Fasting-Induced Transcription Factors Repress Vitamin D Bioactivation, a Mechanism for Vitamin D Deficiency in Diabetes

Vitamin D is an endocrine regulator of calcium homeostasis, but emerging evidence has indicated a role in the regulation of energy homeostasis (1,2). There is strong evidence that vitamin D affects adipogenesis and adipocyte lipid metabolism; however, the effects vary between species and cell models used (1,3). Both Vdr^−/− and Cyp27b1^−/− mice display a lean phenotype with decreased fat mass, probably because of increased energy expenditure (1,4). In contrast, in humans, low 25-hydroxyvitamin D (25-OH-D) associates with obesity (1,2).

Vitamin D deficiency is a widespread medical health problem worldwide (5). Ample epidemiological evidence associates vitamin D deficiency with the prevalence of metabolic diseases, and low 25-OH-D levels have been reported to correlate with the incidence of type 1 and type 2 diabetes (6–9). However, the causal relationship is uncertain, and the mechanisms involved are unclear. Furthermore, the intervention studies aiming at diabetes prevention with vitamin D supplementation have been largely disappointing or inconclusive (6).

Vitamin D is a prohormone activated in two enzymatic steps: 25-hydroxylation in the liver and 1α-hydroxylation in the kidney to produce the main active form 1α,25-dihydroxyvitamin D (1α,25-(OH)₂-D) (10). Vitamin D status is usually assessed by measuring the main circulating vitamin D metabolite (i.e., 25-OH-D) (5). Cytochrome P450 (CYP) 2R1 is the predominant vitamin D 25-hydroxylase in the liver (11,12). A genetic defect in the CYP2R1 gene has been shown to cause an inherited form of vitamin D deficiency and rickets in children (13–15). Large-scale studies have identified CYP2R1 gene variants as one of the major genetic determinants of low 25-OH-D levels (16–18). Additionally, polymorphism in the CYP2R1 gene has been associated with susceptibility to type 1 diabetes (19,20). However, little is known about the regulation of CYP2R1, and vitamin D 25-hydroxylation has even been considered an unregulated step (10).

CYP24A1 enzyme catalyzes 24-hydroxylation of 1α,25-(OH)₂-D and 25-OH-D (10). 24-Hydroxylation is the main inactivation step limiting the effect of the main vitamin D receptor (VDR) ligand 1α,25-(OH)₂-D but also reduces the pool of 25-OH-D available for 1α-hydroxylation. CYP24A1 plays an important role in the regulation of vitamin D action and is under stringent regulatory control (10).

In this report, we show that CYP2R1 and its catalytic activity vitamin D 25-hydroxylation are suppressed in the liver during fasting and in both type 1 and type 2 diabetes mouse models. Mechanistically, we demonstrate involvement of at least two molecular pathways: the peroxisome proliferator–activated receptor γ coactivator 1-α (PGC-1α)/estrogen-related receptor α (ERRα) axis and the glucocorticoid receptor (GR). Furthermore, CYP24A1 is induced during fasting under the control of PGC-1α. Altogether, these results indicate that energy metabolism–regulating factors control vitamin D metabolism and establish repression of vitamin D bioactivation as an important, novel mechanism inducing vitamin D deficiency in diabetes.

All animal experiments were approved by the National Animal Experimental Board, Finland, or the local animal committees (the Animal Studies Committee of Washington University School of Medicine, St. Louis, MO, and the Animal Use and Care Committee of Semmelweis University, Budapest, Hungary) or the local government authorities (Bezirksregierung Köln/Landesamt für Natur, Umwelt und Verbraucherschutz, North Rhine-Westphalia, Cologne, Germany). All animals were housed in standard conditions with a 12-h dark-light cycle. If not otherwise stated, at the end of the experiments, the mice were killed with CO₂ inhalation and neck dislocation, and tissues were collected and frozen in liquid nitrogen. All animals were male.

DBA/2 male mice, obtained from the Laboratory Animal Center, University of Oulu, aged 8–12 weeks, were fasted for 12 or 24 h. The animals had free access to drinking water. At the end of the experiment, the mice were anesthetized with a solution containing fentanyl-fluanisone (Hypnorm) and midazolam (Dormicum) and sacrificed. Independent repetition of the experiment gave similar results.

Pgc-1α^−/− mice (21), aged 3–4 months, and Pgc-1α^+/+ littermates were fasted for 12 h, after which the animals were sacrificed and the tissues collected. Liver-specific Pgc-1β^−/− mice (22) or Pgc-1β^+/+ littermates, aged 8 weeks, were housed individually and fasted for 24 h. Animals were sacrificed and tissues harvested and snap frozen in liquid nitrogen. Both knockout (KO) lines were in the C57BL/6 background. Wistar rats, aged 10 weeks, were fasted for 24 h as has been described previously (23).

C57BL/6 mice, aged 8 weeks, were fed a high-fat diet (HFD) (60% fat, TD.06414; Envigo) or regular chow (5% fat, TD.2018; Envigo) for 16 weeks. The mice were fasted for 12 h overnight, and fasting blood glucose was determined. The mice were sacrificed by CO₂ inhalation, and blood was drawn into an EDTA-primed syringe from vena cava and tissues collected. The chow diet and HFD contained 39.39 μg/kg and 52.35 μg/kg vitamin D, respectively. The estimated amount of the chow diet eaten by a mouse was 4 g/day. However, the HFD-treated mice ate less (i.e., 3.1 g/day). Therefore, the daily intake of the vitamin D was ∼0.16 μg/day for both the chow and the HFD-treated mice.

Experimental type 1 diabetes was induced by intraperitoneal (i.p.) injections of C57BL/6 mice with streptozotocin (STZ) as has been described previously (24,25).

For ERRα inhibition in the kidney, the C57BL/6 mice, aged 8 weeks, were treated i.p. with an ERRα inverse agonist, XCT790 (0.48 mg/kg) dissolved in DMSO/corn oil or vehicle once daily for 3 days. Twelve hours after the last XCT790 injection, mice were either fed or fasted overnight for an additional 12 h, and the tissues were collected.

For the dexamethasone treatment, C57BL/6 mice, aged 8 weeks, were fasted 1 h before an i.p. injection of dexamethasone (3 mg/kg) dissolved in DMSO/corn oil or vehicle only. The mice were fasted 6 h and the tissues collected. In some experiments, simultaneously with the dexamethasone injections, the GR was inhibited with the GR antagonist mifepristone (i.p. 100 mg/kg) dissolved in DMSO/corn oil. In all the treatment groups, the vehicle amount was kept similar.

To inhibit GR in the liver of fasting mice, C57BL/6 mice, aged 8–9 weeks, were injected with mifepristone (i.p. 100 mg/kg) or vehicle (DMSO/corn oil) two times (the first injection at 9 a.m. and the second injection at 9 p.m.), and subsequently, the mice were either fed or fasted overnight for an additional 12 h, and the tissues were collected.

Mouse primary hepatocytes were isolated from male DBA/2 (Figs. 2A and 4A) or C57BL/6 mice (Laboratory Animal Center, University of Oulu), aged 8–10 weeks, as described (26) and cultured in William’s E medium containing 2.0 g/L d-glucose and insulin 5 mg/L, transferrin 5 mg/L, sodium selenate 5 μg/L, and 10% FBS (Sigma-Aldrich, St. Louis, MO). The cultures were maintained for an additional 12–24 h in serum-free William’s E medium before adenoviral infections or chemical treatments. HepG2-cells (ATCC, Manassas, VA) were maintained in basic DMEM (4.5 g/L glucose) supplied with 10% FBS and 1% penicillin-streptomycin (all Gibco, Invitrogen, Carlsbad, CA).

From fasted male DBA/2 mouse livers, RNAs were extracted with the cesium chloride centrifugation method (27). From all the other tissues and cell samples, total RNA extraction was performed using either TRI reagent or RNAzol reagent (Sigma-Aldrich) according to the manufacturer’s protocol. One microgram of RNA was used for cDNA synthesis using p(dN)6 random primers (Roche Diagnostics, Mannheim, Germany) and Maloney murine leukemia virus reverse transcriptase (Promega, Madison, WI) or RevertAid Reverse Transcriptase (Thermo Fisher Scientific, Waltham, MA). The quantitative PCR (qPCR) reactions were performed using SYBR Green chemistry or TaqMan chemistry (Applied Biosystems, Foster City, CA). The sequences for the primers and TaqMan probes are listed in Table 1. The fluorescence values of the qPCR products were corrected with the fluorescence signals of the passive reference dye (ROX). The mRNA levels of target genes were normalized against 18S rRNA, GAPDH, or TBP control levels using the comparative C_T (ΔΔC_T) method.

PGC-1α-2x9 and PGC-1α-L2L3M plasmids were provided by Dr. Donald McDonnell (Duke University School of Medicine, Durham, NC). Recombinant adenoviruses expressing green fluorescent protein (GFP-Ad), LacZ (LacZ-Ad), and PGC-1α (PGC-1α-Ad, PGC-1α-2x9-Ad, PGC-1α-L2L3M-Ad) were prepared as described previously (26). For overexpression of PGC-1α, PGC-1α-2x9, and PGC-1α-L2L3M, multiplicity of infection (MOI) 0.5 was used for each virus. ERRα-Ad was purchased from Vector Biolabs (Malvern, PA). Mouse primary hepatocytes were infected with adenoviruses in William’s E growth medium without serum for the indicated time periods before RNA or protein extractions. Adenoviruses containing scrambled short hairpin RNA (shRNA) (shScr-Ad) and ERRα-targeting shRNA (shERRα-Ad) were purchased from Vector Biolabs. For the shRNA experiments, mouse primary hepatocytes were first infected with either shScr or shERRα at MOI 30 in William’s E medium for 24 h, after which cells were infected with either PGC-1α-Ad or control virus LacZ-Ad at MOI 2. After 48 h, the cells were collected and RNA isolated. The efficiency of the knockdown was tested by measuring ERRα mRNA by qPCR. For inhibition of the ERRα by XCT790 in combination with the PGC-1α overexpression, PGC-1α-Ad was used at MOI 2.

CYP2R1 was detected from mouse liver microsomal fractions. Microsomal fractions were prepared by differential centrifugation (28), and protein content was measured by Bradford reagent (Bio-Rad, Hercules, CA). CYP24A1 was detected from total protein fractions prepared as described previously (29). Protein fractions were subjected to precast SDS-PAGE (10–12% in polyacrylamide) (Bio-Rad) and transferred to polyvinylidene fluoride or nitrocellulose membrane (Millipore, Billerica, MA). Membranes were incubated with appropriate primary antibody in 5% skimmed milk or Amersham ECL Prime Blocking Reagent in Tris-buffered saline with 0.1% Tween, usually overnight, followed by secondary horseradish peroxidase–conjugated antibody incubation. The immunoreactive bands were visualized with Chemiluminescent Peroxidase Substrate-1 reaction (Sigma-Aldrich), and Amersham ECL start Western Blotting Detection Reagent (GE Healthcare, Little Chalfont, U.K.) and quantified by Quantity One or Image Studio software.

Mouse liver microsomal samples (0.5 mg/mL protein) were subjected to incubation with 2 μmol/L cholecalciferol together with 0.1 mol/L PBS and preincubated for 5 min. Enzymatic reactions were initiated with the addition of 0.5 mmol/L NADPH and quenched using ice-cold acetonitrile at a 1:1 volume ratio after 40-min shaking at 37°C. Samples were mixed well and kept at −20°C until analyzed using liquid chromatography–tandem mass spectrometry (LC-MS/MS).

Quantitative analysis of 25-OH-D was performed by reversed-phase LC (ACQUITY UPLC; Waters, Milford, MA) combined with MS detection (Xevo T-QS triple quadrupole mass spectrometer; Waters). The chromatographic separation was carried out with an ACQUITY UPLC BEH Shield RP18 column (2.1 × 50 mm, 1.7 μm) (Waters). The column temperature was set to 45°C, and a gradient elution with mobile phase A (0.5% formic acid) and mobile phase B (15% isopropanol and 85% acetonitrile) at a flow rate of 500 μL ⋅ min⁻¹ was used. The elution gradient consisted of raising the part of mobile phase B from 20% up to 90% in 3.5 min followed by column equilibration until 4.5 min. The injection volume was 4 μL. The retention time of 25-OH-D was 2.82 min. The monitored fragmentation reaction was charge/mass ratio 383 > charge/mass ratio 211. Positive mode of electrospray ionization was used as the ionization source. Data were processed with MassLynx MS version 4.1 software (Waters).

The plasma level of the 25-OH-D was measured using a 25-Hydroxy Vitamin Ds EIA kit (Immunodiagnostic Systems, Tyne and Wear, U.K.) by ValiFinn (Oulu, Finland) according to the manufacturer’s protocol.

The mouse Cyp2r1 gene promoter ERRα binding sites were predicted using MatInspector software (30) (Genomatix, Munich, Germany). The 1.2 kb (−10 to −1,220 bp, relative to the transcription start site [TSS]) promoter region of the mouse Cyp2r1 gene was amplified with PCR from the C57BL/6 mouse DNA by using forward primer TTCTCGAG-CTTCAAGCCTTAAAATGATGTGAG (TT is extranucleotide for efficient binding of the restriction enzyme, CTCGAG is the XhoI restriction site) and reverse primer TTAAGCTT-CTACGAACCAGTCCGGAGC (TT is extranucleotide for efficient binding of the restriction enzyme, AAGCTT is the HindIII restriction site) and inserted upstream of the firefly luciferase gene reporter in the pGL3-basic vector. The pGL3 vector containing the wild-type (WT) promoter construct was subjected to site-directed mutagenesis using a QuikChange II Site-Directed Mutagenesis Kit to generate a construct containing the promoter sequence with mutated ERRα binding site (−1,117 to −1,122 bp, relative to the TSS).

The Cyp2r1 promoter constructs were transiently transfected into HepG2 cells using the FuGene transfection reagent (Promega) together with the pRL-TK Renilla luciferase reporter to normalize for transfection efficiency. Empty pGL3-basic vector was used as a negative control. Twenty-four hours after transfection, PGC-1α and ERRα were overexpressed using the adenovirus at MOI 2. The LacZ-Ad–infected cells were used as a negative control. Cells were incubated for a further 24 h, and the luciferase activities were measured using Dual-Glo Luciferase Assay System (Promega) and Varioskan Flash equipment (Thermo Fisher Scientific). The firefly luciferase values were normalized with the Renilla luciferase values. The data are expressed as relative to the LacZ-Ad–infected cells.

The DNA microarray experiment to study the PGC-1α–regulated genes in mouse (C57BL/6) primary hepatocytes was done as previously described (23). Microarray data can be accessed at the National Center for Biotechnology Information Gene Expression Omnibus (GEO), with the accession number GSE114485.

The published chromatin immunoprecipitation sequencing (ChIP-seq) data PPARGC1A and ESRRA ChIP-seq in HepG2 (accession number GSE31477) (31) and NR3C1 (GR) ChIP-seq in mouse liver (GSE72084) (32) were retrieved and analyzed by using the Cistrome database (33) and visualized using the University of California, Santa Cruz, genome browser.

The statistical data analysis was performed using GraphPad Prism software (GraphPad, La Jolla, CA). Unless otherwise stated, the comparison of means of two groups was done by Student two-tailed t test, whereas multiple groups were compared by one-way ANOVA followed by Tukey post hoc test. Differences were considered significant at P < 0.05.

We observed that fasting represses CYP2R1 expression in the mouse liver in vivo. Remarkably, the expression of CYP2R1 mRNA was strongly repressed by 50% already after 12-h fasting and further suppressed by 80% after 24 h (Fig. 1A). CYP2R1 expression was regulated in negative correlation with the fasting-induced gluconeogenic gene phosphoenolpyruvate carboxykinase (PEPCK) (r = −0.605, P = 0.0011) (Fig. 1B and Supplementary Fig. 1A). Furthermore, CYP2R1 protein was effectively decreased at both time points to 45% after 12 h and 33% after 24 h compared with controls (Fig. 1C). Consistent with the mRNA and protein results, fasting strongly decreased the liver microsomal vitamin D 25-hydroxylase activity to 54% after 12 h and below the detection level after 24-h fasting (Fig. 1D). However, in accordance with the long half-life of the metabolite, the plasma level of 25-OH-D was not affected by short-term, 12-h fasting compared with the fed controls (Supplementary Fig. 1B). In addition to mouse, fasting repressed CYP2R1 expression in rat liver after 24-h fasting (Fig. 1E).

Hepatic CYP2R1 was repressed also in the mouse diabetes models. In the HFD-induced mouse model of obesity and type 2 diabetes (Supplementary Fig. 1C and D), hepatic CYP2R1 mRNA was repressed by 45% (Fig. 1F). Consistent with the CYP2R1 mRNA, the plasma level of 25-OH-D was significantly reduced in the HFD-treated mice compared with the chow-fed controls (Fig. 1G). Also in the type 1 diabetic mouse model (Supplementary Fig. 1E), induced with STZ, CYP2R1 mRNA was repressed by 43% and protein by 29% (Fig. 1H and I). Analysis of published microarray data (accession number GSE39752) supports the finding that CYP2R1 is repressed in the livers of STZ-treated mice (Supplementary Fig. 1F). Altogether, these data show a clear modulation of vitamin D bioactivation by the metabolic state. CYP2R1 expression and the vitamin D 25-hydroxylation were markedly repressed in the livers of fasted as well as diabetic animals.

We next investigated the mechanisms of CYP2R1 repression and hypothesized that nutrition-responsive coactivator PGC-1α would be involved in this process since PGC-1α plays a central role in the fasting response and in uncontrolled diabetes (34). PGC-1α overexpression in mouse primary hepatocytes with PGC-1α-Ad downregulated CYP2R1 strongly and dose dependently, resulting in only an 11% expression at MOI 1 compared with GFP-Ad control (Fig. 2A). To explore the mechanism in more detail, we transduced mutant PGC-1α into hepatocytes (Supplementary Fig. 2A). Gaillard et al. (35) described PGC-1α mutants selective for nuclear receptor interactions; PGC-1α-L2L3M mutant is unable to bind any nuclear receptors, whereas PGC-1α-2x9 mutant interacts selectively with nuclear receptors ERRα or HNF-4α. Interestingly, PGC-1α-2x9 downregulated CYP2R1 expression almost similarly to the WT (Fig. 2B), whereas the L2L3M mutation abolished the CYP2R1 repression and even resulted in weak induction (Fig. 2B). These results indicate that an interaction with a nuclear receptor, most probably ERRα, is indispensable for PGC-1α–mediated CYP2R1 suppression. Supporting this hypothesis, several ERRα target genes (35) were upregulated by the WT and the PGC-1α-2x9 mutant (Supplementary Fig. 2B–D). Moreover, the majority of the PGC-1α-2x9 mutant–induced genes are dependent on ERRα (35).

Several approaches were used to explore the role of ERRα in the regulation of CYP2R1. First, we performed ERRα knockdown by shERRα-Ad virus combined with PGC-1α-Ad treatment in mouse primary hepatocytes. PGC-1α-Ad induced the expression of ERRα about fourfold, as expected (36) (Supplementary Fig. 2E and F). ERRα knockdown abolished CYP2R1 suppression by PGC-1α-Ad (Fig. 2C). Furthermore, we confirmed the role of ERRα by using ERRα inverse agonist XCT790. Indeed, 2 μmol/L XCT790 prevented CYP2R1 repression by PGC-1α (Fig. 2D) without any effect on PGC-1α or ERRα expression (Supplementary Fig. 2G and H). Analysis of public Encyclopedia of DNA Elements data indicates that PGC-1α and ERRα bind to two common regions within the human CYP2R1 gene in HepG2 cells (31) (Supplementary Fig. 2M). We performed a bioinformatics promoter analysis of the mouse Cyp2r1 gene with MatInspector software and identified a potential ERRα binding site in the proximal promoter. When −1.2 kb Cyp2r1-5′-luciferase-reporter construct was transfected into human hepatoma HepG2 cells and the cells were infected with PGC-1α-Ad, luciferase activity was repressed; however, mutation of the ERRα binding site at position −1,117 to −1,122 bp (relative to the TSS) abolished the PGC-1α response (Fig. 2E). Interestingly, ERRα-Ad did not have an effect on the luciferase activity, indicating a crucial need for PGC-1α (Fig. 2E). Altogether, these data indicate that ERRα plays a novel, indispensable role in PGC-1α–mediated downregulation of CYP2R1 expression in mouse hepatocytes.

Next, we investigated the CYP2R1 repression mechanism in vivo by using PGC-1α KO mice fasted for 12 h. Interestingly, PGC-1α mRNA was not significantly induced by fasting in the livers of the Pgc-1α^+/+ animals compared with the fed control animals (Supplementary Fig. 2I). On the other hand, PEPCK expression was significantly induced 4.9-fold by fasting in the Pgc-1α^+/+ mice and 3.4-fold in the Pgc-1α^−/− mice, indicating a fasting response (Supplementary Fig. 2J). CYP2R1 expression was repressed in the fasted Pgc-1α^+/+ mouse livers down to 25% compared with the fed controls (Fig. 2F). Interestingly, PGC-1α KO was not sufficient to abolish the CYP2R1 repression by fasting (Fig. 2F).

Curiously, the expression of PGC-1β was induced more potently in the livers of fasted PGC-1α KO mice (6.1-fold) compared with the WT mice (2.0-fold) (Fig. 2G). In addition, ERRα induction was significant only in the PGC-1α KO mice (Supplementary Fig. 2K). We therefore hypothesized that PGC-1β, a close relative of PGC-1α as well as a fasting-inducible factor (37), could compensate for the chronic loss of PGC-1α in the KO mice or could even play an independent role in the suppression of CYP2R1.

We therefore investigated the fasting effect on CYP2R1 also in the liver-specific PGC-1β KO mice. In 24-h fasted WT mice, CYP2R1 was again repressed by 35% (Fig. 2H). However, PGC-1β KO did not reverse the repression of CYP2R1 (Fig. 2H). PGC-1β KO did not affect the PGC-1α induction by fasting (Supplementary Fig. 2L).

GR activated by cortisol is another key pathway controlling the fasting response in the liver and is also activated in diabetes (38). To investigate the role of GR in the regulation of CYP2R1, we treated mice with a GR agonist, dexamethasone, for 6 h. The treatment decreased liver CYP2R1 mRNA levels by 49% along with changes in the expression of several known GR target genes, including ANGPTL8, NR1D1, and TAT (38,39), and increased PGC-1α levels 3.3-fold (Fig. 3A), suggesting involvement of GR in the regulation of CYP2R1 in vivo. GR is also a known interaction partner for PGC-1α (34). The CYP2R1 protein was decreased 26% by 6-h dexamethasone treatment (Fig. 3B). Furthermore, analysis of the published microarray data (accession number GSE24256) (40) supports the effect of dexamethasone on CYP2R1 expression in mouse liver (Supplementary Fig. 3A).

To verify the role of GR, we cotreated mice with dexamethasone and the GR antagonist mifepristone. Mifepristone completely prevented induction of PGC-1α and TAT by dexamethasone and inhibited repression of CYP2R1 and NR1D1 (Fig. 3C–F). Furthermore, analysis of published ChIP-seq data indicates that GR binds to Cyp2r1 gene proximal promoter in mouse liver (41) (Supplementary Fig. 3B). To evaluate the role of GR in the fasting-mediated repression of CYP2R1, we investigated the effect of pharmacological inhibition of GR by mifepristone during fasting. Mifepristone abolished the repression of the control gene ANGPTL8 (Fig. 3H). Furthermore, mifepristone partially, but significantly, prevented the effect of fasting on the CYP2R1 in the liver (repression decreased from 56 to 18%); however, the fasting effect still remained statistically significant (Fig. 3G).

DNA microarray analysis of PGC-1α–responsive genes in mouse hepatocytes indicated that CYP24A1, the vitamin D 24-hydroxylase and the main inactivator of active vitamin D (42,43), was among the most upregulated genes (Table 2). This was further confirmed by PGC-1α-Ad dose-response experiments using qPCR (Fig. 4A).

CYP24A1 is primarily regulated by VDR (44), and VDR is an interaction partner for PGC-1α (45). We therefore explored the role of VDR in the PGC-1α–mediated induction of CYP24A1. After treating mouse primary hepatocytes with the VDR ligand calcitriol, CYP24A1 was induced 120-fold (Fig. 4B–D). PGC-1α-Ad induced CYP24A1 much more potently, up to 20,000-fold. However, combined treatment with calcitriol did not potentiate CYP24A1 induction by PGC-1α (Fig. 4B). Furthermore, CYP24A1 protein could be detected after PGC-1α-Ad transduction but not after calcitriol treatment (Fig. 4E). PGC-1α-Ad treatment had no effect on VDR expression, but ERRα was induced 11-fold (Fig. 4F and G). Altogether, these results suggest that VDR does not play a major role in the CYP24A1 induction by PGC-1α.

We next investigated the putative role of ERRα in the CYP24A1 induction by PGC-1α. PGC-1α-2x9 mutant induced CYP24A1 similarly to the WT PGC-1α, whereas PGC-1α-L2L3M mutant had no effect, suggesting involvement of ERRα (Fig. 4H). Indeed, induction of CYP24A1 was abolished by ERRα knockdown, indicating a novel PGC-1α-ERRα pathway–mediated regulatory mechanism for CYP24A1 (Fig. 4I).

Although we detected CYP24A1 in hepatocytes after PGC-1α overexpression, we could not detect CYP24A1 in the mouse liver in vivo. Indeed, the main site for CYP24A1 expression and function is the kidney (42). CYP24A1 was induced up to 2.7-fold in the kidney by fasting (Fig. 5A) in agreement with a previous study (46). However, the fasting effect was abolished in the PGC-1α KO mice (Fig. 5B–D), indicating that PGC-1α plays an indispensable role in the CYP24A1 induction by fasting in the kidney. Furthermore, ERRα antagonist XCT790 treatment diminished the fasting-mediated induction of CYP24A1, suggesting that fasting induces CYP24A1 in the kidney through the PGC-1α-ERRα pathway (Fig. 5E–H).

Plasma 25-OH-D level is regularly used as the measure of vitamin D status. The reason behind this is the rather long half-life of 25-OH-D of ∼2 weeks (47). Furthermore, 25-hydroxylation has been considered not to be under efficient metabolic control, and 25-OH-D should thus reflect the intake of vitamin D (48). Several CYP enzymes, including CYP2R1, CYP27A1, CYP2D25, CYP2C11, and CYP3A4, have been reported to be capable of vitamin D 25-hydroxylation in vitro (48). However, genetic studies, expression data, and catalytic properties all suggest that CYP2R1 is the main vitamin D 25-hydroxylase enzyme in liver. Indeed, both the KO studies in mouse and the genetic evidence in humans indicate that defect in the CYP2R1 gene results in vitamin D deficiency (i.e., low plasma 25-OH-D level) (12,13,49). In humans, this has been shown to result in symptomatic rickets, vitamin D–dependent rickets type 1B (15). These patients respond poorly to ordinary vitamin D supplementation but may benefit from 25-OH-D treatment (15).

We now show that the CYP2R1 enzyme may be repressed also functionally at the level of gene regulation. Twelve-hour fasting suppressed liver microsomal vitamin D 25-hydroxylation ∼50%, and after 24-h fasting, we were unable to detect any 25-OH-D formation. Thus, the first vitamin D bioactivation step is under the strict control of nutritional state. Although the acute food deprivation resulted in a strong effect on vitamin D 25-hydroxylase activity, this was not reflected in the plasma 25-OH-D concentration, presumably because of the long half-life of 25-OH-D (47). Therefore, it seems unlikely that short-term fasting would have a significant effect on vitamin D functions at the systemic level. This raises the question of the physiological purpose of the CYP2R1 repression during fasting. A likely explanation is that fasting launches physiological adjustment as precaution for possible longer-term food shortage. This may potentially be related to the role of vitamin D in energy homeostasis (1) and could have been evolutionarily beneficial during periods of starvation. Alternatively, 25-OH-D could have some unknown local function in liver. Furthermore, we observed induction of CYP24A1 in the kidney during fasting. This is a mechanism that limits the level of 1α,25-(OH)₂-D and consequently activation of VDR (10). The CYP24A1 induction and the CYP2R1 repression are expected to suppress vitamin D signaling in a synergistic manner.

The suppression of vitamin D bioactivation by fasting-activated mechanisms has important implications in the context of human metabolic diseases. Hepatic signaling pathways triggered physiologically during fasting display typically prolonged, constant activation in diabetes. The classical consequence is increased activation of gluconeogenesis, resulting in fasting hyperglycemia (50). Since in diabetes, unlike in short-term fasting, the activation of these molecular mechanisms is long-standing, the suppression of vitamin D 25-hydroxylation will eventually lead to a lower plasma 25-OH-D level. In agreement with this theory, we observed reduced plasma 25-OH-D concentrations in the HFD-treated mice. Thus, we propose that repression of vitamin D bioactivation represents a novel mechanism that plays a role in vitamin D deficiency in diabetes.

PGC-1α is one of the major molecular factors regulating gluconeogenesis and other metabolic pathways activated in the diabetic liver (34,50). We now show that PGC-1α, ERRα dependently, also represses vitamin D bioactivation and, thus, establishes regulation of vitamin D metabolism as a novel metabolic function under the control of PGC-1α. Furthermore, the CYP24A1 induction by fasting in the kidney was demonstrated to be under the control of PGC-1α-ERRα. Thus, the PGC-1α-ERRα pathway appears to play a major role in the crosstalk between energy homeostasis and vitamin D metabolism. Interestingly, a recent study showed that Cyp2r1-deficiency in zebrafish affected lipid metabolism through vitamin D–regulated function of PGC-1α (51). PGC-1α and vitamin D metabolism could thus form a regulatory loop.

Although, the PGC-1α-ERRα pathway was found to be an effective regulator of CYP2R1, the PGC-1α KO did not prevent CYP2R1 suppression during fasting. This indicates that additional molecular mechanisms play a role in the regulation of CYP2R1 during fasting. Activation of GR was found to be a second mechanism capable for CYP2R1 suppression. Indeed, cortisol levels are increased during fasting as well as in diabetes (38). By using pharmacological inhibition of GR, we could partially prevent fasting-mediated repression of CYP2R1, suggesting that GR is involved in CYP2R1 repression by fasting. However, we cannot exclude the possibility that additional regulatory mechanisms mediate CYP2R1 repression by fasting. From the point of view of drug therapy, the observed repression of CYP2R1 by pharmacological glucocorticoid treatment may explain the observed association between glucocorticoid use and vitamin D deficiency (52).

In summary, our results reveal a novel crosstalk between energy homeostasis and the vitamin D pathway, suggesting a physiological need for suppression of vitamin D signaling during nutrient deprivation (Fig. 6). This may be related to the role of vitamin D in energy metabolism (1,3). Altogether, our study provides a mechanism that may explain the lower vitamin D levels in patients with diabetes and suggests that vitamin D deficiency is a consequence, not the cause, of diabetes.

Acknowledgments. The authors thank Ritva Tauriainen (University of Oulu) for technical assistance. The help of Dr. Anastasia Georgiadi and Sander Kersten (University of Wageningen, Wageningen, the Netherlands) with the microarray study is acknowledged.

Funding. The study was financially supported by the Scholarship Fund of the University of Oulu (Tyyni Tani Fund) to M.-S.E., the Academy of Finland (grants 267637, 292540) and the Sigrid Juselius Foundation to P.T., the Cologne Excellence Cluster on Cellular Stress Responses in Aging-associated Diseases and Center of Molecular Medicine Cologne of the Medical Faculty to R.J.W., National Institutes of Health grant R01-DK-104735 to B.N.F., and the Academy of Finland (grant 286743) and the Diabetes Research Foundation to J.H.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. S.-M.A. and M.-S.E. performed a majority of the experiments and measurements. S.-M.A., M.-S.E., and J.H. designed the study, analyzed the data, and wrote the manuscript. S.-M.A. and M.B. performed the PGC-1α microarray experiment. O.K. and M.K. performed the HFD experiment and helped with the other animal experiments. P.V. performed the 12- and 24-h fasting experiments in mice and helped with the hepatocyte primary cultures. V.R. performed measurements of the 25-OH-D by LC-MS/MS. M.M. and P.T. performed the PGC-1α KO mouse experiments. A.F. and R.J.W. performed the STZ mouse study. K.T.C. and B.N.F. performed the liver-specific PGC-1β KO mouse study. J.H. supervised the overall conduct of the study. All authors read and approved the final manuscript. J.H. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Data Availability. Data sharing: The data sets generated and/or analyzed during the current study are available in the GEO repository (our microarray for PGC-1α overexpression in mouse primary hepatocytes, accession number GSE114485; published microarray for dexamethasone treatment in mouse livers, GSE24256; published PPARGC1A and ESRRA ChIP-seq in HepG2, GSE31477; and published NR3C1 [GR] ChIP-seq in mouse liver, GSE72084). Resource sharing: The resources generated and/or analyzed during the current study are available from the corresponding author on reasonable request.

Prior Presentation. Parts of this study were presented in abstract form at the 10th International Meeting of the International Society for the Study of Xenobiotics, Toronto, Ontario, Canada, 29 September–3 October, 2013; the 6th Sino-Finn Life Science Forum: From Systems Biology to Translational Medicine, Helsinki, Finland, 17–18 August 2015; Nuclear Receptors: From Molecules to Humans, Ajaccio, France, 24–28 September 2015; the 4th Helmholtz-Nature Medicine Diabetes Conference, Munich, Germany, 18–20 September 2016; and European Molecular Biology Organization/European Molecular Biology Laboratory Symposium: Metabolism in Time and Space: Emerging Links to Cellular and Developmental Programs, Heidelberg, Germany, 11–13 May 2017.