Metformin, 5-aminoimidazole-4-carboxamide ribonucleoside (AICAR), dorsomorphin (Compound C) were purchased from Sigma-Aldrich (St. Louis, MO). Rabbit polyclonal anti-NHE3 antibody EM450 and mouse monoclonal anti-VSVG P5D4 were previously described (Yoo et al., 2012). The following antibodies were commercially obtained: anti-NHE3 (NHE31-A, Alpha Diagnostics, San Antonio, TX)), Anti-phospho-NHE3 (#MABN2415; MilliporeSigma, Burlington, MA)), anti-HA (C29F4, Cell Signaling, Danvers, MA), anti-phospho-S342-Nedd4-2 (12,146, Cell Signaling), anti-phospho-S448-Nedd4-2 (8,063, Cell Signaling), anti-ubiquitin (P5D1, Santa Cruz, Dalla, TX), anti-Nedd4L (46,521, Abcam, Waltham, MA), and anti-β-actin (4,967, Cell Signaling) antibodies.

Caco-2bbe and SK-CO15 cells were cultured as previously described (Yoo et al., 2012). pcDNA3.1 construct carrying human NHE3 (pcDNA-hNHE3V) with a C-terminal vesicular stomatitis virus glycoprotein (VSVG) epitope has been described (No et al., 2014). Caco-2bbe cells transfected with pcDNA-hNHE3V, Caco-2bbe/NHE3, were previously generated (Lin et al., 2010). The lentiviral plasmid pLKO.1 containing short hairpin RNA (shRNA) targeting Nedd4-2 (shNedd4-2) or a scrambled control shRNA (shCon) was purchased from Sigma.

Five-six weeks old male Lepr ^db/db (db/db) mice were obtained from the Jackson laboratory (Bar Harbor, ME). Mice were acclimated and given standard chow and water ad libitum for 2 weeks before the start of experiments. Mice were divided into two groups: the experimental group receiving metformin in their drinking water (1.0 mg/ml) for 30 days, and the control group receiving normal tap water. On day 25, 27 and 29, metformin at the concentration of 500 mg/kg was administered once a day by gavage using a 4 cm-long curved needle with a steel ball at the tip. Blood samples were collected, and glucose levels were measured with a OneTouch Verio glucose meter (LifeScan, Malvern, PA)) every 4–5 days. At the same time, body weights were measured. To determine fecal water contents, feces were collected for 1 h at the end of the experimental period, and total fecal weight was measured. Fecal weight was measured again after overnight dehydration, and fecal water content was determined by (fecal weight before dehydration - fecal weight after dehydration)/fecal weight before dehydration X 100.

Intestinal water flux was measured as previously reported (Lin et al., 2010). Briefly, a mouse anaesthetized with sodium pentobarbital was placed on a 37°C heating block, and a small incision was made in the abdomen. A 5-cm loop of the ileum (between 5 and 10 cm upstream of cecum) cannulated at the proximal and distal ends was flushed with saline for 10 min. This was followed by perfusion of pre-warmed perfusion solution (mM: 118.4 NaCl, 4.7 KCl, 2.52 CaCl₂, 1.18 MgSO₄, 25 Na gluconate, 1.18 KH₂PO₄, pH 7.4) at 1 ml/min for 2 h. The CFTR inhibitor CFTRinh-172 (250 ug/kg) was administered by intraperitoneal injection to each mouse 1 h prior to the start of perfusion. Intestinal water flux was recorded by calculating a change in the volume of perfusion buffer in a reservoir every 10 min over the course of a 90–120 min perfusion period.

The Na⁺-dependent changes in intracellular pH (pH_i) by NHE3 was determined using the ratio-fluorometric, pH-sensitive dye 2′,7′-bis-(2-carboxyethyl)-5-carboxyfluorescein acetoxymethyl ester (BCECF-AM) as previously described (Yoo et al., 2012). Caco-2bbe cells grown for 10–14 days post-confluence were treated with 1 mM metformin or the AMPK activator AICAR (500 μM) 30 min during dye loading. For the measurement of NHE3 activity in mouse intestine, isolation of villi from mice ileum was performed as previously described (He et al., 2011). In brief, mice were euthanized with isoflurane and the ileum was flushed with cold PBS to remove food particles. An equivalent segment of the proximal ileum (approximately 10 cm upstream of the cecum) was opened longitudinally and stabilized on a cooled stage. The villi were dissected under stereomicroscope using sharpened microdissection scissors. Isolated villi were mounted on coverslips and covered with light and solution penetrable polycarbonate membrane (GE, Minnetonka, MN). Coverslips were mounted on a perfusion chamber, placed on an inverted microscope, superfused with NH₄ ⁺ buffer, followed by sequential perfusion with tetramethylammonium (130 mM TMA-Cl, 20 mM HEPES, 5 mM KCl, 1 mM TMA-PO₄, 2 mM CaCl₂, 1 mM MgSO₄, and 25 mM glucose) and Na⁺ buffer that drives Na⁺-dependent pH recovery. Na⁺ buffer was supplemented with 30 μM HOE694 or 2 μM dimethyl amiloride to inhibit NHE1 and NHE2 activities. The microfluorometry was performed using the Metafluor software (Molecular Devices, Sunnyvale, CA) and two to three traces of Na⁺-dependent pH recovery were captured from each coverslip, each trace originating from an independent group of cells. For each experiment, a minimum of four coverslips per group were studied. Na⁺/H⁺ exchange rate was described by the initial rate of pH_i recovery, which was calculated by determining slopes along the pH_i recovery by linear least-squares analysis over a minimum of 7 s.

Quantitative RT-PCR. Total RNA was isolated from ileal mucosa scrapes using RNA Extraction Kit (Qiagen, Germantown, MD), and cDNA was synthesized using SuperScript III First-Strand Synthesis Kit (Life Technologies, Carlsbad, CA). Quantitative RT-PCR was performed as previously described using iQ SYGR Green Supermix (Bio-Rad, Hercules, CA) on the Eppendorf Mastercycler realplex (Eppendorf, Enfield, CT). The primers are as follows:

NHE1: 5′-TCT​TCA​CCG​TCT​TTG​TGC​AG-3′ and 5′-TAA​ACC​GGT​TGA​GCT​TGT​CC-3’. NHE2: 5′-CAG​TGT​CAG​CGA​AAC​CTT​GA-3′ and 5′-GGC​AAT​GAA​TGA​TGA​ACT​GGT​CCT-3’. NHE3; 5′-GAG​CTG​AAC​CTG​AAG​GAT​GC-3′ and 5′-CTC​CAG​AGA​CTG​CAT​GTC​CA-3’: ß-actin; 5′-AGC​CAT​GTA​CGT​AGC​CAT​CC-3′ and 5′-TCT​CAG​CTG​TGG​TGG​TGA​AG-3’.

Caco-2bbe/NHE3 cells were lysed in cold lysis buffer (Cell Signaling) containing 150 mM NaCl, 1 mM ß-glycerophosphate, 2.5 mM sodium pyrophosphate, 1 mM Na₂EDTA, 1 mM EGTA, 1 mM Na₃VO₄, 1 μg/ml leupeptin, 1% Triton X-100, and protease inhibitors cocktail tablets (Roche, Indianapolis, IN). Lysate (500 μg) was pre-cleared by incubation with 30 µl of protein A-Sepharose beads for 1 h and the supernatant was then incubated overnight with anti-VSVG P5D4 antibodies. Immunocomplex was purified by incubating with 50 µl of protein A-Sepharose beads for 1 h, followed by three washes in lysis buffer and two washes in PBS. All the above steps were performed at 4 C or on ice. Beads were eluted with 2.5X Laemmli buffer. Cell lysates and eluted samples were separated by SDS-PAGE. Western immunoblotting was performed as previously described (No et al., 2014) and visualization were performed using the LICOR Odyssey Imaging System (Li-COR corp, Lincoln, NE). Densitometric analysis was performed using ImageJ software (National Institutes of Health).

Caco-2bbe/NHE3 cells were transiently transfected with pMT123 to express HA-Ub (Han and Yun, 2020). Cells were lysed in cold lysis buffer supplemented with 10 µM MG132 to inhibit proteasomal degradation. Equal amounts of cell lysates (typically 300 μg) were incubated overnight with anti-NHE3 EM450 antibody (Yoo et al., 2012), followed by incubation with protein A-Sepharose beads for 1 h. Immunocomplex was washed twice in lysis buffer and once in PBS. NHE3 was eluted, resolved by SDS-PAGE, and immunoblotted with anti-HA antibody.

Surface biotinylation of NHE3 was performed as previously described (No et al., 2014). Briefly, cells were rinsed twice in PBS and incubated in borate buffer (154 mM NaCl, 7.2 mM KCl, 1.8 mM CaCl₂, and 10 mM H₃BO₃, pH 9.0) for 10 min. Cells were then incubated for 40 min with 0.5 mg/ml sulfo-NHS-LC-biotin (Pierce, Rockford, IL) in borate buffer. Unbound sulfo-NHS-LC-biotin was quenched with Tris buffer (20 mM Tris, 120 mM NaCl, pH 7.4). Cells were lysed in lysis buffer (150 mM NaCl, 1 mM ß-glycerophosphate, 2.5 mM sodium pyrophosphate, 1 mM Na₂EDTA, 1 mM EGTA, 1 mM Na₃VO₄, 1 μg/ml leupeptin, 1% Triton X-100, and protease inhibitors cocktail tablets) and sonicated for 2 x 15 s. Lysate was agitated for 30 min and spun at 14,000 x g for 30 min at 4°C to remove the insoluble cell debris. An aliquot was retained as the total fraction representing the total cellular NHE3. Protein concentration was determined and 1 mg of lysate was then incubated with streptavidin-agarose beads (Pierce) for 2 h. The streptavidin-agarose beads were washed 3 times in lysis buffer and twice in PBS. All the above procedures were performed at 4°C or on ice. Biotinylated surface proteins were then eluted by boiling the beads at 95°C for 10 min. Dilutions of the total and surface NHE3 were resolved by SDS-PAGE and immunoblotted with an anti-VSVG antibody as described above.

Caco-2bbe/NHE3 cells grown on Transwells (Corning, Lowell, MA) were fixed using 100% methanol at −20°C for 20 min. Cells were washed three times with PBS, permeabilized using 0.05% Triton X-100 in PBS, and blocked with 5% normal goat serum for 1 h at RT. Cells were then stained with mouse anti-VSVG P5D4 antibody for 16 h at 4°C. Cells were washed with PBS three times and incubated with Alexa 488-conjugated goat anti-mouse IgG (Invitrogen) and Alexa Fluor 647 phalloidin (ThermoFischer) for 1 h at RT. After three 10 min washes with PBS, specimens were mounted with ProLong Glass Antifade Reagent (Invitrogen) and observed under a Nikon A1R HD confocal microscope (Nikon Instruments Inc., NY) coupled to a Plan Apo λ 60x Oli lens.

Statistical analysis was performed using independent samples two-tailed unpaired Student’s t test or ANOVA, followed by the Tukey post hoc analysis. Results are presented as mean ± SEM. Statistical analysis was performed using GraphPad Prism software (La Jolla, CA). A value of p < 0.05 was considered significant.

To assess whether metformin causes diarrhea, we first determined the effect of metformin in db/db T2D mice (Robinson et al., 2012). We started the treatment of mice by providing metformin at the concentration of 1.0 mg/ml in drinking water for 4 weeks. We observed a slight change in the blood glucose levels that did not reach statistical significance compared with the untreated cohort. Takemori et al. (Takemori et al., 2020) recently published their metformin treatment scheme consisting of a stepwise escalation of metformin dosage which lowered plasma glucose levels in mice. Therefore, we incorporated a part of their treatment scheme by administrating high dose metformin (500 mg/kg/day) for three alternative days during the last week of our treatment scheme. This regiment of metformin administration resulted in a significant reduction in glucose levels although still at >400 mg/dL (Figure 1A). There was no significant difference in body weight between control and metformin-treated groups (Figure 1B). Although none of the mice developed watery diarrhea, the feces of metformin-treated mice appeared softer. Quantification of the water content of feces revealed that there was a significant increase in fecal water contents compared with control mice (Figure 1C).

We have previously demonstrated the use of in vivo perfusion through a small loop of intestine to determine NHE3-mediated absorptive process (Lin et al., 2010; He et al., 2015; Jenkin et al., 2018). Hence, to determine whether NHE3-mediated luminal water loss in the intestine, a proxy for diarrhea, is altered by metformin, we determine the rate of water flux in db/db mouse intestine. In addition to NHE3 regulating Na⁺ absorption, Cl⁻ secretion by CFTR is a major determinant of intestinal water loss (Field, 2003). Previous studies have shown that metformin inhibits protein kinase A (PKA)-mediated stimulation of Cl⁻ secretion by CFTR in the intestine (Rogers et al., 2013), suggesting that metformin-induced intestinal water loss is not mediated by CFTR. However, the effect of metformin on the basal CFTR activity is not clear and therefore mice were administered with the CFTR inhibitor CFTRinh-172 intraperitoneally 1 h prior to the start of perfusion to block CFTR-mediated secretion. The similar treatment was shown to block about 90% of CFTR activity (Ma et al., 2002; Thiagarajah et al., 2004). CFTRinh-172 is also shown to inhibit CLC-2 chloride channels that are expressed in IECs (Cuppoletti et al., 2014). Consistent with the increased fecal water content, metformin significantly lowered the rate of fluid absorption to 71.8 ± 5.4 μl/cm/h compared with 90.6 ± 2.9 μl/cm/h for the control group (Figure 1D). To determine whether NHE3 expression is altered by the prolonged treatment with metformin, mouse ileal mucosal layers were collected and NHE3 expression levels were compared with the expression levels of villin, which was used as a loading control for the epithelium. There was a trend that NHE3 expression relative to villin is decreased in metformin-treated mice, but it did not reach statistical significance (Figure 1E). To verify that metformin regulates NHE3, we determined the Na⁺/H⁺ exchange activity of NHE3 in isolated ileal villi as the rate of initial Na⁺-dependent intracellular pH (pH_i) recovery (Lin et al., 2010; He et al., 2015). All NHE3 activity measurements were conducted in the presence of 30 µM Hoe-694, which inhibits endogenous NHE1 and NHE2 activities. NHE3 activity was reduced by about 39% in metformin-treated mouse intestine (Figure 1F). These data indicated that metformin inhibits NHE3 and NHE3-dependent water absorption in the intestine, but the regulation of NHE3 is most likely at the post-translational level.

To confirm the lack of effect on NHE3 expression by metformin, we determined the effect of metformin on NHE3 transcript and protein expression in SK-CO15 cells, which express NHE3 endogenously (Yoo et al., 2012). Cells were treated with 1 mM metformin and NHE3 mRNA levels was determined by quantitative RT-PCR (Figure 2A). For up to 24 h, we did not observe a significant change in NHE3 mRNA levels. We also found that NHE1 and NHE2 transcript levels in SK-CO15 cells were not changed by metformin although the roles of NHE1 and NHE2 in diarrheal disease are unclear. Consistent with the findings in db/db mice, NHE3 protein expression was not significantly altered at 24 h of metformin (Figure 2B). We next determined whether the NHE3 activity is regulated by metformin. For this purpose, we utilized human colonic epithelial Caco-2bbe cells that are stably transfected to express human NHE3 with a VSVG epitope tag (Lin et al., 2010). The advantages of using Caco-2bbe cells are two-fold. First, Caco-2bbe cells lack endogenous NHE3 activity so that the regulation of ectopically expressed NHE3 is independent of transcriptional regulation. Second, the epitope-tagging of the exogenous NHE3 with a VSVG tag allows a better assessment of NHE3 protein by biochemical assays (Lin et al., 2010; Yoo et al., 2012). Na⁺/H⁺ exchange activity of NHE3 was determined in the presence of 30 µM Hoe-694 or 2 µM dimethyl amiloride to block endogenous NHE1 and NHE2 activities. Treating the cells with 1 mM metformin resulted in a significant decrease in NHE3 activity, which was augmented with 5 mM metformin (Figure 2C).

AMPK is the primary effector of metformin, but AMPK-independent action of metformin is also known (Foretz et al., 2010; Rena et al., 2017). We confirmed the activation of AMPK by metformin by determining phosphorylation of AMPK at Thr172 (Zhang et al., 2009), which serves as an indicator of AMPK activation (Figure 3A). To verify that NHE3 regulation by metformin is mediated via AMPK, cells were preincubated with the AMPK inhibitor dorsomophin, also known as Compound C, for 30 min prior to treatment with metformin. NHE3 inhibition by metformin was antagonized by dorsomorphin (Figure 3B), indicating that this regulation is mediated by AMPK. Consistent with these results, 5-aminoimidazole-4-carboxamide ribonucleoside (AICAR), which activates AMPK through accumulation of the AMP mimetic, AICAR monophosphate (Corton et al., 1995), induced inhibition of NHE3 (Figure 3C).

A number of studies have demonstrated that phosphorylation of rat NHE3 is important for inhibition of NHE3 activity (Kurashima et al., 1997; Zhao et al., 1999; Hu et al., 2001). Although the most previous studies have demonstrated phosphorylation of NHE3 at Ser552 and Ser605 by protein kinase A (PKA), recent studies have implied that NHE3 is a subject of phosphorylation at Ser552 by a mechanism independent of PKA (Liu and Jose, 2013; Chen et al., 2015). Therefore, we examined whether NHE3 is phosphorylated by metformin using phospho-specific NHE3 antibodies that recognize NHE3 only when phosphorylated at Ser552 (Kocinsky et al., 2005). We found that metformin increased NHE3 phosphorylation at Ser552 (Figure 3D). Because the anti-phospho-Ser605 antibody does not detect NHE3 in Caco-2bbe cells, we did not determine the status of phosphorylation at Ser605 (Jenkin et al., 2021).

We have recently shown that NHE3s from certain species including human and non-human primates are a subject of ubiquitination (No et al., 2014; Jenkin et al., 2021). We therefore assessed NHE3 ubiquitination in Caco-2bbe cells. To determine ubiquitination levels, Caco-2bbe/NHE3 cells were transiently transfected to express HA-Ub. Following metformin treatment, NHE3 was immunoprecipitated and Ub levels of NHE3 was determined. As shown in Figure 4A, metformin treatment for 30 or 60 min resulted in a significant increase in NHE3 ubiquitination.

Nedd4-2 is a major Ub ligase regulating membrane proteins, including ENaC and NHE3 (Snyder et al., 2002; No et al., 2014). To determine whether NHE3 ubiquitination by metformin is dependent on Nedd4-2, Caco-2bbe cells were transduced with lentiviral shRNA specific for Nedd4-2, shNedd4-2 (Figure 4B) (No et al., 2014; Jenkin et al., 2021). Comparison of NHE3 ubiquitination induced by a 30 min treatment of metformin showed a marked decrease NHE3 ubiquitination levels with Nedd4-2 knockdown (Figure 4C). Importantly, silencing Nedd4-2 expression blocked metformin-induced inhibition of NHE3 (Figure 4D), demonstrating the functional relevance of Nedd4-2 in this regulation.

Inhibition of NHE3 activity involves trafficking of NHE3 protein from the plasma membrane into an intracellular pool (Chow et al., 1999; Hu et al., 2001; Kocinsky et al., 2007; Hu et al., 2013; He et al., 2015). To determine whether metformin induces internalization of NHE3 protein, we determined NHE3 abundance at the apical membrane of Caco-2bbe cells by surface biotinylation before and after metformin treatment. As shown in Figure 5A, the abundance of NHE3 protein in the apical membrane was significantly decreased to 73.6 ± 7.4% of control levels without a change in the total amount of NHE3. In contrast, silencing of Nedd4-2 blocked the change in the apical NHE3 abundance (Figure 5B). This difference was corroborated by immunofluorescence (IF) confocal microscopy (Figure 5C) where Nedd4-2 knockdown mitigated endocytic trafficking of NHE3 by metformin.

Non-primate mammals, including mouse, express NHE3 that lacks a PY motif and hence NHE3 is not regulated via Nedd4-2-dependent ubiquitination (No et al., 2014). However, to determine whether the trafficking of NHE3 by acute treatment of metformin occurs in vivo, we treated WT and db/db mice by oral administration of metformin (500 mg/kg/day) for 3 days. In WT mice, NHE3 was primarily located at the terminal web of the microvilli region where IF signals of NHE3 and F-actin overlapped (Figure 5Di). Metformin resulted in a shift of NHE3 IF signals towards the intracellular region relative to F-actin (Figure 5Dii). In db/db mice, NHE3 IF signals were also detected at the brush border membrane. Unexpectedly, there was more prominent intracellular NHE3 signals in db/db mouse villi compared with WT villi under basal conditions (Figure 5Diii vs. i). Nevertheless, there was a shift of NHE3 fluorescent signals relative to actin by metformin in db/db mice (Figure 5Div), indicating that the endocytic trafficking of NHE3 from the brush border membrane occurs regardless of ubiquitination.

Given the Nedd4-2 dependent ubiquitination of NHE3, we next sought to determine whether AMPK activation promotes the interaction of Nedd4-2 with NHE3. To test this possibility, we expressed HA-Nedd4-2 in Caco-2bbe cells since anti-Nedd4-2 antibody is not suitable for immunoprecipitation based on our prior experience. The specificity of co-immunoprecipitation of HA-Nedd4-2 with NHE3 is demonstrated in Figure 6A where HA-Nedd4-2 co-immunoprecipitated with anti-NHE3 P5D4 antibody but not with IgG. To determine the effect of metformin on the Nedd4-2-NHE3 interaction, cells were treated with metformin for 30 min. Under basal conditions, HA-Nedd4-2 co-immunoprecipitated with NHE3(Figure 6B, C), but metformin did not result in a meaningful change in the amount of HA-Nedd4-2 co-immunoprecipitated with NHE3 (Figure 6B, M), indicating that metformin does not facilitate the interaction between Nedd4-2 and NHE3. Previous studies have demonstrated that a substrate interaction with Nedd4-2 is modulated via phosphorylation of Nedd4-2 at Ser342 or Ser448 (Debonneville et al., 2001; Ichimura et al., 2005; Zhou and Snyder, 2005). Therefore, we determined whether there is a metformin-dependent phosphorylation of Nedd4-2 at Ser342 and Ser448. We observed phosphorylation at Ser342 and Ser448 under basal conditions, but there was no change in phosphorylation levels in response to metformin (Figure 6C). These data indicate that the regulation of NHE3 by metformin is not dependent on a change in Nedd4-2 phosphorylation at Ser342 or Ser448.