reporter gene.
RNA interference. STAT3, ERK, and RelA, or scrambled siRNA lentivirus constructs were obtained from Applied Biological Materials (ABM, Richmond, BC, Canada). Transduced splenic B cells were incubated with 10 μg/ml anti-IgM or LPS for 4 h and then cell lysates was used for western blot.
Chromatin immunoprecipitation. Splenic B cells were enriched from C57BL/6 WT mice and cultured in medium alone or anti-IgM (10 μg/ml) for 8 h. Next, ChIP experiments were performed with ChIP-IT Express kit (53018, Active Motif) according to the manufacturer's protocol. Ten micrograms of anti-pSTAT3 727 , anti-HIF-1α, anti-HIF-1β, or anti-HIF-2α antibodies, as well as control IgG antibody, were used for the immunoprecipitation. Primer sequence for STAT3 binding site was described previously 16 (Supplementary Table 2). HIF-1α binding sites were predicted by JASPAR with the consensus core (A/GCGTG) and primers were designed by Primer-BLAST (Supplementary Table 2).
Glucose uptake and lactate production assays. Glucose uptake was determined using Glucose uptake cell-based assay kit (600470, Cayman) according to the manufacturer's protocol. Lactate production was determined in the supernatant collected from sorted CD1d hi CD5 + CD19 + B cells with or without stimulation for 6 h using the Lactate colorimetric/fluorometric assay kit (K607, Biovision) according to the manufacturer's protocol.
Proliferation and apoptosis assay. Splenic B cells were enriched by CD43 magnetic beads (Militenyi Biotec). The purified B cell population was >98% positive for B220 staining. For proliferation assays, enriched or sorted B cells were labeled with 5 μM CFSE or Celltrace violet (Thermo Scientific) at 37°C for 10 min and then stimulated with anti-IgM, anti-CD40, or LPS for 72 h. For apoptosis assay, enriched B cells were stimulated with anti-IgM, anti-CD40, or LPS for 48 h and then stained with TOPRO (Thermo Scientific) and Annexin-V (ebioscience). For toxic assays of inhibitors, enriched B cells were stimulated with anti-IgM for 4 h with or without doses of inhibitors and then stained with TOPRO and Annexin-V.
Immunization and enzyme-linked immunosorbent assay. In vivo proliferation assay. Mice were fed with 0.8 mg/ml BrdU in drinking water during 7 days before analysis. Cells were then stained with indicated antibodies and prepared according to the BrdU Flow kit (559619, BD Pharmingen) protocol.
CIA and EAE animal models. CIA: Mb1 cre Hif1a f/f mice and WT littermates (8-12 weeks of age) were immunized by intradermal injection in the tail with 100 μg of chicken type II collagen (CII) in complete Freunds' adjuvant (CFA) (Sigma). Twenty-one days after the primary immunization, the mice were boosted with a secondary immunization with same amount of CII emulsified in incomplete Freunds' adjuvant (Sigma) intradermally in the tail proximal to the primary injection site. Next, the clinical scores for each paw were evaluated every other day and scored individually on a scale of 0-4, which results in a maximum score of 16. Each paw is scored as follows: 0, no evidence of erythema and swelling; 1, erythema and mild swelling confined to the tarsals or ankle joint; 2, erythema and mild swelling extending from the ankle to the tarsals; 3, erythema and moderate swelling extending from the ankle to metatarsal joints; 4, erythema and severe swelling encompass the ankle, foot, and digits, or ankylosis of the limb. EAE: Mb1 cre Hif1a f/f mice and WT littermates were immunized subcutaneously with 100 μg MOG peptide  (Charité Berlin) in 50 μl H 2 O emulsified in 50 μl CFA, which was enriched with 10 mg/ml Mycobacterium tuberculosis (H37Ra, Difco/PD PharMingen) on day 0 in order to induce EAE. In addition, 200 ng pertussis toxin (List/Quadratech) was administered i.p. on day 0 and day 2. EAE paralysis of mice was scored as follows: 0, no disease; 1, tail weakness; 2, paraparesis; 3, paraplegia; 4, paraplegia with forelimb weakness; 5, moribund or death.
Histology. On day 45 of CIA model, whole paw joints were fixed in 4% paraformaldehyde, decalcified in EDTA, and then embedded in paraffin. Specimens were longitudinally cut into 4 μm sections, then hematoxylin and eosin (H&E) and tartrate-resistant acid phosphatase (TRAP) stainings were performed.
After completing the EAE experiments, spinal cords were dissected from mice after transcardially perfused with 4% paraformaldehyde and post-fixed them overnight. Paraffin-embedded sections (8 μm) of spinal cords were then stained with H&E and luxol fast blue (LFB) staining.
Isolation of CNS-infiltrated lymphocytes. After cardiac perfusion with PBS, CNS tissues were digested with 2.5 mg/ml collagenase D (Roche) and 1 mg/ml DNaseI (Roche) at 37°C for 45 min. Lymphocytes were isolated by passing the tissue through 70 μm cell strainers, followed by Percoll (Millipore) gradient (70%/37%) centrifugation. Lymphocytes were collected from the interface and washed in PBS. Lentivirus transfection and adoptive transfer. Production of viral supernatants and B cell transduction were described previously 56 . Briefly, lentiviral particles were produced in 293T cells by co-transfection of psPax2 packaging vector (Addgene), VSVG envelope plasmid (Addgene), pDBR (mock), or pDBR-IL-10 plasmid using Lipofectamine 2000 (Invitrogen). After 48 h, supernatants were collected, filtered (0.45 μm), and supplemented with 10 mM HEPES (Invitrogen) and 10 μg/ml polybrene (Millipore). Sorted CD1d hi CD5 + B cells were centrifuged at 3.5 × 10 6 / well in six-well plates with 3 ml of viral supernatants in a total volume of 4 ml at 1800 rpm during 75 min at room temperature, and then washed in PBS. A total of 1.5 × 10 6 transduced CD1d hi CD5 + B cells were transferred intravenously into recipient mice 24 h before EAE induction.
Statistical analysis. For comparison of the two groups, linear regression with a 95% confidence interval, and two-tailed Student's t-test were used. Kaplan-Meier analysis with log-rank test was used to determine the significance of CIA incidence. Two-way analysis of variance with Bonferroni's post test for paired data was used to determine the significance of EAE clinical scores in adoptive transfer experiment. GraphPad Prism software 6.0 was used for statistical analysis. P-value of less than 0.05 was considered statistically significant.
Data availability. The authors declare that all data supporting the findings of this study are available within the article and its Supplementary Information files or are available from the authors on request.