Integration of transcriptome and Mendelian randomization analyses in exploring the extracellular vesicle-related biomarkers of diabetic kidney disease

Abstract

Background

Diabetic Kidney Disease (DKD) is a common complication in patients with diabetes, and its pathogenesis remains incompletely understood. Recent studies have suggested that extracellular vesicles (EVs) may play a significant role in the initiation and progression of DKD. This study aimed to identify biomarkers associated with EVs in DKD through bioinformatics and Mendelian randomization (MR) analysis.

Methods

This study utilized two DKD-related datasets, GSE96804 and GSE30528, alongside 121 exosome-related genes (ERGs) and 200 inflammation-related genes (IRGs). Differential analysis, co-expression network construction, and MR analysis were conducted to identify candidate genes. Machine learning techniques and expression validation were then employed to determine biomarkers. Finally, the potential mechanisms of action of these biomarkers were explored through Immunohistochemistry (IHC) staining, enrichment analysis, immune infiltration analysis, and regulatory network construction.

Results

A total of 22 candidate genes were identified as causally linked to DKD. CMAS and RGS10 were identified as biomarkers, with both showing reduced expression in DKD. IHC confirmed low RGS10 expression, providing new insights into DKD management. CMAS was involved primarily in mitochondria-related pathways, while RGS10 was enriched in the extracellular matrix and associated pathways. Significant differences were observed in neutrophils and M2 macrophages between DKD and normal groups, correlating strongly with the biomarkers.

Conclusion

This study identified two EV-associated biomarkers, CMAS and RGS10, linked to DKD and elucidated their potential roles in disease progression. These results offer valuable insights for further exploration of DKD pathogenesis and the development of new therapeutic targets.

Introduction

Diabetic Kidney Disease (DKD) is a complication related to diabetes that affects approximately 40% of patients with type 2 diabetes (T2D) and 30% with type 1 diabetes (T1D) [1,2]. As the global diabetic population increases, so does the prevalence of DKD. In China, DKD affects between 29.6% and 49.6% of diabetics [3], placing a significant economic burden on both the state and individuals. Currently, clinical practice lacks reliable molecular targets for DKD diagnosis and effective treatment. Although albuminuria is commonly used to assess renal injury and assist in diabetes diagnosis [4], it is not specific to DKD [5]. In practice, significant kidney impairment may be present in some T2D patients despite normal albuminuria levels, highlighting the limitations of albuminuria as a diagnostic tool for DKD [6]. Treatment of DKD remains centered on controlling blood glucose, blood pressure, and lipid levels. Comprehensive management includes renin-angiotensin system blockade and rigorous control of glycemic levels, lipid profiles, and blood pressure. Despite these efforts, the incidence of DKD patients progressing to end-stage kidney disease (ESKD) and requiring renal replacement therapy has continued to rise [7]. The identification of biomarkers offers potential for early DKD diagnosis. Timely interventions can effectively slow DKD progression and prolong patient life expectancy. A study analyzing urine samples from T2D patients categorized into normoalbuminuria, microalbuminuria, and macroalbuminuria, as well as from healthy individuals, successfully identified potential biomarkers including CSPG 4, PLAU, SERPINA3, and ALB [8]. Early interventions were shown to slow renal function decline and improve patient outcomes.

Extracellular vesicles (EVs) are small spherical packages released into the extracellular environment by a variety of cells [9]. They serve as a crucial medium for communication between donor and receptor cells, encompassing exosomes and microvesicles (MV) [10]. Under stress conditions such as hypoxia [11], acidic pH [12], uremic toxin [13], and high sugar levels [14], the secretion of EVs increases and can induce cytokine release. This promotes the aggregation of inflammatory cells, and EVs secreted by damaged kidney cells can transfer to other normal kidney cells, altering their phenotype and inducing intercellular crosstalk [15], thus emerging as a novel vector for cell-cell communication [16]. In the kidney, under high glucose conditions, glomerular endothelial cells (GEC) release TGF-β1 mRNA-rich EVs. These EVs regulate communication between GEC and podocytes, promote epithelial-mesenchymal transition (EMT), and cause podocyte injury, affecting renal function [17]. Exosomes, a subgroup of EVs approximately 40-100 nm in diameter [18], mediate intracellular communication by delivering cell-specific cargoes containing proteins, lipids, and genetic material to recipient cells [19]. In DKD patients, the level of Dll4 protein in urinary exosomes is elevated compared to controls [20], indicating that exosomes may play a significant role in the development of DKD Urinary exosomes (UEs) enrich low-abundance proteins, which are not contaminated with urinary sediment, thus enhancing the potential for accelerating the search for diagnostic and/or prognostic biomarker [21]. Multiple studies have identified specific miRNAs, mRNAs, and proteins in uEVs that can be useful in diagnosing diabetic kidney disease [22]. Several studies have profiled various miRNAs in different diabetic patient subgroups or glucose conditions. Jia et al. showed that the levels of miR-192, miR-194, and miR-215 are increased in type 2 diabetic patients with microalbuminuria compared to patients with normoalbuminuria and control subjects, but are decreased in patients with macroalbuminuria [23]. Prunotto et al. intriguingly identified nearly 1,200 proteins, including 14 previously undiscovered urinary proteins [24]. Subsequent gene ontology analysis categorized many of these proteins as brain-specific, and their expression in the kidney was later confirmed [24]. Consequently, urinary exosomes have emerged as a promising new avenue for identifying nephron segment-specific molecules and detecting disease-induced changes [24].

Identifying the expression of specific genes in diseases is crucial for understanding their mechanisms. To this end, bioinformatics software and databases such as GWAS, the Kyoto Genome and Genome Encyclopedia (KEGG) Enrichment Analysis, and Gene Aggregation Analysis have been developed [25]. Mendelian randomization (MR), which uses genetic variations as instrumental variables, is widely used in disease research to assess causal relationships between exposures and outcomes [26]. Therefore, MR was employed to identify EV-related genes and new biomarkers or mechanisms causally linked to DKD.

In this study, data from the Gene Expression Omnibus (GEO) and GWAS databases were used, considering exosome-related genes as exposure factors, with DKD as the outcome. Machine learning techniques identified DKD-related exosome biomarkers. The study identified 22 candidate genes potentially causally connected to DKD, with CMAS and RGS10 subsequently recognized as biomarkers. Overall, understanding the molecular targets of DKD, and the interplay between EVs, genes, and their mechanisms, may facilitate the development of effective diagnostic and therapeutic strategies for this debilitating disease.

Methods

Data acquisition

Gene expression profile datasets, comprising GSE96804 and GSE30528, were sourced from the GEO database (https://www.ncbi.nlm.nih.gov/). The training cohort (GSE96804) [27,28] included 41 DKD and 20 control glomerulus tissue samples, while the validation cohort (GSE30528) [29] consisted of 9 DKD and 13 control samples. A total of 121 exosome-related genes (ERGs) and 200 inflammation-related genes (IRGs) were derived from the literature [30] and the ‘HALLMARK_INFLAMMATORY_RESPONSE’ pathway in MSigDB (http://www.gsea-msigdb.org/), respectively. Expression quantitative trait loci (eQTL) data for candidate genes and GWAS data for DKD were obtained from the Integrative Epidemiology Unit (IEU) Open GWAS database (https://www.ebi.ac.uk/gwas/). The dataset ‘finn-b-DM_NEPHROPATHY_EXMORE’ contained 3,283 samples and 16,380,336 SNPs. Additional file 1 illustrated the flow of this study.

Differential expression analysis

The limma package (v 3.62.2) was used to identify differently expressed genes (DEGs) between DKD and control samples in the training dataset (|Log₂FC| > 0.5, p < 0.05) [31]. Volcano and heat maps were constructed using the ggplot2 package (v 3.4.3) and the ComplexHeatmap package (v 2.16.0), respectively [32,33].

Weighted gene co-expression network analysis (WGCNA)

Analysis was performed using the WGCNA package (v 1.72.1) in the training cohort [34]. Initially, samples were clustered to identify and eliminate any outliers. A soft threshold was then selected to ensure optimal alignment of gene interactions with a scale-free distribution. Gene similarity was calculated using neighbor-joining to generate a systematic clustering tree. The co-expression network was established by applying the hybrid dynamic tree cutting algorithm, with a minimum of 100 genes per gene module. Finally, modules were correlated with traits to identify key modules, and genes within these modules were designated as key module genes.

Identification and function analysis of candidate genes

Candidate genes related to exosomes were identified by intersecting DEGs with key module genes associated with DKD and exosomes. Gene ontology (GO) and kyoto encyclopedia of genes and genomes (KEGG) enrichment analyses were conducted using the clusterProfiler package (v 4.8.3) to explore the biological functions of these gene (p < 0.05), using the background gene set in the org.Hs.eg.db package (v 3.17.0) [35]. A visual indication of gene expression trends under specific GO terms is provided by the z-score, calculated with the following formula:

$$\mathit{\text{zscore}}=\frac{(\mathit{\text{up}}-\mathit{\text{down}})}{\sqrt{\mathit{\text{count}}}};$$

‘up’ and ‘down’ represent the number of genes classified as up-regulated (log₂FC > 0) and down-regulated (log₂FC < 0), respectively, within a GO term; ‘count’ indicates the total number of genes within that term. Moreover, a protein-protein interaction (PPI) network was constructed with a confidence score of 0.7 using the STRING database to examine interactions among proteins related to exosomes.

Mendelian randomization (MR) analysis

In pursuing biomarkers, MR analysis incorporated genes related to exosomes. Harmonization of effect alleles and sizes was performed using the harmonize data function within the TwoSampleMR package (v 0.5.7) [36]. Instrumental variables (IVs) were screened under conditions ensuring a strong association with the exposure, absence of confounders, and influence on the outcome solely through the exposure. The function ‘extract instruments’ of the TwoSampleMR package was utilized to screen SNPs with significant relationships to exposure (p < 5 × 10⁻⁶). Furthermore, IVs for linkage disequilibrium (LD) were removed (clump = TRUE, R²=0.001; kb = 10000). F-statistics of SNPs were calculated to exclude weak IVs (F > 10) and minimize the influence of confounding factors on MR results. These steps ensured that the IVs met the assumptions required for correlation analysis. Subsequently, MR analysis was conducted using five methods: MR Egger, Weighted median, Inverse variance weighted (IVW), Simple mode, and Weighted mode [37–40]. The IVW approach proved crucial in establishing causal relationships, identifying genes with a causal link to the outcome as feature genes (P of the IVW method was less than 0.05). Sensitivity analysis and the Steiger test were employed to assess the reliability of the MR analysis and to confirm that the screened SNPs met the three assumptions of Mendelian randomization. Finally, feature genes were analyzed in multivariate MR to determine their significance.

Machine learning

In this study, models were constructed using the XGBoost algorithm in the training set. Feature genes were ranked based on their gain scores from 100 iterations, and the top 10 models with the largest gains were selected as key genes1. These feature genes were then employed to build a random forest (RF) model in the training set using the RF algorithm’s train function. Subsequently, the DALEX package (v 2.4.3) was used to analyze the RF models, and the top 10 models were identified as key genes2.

Identification and function analysis of biomarkers

Key genes were determined by identifying the intersection of key gene 1 and key gene 2. Expression validation analyses were conducted in both the training and validation cohorts to identify key genes with consistent and significant expression trends, which were subsequently selected as biomarkers. These biomarkers were matched in the Vesiclepedia database (http://microvesicles.org/index.html) to confirm their association with exosomes. To investigate the role of biomarkers in diagnosing DKD, the rms package (v 6.7.1) was used in the training cohort to construct a nomogram model to predict the risk of developing DKD [41]. Additionally, the validity of the nomogram was verified by plotting calibration curves, receiver operating characteristic (ROC) curves, and decision curve analysis (DCA) curves. Single-gene gene set enrichment analysis (GSEA) was executed using the clusterProfiler package (v 4.8.3) to identify regulatory pathways and biological functions of the identified biomarkers [35]. The background gene set used for this analysis was org.Hs.eg.db. To calculate correlations, each biomarker was compared to other genes in the training set, and their correlation rankings with other genes were determined. GO and KEGG analyses were then executed to investigate significant pathways associated with the biomarkers (P.adj < 0.05, q < 0.2). Finally, The GeneMANIA database (http://genemania.org/) was utilized for predicting genes associated with biomarkers function and their relevant functions.

Immune infiltration analysis

To assess the difference in immune cell infiltration between DKD samples and normal control samples, the CIBERSORT algorithm and LM22 immune cell genome were employed. The infiltration abundance of 22 immune cells in the training set was calculated, and the Wilcoxon rank sum test was used to determine the differences in immune cell infiltration between the DKD and control groups. Additionally, the Spearman correlation between differential immune cells and biomarkers was investigated.

Inflammation index analysis and network construction

ssGSEA was conducted using IRGs in the training set with the GSVA package (v 1.48.3), and correlations were calculated between biomarkers and inflammation index scores [42]. Additionally, miRNAs of biomarkers were predicted using the TargetScan, miRDB, microT, and miRWalk databases, and miRNAs of 35,195 EVs were downloaded from the Vesiclepedia database (http://www.microvesicles.org/). miRNAs predicted from the five databases were intersected to obtain the shared miRNAs. The lncRNAs of the shared miRNAs were predicted using the StarBase database. Subsequently, a competing endogenous RNA (ceRNA) network was constructed. Searches were conducted in the DrugBank database for potential therapeutic agents targeting the biomarkers.

Immunohistochemistry (IHC) staining

Human kidney tissue from the Department of Nephrology at Chengdu University of Chinese Medicine was approved by the Ethics Committee. The validation test collected 8 samples, 4 from DKD patients and 4 from the blank group (II membranous nephropathy), with ages ranging from 45 to 69 years, equally distributed among males and females. All diagnoses were based on renal needle biopsy. Specific case data are provided as supplementary material. Following standard dewaxing and hydration, antigen retrieval was performed using citrate buffer (Beijing Zhongshan Jinqiao Biotechnology Co., LTD), and the samples were incubated with 3% H2O2 at room temperature for 10 min to inactivate endogenous peroxidase. Blocking was performed using bovine serum albumin V. The primary antibody was incubated overnight, followed by incubation with a universal secondary antibody for 30 min. Subsequently, 3,3′-Diaminobenzidine Tertrahydrochloride (DAB) reaction solution (Beijing Zhongshan Jinqiao Biotechnology Co., LTD) was applied for staining, and observations were made under a microscope. Hematoxylin (Wuhan Seville Biotechnology Co., LTD) was used for counterstaining for at least 15 min. The dehydration and sealing processes followed those of Nissl staining, with each step interspersed with three 5-min PBS rinses.

Statistical analysis

All analyses were conducted in R (v 4.2.3). Differences between groups were analyzed using the Wilcoxon test. p < 0.05 was considered statistically significant.

Results

Identification of 2,693 DEGs in the trainning set

A total of 2,693 DEGs were identified in the training set, comprising 1,290 up-regulated and 1,403 down-regulated DEGs (Figure 1a,b). Key module genes related to DKD and exosomes were identified using WGCNA. The results, with DKD and control grouping as traits, revealed no outlier samples in the training set and determined the optimal soft threshold to be 7 (R² = 0.85) (Figure 1c,d). A total of 9 modules were obtained. MEblack, MEpink, MEmagenta, MEbrown, MEtan and MEturquoise were highly correlated with DKD as key modules (|cor| > 0.3, p < 0.05), comprising 16,100 key module genes (Figure 1e,f). In addition, a total of 3 key modules (|cor|>0.3, p < 0.05) with high relevance to exosomes were obtained using the same method, including 25,832 key module genes (Additional file 2a–e).

Figure 1.

Identification of DEGs and construction of co-expression networks. (a,b) A total of 2,693 DEGs were identified in the training set. (c,d) Key module genes related to DN and exosomes were identified using WGCNA. Analysis of the results, revealed no outlier samples in the training set. (e,f) A total of 9 modules were obtained. MEblack, MEpink, MEmagenta, MEbrown, MEtan and MEturquoise were highly correlated with DN as key modules, which included 16,100 key module genes.

The 22 candidate genes were causally associated with DKD

In this study, 1,165 genes related to exosomes were acquired by intersecting DEGs and the key module genes related to DKD and exosomes (Figure 2a). Enrichment analysis showed that genes related to exosomes were primarily associated with extracellular matrix organization, extracellular structure organization, and external encapsulating structure organization, among others (Figure 2b). They were also enriched in KEGG pathways such as Focal adhesion, ECM-receptor interaction, and the AGE-RAGE signaling pathway in diabetic complications, etc. (Figure 2c). The PPI network revealed complex interactions between related to exosomes (Figure 2d).

Figure 2.

Identification of candidate genes, enrichment analysis and construction of PPI networks. (a) 1,165 candidate genes were obtained using the intersection of DEGs and key module genes associated with DN and exosomes, and (b) GO enrichment analysis. (c) The chord diagram of KEGG (top 6). (d) Expression of PPI network.

MR analysis revealed a total of 22 related to exosomes causally associated with DKD, defined as feature genes (Additional file 3). Of these, 10 were protective factors (TRAPPC6A, GALC, and RGS10, etc.) and 12 were risk factors (DCAF11, COQ5, and CMAS, etc.) (Additional file 4a). Steiger test and sensitivity analysis confirmed the reliability of the MR results (Additional file 5–7). Multivariate MR analysis showed that a total of 17 feature genes (EPDR1, CMAS, and RGS10, etc.) as significant in the development of DKD (Additional file 4b).

CMAS and RGS10 identified as biomarkers

A total of 6 key genes were identified by intersecting genes from two machine learning algorithms (Figure 3a–c). CMAS and RGS10 were selected as biomarkers due to their significant differential expression and consistent trends between the training and validation sets (Figure 3d). Additionally, they were confirmed in the Vesiclepedia database, indicating their relevance to exosomes (Additional file 8). Since the BOK gene was not detected in the validation set GSE30528, its expression could not be validated in this dataset. A nomogram was constructed based on the biomarkers, and its performance was evaluated using calibration curves, ROC curves (AUC = 0.930), and DCA curves. The nomogram demonstrated improved predictions (Figure 3e–h), suggesting it is a robust tool for predicting outcomes in our study population. To further analyze the functions of the biomarkers, enrichment analysis and GeneMANIA analysis were conducted. CMAS was predominantly enriched in GO categories related to mitochondrial protein-containing complexes and in KEGG pathways, notably oxidative phosphorylation (Additional file 9a). RGS10 was predominantly enriched in GO categories related to structural constituents of the extracellular matrix, and also within KEGG pathways, specifically those involving protein processing in the endoplasmic reticulum (Additional file 9b). The Gene-Gene Interaction (GGI) network revealed many genes associated with the biomarkers’ functions, such as GNAL3 and NAA10 (Additional file 9c).

Figure 3.

Machine learning algorithms gets key genes. (a–c) Gene crossover by machine learning algorithm identified 6 key genes. (d) CMAS and RGS10 as biomarkers. (e) Nomogram of biomarkers. (f–h) The calibration curves, ROC curves and DCA curves for nomogram.

Immune cell prevalence in DKD

The infiltration levels of 22 immune cells in individuals with DKD and normal groups were depicted (Figure 4a). The Wilcoxon test showed a significant difference in the levels of nine immune cell types, including memory B cells and CD8 T cells, between the DKD and normal groups (p < 0.05), with M2 and M1 Macrophages more prevalent in DKD while Neutrophils were more prevalent in control (Figure 4b). The correlation analysis revealed that the biomarkers exhibited a strong positive correlation with Neutrophils and a significant negative correlation with M2 Macrophages (Figure 4c). Analysis of biomarker correlation with the inflammation index showed a correlation of −0.39 between CMAS and HALLMARK_INFLAMMATORY_RESPONSE (Figure 4d).

Figure 4.

Estimation of immune cell infiltrations in DN. (a) Infiltration levels of 22 immune cells in DN individuals and normal groups. (b) Comparisons using Wilcoxon tests showed significant differences in levels of 9 immune cell types. (c) The biomarkers were strongly positively related to neutrophils and significantly negatively related to M2 macrophagess. (d) The correlation between CMAS and HALLMARK_INFLAMMATORY_RESPONSE is negative.

Potential therapeutic efficacy of drugs in DKD

In the database, CMAS predicted seven shared miRNAs and RGS10 predicted four shared miRNAs (Figure 5a,b). A ceRNA network containing 337 nodes and 377 edges was obtained, featuring important miRNAs such as hsa-miR-199b-5p, hsa-miR-2355-5p, and hsa-miR-3200-5p (Figure 5c). Drugs including calcium lactate, didanosine, tolmic acid, and alpha-tocopherol succinate were identified as potentially therapeutic for DKD (Figure 5d).

Figure 5.

Regulatory network construction and potential drug prediction. (a,b) CMAS predicted 7 shares miRNAs, RGS10 4 shared miRNAs. (c) The important miRNAs obtained are hsa-miR-5p, hsa-miR-2355-5p and hsa-miR-3200-5p. (d) Drugs such as calcium lactate, adenosine, formic acid and succinate fertilizer have been found to have potential therapeutic effects.

Expression of CMAS and RGS10 in DKD

Compared with the control group (Figure 6a–c), the IHC results showed a decrease in RGS10 and CMAS proteins in the kidney tissue of DKD subjects, with RGS10 decreasing significantly. This is consistent with our validation results.

Figure 6.

Validation of biomarker expression. (a–c) The IHC results showed a significant decrease of RGS 10 and CMAS proteins in the renal tissues of DN subjects, in which RGS 10 was significantly decreased (n = 4).

Discussion

DKD is a major complication of diabetes, a chronic progressive disease that can lead to end-stage kidney disease and increase the incidence and mortality of cardiovascular disease. DKD is characterized by its complexity and strong genetic component [43]. EVs, released by most cell types and circulating in body fluids, act as mechanisms for long-distance intercellular communication and regulate gene expression profiles and the fate of target cells [44]. Studies have demonstrated that exogenous EVs promote renal regeneration and reduce inflammation and fibrosis in chronic kidney disease [45]. The study utilized two DKD-related datasets, GSE96804 and GSE30528, along with 121 ERGs and 200 IRGs. Methods such as differential analysis, co-expression network construction, and MR analysis were employed to identify candidate genes, while machine learning and expression validation were conducted to identify biomarkers. The potential mechanisms of these biomarkers were investigated through IHC staining, enrichment analysis, immuno-infiltration analysis, and regulatory network construction. Ultimately, two EV-related biomarkers associated with DKD (CMAS and RGS10) were identified, revealing their potential roles in disease progression. These findings provide valuable insights into the pathogenesis of DKD and aid in the search for new therapeutic targets.

Analysis by machine calculation identified two biomarkers: CMAS and RGS10. CMAS, an activated glycosyltransferase, plays a crucial role in the biosynthesis of sialic acid oligosaccharides [46]. Recent studies suggest its significant involvement in multiple diseases. For instance, CMAS has been implicated in various cancers including colon, breast, and pancreatic cancers, where it promotes the proliferation and migration of cancer cells [47,48]. Conversely, studies on CMAS in diabetes are limited, revealing decreased expression of CMAS in DKD based on kidney tissue datasets and clinical samples. Additionally, this study found an inverse association between the expression level of CMAS and the inflammatory index score. Numerous studies have confirmed a close link between DKD and inflammation. Hyperglycemia can induce the expression of inflammatory mediators, such as chemokines and cytokines, and exacerbate renal injury through various mechanisms [29,49]. Therefore, we hypothesize that low expression of CMAS may promote the production of inflammatory mediators, adversely affecting the development of DKD. However, more in-depth experimental and clinical studies are required to verify this mechanism and define the precise role of CMAS in DKD. In conclusion, given its key roles in multiple diseases, the potential mechanisms and therapeutic target value of CMAS deserve further investigation in future studies.

RGS10, a member of the RGS protein family, is known for promoting GTP enzyme activity, accelerating cytokine triphosphate hydrolysis, and regulating G protein signaling pathways [50]. Several studies have demonstrated RGS10’s negative regulation of the nuclear factor NF-κB pathway, limiting the production of pro-inflammatory cytokines in microglia [51]. Additionally, RGS10-deficient macrophages produced higher levels of pro-inflammatory cytokines including TNF, IL-1β, and IL-12p70 in response to lipopolysaccharide (LPS) treatment, indicating that RGS10 acts as a negative regulator of the inflammatory response [52]. There is a tight association between insulin resistance, metabolic syndrome, and chronic inflammation, in which RGS10, as a key negative inflammatory mediator, plays a pivotal role. This study showed a significant negative correlation between RGS10 and M1 macrophages. It is suggested that RGS10 negatively regulates the activation of M1 macrophages by effectively inhibiting the NF-κB signaling pathway, a mechanism that further limits the production of proinflammatory cytokines during inflammatory stress, thus exhibiting a significant anti-inflammatory effect in activated macrophages [50]. This finding is consistent with the results of this study. Furthermore, with the observed decrease in RGS10 expression in DKD, it is speculated that reduced RGS10 expression may diminish the anti-inflammatory effect, ultimately exacerbating the inflammatory impact on the development of DKD.

The results of the enrichment analysis revealed that CMAS and RGS10 were primarily enriched in mitochondrial protein complexes and the oxidative phosphorylation pathway. Oxidative stress, characterized by an imbalance between excessive oxygen radicals and insufficient antioxidant substances, is a key mechanism contributing to DKD [53]. Excessive production of reactive oxygen species inflicts significant damage on cells and tissues. The mitochondrial respiratory chain serves as the primary pathway for sustained ROS production [54]. Oxidative stress resulting in mitochondrial dysfunction induces ATP depletion as well as mitochondrial membrane potential loss. Depleted ATP synthesis causes reduction in glutathione levels as its rate of formation is ATP mediated, occurring in all segments of nephron except proximal tubule. Thus the mitochondrial membrane potential loss results in increase in mitochondrial permeability which in turn releases cytochrome C (CytC) [55]. Cytochrome-C enters into the cytoplasm and binds to Apaf-1 to form the oligomer under the synergistic effect of ATP/dATP, which activates caspase-9 and downstream caspase-3, leading to cell apoptosis [56]. Moreover, Mitochondria play a fundamental and multifaceted role in the regulation and orchestration of cellular Ca2⁺ signaling [56]. Research indicates that an increase in intracellular Ca²⁺ concentration is a key factor triggering the release of EVs [45]. Mitochondrial dysfunction may lead to abnormal Ca²⁺ concentrations, which could negatively impact the EV release process. Given the role of EVs in the pathophysiology of DKD [57,58], understanding the complex interactions between CMAS and RGS 10, mitochondrial function, Ca²⁺ concentration regulation, and EV release is vital for fully comprehending the pathogenesis of DKD.

Although DKD is not traditionally viewed as an immune-mediated disease, increasing evidence suggests that immunity plays a significant role in its progression, involving a complex network of molecules and biological processes [59]. For instance, IL-8 is elevated in the serum of type 2 diabetic patients and causes podocyte damage via the IL-8-CXCR1/2 axis, further aggravating renal damage in the presence of other toxic and proinflammatory factors [60,61]. Studies have indicated that various leukocyte subsets, including macrophages and T lymphocytes, are involved in the pathological process of DKD [62]. In this study, M2 and M1 macrophages were more prevalent in the DKD group, and the phenotypic activation imbalance of M1/M2 macrophages was a critical factor in DKD [63]. Sirt6 molecules protected podocytes from injury in simulated diabetic kidney environments by activating M2 macrophages [64]. Additionally, HPS improved the status of DKD in mice by promoting the polarization transition from M1 to M2 phenotype and inducing CD4 T cells into Th2 and Treg cell populations [65]. These findings suggest that alterations in the type and quantity of immune cells may play a crucial role in the progression of DKD, offering valuable insights for better understanding and intervention in its pathogenesis.

In the database, important miRNAs identified include hsa-miR-199b-5p, hsa-miR-2355-5p, and hsa-miR-3200-5p. In a DKD model mouse, hsa-miR-199b-5p is overexpressed, regulating related signaling pathways by inhibiting the expression of the Klotho gene, thus affecting the development of DKD. It also plays a role in maintaining actin skeleton stability in glomerular podocytes, regulating extracellular matrix signal transduction, and processing proteins in the endoplasmic reticulum [66]. hsa-miR-2355-5p inhibits bladder cancer by regulating the expression of DDX11-AS1 and LAMB3, and in chondrosarcoma cells, it promotes VEGFR2-mediated angiogenesis through exosomal spongification of RAMP2-AS1 [67]. HULC, as a competitive endogenous RNA of miR-3200-5p (ceRNA), regulates ferroptosis in HCC cells treated with si-HULC by activating transcription factor 4 (ATF4), which plays a role in various stress responses [68]. Reports on hsa-miR-2355-5p and hsa-miR-3200-5p in DKD studies are scarce. Furthermore, data analysis has revealed that drugs such as calcium lactate, didanosine phthalate, and alpha-tocopherol succinate show potential therapeutic effects. Oxidative stress is central to the onset and progression of DKD, triggering pathological processes and exacerbating inflammatory responses, thus creating a detrimental oxidative stress-inflammatory cycle [69]. α-Tocopherol succinate, as a naturally derived antioxidant, significantly reduces ROS levels induced by cisplatin [70,71].

Conclusion

Two biomarkers related to EVs (CMAS and RGS10) were identified through transcriptome analysis combined with Mendelian analysis, revealing their potential mechanisms of action in DKD, which was further verified through clinical experiments. However, several limitations were noted. First, the different stages of the disease were not considered, which may lead to a less comprehensive understanding of the changing expression levels of biomarkers. Secondly, besides inflammatory factors, the pathogenesis of DKD may involve other influencing factors, such as genetic, metabolic, and environmental factors, which were not fully explored in this study. Moreover, we noted that both biomarkers, CMAS and RGS10, were associated with multiple diseases, suggesting the specificity of our single marker may be limited. Additionally, the lack of direct comparison data between DKD and pure diabetes partly hinders our deep understanding of the differential expression of biomarkers between these two disease states. In the future, we plan to expand the scope of data collection to include more disease stage information to fully reveal the expression dynamics of biomarkers; second, and consider genetic factors, metabolism, and environment to explore the pathogenesis of DKD. Moreover, strengthen the specificity verification of biomarkers, and build a multi-marker detection model, to improve the accuracy and specificity of diagnosis and effectively overcome the limitation of single marker. In addition, we will conduct a comparative study between DKD and pure diabetes to reveal the subtle differences more precisely in the expression of biomarkers, thus promoting a deep understanding of the pathogenesis of DKD.

Glossary

Abbreviations

DKD

Diabetic Kidney Disease

EVs

Extracellular Vesicles

MR

Mendelian Randomization

ERGs

Exosome-Related Genes

IRGs

Inflammation-Related Genes

IHC

Immunohistochemistry

T2D

Type 2 Diabetes

T1D

Type 1 Diabetes

MV

Microvesicles

KEGG

Kyoto Encyclopedia of Genes and Genomes

GEO

Gene Expression Omnibus

eQTL

Expression Quantitative Trait loci

IEU

Integrative Epidemiology Unit

GO

Gene Ontology

PPI

Protein-Protein Interaction

LD

linkage Disequilibrium

IVW

Inverse Variance Weighted

RF

Random Forest

ROC

Receiver Operating Characteristic

DCA

Decision Curve Analysis

GSEA

Gene Set Enrichment Analysis

GGI

Gene Gene Interaction

Funding Statement

This work was supported by the National Natural Science Foundation of China [81774279]; Key research and development project of Sichuan Provincial Science and Technology Department [2022YFS0389]; Chengdu University of Traditional Chinese Medicine Xing lin Scholar Nursery talent special fund [MPRC2023024].

Ethics approval and consent to participate

The study was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of the Affiliated Hospital of Chengdu University of Traditional Chinese Medicine. (Approval date is June 14,2023). All participants provided written informed consent.

Authors’ contributions

The corresponding author attests that all of the listed authors meet the authorship criteria and that no others meeting the criteria have been omitted. Y X and W Q Y performed the main statistical analysis including data curation, formal analysis, and investigation. contributed to the investigation and methodology. Y R S and Z L B contributed to the conceptualization and design of the study, advised on statistical aspects and interpreted the data. Y X offer advice regarding the data interpretation and supervised. W Q Y obtained funding and supervised the overall project. All of the authors participated in drafting the manuscript. All of the authors reviewed the manuscript and approved the final version to be published.

Disclosure statement

The authors declare that they have no competing interest.

Data availability statement

The datasets analyzed during the current study are available in the Gene Expression Omnibus repository, http://www.ncbi.nlm.nih.gov/geo/ and the datasets Integrative Epidemiology Unit repository, https://www.ebi.ac.uk/gwas/.