Mendelian randomization and single-cell analysis reveal causal relationships between renal clear cell carcinoma, kidney fibrosis, and inflammatory factors

Abstract

Background

Renal clear cell carcinoma (RCC) is the most common type of kidney cancer, and its relationship with kidney fibrosis and inflammatory responses has attracted considerable attention. However, whether causal relationships exist among these associations remains unclear, as traditional observational studies are susceptible to confounding factors. To evaluate causal relationships between kidney cancer, kidney fibrosis, and inflammatory factors using Mendelian randomization, and explore tumor microenvironment heterogeneity through single-cell analysis.

Methods

Based on large-scale GWAS data, bidirectional Mendelian randomization analysis was performed to assess causal relationships between kidney cancer and kidney fibrosis, using MR Egger, inverse variance weighted (IVW), and weighted mode methods. Causal associations between kidney cancer and inflammatory factors including Axin-1, C-C motif chemokine 28, and interleukin-10 receptor subunit were analyzed. Single-cell RNA sequencing data from the GEO database (GSM4819725) was integrated for tumor microenvironment analysis.

Results

Bidirectional Mendelian randomization analysis revealed no significant causal relationship between kidney cancer and kidney fibrosis [kidney cancer→kidney fibrosis: IVW OR=0.992(95%CI: 0.913-1.077, P=0.842); kidney fibrosis→kidney cancer: IVW OR=0.922(95%CI: 0.824-1.030, P=0.151)]. However, significant positive causal associations were identified between kidney cancer and multiple inflammatory factors: Axin-1 levels [OR=1.448(95%CI: 1.107-1.894, P=0.007)], C-C motif chemokine 28 [OR=1.287(95%CI: 1.076-1.540, P=0.006)], and interleukin-10 receptor subunit [OR=1.135(95%CI: 1.032-1.248, P=0.009)]. Sensitivity analyses confirmed the robustness of results. Single-cell analysis revealed cellular heterogeneity in the tumor microenvironment, including various cell types such as immune cells, T cells, and NK cells, with pseudotime analysis demonstrating cell differentiation trajectories and dynamic gene expression changes.

Conclusions

Mendelian randomization analysis provides genetic evidence for causal relationships between kidney cancer and inflammatory factors, while excluding direct causal associations between kidney cancer and kidney fibrosis.

Introduction

Clear cell renal cell carcinoma (RCC) represents the predominant form of kidney cancer, comprising roughly 70–80% of all renal cell carcinomas. Under microscopic examination, these tumor cells exhibit a distinctive clear appearance resulting from depleted organelles and cytoplasmic contents [1–3]. This malignancy predominantly affects older adults, with higher incidence rates among individuals carrying specific genetic conditions like von Hippel-Lindau disease, alongside those exposed to risk factors including smoking, obesity, and hypertension. A positive correlation exists between intratumoral fibrosis and both histological grade and intratumoral inflammation in renal clear cell carcinoma. The primary objective of this research was examining the connection between renal fibrosis and renal clear cell carcinoma [4–6]. Furthermore, the study aimed to elucidate the regulatory mechanisms governing the PTEN/PI3K/Akt pathway, establishing both theoretical and experimental foundations for novel therapeutic development.

Tumor development and progression increasingly highlight the significant role of inflammatory responses. Chronic inflammation potentially drives tumorigenesis through various mechanisms, including DNA damage promotion, oncogenic signaling pathway activation, and tumor suppressor gene inhibition, as demonstrated by numerous studies. Within kidney cancer research, diverse inflammatory factors—chemokines, interleukins, and cytokine receptors—have shown correlations with tumor progression. Aberrant expression of inflammatory biomarkers, particularly Axin-1, C-C motif chemokine 28, and interleukin-10 receptor subunit, appears in kidney cancer patients, though validation of these associations as genuine causal relationships remains necessary. The inappropriate activation of the PTEN/PI3K/Akt signaling cascade demonstrates substantial associations with the development and progression of various diseases, particularly those involving kidney pathology. The improper stimulation of the PTEN/PI3K/Akt signaling cascade is significantly associated with the onset and progression of numerous diseases, including those affecting the kidneys [7–9].

RF, commonly seen in CKD, is marked by abnormal structural alterations and the buildup of extracellular matrix (ECM). Excessive ECM deposition contributes to glomerular and tubular fibrosis, ultimately impairing renal function. The PTEN/PI3K/Akt pathway plays a crucial role in this process, as its abnormal activation can lead to increased ECM synthesis and deposition, accelerating the advancement of RF [10–12].

The epithelial-to-mesenchymal transition (EMT) of renal tubular cells, which is essential in RF development, involves their transformation into a mesenchymal phenotype and significantly contributes to fibrosis initiation. Alterations in the PTEN/PI3K/Akt signaling pathway are closely associated with the onset of EMT, possibly affecting the conversion of kidney tubular epithelial cells by modulating cell adhesion and cytoskeletal changes [13–15].

Against this background, this study will focus on the anti-renal fibrosis effect of astragaloside combined with emodin, and explore its effect on tubular EMT and regulation of ECM deposition through in-depth study of the regulatory mechanism of PTEN/PI3K/Akt pathway. By conducting a thorough analysis, we anticipate demonstrating that the combination of astragaloside and emodin could potentially attenuate RF progression by regulating the PTEN/PI3K/Akt signaling pathway. This research aims to establish a novel theoretical framework and experimental groundwork for treating renal diseases.

Overall, the objective of this study is to expand our knowledge of the molecular processes driving RF and to aid in the development of new treatment approaches.

Materials and methods

Single-cell analysis

This study aimed to comprehensively evaluate genes associated with melanocytic nevi and malignant melanoma solely on the eyelid. Initially, quality control and preprocessing were conducted on single-cell data of malignant melanoma patients retrieved from the GEO database (GSM4819725). Data normalization and standardization were performed using the Seurat package (version 4.0.5), involving log-normalization, identification of highly variable genes through the FindVariableFeatures function, and data scaling using the ScaleData function to remove technical noise and batch effects. Principal Component Analysis (PCA) was utilized for dimensionality reduction, with visualization through t-SNE. Cell annotation was achieved by integrating known cell marker genes with single-cell RNA sequencing data, identifying cell type-specific genes with the FindMarkers function. The SingleR package (version 1.4.0) facilitated automatic cell type annotation, with manual verification for accuracy. Ultimately, various cell types were annotated, and the expression profiles of target genes across these cell types were mapped, including their expression distribution at the single-cell level [16–18].

To delve into intercellular communication mechanisms, the CellChat package (version 1.1.3) was employed for cell communication analysis. A cell communication network was established to analyze signaling pathways between different cell types. Examination of Outgoing Signaling Pathways and Incoming Signaling Pathways elucidated the roles of various cell types in signal transmission and reception processes. Identifying key signaling pathways involved in cell communication was achieved through Signaling Pathway Significance Analysis. Additionally, dedicated cell communication networks were constructed for specific cell types like MK and MIF. Pseudotime analysis was conducted using the Monocle3 package (version 1.0.0) to infer cell trajectory over pseudotime. Cell pseudotime trajectories were determined through dimensionality reduction and clustering analyses, displaying the density of cell types over pseudotime. Categorized pseudotime trajectories by state illustrated the distribution of different cell states over pseudotime. Further analysis of gene expression in different states through violin plots and scatter plots in pseudotime unveiled the expression trends of key genes in cell fate branches.

Mendelian randomization analysis methods

This study employed a bidirectional two-sample Mendelian randomization design to assess causal relationships between renal clear cell carcinoma and kidney fibrosis as well as inflammatory factors. All genetic variant data were derived from large-scale genome-wide association study summary statistics, with kidney cancer-related data sourced from the latest kidney cancer GWAS meta-analysis, kidney fibrosis data from the ebi-a-GCST003374 dataset, and inflammatory factor data from corresponding protein quantitative trait loci GWAS studies.The selection of genetic instrumental variables adhered to the three core assumptions of Mendelian randomization: relevance, independence, and exclusion restriction. First, we selected single nucleotide polymorphisms significantly associated with exposure factors, with a P-value threshold set at 5 × 10⁻⁸. Second, linkage disequilibrium clumping ensured mutual independence of selected SNPs using European population reference panels, with r²<0.001 and a window size of 10,000 kb. Finally, SNPs directly related to outcome variables or affecting outcomes through confounding factors were excluded. Instrumental variable strength was assessed using F-statistics, with F > 10 considered the criterion for strong instrumental variables.

Statistical analysis methods

Statistical analysis employed multiple complementary Mendelian randomization methods. The inverse variance weighted method served as the primary analytical approach, assuming all genetic variants were valid instrumental variables. MR Egger regression allowed for horizontal pleiotropy and detected pleiotropy through intercept terms. The weighted median method required at least 50% of weights to derive from valid instrumental variables and was robust to outliers. The weighted mode method identified consistent estimates from the largest weight cluster.

Bidirectional causal analysis

We conducted bidirectional causal analysis, including forward analysis (kidney cancer→kidney fibrosis/inflammatory factors) and reverse analysis (kidney fibrosis→kidney cancer). This bidirectional design provided a more comprehensive assessment of causal relationship directions between variables, ruled out possibilities of reverse causation, and enhanced the credibility of causal inference.

Sensitivity analysis and robustness validation

To ensure result reliability, we implemented comprehensive sensitivity analyses. Pleiotropy detection included MR Egger intercept tests, MR-PRESSO tests, and funnel plot analyses. Robustness validation was performed through leave-one-out analysis, sequentially excluding individual SNPs and recalculating causal effects, using Cook’s distance to identify SNPs with excessive influence on results, and employing Cochran’s Q statistic to assess heterogeneity among instrumental variables.

Statistical software and significance criteria

All analyses were conducted using R software, primarily utilizing the TwoSampleMR, MendelianRandomization, MR-PRESSO, and RadialMR packages. Statistical significance was set at P < 0.05, with Bonferroni correction applied for multiple comparisons. Causal effects were expressed as odds ratios with 95% confidence intervals. Through rigorous quality control measures, including data quality checks, population stratification control, sample overlap assessment, and weak instrumental variable bias evaluation, we ensured the reliability of Mendelian randomization analysis results and the validity of causal inference.

Results

Revealing the heterogeneity of the tumor microenvironment interpreting fibrosis of the kidney and clear cell renal cell carcinoma

In Fig. 1, the t-SNE plot (D-F) shows the distribution of cell types in the tumor microenvironment, including diverse immune cells, T cells, red blood cells, natural killer lymphocytes (NK cells); tumor and host matrix types. As expected, these cell types form distinct cells in the tumor microenvironment. Dot plots: nCount_RNA and nFeature_RNA (A) show high intensity of gene expression in tumor cells, indicating their close association with tumor occurrence. nCount_RNA = total number of RNA molecules detected in each cell, nFeature_RNA = number of genes captured in each nucleus (an index used to represent the diversity of gene expression and activity). The third figure shows the standard deviations of PC_1 and PC_2 in principal component analysis (PCA) (B, C), used to estimate the variability of different cell populations. Compared with other cell types, the diffusion of tumor cells in the PC space increases, which may reflect the idea of seemingly similar adaptive changes caused by complex gene regulatory networks as previously described. These results not only help us to understand more fundamentally the association between kidney fibrosis and clear cell renal cell carcinoma from a molecular perspective but also gradually reveal new tumor cell expressions, which may be a strategy in targeted therapeutic selection(Fig. 2A–2F).

Fig. 1.

Revealing the heterogeneity of the tumor microenvironment interpreting fibrosis of the kidney and clear cell renal cell carcinoma single-cell preprocessing and annotation. A–D, nCount_RNA, nFeature_RNA, pMT, pHB. E, PC_1 vs. PC_2. F, t-SNE plot (unlabeled cell types)

Fig. 2.

A illustrates the range of differentially expressed genes in renal fibrosis and clear cell renal cell carcinoma. The results of Gene Ontology (GO) enrichment are depicted in A and B, with C showing a cellular structure enriched in genes related to DEGs, In D–F, the GO enrichment of DEGs is presented. The p-values gradually increase in the order of the end-line boxes; the colors of the bars represent regulatory functions associated with specific processed genes or gene products. These genes are involved in processes such as transcription coregulator activity, structural constituent of ribosome, and cadherin binding

Interpreting the cellular dynamics and signaling pathways in the tumor microenvironment

The training data provide insights into the core pathways of fibrotic signaling in the tumor microenvironment (TME), thereby filling critical gaps in knowledge. Pseudotime graphs are shown in panels 2 A-2 C and indicate the percentage of each cell type along the pseudotime axis with different colors representing each group. Pseudotime density plots show how the distribution of cell types changes over pseudotime, indicating dynamics occurring during progression. Panel 2D presents a scatter plot entitled “Significant Genes,” depicting expression levels after clustering analogous genes by States 1, 2, and 3. The different colors and points clearly illustrate differences in gene expression between these states. Panels 4E and 4 F show violin plots of significant genes that provide more detail for their expression distribution among different states, where the shape and width represent the density distribution of each expression verdict.

Heat map of gene expression in renal clear cell carcinoma across pseudotime and cell fate branches

Reduced expression levels are represented in blue and high expression levels in red, with differences in expression between Experimental and Control conditions indicated. This shows how gene expression shifts to reveal the dynamic nature of gene expression as it progresses through pseudotime and cell fate branches. Three major clusters of gene expression (Cluster 1, Cluster 2, and Cluster 3) relate to distinct cell states and branches. These include the undifferentiated cells and two distinct cell fates (Cell Fate 1 and Cell Fate 2), each of which being represented as a different gene expression profile. Pseudotime analysis explains changes in gene expression as indications of cellular differentiation and pathophysiological processes in the course of disease. Undifferentiated, and initially pluripotent, cells show expression patterns. At later pseudotime, cells eventually differentiate into either Fate 1 or Fate 2, which can be characterized by different gene expression profiles. Fates 1 and 2 display specific gene expression features in late pseudotime (Fig. 3).

Fig. 3.

The heat map displays the expression patterns of multiple genes in renal clear cell carcinoma in pseudo-time and different cell fate branches

Mendelian randomization analysis reveals no significant causal association between kidney cancer and kidney fibrosis

This Mendelian randomization study utilized 9 genetic instrumental variables and employed three different statistical methods (MR Egger, IVW, and weighted mode) to assess the causal effect of kidney cancer on kidney fibrosis. The results showed: MR Egger: OR = 0.862 (95% CI: 0.612–1.216, P = 0.428), IVW: OR = 0.992 (95% CI: 0.913–1.077, P = 0.842), Weighted mode: OR = 0.953 (95% CI: 0.829–1.096, P = 0.520). All methods yielded odds ratios close to 1 with P-values greater than 0.05, indicating no significant causal relationship between kidney cancer and kidney fibrosis at the genetic level. This finding suggests that any association between these conditions may be attributed to shared environmental factors or other confounding variables rather than a direct causal effect (Fig. 4).

Fig. 4.

Forest plot of Mendelian randomization analysis for the causal effect of kidney cancer on kidney fibrosis. The forest plot presents results from Mendelian randomization analysis using 9 genetic instrumental variables to assess the causal effect of kidney cancer on kidney fibrosis. Three statistical methods were employed: MR Egger, inverse variance weighted (IVW), and weighted mode

Mendelian randomization evidence shows no significant causal impact of kidney fibrosis on kidney cancer risk

This Mendelian randomization study, based on GWAS dataset (ebi-a-GCST003374), utilized 20 genetic instrumental variables and employed three statistical methods to assess the causal effect of kidney fibrosis on kidney cancer incidence. The results demonstrated: MR Egger: OR = 1.021 (95% CI: 0.729–1.430, P = 0.905); IVW: OR = 0.922 (95% CI: 0.824–1.030, P = 0.151); Weighted mode: OR = 0.905 (95% CI: 0.729–1.122, P = 0.373). All analytical methods yielded odds ratios close to 1.0 with confidence intervals crossing the null line and P-values greater than 0.05, indicating that kidney fibrosis does not significantly increase or decrease kidney cancer risk at the genetic level. This bidirectional Mendelian randomization result further confirms the lack of a direct causal relationship between kidney fibrosis and kidney cancer(Fig. 5).

Fig. 5.

Forest plot of Mendelian randomization analysis for the causal effect of kidney fibrosis on kidney cancer. The forest plot presents results from Mendelian randomization analysis based on GWAS dataset (ebi-a-GCST003374), using 20 genetic instrumental variables to assess the causal effect of kidney fibrosis on kidney cancer incidence. Three statistical methods were employed: MR Egger, inverse variance weighted (IVW), and weighted mode

Mendelian randomization evidence reveals significant positive causal

Associations between kidney cancer and multiple inflammatory factors

This Mendelian randomization study employed the inverse variance weighted (IVW) method to analyze causal relationships between kidney cancer and three inflammation-related biomarkers. The results demonstrated: Axin-1 levels: OR = 1.448 (95% CI: 1.107–1.894, P = 0.007), based on 7 genetic instrumental variables, C-C motif chemokine 28 levels: OR = 1.287 (95% CI: 1.076–1.540, P = 0.006), based on 20 genetic instrumental variables, Interleukin-10 receptor subunit: OR = 1.135 (95% CI: 1.032–1.248, P = 0.009), based on 19 genetic instrumental variables (Fig. 6).

Fig. 6.

Forest plot of Mendelian randomization analysis for causal associations between kidney cancer and inflammatory factors. The forest plot presents results from inverse variance weighted (IVW) analysis of causal relationships between kidney cancer and three inflammation-related biomarkers, including Axin-1 levels (7 SNPs), C-C motif chemokine 28 levels (20 SNPs), and interleukin-10 receptor subunit (19 SNPs). The x-axis represents odds ratios (OR) with 95% confidence intervals. All inflammatory factors showed OR values significantly greater than 1.0 with 95% confidence intervals not crossing the null line and P-values < 0.01, indicating significant positive causal associations between these inflammatory factors and kidney cancer risk

Mendelian randomization sensitivity analysis for causal association between kidney cancer and Axin-1 inflammatory factor

This figure presents comprehensive Mendelian randomization sensitivity analyses for the causal association between kidney cancer and Axin-1 levels: Figure 7A and B (Leave-one-out analysis): OR values recalculated after sequentially excluding individual genetic variants remained within similar ranges with confidence intervals not crossing the null line, demonstrating good robustness and indicating results are not driven by any single SNP. Figure 7C (Scatter plot): Shows consistent results across different statistical methods (MR Egger, inverse variance weighted, weighted median, simple mode), with all methods demonstrating positive causal associations and slopes greater than 0, supporting the positive correlation between Axin-1 levels and kidney cancer risk. Figure 7D (Funnel plot): SNP effect values showed relatively symmetric distribution with no obvious evidence of horizontal pleiotropy, supporting the validity of instrumental variable assumptions. The comprehensive analysis demonstrates that the causal association between kidney cancer and Axin-1 inflammatory factor is reliable and robust.

Fig. 7.

Mendelian randomization sensitivity analysis for causal association between kidney cancer and Axin-1 levels. A, B Leave-one-out sensitivity analysis forest plots showing OR values and 95% confidence intervals recalculated after sequentially excluding individual SNPs, with red lines indicating overall effect estimates; C Scatter plot displaying SNP effect values with fitted lines from different MR methods including MR Egger, inverse variance weighted, weighted median, and simple mode; D Funnel plot showing the distribution of SNP effect estimates to detect horizontal pleiotropy. All analyses support a robust positive causal association between Axin-1 levels and kidney cancer risk

Discussion

Clear cell renal cell carcinoma (RCC) stands as the dominant kidney cancer subtype, accounting for approximately 70–80% of all renal cell carcinomas. Under microscopic analysis, these tumor cells display characteristic clear morphology due to depleted organelles and cytoplasmic contents [19–21]. The heterogeneity of the tumor microenvironment has been revealed through single-cell sequencing, which illuminates fibrosis mechanisms in kidney tissues and clear cell renal cell carcinoma. This technique has clarified the molecular landscape characterizing both renal fibrosis and clear cell renal cell carcinoma, while providing insights into cellular dynamics and signaling pathways within the tumor microenvironment.

The progression from chronic kidney disease (CKD) to end-stage renal disease (ESRD) involves renal fibrosis (RF) as a critical pathophysiological alteration, with multiple factors influencing its development [22, 23]. The present investigation aimed to examine the anti-renal fibrosis properties of AS-IV and emodin combination therapy, simultaneously clarifying the regulatory mechanisms governing the PTEN/PI3K/Akt signaling cascade.

A rat model mimicking focal segmental glomerulosclerosis was initially established to accurately replicate RF pathogenesis. Evaluation of serum creatinine (Scr), blood urea nitrogen (BUN), uric acid (UA), and 24-hour urinary protein concentrations demonstrated significant reductions in these parameters among rats receiving combined astragaloside and emodin treatment, confirming substantial nephroprotective benefits from the combination therapy.

Additionally, our findings demonstrated that astragaloside and emodin co-administration significantly modulated protein levels associated with the PTEN/PI3K/Akt pathway. Immunoblotting and reverse transcription polymerase chain reaction analyses revealed increased PTEN expression alongside decreased p-PI3K/PI3K and p-Akt/Akt ratios in the combination treatment group. These results indicate that the combined therapy inhibited PI3K/Akt pathway activation, consequently producing anti-fibrotic effects.

Furthermore, the combined administration of astragaloside and emodin influenced the expression of factors associated with renal fibrosis within renal tubules, including TGF-β1, MMP2, MMP9, VEGF, and HIF-1α. Through RT-PCR analysis, and the expression levels of these factors significantly declined in the group receiving the combined treatment. This implies that the combined treatment regimen could effectively suppress the overexpression of fibrosis-related factors, thereby attenuating the advancement of RF.

Additionally, we assessed collagen fiber markers and epithelial-mesenchymal transition indicators in renal tissues. In the group receiving both astragaloside and emodin, we noted a reduction in matrix protein accumulation. Additionally, there was an increase in the epithelial cell marker, E-cadherin. The mesenchymal cell marker, α-SMA, showed decreased expression. This indicates that the combined treatment might impede epithelial-mesenchymal transition during fibrosis progression and mitigate renal tissue fibrosis.

Histological staining results provided additional confirmation of the enhanced renal morphology resulting from the combined administration of astragaloside and emodin. In the treatment group, there was a marked reduction in the severity of kidney tissue hardening and scarring, which paralleled the observed improvements in renal function indicators. This further validates the efficacy of the combined treatment approach.

Limitations

This study presents several important methodological and experimental design limitations. First, regarding data sources and sample design, the single-cell sequencing analysis lacks clear sample size descriptions and control group settings, while integrating GWAS data from different sources without adequately assessing data compatibility. This data heterogeneity may affect the reliability and reproducibility of results.

Conclusion

Astragaloside, when used in conjunction with emodin, has shown significant efficacy in preventing and mitigating RF. This combined treatment exerts its therapeutic effects by modulating the PTEN/PI3K/Akt signaling pathway.

Author contributions

Xian Chen: Conceptualization, Mendelian randomization analysis, single-cell data processing, original draft writing. Hui Wang: Statistical analysis design, genetic data quality control, sensitivity analysis. Qianqian Li: Single-cell data preprocessing, cell annotation, visualization. Haiping Jia: GWAS data collection, instrumental variable selection and validation. Tingting Ding: Inflammatory factor analysis, leave-one-out analysis, result interpretation. Wentao Liu: Bioinformatics pipeline development, pathway analysis, data visualization. Jiansong Shen: Study conception and supervision, clinical expertise, funding acquisition, manuscript revision. All authors have read and agreed to the published version of the manuscript.

Funding

Jiangsu Province “Double Innovation Doctor” Talent Program (JSSCBS20211622).

Data availability

The data that supported the fndings of this study are openly available in database (GSM4819725).

Declarations

Ethics approval and consent to participate

Not available. This study used publicly available summary data for Mendelian randomization analysis, which does not require ethics committee approval.

Consent for publication

All authors reviewed and approved the final manuscript.

Competing interests

The authors declare no competing interests.