The use of C57BL/6 mice and the study protocol were approved by the Animal Care of Peking Union Medical College (PUMC). All mice were housed in a barrier facility under standard conditions according to the protocols of the PUMC vivarium. All researches were carried out with the approval of the PUMC ethics committee.

Herein, 8–10-week-old C57BL/6 mice were 1) given 1.5% DSS in their drinking water for 6 days in the colitis group (n = 10), 2) fed in a hypoxic chamber (oxygen content 8%) for 18 h in the hypoxia group (Cosin-Roger et al., 2017) (n = 12), which was generally well-tolerated, and 3) not administered either treatment in the control group (n = 6). We conducted another experiment of DSS-induced colitis (2.5% DSS for 7 days) treated with Dimethyloxallyl Glycine (DMOG) (Milkiewicz et al., 2004; Cummins et al., 2008) (#sc-200755, Sigma-Aldrich, United States). On days 0, 2, 4, 6, and 8, the DSS + DMOG group was given DMOG i. p. (8 mg in 0.5 ml saline) while DSS colitis group were injected with sterile saline (0.5 ml) at the same intervals (n = 5). Control group (Ctrl) receive normal water and sterile saline (n = 5).

Establishment: We used the clustered regularly interspaced short palindromic repeats (CRISPR)/spCas9 technology to obtain VDR knockout (KO) homozygous mice. The conserved functional region Exon3 (Exon3) in the VDR genome structure was selected as the target.

Identification: After weaning, the mice tail or toes were sampled and lysed for polymerase chain reaction (PCR) to obtain the genotype. The PCR primer design strategy and identification method for genotyping were shown in the Supplementary Figure S1.

Rearing and phenotype: Reared in the Animal Laboratory of the Chinese Academy of Medical Sciences, VDR heterozygous mice were bred among close relatives, and VDR-KO mice were used for subsequent experiments. Compared with their wild type (WT) siblings, VDR knockout mice showed no obvious differences before weaning. After weaning, VDR knockout mice gradually showed hair loss, bone thinning, atrophy of skin, and reduced fertility began from 3 to 4 months of age (as showed in Supplementary Figure), which were consistent with previous literature (Zhang et al., 2005).We used high-lactose feed (main ingredients: lactose, 20%; vitamin D, 2.2 IU/g; calcium, 2%; and phosphorus, 1.25%) for routine feeding according to the literature (Zhang et al., 2005), which can prolong the lifetime of VDR knockout mice and alleviate systemic manifestations of vitamin D deficiency, such as rickets and skeletal dysplasia; the mice were specific pathogen-free and exposed to a 12-h/12-h night/day cycle.

Intestinal barrier permeability was measured using Cellvizio® small animal live confocal fluorescence microscope (Mauna Kea Technologies, France). The azo dye Evans Blue (#E8010, Solarbio, China) was injected through venae angularis to locate the blood vessels. Fluorescein isothiocyanate (FITC)–dextran (70 kD) (#46945, Sigma-Aldrich, United States) was administered through the anus to check the permeability of the intestinal barrier by observing whether it penetrated into the microvessels of the intestinal mucosa. The probe (Z-1800) was advanced 2–3 cm into the anus, and 10 fields of view were observed. The collected image depth was 70 μm; the field of view was 600 × 600 μm; and the laser wavelengths were 488 and 660 nm. Red fluorescence represented blood vessels and green fluorescence represented the presence of FITC–dextran. A mix of red and green fluorescence indicated the leakage of 70-kD FITC–dextran into the microvessels and reflected increased intestinal barrier permeability.

The mice were subjected to CO₂ anesthesia and then euthanized. The colon was removed, and the length was measured. Then, the whole colonic “Swiss roll” was embedded and sliced for hematoxylin and eosin (HE) staining, and the pathological score was calculated according to the literature (Zhao et al., 2012).

After colon removal, a 1-cm tissue was quickly cut near the anus with a razor blade and immediately put in 2.5% glutaraldehyde solution for subsequent operations. Then, tight junction structure was observed under an electron microscope.

After weaning, the mice were genotyped and confirmed. Then, mice with wild-type VDR ( +, + ) and VDR-KO homozygous VDR (-, -) were housed in separated cages. Feces were collected from each group after 4 weeks of separate housing. Sequencing and data analysis of Miseq (PE300) were performed in the 16S V3–V4 area of the fecal flora.

The human colon cancer cell line DLD-1 and naive colon cell line NCM460 were purchased from the American Type Culture Collection (United States) and cultured in RPMI 1640 (#72400047, Gibco, United States) or DMEM(#11995065, Gibco, United States) medium with 10% fetal bovine serum (FCS, Invitrogen, San Giuliano Milanese, Italy). The normal incubator was set to 5% CO₂ at 37°C, and the hypoxic incubator was set to 94% N₂, l% O₂, and 5% CO₂ at 37°C. The hypoxia group was given 1% O₂ for 48 h, and the solution was changed once every 24 h. Dimethyloxallyl Glycine (DMOG) (#sc-200755, Sigma-Aldrich, United States) was used at concentrations of 1 mmol/l under normal incubator for 24 h (Heinis et al., 2010).

The following plasmids (#HSH018474-LVRU6GP,# HSH100153-LVRU6GP, #EX-T0096-LV105, GeneCopoeia, MD, United States) were co-transfected into 293 T cells:1) the VDR interference plasmid (sh-VDR) (small hairpin RNA, sh-RNA), 2) the HIF-1α interference plasmid (sh-HIF-1α),3) empty plasmid sh-NC (normal control, NC), 4) HIF-1α overexpression (OE) plasmid (OE- HIF-1α),5) empty vector (EV) and lentivirus packaging plasmids 6) pMD2. G and 7) psPAX2. Cell supernatants were collected at 48 and 72 h to obtain sh-VDR, sh- HIF-1α,sh-NC, OE- HIF-1α and EV viruses. Then, DLD-1 or NCM460 were seeded at 20%–30% density in culture dishes. After 48 h of virus infection, they were screened with 6 μg/ml or 2ug/ml puromycin. After 1 week of continuous culture, stable clones were obtained, and subsequent experiments were carried out.

DLD-1 or NCM460 cells in good condition were digested and counted. Then, they were inoculated (100 μL, comprising approximately 1 × 10⁵ cells) in the Transwell chamber. Next, transmembrane resistance of the cells was tested after continuous culture for 1 week.

The cells or colonic tissues were fully lysed, and total protein was extracted, which was separated on SDS–PAGE gel and transferred to a membrane. The primary antibodies ZO-1 (#ab96587,Abcam, UK), E-cadherin (#3195S, Cell Signaling Technology, MA, United States), occludin (#ab216327,Abcam, UK), claudin-1 (#A2196,ABclonal, China), Anti-Carbonic Anhydrase 9(CA9) (#ab243660,Abcam, UK), HIF-1α (#ab228649,Abcam, UK) and Histon H₃ (#4499S, Cell Signaling Technology, MA, United States) were added after blocking nonspecific binding sites with a protein blocker at 4°C for 24 h. Then, they were washed with the Tris–Tween buffer, and secondary antibody was added. This was followed by incubating for 1 h and washing. Then, a chemical luminescent substrate (#A38555, Thermo Fisher Scientific, MA, United States) was added and visualized with a fluorescent chemiluminescence imager. Densitometric analysis of bands was performed using the ImageJ software (version 1.47, National Institutes of Health, NIH, United States).

Cell RNA was extracted, and the genomic DNA was removed. Then, cDNA was synthesized by reverse transcription. PCR reaction cycles were as follows: pre-denaturation at 95°C for 30 s, denaturation at 95°C for 10 s, annealing at 60°C for 30 s, and extension at 95°C for 15 s; this was repeated for 45 cycles. The relative expression level was calculated based on the internal reference.

Chromatin was extracted from DLD-1 cells under normoxia or hypoxia (1% O₂ for 48 h) separately. Rabbit anti-human HIF-1α antibody and rabbit IgG as a negative control were used to immunoprecipitate chromatin. Immunoprecipitated DNA was analyzed by qPCR using the following sets of primers.

VDR.

Forward: 5ʹ–GCC​ACG​CTG​TAG​CCT​TAG​AT–3ʹ.

Reverse: 5ʹ–GTAGAACCACGGCAGGAAGG–3ʹ

Data are presented as fold enrichment compared to IgG control.

The human VDR promoter was cloned by PCR to generate a VDR promoter reporter gene vector (HPRM30353-PG04, GeneCopoeia, MD, United States). HIF-1α transcription factor plasmid (EX-T0096-M02, GeneCopoeia) and control plasmid (NRG-PG04, GeneCopoeia) were used in this experiment. HT-29 cells were co-transfected with NRG-PG04 + EX-T0096-M02, EX-T0096-M02, and HPRM30353-PG04. After 24 h of culture under normoxia or hypoxia, the cells were lysed and analyzed for luciferase activity using Secrete-Pair™ Gaussia Luciferase Assay Kit (GeneCopoeia).

SPSS 21.0 software was used for statistical processing of the results. T-tests or one-way analysis of variance were used to compare the differences in relative expression between groups. The least significant difference method (equal variance) or Dunnett’s T3 method (uneven variance) was used for multiple-group comparisons; p < 0.05 indicated statistically significant difference.

As shown in Figure 1, compared with the DSS group (DSS), the weight loss rate slowed down in the DSS + hypoxia group (Figure 1A), and colon inflammation in the DSS + hypoxia group was lesser, which translated into improved colon length (Figure 1C) and better pathology score (Figure 1D). The expression of HIF-1α and CA9 in the intestinal mucosal were increased after hypoxia treatment (Figure 1B). The destruction of the tight junction structure under electron microscope was mitigated in the DSS + hypoxia group when compared with the DSS group (Figure 2A Figure 2s,t). In addition, IHC also showed increased expression of occludin in the DSS + hypoxia group. Fluorescein leakage experiment in vivo showed blurred blood vessel texture and fused FITC-70 kD (green fluorescein) into blood vessels in the DSS group [Figure 2A (e,i,m,g,k and o)], while the leakage of FITC-70 kD in the capillaries was reduced in the DSS + hypoxia group [Figure 2A (g,k,o,h,l and p)]. Western blotting showed that the expression of tight junction proteins (e.g. ZO-1 and occludin) was higher in the DSS + hypoxia group than in the DSS group (Figure 2B). Furthermore, VDR was elevated in the DSS + hypoxia group compared with the DSS group (Figure 2C).

VDR was indispensable for mucosal barrier function under hypoxia in VDR-KO mice with DSS-induced colitis.

We used VDR-KO mice to further investigate if the absence of VDR abolishes the effect of hypoxia on acute colitis. As shown in Figure 3, compared with the VDR-KO control group, the VDR-KO colitis group (KO + DSS) showed obvious weight loss (Figure 3A), shortened colon (Figure 3B). Histologically, severe inflammation and destroyed colonic epithelium could be seen after DSS treatment (Figure 3C). Previously, we proved that hypoxia can alleviate colitis and intestinal mucosal barrier damage in DSS mice; however, after VDR knockout, hypoxia treatment could not reduce the rate of weight loss (Figure 3A), improve the colon length (Figure 3B), and improve the pathological score (Figure 3C) in VDR-KO colitis mice. Moreover, there were no differences in occludin expression between the KO + DSS group and KO + DSS + hypoxia group (Figure 3D (c,d) and Figure 3E). Under an electron microscope, the VDR-KO group and KO + hypoxia group showed incomplete tight junction structure and slightly blurred edges, which reflected the loss of the barrier structure [Figure 3D (q,r,s and t)]. We performed an in vivo intestinal fluorescein leakage test and identified that the barrier function of the above two groups was also impaired, and some FITC-70 kD (green fluorescein) had penetrated into the blood vessels (red fluorescein) (Figure 3D). After DSS treatment, the red blood vessel texture disappeared, and the green fluorescence and red fluorescence almost completely merged, indicating that the barrier function was lost and FITC had completely infiltrated the blood vessels, and the above situation did not improve after hypoxia treatment (Figure 3D).

To further confirm the role of VDR on the intestinal epithelial barrier under hypoxia, we tested VDR mRNA expression in DLD-1 sh-VDR cells and their control sh-NC cells under 1% O₂ incubation for 48 h. The results showed that hypoxia treatment significantly increased VDR expression in DLD-1 sh-NC cells, as shown in Figure 4D. However, hypoxic stimulation could not induce VDR mRNA expression after VDR knockdown. Furthermore, as showed in Figure 4, in DLD-1 sh-NC cells, hypoxia treatment increased transmembrane electrical resistance value (n = 9, 157.33 ± 3.88 vs 299.89 ± 26.50, p = 0.003); however, there was no difference in the TER value after 1% O₂ incubation in sh-VDR cells (n = 9, 89.44 ± 2.49 vs 90.89 ± 4.95, p = 1.0) (Figure 4E). Accordingly, in NCM460 sh-NC cells, hypoxia treatment increased TER value (n = 6, 93.67 ± 1.21 vs 105.0 ± 3.46, p < 0.001); however, there was no difference in the TER value after 1% O₂ incubation in sh-VDR cells (n = 6, 81.5 ± 4.32 vs 85.67 ± 1.86, p = 0.15), as shown in Supplementary Figure S2C.

Next, we confirmed the interaction between hypoxia and VDR. The Transcription Factor Affinity Prediction software sTRAP (available at http://trap.molgen.mpg.de/cgi-bin/trap_two_seq_form.cgi) suggested the presence of HIF-binding sites in the VDR promoter region. chromatin immunoprecipitation (CHIP) showed (Figures 5A,B) that the enrichment ratio of the hypoxia group was significantly higher than that of the control group, suggesting that HIF-1α can bind to the VDR promoter and mediate downstream signal transcription. To investigate whether the activation of VDR by hypoxia is associated with HIF-1α, the VDR promoter reporter gene vector (HPRM30353-PG04) was co-transfected with the HIF-1α expression vector (EX-T0096-M02) into 293 T cells (Figure 5C). The luciferase assay showed increased luciferase activity with this co-transfection when compared with single transfection of HPRM30353-PG04 (12.065 ± 0.841 vs 9.293 ± 0.166, p < 0.0001; lanes seven and eight of Figure 5D). Moreover, co-transfection of HPRM30353-PG04 and EX-T0096-M02 presented better luciferase activity under hypoxia than under normoxia (12.065 ± 0.841 vs 8.144 ± 0.233, p < 0.0001; lanes four and eight of Figure 5B), which was attributed to the destabilization of HIF-1α in normoxia. In addition, we established HIF-1α-knockdown (Sh-HIF-1α) and HIF-1α over-expression (OE-HIF-1α) cell lines of DLD-1 using lentiviral plasmids. As shown in Figure 5C, over-expression of HIF-1α in DLD-1 can increase the expression of VDR, whereas HIF-1α-knockdown reduces the expression of VDR directly. Moreover we used DMOG to further investigate its effect on HIF-1α and VDR. The results showed that over-expression of HIF-1α by Dimethyloxallyl Glycine (DMOG) in NCM460 and DLD-1 increased the expression of VDR (as in Figure 5D and Supplementary Figure S2D). While HIF-1α-knockdown by lentiviral plasmids reduced the VDR expression in NCM460(as in Supplementary Figure S2E). Furthermore, we conducted another experiment of DSS-induced colitis treated with DMOG. Compared with the DSS group (DSS), the weight loss rate tended to slowed down in the DSS + DMOG group (Supplementary Figure S3A), and colon inflammation in the DSS + DMOG group was lesser, which translated into improved colon length (Supplementary Figures S3B, SC) and better pathology scores (Supplementary Figures S3D, SE). The expression of VDR in the intestinal mucosal tendeded to increase after DMOG treatment (Supplementary Figure S3F).

Previously, we demonstrated that the environment (hypoxia) can affect gene expression. Since many studies have confirmed that genes interact with the environment, VDR is very likely to affect aspects of the intestinal environment, such as intestinal flora. To minimize the interference factors, we used siblings of VDR-KO mice (n = 5) and wild-type mice (n = 5) to identify the changes in the fecal flora of VDR-KO mice. The results showed that six bacteria species were reduced in the feces of VDR-KO mice (Figure 6). These species comprised Desulfovibrio spp. and Odoribacter spp., which are obligate anaerobic bacteria, Ruminiclostridium spp., which are related to fiber degradation and can produce short-chain fatty acids (SCFAs), and Anaerotruncus spp. and Ruminococcaceae spp., which belong to the family Clostridium, and can also produce butyrate, acetic acid, and other SCFAs. Intestinimonas spp. is known for its unique metabolic features of producing butyrate from both sugars and amino acids. Other two bacteria species that were increased in the feces of VDR-KO mice were Parasutterella spp. and Prevotellaceae spp. Parasutterella spp. has recently been found to be related with chronic intestinal inflammation in patients with irritable bowel syndrome (Chen et al., 2018). It shows that the lack of the VDR gene affects the number and composition of intestinal flora and reduces the abundance of anaerobic bacteria and SCFA-producing bacteria.