Introduction

The global incidence of diabetes has been steadily increasing, as it is a metabolic disorder characterized by hyperglycemia. As of 2021, the global prevalence of diabetes was 6.1%, with a staggering 537 million affected individuals. It is projected that by 2050, the overall prevalence of diabetes will exceed 10%, with over 13.1 billion people affected, imposing a significant burden worldwide1,2,3. Characterized by the accumulation of extracellular matrix (ECM), mesangial expansion, thickening of the basement membrane, glomerulosclerosis, tubular inflammation, atrophy, and interstitial fibrosis, diabetic nephropathy (DN) stands as the prevailing microvascular complication of diabetes mellitus4. DN serves as a major cause of end-stage renal disease (ESRD) globally5. Research indicates that ESRD has a profound impact on global health, directly contributing to increased morbidity and mortality rates, as well as serving as a significant risk factor for cardiovascular diseases. Approximately 30–50% of ESRD cases worldwide can be attributed to DN. Although current treatments for DN partially alleviate patients’ conditions, they fail to completely halt the progression of DN to ESRD. The pathogenic mechanisms of DN are still complex and poorly understood, and effective therapeutic approaches remain inadequate. Therefore, there is an urgent need to explore prospective treatment strategies for DN6.

In the realm of metabolic diseases, the connection between the gut microbiota and these ailments has emerged as an influential and extensively investigated area of study7. Over the past years, a considerable volume of supporting data has underscored a noteworthy link between the gut microbiota and metabolic conditions, encompassing obesity, non-alcoholic fatty liver disease, and diabetes8,9,10. Moreover, recent studies have demonstrated that an imbalance in gut bacteria can impact the progression of chronic kidney disease (CKD) through pathways involving the production of uremic toxins, regulation of inflammation, and modulation of the immune system11. In the occurrence and development of DN, alterations in the gut microbiota play a crucial and indispensable role12. A profound disparity in the composition of gut microbiota has been unveiled through research, delineating a strong link between gut dysbiosis and the initiation as well as advancement of DN, thereby implying a close correlation between individuals afflicted by DN and those in a state of good health13. Additionally, some studies have shown a correlation between specific microbial abundances and clinical indicators of DN, such as urine protein levels and the extent of renal impairment14. These findings offer valuable insights that can guide further exploration into the mechanistic roles played by the gut microbiota in the development of DN, warranting deeper investigation. However, a critical question that remains to be answered is how the gut microbiota leads to renal damage in individuals with diabetes. The characteristics of gut microbiota imbalance include disruptions in homeostatic equilibrium and abnormal generation of bacterial metabolites. While certain microbial metabolites (both beneficial and harmful) have been implicated in the pathological progression of DN15, the specific causal relationships between various microbial communities and DN are still unclear due to the diverse array of gut microbiota species. The quest to unveil the underlying pathological mechanisms encountered various obstacles, demanding additional research to surmount these current challenges and attain a more holistic comprehension of the intricate correlation between DN and the gut microbiota.

The utilization of genetic variation as instrumental variables characterizes the research approach known as mendelian randomization (MR), facilitating the examination of whether an exposure factor possesses a causative impact on the outcome of interest16. Compared to conventional methods, MR could avoid reverse causation bias and attenuates the interference of confounding factors, so it is becoming increasingly popular in epidemiologic studies17. In recent years, genome-wide association studies (GWAS) have detected genetic variants associated with plasma proteome levels18,19, which provide an opportunity to utilize MR to further reveal the underlying mechanisms behind the causal relationship between exposure and outcome, as many circulating proteins always act as the principal regulators of molecular pathways20.

Hence, in this study, we performed a comprehensive two-sample MR analysis to systematically explore the causal relationship between gut microbiota and DN, and further explored the mediating role of plasma proteome between them. Through the elucidation of these mediating effects, we can acquire valuable insights into the mechanistic pathways through which the gut microbiota may exert influence on the risk of DN, thereby opening avenues for potential therapeutic interventions.

Methods

Data source

Research design of this study is shown in Fig. 1. The GWAS summary data related to gut microbiome abundance were published by Kurilshikov et al. in 2021. This study conducted a genome-wide association study on 211 types of gut microbiome abundance involving 18,340 participants (14,306 as European population) from 25 cohorts of diverse ethnicities and ages21. Genetic variation data related to DN were obtained from FinnGen. The FinnGen study is a large-scale genomics project that analyzed over 500,000 samples from Finnish biobanks, associating genetic variations with health data to understand disease mechanisms and susceptibilities. The project is a collaboration among research organizations, biobanks within Finland, and international industry partners. On May 11, 2023, the database released the R9 version of GWAS data, from which GWAS summary data for 4,111 DN cases and 312,650 controls were obtained based on ICD-10 code “N08.3”22. GWAS data based on the plasma proteome were published by Sun et al. in 2018. The study included approximately 50,000 participants in a randomized trial with different blood donation intervals. From mid-2012 to mid-2014, 25 centers of the NHS Blood and Transplant (NHSBT) in England recruited blood donors aged 18 and above, excluding those with a history of major diseases such as myocardial infarction, stroke, cancer, Human immunodefciency virus (HIV), Hepatitis B or C virus, or recent illness or infection. A multiplexed aptamer-based method (SOMAscan assay) was used to measure the relative concentrations of 3,622 plasma proteins or protein complexes, assessed using 4,034 modified aptamers18. Prior to analysis, all the aforementioned data underwent a rigorous ethical review and encompassed a diverse range of participants from European populations, encompassing both males and females, thus effectively mitigating the potential for bias resulting from population stratification. The information about the included GWAS datasets is provided in Table 1 and Supplementary Tables 1, and there is no sample overlap in the MR analysis of this study based on the information from the data set sources.

Fig. 1
figure 1

Study design flowchart.

Full size image
Table 1 Details of GWAS datasets included in MR studies.
Full size table

Selection of instrumental variables

To perform MR analysis, the instrumental variables must satisfy three assumptions: the relevance assumption, the independence assumption, and the exclusion restriction assumption. Based on these assumptions, we established criteria for selecting instrumental variables. In accordance with the relevance assumption, all eligible single nucleotide polymorphisms (SNPs) must be significantly associated with the exposure. We identified SNPs linked to exposure using a genome-wide significance level as our selection criterion (P-value < 5 × 10−8). In the mediation MR analysis, due to varying associations between different plasma proteins and genetic variants, extracting too many SNPs can enhance heterogeneity, while too few can result in insufficient explanatory variance, thus a second threshold was set (P-value < 1 × 10−5). SNPs exhibiting linkage disequilibrium (LD) were excluded based on an r2< 0.001 and a window size > 10,000 kb. Data on outcome-related SNPs were extracted based on SNP identifiers within the outcome dataset. During this process, ambiguous SNPs and palindromic SNPs were removed, matching SNPs across both datasets. We calculated the F-statistic for each SNP. The F-statistic, an intermediate measure in variance analysis, gauges the strength of association that instrumental variable SNPs have with exposure23. SNPs with an F-statistic less than 10 were considered weak instrumental variables and excluded. To satisfy the independence assumption, i.e., to ensure that SNPs are not associated with any confounders, the MR-PRESSO test, a method used to detect and correct outliers in Mendelian randomization caused by external factors such as pleiotropy of genes, was conducted to identify and exclude potential pleiotropic SNPs. PhenoScanner, an online database for retrieving phenotype information associated with specific genes, was also used to meticulously examine each SNP to assess its potential association with confounders, effectively removing SNPs that demonstrated potential violations of the independence assumption24. To fulfill the exclusion restriction assumption, i.e., to ensure that SNPs exert no direct effect on the outcome but only affect it indirectly via the exposure, the MR-Steiger test, a test employed to determine the correct causal direction in Mendelian randomization analyses, was used to verify the causal inference direction of each SNP, excluding those with incorrect directions25. Following rigorous screening, the remaining SNPs were deemed valid instrumental variables. In summary, the MR processes for direct and mediated effects were consistent, except for the different extraction thresholds for SNPs.

Mendelian randomization analysis

Classic MR requires the use of inverse variance weighting (IVW). To evaluate the robustness of the results, we also utilized MR-Egger regression, weighted median (WM), weighted mode approach, and MR robust adjusted profile score (MR-RAPS) to comprehensively assess potential biases. The IVW method employs a form of variance-inverse weighting to combine the causal estimates of the Wald ratio for each SNP26. The Wald ratio estimate measures the ratio of the impact of a single SNP on the outcome to its impact on exposures, assuming all associations conform to a logarithmic linear relationship27. In the realm of MR analysis, the utilization of MR-Egger regression emerges as an invaluable instrument for establishing a weighted linear regression model that examines the relationship between exposure coefficients and outcomes28. Similar to the IVW method, MR-Egger regression allows for the assessment of horizontal pleiotropy through the significance of its intercept. The foundation of the MR-Egger method rests upon the underlying assumption of No Measurement Error (NOME). We also calculated the I2statistic to quantify the degree to which MR-Egger violates the NOME assumption. Results should be adjusted when I2is less than 90%. When multiple variants are invalid, the results from the above methods may lack robustness; in such cases, the WM method and weighted median approach demonstrate greater robustness. The median weights for individual genetic variants are computed by the WM method, which combines these weights to generate an estimate. Significantly, the WM method can yield dependable estimates of causal effects, provided that at least 50% of the weights utilized in the analysis originate from valid instrumental variables29. Even in the presence of some invalid instrumental variables, the WM method can accurately estimate causal relationships and enhance precision. The weighted median method remains robust even with a higher number of invalid instrumental variables. According to the MR analysis guidelines, when pleiotropy is present, MR-Egger regression is employed as the primary analytical method, when heterogeneity is present, IVW random effects model with reference to the weighted median method is employed, and if no heterogeneity is detected, the IVW random effects model is used as the main analysis method. In addition, this study incorporates the recently developed MR-RAPS technique, which employs a random effects distribution to directly model the pleiotropic effects of genetic variants. Compared to traditional Mendelian randomization techniques, this novel strategy offers enhanced robustness. In the analysis of the gut microbiome, to avoid errors due to multiple comparisons (i.e., Type I statistical errors, where the risk of bias increases as the number of comparisons rises, leading to false positives at a P-value threshold of 0.05), we applied the false discovery rate (FDR) method to adjust the P-values. This approach helps to reduce the false positive rate that may arise from spurious statistical significance when performing numerous statistical tests. Statistical significance was considered when the adjusted P-value was < 0.05.

Sensitivity analysis

Pleiotropy includes horizontal pleiotropy and vertical pleiotropy. Typically, vertical pleiotropy does not affect the reliability of conclusions; however, horizontal pleiotropy should be eliminated. To gauge the degree of horizontal pleiotropy, our primary approach involves the utilization of the MR-Egger method for estimation. If the P-value of the MR-Egger intercept is less than 0.05, the instrumental variable is considered to be severely affected by horizontal pleiotropy, rendering the results unreliable. In the context of MR analysis, heterogeneity may still be present among SNPs, even if they are all deemed valid instrumental variables. The presence of substantial heterogeneity can render results unreliable; thus, heterogeneity testing is necessary to enhance the credibility of the outcomes. We employ the IVW method to calculate heterogeneity among SNPs and utilize Cochran’s Q test. If P< 0.05, heterogeneity is considered present. Additionally, as per convention, we use the leave-one-out method and have produced a funnel plot. We conducted a comprehensive MR-Steiger test to ensure the correct overall causal direction30. Finally, we calculated the statistical power to ascertain the reliability of the negative results. The calculation method has been described in detail in previous literature31.

Proteomics mediation analysis and enrichment analysis

Using GWAS summary data, a two-step mediated MR analysis was conducted to determine whether plasma proteins are intermediary factors in gut microbiota-induced DN, and the process of MR analysis was described as previously. The first step involved a two-sample MR analysis between significant gut microbiota and the plasma proteome, followed by a two-sample MR analysis between the plasma proteome and DN. Proteins significant in all three analyses exhibit partial mediation effects; those significant only in the two-step mediation analysis display complete mediation effects, while proteins not consistently significant across the analyses show no mediation effects. Indirect effects are calculated using the formula β1*β2, while direct effects are determined by subtracting the indirect effects from the total effects. The Reactome database was used to analyze the list of intermediary plasma proteins to understand enriched pathways. The Reactome database serves as a peer-reviewed repository of human biological pathways and reactions. To assess the potential enrichment of specific Reactome pathways within the gene list, over-representation analysis was performed, resulting in the generation of probability scores and significance P-values.

Statistical software and plotting

The conclusions section features various visualizations to depict the findings. For each SNP, scatter plots were generated (Supplementary Fig. 1) to illustrate the associations between exposure factors and outcome effects, accompanied by regression curves to visualize causal estimates. The results of the MR analysis were displayed using a significance heatmap. Funnel plots were employed to assess potential directional effects, pleiotropy, and data distribution (Supplementary Fig. 2). Forest plots were created using the final causal estimates, showcasing the results for each SNP as well as the overall outcomes of the MR analysis (Supplementary Fig. 3). Leave-one-out analysis plots were used to identify whether the SNP was a strong influence utility on the results (Supplementary Fig. 4). The statistical analyses in this study were performed utilizing R (version 4.2.3) and three R packages, namely “Two Sample MR,” “MR-PRESSO,” and “mr.raps.”

Results

Datasets information

SNPs associated with the abundance of Catenibacterium primarily originate from chromosomes 1/4/6/7 (Fig. 2-A), with the top SNP being rs12404911 (Fig. 2-B). SNPs linked to the abundance of Parasutterella are found across multiple chromosomes (Fig. 3-A). The top SNP is rs2387977 (Fig. 3-B). SNPs associated with DN are predominantly located on chromosome 6 (Fig. 4-A). The top SNP is rs9273363 (Fig. 4-B).

Fig. 2
figure 2

Analysis related to Catenibacterium. Note: (A) Manhattan plot of the summary data from the genome-wide association study on Catenibacterium; (B) The top 5 single nucleotide polymorphisms (SNPs) most significantly associated with Catenibacterium; (C) Scatter plot and regression curve of Mendelian randomization exploring the causal association between Catenibacterium and diabetic nephropathy; (D) The top 5 plasma proteins most significantly influenced by the abundance of Catenibacterium; (E) Scatter plot of the significance level of Catenibacterium’s effects on plasma proteins, with specified proteins indicating mediation effects.

Full size image
Fig. 3
figure 3

Analysis related to Parasutterella. Note: (A) Manhattan plot of the summary data from the genome-wide association study on Parasutterella; (B) The top 5 single nucleotide polymorphisms (SNPs) most significantly associated with Parasutterella; (C) Scatter plot and regression curve of Mendelian randomization exploring the causal association between Parasutterella and diabetic nephropathy; (D) The top 5 plasma proteins most significantly influenced by the abundance of Parasutterella; (E) Scatter plot of the significance level of Parasutterella’s effects on plasma proteins, with specified proteins indicating mediation effects.

Full size image
Fig. 4
figure 4

Analysis related to diabetic nephropathy. Note: (A) Manhattan plot of the summary data from the genome-wide association study on diabetic nephropathy; (B) The top 5 single nucleotide polymorphisms (SNPs) most significantly associated with diabetic nephropathy; (C) The top 5 plasma proteins most significantly affecting the incidence of diabetic nephropathy; (D) Scatter plot of the significance level of the effects of plasma proteins on diabetic nephropathy, with specified proteins indicating mediation effects; (E) Schematic diagram of the interaction effects of plasma proteins on the incidence of diabetic nephropathy.

Full size image

Screening of instrumental variables

In the MR analysis, initially, a total of 1823 exposure-related SNPs were screened, and no weak instrumental variables were identified; 92 SNPs were excluded from the outcome database due to missing data. After screening with Phenoscanner, a total of 280 SNPs were identified as ambiguous or palindromic SNPs, while 36 SNPs were found to be associated with confounding factors. The MR-PRESSO test revealed no SNPs with horizontal pleiotropy; the MR-Steiger test found no SNPs with incorrect causal direction. After Bonferroni correction, 5 SNPs directly related to the outcome were removed, resulting in 1410 qualified SNPs included in the study. In the mediation MR analysis, initially, 177,296 SNPs associated with exposures were screened, with no weak instrumental variables identified; 15,420 SNPs were removed from the outcome database due to missing data. A total of 20,333 ambiguous or palindromic SNPs were removed during the data merging process (Supplementary Table 2). The MR-PRESSO test revealed no SNPs with horizontal pleiotropy; the MR-Steiger test found no SNPs with incorrect causal direction. After Bonferroni correction, 2,685 SNPs directly related to the outcome were removed, resulting in 138,858 qualified SNPs included in the study. Detailed included SNP data are shown in Supplementary Table 3.

Mendelian randomization analysis

As shown in Fig. 5, in the MR analysis, 21 gut microbiotas were found to have a causal relationship with the onset of DN, with 2 remaining significant after adjusting the P-values. Specifically, an increased abundance of Catenibacterium [OR (95% CI): 1.282 (1.178 ~ 1.395), P = 9.545 × 10−09, FDR < 0.001, Pheterogeneity = 0.592, Ppleotropy = 0.903, PMR−Steiger = 0.733, PMR−PRESSO = 0.910] and Parasutterella [OR (95% CI): 1.270 (1.132 ~ 1.426), P = 4.945 × 10−05, FDR = 0.004, Pheterogeneity = 0.385, Ppleotropy = 0.952, PMR−Steiger = 0.501, PMR−PRESSO = 0.945] was both associated with an increased risk of DN, as shown in Figs. 2-C and 3-C respectively. These findings showed no evidence of pleiotropy or heterogeneity. Detailed MR results are presented in Supplementary Table 4. Detailed results of all sensitivity analysis results for this study are presented in Supplementary Tables 5–7. In the mediation MR analysis, causal effects were observed for Catenibacterium on 291 proteins such as Stromelysin-2 after adjusting P-values (Fig. 2-D, E), and for Parasutterella on 81 proteins such as 3-hydroxyanthranilate 3,4-dioxygenase (Fig. 3-D, E), and changes in the levels of 253 proteins such as Protein kinase C iota type were causally related to the onset of DN (Fig. 4-C, D). Subsequently, a protein-protein interaction analysis was conducted for plasma proteins inducing DN (Fig. 4-E), and finally, 22 plasma proteins with mediating effects were identified (Table 2; Fig. 6-A, B, C), which mediated the effects of Catenibacterium and Parasutterella on the onset of DN, with mediation effect sizes ranging from 72.9 to 2.8%. Detailed intermediation analysis data are presented in Supplementary Table 8.

Fig. 5
figure 5

Heatmap of the significance of the causal relationship between gut microbiota and diabetic nephropathy. Note: From the outer to the inner layers are the IVW random effects model, MR-Egger model, weighted median model, weighted mode model, and MR-RAPS model.

Full size image
Table 2 Significant mediating metabolism or signal transduction pathways.
Full size table
Fig. 6
figure 6

Results of mediation and enrichment analyses. Note: (A) Forest plot of the total and two-step mediation effects of Catenibacterium; (B) Forest plot of the total and two-step mediation effects of Parasutterella; (C) Scatter plot of the two-step effects of mediation proteins; (D) Pathways of metabolic or signaling transduction mediated by Parasutterella in the incidence of diabetic nephropathy; (E) Pathways of metabolic or signaling transduction mediated by Catenibacterium in the incidence of diabetic nephropathy.

Full size image

Enrichment analysis

In the enrichment analysis, 27 significant pathways mediating the effects of Catenibacterium and Parasutterella on the onset of DN were identified. Specifically, 11 significant pathways were associated with the pathogenesis of DN due to Catenibacterium, with the most significant being activation of transcription factor-6 (ATF6-alpha) activates chaperone genes (Reactions 1/5, Entities 1/10, P-value 0.014); 16 significant pathways were linked to the pathogenesis due to Parasutterella, with the most significant being Pyruvate metabolism (Reactions 2/15, Entities 1/31, P-value 0.016). Table 3; Fig. 6-D, E display the enriched information on mediating pathways, with detailed data available in the Supplementary Table 9.

Table 3 Activated mediating metabolism or signal transduction pathways.
Full size table

Discussion

The pivotal role of gut microbiota in metabolic disease development cannot be overlooked, as disruptions in gut microbiota are intricately linked to various metabolic disorders such as obesity, non-alcoholic fatty liver disease, and diabetes mellitus8,9,10. Consequently, targeting gut microbiota has emerged as a significant intervention strategy for managing diverse ailments7. It has been found that there is a strong association between gut microbiota and DN12. The imbalance of gut microbiota, changes in the abundance of specific flora and the role of metabolites of gut microbiota have important effects on the occurrence and development of DN13,14,15. Therefore, a systematic study of the causal relationship between gut microbiota and DN as well as the mechanism of intestinal flora disturbance leading to DN is very important for the prevention and treatment of DN. Several recent studies have only explored the gut microbiota that influence the development of DN, without further investigation of the mechanisms behind it32,33. Within this study, a thorough two-sample MR analysis was executed, revealing a positive association between the augmentation of Catenibacterium and Parasutterella intestinal bacteria and the heightened risk of DN. Moreover, subsequent intermediate proteomic analysis was carried out to delve deeper into the mechanisms involved. The potential biological mechanism of Catenibacterium and Parasutterella intestinal bacteria causing DN has been explored.

The Catenibacterium genus belongs to the Erysipelotrichaceae family30, Parasutterella is a gram-negative, strictly anaerobic genus of micrococcus from the phylum Proteus. To date, the conventional epidemiological investigation of the correlation between these two genera and DN remains limited. The results of this study elucidate that an increased abundance of both bacterial genera in the gut is a risk factor for DN and is associated with an increased risk of DN. Studies have found that Catenibacterium is related to metabolic diseases and can lead to various metabolic disorders34. A meta-analysis study on intestinal bacteria related to obesity found that Catenibacterium is significantly higher in obese people than non-obese people, and intake of lactic acid bacteria can significantly reduce the number of the bacterium. It may be one of the treatments for obesity to regulate intestinal microflora through intake of lactic acid bacteria34,35. In a study investigating the relationship between microbiota and metabolic markers in African American men (AAM) with prediabetes and vitamin D deficiency, a high abundance of Catenibacterium was found, and the number of the genus changed with worsening blood sugar36. Furthermore, a rise in the abundance of this particular genus was observed in stool samples obtained from individuals with ESRD, thereby indicating a plausible association between the levels of this genus and the progression of kidney disease37. Parasutterella plays an important role in the development of type 2 diabetes and obesity. In a transformational human study, Parasutterella abundance was found to be positively correlated with BMI and type 2 diabetes, and positively correlated with dietary carbohydrate intake, but not with reduced microbiome α/β diversity and low inflammation in obesity. The association with L-cysteine may be related to the development of type 2 diabetes. And has been linked to fatty acid biosynthetic pathways that lead to weight gain from carbohydrate-rich diets during the development of obesity38. In an animal study, Lactobacillus paracasei NL41 was found to reduce the number of Parasutterella by decreasing serum lipopolysaccharide (LPS), free fatty acid (FFA), tumour necrosis factor-α (TNF-α), interleukin (IL)−6, and IL-8, and by increasing the levels of IL-10 modulates the gut microbiota and improves inflammation in type 2 diabetic rats39. During the investigation of renal fibrosis, a positive correlation was discovered between the relative abundance of Parasutterella and the disease’s progression. Additionally, the augmentation of Parasutterella appears to facilitate the onset and advancement of renal fibrosis caused by unilateral ureteral occlusion (UUO)40. The increased abundance of Parasutterella in UUO mice is consistent with earlier findings in CKD patients41, CKD rats42, and is associated with chronic intestinal inflammation43. Within a two-sample MR investigation examining the relationship between gut microbiota, CKD, and chronic systemic inflammation, a noteworthy discovery emerged. The heightened abundance of Parasutterella was identified as a risk factor for proteinuria, exhibiting a positive association with elevated levels of the urinary albumin: creatinine ratio (UACR)44. This finding aligns with earlier studies wherein disruptions in intestinal flora, exemplified by increased Parasutterella abundance, were positively linked to escalated proteinuria in rat models of CKD42. The mechanism may involve a enteric-borne uremic toxin, phenyl-sulfate (PS)45, which is also one of the causes of DN albuminuria46, which can increase albuminuria by damaging podocytes47. The association between phenyl sulfate (PS) levels and the urinary albumin: creatinine ratio (UACR) was notably significant among individuals with microalbuminuria. Furthermore, among patients with microalbuminuria, PS emerged as the sole predictive factor for UACR progression over a two-year period48. Therefore, intervention on Catenibacterium and Parasutterella intestinal bacteria has the potential to become a new way to prevent and treat DN.

Many circulating proteins are key regulators of molecular pathways, so exploring their potential mediating effect in the association of gut microbiota and DN could help to understand the underlying mechanisms of intestinal flora disorder on DN. In this study, a total of 16 plasma proteins mediating between Catenibacterium and DN were identified, and the highest mediating effects was vinculin. It has been found that vinculin is an important protein in podocytes that is involved in the attachment of extracellular matrix to intracellular actin fibers49. Vinculin is able to recruit other proteins in podocytes, such as α-actinin and Arp2/3, thus participating in cell adhesion and force transmission processes50. In podocytes, vinculin plays an important role in maintaining the normal morphology of foot processes51,52, and this process may be affected by palladin. Palladin, a ubiquitously expressed actin-related protein, is a key component involved in cytoskeleton construction in cells. Vinculin expression is significantly downregulated when palladin is inhibited, and palladin may affect podocyte adhesion and extracellular signaling by regulating vinculin interactions with these proteins53,54. DN is associated with podocyte injury and dysfunction55. It has been found that the induction of aptamers and cytoskeletal proteins such as vinculin in podocytes is inhibited under diabetic conditions, the levels of focal adhesion components are decreased in DN, however, phosphorylated vinculin is upregulated under hyperglycemic conditions, and podocyte adhesion to the basement membrane is impaired. DN is also associated with the activation of the Rho-ROCK signaling pathway, which is involved in the regulation of the cytoskeleton, and its inhibition prevents proteinuria and renal failure. Increased GSK3β induction and decreased vinculin induction under diabetic conditions suggest that GSK3β stimulates disordered adhesion and remodeling of the cytoskeleton in podocytes through the Rho A-GSK3β pathway56. Vinculin expression is increased in patients with DN who have type 1 diabetes57. Consistent with our study, the increased abundance of Catenibacterium increases the plasma protein vinculin, which contributes to the increased risk of DN. In addition, vinculin may be involved in the regulation of renal inflammation through several pathways, including increased vinculin strengthening the connection between renal tubular epithelial cells and interstitial cells, thereby driving renal interstitial inflammation58. Cyclosporine A can cause an increase in vinculin expression in the mouse kidney, and this increase is positively correlated with the severity of cyclosporine A-induced kidney injury59. By establishing a 3D in vitro tissue model, Cha et al. found that intracellular and extracellular fibronectin can regulate the polarization state of macrophages through vinculin, thereby affecting the inflammatory response60. Porst et al. found that cellulose-1 can affect the function of mesangial cells by regulating the attachment, spreading, and proliferation of interstitial cells, and vinculin is involved in this process61. Bains et al. quantitatively showed increased expression of cell-junctional proteins such as vinculin in glomerular cells in cases of proteinuria, suggesting their involvement in renal pathology62. Oxidative stress can also lead to renal cell injury and apoptosis, and vinculin, a member of the cell-matrix junction component, may also be involved in oxidative stress-induced apoptosis because of its role in maintaining cytoskeletal structure63. Studies have shown that the intercellular junction protein vinculin can promote the development of renal fibrosis by regulating the EMT process of renal tubular epithelial cells, and the mechanism involves the activation of p38 signaling pathway induced by acetic acid58, the action of angiotensin II (Ang II)64, and T cell death-associated genes 51 (TDAG51) on vinculin regulation65and other levels. As a pivotal cytoskeletal protein, vinculin plays an indispensable role in maintaining cellular morphological and structural integrity. Dysregulation of vinculin’s biological function may be intricately associated with inflammatory responses, apoptosis, oxidative stress-induced injury, and renal fibrosis in DN. Existing research has demonstrated that vinculin represents a promising therapeutic target for conditions such as age-related heart failure, impaired angiogenesis due to dysfunctional high-density lipoprotein (dHDL), atherosclerosis, and other diseases66,67,68. This insight offers a novel research avenue for elucidating the molecular mechanisms underlying DN and holds potential as a significant therapeutic target for DN.

A total of 6 plasma proteins mediating between Parasutterella and DN were identified, and the highest mediating effects was Multiple PDZ domain protein (MPDZ also named MUPP1). MUPP1 is an osmotic response protein in kidney cells. Under hypertonic conditions induced by sucrose and mannitol, the tight junction protein MUPP1 is significantly up-regulated and plays an important role in the osmotic stress response of renal cells69. This up-regulation was found to be necessary for the maintenance of tight epithelial properties in these cells. In line with our study, Parasutterella increased plasma MUPP1 abundance in hyperosmolar hyperglycemia, which contributed to the increased risk of DN. The role of MUPP1 in the kidney may be mediated by its interaction with Cldn4, MUPP1 contributes to the maintenance of a tight epithelium in the medulla of the kidney under hypertonic stress by correctly localizing Cldn4 to the tight junctions70. MUPP1 could interact with renal tubule-associated K (+) channel Kir4.2, and the interaction played an important role in the regulation of K (+) transport in renal tubules. The study found that, MUPP1 can bind to Kir4.2 and reduce its cell membrane expression. Thus MUPP1 and Kir4.2 may participate in a protein complex in the nephron that could regulate transport of K (+) as well as other ions71. In addition, Hepatocyte growth factor activator inhibitor-1 (HAI-1) is a Kunitz-type serine protease inhibitor. HAI-1 led to apoptosis through a reduction in the levels of MUPP1 in cervical cancer cell lines72. MUPP1 plays a crucial role in establishing cell polarity, stabilizing gap junctions, and maintaining its physiological functions in the kidney by regulating the localization and function of various associated proteins. Consequently, therapeutic strategies targeting MUPP1 may hold potential value in treating or delaying the progression of DN through the restoration of intercellular junctions, repair of the glomerular filtration barrier, and anti-inflammatory and anti-fibrotic effects. However, current research on MUPP1 in DN remains in its preliminary stages, with its specific mechanisms of action yet to be fully elucidated. Further in-depth studies are urgently required to validate its clinical application potential. Investigating the mechanisms of MUPP1 in DN will facilitate its clinical translation as a therapeutic target and holds significant research and clinical application value.

In this study, we conducted an enrichment analysis of plasma proteins associated with intestinal microflora in DN patients. Our findings suggest that 11 pathways may mediate the effects of Catenibacterium on DN, including the ATF6 (ATF6-alpha) activated chaperone genes pathway. The ATF6 pathway is a critical component of the endoplasmic reticulum (ER) stress response mechanism. Previous research has established a close association between ER stress and the development of DN73. Over-activation of the ATF6 pathway may impair renal cell function and survival through the following mechanisms: (1) Exacerbation of inflammatory responses: Zhu et al. demonstrated that (+)-catechin reduces the activation of ER stress-related NLRP3 inflammasome by downregulating ATF6 and other ER stress-related factors, thereby alleviating DN injury74. (2) Induction of apoptosis: Studies have shown that lipotoxicity-induced protein arginine methyltransferase 1 (PRMT1) increases mesangial cell apoptosis and promotes DN progression by activating the PERK and ATF6-mediated ER stress pathway75. Additionally, ursodeoxycholic acid (UDCA) and 4-phenylbutyric acid (4-PBA) mitigated renal lesions in diabetic mice by reducing macrophage injury and renal tubular epithelial cell apoptosis, achieved through inhibition of the ATF6 signaling pathway and prevention of ER stress activation76. Given its involvement in DN pathogenesis, inhibiting ATF6 activity may help alleviate renal injury77. In this study, enrichment analysis revealed association between Catenibacterium and ATF6 pathway activation, suggesting that Catenibacterium may play a crucial role in DN development via ATF6 pathway activation. Therefore, inhibiting Catenibacterium could offer a novel therapeutic strategy for DN.

In the enrichment analysis conducted in this study, we identified that Parasutterella influences DN via 16 pathways, including pyruvate metabolism and the citric acid cycle (TCA cycle). Pyruvate, as an end product of glycolysis, plays a pivotal role in cellular energy metabolism. The onset and progression of DN are closely associated with impaired glucose metabolism. Research has demonstrated that the enzymatic activity of the glycolytic pathway in kidney tissues of DN patients is diminished, leading to reduced pyruvate production. Concurrently, the enzymatic activity of the TCA cycle is also suppressed, hindering the oxidative metabolism of pyruvate. These abnormalities in glucose metabolism result in inadequate energy supply to renal cells, triggering pathological processes such as cell dysfunction, inflammatory responses, and apoptosis, ultimately promoting the development and progression of DN78. Pyruvate kinase M2 (PKM2), a key enzyme in pyruvate metabolism, exhibits altered expression and activity in DN, potentially contributing to its pathogenesis through the following mechanisms: (1) Increased glycolytic flux: PKM2 activation can enhance the glycolytic pathway, reducing intracellular accumulation of free glucose and other metabolites, thereby mitigating high glucose toxicity and inhibiting glomerular pathology79; (2) Maintenance of mitochondrial function: PKM2 expression and activity can influence mitochondrial morphology and function, increasing ATP production and sustaining energy metabolism and normal kidney cell function80; (3) Improvement of glomerular function and structure: PKM2 activation can reduce glomerular hypertrophy, improve glomerular structure, and increase glomerular filtration rate, thus delaying renal function decline81. In this study, Parasutterella was identified as being associated with the pyruvate metabolism pathway, indicating its potential role in the pathogenesis of DN through the modulation of pyruvate metabolism. Intervention in this pathway may offer a novel therapeutic strategy for the treatment of DN82.

By employing rigorous criteria, our two-sample MR analysis offers compelling insights into the complex interplay between gut microbiota and DN. The findings suggest that an augmentation in the abundance of intestinal bacteria, specifically Catenibacterium and Parasutterella, may heighten the risk of developing DN. Furthermore, our mediation analysis revealed the important role of plasma proteome as mediators, depicting the causal trajectories from the abundance of gut microbiota increases to the progression of DN, which has profound significance in constructing a blueprint from the abundance of gut microbiota in vivo to the pathogenesis of DN, thereby effectively guiding the prevention of DN and delaying its progression.

However, it is crucial to acknowledge certain limitations in our study. First, the genetic analyses in this study were performed on individuals of European ancestry, but the fact that the gut microbiota varies significantly across geographic and ethnographic groups challenges the generalizability of our finding83. For example, Asian populations consume more fermented foods such as yoghurt and kimchi, which contributes to an increase in the abundance of lactic acid bacteria (e.g. Lactobacillus and Bifidobacterium) in the gut. Whereas in Europe, although fermented foods are also consumed, they are usually dominated by cheese and beer, which are more likely to affect changes in Lactobacillus casei, and Saccharomyces cerevisiae. This diversity means that the relationship between gut microbiota and DN that we obtained based on European populations may be limited when generalized to other populations, suggesting that future studies should consider diverse sample sources to explore the characteristics of gut microbiota and their impact on DN in different ethnic and geographical contexts. Second, our data were derived from publicly available aggregated statistics and the raw clinical outcome data for each individual were not available, thus preventing further population stratification analyses. Lastly, we did not perform experimental validations for the identified mediating proteins and potential mechanisms by which gut microbiota affects DN.

Conclusions

Through our extensive MR analysis, a multifaceted association between gut microbiota and DN has been uncovered. Remarkably, we have identified the augmentation of Catenibacterium and Parasutterella intestinal bacteria as causal contributors to the development of DN. More importantly, we reveal the underlying mechanism by which the increased abundance of Catenibacterium and Parasutterella intestinal bacteria lead to DN, providing a blueprint for the involvement of gut–kidney axis in the pathogenesis of DN and paving the way for future studies.