Vitamin D3 receptor polymorphisms regulate T cells and T cell-dependent inflammatory diseases

Significance

Conclusive identification of single genes contributing to complex autoimmune diseases has been challenging. Here, we positionally identify the vitamin D3 receptor gene (Vdr) as a driver of T cell-dependent inflammatory diseases using forward genetics. In the process, we generated congenic mice that overexpress Vdr in T cells as a consequence of natural polymorphisms in its promoter. These mice present a unique opportunity to study the immunomodulatory properties of VDR in a physiological setting. Moreover, the restricted overexpression of VDR to immune cells allows discrimination between immune-acting and confounding musculoskeletal effects of VDR. Our results suggest that VDR plays a role in T cell activation in parallel to the antiinflammatory actions mediated by its ligand.

Abstract

It has proven difficult to identify the underlying genes in complex autoimmune diseases. Here, we use forward genetics to identify polymorphisms in the vitamin D receptor gene (Vdr) promoter, controlling Vdr expression and T cell activation. We isolated these polymorphisms in a congenic mouse line, allowing us to study the immunomodulatory properties of VDR in a physiological context. Congenic mice overexpressed VDR selectively in T cells, and thus did not suffer from calcemic effects. VDR overexpression resulted in an enhanced antigen-specific T cell response and more severe autoimmune phenotypes. In contrast, vitamin D3-deficiency inhibited T cell responses and protected mice from developing autoimmune arthritis. Our observations are likely translatable to humans, as Vdr is overexpressed in rheumatic joints. Genetic control of VDR availability codetermines the proinflammatory behavior of T cells, suggesting that increased presence of VDR at the site of inflammation might limit the antiinflammatory properties of its ligand.

Complex autoimmune diseases affect 5 to 8% of the world population and present a major health and socio-economical problem. These diseases are of multifactorial origin and are often associated with a strong genetic component (1). Over the last decades, human association studies have been successful in identifying a discrete number of strong risk loci (e.g., HLA) together with a large number of weaker loci (1, 2). However, it has not been possible to conclusively identify the underlying genes due to technical limitations, which include linkage disequilibrium, sample heterogeneity, in vitro artifacts, and the loss of biological context to carry out proof-of-concept studies. Combined analyses of epidemiological association studies have been supportive but not conclusive. One such example is the association between the circulating form of vitamin D₃ [25(OH)D₃] and incidence of various autoimmune diseases (3), which proposed a promising role for vitamin D3 in the regulation of autoimmunity.

In contrast to genome-wide association studies in humans, forward genetics-based animal studies enhance the possibility to identify (immune-related) quantitative trait loci (QTL) in a hypothesis-free manner, and to test their functional importance. Collagen-induced arthritis (CIA) is one of the most extensively studied rodent models of rheumatoid arthritis (RA), sharing several pathological features with the human autoimmune condition, thereby making it a suitable model for screening autoimmune phenotypes (4). CIA is a complex, polygenic disease dependent on T and B cell reactivities, similar to RA. Using susceptibility to CIA as a trait for genetic linkage analysis, we have previously identified several arthritis-regulating QTL and positioned the underlying polymorphisms in some of them (5–7). In the present study we set out to fine-map the Cia37 QTL (54 cM) on mouse chromosome 15qF1, previously identified by Ahlqvist et al. (5), with the aim of identifying the genetic polymorphisms regulating this QTL.

We found that a small subfragment (0.46 Mbp) of the Cia37 locus containing natural polymorphisms in the vitamin D receptor gene (Vdr) increases susceptibility to autoimmunity in congenic mice. These naturally occurring polymorphisms are exclusively active in immune cells, offering a unique platform to study the immunomodulatory properties of VDR in a physiological context, without systemic vitamin D-related effects. Overexpression of VDR enhances T cell-driven inflammation in mice and associates with autoimmune arthritis in humans (8–10). These observations suggest that even highly localized VDR expression levels in inflammatory conditions, rather than 25(OH)D₃ availability, could be associated with a poor disease prognosis in autoimmune patients.

Results

A 0.5-Mbp Fragment of the Cia37 Locus Controls Susceptibility to the Autoimmune Model CIA.

The Cia37 locus was originally described as a 54-cM genetic fragment from DBA/1J introgressed into the C57BL/10 mice expressing the H-2^q MHC haplotype (C57BL/10.Q, simply referred to as BQ) mouse strain (5). We found that the reported arthritis-promoting phenotype could be confined to a 4-Mbp subcongenic fragment within the Cia37 QTL that we termed C41 (Fig. 1 A and B). Through intercross breeding of heterozygous C41 mice and recombinant selection, we were able to divide the C41 fragment into two subcongenic fragments, termed C412A and C412B (Fig. 1A). The C412A QTL lost the disease phenotype observed in the parental C41 QTL, whereas the 0.5-Mbp C412B sub-QTL, containing eight protein-coding genes, retained the original CIA phenotype (Fig. 1 C and D). C412B was then made into a congenic line, which we used for more detailed analysis of the QTL-associated phenotype. BQ mice (the background strain) were used as wild-type controls. Since the development of CIA depends on both the innate and adaptive immune cell compartments, we addressed whether the C412B QTL affected either one or both immune branches. To assess this, we tested two inflammatory disease models that are independent of T and B cells, namely dextran sodium sulfate (DSS)-induced colitis and collagen antibody-induced arthritis (CAIA) (11, 12). We did not observe any phenotypic differences between C412B and BQ animals in either of these disease models (Fig. 1 E and F). Together, these observations indicate that the C412B QTL plays a minimal or no role in the innate immune compartment, but instead regulates the adaptive immune response.

Fig. 1.

The C412B fragment regulates collagen-induced arthritis. (A) Schematic representation of the C41 QTL on chromosome 15 and derived subfragments (C412A and C412B). Congenic mice carry the DBA1J allele of the indicated fragments on an otherwise C57BL10.Q (BQ, wild-type) background. The C412B fragment contains the Vdr gene. (B) Collagen-induced arthritis in C41 (B), C412B (C), or C412A (D) congenic mice (open) compared to BQ wild-type controls (closed). Arthritis incidence and total number of mice are indicated in parenthesis. (E) CAIA in C412B and BQ mice. (F) DSS-induced colitis disease activity index and colon length at endpoint (day 10). All data points display mean (SEM). *P < 0.05, **P < 0.01.

Positioning of Vdr as the Causative Gene.

To identify possible causative polymorphisms within the C412B fragment, we compared the genomes of both founder strains (DBA/1J and C57BL/10.Q) using the genome sequences available at the Mouse phenome database [RRID:SCR 003212 (13)]. The C412B fragment contained eight polymorphic genes, of which five contained (coding) nonsynonymous single nucleotide polymorphisms (SNPs), resulting in the following amino acid exchanges: HDAC7 (L502M), VDR (L276M), COL2A1 (T539A), SENP1 (R432H), and ZFP641 (L25F, H347R).

We first addressed potential functional consequences of the listed amino acid substitutions by considering their position in the respective proteins. None of the amino acid substitutions overlapped with critical protein sites, such as catalytic domains or DNA binding motifs (14). We concluded that these amino acid substitutions were unlikely to have functional consequences but addressed our main concerns regarding each of these proteins in a series of experiments that follow below.

COL2A1 (collagen type II α-chain) is an important component of the cartilage, and the antigen targeted in CIA. However, T539A on COL2A1 does not overlap with any of the known T or B cell epitopes (15, 16). Moreover, CIA mice had no differences in the antibody response toward this particular region of the CII molecule (N-GERGPSGLAGPKGANGDPGRPGEP-COOH, peptide #27) (SI Appendix, Fig. S1A) that could help explain the increased disease severity.

SENP1 affects maturation of M1 macrophages by interfering with IFN-γ signaling (17). However, we found no phenotypic differences in the DSS-colitis or CAIA models (Fig. 1 E and F), where macrophages are critical (18, 19).

HDAC7 is a regulator of apoptosis during thymic negative selection and important for early B cell development (20, 21). Using flow cytometry, we did not find a significant skewing in the frequency of thymocyte T cell populations, recent thymic emigrants, or B cells (SI Appendix, Fig. S1 B–D).

Finally, ZFP641 is not expressed in primary or secondary lymphoid organs (ENCODE, BioProject: PRJNA66167, PMID:25409824), and there is a lack of literature detailing potential effects of this protein on the immune system. All data considered, we concluded that the listed amino acid substitutions had no functional consequences.

We thus focused our attention on noncoding SNPs, which can cause critical transcriptional changes. For this purpose, we compared the expression levels of all polymorphic genes, in the secondary lymphoid organs of BQ and C412B mice. Surprisingly, stimulation of splenocytes with Con A revealed a selective overexpression of the Vdr gene in cells from C412B mice (Fig. 2A). This expression phenotype could be recapitulated by stimulating cells with anti-CD3 and anti-CD28 antibodies, indicating a T cell-associated phenotype (Fig. 2B). We found that the elevated Vdr expression translated into higher VDR intracellular protein levels (Fig. 2C). Vdr overexpression enhanced VDR activity in the cell, as demonstrated by increased expression of Cyp24a1 (Fig. 2D) (22). Furthermore, stimulation of cells in charcoal-stripped FBS (Fig. 2D) revealed that overexpression of Vdr required the presence of its ligand [1,25(OH)₂D₃].

Fig. 2.

Polymorphisms in the Vdr promoter drive overexpression in leukocytes of congenic mice. (A) Expression of genes within C412B fragment in unstimulated or Con A-stimulated (1 µg/mL) splenocytes for 48 h. Gene expression is expressed as fold-change over BQ in all graphs (n = 5 per group). (B) Kinetics of Vdr gene expression after stimulation with anti-CD3 and anti-CD28 (1 µg/mL) antibodies for indicated time points (n = 3 per group). (C) Representative Western blot showing VDR protein expression under same conditions as in A (Left) and quantification (Right). (D) Gene expression of Vdr, Cyp24a1, and Cyp27b1 after stimulation of splenocytes with anti-CD3/CD28 in the presence of standard serum (FBS), charcoal-stripped FBS, or stripped FBS supplemented with 10 nM 1,25(OH)₂D₃ (VD3) (n = 5 per group). (E) Luciferase reporter assay showing activity of BQ or C412B Vdr promoter allele after transfection to EL-4 or HEK-293T cell lines (n = 5 per group). Data shown represents two independent experiments. (F) Phylogenetic tree showing clustering of common mouse inbred strains based on allelic differences in Vdr gene [Mouse Phylogeny Viewer (87)]. All data are summarized as mean (SEM); n.d. indicates not detected. *P < 0.05, **P < 0.01.

Considering the enhancement of Vdr expression in congenic mice, we screened the Vdr locus (±1.5 kbp) for potential SNPs regulating transcriptional activity. Comparing both strains, we found a cluster with three SNPs (rs31687858, C/G, frequency: 0.4/0.6; rs16783528, A/G, 0.9/0.1; and rs16804737, T/A, 0.5/0.5) in the Vdr promoter (23). We cloned the Vdr promoter in their DBA/1J (C412B) and BQ allelic forms into luciferase reporter constructs and transfected them into the mouse EL-4 and the human HEK-293T cell lines. We observed that the DBA/1J allele indeed led to higher transcriptional activity of Vdr than the BQ allele (Fig. 2E). Comparison of the Vdr gene sequence among various inbred mouse strains revealed a total of seven different haplotypes, showing that it is a highly polymorphic region (Fig. 2F). We thus concluded that conserved polymorphisms in the Vdr promoter were driving Vdr gene overexpression in congenic mice.

Promoter Polymorphisms Affect Vdr Expression Primarily on T Cells without Affecting Systemic Vitamin D-Related Parameters.

As part of the vitamin D endocrine system, VDR is an important regulator of calcium and bone density. Considering that we found Vdr expression to be altered in immune cells, we explored the possibility of a systemic dysregulation of vitamin D biology in C412B mice. We thus compared bone mineral density (BMD), as well as circulating levels of 25(OH)D₃ (25D3), parathyroid hormone (PTH), and calcium between the strains. Calcium was of special interest because it is an essential second messenger during the activation of immune cells. We did not observe significant differences between BQ and C412B naïve mice across any of the parameters tested (Fig. 3 A–F). We tested bone parameters and serum 25D3 in arthritic mice, which were also similar. Moreover, we did not find different VDR protein levels in the kidney, a tissue that expresses this protein constitutively (Fig. 3G). Taken together, these data suggest that Vdr overexpression in C412B mice is restricted to immune cells, and thus does not affect systemic vitamin D-related parameters.

Fig. 3.

Systemic vitamin D-dependent parameters are similar in congenic and BQ mice. (A) Illustration of the BMD parameters evaluated by high-resolution micro-CT. Bone cortical parameters were measured in the mid-diaphysis (D); trabecular bone parameters stem from the metaphysis (M). Scale bar represents 1 mm. (B) BMD parameters of femur and tibia of naïve mice, and femurs of arthritic mice (C). (D) Serum levels of 25(OH)D₃ in naïve and arthritic mice. (E) Serum levels of PTH and calcium (F) in naïve mice. (G) VDR protein expression in kidney lysates as determined by Western blot. β-Actin was used as the loading control. Data are summarized as mean (SEM); n.s. indicates not significant.

Considering these results, we carried out a more detailed analysis of the Vdr expression changes found in the immune system. For this purpose, we sorted and stimulated various immune cell types in vitro and analyzed Vdr expression after stimulation, as shown in Fig. 4A. Both CD4⁺ and CD8⁺ C412B T cells showed marked overexpression of the Vdr gene compared to BQ controls after anti-CD3/CD28 stimulation. C412B bone marrow-derived dendritic cells (bmDCs, derived as in ref. 24) and peritoneal macrophages (isolated as in ref. 25) also overexpressed Vdr, albeit to a milder degree, in response to LPS stimulation (26) (Fig. 4A and SI Appendix, Fig. S3A). Finally, in anti-IgM stimulated CD19⁺ B cells (27), we found minimal Vdr expression and no difference between the strains. Collectively, these results suggest that C412B mice overexpress Vdr predominantly in T cells. We therefore focused on the effect of Vdr overexpression on T cell activation.

Fig. 4.

VDR is up-regulated in T cells in response to inflammation and enhances T cell activation. (A) Gene expression of Vdr gene in the indicated cell populations. T cells were stimulated for 48 h with 1 µg/mL of anti-CD3/CD28, B cells with 1 µg/mL of anti-IgM; peritoneal macrophages and bmDCs for 24 h with 1 µg/mL of LPS 055:B5. Gene expression is expressed as fold-change over BQ. Mean (SEM) from n = 4 to 5 mice per group. (B) Representative flow cytometry stain of VDR (fmo, fluorescence minus one control; istp, isotype control) and VDR mean fluorescence intensity (MFI) in the indicated T cell subsets 10 d after immunization of mice with ovalbumin (ova). (C) Expression of CD40L on CD4⁺ T cells from ova-immunized mice after in vitro recall with ova (100 µg/mL) in the presence or absence of 10 nM 1,25(OH)₂D₃ (D3). n.s. indicates not significant. (D) Vdr gene expression in lymph node cells from ova-immunized mice after recall with ova as in C. Mean (SEM) of n = 4 mice per group. (E) VDR expression in indicated immune cell types from draining lymph nodes of naive (closed) or ova-immunized (open, day 10) mice (flow cytometry). (F) Activation of HCQ3-tg CD4⁺ T cells (reactive to rCII259-273) transfected with plasmids constitutively expressing VDR (pmax.VDR/VDR⁺) or GFP (pmax.GFP/GFP⁺). The transfection efficiency is shown on the left. Once transfected, the cells were activated with rCII259-273 in the presence or absence of 1,25(OH)₂D₃ 10 nM (D3) for 48 h, at which timepoint expression of CD44, CD69, and IFN-γ was assessed using flow cytometry (Right). (G) Vdr expression in synovial tissue from healthy donors, patients with osteoarthritis (OA), or rheumatoid arthritis (RA, GEO dataset id: GDS5401/204254_s_at). (H) Vdr expression in mononuclear cell from juvenile RA patients (GDS711/1388_g_at). *P < 0.05, **P < 0.01.

VDR Is Up-Regulated in Response to Inflammation and Enhances T Cell Activation.

Flow cytometry analysis of VDR (protein) confirmed our gene-expression results, showing that activated effector (CD44⁺, CD69⁺), but not naïve (CD62L⁺), C412B CD4⁺ T cells overexpressed VDR at the protein level (Fig. 4B). Interestingly, C412B mice immunized with ovalbumin showed a higher frequency of CD4⁺CD40L⁺ antigen experienced T cells (28) that overexpressed Vdr after antigen recall (Fig. 4 C and D, respectively). Comparing VDR expression in immune cells of BQ naïve and immunized mice, we found that T cells, and to a lesser extent myeloid cells, up-regulated VDR in response to inflammation (Fig. 4E). In contrast, B cells only poorly expressed the receptor. In conclusion, the data suggest a positive correlation between VDR and T cell activation.

We thus tested whether recombinant VDR expression would enhance T cell activation, by transfecting transgenic T cells specific to collagen type II amino acids 259 to 273 (HCQ3-tg), with a plasmid expressing either VDR or GFP (as control) under a constitutive promoter. Once more, VDR overexpression correlated with the expression of T cell activation markers CD44 and CD69, as well as IFN-γ production after activation with cognate antigen (Fig. 4F).

Collectively, these results indicate that the identified promoter polymorphisms primarily affect Vdr expression in T cells, without affecting its expression in other tissues where VDR exerts primordial functions, such as regulation of serum calcium levels and bone resorption. Moreover, these observations suggest that antigen-specific T cells with high levels of VDR constitute a potential driver of inflammation in mice, influencing disease development in a pathogenic or self-reactive setting.

In order to relate our observations to a human setting, we addressed whether Vdr expression could also be found up-regulated in samples of RA and juvenile idiopathic arthritis (JIA) patients, using publicly available gene-expression datasets (GDS5401/204254_s_at; GDS711/1388_g_at) (8, 29, 30). We found that this was indeed the case. Vdr expression was higher in synovial tissues from RA patients compared to that of healthy controls (Fig. 4G). Interestingly, Vdr expression was also elevated in synovial fluid of JIA patients, which in turn was significantly higher than in peripheral blood mononuclear cells (PBMCs) from the same patient or from healthy controls (Fig. 4H), showing that expression of Vdr is up-regulated specifically at the site of inflammation.

VDR Overexpression Enhances Autoimmune T Cell Response and Worsens Inflammation.

Our results so far indicated that the C412B locus predominantly affects the adaptive immune compartment, particularly T cells. However, both T cells and antigen presenting cells (APCs) up-regulated VDR in response to inflammation (Fig. 4E). APCs are known targets of 1,25(OH)₂D₃ (1,25D3), and in our hands treatment with 1,25D3 inhibited bmDC maturation as well as bmDC-mediated T cell activation (SI Appendix, Fig. S2). As a result, we wanted to rule out potential effects of increased VDR (Fig. 4A) on the function of C412B APCs.

To this end we compared the function of BQ and C412B APCs in vitro and in vivo (SI Appendix, Fig. S3). Although C412B bmDCs significantly overexpressed Vdr (Fig. 4A and SI Appendix, Fig. S3A), they were comparable to their BQ counterparts in their capacity to activate T cells in antigen presentation assays in vitro (SI Appendix, Fig. S3 B and C). In vivo, C412B migratory DCs (31) did not overexpress VDR (SI Appendix, Fig. S3D), and furthermore expressed similar levels of activation markers (SI Appendix, Fig. S3E). Nevertheless, we tested for possible functional differences in vivo by transferring labeled CD45.1.HCQ3-tg T cells to either BQ or C412B recipients, and measuring T cell proliferation after immunization of the recipient mice with cognate antigen (SI Appendix, Fig. S3F). We found that antigen presentation and T cell activation in vivo was comparable between C412B and BQ mice. We thus concluded that the moderate Vdr expression differences observed in APCs do not significantly affect their maturation and migration upon antigen encounter, nor the consequent antigen availability, presentation, and activation of T cells.

Because the data suggested a direct effect of VDR on T cells, we investigated the impact of heightened VDR expression on autoreactive T cells in different models of autoimmunity (Fig. 5). During CIA, C412B mice had increased lymph node cellularity (Fig. 5A), indicating a more severe (auto)immune response. Indeed, T cell restimulation assays showed that antigen-specific T cells from C412B mice had a more proinflammatory phenotype than BQ cells (Fig. 5B), evidenced by significantly higher numbers of antigen-specific cells producing IL-2, IFN-γ, and IL-17A. In terms of regulatory T cells (Tregs), we did not find significant differences (Fig. 5C). In line with the increased proinflammatory activity, C412B mice also displayed elevated serum titers of anti-CII antibodies (Fig. 5D). Treatment of T cells from CIA mice with 1,25D3 did not significantly reduce their effector properties during antigen recall (Fig. 5B). Interestingly, 1,25D3 did reduce cytokine production in congenic T cells stemming from naïve mice, particularly when at high (100 nM) concentrations (Fig. 5E and SI Appendix, Fig. S6B). Our observations held true in a second T cell-dependent autoimmune model, experimental autoimmune encephalomyelitis (EAE). As in CIA, C412B mice also developed more severe symptoms of EAE (Fig. 5F). In line with our earlier data (Fig. 5B), C412B CD4⁺ T cells from lymph nodes of EAE mice displayed a more proinflammatory profile (Fig. 5G).

Fig. 5.

VDR overexpression enhances autoimmune T cell response and worsens inflammation. (A) Lymph node cell count at CIA (Fig. 1C) endpoint (d70). (B) Antigen-specific T cell response in lymph node cells from CIA mice. Lymph node cells were cultured with the immunodominant CIA T cell epitope (rCII259-273) to measure the antigen-specific T cell response. From left to right, quantification of cells producing IL-17A, IFN-γ, and IL-2 using ELISpot. (C) Treg (CD4⁺CD25⁺FoxP3⁺) frequency in lymph nodes of CIA mice. (D) Serum levels of anti-CII IgG1 antibodies during CIA (day 35). (E) Frequency of CD4⁺ T cells from naïve mice producing IL-17 after activation with anti-CD3/CD28 for 48 h in the presence or absence of 1,25(OH)₂D₃ (D3) (n = 5 per group). (F) Disease course (Left) and sum score (Right) of MOG79-96 induced EAE in BQ and C412B mice. (G) Expression of IL-17A, TNF-α, IFN-γ, and GM-CSF in PMA/ionomycin stimulated CD4⁺ T cells from lymph nodes of EAE mice (representative stain on the left). All data are summarized as mean (SEM); n.s. indicates not significant. *P < 0.05, **P < 0.01.

In conclusion, our results suggest that the identified naturally occurring Vdr promoter polymorphisms increase VDR presence in T cells, potentiating T cell-dependent inflammation and worsening autoimmune conditions.

VDR Signaling Enhances Proliferation of T Cells.

Our data strongly support the notion that T cell activation mediates an increase in VDR, which directly associates with the proinflammatory effector function of these cells. To more deeply characterize this relationship, we analyzed and compared proteomic changes elicited by either VDR overexpression or 1,25D3 treatment in activated CD4⁺ T cells.

First, the results confirmed drastic overrepresentation of VDR in C412B T cells. ACPP and CYP24A1, which are known transcriptional targets of VDR (22, 32), were also highly overrepresented in C412B T cells (Fig. 6A). Surprisingly, treatment with 1,25D3 but not overexpression of VDR resulted in increased expression of several proteins associated with a regulatory T cell phenotype [Fig. 6B: CTLA4, THEMIS (33), ENTPD1 (34), IZUMO1R (35), and JAK2 (36)], while at the same time reducing the expression of STAT4, reported to drive Th1/Th17 T cell phenotypes (37).

Fig. 6.

1,25D3 signaling through VDR enhances proliferation of T cells. (A) Volcano plot comparing the proteomic profile of C412B (overexpress VDR) and BQ CD4⁺ T cells. T cells were stimulated for 72 h in vitro with 1 µg/mL of anti-CD3/CD28 antibodies. (B) Volcano plot comparing the proteomic profile of BQ CD4⁺ T cells stimulated as in A in the presence of 10 nM of 1,25(OH)₂D₃ or vehicle (ethanol). Four mice were used per group for proteomic analysis and P values were calculated using Welsh’s t test. (C) Comparison of protein expression from A and B. Proteins that were significantly and substantially up- or down-regulated (and detected) in both cases are highlighted in red. (D) Proliferation of CD4⁺ T cells in lymph nodes from ova-immunized mice (d10) as determined by the expression of Ki-67 and representative flow cytometry dot plot (Left). (E) Proliferation of CD4⁺ T cells from naïve mice as determined by dilution of CellTrace (CTV) after 72 h of in vitro stimulation as in A. Representative flow cytometry plot on the left and quantification on the right. (F) Thymidine incorporation assay showing proliferation of lymph node cells stimulated for 48 h as in A in the presence of 1,25D3 or vehicle (ethanol). (G) CIA disease course in QB mice fed standard (open) or cholecalciferol (vit.D)-deficient (closed) diet, with (square) or without (circle) subcutaneous supplementation of D3. D3 or vehicle (polyethylene glycol) were administered using implanted subcutaneous osmotic pumps (Alzet) delivering an equivalent of 1,000 IU/kg of ingested food. (H) ELISA-based T cell recall assay (rCII259-273) showing secretion of IL-17A and IL-2 after ex vivo restimulation of lymph node cells from CIA mice from G. Data are summarized as mean (SEM) and number of mice in parenthesis. *P < 0.05, **P < 0.01.

1,25D3 has several mechanisms of action, not all of them being mediated by VDR (38). In Fig. 6C we compared the proteomic changes resulting from VDR overexpression with those resulting from 1,25D3 treatment. In this way, we used C412B mice as a unique tool to pinpoint which targets were a product of 1,25D3 signaling, specifically through VDR. Interestingly, five of six overlapping up-regulated proteins were positive regulators of cell proliferation [Fig. 6C: PLS1 (39), DIXDC1 (40), SDCBP2 (41), RUNDC1 (42), RIPK3 (43)]. In line with these findings, we observed that both VDR overexpression and 1,25D3 treatment led to increased proliferation of T cells both in vivo and in vitro (Fig. 6 D–F). While up-regulation of the inhibitory protein CTLA4 insinuated regulatory properties, many of the identified proteins have also been linked T cell activation [PRNP (44), PLS1 (39)], Th17 development [SDCBP2 (45)], and cytokine production in different autoimmune models [RIPK3 (46)]. Taken together, our data indicate that VDR signaling enhances proliferation and differentiation of activated T cells toward proinflammatory phenotypes, and thereby exacerbates inflammation.

Nonphysiological Levels of 25D3 Limit the Development of T Cell-Dependent Arthritis.

Fluctuations in serum 25D3 levels have been proposed to affect the immune response and the development of several autoimmune diseases. Considering our previous results, particularly the differing effects of 1,25D3 exposure versus VDR overexpression, we decided to compare the effects of supra- and subphysiological levels of 25D3 in conditions of physiological genetic regulation of the VDR. Hence, we divided BQ mice into two dietary regimens: Standard rodent chow or cholecalciferol (D3; precursor to 25D3)-deficient diet. We further divided each of these dietary groups into two groups: Subcutaneous implantation of osmotic pumps supplying mice with 1) the equivalent of 1,000 IU D3 per kilogram of ingested food or 2) propylene glycol as vehicle control. The effectiveness of each regimen was controlled by measuring serum 25D3 in all groups (SI Appendix, Fig. S6A). When inducing CIA in all of the four experimental groups simultaneously, we observed that mice on a D3-deficient diet were completely protected from CIA, whereas D3-supplementation was enough to reverse this protection (Fig. 6G). At the same time, mice fed a normal diet and supplemented with D3 (i.e., D3 in excess) had 50% lower prevalence of the disease when compared to the vehicle control group fed a standard diet. Interestingly, both mice supplemented with D3 and mice fed on a D3-deficient diet failed to mount a collagen-specific T cell response in antigen recall assays (Fig. 6H). These observations suggest that disturbances in the physiological levels of 25D3 can be detrimental to the healthy immune response.

Discussion

In this study, we positionally identified the VDR as a mediator of proinflammatory T cell effector functions, with severe consequences for the development of T cell-dependent autoimmunity. We show here that the naturally occurring DBA/1J Vdr promoter allele increases Vdr expression, and consequently VDR activity, in stimulated T cells. This genetic and regulatory effect could be of importance for the triggering and progression of human autoimmune diseases, particularly due to the high expression of VDR in the inflamed tissue (8, 30, 47).

The vitamin D endocrine system plays an essential role in the regulation of calcium homoeostasis and bone mineralization. VDR is a nuclear hormone receptor that mediates the genomic actions of 1,25-dihydroxycholecalciferol [1,25(OH)₂D₃; 1,25D3], the active form of vitamin D3 (cholecalciferol; D3), and regulates transcription of up to 3% of the human genome (48). Over the last decades, data have emerged attributing an immunomodulatory function to the VDR and 1,25D3 (49). In humans, serum levels of 25(OH)D₃ (circulating precursor to 1,25D3; 25D3) (3) and polymorphisms of Vdr have been linked to the incidence or severity of several complex (auto)immune diseases (10, 50, 51). Indeed, VDR is expressed in most immune cell types (52), and consequently the 1,25D3/VDR axis has generated interest as an immunotherapeutic target. It is the regulatory properties during the adaptive immune response have attracted much attention in the field of autoimmunity. 1,25D3 prevents full maturation of bmDCs, leading to impaired T cell activation (53). Furthermore, 1,25D3 acts directly on T cells to negatively affect their proliferative capacity (54), while favoring regulatory phenotypes (55, 56).

However, 1,25D3 was originally discovered to enhance bacterial clearance (26, 57), and treatment of DCs with 1,25D3 increases secretion of proinflammatory cytokines (58). At the same time, VDR is also required for T cell activation and proliferation (59). Vdr KO mice were shown to have an impaired Th1 response (60) and normal Treg frequencies (61), while completely failing to develop EAE (62). Furthermore, 1,25D3-deficient mice have reduced amounts of Th17 cells (43) and display a delayed onset of EAE (63). Preclinical findings on the immunomodulatory effects of the 1,25D3/VDR axis are ambiguous and, in humans, studies that relate 25D3 status to autoimmunity are of observational nature (64). Recent double-blinded placebo-controlled vitamin D3 interventional studies on autoimmune patients showed no considerable effects of 25D3 on disease activity (65–67). This may be linked to the fact that activated T cells from an inflammatory environment have decreased sensitivity to 1,25D3 (47, 68). Together, the existing data suggest that studies may overestimate the ability of 1,25D3 to inhibit T cell-mediated inflammation in vivo (47).

Specifically, it should be considered that the regulatory properties of 1,25D3 are often demonstrated in artificial settings, frequently opting for 1,25D3 concentrations in the range of 10 to 100 nM. These doses are orders-of-magnitude above the physiological concentration of 1,25D3 (i.e., 40 to 140 pM) (69, 70) and affect cell viability (71). In vivo, supplementation of 1,25D3 can produce important calcemic effects (72–75), to the point that diabetes mellitus in VDR knockout mice can be reversed by correcting hypocalcaemia (76). It is apparent that excess or absence of 1,25D3 artificially modulate the immune response. We observe that excess of 25D3 suppresses the immune response, although its presence in physiological amounts is required for a healthy immune response. Complete 25D3 deficiency prevented mice from mounting an antigen-specific T cell response and consequently inhibited development of autoimmunity. Overall, and together with our observations, the vast existing data relating vitamin D metabolites or VDR with immune activity demonstrate that artificial models are not particularly informative. In studying congenic mice with natural Vdr polymorphisms, we were able to analyze the effects of VDR on the immune system in a physiologic context, seemingly without any systemic vitamin D-related effects. We measured several serum key vitamin D parameters to be stable in C412B mice (i.e., 25D3, BMD, PTH, and serum calcium); nevertheless, we have no data to exclude possible fluctuations in the active but short-lived form of vitamin D3, 1,25D3.

Using the congenic mice, we find that mice that overexpress Vdr in activated T cells are more susceptible to T cell-dependent autoimmunity (CIA, EAE), and develop significantly higher numbers of proinflammatory antigen-specific T cells during the disease course. We thoroughly addressed possible consequences of Vdr overexpression in other immune cell types relevant to CIA. Apart from T cells, APCs were the only cell type that indicated overexpression of Vdr in vitro. Nevertheless, Vdr overexpression in congenic APCs was mild, and did not alter their capacity to stimulate and activate T cells. This was a surprising observation, considering that APCs are important targets of 1,25D3 and important mediators of 25D3 activation. Likely, the mild nature of Vdr overexpression in congenic APCs is not enough to significantly affect their maturation and downstream functions. Moreover, neither congenic, nor wild-type B cells in general, up-regulated Vdr under inflammatory (activated) conditions. In addition, we found no phenotypic differences in T cell- and B cell-independent inflammatory disease models (11, 12, 77). Taken together, our observations strongly suggest that the expression levels of Vdr can have direct implications on T cell activity in an APC-independent manner, enhancing their proinflammatory behavior. Indeed, it has been previously reported that VDR is required in both human and mouse T cells for their full activation and proliferation (59, 76). Similarly, we demonstrate that VDR signaling enhances proliferation and activation of murine T cells, and that Vdr expression is increased in the T cell-infiltrated synovial tissue of RA patients (8, 29, 30). Interestingly, the increase in Vdr expression is specific to the site of inflammation (synovial joint), as PBMCs from the same individual showed reduced Vdr levels similar to those in healthy individuals. Together, these data support the idea that VDR signaling in T cells is crucial for their full activation, proliferation, and survival.

In line with previous literature, we found that 1,25D3 has a clear suppressive effect on the activation of naïve T cells. However, overexpression of VDR on T cells favored up-regulation of proliferative markers. Moreover, we find that the antiinflammatory net effect of 1,25D3 on T cells stemming from an inflammatory setting with high VDR expression seems to be dampened. In this context, it is important to highlight that effector T cells from synovial fluid of RA patients express high levels of VDR but fail to respond to the anti-inflammatory properties of 1,25D3 (47). It is possible that VDR plays a role in T cell activation in parallel to the antiinflammatory actions mediated by its ligand, particularly since the actions of 1,25D3 can be mediated by other receptors than the VDR (38). It is also possible that VDR may have alternative ligands that when present (such as in an inflammatory milieu) may drive the immune response in a proinflammatory direction. Thus, it is conceivable that 1,25D3 signaling favors regulatory properties, while VDR signaling favors proliferative characteristics. Consequently, it is plausible that high VDR expression may limit the antiinflammatory properties of 1,25D3.

The evolutionary pressure to maintain Vdr promoter polymorphisms suggests that regulation of Vdr expression in T cells is a key mechanism to regulate their activation. In fact, several studies have associated Vdr promoter polymorphisms with different T cell-dependent (auto)immune disorders (10, 50, 51, 78). Polymorphisms favoring Vdr expression might provide an advantage during infections, at the cost of an increased risk of autoimmunity. Indeed, and in line with our data, Vdr alleles that increase Vdr expression in PBMCs (9) reduce the risk of tuberculosis (79), while increasing the risk of autoimmunity (10, 80).

The present study takes advantage of natural polymorphisms to provide a more physiological perspective on the immune-modulatory actions of VDR. In conclusion, we demonstrate that naturally occurring promoter polymorphisms of Vdr control its availability in T cells. VDR regulates the magnitude of T cell receptor signaling and increased levels can be detrimental in an autoimmune scenario. In this context, it should be considered that the immune-regulatory properties of 1,25D3 are reliant on the genetic variability of its receptor. Together with other published studies, our data suggest that while excess of 1,25D3 signaling can generate antiinflammatory properties, physiological amounts of both 1,25D3 and VDR are required to mount an effective immune response.

Materials and Methods

Animals.

C57BL/10.Q (BQ) founders originated from J. Klein (Tuebingen University, Germany). The Cia37 fragment originated from the DBA/1J strain and has been back-crossed to the BQ strain for over 14 generations, and mapped by microsatellite markers. QB (BALB/c × BQ) F1 mice were used to study the effect of cholecalciferol on collagen-induced arthritis, and are described elsewhere (81). HCQ3-tg mice (generated as described in ref. 82) express a T cell receptor specific to the collagen type II (rCII) epitope amino acids 259 to 273 (rCII259-273) (15). Mice were kept under specific pathogen-free conditions in the animal house of the Section for Medical Inflammation Research, Karolinska Institute in Stockholm. Animals were housed in individually ventilated cages containing wood shavings in a climate-controlled environment with a 14-h light-dark cycle, fed with standard chow and water ad libitum. Cholecalciferol-deficient diet was obtained from TestDiet. All of the experiments were performed with age-, sex-, and cage-matched mice and in a blinded fashion and all of the genetic experiments were performed with littermate controls. All of the experimental procedures were approved by the ethical committees in Stockholm, Sweden. Ethical permit numbers were 12923/18 and N134/13 (genotyping and serotyping), N35/16 (CIA), N83/13 (EAE), and N181/13 (colitis).

Cell Culture.

For cell culture, 10⁶ splenocytes or 5 × 10⁵ lymph node cells were cultured in 200 µL of complete DMEM per well in U-shaped bottom 96-well plates (Nunclon). Cells were incubated at 37 °C and 5% CO₂ without using the outermost wells to avoid artifacts related to evaporation. Complete DMEM: DMEM+Glutamax (Gibco), 5% FBS (Gibco), 10 µM Hepes (Sigma), 50 µgmL⁻¹ streptomycin sulfate (Sigma), 60 µgmL⁻¹ penicillin C (Sigma), 50 µM β-mercaptoethanol (Gibco). FBS was heat-inactivated for 30 min at 56 °C. Charcoal stripped FBS was obtained from Gibco. Depending on the experiment, cells were stimulated with: Con A (conA, 1 µgmL⁻¹, Sigma); anti-mouse CD3 (1 µgmL⁻¹; 500A2, BD Pharmingen); anti-mouse CD28 (1 µgmL⁻¹, 37.51; BD Pharmingen); F(ab′)2 anti-mouse IgM, µ chain specific (5 µgmL⁻¹; eBioscience); LPS (1 µgmL⁻¹, O55:B5 from Escherichia coli; Sigma); rCII259-273 (10 µgmL⁻¹).

ELISA.

Flat 96-well plates (Maxisorp, Nunc) were coated overnight at 4 °C with the capture antibody (Ab, listed below) in PBS. Coating solution was decanted and supernatants from cell culture were added. Plates were incubated for 3 h at room temperature before washing (0.05% PBS-Tween) and adding the biotinylated detection Ab in PBS (1 h at room temperature). Plates were washed and incubated 30 min at room temperature with Eu-labeled streptavidin (PerkinElmer; 1:1,000) in buffer E (50 mM Tris⋅HCl, 0.9% [wt/vol] NaCl, 0.5% [wt/vol] BSA, 0.1% Tween 20, 20 µM EDTA). After washing, DELFIA Enhancement Solution (PerkinElmer) was added and fluorescence read at 620 nm (Synergy 2, BioTek). Antibodies (Ab) were: IL-2 (capture Ab 5 µgmL⁻¹ JES6-IA12; detection Ab 2 µgmL⁻¹ biotinylated-JES6-5H4, in-house produced); IL-17A (capture Ab 5 µgmL⁻¹ TC11-18H10.1; detection Ab 2,5 µgmL⁻¹ TC11-8H4, Biolgend); IFN-γ (capture Ab 5 µgmL⁻¹ AN18; detection Ab 2,5 µgmL⁻¹ biotinylated R46A2, in-house produced).

ELISPOT.

EMD Millipore MultiScreen 96-well assay plates were coated as described above. Coating solution was decanted and 10⁶ splenocytes or 5 × 10⁵ lymph node cells were plated per well. After culture, plates were washed (0.01% PBS-Tween) and biotinylated detection antibodies were added in PBS as described above. Plates were washed and Extravidin alkaline phosphatase (Sigma) was added at a 1:2,500 dilution in PBS (30 min, room temperature). Plates were washed before adding Sigmafast BCIP/NBT (Sigma) substrate solution and incubating for 5 to 10 min. When spots became visible, plates were washed in water and counted using a CTL ImmunoSpot Analyzer.

RNA Isolation and mRNA Expression.

RNA was extracted from 1 × 10⁶ cells using Qiagen RNeasy columns according to the manufacturer’s instructions without additional DNase digestion. RNA concentration was determined using a NanoDrop 2000 (Thermo Scientific). Sample concentrations were normalized before proceeding with reverse transcription. Samples were stored at −20 °C for short-term storage. cDNA synthesis was carried out using the iSrcipt cDNA synthesis kit (Bio-Rad) according to manufacturer’s instructions. For qRT-PCR, primers were designed to cover an exon–exon junction in order to minimize amplification of genomic DNA and were used at a final concentration of 300 nM. The qPCR reaction was carried out using the iQSYBR Green Mix (Bio-Rad) in white 96-well plates using a CFX96 real-time PCR detection system (Bio-Rad). Actb or Gapdh were used as an endogenous control. Primer sequences are listed in SI Appendix, Table S1. Data were analyzed according to the ∆∆Ct method (83), assuming equal efficiency for all of the primer pairs.

Flow Cytometry.

All centrifugation steps were carried out a 350 × g for 5 min at room temperature; 1 × 10⁶ cells were blocked in 20 µL of PBS containing 5 µg in-house produced 2.4G2 in 96-well plates for 10 min at room temperature. Samples were washed with 150 µL of PBS and subsequently stained with the indicated antibodies in 20 µL of PBS diluted 1:100 or 1:200 at 4 °C for 20 min in the dark (see Ab list). Cells were washed once, fixed, and permeabilized for intracellular staining using BD Cytofix/Cytoperm (BD) according to manufacturer’s instructions. Cell were stained intracellularly with 20 µL of permeabilization buffer (BD), using the antibodies at a 1:100 final dilution, for 20 min at room temperature. FoxP3 staining required nuclear permeabilization and was carried out using Bioscience Foxp3/Transcription Factor Staining Buffer. For intracellular cytokine staining, cells were stimulated in vitro with phorbol 12-myristate 13-acetate (PMA) 10 ngmL⁻¹, ionomycin 1 µgmL⁻¹, and BFA 10 µgmL⁻¹ for 4 to 6 h at 37 °C prior to fixation, permeabilization, and staining.

Protein Isolation and SDS/PAGE.

Total protein was isolated from 2 × 10⁶ splenocytes using 30 µL of lysis buffer (4% SDS, 1 mM EDTA, 20 mM Hepes, 50 mM Tris pH 6.8) with freshly added protease inhibitors (cOmplete, Roche). Samples were heated to 95 °C for 15 min. Lysates were centrifuged for 15 min at top speed and supernatants were used for SDS/PAGE (4–12 NuPAGE Bis-Tris gel; Thermo Scientific) according to the manufacturer’s instructions (45 min, 200 V, Mops buffer). Protein concentrations were measured using NanoDrop 2000.

For Western blot, proteins were blotted onto a PVDF membrane (Millipore) for 1 h at 30 V in NuPAGE transfer buffer (Thermo Fisher). Membranes were blocked for 1 h at room temperature in blocking solution (0.05% PBS-Tween, 5% milk powder). Incubation with the primary antibody (mouse anti-VDR, clone D-6 [Santa Cruz], final dilution 200 ngmL⁻¹) was performed overnight at 4 °C in blocking solution. β-Actin was used as a loading control (Abcam). After incubation, membranes were washed in PBS-T and incubated with Affinipure peroxidase-coupled goat anti-mouse IgG(H + L) (final concentration, 40 ngmL⁻¹; Jackson Laboratories) for 1 h at room temperature. Membranes were washed and coated with 1 mL ECL substrate solution (GE Helathcare) before developing. X-ray films were exposed for up to 15 min and developed on a Curix 60 X-ray film processor (Agfa).

Recombinant Expression of VDR in HCQ3-tg Cells.

For the expression of VDR in mammalian cells, the GFP sequence from the Lonza pmaxGFP plasmid was replaced by the VDR sequence (sequence accession P48281) shown below. The plasmid was digested using KpnI and XhoI restriction enzymes according to the manufacturer (Thermo Scientific). Primary transgenic CD4⁺ cells (HCQ3-tg, described in ref. 15) were sorted from spleen and lymph nodes of HCQ3-tg mice using Dynabeads Untouched Mouse CD4 Cells Kit (Thermo Fisher); 2 × 10⁶ CD4⁺ T cells were transfected using P3 Primary Cell 4D-Nucleofector X Kit (Lonza) according to the manufacturer (program DN-100) with 2 µg of pmaxGFP or pmaxVDR. Cells were cultured for 48 h to maximize protein expression before further analysis. Transfection efficiency was 30 to 40%.

Collagen-induced arthritis (CIA).

Twelve-week-old mice were immunized with 100 µg of rat collagen type II (rCII) in 100 µL of a 1:1 emulsion with complete Freund’s adjuvant (BD, Difco) and PBS intradermally at the base of the tail. Mice were challenged at day 35 with 50 µg of rCII in 50 µL of IFA (BD, Difco) emulsion. Mice were monitored for arthritis development every second to third day using the following scoring system: Each visibly inflamed (i.e., swelling, redness) ankle or wrist was given 5 points, whereas inflamed knuckles were given 1 point, resulting in a total of 60 possible points per mouse.

Studies on the Effect of Cholecalciferol (D3) on CIA.

QB mice were fed either a normal or D3-deficient diet (TestDiet) from the time of weaning until termination of the experiment. Since the time of weaning is around 28 d of age, mice were on the specified diets for 2 mo before disease induction. For D3 supplementation, mice were implanted with subcutaneous osmotic pumps (Alzet) delivering the equivalent of 1,000 IU D3 per kilogram ingested food or vehicle (propylene glycol). Pumps were implanted at 10 wk of age, 1 d before immunization with rCII.

Antibody Titer Determination.

Antibody titers were determined using ELISA. Briefly, plates were coated with rCII 10 µgmL⁻¹ overnight; the next day serum samples were added (typically 1:1,000 to 10,000 in PBS) and incubated for 3 h at room temperature. Anti-CII antibodies were detected using HRP-coupled anti-mouse IgG1, IgG3, or IgGκ antibodies (from Southern Biotech).

CAIA.

CII-specific antibodies (M2139, CIIC1, CIIC2, and UL1) were generated and purified as previously described (84). The sterile mixture of M2139, CIIC1, CIIC2 and UL1 mAbs (4 mg per mouse) was injected intravenously. On day 7, lipopolysaccharide (O55:B5 LPS from Merck; 25 μg in 200 μL per mouse) was injected intraperitoneally to all mice to increase severity of the disease.

DSS-Induced Colitis.

Mice were allowed to drink water containing 3% DSS (Sigma, ∼40 kDa) ad libitum for 10 d and were scored daily for symptoms of colitis. Mice were scored according to Kim et al., (85): Weight loss: 0 (no loss), 1 (1–5%), 2 (5–10%), 3 (10–20%), and 4 (>20%); stool consistency: 0 (normal), 2 (loose stool), and 4 (diarrhea); and bleeding: 0 (no blood), 2 (slight visual bleeding), and 4 (gross bleeding, blood around anus). The disease activity score represents the sum of the individual values for stool, blood, and weight loss.

MOG79-96–Induced EAE.

Mice were immunized with a 100 µL of emulsion containing 100 µg of myelin-oligodendrocyte glycoprotein (MOG) 79-96 peptide in PBS and 50 µL of complete Freund’s adjuvant (Difco). All of the animals were boosted with 200 ng of Bordatella pertussis toxin (Sigma Aldrich) in PBS intraperitoneally on day 0 and 48 h after initial immunization. From day 7 onward, mice were monitored for disease development as follows: 0, no clinical signs of disease; 1, tail weakness; 2, tail paralysis; 3, tail paralysis and mild waddle; 4, tail paralysis and severe waddle; 5, tail paralysis and paralysis of one limb; 6, tail paralysis and paralysis of two limbs; 7, tetraparesis; 8, moribund or deceased.

Assessment of BMD.

One femur was subjected to a peripheral quantitative computed tomography (pQCT) scan with a Stratec pQCT XCT Research M, software v5·4B (Norland) at a resolution of 70 mm. Trabecular BMD was determined with a metaphyseal scan positioned proximal from the distal growth plate at a distance corresponding to 3% of the length of the femur. The inner 45% of the area was defined as the trabecular bone compartment. Cortical BMD was determined with a middiaphyseal scan.

Serum 25(OH)D₃.

Serum 25(OH)D₃ was quantified using Mouse Rat 25-OH Vitamin D ELISA (Eagle Biosciences).

Serum Calcium.

Calcium ion concentration in serum was determined using the Calcium Colorimetric Assay Kit (Sigma).

Serum PTH.

PTH levels were measured from serum using the PTH/Parathyroid Hormone EIA Kit (Sigma).

Statistical Analysis.

Statistical analysis was performed using GraphPad Prism v6.0. Statistical comparison of two unpaired groups was carried out using Mann–Whitney U nonparametric test unless stated otherwise. P values under 0.05 were considered statistically significant and are denoted with *P < 0.05 or **P < 0.01.

Data Availability.

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository (86), http://proteomecentral.proteomexchange.org/cgi/GetDataset?ID=PXD019876. All other data have been deposited at https://figshare.com/articles/PNAS_MS_2020-01966R_data_xlsx/12496925.