Cholesterol metabolism drives regulatory B cell function

Regulatory B cells restrict immune and inflammatory responses across a number of contexts. This capacity is mediated primarily through the production of IL-10. Here we demonstrate that the induction of a regulatory program in human B cells is dependent on a metabolic priming event driven by cholesterol metabolism. Synthesis of the metabolic intermediate geranylgeranyl pyrophosphate (GGPP) was required to specifically drive IL-10 production, and to attenuate Th1 responses. Furthermore, GGPP-dependent protein modifications controlled signaling through PI3Kδ-AKT-GSK3, which in turn promoted BLIMP1-dependent IL-10 production. Inherited gene mutations in cholesterol metabolism result in a severe autoinflammatory syndrome, termed mevalonate kinase deficiency (MKD). Consistent with our findings, B cells from MKD patients induced poor IL-10 responses and were functionally impaired. Moreover, metabolic supplementation with GGPP was able to reverse this defect. Collectively, our data define cholesterol metabolism as an integral metabolic pathway for the optimal functioning of human IL-10 producing regulatory B cells. Graphical abstract MKD = Mevalonate kinase deficiency


Introduction
Immunosuppressive B cells form a critical component of the immune regulatory compartment 1,2 . It is thought that their suppressive capacity derives mainly from their ability to produce IL-10, and in the absence of any lineage marker, this is considered a hallmark of regulatory B cells 3-6 . Their functional importance has been well described in murine models of disease, demonstrating a potent regulatory capacity across a number of contexts including infection, cancer, and autoimmune disease 3,7-10 . 5 recent studies, as well as those of others, suggest that cholesterol metabolism plays a role in restricting inflammatory responses [22][23][24] . These data are consistent with patients carrying mutations in the pathway, who develop severe and recurring autoinflammatory fevers, associated with dysregulated B cell responses, manifest by elevated serum immunoglobulin levels 25 .
Given the immunoregulatory associations with cholesterol metabolism, we set out to address its role in IL-10 producing B cells. In doing so, we revealed that cholesterol metabolism is a central metabolic pathway required for the induction of a regulatory phenotype in human B cells, defining the specific

Cholesterol metabolism drives IL-10 independent of B cell population
We next aimed to understand how cholesterol metabolism was able to mediate IL-10 production.
Certain populations of human B cells have been proposed as primary producers of IL-10. The most 105 well characterized of these are CD24 hi CD27 + (B10) and CD24 hi CD38 hi B cells 5,6 . In agreement with IL10 mRNA was measured by qRT-PCR, and calculated relative to 18S by the formula 2-∆Ct. C. Schematic protocol for the coculture for T and B cells. Briefly, human CD19 + B cells were stimulated with CpG ± atorvastatin (AT) ± mevalonate (MA) overnight, before washing and addition to αCD3/28 stimulated human CD4 + T cells for 4 days ± αIL-10 antibody. D. IFNγ production in human CD4 + T cells after co-culture with autologous TLR9-activated B cells. IFNγ suppression was calculated by: ((x-y)/x)*100 where x = IFNγ production by T cells alone, y = IFNγ production in co-cultured T cells.  Figure 4G). Therefore, these data indicated a role for cholesterol metabolism in regulating IL-10 production that is shared across B cell populations, rather than an effect on specific populations.

120
To more precisely understand the mechanistic control by cholesterol metabolism, we next sought to investigate if a specific pathway metabolite downstream of HMG-CoA was regulating IL-10.
Cholesterol metabolism encompasses a number of metabolic pathways implicated in immune function including mevalonate, isoprenyl and sterol metabolism (Supplementary Figure 1), all of which are attenuated by HMG-CoA reductase inhibition to varying degrees. Given that defects in the isoprenyl 125 branch have been demonstrated to result in hyperinflammatory responses in vivo 23,26 , we investigated if isoprenylation was regulating IL-10. To this end, we targeted geranylgeranyltransferase (GGTase) and farnesyltransferase (FTase), enzymes known to post-translationally modify proteins with geranylgeranyl pyrophosphate (GGPP) or farnesyl pyrophosphate (FPP) groups respectively (enzymes and metabolites outlined in Figure 2A). Inhibition of GGTase, but not FTase, prevented 130 TLR9-induced IL-10 production, indicating that geranylgeranyl dependent modifications regulate IL-10 expression ( Figure 2B, Supplementary Figure 5A). In keeping with the effects of HMG-CoA reductase inhibition, inflammatory cytokine production was preserved ( Figure 2C). To support the notion of GGPP dependency, we depleted metabolites within the pathway with atorvastatin, and found that specifically supplementing the geranyl branch with exogenous GGPP prevented the 135 blockade of IL-10 production ( Figure 2D, Fig S5B). Finally, to understand if cellular utilization of GGPP was dependent on the enzymatic activity of GGTase, we supplied cells with exogenous GGPP together with inhibition of GGTase. We found that GGPP was unable to rescue levels of IL-10 ( Figure 2E), consistent with the idea that both GGTase activity and GGPP sufficiency are required for TLR9 induced expression of IL-10.

Geranylgeranyl pyrophosphate regulates signaling through TLR9
Isoprenyl modifications almost exclusively regulate the localization of Ras superfamily proteins 27 .
Ras-dependent pathways downstream of TLR9 include Raf-MEK-ERK and PI3K-AKT cascades ( Figure 3A). Signaling through both pathways is critically required for IL-10 production, as inhibition 145 of either pathway is sufficient to block TLR9-dependent IL-10 induction ( Figure 3B-C). Therefore, we sought to address if activation of these pathways is dependent on GGPP. Following GGTase inhibition, AKT phosphorylation on Ser473 was severely impaired, whereas ERK phosphorylation was modestly reduced at early timepoints ( Figure 3D targeting Ser9 28 . In keeping with the idea that GSK3 suppression is required for IL-10 production, chemical inhibition of GSK3 enhanced TLR9-induced IL-10 expression ( Figure 3E). Blocking GGTase activity resulted in reduced Ser9 phosphorylation on GSK3. This indicated preserved activation of GSK3, and that GGTase activity negatively regulates GSK3, which in turn is necessary for IL-10 production ( Figure 3F). We then examined if inhibition of GSK3 was sufficient to rescue an 155 upstream perturbation of geranylgeranylation. Indeed, bypassing GGTase through GSK3 inhibition was able to fully rescue IL-10 expression, without affecting TNFa ( Figure 3G). Together these data  *** suggest that following TLR9 engagement, IL-10 induction through AKT-GSK3, and possibly ERK, is dependent on GGTase activity.

PI3Kd regulates IL-10 expression in human and murine B cells
We next aimed to determine how geranylgeranyl-driven phosphorylation cascades regulate AKT-GSK3 signaling. PI3K mediates Ras-dependent AKT signaling, which suggests that isoprenylationdriven phosphorylation cascades through AKT are dependent on PI3K activity. Accordingly, pan inhibition of PI3K blocked expression of IL-10 upon TLR9 stimulation ( Figure 4A). We found, by 165 selective inhibition of either d, a, or g isoforms of PI3K, that IL-10 is primarily regulated through inactive PI3Kd resulted in almost a complete loss of IL-10 production ( Figure 4C). Furthermore, and in agreement with our previous observations, TNFa production was preserved ( Figure 4C). These data suggest that isoprenylation-dependent interactions between Ras and PI3Kd are required for IL-10 production, and likely underpin the regulatory function of B cells.

Geranylgeranyl pyrophosphate regulates IL-10 induction via BLIMP1
Currently there is no defined transcription factor that regulates IL-10 in human B cells. Therefore, we sought to understand how IL-10 is transcriptionally regulated, and to clarify the role for GGPP in this process. Stimulation of B cells in the presence of actinomycin D indicated that a transcriptional event within the first 24 hours was necessary for IL-10 production (Supplementary Figure 7A). Therefore,  Figure 7D). Consistent with these data, we observed increased expression of BLIMP1 protein within the IL-10 + B cell populations ( Figure 5E).
To more thoroughly address its role, we performed gene targeting experiments on BLIMP1 in primary human B cells. TLR9 stimulation strongly upregulated BLIMP1 expression, whilst siRNA 195 knockdown was able to reduce TLR9-dependent protein level expression by ~60% ( Figure 5F-H). In agreement with a central role in regulating IL-10, siRNA-mediated knockdown of BLIMP1 reduced IL-10 expression by 50-90%, whereas TNFa production was preserved ( Figure 5I). Together, these data demonstrate that BLIMP1 regulates IL-10 in human B cells, and its expression is dependent on cholesterol metabolism for its provision of GGPP.

Mevalonate kinase deficient patients generate poor regulatory B cell responses
Our data demonstrated that cholesterol metabolism is essential for B cell derived IL-10 production.

210
This is in line with a general shift from a naïve to memory like B cell phenotype in MKD patients (Supplementary Figure 8B). We next examined if dysregulated cholesterol metabolism in patients with MKD leads to an inability to mount an IL-10 response, which could contribute to disease persistence or exacerbation. In line with this hypothesis, MKD patients generated poor IL-10 responses after stimulation through TLR9 (30-70% reduction versus controls, Figure 6C), whilst

215
TNFa expression was enhanced in 2 of 4 donors (Supplementary Figure 8C). Interestingly, the defect in IL-10 production could be reversed by the addition of GGPP, indicating that at least in part, this defect was due to a relative deficiency of the isoprenyl metabolite ( Figure 6C).
As with our previous observations, IL-10 expression was reduced across all B cell phenotypes, 220 including a ~3-fold reduction within B10 and CD24 hi CD38 hi B cells ( Figure 6D, gating in Supplementary Figure 4B). This defect in IL-10 production was associated with functional impairment, as MKD B cells were unable to suppress IFNg production by autologous CD4 + T cells, when compared to B cells from healthy controls ( Figure 6E). Moreover, pre-treatment of MKD B cells with exogenous GGPP prior to co-culture, was able to restore regulatory capacity in 2 of 2 225 donors tested (Supplementary Figure 8D This study provides the first description of the metabolic requirements for IL-10 producing B cells, arguing for a central reliance on cholesterol metabolism. Our data point to a model in which synthesis of GGPP prior to stimulation permits transduction of receptor signaling cascades necessary for IL-10 expression. As a consequence, the transcription factor BLIMP1 is induced, which then promotes IL-10 gene expression. We propose that this has direct relevance in vivo, as IL-10 producing B cells from 245 patients who carry mutations in the cholesterol metabolic pathway phenocopy our in vitro findings.
Investigations into B cell metabolism have focused primarily on either antibody production or activation induced metabolic reprogramming 18-20,33 . As alluded to above, there was no understanding of the metabolic requirements for regulatory B cells. Here we demonstrate that cholesterol 250 metabolism is critical in mediating the regulatory capacity of human B cells through its control of IL-10. Interestingly, cholesterol metabolism has been implicated in regulatory T cell function through a mechanism dependent on ICOS and CTLA-4, whilst having no effect on IL-10 expression 34 .
Although the authors did not explore this, we anticipate that regulatory T cells may also possess significant GGPP dependency, as regulatory T cell function is especially reliant on PI3Kd activity 35 .

255
Therefore, whilst cholesterol metabolism appears to regulate different effector molecules between the cell types, it may be that regulatory T and B cells rely more heavily on cholesterol metabolism due to the necessity for potent GGPP-dependent PI3Kd activity.
Much of the experimental data regarding immunity and metabolism have suggested a paradigm in 260 which the integration of metabolic pathways controlling cell fate arises as a direct consequence of immune cell activation. Based on the ability of cholesterol metabolism to control induction of a regulatory program in human B cells by modulating TLR9 signaling, we propose that the programming of immune responses through cholesterol metabolism may differ in this regard. We propose that isoprenyl modifications constitute a metabolic pre-programming event. This pre-direct consequence of stimulation), that has the effect of fine-tuning signaling cascades. This notion of pre-programming is consistent with data showing the GGPP-dependent constitutive localization of Ras family proteins at the cell membrane and endosomes, and that blockade of cholesterol metabolism retains Ras in the cytoplasm 36,37 . In other words, the state of signaling intermediates is 270 preset by the state of cholesterol metabolism of the quiescent cell at any given point.
Our finding that either GGPP deficiency or inhibition of its cognate enzyme GGTase was sufficient to block IL-10 production suggested that both the metabolite and its enzyme are absolutely required for the induction of a regulatory phenotype. Notably, inhibition of FTase was unable to inhibit IL-10 275 production, likely due to the differing targets for geranylgeranylation and farnesylation 27 . This suggests that cholesterol metabolism drives the anti-inflammatory function of B cells primarily through the synthesis of the isoprenyl group GGPP. These observations contrast with previous reports suggesting that isoprenylation largely mediates the restriction of pro-inflammatory cytokines, including TNFa, IL-6, and IL-1b. It should be noted however, that these data were derived from 280 studies of murine macrophages and intestinal epithelial cells 23,38,39 . Nonetheless, whilst human B cells are known to co-express inflammatory cytokines TNFa and IL-6 alongside IL-10 40 , we saw no alteration in their capability to produce these upon blockade of GGTase. This suggests that proinflammatory cytokine production relies less heavily on cholesterol metabolism. Our finding that cholesterol metabolism regulates PI3K signaling may somewhat explain this, given that PI3K can 285 differentially regulate pro-versus anti-inflammatory cytokine production upon TLR ligation [41][42][43] .
We found that the mechanistic control of cholesterol metabolism revolves around its ability to regulate PI3Kd-AKT signaling downstream of TLR9. Although we documented expression of all isoforms of PI3K, we observed that PI3Kd, to a greater extent than PI3Ka or PI3Kg, contributed to 290 IL-10 expression in human B cells upon TLR9 ligation. This finding is consistent with a recent observation identifying IL-10 producing B cells as the primary pathological cell type in a model of activated PI3Kd syndrome, driving Streptococcus pneumoniae persistence 14 . We further suggest that GSK3 inactivation downstream of GGPP-dependent PI3Kd-AKT signaling is also required for optimal IL-10 expression. Consistent with this, AKT-driven phosphorylation of GSK3 on Ser9 is 295 known to differentially regulate cytokine production in monocytes 28 . Moreover, we observed that reduced IL-10 expression induced in the context of defective isoprenylation could be rescued through inhibition of GSK3. In line with our findings, GSK3 inhibition in both innate cells 28   To validate our findings in the context of human disease, we investigated patients with dysregulated cholesterol metabolism. MKD patients carry a mutation in the mevalonate kinase gene. As a 315 consequence, their ability to convert mevalonate to mevalonate-5-phosphate is severely impaired. In keeping with our findings following perturbation of cholesterol metabolism in healthy B cells in vitro, we observed poor regulatory responses in B cell from these patients. This anti-inflammatory defect uncovered an unappreciated dimension to the spectrum of MKD. This included a reduced ability to produce IL-10, associated with a functional impairment in restricting T cell responses. In both cases, through the pathway might be contributing. In agreement with this, MKD patients have been demonstrated to accumulate unprenylated Ras proteins 48 . To date, the causative factor driving disease pathology has been defined as excessive inflammatory cytokine production, driven through increased macrophage driven IL-1b production 23,49 , but also through the induction of a trained immunity 325 phenotype driven by accumulated mevalonate 50

Declaration of interests
The authors declare no competing interests   Paired-end sequencing was then conducted using the HiSeq 2500 platform (Illumina). Raw data was Statistics All statistical analysis was conducted using GraphPad Prism v7.0 (GraphPad, San Diego, CA, USA).