We obtained summary-level data from publicly available genome-wide relationship studies (GWAS). The GWAS study that provided data on GM was initiated by the international MiBioGen consortium (12), while the GWAS study offering data on T2DM microvascular complications was primarily sourced from the FinnGen database. The MR analysis was grounded on three pivotal preconditions: Firstly, the instrumental variables (IVs) chosen must exhibit a robust association with GM taxa. Secondly, these IVs operate independently from any confounders affecting both GM taxa and T2DM microvascular complications. Lastly, there should be no horizontal pleiotropy, meaning the IVs influence T2DM microvascular complications strictly through their effect on GM taxa (13). Ethical approval and consent to participate were secured for each cohort included in the GWAS studies, and the summary-level data were released for analysis ( Figure 1 ).

Drawing on the resources of the MiBioGen consortium, Kurilshikov et al. analyzed 16S rRNA gene sequencing profiles and genotyping data from 18,340 samples to delve into the interplay between genetic variations and GM. These samples represented subjects from 25 cohorts across 11 countries, all of European descent. From this effort, the GWAS study pinpointed 122,110 variant sites spread across 211 taxa, from genus to phylum levels, highlighting the variances in GM taxa across populations.

We utilized the R package “ieugwasr” to extract three GWAS datasets from the IEU platform (https://gwas.mrcieu.ac.uk/), their IDs are finn-b-E4_DM2OPTH, finn-b-E4_DM2NEU, and finn-b-E4_DM2REN. These datasets originate from the FinnGen research projects (https://r8.risteys.finngen.fi/phenocode/E4_DM2OPTH, https://r8.risteys.finngen.fi/phenocode/E4_DM2NEU, https://r8.risteys.finngen.fi/phenocode/E4_DM2REN). The source of the dataset is detailed in Supplementary Table S11 .

Statistical analyses were executed using R software (Version 4.1.1). We employed the “TwoSampleMR” R package to investigate the potential causal relationship between GM taxa and microvascular complications of T2DM. P value < 0.05 was deemed indicative of statistically significant evidence for a potential causal effect. To bolster the integrity of the data and the precision of the findings, SNPs associated with GM taxa and T2DM microvascular complications that met a genome-wide significance threshold (P < 1×10–⁵) were selected, these SNPs were further confirmed to be uncorrelated by a distance cut-off of 10,000 kilobases apart and a correlation index R² ≤ 0.001. In the absence of horizontal pleiotropy, the inverse variance weighted (IVW) test was used as the primary method for calculating the causal effect values to obtain unbiased estimates. A fixed/random effects model was selected for the IVW test based on the presence or absence of heterogeneity. OR and 95% confidence interval (CI) showed the effect size. The weighted median (WM) method and the MR-Egger test were utilized as additional methods for MR analysis. WM results were used as the significant causal effect values if the number of SNPs with heterogeneity exceeded 50%. MR Egger’s results remained valid if SNPs with pleiotropy were above 50%. Cochrane’s Q test was applied to test for heterogeneity. IVs with P<0.05 were considered heterogeneous. The intercept of MR Egger regression assessed the presence of potential pleiotropy in IVs. Horizontal pleiotropy was deemed to be non-existent if P >0.05. To ensure the accuracy of results for GM taxa causally related to T2DM microvascular complications (based on IVW results), the multipotency was further analyzed using the MR-Pleiotropy RESidual Sum and Outlier (MR-PRESSO) test (R package “MRPRESSO”). And, the leave-one-out method was used to validate data robustness. Additionally, we performed a reverse Mendelian randomization analysis to further assess whether there was a causal relationship between microvascular complications of type 2 diabetes and identified significant bacteria.

In our analysis, we employed three well-established MR methods. Utilizing the IVW approach, we pinpointed five specific microbial species with a causal association to T2DM retinopathy, as illustrated in Figure 2A . Notably, the Family Clostridiales vadin BB60 group (OR = 1.42, 95% CI: 1.11–1.81, P = 0.01), genus Actinomyces (OR = 1.43, 95% CI: 1.16–1.92, P = 0.02), and genus Lactobacillus (OR = 1.33, 95% CI: 1.05–1.68, P = 0.02) were identified as risk factors for T2DM retinopathy. Conversely, genus Gordonibacter (OR = 0.82, 95% CI: 0.69–0.98, P = 0.03) and genus Ruminococcaceae UCG010 (OR = 0.58, 95% CI: 0.39–0.88, P = 0.01) were recognized as protective factors ( Figure 2B ). Furthermore, the WM and MR-Egger methods yielded results consistent in both size and direction with those obtained through the IVW method, as depicted in Figures 3A–E .

The results presented in Table 1 , derived from the IVW analysis, provide compelling evidence suggesting that family Rhodospirillaceae (OR = 0.58, 95% CI: 0.40–0.85, P < 0.01), genus Dialister (OR = 0.53, 95% CI: 0.33–0.85, P = 0.01), and genus Sutterella (OR = 0.60, 95% CI: 0.37–0.96, P = 0.03) exhibit a protective effect against the development of T2DM neuropathy ( Figure 2D ). This underscores the potential significance of these microbial taxa in mitigating the risk of T2DM neuropathy. Figure 2C visually portrays the relationships between 211 GM and T2DM neuropathy. Notably, except for Rhodospirillaceae, the results from all three MR methods for the remaining species displayed similar effect sizes and orientations, as depicted in Figures 3F–H .

The results depicted in Figure 2E reveal that seven microbial species exhibit causal associations with T2DM nephropathy. Among these, Class Verrucomicrobiae (OR = 1.82, 95% CI: 1.24–2.68, P <0.01), family Verrucomicrobiaceae (OR = 1.82, 95% CI: 1.24–2.68, P <0.01), genus Akkermansia (OR = 1.82, 95% CI: 1.24–2.68, P <0.01), genus Flavonifractor (OR = 1.91, 95% CI: 1.11–3.30, P = 0.02), and order Verrucomicrobiales (OR = 1.82, 95% CI: 1.24–2.68, P <0.01) were identified as risk factors for T2DM nephropathy. In contrast, the genus Sutterella (OR = 0.63, 95% CI: 0.42–0.94, P = 0.02) and phylum Proteobacteria (OR = 0.52, 95% CI: 0.33–0.79, P <0.01) were identified as protective factors against T2DM nephropathy ( Figure 2F ). The consistency of the results across the three MR methods ( Figures 3I–O ) adds further strength to these findings, indicating that these microbial species and taxonomic groups play a consistent role in either increasing or decreasing the risk of T2DM nephropathy.

We use IVW and MR-Egger tests to detect heterogeneity, and the results indicate no heterogeneity among all the IVs (P <0.05, Table 1 ). In addition, the MR-PRESSO test the MR-Egger regression intercept test also showed no horizontal pleiotropy ( Table 1 ). And leave-one-out analyses suggested that the MR results are robust and not influenced by specific SNPs ( Supplementary Figure S1 ).

We used type 2 diabetes microvascular complications as an exposure factor and the fifteen identified significant gut microbial taxa as the outcomes for reverse MR analysis. The results showed that no significant causality was detected, except for genus Sutterella (OR = 1.05, 95% CI: 1.01–1.09, P = 0.02), which was reciprocally associated with type 2 diabetic nephropathy. The results of the reverse MR analysis are shown in Supplementary Table S10 .

Our discovery unequivocally demonstrates a causal relationship between retinopathy and specific taxonomic groups within the microbial community. Notably, the Family Clostridiales vadin BB60 group, as well as the genera Actinomyces, Gordonibacter, Lactobacillus, and Ruminococcaceae UCG010 were found to be intricately linked to the development of retinopathy. This insight contributes significantly to our understanding of the role these microbial entities play in the occurrence and progression of retinopathy.

Research into the connection between Clostridiales and retinopathy has been relatively limited. In a previous clinical trial conducted by Huang et al (14), the diabetic retina group exhibited decreased levels of Clostridium genera and increased levels of Lactobacillus when compared to the normal control group. However, in a foreign study (15), it was observed that patients with both (T2DM) and metabolic syndrome had elevated levels of Clostridiales. Concerning Actinomyces, its role in GM has been a subject of debate. Some studies have depicted it as beneficial for human health (16), while others have contested this notion. Relevant animal studies have shown that a high-fat diet can significantly elevate the presence of intestinal Actinomyces ( 17), and given that obesity was a primary contributor to the high prevalence of T2DM, which can lead to severe complications. Li et al. found that obesity can exacerbate retinopathy by increasing Actinobacteria levels, further supporting this view (18). The impact of Lactobacillus on diabetes remains unclear. As mentioned earlier, it can be increased in diabetic retinopathy patients (14). Nevertheless, in another study conducted by Li et al (19), it was found that Lactobacillus was reduced in the diabetic retina group compared to the diabetic group without retinopathy. Ruminococcaceae UCG010, belonging to the phylum Firmicutes, has shown potential in ameliorating metabolic disorders and inflammation during diabetes by influencing the production of Short-chain fatty acids (SCFAs) and the conversion of primary to secondary bile acids (BAs) (20, 21). In a MR study exploring the gut microbiome and T2DM, it was revealed that the risk of T2DM was inversely associated with Ruminococcaceae UCG010, consistent with our study, thus reinforcing the reliability of our findings (21). Furthermore, our research extends the understanding of the relationship between Ruminococcaceae UCG010 and diabetic retinopathy, providing a promising direction for future investigations in this field.

To date, no relevant studies have explored the association between Gordonibacter and T2DM retinopathy.

Our research also uncovered a significant causal relationship between the microbial taxa Rhodospirillaceae, Dialister, and Sutterella with respect to T2DM neuropathy.

Remarkably, most studies focusing on Rhodospirillaceae have been conducted in animal models. Pei et al. conducted an animal study in which mice were divided into a normal diet group and a high-sugar diet group (HGD) (22). After 16 weeks of feeding, the HGD group exhibited a notable increase in the abundance of Rhodospirillaceae in the GM, accompanied by the development of neuromuscular dyskinesia. The findings from this study slightly differ from our research, and it’s possible that the disparity can be attributed to the variation in subjects studied, i.e., animals versus humans. The role of Dialister in T2DM patients has been a subject of contention. A Romanian cohort study suggests that Dialister may potentially have deleterious effects in T2DM (23). However, a systematic review has indicated a decrease in Dialister abundance in patients with newly diagnosed T2DM (24, 25). On the other hand, Sutterella, a gram-negative bacterium, has shown associations with diabetes (26). In their study, Gaike et al. observed that, in comparison to the healthy group, the abundance of Sutterella was reduced in newly diagnosed T2DM patients, hinting at a potential protective role in the development of T2DM (27). Nonetheless, none of the prior studies have delved further into the relationship between these two microbial species and T2DM neuropathy. This innovative aspect of our study not only sheds new light on this specific connection but also paves the way for future research in this domain.

According to our research, we have identified a causal relationship between several microbial taxa and T2DM nephropathy, including the class Verrucomicrobiae, orders, families, the genera Akkermansia, Flavonifractor, and Sutterella, as well as the phylum Proteobacteria.

In a comparative analysis of the gut microbiome composition between patients with T2DM and healthy individuals, a significant increase in the abundance of Verrucomicrobia was observed at the phylum level (28). Additionally, research conducted in T2DM mouse models indicated a substantial increase in Verrucomicrobia at multiple taxonomic levels, including phylum, class, and order, implying its potential involvement in accelerating the development of T2DM (29). Salguero et al. conducted a study comparing the T2DM chronic kidney disease group with healthy controls and found that Verrucomicrobia exhibited increased abundance in patients within the kidney disease group (30). These findings align with the results of our study, supporting the notion that Verrucomicrobia may play a role in the progression of T2DM nephropathy. Surprisingly, investigations into the relationship between Akkermansia and nephropathy have predominantly centered around animal experiments, yet consistent results remain elusive. In the study conducted by Shi et al (31), it was discovered that Huangkui capsules could enhance the metabolic profiles of mice with diabetic nephropathy by modulating the GM, notably by decreasing the abundance of Akkermansia. Conversely, Luo et al. found that inulin-type fructans could lead to the enrichment of Akkermansia. This enrichment, in turn, ameliorated glomerular injury and renal fibrosis by elevating serum acetate concentration (32). These contrasting findings underscore the complexity of the relationship between Akkermansia and nephropathy, indicating the need for further research to unravel the underlying mechanisms and potential therapeutic applications in this context. Regarding Flavonifractor, in an experiment that involved the comparison of GM between individuals with diabetic nephropathy (DN) and those with diabetes alone, a noteworthy observation emerged. It was evident that Flavonifractor exhibited a marked increase in the DN group in contrast to the diabetic-alone group. Furthermore, subsequent investigations have robustly supported these findings, revealing a significant and positive correlation between Flavonifractor and parameters such as urinary albumin/creatinine ratio and thylakoid stromal hyperplasia (33). The consistency of these results with our own study highlights the potential significance of Flavonifractor in the context of diabetic nephropathy. Research on the causal relationship between Proteobacteria and nephropathy has been somewhat limited. In a study by Li et al., it was shown that obesity could worsen nephropathy by disrupting the balance of the intestinal microbiota, resulting in an increase in the relative abundance of Proteobacteria (18). However, it’s worth noting that their findings contradicted our own results, and this discrepancy might be attributed to the fact that their study was conducted on mice.

Regrettably, no investigation has evidence supporting a genetic correlation between Sutterella and nephropathy.

Our research boasts several notable advantages. First and foremost, our study benefitted from a large sample size and was conducted using a comprehensive GWAS database, ensuring the robustness and reliability of our conclusions. This was particularly advantageous when investigating GM, where previous research had often yielded conflicting results. Second, our MR analyses uncovered previously unknown and meaningful microbial species, such as Gordonibacter and two unidentified species. Additionally, we shed light on microbial species that had not been thoroughly investigated in relation to type 2 diabetes complications, such as Ruminococcaceae UCG010 and Dialister. These discoveries provide valuable directions for future research, offering fresh insights into the complex relationship between GM and type 2 diabetes complications.

There were also some limitations to our study. Firstly, the data used in our study primarily pertained to European populations, which may exhibit distinct microbiota compositions compared to other populations, so our conclusions may not be applicable to other populations. Secondly, the threshold of 5x10^-8 was considered the gold standard for selecting p-values in MR analyses. However, in our study, setting the GWAS p value at 5 × 10^-8 during the screening process resulted in an insufficient number of SNPs for the analysis. Therefore, we established the GWAS p value threshold at 1 × 10^-5. Thirdly, numerous previous studies have suggested a potential link between these three comorbidities (34–36), but in our study, the SNP data for these three complications were sourced directly from the FinnGen database. Each GWAS was conducted independently, and the associations with gut microbiota were analyzed separately.

Moreover, it is crucial to consider the stage, severity, and duration of T2DM in the individuals studied. The gut microbiota can change considerably from the onset and development to the long-term persistence of T2DM and the progression of related complications. Early-stage disorders might exhibit dysbiosis as a causative factor in T2DM and microvascular complications. Conversely, in long-term disorders, alterations in bacterial levels might represent an adaptive response of the microbiota to mitigate systemic dysfunction. These are essential for interpreting our data accurately, considering the significant confounding effects these factors may impose.

Also, the influence of the gut microbiome on pancreatic function and microvascular metabolism is complex and mediated indirectly through an array of bacterial bioactive factors and metabolites. These factors likely act through multiple stages, modulating various aspects of host metabolism, which, in turn, might contribute to the development of retinopathy, neuropathy, or nephropathy. The host’s health and nutritional state also play critical roles in these interactions, potentially influencing the gut microbiome’s impact on disease progression. Individual differences in these host states might explain some of the contradictory findings observed between our study and others. Acknowledging these potential confounding factors provides a more comprehensive understanding of the complex interplay between the gut microbiome and T2DM complications.

The gut microbiota was significantly associated with T2DM microvascular complications through the production of metabolites such as short-chain fatty acids (SCFAs), amino acids, trimethylamine N-oxide (TMAO), bile acids, and indolepropionic acid (37). However, in our study, we focus on the results and discussed the partly metabolites. The remaining gut microbiota will influence the development of microvascular pathologies by releasing which metabolites can be used as a direction for future research. In addition to the well-characterized bacterial constituents of the gut microbiota, other lesser-studied microorganisms such as archaea, mycobionts, and viruses also reside in the gut and may play roles in the pathogenesis of T2DM and its associated microvascular complications, including retinopathy, neuropathy, and nephropathy (38–40). While bacteria dominate the research landscape, these studies suggest that the diverse non-bacterial components of the microbiome could also influence metabolic processes and inflammatory pathways implicated in T2DM. Given the complexity and multifactorial nature of T2DM and its complications, future research should consider these organisms to gain a more comprehensive understanding of the microbial influences at play. Exploring the interactions between these microorganisms and the host may reveal new mechanisms of disease progression and potential therapeutic targets.