HDAC6/HNF4α loop mediated by miR-1 promotes bile acids-induced gastric intestinal metaplasia

GES-1, an immortalized gastric epithelial cell line, was used mainly to establish the IM model which was described previously [26]. AGS, MKN45 and AZ521 are gastric cancer cell lines. HCT-116 is one kind of colon cancer cell line. All of the cell lines were purchased from ATCC and resuscitated within 6 months and they were also tested negative for mycoplasma contamination. Gastric cells were cultured in PRIM-1640 medium (Gibco, US), while colon cells were cultured in DMEM medium (Gibco, US) and all cell lines were cultured with 10% fetal bovine serum (Biological Industries, Israel), 100 mg/ml streptomycin and 100 U/ml penicillin. Deoxycholic acid (DCA) is a major hydrophobic BA with strong cytotoxicity and it was purchased from BiocytoSci (USA).

A normal tissue microarray containing 24 cases (BN01011b) and a gastric IM tissue microarray (ST8017a) containing 80 cases were purchased from Alenabio Biotech (China). In addition, 24 normal cases, 15 gastritis cases, and 39 gastric IM cases were also included. Ten paired gastric IM specimens were obtained from patients who underwent endoscopy. To exclude the effect of Hp infection, all selected patients were Hp negative. All patients signed the informed consent forms before the specimens were obtained. Our study was approved by the Human Subjects Committee of Xijing Hospital. The pathological status of these specimens was collected from the Department of Pathology.

The generation of Lgr5-Cre and LSL-Hnf4α mice were generated on a C57BL6 background. Lgr5-Cre mice were bred with LSL-Hnf4α mice to activate Hnf4α in gastric cells. To activate Cre recombinase in Rosa26^Hnf4α mice, the animals were intraperitoneally injected with 5 mg tamoxifen dissolved in corn oil once a day for three consecutive days at 4 weeks of age. The treatment, maintenance and care of mice in this study followed the protocols of the Animal Research Committee of Xijing Hospital.

Immunohistochemistry (IHC) was performed using anti-HDAC6 (1:400, Cell Signaling Technology, #7558) and anti-HNF4α (1:100, Abcam, #ab92378) following the manufacturer’s instructions. In situ hybridization (ISH) was carried out by using a 5-digoxigenin (DIG) and 3′-DIG-labeled locked nucleic acid-based probe specific for miR-1 (Exiqon, Denmark) on tissue microarray. The IHC and ISH staining results were independently evaluated by two professional pathologists independently.

TRIzol^® reagent (Invitrogen, USA) was used to extract the total RNA from cell lines and human tissue samples according to a standard protocol. Then RNA was reverse-transcribed into cDNA using the PrimeScript^® RT Reagent kit (Takara Biotechnology, Japan) and qPCR was conducted using SYBR Premix Ex Taq II (Takara Biotechnology, Japan) on a CFX96™ Real-Time PCR Detection system (Bio-Rad Laboratories, USA). β-Actin and U6 were used as the internal control and the 2^∆∆Cq method was used to quantify the relative mRNA expression of each gene. Sequences of gene-specific primers are provided in Table 1.

Total protein was extracted by mixing cells with a radioimmunoprecipitation assay (RIPA) buffer (Beyotime, China) supplemented with protease and phosphatase inhibitors. Western blot analysis was carried out following standard procedures. The following antibodies were used for this experiment: anti-HDAC6 (1:1000, Cell Signaling Technology, #7558), anti-HNF4α (1:2000, Abcam, #92378), anti-MUC2 (1:4000, Abcam, #ab133555), anti-KLF4 (1:1000, Cell Signaling Technology, #4038), anti-CDX2 (1:1000, Cell Signaling Technology, #12306), and anti-β-actin (1:5000, Bioworld, #AP0060). The primary antibody dilution buffer, PREstain Protein Ladder and Western SuperSensitive Substrate were purchased from BioCytoSci (USA).

Immunofluorescence (IF) staining for HDAC6 and HNF4α was performed in GES-1 cells. The primary antibodies were a rabbit anti-human HDAC6 antibody (1:200, Cell Signaling Technology, #7558), a rabbit anti-human HNF4α antibody (1:200, Abcam, #92378), a rabbit anti-mouse LGR5 antibody (1:25, Abcam, #75732) and a mouse anti-mouse HNF4α antibody (1:50, Abcam, #374229).

Synthetic miR-1 agomir, antagomir, the corresponding negative control oligonucleotides, plasmid and small interfering RNA (siRNA) were purchased from Genepharma (China), and their sequences are shown in Table 2. The transfection reagent was purchased from Thermo Fisher Scientific (USA) and used following the manufacturer’s protocol. HDAC6 and HNF4α overexpression lentiviral vectors were designed and provided by Genechem Co. Ltd. (China).

We amplified a 2000 bp sequence upstream of the transcription start sites of the HDAC6 promoter by PCR and then cloned it into a pGL3-basic luciferase vector (Invitrogen) to generate pHDAC6/2000-Luc, which was used to generate HDAC6-p1 (− 2000 to − 1 bp), HDAC6-p2 (− 1589 to − 1 bp), HDAC6-p3 (− 1544 to − 1 bp), HDAC6-p4 (− 1432 to − 1 bp) and HDAC6-p5 (− 1333 to − 1 bp). The 3′-UTR reporter plasmids of HDAC6 and HNF4α for miR-1 were constructed using the chemically synthesized DNA oligos: HNF4α-wtUTR and HDAC6-wtUTR containing the full-length cDNA sequence, and HNF4α-mutUTR and HDAC6-mutUTR expression vectors lacking the 3′-UTR were also constructed.

GES-1 cells were transiently transfected with an HDAC6 promoter vector. Then, chromatin immunoprecipitation (ChIP) analysis was carried out according to the standard method of the Magna ChIP G Assay kit (EMD Millipore. USA). Chromatin was immunoprecipitated with anti-HNF4α (Abcam, #ab181604) or IgG as a negative control. Finally, immunoprecipitated DNA–protein complexes were isolated and a real-time PCR assay was carried out to examine the quantity of the specific proteins. The primers for the HDAC6 promoters are listed in Table 1.

The SPSS software (V.19.0, SPSS, Chicago, Illinois, USA) statistical package was used to conduct all statistical analyses. All continuous data are expressed as the mean ± SD. The χ² test was used to compare the frequencies of categorical variables. Mutual associations among clinical results were assessed by using Spearman’s rank correlation. Statistical comparisons between two groups were analyzed with the Mann–Whitney U test. Multiple comparisons were performed via a one-way analysis of variance (ANOVA) with the Bonferroni post hoc test. P values less than 0.05 were considered statistically significant.

To better understand the prometaplastic role of Hnf4α in vivo, we constructed an Hnf4α transgenic mouse model that expresses the active form of Hnf4α in Lgr5⁺ gastric stem cells upon tamoxifen exposure (Figs. 1a, S1). The mice were killed and gastric tissues were separated after tamoxifen treatment for 0, 6 and 12 months, respectively. We found that the gastric mucosa of wild-type (WT) mice and Rosa26^Hnf4α mice were both mostly normal at 0 and 6 months (data not shown). Nevertheless, compared with that of WT mice, at 12 months post-tamoxifen, the expression level of Hnf4α in gastric tissues was significantly higher and structural abnormalities were observed in the gastric mucosa of Rosa26^Hnf4α mice (Fig. 1c, d). More importantly, electron microscopy studies revealed that mucin increased in the gastric cells of Rosa26^Hnf4α mice (Fig. 1b). Further, AB-PAS staining showed obvious positive staining in the fovea of gastric mucosa and the bottom of glands in Rosa26^Hnf4α mice, while weak staining was found only in the pits in WT mice (Fig. 1e). Although no obvious IM cells were observed in the gastric mucosa of mice, these changes may still indicate that Hnf4α might affect the development of gastric mucosa and promote the secretion of intestinal mucus.

Initially, we treated GES-1 cells with DCA and extracted RNA for RNA-seq. Figure 2a shows the heat map of differentially expressed molecules, and HDAC6 is one of the significantly increased molecules. KEGG data showed that DCA is closely related to tumor and signal transduction (Fig. 2b). Next, GES-1 cells were treated with a gradient concentration of DCA, and the results showed that the mRNA of HDAC6 and HNF4α increased obviously (Figs. 2c, S2A). We also detected that DCA enhanced the HDAC6 protein along with intestinal markers CDX2, KLF4, MUC2 and HNF4α in GES-1 cells and AZ521 cells (Figs. 2d, S2B). The results of IF further confirmed that DCA caused the enhancement of HDAC6 and HNF4α in GES-1 cells and their expression in HCT-116 cells was used as a positive control (Figure S2C). More importantly, similar results were found in mice primary gastric mucosa cells (Figs. 2e, f). Together, these results suggest that HDAC6 increased significantly in BA-induced IM cells.

To investigate the relevance of HDAC6 and HNF4α with gastric IM clinically, we examined the expression of both in normal, gastritis and IM tissues (Fig. 3a, b). Both HDAC6 and HNF4α increased in IM tissues, and the number of samples showing high expression was greater than that showing low or moderate expression in gastric IM tissues (Fig. 3c–e). Additionally, negative expression of HDAC6 and HNF4α were mainly observed in normal tissues, while positive expression was mostly observed in gastric IM tissues (HDAC6: 85/119, 71.4%, P = 0.0302; HNF4α: 102/119, 85.7%, P = 0.0001). The correlation analysis showed that the expression levels of HDAC6 in IM tissues were positively associated with HNF4α (Fig. 3f and Table 3). Consistently, the staining results of tissue microarrays indicated that most of the HDAC6-positive tissues were also positive for HNF4α. Overlapping expression of the two molecules was seen in 80.67% (96/119) of IM tissues, which further suggested a close correlation between HDAC6 and HNF4α (Fig. 3g and Tables 4, 5). Moreover, as shown in Fig. 3h, the mRNA level of both is higher in IM tissues than in normal tissues. In summary, these findings imply a potential prometaplastic role of HDAC6 and HNF4α in human gastric tissues.

To further study the role of HDAC6 in IM, we performed gain-of-function and loss-of-function experiments. First, we infected GES-1 cells with an HDAC6 overexpression vector, and the results showed that the upregulation of HDAC6 caused an increase of CDX2, KLF4 and MUC2 protein levels (Fig. 4a).Subsequently, AGS cells, which highly express HDAC6 and HNF4α, were transfected with HDAC6-specific siRNA and the results revealed that the downregulation of HDAC6 inhibited the expression of downstream intestinal markers (Fig. 4b). Further, we investigated the regulatory effect of HNF4α on HDAC6 and found that HNF4α overexpression caused a significant increase in HDAC6 in GES-1 cells (Fig. 4c), while siHNF4α resulted in a decrease in HDAC6 in AGS cells, which indicated that HNF4α could positively regulate the expression of HDAC6 in gastric cells (Fig. 4d). Interestingly, HDAC6 was found to also have a positive regulatory effect on HNF4α, which may lead to the formation of a positive feedback regulatory loop. The loop is still effective in the presence of BA as shown in Fig. 4e because both siHDAC6 and siHNF4α could reverse the upregulation of each other and intestinal markers triggered by DCA. Next, to reveal the regulatory mechanism of HNF4α on HDAC6, the JASPAR (https://​www.​https://​jaspar.​binf.​ku.​dk/​) and PROMO (https://​alggen.​lsi.​upc.​es/​cgi-bin/​promo_​v3/​promo/​promoinit.​cgi?​dirDB=​TF_​8.​3) databases were used to predict whether HNF4α had a transcription binding site in the HDAC6 promoter region. Then we constructed the reporter genes for truncated bodies containing different HDAC6 binding sites according to the predicted results of JASPAR and cotransfected them with the HNF4α expression vector in GES cells (Table 1). The data showed that the activity of HDAC6-p4 plasmids increased most obviously, suggesting that the sequence (− 1432 to − 1332 bp) might be the active fragment of HNF4α on the HDAC6 promoter (Fig. 4f). The ChIP assays further verified that HNF4α could combine with the HDAC6 promoter (AGGATCAGAGGGCAA), and this combination was strengthened under DCA (Fig. 4j, h). Additionally, Hdac6 was found to be enhanced by Hnf4α in Rosa26^Hnf4α mice and both were mainly located at the bottom of the gland in the gastric antrum (Fig. 4i). In summary, these results indicated that HDAC6 positively regulated intestinal markers and HNF4α and showed that HNF4α could transcriptionally activate HDAC6 in gastric cells.

All the above results indicated that DCA promotes the development of IM through a closed HDAC6/HNF4α loop, although the role of BA in this signaling pathway remains unclear. In a previous study, we sequenced microRNA in GES-1 cells after BA treatment [26]. Combined with a bioinformatics analysis, we found that among several microRNAs that were significantly reduced, miR-1 might be closely related to this loop (Fig. 5a). We speculated that miR-1 might target HDAC6 and HNF4α posttranslationally in IM cells. As expected, the treatment with DCA led to an obvious decrease in miR-1 in GES-1, MKN45 and AZ521 cells (Fig. 5b), and similar results were observed in primary cells post-DCA (Fig. 5c). Furthermore, we found that the level of miR-1 in gastric IM tissues was markedly lower than that in peripheral normal tissues (Fig. 5d). Additionally, microarray in situ hybridization showed that the level of miR-1 in IM was obviously lower than that in normal tissues and the location of miR-1 was in the cytoplasm and nucleus (Fig. 5e). Moreover, the staining indicated that the expression level of miR-1 in IM was remarkably lower than that of HDAC6 or HNF4α and miR-1 was negatively correlated with HDAC6 and HNF4α (Figure S3A, S3B). Collectively, these data suggest that miR-1 decreased in IM tissues and BA-induced IM cells.

To confirm that miR-1 could target HDAC6 and HNF4α, we treated GES-1 cells with a concentration gradient of anti-miR-1. We found that 200 nM anti-miR-1 remarkably reduced the level of miR-1 (Fig. 6a). Then, we used 200 nM anti-miR-1 to transfect GES-1 and BGC-823 cells, and found a robust enhancement of HDAC6 and HNF4α levels accompanied by an increase in intestinal markers (Fig. 6b). Moreover, we transfected AGS and BGC-823 cells with ago-miR-1 and observed that 100 nM ago-miR-1 caused an obvious increase in miR-1 (Fig. 6c). Additionally, HDAC6 and HNF4α decreased substantially with the reduction of CDX2, KLF4 and MUC2 after transfection (Fig. 6d), which is similar to the results from HDAC6 or HNF4α downregulation. These results suggest that the expression of miR-1 could inversely affect the expression levels of HDAC6 and HNF4α. To show that the regulation of the two molecules by miR-1 is achieved by acting on their 3′-UTRs, we carried out luciferase reporter assays (Fig. 6e, 6f). In the GES-1 and BGC-823 cells, the activity of the HDAC6 wt3′-UTR was decreased by miR-1 restoration, while the HDAC6 mut3′-UTR with a mutated binding site sequence was not affected by miR-1. Similarly, the activity of the HNF4α wt3′-UTR was also significantly inhibited by miR-1 (Fig. 6g, h). A quantitative protein analysis showed that HDAC6 levels increased in cells transfected with the WT or mutant HDAC6 plasmids, but were attenuated in cells cotransfected with the WT HDAC6 and miR-1 (Fig. 6i). Similarly, cotransfection of both HNF4α and miR-1 did not cause an increase in HNF4α (Fig. 6j). In addition, miR-1 attenuated the activation of HDAC6 and HNF4α by DCA (Figure S3C), which indicated that miR-1 is crucial in this process. Our findings demonstrate that the ectopic expression of HDAC6 and HNF4α induced by BA occurs mainly via miR-1 silencing.