Type 2 diabetes mellitus (T2DM) is one of the most common chronic metabolic disorders and is characterized by hyperglycemia resulting from the combination of insulin resistance and inadequate insulin secretion [1,2,3,4]. The number of people with T2DM has drastically increased over the past several decades [1]. Metformin, a biguanide class drug, is recommended by the American Diabetes Association and European Association for the Study of Diabetes as a first-line medicine for the treatment of T2DM [2]. Metformin is a derivative of phenformin and buformin from galegine in Galega officinalis, traditionally used to decrease blood sugar and relieve the symptoms of diabetes (Figure 1) [3,4]. Among the three biguanides, phenformin and buformin were withdrawn from the market due to the high frequency of lactic acidosis in the 1970s. However, metformin showed superior safety and better efficacy in the treatment of T2DM [5,6,7,8,9]. These advantages for clinical use have resulted in metformin being widely used for more than 60 years [5]. Metformin does not target a specific pathway or disease mechanism [4]; therefore, studies have aimed to reveal the mechanism of action of metformin related to the treatment of cancer and cardiovascular diseases [1,2,3,4,5]. Metformin exhibits the peak plasma concentrations in 3 h with C_max 1.0–1.6 mg/L for dose of 500 mg and approximately 55% of bioavailability ([6] and refences therein). After absorption, metformin is distributed in the liver, kidneys, adrenal glands, and pancreas at about seven-fold higher concentration than that of the serum [7,8]. Based on the evidence suggesting a higher accumulation of metformin in the liver as well as another report by Rena et al. [9], the liver is a potential target organ of metformin [10,11,12]. Several studies have suggested that metformin suppresses the hepatic gluconeogenesis resulting from glucose tolerance modulation mediated by the adenosine monophosphate-activated protein kinase (AMPK) activity [3,13,14]. Recent evidence from three studies suggests that the gut is a major target of metformin action and not the liver. First, metformin when administered intravenously, instead of orally, demonstrated no glucose-lowering effects [15,16,17]. Further, the jejunum tissue was found to exhibit a metformin concentration of up to 2000 μmol/kg of tissue, which was 30–300 times higher than the plasma concentrations [18,19,20]. The jejunum biopsy under pre-dose and post-dose of metformin demonstrated the gastrointestinal tract as a prominent target of metformin [18]. Second, the organic cation transporter (OCT) 1, expressed in the membrane of enterocytes, might be possibly involved in the absorption of metformin from the intestinal lumen [10,21]. According to Dujic et al. [22], a reduced function of OCT1 might increase the intestinal metformin concentration and the risk of gastrointestinal intolerance in the metformin-treated patients. Finally, the gut-restricted glucose-lowering effect of metformin was observed for intermediate-release metformin, extended-release metformin, and delayed-release metformin, and the same dose of metformin was more effective through those dosage forms than extended-release form [23]. Although various putative mechanisms of glucose homeostasis modulation in the gut by metformin have been proposed, more studies are needed to establish these hypotheses.

Microbiome in the human body assist in the expansion of host genomes, by facilitating the host’s metabolism and physiology [24,25]. Over the last few decades, the development of sequencing technologies and drastic progress in population-scale studies have revealed the host and microbiome relationship. Large-scale research projects on the microbiome have been actively conducted, such as the Human Microbiome Project (HMP) consortium funded by the United States National Institutes of Health (NIH) and the Metagenomics of the Human Intestinal Tract (MetaHIT) consortium funded by the European Commission [24]. A microbiome study demonstrated that the human gut microbiome abundance correlates with metabolic markers, such as adiposity, insulin resistance, and dyslipidemia [26]. Furthermore, gut dysbiosis has also been observed in T2DM patients [27,28,29,30,31,32,33,34,35]. Based on the hypothesis that metformin targets the human gastrointestinal tract, the gut microbiome has attracted attention as a key factor in the treatment of T2DM [36,37,38,39,40,41]. Thus, this review focused on the various studies related to the gut microbiome and its association with the anti-diabetic effects of metformin.

Over the past decade, several studies have demonstrated that patients with T2DM, obesity, or inflammatory bowel diseases often show dysbiosis in the gut microbiota [42,43,44,45]. The report by Larsen et al. [30] differentiated the composition of the gut microbiota in the T2DM patients from that in the non-diabetic adults (Table 1), and other studies have demonstrated dysbiosis in T2DM patients under different conditions, such as subject’s race and co-administration with other drugs. According to Larsen et al. [30], at the phylum level, the abundance of Firmicutes in T2DM patients was lower than that in the control group, and Bacteroidetes and Proteobacteria were more abundant than in the control group. The tendency of abundance at the phylum level was similar among the other clinical trials [29,30,33,34,35]. Furthermore, at the genus level, Roseburia, a butyrate-producing bacterium, was less abundant in the T2DM patients [27,29,30,32]. These results were in line with the other studies showing an increase in the abundance of Roseburia and insulin sensitivity after intestinal microbiota transplantation from lean donors to recipients with metabolic syndrome [46]. In addition, the abundance of Lactobacillus spp. was higher in T2DM patients than in the control groups [28,29,30]. The abundance of Lactobacillus spp. was positively correlated with blood glucose levels in the two clinical trials [29,30] and these results were consistent with those evident in a mice study [47]. The positive correlation between Lactobacillus spp. and the glucose levels might be due to the immunomodulatory role of Lactobacillus spp. [48]. Similarly, dysbiosis in T2DM patients might be due to the interaction of the gut microbiota with the host immune system, which was supported by several animal studies. In particular, the gut microbiota, which communicates with the host through pattern recognition receptors, such as toll-like receptors (TLRs), contributes to the development of insulin resistance with increased plasma LPS concentration [49,50]. According to Larsen et al. [30], the abundance of Gram-negative bacteria, which can stimulate the immune system like TLRs, was increased in T2DM patients. The role of TLRs in insulin resistance has been established through various studies. The TLR-5 deficient mice became obese and exhibited a metabolic syndrome. Further, when the gut microbiome from the TLR-5 deficient mice was transplanted to the germ-free mice, the germ-free mice showed a similar phenomenon as the TLR-5 mice [51]. In addition, Song et al. [52] reported that TLR-4 activation is associated with insulin resistance in adipocytes. Previously cited clinical studies have identified SCFA-producing bacteria as the key for dysbiosis in T2DM patients in response to the immune responses [27,28,29,32,33,34]. The gut microbiota has been considered as one of the factors affecting T2DM; thus, the gut microbiota might be considered a potential target for the treatment of T2DM. Several studies have demonstrated the positive effects of probiotics for the treatment of T2DM, such as the decrease in systemic LPS levels and improvement in insulin resistance [53,54].

The glucose-modulating effect of metformin on the gut microbiome has been evaluated in various clinical trials. The first clinical study that observed the relationship between metformin and the gut microbiome was conducted as an open-label, single-group study in T2DM patients [55]. In this study, they demonstrated alterations in the composition of the gut microbiome, glucose hormone, glucose-related parameters, and bile acid concentration in feces. A similar tendency was observed in the present study, despite minor differences in the gut microbiome composition (Table 1).

First, at the phylum level, the alterations in the abundance of Firmicutes and Bacteroidetes were remarkable on comparing visits 3 (non-treatment) and 4 (metformin treatment). Although there were differences among subjects, the abundance of Firmicutes was commonly increased, whereas that of Bacteroidetes was decreased after metformin treatment. This result is in line with the previous finding that the Firmicutes/Bacteroidetes ratios were considered a predictor for metabolic disease such as T2DM or obesity in several human studies [30,172,173]. The Firmicutes/Bacteroidetes ratios were decreased in the T2DM patients, and this phenomenon was recovered by metformin treatment in several clinical studies [55,56]. In contrast to these results, some studies did not show an alteration in the ratio of Firmicutes to Bacteroidetes [60]. Inconsistencies among studies could be considered for the following reasons. The phyla Firmicutes and Bacteroidetes are the most abundant bacteria in the human gut and include a large number of bacterial species. Thus, a comparison of the Firmicutes and Bacteroidetes ratio is considered too simple to evaluate metabolic disease or improvement. In addition, the difference might be attributed to the compositional difference between the stool and biopsy specimens. Clinical studies in this review used fecal samples to analyze the gut microbiome, but previous studies have shown differences in the microbial compositions of biopsy and fecal samples [42,174]. In particular, the mucosa-associated microbiota, to which the phylum Firmicutes is enriched, exhibited compositional differences in biopsies derived from colon and stool samples [175,176]. Thus, Hollister et al. [174] suggested biopsy or surgical specimens for the evaluation of mucosa-associated microbiota. For these reasons, further investigations require the validity of the Firmicutes/Bacteroidetes ratio as a relevant marker for metabolic diseases. At the genus level, it is noteworthy that Escherichia, including Escherichia/Shigella and Escherichia coli, exhibited a significant increase in T2DM patients upon metformin treatment. An increase in the abundance of Escherichia/Shigella upon metformin treatment was also observed in the other clinical trials including healthy volunteers [29,31,56,60,61]. Forslund et al. [31] suggested that metformin administration creates a competitive environment for Escherichia coli using nitrate or other energy sources, resulting in changes in the abundance of the gut microbiome [177]. Wu et al. [56] also demonstrated a change in the abundance of E. coli as an indirect effect of metformin treatment in the in vitro gut simulation. Elbere et al. [60] demonstrated that the abundance of Escherichia/Shigella before metformin treatment is associated with side effects. In this study, the increased presence of Escherichia/Shigella showed mild and severe side effects; however, this level was lower than the detection limit in the no-side-effect group. Thus, increased abundance of Escherichia was considered as a marker for the gastrointestinal side effects of metformin. Forslund et al. [31] suggested that side effects derived from Escherichia are due to an increase in lipopolysaccharide synthesis or sulfate metabolism potential, known to contribute to intestinal bloating [31,178,179,180,181].

In contrast, the abundance of Intestinibacter spp. decreased in T2DM patients treated with metformin in several clinical studies [31,56,61]. Until now, the role of Intestinibacter is still unclear, Forslund et al. [31] suggested that Intestinibacter showed resistance to oxidative stress and degradation of fucose, indicating indirect mucus degradation through analysis of SEED (http://pubseed.theseed.org/, accessed on 30 March 2021) [182] and gut microbial modules (GMM) functional annotations.

In addition, A. muciniphila, which is positively correlated with metformin treatment, showed a less clear link in human studies. Although Wu et al. [56] demonstrated an increase in A. muciniphila in the in vitro pure cultures, there was no correlation between the abundance of A. muciniphila and % hemoglobin A1c. Furthermore, clinical studies in healthy volunteers showed no change in the abundance of A. muciniphila when they were treated with metformin [60,61]. The reasons for these differences might be considered to be affected by factors dependent on individuals, such as fibers [183], polyphenol availability [184,185], immune response [186,187], and age [188,189]. Thus, it might be difficult to conclude the role of A. muciniphila in humans as a major contributor to the anti-diabetic effect of metformin, although improvements in the metabolic parameters were observed in the A. muciniphila-treated human studies.

From the perspective of biochemical alterations upon metformin treatment, there were some differences in the bile acid and SCFA concentrations in the feces. They found that metformin exposure increased the excretion of bile acid in feces, consistent with the inhibitory effect of metformin on the resorption of bile acids [161,162,163]. In addition, the abundance of Firmicutes and Bacteroidetes correlated with the bile acid concentration and gut peptide, suggesting that metformin indirectly regulates the secretion of gut hormones via bile acid metabolism. Increased SCFA concentration in feces or an increase in the abundance of SCFA-producing bacteria has been observed in human studies [31,56,57,58]. In particular, Wu et al. [56] only demonstrated that the concentration of SCFA in fecal samples, resulting in formation of butyrate and propionate, substantially increased on metformin treatment. This result is consistent with animal studies that showed that metformin increases SCFA-producing bacteria [88,89,90,91,109,110]. Thus, these clinical results support the hypothesis that metformin exerts beneficial effects via bile acids and SCFAs.

The clinical studies discussed in this review exhibited differences among studies, including observed taxonomic groups in metformin treatment and diversity in abundance. As far as diversity is concerned, only a few subjects were engaged in the clinical study, resulting in no statistical difference in the diversity of the gut microbiome. This issue has been inconsistent in clinical trials (Table 1). Forslund et al. [31] conducted a meta-analysis of metagenomic data from Swedish, Danish, and Chinese individuals. In this study, gut microbiome was less rich T2DM patients without metformin treatment; this richness slightly recovered, almost as much as that in the control group, in T2DM patients on metformin treatment [31]. The diversity was also shown to decrease in the Chinese T2DM patients [79], which was consistent with the results from the study by Forslund et al. [31].

These differences might be derived from the dosage, study duration, disease state, race differences, and sample size. Thus, to elucidate the anti-diabetic effect of metformin via modulation of the gut microbiome, clinical studies in ethnic-controlled environments or comparisons among ethnicities are required. Indeed, clinical studies in various populations have been conducted at ClinicalTrial.gov (assessed on 30 March 2021) (Table 3). Most of the research to date has revealed taxonomic groups in the gut at the genus level, and not at the species level, due to technical limitations. To counter this limitation, recent studies introduced a gut microbiome analysis method to make a possible profile at the species level [190,191,192]. In the future, these methods to analyze the gut microbiome could help clarify the relationship between metformin and the gut microbiome.