Gut microbiota differences in chronic kidney disease patients with and without diabetes

To investigate whether gut microbiota is involved in the regulation of diabetes, we collected fecal samples from patients with diabetic kidney failure (DKD) and age-matched non-diabetic kidney disease patients (NDKD) for 16S rRNA sequencing. The analysis revealed 250 specific OTUs in the NDKD group, 32 unique OTUs in the DKD group, and 384 OTUs shared between both groups (Fig. 1A). Please note that the denominator of the percentage is the total sequencing count. Although 32 new OTUs were detected in the DKD sample group, their biomass proportion is negligible. Of course, the 250 OTUs reduced in DKD compared to NDKD are also not dominant bacteria, collectively accounting for only 0.7% of the total microbial count (more precisely,"their genetic material only accounts for 0.7% of the total genetic material"). The shared 384 OTUs constitute the overwhelming majority in terms of biomass (genetic material). In terms of OTU count, DKD patients exhibited a one-third reduction in OTUs compared to NDKD patients, indicating a significantly lower gut microbiota abundance in the DKD group. As well as the Chao1 index, an alpha diversity index, revealed a significant decrease in community richness and evenness between the groups, indicating an alteration in microbial alpha diversity due to diabetes (Fig. 1B). At the phylum level, both NDKD and DKD groups were predominantly composed of Firmicutes, Bacteroidota, Proteobacteria, and Fusobacteriota, with no significant differences in phylum composition but differences in microbial abundance. At the class level, the dominant groups were Fusobacteriia, Bacteroidia, and Clostridia. At the order level, Bacteroidales, Fusobacteriales, and Enterobacteriales were predominant. At the family level, Fusobacteriaceae, Enterobacteriaceae, and Bacteroidaceae were identified as the dominant taxa. While the overall microbial composition did not exhibit significant differences between the NDKD and DKD groups at the phylum, class, order, or family levels, notable variations in microbial abundance were detected (Supplementary Fig. 1A-D). Phylum-level analysis further revealed that diabetes was associated with a reduction in Bacteroidota and Proteobacteria, whereas the proportions of Firmicutes and Fusobacteria were elevated (Fig. 1C). Differential analysis revealed significant differences in the relative abundance of Firmicutes and Proteobacteria between the DKD and NDKD groups (Fig. 1D). These findings suggest that, although the dominant microbial species at the phylum, class, order, and family levels did not significantly differ between the two groups, microbial abundance exhibited notable variation.

Fig. 1

Intestinal microbiota between NDKD and DKD groups. A Venn diagram of gut microbial species between NDKD and DKD groups. Numbers represent OTU counts, with percentages in parentheses indicating the sequencing number relative to the total sequencing number. B α-diversity of the microbial community. C Community composition at the phylum level for NDKD and DKD groups. D Differential analysis results of different bacteria between NDKD and DKD groups. Results are presented as LDA score

Differential metabolites in the NDKD and DKD groups

Serum metabolites were used to analyze the differential metabolites between NDKD and DKD groups, and the screening thresholds were set at VIP > 1 and P < 0.05. The results indicated significant differences in both positive and negative ion metabolites between the two groups (Table 1).

Table 1 Metabolites between NDKD and DKD groups showed a total of 10 metabolites

A total of 15 significantly altered metabolites were identified in negative ion metabolism, with 6 downregulated and 9 upregulated (Fig. 2A, Left panel). In positive ion metabolism, 5 metabolites exhibited significant differences, among which 2 were downregulated and 3 were upregulated (Fig. 2B, Left panel). Further classification of metabolites based on compound types revealed notable changes in superclass metabolites, primarily including benzene derivatives, organic acids and their derivatives, organic oxides, organic heterocyclic compounds, and organic nitrogen compounds (Fig. 2A and B, Right panel). These metabolites include dihydroxyacetone, D-fructose, D-tagatose, pyruvaldehyde, L-gulonic gamma-lactone, m-chlorohippuric acid, 3-phosphoserine, confertifoline, 3-methoxy-4-hydroxyphenylglycol sulfate, L-threonate, 2E-eicosenoic acid, p-cresol, N-acetyl-D-glucosamine, myo-Inositol, 9(S)-HODE, D-mannose, guanidoacetic acid, 1-palmitoyl-sn-glycero-3-phosphocholine, TMAO, and DL-indole-3-lactic acid (Table 1). Among above metabolic components, DKD and NDKD exhibited significant differences in fold change and VIP values for D-fructose, 3-methoxy-4-hydroxyphenylglycol sulfate, and 1-palmitoyl-sn-glycero-3-phosphocholine, potentially serving as biomarkers.

Fig. 2

Differential metabolite analysis between NDKD and DKD groups. A Significant differential analysis of metabolite expression in positive-ion mode, (Left panel) ranked by fold change, (Right panel) categorized by superclass. Blue bars indicate downregulation, while red bars signify upregulation. The color of the molecular name represents the type of compound, as referenced in the legends on the right. B Significant differential analysis of metabolite expression in negative-ion mode, (Left panel) ranked by fold change, (Right panel) categorized by superclass. Blue bars indicate downregulation, while red bars signify upregulation. The color of the molecular name represents the type of compound, as referenced in the legends on the right. C The correlation heatmap for all differential metabolites, with color intensity indicating the strength of the correlation: red for positive correlation and blue for negative correlation

Subsequently, correlations were determined between 15 positive ion metabolites and 5 negative ion metabolites (Fig. 2C, Left panel). Among the positive ion metabolites, p-Cresol, 2E-eicosenoic acid, N-acetyl-D-glucosamine, L-gulonic gamma-lactone, m-chlorohippuric acid, dihydroxyacetone, D-fructose, D-tagatose, and pyruvaldehyde, as well as 9(S)-HODE, 3-methoxy-4-hydroxyphenylglycol sulfate, myo-inositol, 3-phosphoserine, and L-threonate, showed positive correlations. Especially dihydroxyacetone, D-fructose, D-tagatose, and pyruvaldehyde showed a strong positive correlation (Fig. 2C, Left panel). In the realm of anionic metabolites, guanidoacetic acid and 1-palmitoyl-sn-glycero-3-phosphocholine, as well as D-mannose and TMAO, exhibited positive correlations (Fig. 2C, Right panel).

DEGs analysis of DKD

The limma package was employed for differential expression analysis of the publicly available dataset GSE166239 (Nordbø et al. 2023). The DEGs between the HN and T2DN patients were visualized using a heatmap (Fig. 3A), and their expression levels were displayed in a volcano plot (Fig. 3B). Enrichment analysis of the DEGs showed that they are related to amino acid metabolism, redox enzyme activity, cell activation, and the extracellular matrix (Fig. 3C).

Fig. 3

Differential expression analysis and functional enrichment analysis. A Heatmap showing DEGs between hypertensive nephrosclerosis and diabetic nephropathy patients. B Volcano map showing DEGs. C Differential gene enrichment analysis. DEGs were subjected to enrichment analysis according to Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO). GO functions included biological process (BP), cellular component (CC), and molecular function (MF)

The serum differential metabolite TMAO targets the gene PRSS3

Based on the differential bacteria identified by 16S rRNA sequencing and the gutMGene database, we utilized non-targeted metabolomics to analyze metabolites influenced by these microbial communities. Six potential metabolites most affected by the differential bacteria were identified: phenylacetylglutamine, indoxyl sulfate, p-cresol sulfate, trimethylamine oxide, 3-indolepropionic acid, and p-cresol glucuronide (Fig. 4A). These metabolites include common uremic toxins. The relationship between the differential serum metabolites and their target genes indicated that these metabolites were not influenced by a single gene (Fig. 4B). A Venn diagram comparing differential metabolites from metabolomics and those from 16S differential bacteria revealed that only TMAO was present in the intersection. Further analysis of DEGs associated with diabetic nephropathy and TMAO target genes identified one gene, PRSS3 (Fig. 4C). Target gene analysis showed that the genes influencing TMAO included FMO3, CRYAA, EBF1, HSPA8, HSPA2, RNASE1, HSPA4, and PRSS3 (Fig. 4D). A significant increase in gut microbiota led to a notable increase in TMAO, which subsequently inhibited PRSS3 expression (Figs. 4E). These results suggest that the serum differential metabolite TMAO specifically targets the differential gene PRSS3.

Fig. 4

Cross-analysis of differential metabolites and their associated genes, along with differentially expressed genes, in patients with DKD revealed TMAO as a specific target of PRSS3

A Sankey diagram depicting associations between specific microbial taxa and metabolites. Microbial genus (Left): Lachnospiraceae: GCA-900066755, Hungatella, Eisenbergiella; Oscillospiraceae: Oscillibacter; Ruminococcaceae: Harryflintia, Candidatus_Soleaferrea, Anaerotruncus, UBA1819. Metabolites (Right): phenylacetylglutamine, indoxyl sulfate, p-cresol sulfate, trimethylamine oxide, 3-indolepropionic acid, p-cresol glucuronide. Colored lines represent associations between specific microbial genus and metabolites. B Circos plot depicting associations between specific genes and metabolites. Colored lines represent associations between specific genes and metabolites. C Venn diagram of differential metabolites and 16S differential microbial genus, DEGs of diabetic nephropathy, and TMAO target genes. Intersection of differential metabolites in metabolomics (green) with corresponding metabolites of differential bacteria in 16S analysis (blue) (left); intersection of diabetic nephropathy DEGs (green) with TMAO target genes (blue) showing the common gene PRSS3 (right). In each Venn diagram, the overlapping region indicates the common elements between the two datasets, with the quantity highlighted in the red box. The bar charts below display the total number of elements in each dataset and the number of shared elements between the two datasets. D Relationship between diabetic nephropathy-associated DEGs and TMAO target genes. The figure illustrates eight TMAO target genes (FMO3, CRYAA, EBF1, HSPA8, HSPA2, RNASE1, HSPA4, PRSS3), which are also DEGs associated with diabetic nephropathy. The gene names are listed on the left, and the bar chart on the right indicates that all eight genes are related to TMAO. The colored bands represent connection pathways between each gene and TMAO. E Connections between differential bacteria, metabolites, and target genes. The figure illustrates how four differential bacteria (Lachnospiraceae|GCA-900066755, Lachnospiraceae|Hungatella, Lachnospiraceae|Eisenbergiella, Oscillospiraceae|Oscillibacter) influence the downstream target gene PRSS3 through the metabolite TMAO

Molecular docking analysis of TMAO with PRSS3

The molecular docking analysis between TMAO and PRSS3 yielded a docking score of −2.7, indicating a moderate binding affinity. The study identified several key residues in Chain A of PRSS3 that interact with TMAO. These residues include Phe104, Cys105, His120, Asp251, Ser252, Cys253, Gln254, Arg255, Asp256, Ser257, Val271, Ser272, Trp273, Gly274, Gly276, Cys277, Gly284, Val285, and Tyr286 (Fig.5). These interactions suggest that TMAO binds to PRSS3 at multiple sites, potentially affecting protein stability and influencing its function.

Fig. 5

Molecular docking of TMAO with PRSS3. Molecular docking result between TMAO and PRSS3 protein

High glucose (HG) stress downregulates PRSS3 expression in renal cells

Based on bioinformatic analyses, PRSS3 was identified as a DEG in DKD, suggesting that its expression might be responsive to HG induction. To investigate this, we treated HK-2 cells with HG found that PRSS3 expression was significantly downregulated compared to the control group (Fig. 6A-B). CFDA staining of live cells indicated a significant decline in HK-2 cell viability under HG conditions (Fig. 6C).

Fig. 6

HG environment impairs PRSS3 expression in vitro and in vivo. A Expression of PRSS3 after HG treatment. B Immunofluorescence staining of PRSS3 in HK-2 cells. C CFDA staining indicating the viability of HK-2 cells after HG treatment. D H&E and MASSON staining of kidneys from control and DKD mice. E-I Renal function indicators in control and DKD mice: blood glucose levels, serum urinary creatinine, serum BUN, urinary protein levels, and urinary albumin excretion rate. J Levels of inflammatory factors in kidneys of control and DKD mice: IL-1β, IL-6, IL-18, and TNF-α. K Expression of PRSS3 in kidneys of control and DKD mice. **p < 0.01, * * * p < 0.001

To explore these findings in vivo, streptozotocin (STZ) was employed to establish a DKD animal model. Hematoxylin and eosin (H&E) staining revealed that renal tissues from DKD mice exhibited more fissures compared to healthy controls, indicating the onset of fibrosis. Masson's trichrome staining showed a broader distribution of Masson-positive areas in the renal tissues of DKD mice, suggesting the presence of interstitial fibrosis (Fig. 6D). Biochemical analysis of blood samples demonstrated significant elevations in blood glucose, serum creatinine, and blood urea nitrogen levels. Urine samples showed markedly increased excretion rates of urinary protein and albumin (Fig. 6E-I), confirming that STZ treatment successfully induced DKD symptoms. Inflammatory cytokines, including interleukin-1β (IL-1β), IL-6, IL-18, and tumor necrosis factor-α (TNF-α), were markedly elevated in tissue samples from DKD mice (Fig. 6J). Western blot analysis further confirmed a substantial reduction in PRSS3 protein expression levels in the DKD animal model (Fig. 6K). These findings collectively imply that PRSS3 expression is positively correlated with renal cell health and normal renal morphology and function.

TMAO treatment of HK-2 cells leads to downregulation of PRSS3 expression

Bioinformatic analyses indicated that alterations in the gut microbiota of diabetic patients ultimately target PRSS3 via the microbial metabolite TMAO, exacerbating renal injury. To determine whether the downregulation of PRSS3 is regulated by TMAO, we treated cultured renal proximal tubular epithelial cells with TMAO. The treatment led to a significant reduction in PRSS3 expression levels in HK-2 cells (Fig. 7A-B). Additionally, TMAO inhibited cell viability (Fig. 7C).

Fig. 7

TMAO treatment disrupts PRSS3 expression in vitro and in vivo. A Expression of PRSS3 after TMAO treatment. B CFDA staining indicating the viability of HK-2 cells after TMAO treatment. C Immunofluorescence staining of PRSS3 in HK-2 cells. D H&E and MASSON staining of kidneys from control and TMAO-treated mice. E-I Renal function indicators in control and TMAO-treated mice: blood glucose levels, serum urinary creatinine, serum BUN, urinary protein levels, and urinary albumin excretion rate. J Levels of inflammatory factors in kidneys of control and TMAO-treated mice: IL-1β, IL-6, IL-18, and TNF-α. K Expression of PRSS3, KIM-1, and NGAL in kidneys of control and TMAO-treated mice. **p < 0.01, * * * p < 0.001

Earlier research has documented that oral administration of TMA, the precursor of TMAO, induces renal injury in rodents (Maksymiuk et al. 2022). Consistently, in our study, administration of TMAO to mice also resulted in renal injury. H&E and Masson's trichrome staining both displayed fibrotic phenotypes similar to those observed in STZ-treated mice (Fig. 7D). Blood biochemical analysis demonstrated markedly elevated levels of blood glucose, serum creatinine, and blood urea nitrogen. Urine samples exhibited markedly elevated excretion rates of urinary protein and albumin (Fig. 7E-I). These indicators suggest that TMAO treatment induced symptoms characteristic of DKD. A significant elevation in the levels of inflammatory cytokines was observed in tissue samples (Fig. 7J). Western blot analysis revealed that TMAO treatment significantly reduced PRSS3 protein levels in mouse kidneys, while markedly increasing the expression of KIM-1 and NGAL, markers of early renal injury (Fig. 7K). These findings confirm that TMAO is toxic to renal cells both in vitro and in vivo. And endogenous TMAO derived from the gut microbiota may exacerbate renal injury in diabetic patients.

PRSS3 antagonizes TMAO-induced renal injury

In previous experiments, we observed that PRSS3 expression was downregulated in renal cells by either HG or TMAO, indicating a general positive correlation between PRSS3 expression and renal cell health. To determine whether PRSS3 merely serves as an indicator of renal cell health or plays an essential active role in its maintenance, we examined the correlation between PRSS3 expression and renal cell viability. We overexpressed PRSS3 in HK-2 cells under HG conditions and added TMAO to simulate patients with altered gut microbiota (Fig. 8A-B). Western blot analysis of KIM-1 and NGAL revealed that PRSS3 overexpression effectively mitigated HG-induced damage in HK-2 cells, while TMAO treatment counteracted the protective effects of PRSS3, suggesting opposing roles between TMAO and PRSS3 (Fig. 8A). Cell viability assays showed that PRSS3 overexpression improved the viability of HK-2 cells under HG conditions (Fig. 8C). Although TMAO treatment downregulated PRSS3 expression and attenuated its protective effect on cell viability, the viability remained higher than in the control group without PRSS3 overexpression. This supports the potential of PRSS3 overexpression as a therapeutic strategy to counteract TMAO-induced cellular damage.

Fig. 8

PRSS3 overexpression mitigates HG- and TMAO-induced renal cell damage in vitro and in vivo. A Expression of PRSS3, KIM-1 and NGAL in HK-2 cells. B CFDA staining indicating the viability of HK-2 cells. C Immunofluorescence staining of PRSS3 in HK-2 cells. D H&E and MASSON staining of kidneys from PRSS3-overexpression (AAV-PRSS3) and TMAO-treated mice. E-I Renal function indicators in PRSS3-overexpression and TMAO-treated mice: blood glucose levels, serum urinary creatinine, serum BUN, urinary protein levels, and urinary albumin excretion rate. J Levels of inflammatory factors in kidneys of PRSS3-overexpression and TMAO-treated mice: IL-1β, IL-6, IL-18, and TNF-α. K Expression of PRSS3, KIM-1, and NGAL in kidneys of PRSS3-overexpression and TMAO-treated mice. *p < 0.05, **p < 0.01, * * * p < 0.001

Subsequently, we overexpressed PRSS3 in DKD mice via adeno-associated virus (AAV)-mediated gene delivery. Histological analysis revealed that PRSS3 overexpression significantly reduced fibrosis in renal sections (Fig. 8D). Given the pro-fibrotic effect of TMAO observed in previous experiments and the opposing physiological effects of TMAO and PRSS3 in nephropathy, we directly tested this antagonistic relationship by administering TMAO to DKD mice receiving PRSS3 overexpression therapy. The results showed that TMAO negated the ameliorative effects of PRSS3 overexpression in DKD mice (Fig. 8D). Functionally, PRSS3 overexpression improved renal indicators in blood and urine samples of DKD mice (Fig. 8E-I). Notably, even in the TMAO-treated group, which theoretically should exhibit worsened indicators, the parameters remained better than those in the control group, indicating a significant protective effect of PRSS3 overexpression. Similarly, in terms of inflammatory markers, although TMAO treatment impaired the improvement conferred by PRSS3, their levels were still notably reduced relative to the control group (Fig. 8J). Measurements of KIM-1 and NGAL in mouse kidneys yielded results consistent with those observed in cultured cells (Fig. 8K). This may be attributed to the rapid response characteristics of KIM-1 and NGAL as injury reactants, which are sensitive to cellular damage but may not linearly reflect the antagonistic effect of PRSS3 on TMAO in this experiment.