Abnormal eating behavior is a risk factor of metformin-induced diarrhea in mice

In our previous studies, metformin-induced diarrhea was observed only in obese diabetic mice (db/db) when used as model subjects, and the metformin dose was set at 500 mg/kg/bid (1000 mg/kg/day) [8]. In addition, the administration of a single doses ≥ 900 mg/kg/day has been reported to cause moribundity, mortality, and clinical signs of toxicity in rats [23]. Therefore, in this study, we utilized healthy C57BL/6 J mice, administered metformin at a dose of 500 mg/kg/bid, and concentrated on feeding patterns. To reproduce the eating habits of obese peoples, C57BL/6 J mice (healthy mice) were initially fed a high-fat/high-sucrose diet for three weeks (Fig. S1A). The mice which were fed a high-fat, high-sucrose diet consumed slightly less food per day, but their body weight gain tended to increase (Fig. S1B and C).

Metformin treatment (500 mg/kg/bid) did not induce diarrhea, and the fecal moisture was slightly reduced in the mice of the high-fat/high-sucrose group (Fig. S1D and E). Although we expected that reduced food intake was due to increased levels of total serum GLP-1, the levels were reduced and not recovered by the metformin treatment (Fig. S1F). These results suggest that high-fat/high-sucrose diet feeding cannot not be replaced a model for metformin-induced diarrhea as observed in db/db. Interestingly, the atrophy of the colon was observed in mice feeding the high-fat/high-sucrose diet (Fig. S1D and G).

Next, instead of a high-fat/high-sucrose diet, the mice were fed a low-calorie diet containing 10% cellulose. In preliminary tests, (a) and (b) in Fig. 1A and B, although no significant differences in gut appearance were observed, some mice in group (b) excreted feces that were difficult to handle with forceps. Therefore, we examined the differences in feeding patterns, specifically focusing on whether the mice ingested food all at once by setting a fasting period for 12 h in night (c) and (d). Although the total food intake during the night (12 h) and in the morning (3 h) was lower in the fasting groups (c and d) compared to the no-restriction group (b), the fasting groups showed a higher intake during the refeeding period in the morning (3 h) (Fig. 1C). The administration of metformin to mice fed a 10% cellulose diet and subjected to fasting resulted in high fecal moisture (Fig. 1D). When mice were not subjected to fasting, fecal moisture did not change (around 55%), regardless of whether they were fed a normal diet or a 10% cellulose diet.

Fig. 1

Overeating increases fecal water content after metformin treatment A Six-week-old male C57BL/6 J mice (n = 8 each) were fed a normal diet for 1 week (gray arrow). Subsequently (from seven-week-old), one group (a) continued on the normal diet, while the other groups (b–d) were switched to a 10% cellulose diet (blue arrow) for three weeks. On the night prior to the final day, metformin (Met: 500 mg/kg) was orally administered, except for group (c), which did not receive the administration. Groups (c and d) on the 10% cellulose diet underwent a 12-h fasting period (gray bar). Finally, all mice received metformin administration (Met: 500 mg/kg) and were fed without restriction. Images of the mice gut (B) are shown. C. Food consumption of the final 15 h (12 h during the night: black) and 3 h in the morning was monitored. Means and S.D. are shown. D. Water content in mice feces is shown. ** p < 0.01

To determine whether diarrhea-like feces were caused by the amount of diet consumed by mice following the second administration of metformin, a restricted diet (0.5 g/mouse: refer from Fig. 1C) was provided (Fig. S2A). The dietary restriction improved fecal firmness and reduced water content, which was accompanied by a decrease in fecal size (Fig. S2B and S2C).

Overeating is a risk factor for metformin-induced diarrhea in mice

To examine whether the 10% cellulose-containing diet increased appetite, which could potentially lead to overeating even in healthy mice, the number of mice was increased, and food intake was monitored. A more than 20% increase (Fig. 2A), followed by enhanced weight gain (Fig. 2B), might be attributed to lowered GLP-1 levels (Fig. 2C), although no significant differences in glucose tolerance were observed (Fig. 2D). These results suggest that overeating is a risk factor for metformin-induced diarrhea in mice, which may be reproduced even in mice that have not developed diabetes or severe obesity. Clear differences in colon appearance and fecal moisture were observed when mice that had been fed the 10% cellulose diet were treated with metformin (Fig. 2E–H). Moreover, the mice adapted to a 10% cellulose diet consumed approximately 1.5 times more food after fasting compared to the normal diet group (Fig. S3), suggesting that overeating may be a trigger of metformin-induced diarrhea.

Fig. 2

10% cellulose diet and overeating increase fecal water content after metformin treatment A Six-week-old male C57BL/6 J mice [Each group consisted of 4 cages (n = 4 per cage, total 16 mice)] were fed a normal diet for 1 week. Subsequently, one group continued the normal diet, while the other groups were switched to a 10% cellulose diet for three weeks. Daily intake was monitored every 3 or 4 days from day 7 to 21. B. Bodyweight change monitored. n = 16. * p < 0.05. (p = 0.052 at day 21.) At day 21, half of mice (n = 8) were subjected to blood collection for total GLP-1 measurement by ELISA (C) followed by a glucose tolerance test performed by orally administering 1 g of glucose (D). The remaining mouse groups (n = 8 each) were subjected to metformin assays. E. The experimental scheme is shown. Metformin (Met: 500 mg/kg/bid, twice a day) was orally administered before and after fasting (12 h during the night). (-) means without metformin (only water). Images of mice colon (F) and feces (G) after metformin treatment are shown. H. Water content in mice feces is shown. Means and S.D. are shown. N.S., not significant. ** p < 0.01

To conclude that overeating is a risk factor in metformin-induced diarrhea, the mice were provided with a larger food intake after the second administration of metformin compared to the conditions shown in Fig. S2. The supplementation of even 1.5 g of diet was insufficient to induce diarrhea caused by metformin (Fig. 3A and B), which was also evaluated based on fecal moisture content (Fig. 3C). Considering that the mice did not consume all the supplied diet (Fig. 3D), at least less than 1.4 g of diet, which is almost the same amount as that consumed by mice fed with a normal diet (Fig. S3), may be insufficient to induce diarrhea.

Fig. 3

Dietary restriction prevents metformin-induced diarrhea. A. Seven-week-old male mice (n = 4/group) were fed a 10% cellulose diet for three weeks and treated with metformin (Met: 500 mg/kg) administered orally (P.O.). One group was allowed ad libitum feeding, while another group was subjected to 12-h nighttime fasting. The fasting groups were further divided into three categories: no dietary restriction or a specified amount of diet per mouse was provided. B Images of the colon and feces after 3-h feeding post the 2nd metformin treatment. C Fecal moister were measured. Means and S.D. are shown. ** p < 0.01. D. Amount of food actually consumed, excluding spills, during the 3-h feeding post the 2nd metformin treatment

Next, we examined the involvement of indigestible cellulose and the volume of daily food intake in metformin-induced diarrhea. Since mice consume approximately 4 g/day of the 10% cellulose diet (Fig. 2A), the daily supply of the 10% cellulose diet was limited to 3.5 g/day during the final week for one group. After a 12-h fasting period with metformin administration, the mice were given ad libitum access to either a normal diet or the 10% cellulose diet (Fig. S4A). Mice that had been fed the 10% cellulose diet for 3 weeks under ad libitum conditions exhibited diarrhea-like feces (Fig. S4B) and increased fecal moisture content (Fig. S4C), regardless of the diet type provided during the 3-h refeeding period following the 12-h fast with metformin. Under restricted feeding conditions during the final week, the mice consumed approximately 1.6 g of diet during the 3-h refeeding period (Fig. S4D). These findings suggest that the level of dietary intake, possibly reflecting habitual feeding behavior, may also contribute to metformin-induced diarrhea. However, the absolute amount of food consumed over a short period appears to be the more critical factor.

Wood creosote alleviates metformin-induced diarrhea independent of the expression of endocrine markers

Science wood creosote is a candidate for a therapeutic material to treat metformin-induced diarrhea in db/db mice, which is also tested in the present overeating model (Fig. 4A). Fecal moisture in mice treated with metformin was significantly recovered by the co-administration of wood creosote (Fig. 4B). Suppressed blood GLP-1 levels increased two-fold with the administration of metformin, but the levels did not vary with wood creosote (Fig. 4C). These results suggested that the increase in GLP-1 secretion in the gut might be a cause of metformin-induced diarrhea, but the mechanism by which wood creosote ameliorate the diarrhea symptom might be differ from the overeating model.

Fig. 4

Wood creosote ameliorates diarrhea induced by metformin A Seven-week-old male mice that had been fed normal or 10% cellulose diets for three weeks were treated with or without metformin (Met: 500 mg/kg/bid) and wood creosote (5 mg/kg). (-) means without metformin (only water). An image of the colon is shown. Facial moisture (B) and blood GLP-1 levels (C) were measured. Gray bars indicate the mouse group fed a normal diet and subjected to 12 h fasting [without metformin, (-) in A]. Sky blue bars indicate the mouse group fed a 10% cellulose diet and subjected to 12 h fasting [without metformin, (-) in A]. Dark blue bars indicate the mouse group fed a 10% cellulose diet and subjected to metformin treatment before and after 12 h fasting [Met in A]. Yellow bars indicate the mouse group that was fed a 10% cellulose diet, treated with metformin before and after 12 h of fasting, and administered wood creosote alongside the second metformin treatment [Met + W. Creosote in A]. n = 8. Means and S.D. are shown. * and ** indicate p < 0.05 and p < 0.01, respectively. D. mRNA expression was examined by quantitative PCR analyses. The expression levels are shown as relative levels normalized by Gapdh mRNA level. n = 6. E. Total bile acid in the ileum was measured using the bile acid quantification kit

The markers associated with metformin-induced diarrhea in a mouse model of obesity and type 2 diabetes (db/db) [7] were also tested in the present overeating model. The mRNA expression of the chloride channel Cftr in the ileum was increased by metformin (Fig. 4D), suggesting that, as in db/db, an increase in luminal osmolarity is also a cause of diarrhea in the overeating model. Although both Cftr and the water channel aquaporin 3 (Aqp3) has been reported induced by cAMP/CREB signaling, the Aqp3 mRNA levels was decreased by metformin. In contrast, the Aqp8 mRNA level was increased by metformin. The bile acid accumulation in the ileum (Fig. 4E) can be explained by suppression of mRNA expression of two genes, Asbt and Oatpb, involved in bile acid reuptake.

The treatment with wood creosote did not affect the mRNA expression patterns modified by metformin. These results lead to a speculation that the recovery or amplification rate in GLP-1 expression may be a primary cause of metformin-induced diarrhea and that the combination of dysregulation in several factors involved in osmolality and water balance may impair water absorption in the gut. However, wood creosote may improve water absorption in the colon without affecting the expression of metformin-associated genes, which was also observed in db/db mice.

Loperamide decreases the amount of digestate entering the cecum

Loperamide, a μ-receptor agonist, suppresses acetylcholine secretion in the parasympathetic peripheral neurons, thereby decreasing peristalsis in the digestive tract and relieving diarrhea symptoms [24]. By comparing the actions of wood creosote and loperamide in relieving diarrhea symptoms, the commonalities and differences between the two agents were examined. Like wood creosote, loperamide ameliorates diarrhea symptoms in mice subjected to metformin-induced diarrhea in an overeating model (Fig. S5A). Water content in feces was reduced by loperamide treatment (Fig. S5B). Interestingly, the influx of digestate into the cecum was reduced in the loperamide group (Fig. S5C), while no difference in digestate influx was observed in the wood creosote group. These results suggest that both wood creosote and loperamide could ameliorate diarrhea symptoms induced by the combination of overeating and metformin, but the precise mechanisms by which these drugs reduce water content in feces may not be identical.

The GLP-1 receptor antagonist Exendin-3 (9–39) prevents diarrhea induced by metformin

We investigated the role of GLP-1 in metformin-induced diarrhea, as increased GLP-1 levels have been linked to diarrhea symptoms not only in this study but also in our previous study using db/db mice [8]. The mice were fed a 10% cellulose diet for three weeks and then intraperitoneally administered the GLP-1 receptor antagonist Exendin-3 (9–39) prior to the second metformin administration after 12 h fasting. While Exendin-3 (9–39) showed minimal or no effects on the condition of the colon or feces (including fecal moisture) in mice without metformin treatment, it effectively prevented diarrhea-like symptoms in mice treated with metformin (Fig. 5A and 5B). The cecum weight was also reduced by Exendin-3 (9–39) in mice treated with metformin (Fig. 5C). In contrast, the administration of the GLP-1 receptor agonist Exendin-4 was insufficient to induce diarrhea in mice without metformin, although there was a tendency for increased fecal moisture in mice treated with metformin.

Fig. 5

Exendin-3 (9–39) prevents diarrhea induced by metformin A Seven-week-old male mice (n = 4) fed either 10% cellulose diet for three weeks were treated with metformin (Met: 500 mg/kg) administered orally (P.O.). Following a 12-h fasting period, the mice received Exendin-3 (9-39) or Exendin-4 (1 μg each in PBS) via intraperitoneal administration (I.P.). Then, the mice were administered metformin (500 mg/kg). After 3 h of re-feeding, the colon was collected (upper), and feces were recovered (lower). (-) means without GLP-1 receptor agonist/antagonist (only PBS). Facial moisture (B) and cecum weight (C) were measured. Means and S.D. are presented, with * and ** indicating p < 0.05 and p < 0.01, respectively. D Total RNA was extracted from the ileum, and mRNA levels of Glucagon [Gcg (Glp-1)], Cftr, and Il-6 were quantified by qPCR, normalized to Gapdh mRNA levels

Diarrhea induced by metformin in db/db mice is associated with increased mRNA expression of Glucagon (Gcg, which encodes GLP-1) and the chloride channel Cftr [8]. The potent biological effects of Exendin-3 and Exendin-4 were evaluated based on changes in these mRNA levels (Fig. 5D). Notably, the reduction of diarrhea symptoms was accompanied by a decrease in Cftr mRNA expression in metformin treated mice. Since the pro-inflammatory cytokine IL-6 and its signaling have been reported to be associated with GLP-1 signaling and metformin signaling [21, 25, 26], the mRNA levels of Il-6 were also examined. Exendin-3 (9–39) and Exendin-4 modulated Il-6 mRNA levels up and down, respectively, in metformin-untreated mice. In contrast, neither Exendin-3 nor Exendin-4 affected Il-6 mRNA levels in metformin-treated mice, despite the enhanced expression induced by metformin. These results suggest that inflammation may not be linked to metformin-induced diarrhea.

To compare the present overeating model with db/db mice [7], fecal microbes were also examined. Similar to metformin-induced diarrhea in db/db mice, the population of the Firmicutes family in feces was reduced when overeating mice exhibited diarrhea (Fig. S6A and B). These results suggest that, in the context of a metformin-induced diarrhea model, the present transient overeating model may resemble the genetic overeating mouse model (db/db).