Six to eight-week male C57BL/6 mice were obtained from the Model Animal Genetics Research Center of Nanjing University (Nanjing, China). Animal welfare and experimental procedures were carried out strictly in accordance with the recommendation of Guide for the Care and Use of Laboratory Animals (Ministry of Science and Technology of China, 2006), and the Nanjing University Animal Care and Use Committee (NJU-ACUC). The protocol was approved by the Nanjing University Animal Care and Use Committee (NJU-ACUC) and to minimize suffering and to reduce the number of mice used.

Dextran sulfate sodium was purchased from MP Biomedicals (Aurora, OH, United States). Compound C and AICAR were purchased from MedChemExpress (MCE, United States). Antibody for phosphorylated AMPK (Thr172) was purchased from Cell Signaling Technology (Beverly, MA, United States), antibodies for AMPKα1/2, ZO-1, occludin, claudin-1, Actin, Tubulin were purchased from Santa Cruz (Santa Cruz, CA, United States), IgG for rabbit and mouse were purchased from (Beyotime Biotechnology, China). Human recombinant interferon-γ (IFN-γ) was purchased from R&D Systems (Minneapolis, MN, United States). Lipopolysaccharide (LPS), metformin, mesalazine and other chemicals were purchased from Sigma-Aldrich (St. Louis, MO, United States). Alexa Fluor 488-conjugated anti-rabbit IgG was purchased from Life Technology (Carlsbad, CA, United States)

2.5% (wt/vol) DSS was utilized to induce acute colitis in mice. Mice administrated with DSS dissolved in drinking water continuously from day 0 to day 7. Normal mice received the same drinking water without DSS. Metformin, mesalazine and AICAR were dissolved in normal saline, Compound C was diluted in normal saline from stock solution, which was dissolved in dimethyl sulfoxide (DMSO). Metformin (125, 250, 500 mg/kg, p.o.), mesalazine (200 mg/kg, p.o.), AICAR (100 mg/kg, i.p.) and Compound C (10 mg/kg, i.p.) were administered once a day from day 0 to day 11 (n = 6–8 mice per group). Weight, morbidity and the presence of gross blood in feces and at the anus were observed daily. The disease activity index (DAI) was calculated by assigning well-established and validated scores as previously described (Wu et al., 2014). That is, the following parameters were used for calculation: (a) diarrhea (0 points = normal, 2 points = loose stools, 4 points = watery diarrhea); (b) hematochezia (0 points = no bleeding, 2 points = slight bleeding, 4 points = gross bleeding). At day 9 following induction with DSS, the animals were sacrificed, the entire colon was quickly removed for ex vivo study. Segments of the colon taken for histopathological essay were fixed in 10% normal buffered formalin, embedded in paraffin. Sections were stained with hematoxylin and eosin and histological score was evaluated (blinded) as follows: 0, no signs of inflammation; 1, low leukocyte infiltration; 2, moderate leukocyte infiltration; 3, high leukocyte infiltration, moderate fibrosis, high vascular density thickening of the colon wall, moderate goblet cell loss, and focal loss of crypts; and 4, transmural infiltrations, massive loss of goblet cell, extensive fibrosis, and diffuse loss of crypts.

Mice were sacrificed after anesthesia and a portion of the colons close to the rectum were immediately removed. Epithelial cells were obtained from freshly isolated colons and flushed from both ends with sterile phosphate buffer (PBS). Colons were then opened longitudinally and thoroughly washed in cold PBS. The colon was then incubated in 10 ml 30 mM ethylene diamine tetraacetic acid (EDTA) and 1.5 mM dithiothretol (DTT) on ice for 20 min and then removed and washed in cold PBS and incubated in 10 ml PBS containing 30 mM EDTA at 37°C at 200 RPM for 10 min. The cells were shaking vigorously for 30 s and centrifuged at 1000 g for 5 min at 4°C, washed in PBS containing 10% FBS and spun for a further 5 min at 4°C at 1000 g. Cell pellets constituted isolated colonic epithelial cell fractions.

Samples were collected and lysed in a lysis buffer containing protease and phosphatase inhibitors (Pierce). The proteins were fractionated by SDS-PAGE and electrophoretically transferred onto polyvinylidene fluoride membranes. The membrane was blocked with 5% BSA for 2 h at room temperature. Different antibodies were incubated overnight at 4°C. After washing with PBST for 3 times, 10 min/time, membranes were incubated with secondary antibody for 2 h at room temperature. Finally, membranes were washed by PBST for 3 times, 10 min/time, and the software Quantity One (Bio-Rad Laboratories, Hercules, CA, United States) was used for densitometric analysis.

Real-time PCR was performed as described previously (Schmittgen and Livak, 2008). Total RNA of colon tissue was reverse transcribed to cDNA and subjected to quantitative PCR, which was performed with the BioRad CFX96 TouchTM Real-Time PCR Detection System (BioRad, CA, United States) using iQTM SYBR^® Green Supermix (BioRad, CA, United States), and threshold cycle numbers were obtained using BioRad CFX Manager software. The program for amplification was 1 cycle of 95°C for 2 min followed by 40 cycles of 95°C for 10 s, 60°C for 30 s, and 95°C for 10 s. The relative gene expression was the comparative C_T method. The primer sequences used in this study were as follows:

For paraffin-embedded colonic tissue, the sections (4–5 μm) were deparaffinized, rehydrated and washed in 1% PBS-Tween 20 (PBST). Then they were blocked with 3% bovine serum albumin (BSA) and incubated for 2 h at room temperature, and incubated with anti-phosphorylated AMPK antibody (1: 100) and IgG (1:100) overnight at 4 C. After washing with PBST for 3 times, 10 min/time, slides were incubated with fluorescein isothiocyanate (FITC)-conjugated anti-rabbit IgG (1:200, at room temperature for 2 h and then stained with 1 μg/mL DAPI (4′,6-diamidino-2-phenylin- dole) for 1 min. Likewise, for Caco2 cells, after treated with 2.5% DSS for 48 h with or without metformin (30 μM), cells were fixed with ice-cold methanol and were permeabilized with 0.3% Triton X-100/PBST. After blocking with 3% BSA for 2 h, the Caco2 cells were incubated with the ZO-1, claudin-1 and occludin antibody (1:100) overnight at 4°C. After washing with PBST for 3 times, 10 min/time, cells were exposed to the FITC-conjugated secondary antibody (1:200, at room temperature for 2 h). Finally, the coverslips were stained with DAPI for 1 min and washed with PBST for 3 times, 10 min/time. Confocal microscopy analyses were carried out using Olympus FV1000 confocal system (Olympus, Japan).

Immunohistochemical analysis was performed on paraffin-embedded colon tissue sections (4–5 μm) to evaluate the expression of ZO-1, claudin-1 and occludin. Slides were deparaffinized, rehydrated and blocked as described above, and incubated with ZO-1, claudin-1 and occludin antibody (1:100) overnight at 4°C. After washing with PBST for 3 time, 10 min/time, slides were incubated with streptavidin-HRP (Shanghai Gene Company, GK500705, Shanghai, China) for 40 min, then stained with DAB (Shanghai Gene Company, GK500705, Shanghai, China) substrate for 2–10 min and counter-stained with hematoxylin.

Caco2 cells were purchased from Shanghai Institute of Cell Biology (Shanghai, China) and maintained in DMEM medium, supplemented with 100 U/ml of penicillin, 100 μg/ml of streptomycin and 10% fetal calf serum under a humidified 5% (v/v) CO₂ atmosphere at 37°C.

Results were expressed as mean ± SEM. of three independent experiments and each experiment included triplicate sets. Data were statistically evaluated by one-way ANOVA followed by Dunnett’s test between control group and multiple dose groups. The level of significance was set at a P-value of 0.05.

To investigate the potential role of AMPK during the progression of colitis, the activation of AMPK, which is characterized as phosphorylation on Thr172 in α subunit, was evaluated during the development of colitis induced by DSS. As shown in Figure 1A, mice were administrated with 2.5% DSS for 7 consecutive days, and then substituted with water for another 4 days. Colon length was substantially reduced during the progression of colitis, indicating the worsened condition of inflammation (Figure 1B). Meanwhile, epithelium was isolated and collected from colons in mice after DSS administration on Day 0, 5, 7 and water administration on Day 9, 11, respectively. It was shown that the activation of AMPK was gradually decreased during DSS administration and then slightly increased after drinking water (Figures 1C,D). Consistent with this result, the immunofluorescence of phosphorylated AMPK showed the same tendency (Figures 1E,F).

According to the results in Figure 1, we found an inverse correlation between AMPK phosphorylation level and colitis severity. Therefore, we wonder whether promoting AMPK activation could prevent the development of colitis. Since metformin has previously been demonstrated to active AMPK and is considered as a safe drug in clinical application, we use metformin to further study the role of AMPK in colitis. Mesalazine, a cyclooxygenase (COX) inhibitor, is widely used for IBD treatment (Frieri et al., 2017). In the present study, we used mesalazine (200 mg/kg) as a positive control. Mice were divided into 6 groups (n = 6–8 per group): normal, DSS alone, DSS with metformin (125, 250, and 500 mg/kg, p.o.) and DSS with mesalazine (200 mg/kg, p.o.). Colonic epithelial cells were isolated for western blot. As shown in Figures 2A,B, compared with DSS group, metformin (125, 250, and 500 mg/kg) exhibited significant effect on AMPK activation. Simultaneously, compared with normal group, mice administrated with 2.5% DSS showed typical symptoms of colitis, characterized as significant weight loss, shortened colon length, and elevated disease activity index (DAI) which reflected severity of colitis (including diarrhea, intestinal bleeding). In contrast, these responses in mice treated with metformin (125, 250, and 500 mg/kg) were attenuated markedly in a dose-dependent manner. In addition, mesalazine showed weaker effects than metformin (500 mg/kg) (Figures 2C–F).

To further confirm the inhibitory effect of metformin on colitis, histological changes and cytokine profiles were examined. As shown in Figures 3A,B, sections from colon tissue stained with hematoxylin and eosin (H&E) showed a loss of architecture and immune cell infiltration. However, the inflammatory responses were obviously reduced following metformin (125, 250, and 500 mg/kg) or mesalazine (200 mg/kg) treatment. The blinded histological score also revealed that oral administration of metformin was associated with a significant down-regulation in the severity of colitis. Furthermore, inflammation-associated factors such as interleukin-18 (IL-18), interleukin-1β (IL-1β), interleukin-6 (IL-6), Cyclooxygenase2 (COX2), intercellular cell adhesion molecule-1 (ICAM1), and inducible nitric oxide synthase (iNOS) were examined in mRNA level. As showed in Figure 3C, IL-18, IL-1β, IL-6, COX2, ICAM1, iNOS were upregulated in DSS-induced inflammation. By contrast, mRNA levels of these factors were significantly decreased in mice treated with metformin or mesalazine.

To further investigate the effect of metformin on tight junctions (i.e., ZO-1, claudin-1 and occluding proteins) during experimental colitis, colonic epithelium of mice from each group was isolated and the expression of tight junction proteins were examined by western blot and immunohistochemistry. As shown in Figures 4A,B, there was a dramatic loss of tight junction proteins staining in mice treated with DSS. Interestingly, concomitant administration of metformin (125, 250, and 500 mg/kg) significantly prevented the loss of tight junction proteins. The immunohistochemistrical results also showed that substantial decrease of these three tight junction proteins was regained in mice treated with metformin or mesalazine (Figures 4C–F and Supplementary Figure S1).

Since metformin has been reported to have some other targets (Kim et al., 2016), we further used Compound C, a specific inhibitor of AMPK and another AMPK activator, AICAR, to examine whether metformin functioned via AMPK activation. Mice were divided into 5 groups (n = 6–8 per group): normal, DSS with vehicle, DSS with metformin (500 mg/kg, p.o.), DSS with metformin (500 mg/kg, p.o.) and Compound C (10 mg/kg, i.p.), DSS with AICAR (100 mg/kg, i.p.). Intestinal epithelial cells were isolated from colon tissue for western blot tests (Figures 5A,B), we found that both metformin and AICAR activated AMPK significantly, while Compound C could reverse AMPK activation induced by metformin. Similarly, AICAR could also attenuate DSS-induced colitis and promoted the expression of tight junctions. However, the therapeutic effect of metformin on experimental colitis and damaged integrity of intestinal epithelium induced by DSS was almost reversed by Compound C. These data suggested that regulation of tight junctions by metformin occurs in an AMPK-dependent manner (Figures 5C–H).

To further study the mechanism underlying the effect of metformin on tight junction proteins, we used Caco2 cells which could form tight junctions when grew into monolayer. To mimic the circumstance of epithelial cells during colitis, 2.5% DSS, IFN-γ (1000 U) or LPS (500 ng/ml) were applied to mimic epithelial cell damage in colitis in vitro respectively (Araki et al., 2006; Cao et al., 2013; Satyam et al., 2017). As shown in Figures 6A–I and Supplementary Figure S2, cells were incubated with 2.5% DSS for 48 h, and with IFN-γ (1000 U) or LPS (500 ng/ml) for 72 h respectively, a significant down-regulation of p-AMPK with decreased expression of tight junction proteins (i.e., ZO-1, claudin-1 and occludin) was observed by western blot. However, protein level of p-AMPK, tight junction proteins was markedly increased in the presence of metformin (10, 30, and 100 μM).

Next, we used Compound C and AICAR to test whether the effect of metformin was dependent on AMPK. As shown in Figures 7A–C, AICAR (0.3, 1 mM) significantly up-regulated the decreased expression of tight junction proteins induced by 2.5% DSS in Caco2 cells, which exhibited a similar effect of metformin. Nevertheless, AMPK phosphorylation caused by metformin was markedly inhibited by compound C (1 μM) in Caco2 cells, with a consequent reverse on the increased expression of ZO-1, claudin-1 and occludin mediated by metformin (Figures 7D–F).

It is shown in previous studies that AMPK facilitates the assembly of tight junctions in epithelial cells (Calder et al., 2017; Prasad et al., 2017; Rowart et al., 2017; Wang et al., 2017). Here we observed the distribution of ZO-1, claudin-1 and occludin in the presence or absence of metformin (30 μM) in Caco2 cells after 2.5% DSS treatment for 48 h with immunofluorescence staining and confocal microscopy. As shown in Figures 8A–D, DSS treatment induced irregular morphology and distribution of ZO-1, claudin-1 and occluding. However, the membrane-distribution and assembly of these tight junction proteins were significantly enhanced by metformin. These data demonstrated that AMPK activation by metformin maintained the structure of tight junctions in colonic epithelial cells and increased their resistance to DSS damage.