Metformin, the most commonly used glucose-lowering drug, can alleviate patients’ hyperglycemia via the suppression of hepatic glucose production and the increase of glucose uptake and utilization in adipocytes and muscle cells (63). A vast body of studies revealed that metformin altered the gut microbiota community in T2DM (11, 64–66). A multicenter, randomized clinical trial suggested that metformin ameliorated hyperglycemia and hyperlipidemia in T2DM patients via increasing beneficial bacteria, such as Blautia and Faecalibacterium (64). Another randomized, placebo-controlled study showed that metformin perturbed the gut microbiome in individuals with treatment-naive T2DM (11). The authors also transplanted the fecal samples from donors (treated with metformin for 4 months) to germ-free mice and observed that glucose tolerance was improved by increasing the production of short-chain fatty acids (SCFAs) or altering plasma bile acid composition, suggesting a direct metabolic benefits of metformin. Similarily, Sun et al. demonstrated that Bacteroides fragilis was decreased and the bile acid glycoursodeoxycholic acid (GUDCA) was increased in newly diagnosed T2DM individuals treated with metformin, and the benefits of metformin were abrogated in HFD-fed mice colonizaed with Bacteroides fragilis, implicating that Bacteroides fragilis–GUDCA–intestinal farnesoid X receptor (FXR) axis mediated the glucose-lowering effect of metformin (67). A recent systematic review disclosed that pre-diabetes and newly diagnosed T2DM patients treated with metformin were correlated with increases in specific taxa associated with metabolic control, such as Enterobacteriales and Akkermansia muciniphila (68). In line with clinical research, studies in animal models further confirmed that metformin increased SCFAs production, reduced circulation lipopolysaccharides and inhibited intestinal proinflammatory signaling activities (65, 69, 70), thus contributing to improving metabolic disoders. Notably, gut microbiota could also mediate the side effects of metformin. Forslund et al. emphasized that a relative increase in abundance of Escherichia could enrich virulence factors and gas metabolism genes (10), which contributed to the gastrointestinal side effects of metformin.

TZD drugs belong to peroxisome proliferative activated receptor (PPARG) agonists and improve insulin sensitivity for T2DM (71). It reduced hepatic glucose production and increased peripheraltion of glucose and lipid metabolism, thus improving glycemic control. Few studies have discussed the interaction between gut microbiome homeostasis and insulin sensitizers and insulin in T2DM (72, 73). Full-length bacterial 16S rRNA sequencing and RNA sequencing analysis presented that rosiglitazone improved insulin sensitivity without altering the composition of gut microbiome but modifying gene expression signatures associated with lipid and carbohydrate metabolism as well as immune regulation in diabetic mice (72). Moreover, insulin improved taurine and hypotaurine metabolism via increasing Fusobacterium and up-regulating the genes involved in triglyceride and arachidonic acid metabolism (73).

α-GIs, including acarbose, voglibose and miglitol, are considered to postpone the digestion of carbohydrates in the intestinal tract and reduce postprandial hyperglycemia in noninsulin-dependent T2DM (74). They are commonly used oral glucose-lowering drugs in China and many Asian countries. A randomized clinical study revealed that acarbose increased the abundance of Bifidobacterium, Eubacterium and Lactobacillus, and lowered the abundance of Bacteroides in Japanese patients with T2DM (75). Likewise, in Chinese patients with T2DM, Su et al. found that acarbose treatment increased the content of Bifidobacterium and Enterococcus, as well as decreased some inflammatory cytokines (76). Mechanistically, Gu et al. highlighted that acarbose altered the relative abundance of microbial genes involved in bile acid metabolism and improved metabolic parameters (12). Interestingly, acarbose also increased the relative abundance of Ruminococcus and Bifidobacterium in ZDF rats (77). On the other hand, acarbose was an inhibitor of both human and bacterial α-glucosidases, which might limite the ability of the target microbiome to metabolize complex carbohydrates, thus leading to the resistance of acarbose (78). Additionally, due to the weakened microbial enzyme activities, the metabolism of voglibose was reduced, along with significantly glucose-lowering effects were presented in antibiotic pretreatment mice (79), suggesting that gut microbiota mediated the effect of α-glucosidase inhibitors.

SGLT2 is expressed in the renal proximal tubule and accounts for reabsorbing the filtered glucose. SGLT2 inhibitors exert the glucose-lowering effect by blocking glucose reabsorption in the renal proximal tubule and increasing urinary glucose excretion, accompanied with pleiotropic benefits in cardiovascular and renal protection (80, 81). Several studies have explored the alteration of gut microbiota with SGLT2 inhibitor treatment (82–85). After a 3-month intervention, empagliflozin improved cardiovascular disease (CVD) risk factors in patients with T2DM, which might be attributed to the significantly altered gut microbiota, including the elevated levels of SCFA-producing bacteria (Roseburia, Eubacterium, and Faecalibacterium) and a reduction in several harmful bacteria (Escherichia–Shigella, Bilophila, and Hungatella) (82). Whereas, van Bommel et al. reported that 2-week treatment with dapagliflozin and gliclazide did not affect either gut microbiome alpha diversity or composition in T2DM patients treated with metformin (83). This discrepancy might be due to the fact that all the participants had already been treated with metformin, which could overshadow the possible impact of dapagliflozin on the gut microbiome. In T2DM mice, dapagliflozin treatment showed a trend for increased Akkermansia muciniphila and decreased Oscillospira and Firmicutes/Bacteroidetes ratios (84). However, another study demonstrated that dapagliflozin did not increase the abundance of beneficial bacteria (85). Therefore, more rigorous clincial studies with greater sample size are needed to figure out the interactions between SGLT2 inhibitors and gut microbiota.

GLP-1 secreted by enteroendocrine L cells is an incretin hormone and stimulates glucose-dependent insulin secretion. GLP-1 RAs, a new type of hypoglycemic drugs, mimic the effects of endogenous GLP-1, as well as improve glycemic control and cardiovascular outcomes for T2DM patients (86, 87). Accumulating evidence reported that GLP-1 RAs were linked to the changed composition of gut microbiota (88–91). In liraglutide-treated diabetic male rats, several SCFA-producing bacteria, such as Bacteroides, Lachnospiraceae, and probiotic bacteria, Bifidobacterium, were selectively enhanced, which might alleviate systemic inflammation and improve glucose control (88). Besides, germ-free mice colonized with microbiota from liraglutide-treated diabetic mice were shown to improve glucose-induced insulin secretion and regulate the intestinal immune system (91). Also, in T2DM patients, microbial interaction network was altered in patients treated with liraglutide. The distribution of community structure differed between the pre-liraglutide-treatment group (21 species of bacteria were abundant) and post-liraglutide-treatment group (15 species were abundant) (89). Nevertheless, a recent study enrolling 51 T2DM adults with initial therapy of metformin and/or sulphonylureas showed that the diversity and composition of the intestinal microbiota did not change after 12-week liraglutide intervention (92). This inconsistency might be attributed to the initial therapy of metformin and/or sulphonylureas, which could counteract the effect of liraglutide. Recently, Tsai et al. found that gut microbiota contributed to the heterogenicity of GLP-1 RA responses in T2DM patients (90). To sum up, the positive microbial signatures, mainly including Bacteroides and Roseburia, with immunomodulation effects were dominant in GLP-1 RA responders, while the negative microbial signatures, such as Prevotella, Butyricimonas, Mitsuokella and Dialister, with pro-inflammatory properties were dominant in GLP-1 RA non-responders (90). Thus, gut microbiota may be a potential target to improve the GLP-1 resistance.

DPP-4 inhibitors improve hyperglycemic conditions by stabilizing GLP-1 and glucose-dependent insulinotropic polypeptides (93). A series of studies considered that DPP-4 inhibitors reshaped the microbial composition and increased fecal SCFAs to improve metabolic homeostasis (94–97). Liao et al. demonstrated that DPP-4 inhibitors promoted a functional shift of the altered microbiome induced by HFD, especially increasing the abundance of Bacteroidetes, which contributed to improving glucose homeostasis (94). Another study revealed that DPP-4 inhibitors displayed significantly decreased Firmicutes/Bacteroidetes ratios, and elevated levels of butyrate-producing Ruminococcus and Dorea in HFD-induced mice (97). Likewise, vildagliptin treatment also reduced the ratio of Fimicutes/Bacteroidetes, and increased butyrate-producing bacteria, including Baceroides and Erysipelotrichaeae, in HFD-induced SD rats (96). Furthermore, vildagliptin significantly reduced DPP-4 activity mainly by decreasing Oscillibacter and increasing Lactobacillus (95), which provided new therapeutic uses of DPP-4 inhibition to tackle gut microbiome dysfunctions in T2DM.

TCMs, known as botanical medicine or phytomedicine, could significantly improve glucose control by enhancing insulin sensitivity, simulating insulin secretion and protecting β-cells (98). In recent years, increasing evidence confirmed that TCMs improved glucose metabolisms and alleviated T2DM at least partly by modulating gut microbiota (99).

A number of studies in animal models of T2DM have extensively explored the interactions between TCMs and gut microbiota (100–107). Zhou et al. found that ginsenoside Rb1, one of the most valuable herbal medicine, increased the abundance of Parasutterella, and decreased Alistipes, Prevotellaceae, Odoribacter and Anaeroplasma in T2DM mice model, thus attenuating IR and metabolic disorders (100). In T2DM rats model, Baihu Rensheng decoction (BHRS) increased the relative abundance of Lactobacillus, Blautia, and Anaerostipes, as well as decreased the Allobaculum, Candidatus Saccharimonas, and Ruminococcus (101). Mechanically, BHRS was considered to repair gut barrier and inhibit TLR4/NF-κB-mediated inflammatory response. Similarly, Buyang Huanwu decoction (BYHWD), a widely used TCM formula, decreased the Firmicutes/Bacteroidetes ratio and increased the abundance of Lactobacillus and Blautia (102). Another study suggested that Fufang Fanshiliu decoction enriched the abundance of Lactobacillus, Akkermansia, and Proteus, and reduced the abundance of Alistipes, Desulfovibrio, and Helicobacter in T2DM rats model (104). Moreover, Liu-Wei-Di-Huang Pills improved glucose metabolism by promoting the abundance of Lactobacillus, Allobaculum, and Ruminococcus, and increasing SCFAs levels in T2DM rats model (106), which might be related to the SCFAs-GPR43/41-GLP-1 pathway. Shenqi compound (SQC), a TCM formula, has been widely used for T2DM. It showed that SQC exerted a beneficial role by decreasing the Firmicutes/Bacteroidetes ratio and modulating metabolites in different pathways (107). Gegen Qinlian Decoction exerted the glucose-lowering effect by significantly modulating the overall gut microbiota structure and enriching butyrate-producing bacteria, including Faecalibacterium and Roseburia, which subsequently attenuated intestinal inflammation (108). Ge-Gen-Jiao-Tai-Wan formula alleviated symptoms ofT2DM rats by increasing the beneficial phylum Firmicutes and bile-acid-related genus Lactobacillus, thus promoting the production of primary bile acids, and upregulating the PBA-FXR/TGR5-GLP-1 pathway (109). Intriguingly, a current study suggested that the Scrophulariae Radix and Atractylodes sinensis (XC) pair could assist metformin in improving postprandial hyperglycemia by inhibiting the increase of Bacteroides in T2DM rats model (105), which could effectively apply to clinical practice in treating T2DM. In addition to the animal studies, a clinical trial in newly diagnosed T2DM patients also underlined that the hypoglycemic effect of berberine was related to the inhibition of DCA biotransformation by Ruminococcus bromii (110). Collectively, these findings address the effect of antidiabetic drugs on gut microbiota ( Figure 3 ) and emphasize the host-microbe-drug interactions, providing promising microbiome-targeting approaches to treat T2DM.

With the completion of the human genome and the human microbiome projects (111), a number of large biobanks including gut microbiome and multiple omics data (information on genetics, transcriptome, proteome and metabolome) had been established, such as UK biobank and TwinsUK cohort (112, 113). These biobanks utilized clinical studies, involving well-characterized human cohorts with extensive clinical and demographic details, exploring the host-microbe-drug interactions. In parallel with the existing data, there was continuous need for digging deeper into the unknown filed of drug-microbiome interactions. The accumulated data in literature calls for the construction of predictive tools or models that consider all such parameters to provide accurate predictions (114, 115). Machine learning methods and network analyses, including decision-tree algorithms and random forest, could then be applied to create a predictive tool for the efficacy and toxicity of antidiabetic drugs in T2DM ( Figure 4A ). For instance, the T2DM prediction model based on the characteristics of the salivary microbiota (microbiome data) was established by random forest in elderly patients with T2DM (116). Another study showed that machine learning tools with gut microbiome profiling exhibited the highest overall predictive power for improving early prediction of T2DM (117). These findings not only had the ability to rapidly inform clinical practice but also elucidated hypotheses regarding the mechanisms in which microbial transformations of drugs changed their pharmacokinetic properties.

Given the interplay between the host, gut microbiome and drug metabolism, there is increasing awareness that we should take microbiome profile based on novel technologies into account when considering personalized medicine.

Considering that microbiome profiling of human samples provided evidence for microorganism-mediated drug metabolism, further experimental studies are required to identify the specific microbiome responsible for drug biotransformation. Experimental manipulations of gut microbiota incorporated the use of humanized gnotobiotic mice models to further investigate the specific role of the microbiota in modulating drug pharmacokinetics (118). Humanized gnotobiotic mice are typically either germ-free animals or those colonized with defined microbiota and achieved by transplanting human faecal microbiota into germ-free mice (119). As discussed, these models have already proven successful for the treatment of several diseases, including T2DM (120). In recent years, organs-on-chips and bacterial culturomics as emerging technologies also have been developed (121, 122), making functional validation of gut microbiome finally possible. Antidiabetic drugs of interest can be incubated via these technologies to assess their biotransformation by gut microbiome, enabling the development of personalized medicine in T2DM ( Figure 4B ).