Single-cell multi-omics and spatial transcriptomics reveal the transcriptional regulatory landscape of clear cell renal cell carcinoma

Abstract

Background

Clear cell renal cell carcinoma (ccRCC) represents the most aggressive form of renal cell carcinoma (RCC), distinguished by pronounced intratumoral heterogeneity, extensive metabolic reprogramming, and marked resistance to conventional therapeutic approaches. This study aimed to comprehensively characterize the cellular heterogeneity, epigenetic regulation, and transcription factor (TF) networks in ccRCC by integrating multi-omics data, and to identify functional key genes with prognostic and therapeutic significance.

Methods

Single-cell RNA sequencing (scRNA-seq), single-cell assay for transposase-accessible chromatin using sequencing (scATAC-seq), and spatial transcriptomics (ST) were integrated to comprehensively explore cellular heterogeneity, epigenetic regulation, and TF networks in ccRCC. To uncover dynamic alterations in gene expression during cellular differentiation, single-cell pseudotime analysis and gene set enrichment analysis (GSEA) were performed. Furthermore, the functional significance of Y-box binding protein 3 (YBX3) in ccRCC cells was experimentally validated.

Results

Single-cell transcriptomic profiling revealed 16 distinct cell populations within the ccRCC tumor microenvironment (TME), including ccRCC tumor cells, exhausted CD8⁺ T cells (Exhau CD8⁺ T cells), and macrophages. The scATAC-seq analysis demonstrated cell type-specific chromatin accessibility in immune cells, whereas ccRCC tumor cells exhibited reduced accessibility at immune-related genes, such as cluster of differentiation 2 (CD2). Epigenetic profiling further indicated that differentially accessible chromatin peaks in ccRCC cells were primarily enriched within intronic and exonic regions, implicating key TFs, including hepatocyte nuclear factor 1-beta (HNF1B) and the FOS-JUNB complex. An integrated analysis of scRNA-seq and scATAC-seq datasets identified five critical genes, YBX3, cubilin (CUBN), small nucleolar RNA host gene 8 (SNHG8), acetyl-CoA acyltransferase 2 (ACAA2), and protein kinase AMP-activated catalytic subunit α2 (PRKAA2), that were significantly associated with ccRCC prognosis. Pathway enrichment analysis revealed their involvement in metabolic reprogramming and tumor progression. Functional assays further confirmed that YBX3 knockdown inhibited ccRCC cell proliferation and migration.

Conclusions

This study elucidates cellular heterogeneity, the epigenetic regulatory landscape, and the key genes driving ccRCC progression. The integration of multi-omics data offers novel insights into precise diagnostic strategies and therapeutic interventions, highlighting the pivotal role of genes such as YBX3.

Highlight box.

Key findings

• Integrated multi-omics approach [single-cell RNA sequencing, single-cell assay for transposase-accessible chromatin using sequencing, and spatial transcriptomics (ST)] reveals cellular heterogeneity and transcriptional regulation in clear cell renal cell carcinoma (ccRCC).

• 16 distinct cell populations identified within the ccRCC tumor microenvironment, including ccRCC tumor cells, exhausted CD8⁺ T cells, and macrophages.

• Key prognostic biomarkers, YBX3, CUBN, SNHG8, ACAA2, and PRKAA2, identified with survival significance.

• YBX3 found to drive ccRCC progression and tumorigenic traits, with knockdown inhibiting cell proliferation and migration.

What is known and what is new?

• ccRCC is marked by intratumoral heterogeneity, immune evasion, and metabolic reprogramming. Epigenetic regulation in ccRCC remains underexplored.

• This study integrates single-cell RNA, chromatin accessibility, and ST to uncover novel key transcription factors and regulatory networks driving ccRCC.

What is the implication, and what should change now?

• The findings emphasize YBX3 as a central oncogenic driver in ccRCC. Further studies should explore its role in regulating tumor progression and immune evasion, and investigate potential therapeutic targeting of YBX3 and other biomarkers in ccRCC treatment strategies.

Introduction

Clear cell renal cell carcinoma (ccRCC) is the most common and a highly aggressive subtype of kidney cancer, accounting for approximately 70–80% of renal cancer cases, with a 5-year survival rate of less than 10% in patients with metastatic disease (1). With advances in diagnostic methods and surgical techniques, patients with small renal masses can achieve favorable outcomes, with 5-year cancer-specific survival rates exceeding 94% (2). Unfortunately, the prognosis for advanced or metastatic renal cell carcinoma (mRCC) remains poor, with an overall 5-year survival rate of less than 20% (3). Over the past decade, the first-line treatment of advanced renal cell carcinoma (RCC) has gradually shifted toward immune checkpoint inhibitor (ICI)-based combination regimens. For example, in a phase III trial with a median follow-up of 14.6 months, toripalimab plus axitinib reduced the risk of disease progression or death by 35% compared with sunitinib, as assessed by an independent review committee (IRC), and the IRC-assessed objective response rate (ORR) was significantly higher in the toripalimab-axitinib group than in the sunitinib group (56.7% vs. 30.8%) (4). Moreover, population-based analyses have demonstrated that the median overall survival (OS) of patients with mRCC improved from 8 months in the pre-ICI era [2011–2015] to 11 months in the ICI era [2016–2020] (5). ccRCC is characterized by marked heterogeneity and profound metabolic reprogramming, involving glycolysis (6), mitochondrial bioenergetics (7), and lipid metabolism (8), and broad resistance to conventional therapies (9). Recent investigations have revealed that the tumor microenvironment (TME) of ccRCC constitutes a highly intricate ecosystem composed of tumor cells, immune cells, and diverse stromal components (10). Within this environment, complex intercellular communication networks foster a milieu characterized by profound immunosuppressive features and aberrant metabolic activity, both of which are regarded as key drivers of ccRCC progression and therapeutic resistance.

The rapid advancement of single-cell sequencing technologies has enabled researchers to dissect the cellular composition and functional states of ccRCC at single-cell resolution. Nevertheless, transcriptomic data alone are insufficient to fully elucidate the regulatory mechanisms governing gene expression. The advent of single-cell assay for transposase-accessible chromatin using sequencing (scATAC-seq) has facilitated the identification of critical regulatory elements and transcription factor (TF) binding sites by profiling chromatin accessibility (11). Integrating these approaches provides a more comprehensive framework for understanding transcriptional regulatory networks in ccRCC. Moreover, the development of spatial transcriptomics (ST) has offered novel perspectives on the spatial organization of gene expression and cell-microenvironment interactions, as the microenvironment heavily affects disease biology and may affect responses to systemic therapy (12). Despite substantial progress in the molecular characterization of ccRCC, knowledge of its epigenetic regulatory mechanisms remains limited (13). Significant gaps persist in clarifying how chromatin accessibility dynamics and transcriptional regulators collectively shape tumor metabolic reprogramming, immune evasion, and divergent clinical trajectories (14). Key questions also remain regarding the regulation of metabolic plasticity by ccRCC-specific TF networks, the identification of epigenetic drivers underlying tumor heterogeneity and therapeutic resistance, and the contribution of spatiotemporal cis-regulatory remodeling to reinforcing transcriptional dependencies during tumor progression.

In this study, a multi-omics integration strategy combining single-cell RNA sequencing (scRNA-seq), scATAC-seq, and ST was employed to systematically investigate cellular heterogeneity, epigenetic regulation, and the functional roles of key driver genes in ccRCC (Figure 1). The analysis addressed three fundamental scientific questions: (I) how the cellular composition and intercellular interactions within the ccRCC microenvironment contribute to tumor progression; (II) which TFs and chromatin remodeling events regulate ccRCC transcriptional programs and how these changes dynamically within the TME; and (III) how spatiotemporal gene expression patterns shape tumor biology. Integration of scRNA-seq and scATAC-seq data identified 380 genes significantly upregulated in ccRCC. Random survival forest (RSF) analysis highlighted Y-box binding protein 3 (YBX3) as a key predictor of poor prognosis, while cubilin (CUBN), small nucleolar RNA host gene 8 (SNHG8), acetyl-CoA acyltransferase 2 (ACAA2), and protein kinase AMP-activated catalytic subunit α2 (PRKAA2) were associated with favorable outcomes. Pseudotime analysis further revealed sustained upregulation of YBX3 during tumor progression. ST provided spatial mapping of TME organization, and functional assays demonstrated that YBX3 knockdown suppressed the proliferation and migration of 786-O cells, thereby implicating these signaling pathways in YBX3-mediated oncogenic activity. The TME plays a pivotal role in driving ccRCC progression and therapeutic resistance (7). To elucidate these processes, scRNA-seq, scATAC-seq, and ST were employed to investigate gene activity, epigenetic modifications, and cellular interactions within the TME (8). This integrated multi-omics strategy enabled the identification of critical factors associated with ccRCC progression and provided deeper insight into the molecular mechanisms underlying disease pathogenesis and resistance. We present this article in accordance with the STREGA reporting checklist (available at https://tau.amegroups.com/article/view/10.21037/tau-2025-328/rc).

Figure 1.

Workflow diagram. GO, Gene Ontology; GSEA, gene set enrichment analysis; KEGG, Kyoto Encyclopedia of Genes and Genomes; RSF, random survival forest; scATAC-seq, single-cell assay for transposase-accessible chromatin using sequencing; scRNA-seq, single-cell RNA sequencing.

Methods

Data sources

ScRNA-seq data (19 samples), scATAC-seq data (19 samples, fragment files only), and ST data (5 samples) were obtained from the Gene Expression Omnibus (GEO) database (GSE207493 and GSE250163; https://www.ncbi.nlm.nih.gov/geo/). In addition, gene expression profiles and clinical information for kidney renal clear cell carcinoma (KIRC) were retrieved from The Cancer Genome Atlas (TCGA) (https://portal.gdc.cancer.gov/). The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Data quality control and normalization

The quality control of scRNA-seq data, shown in Figure S1, was performed using the Seurat package (15). Low-quality cells were excluded based on total unique molecular identifier (UMI) counts, number of expressed genes, and the proportions of mitochondrial and ribosomal gene expression. Outliers were removed using the median absolute deviation (MAD) method, and doublets were detected and eliminated with the DoubletFinder package, ensuring retention of high-quality single-cell data. Normalization was conducted using the LogNormalize method, which scales total expression per cell to 10,000 followed by logarithmic transformation. Cell-cycle scores were calculated with CellCycleScoring, and highly variable genes were identified using FindVariableFeatures. Variability arising from cell-cycle effects and differences in gene expression ratios was regressed out with ScaleData. Dimensionality reduction was first carried out using principal component analysis (PCA), followed by batch-effect correction with Harmony and nonlinear dimensionality reduction through uniform manifold approximation and projection (UMAP). Finally, cell types were annotated through a combination of manual curation and automated classification using SingleR.

ScATAC-seq data processing and integration with scRNA-seq

ScATAC-seq data were processed using the Signac (v1.10.0) R package (16). Low-quality cells were filtered out based on the following criteria: peak counts per cell (nCount_peaks) between 1,000 and 20,000, blacklist_fraction <0.05, nucleosome_signal <4, and transcription start site (TSS) enrichment score >1. Data normalization was performed with the term frequency-inverse document frequency (TF-IDF) method, and all peaks were used as features for dimensionality reduction through latent semantic indexing (LSI). Nonlinear dimensionality reduction was then applied using the RunUMAP function in the Seurat package. Gene activity scores were calculated with the GeneActivity function, and the resulting matrix was incorporated into the Seurat object.

Integration of scRNA-seq and scATAC-seq datasets was carried out within the Seurat framework (17). The FindTransferAnchors and TransferData functions were employed to identify anchors and predict cell-type labels across datasets, enabling accurate data integration and robust annotation of cell types.

TME cell interactions and functional analysis

The intercellular communication network within the ccRCC TME was analyzed using high-quality single-cell data. Ligand-receptor interactions were inferred with the CellChat package (18), which identifies signaling inputs, outputs, and cell-cell interactions through network analysis and pattern recognition. Interactions between cell subtypes were quantified by both interaction strength (weight) and frequency (count) to assess their contributions to disease progression. Functional annotation of genes was performed using the ClusterProfiler package, which enabled Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses to characterize the biological functions of key genes. Pathways with both a P value and a false discovery rate (FDR)-adjusted P value (q-value) below 0.05 were considered significantly enriched.

Epigenetic regulation and TF networks

Differential peak analysis of scATAC-seq data was performed using the FindAllMarkers function in the Seurat package (16), with thresholds set at average log₂ fold change (avg_log₂FC) >0.25 and adjusted P value (q-value) <0.05. Cell type-specific differential peaks were visualized through chromatin accessibility heatmaps, normalized by Z-score scaling across genomic regions. Differential peaks were annotated to genomic features, including exons and introns, using the ChIPseeker package. Enrichment analysis was then applied to evaluate biological pathways associated with these peaks. TF motifs were incorporated with the AddMotif function, and significantly enriched motifs were identified using FindMotifs. Key TFs were selected based on both fold change and biological relevance. Furthermore, TF footprint analysis was conducted to assess enrichment patterns of binding sites for critical TFs across different cell populations.

Key gene screening, RSF, and survival data analysis

Upregulated marker genes in ccRCC were first identified from scRNA-seq data. Intersection analysis of these genes with upregulated differential peaks from scATAC-seq data produced a gene set comprising 380 candidates. RSF analysis (19), implemented with the randomForestSRC package, was then applied to identify prognosis-related genes and rank them by importance scores. Seven genes with importance scores greater than 0.5 were selected as candidate biomarkers. Survival analyses of these genes were conducted using the TCGA-KIRC dataset, and Kaplan-Meier survival curves combined with log-rank tests revealed statistically significant prognostic associations (P<0.05) for YBX3, CUBN, SNHG8, ACAA2, and PRKAA2, supporting their potential as key genes for further investigation.

Single-cell pseudotime analysis

Pseudotime analysis of the five key genes was conducted on scRNA-seq data using the Monocle algorithm (20), which projects high-dimensional gene expression profiles onto a one-dimensional pseudotime trajectory. This approach enabled the inference of cellular fate and developmental pathways by ordering individual cells along a continuum of transcriptional progression, thereby clarifying the roles of these key genes in regulating cellular differentiation.

ST data analysis and developmental trajectory construction

UMI count distributions across five ST samples were assessed, revealing higher counts predominantly in epithelial regions. The data were subsequently normalized and standardized, followed by linear dimensionality reduction using PCA and nonlinear reduction with UMAP. Application of the Louvain algorithm identified nine distinct cell subgroups across the five samples.

Single-cell data were integrated with ST data using the spacexr package (21) for deconvolution analysis, enabling the identification of dominant cell types within each spatial spot. Expression levels of PRKAA2, CUBN, and ACAA2 were evaluated across spatial regions, whereas YBX3 and SNHG8 were notably undetectable in the ST dataset. To further assess tumor-immune interactions, particularly with T lymphocytes, we analyzed GSM7974886 using both unsupervised approaches and targeted ligand-receptor pairs [CD27-CD70 and programmed death-ligand 1 (PD-L1) and programmed cell death 1 (PD-1)], which confirmed direct interactions between tumor cells and T-cell subsets (22). Pseudotime analysis was then performed on selected cells extracted from ST data, with differentiation trajectories constructed based on calculated cell similarities to generate a trajectory map illustrating gene expression dynamics at distinct pseudotime points. Cells were colored according to pseudotime values (calculated by Monocle2 using gene expression data) and Seurat clusters, thereby highlighting transcriptional changes from initial to terminal states. Furthermore, pseudotime trajectories were mapped onto corresponding hematoxylin and eosin (H&E)-stained images to provide spatial context within the analysis.

Gene set enrichment analysis (GSEA)

GSEA was performed to compare signaling pathway differences between high- and low-expression groups of the five key genes. Pathway annotations were derived from the molecular signatures database (MsigDB, version 7.0), and significantly enriched gene sets (q-value <0.05) were ranked based on consistency scores.

YBX3 functional validation

The 786-O human clear cell renal carcinoma cell line (Shanghai Cell Bank, Chinese Academy of Sciences, Shanghai, China) was cultured in RPMI-1640 medium supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin at 37 °C in a 5% CO₂ atmosphere. Cells were transfected with YBX3-targeting small interfering RNAs (siRNAs) using Lipofectamine 2000 (Invitrogen, Carlsbad, USA) and subsequently seeded into 96-well or 6-well plates for functional assays. Cell proliferation was assessed using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay at 24 and 36 h post-transfection. Briefly, cells were incubated with MTT solution (5 mg/mL) for 4 h, the resulting formazan crystals were dissolved in dimethyl sulfoxide (DMSO), and absorbance was measured at 570 nm. For wound-healing assays, linear scratch wounds were generated in cell monolayers (80–90% confluence) using a pipette tip, and wound closure was evaluated at 24 and 36 h post-injury. The siRNA target sequence for YBX3 was siYBX3-1:

5'-CGGUUCAUCGAAAUCCAACUUTT-3' (sense) and 5'-AAGUUGGAUUUCGAUGAACCGTT-3' (antisense); siYBX3-2: 5'-CCGUCUGUUCGCCGUGGAUAUTT-3' (sense) and 5'-AUAUCCACGGCGAACAGACGGTT-3' (antisense); siYBX3-3: 5'-GAGAUGGAGAAACUGUAGAGUTT-3' (sense) and 5'-ACUCUACAGUUUCUCCAUCUCTT-3' (antisense); siCtrl: 5'-UUCUCCGAACGUGUCACGUTT-3'.

To investigate the effects of YBX3 knockdown on key signaling pathways, 786-O cells were harvested 24 h after siRNA transfection. Total protein was extracted using radioimmunoprecipitation assay (RIPA) buffer supplemented with protease and phosphatase inhibitors, and concentrations were determined by a bicinchoninic acid (BCA) assay. Equal amounts of protein were resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), transferred onto polyvinylidene fluoride (PVDF) membranes, and probed with the respective primary antibodies. Protein bands were visualized using enhanced chemiluminescence (ECL) reagents (Thermo Fisher Scientific, Waltham, MA, USA) and quantified with ImageJ software (version 1.54f). All experiments were performed in triplicate, and statistical significance was defined as P<0.05.

Statistical analysis

Statistical analyses and visualization of differential expression were conducted using R software (version 4.3.0). Multiple hypothesis testing was adjusted using the Benjamini-Hochberg procedure to control the FDR. A P value <0.05 was considered statistically significant. Significance levels were annotated as follows: *, P<0.05; **, P<0.01; ***, P<0.001; ****, P<0.0001; ns, not significant (P>0.05).

Results

Single-cell transcriptomics reveals cellular composition and functional heterogeneity in ccRCC

Clustering analysis of cells within the ccRCC TME was performed using UMAP dimensionality reduction, which identified 16 distinct cell populations (Figure 2A,2B). These populations included ccRCC tumor cells, exhausted CD8⁺ T cells (Exhau CD8⁺ T cells), tumor-associated macrophages (TAMs), and natural killer (NK) cells. Marker gene analysis (Figure 2C) revealed distinct expression patterns for ccRCC cells and Exhau CD8⁺ T cells, underscoring their critical roles within the TME. Further marker-based clustering demonstrated unique gene expression profiles in ccRCC cells, highlighting extensive interactions with immune cells. Notably, ccRCC cells constituted the dominant population across multiple samples and TME regions (Figure 2D).

Figure 2.

Single-cell transcriptomic clustering and identification of cell populations. (A) UMAP visualization presenting 16 distinct cell clusters derived from PCA-based dimensionality reduction. (B) Identification of 12 cell populations based on canonical marker genes. (C) Bubble plot displaying marker gene expression profiles specific to each cell type (bubble size: expression proportion; color intensity: average gene expression). (D) Distribution of cell populations across 19 samples. ccRCC, clear cell renal cell carcinoma; PCA, principal component analysis; UMAP, uniform manifold approximation and projection.

A cell-cell communication network was subsequently constructed within the ccRCC TME (Figure 3A,3B). Strong interactions were observed between ccRCC cells and both TAMs and cancer-associated fibroblasts (CAFs), predominantly mediated by ligand-receptor pairs such as CCL5-CCR1 (Figure 3C). Differential expression analysis of ccRCC cells revealed significant gene upregulation and downregulation (Figure 3D). GO analysis (Figure 3E) indicated that these genes were enriched in metabolic and stress-response processes, particularly those associated with hypoxia. Pathways related to glycolysis and pyruvate metabolism were significantly enriched, suggesting metabolic adaptation of ccRCC cells within the hostile TME. Additionally, enrichment of cell structure and adhesion pathways implicated ccRCC cells in metastatic progression. Consistent with these findings, KEGG pathway analysis (Figure 3F) demonstrated enrichment of differentially expressed genes in carbon metabolism and cancer-related stress-response pathways. Enhanced glycolysis and gluconeogenesis supported increased energy demands of tumor growth, while activation of the hypoxia-inducible factor 1 (HIF-1) signaling pathway and central carbon metabolism pathways highlighted mechanisms of hypoxia adaptation and metabolic reprogramming in ccRCC.

Figure 3.

Cell interactions and differential gene analysis. (A) Network diagram illustrating interactions among 12 identified cell populations; edge thickness indicates interaction probability and strength. (B) Cell populations ranked by interaction strength, showing the highest levels in ccRCC cells. (C) Ligand-receptor interaction map. (D) Heatmap of differentially expressed genes. (E) GO functional enrichment analysis. (F) KEGG pathway enrichment analysis. BP, biological process; CC, cellular component; ccRCC, clear cell renal cell carcinoma; GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; MF, molecular function.

Epigenetic regulation analysis reveals chromatin accessibility patterns and TF networks in ccRCC

Tumor samples were subjected to scATAC-seq analysis, and quality control metrics confirmed the suitability of the data for downstream analyses (Figure 4A). Chromatin accessibility clustering identified 27 distinct clusters corresponding to 12 cell types, consistent with scRNA-seq findings (Figure 4B). UMAP visualization (Figure 4C) demonstrated the distribution of key cell populations, including ccRCC cells, TAMs, and exhausted CD8⁺ T cells. Integration of scRNA-seq and scATAC-seq datasets through canonical correlation analysis (CCA) revealed co-regulatory relationships between gene expression and chromatin accessibility (Figure 4D). Given the critical immune function of cluster of differentiation 2 (CD2) (23), chromatin accessibility within the CD2 gene locus was further examined across immune cell subsets (Figure 4E). Notably, higher accessibility was detected in CD4⁺ T cells, exhausted CD8⁺ T cells, and pro-CD8⁺ T cells, whereas ccRCC cells exhibited significantly reduced accessibility.

Figure 4.

Single-cell multi-omics integration analysis. (A) Quality assessment metrics for scATAC-seq data, including total peak counts, TSS enrichment, proportion of peaks within blacklist regions, and nucleosome signal strength. (B) UMAP visualization of 27 chromatin accessibility-based cell clusters. (C) Comparative annotation of cells identified from scRNA-seq (left) and scATAC-seq (right). (D) UMAP representation integrating single-cell RNA and ATAC sequencing datasets. (E) Chromatin accessibility patterns at the CD2 genomic locus in ccRCC cells, immune cell populations, and other cell subsets. ATAC, assay for transposase-accessible chromatin; ccRCC, clear cell renal cell carcinoma; scATAC-seq, single-cell assay for transposase-accessible chromatin using sequencing; scRNA-seq, single-cell RNA sequencing; TSS, transcription start site; UMAP, uniform manifold approximation and projection.

Heatmap analysis (Figure 5A) demonstrated increased chromatin accessibility at specific differential peaks in ccRCC cells compared with other cell types, highlighting distinct functional and epigenetic regulatory features. Annotation of these peaks using the ChIPSeeker R package (Figure 5B) revealed that the majority were localized within intronic and exonic regions, whereas promoter and distal intergenic regions were less frequently represented, suggesting a relatively minor role in transcriptional regulation. Pathway enrichment analysis (Figure 5C) further indicated significant associations between ccRCC-specific differential peaks and pathways related to endocytosis, endoplasmic reticulum (ER) protein processing, and autophagy. Motif analysis identified binding motifs for six TFs, hepatocyte nuclear factor 1-beta (HNF1B), HNF1A, HNF4G, HNF4A, FOS, and JUNB (Figure 5D). Detailed visualization of chromatin accessibility within the CD2 gene locus (chr1:116755000-116770000) revealed marked differences among cell types (Figure 5E): accessibility was elevated in CD4⁺ T cells, exhausted CD8⁺ T cells, and pro-CD8⁺ T cells, whereas ccRCC tumor cells exhibited significantly reduced accessibility.

Figure 5.

Differential accessibility regions and key TF analysis. (A) Heatmap (Z-score normalized) highlighting differential chromatin accessibility across cell populations. (B) Genomic distribution of differentially accessible chromatin regions (promoters, introns, exons), illustrated by pie and bar charts. The y-axis represents the number of differential chromatin accessibility regions identified in each genomic feature. (C) Pathway enrichment analysis identifying significant biological pathways related to differentially accessible chromatin (ER protein processing, autophagy, tight junction pathways). (D) Motif enrichment analysis revealing TFs preferentially binding to differential chromatin peaks. (E) Aggregated signal profiles of representative transcription factors (e.g., HNF1B, HNF4A, FOS-JUNB complex) showing enrichment around differential chromatin accessibility peaks. (F) Venn diagram displaying the overlap (380 genes) between ccRCC-specific marker genes (scRNA-seq) and genes associated with elevated chromatin accessibility (scATAC-seq). ccRCC, clear cell renal cell carcinoma; DARs, differentially accessible regions; DEGs differentially expressed genes; ER, endoplasmic reticulum; scATAC-seq, single-cell assay for transposase-accessible chromatin using sequencing; scRNA-seq, single-cell RNA sequencing; TF, transcription factor; UTR, untranslated region.

Key gene identification and prognostic significance

Intersection analysis of upregulated marker genes (scRNA-seq) with upregulated differential peaks (scATAC-seq) identified 380 shared genes (Figure 5F). RSF analysis highlighted several prognostic candidates in ccRCC, with YBX3 showing the highest relative importance, followed by GPX3 and CUBN; SNHG8, PAX2, PRKAA2, and ACAA2 also demonstrated substantial importance (Figure 6A). Based on RSF modeling, seven genes were prioritized for survival analysis using TCGA data. Elevated YBX3 expression was strongly associated with poor prognosis [P<0.001, hazard ratio (HR) =1.715; Figure 6B]. In contrast, high expression of CUBN, SNHG8, PRKAA2, and ACAA2 significantly correlated with improved survival outcomes (all P<0.001; Figure 6C-6F). Neither GPX3 nor PAX2 showed statistically significant associations with prognosis (both P>0.05; Figure 6G,6H). Collectively, YBX3, CUBN, SNHG8, PRKAA2, and ACAA2 were identified as key prognostic biomarkers warranting further investigation.

Figure 6.

RSF analysis and key gene survival curves. (A) Ranking of variable importance from RSF analysis. (B-H) Kaplan-Meier survival curves for key genes (YBX3, CUBN, SNHG8, PRKAA2, ACAA2, GPX3 and PAX2). HR, hazard ratio; CI, confidence interval; RSF, random survival forest.

Decoding cell differentiation trajectories and dynamic expression changes of key genes in ccRCC

Pseudotime analysis using the Monocle algorithm was performed to reconstruct cellular differentiation trajectories from single-cell expression data (Figure 7A). Pseudotime reflects the continuum of cellular progression and transitional states during differentiation. Cells were classified according to differentiation pathways (Figure 7B) and further grouped into distinct clusters, each uniquely color-coded to highlight their distribution along differentiation trajectories (Figure 7C). Genes with significant pseudotime-dependent expression were identified and clustered into three major groups based on expression dynamics (Figure 7D). For example, WFDC2 was predominantly expressed at early stages of differentiation, whereas MT-ND3 expression peaked at later stages, underscoring their stage-specific regulatory roles.

Figure 7.

Cell pseudotime and key gene expression patterns. (A) Pseudotime trajectory showing cell progression, color-coded by pseudotime values. (B) Cell distribution categorized by differentiation state, represented by distinct colors. (C) Sample-wise comparison of cell distribution, indicated by different colors for each sample. (D) Heatmap showing gene expression dynamics across pseudotime; x-axis represents pseudotime progression, and y-axis shows genes grouped by similar expression patterns. (E) Expression profiles of key genes (ACAA2, CUBN, PRKAA2, SNHG8, YBX3) along differentiation trajectories. RCC, renal cell carcinoma.

Dynamic expression changes of prognostic genes were also observed along differentiation trajectories. ACAA2 and SNHG8 exhibited transient upregulation followed by a decline, whereas CUBN, PRKAA2, and YBX3 showed sustained increases throughout differentiation (Figure 7E). These findings suggest that prognostic genes exert distinct, stage-specific regulatory functions during ccRCC cell differentiation.

ST reveals cell type distribution and spatial gene expression patterns

As shown in Figure 8, ST analysis was performed across five samples (GEO accession: GSE250163). UMAP projection revealed nine initial cell subgroups, which were further refined into 12 populations, including ccRCC cells, TAMs, and exhausted CD8⁺ T cells (Figure 8A). ccRCC cells represented the dominant spatial population in most samples, except GSM7974886. Gene-level visualization highlighted variations in both expression and spatial abundance (Figure 8B). Notably, ACAA2 expression was significantly elevated in tumor regions, whereas PRKAA2 was expressed at lower levels. Figure S2A shows that CD70 expression was predominantly localized in tumor cells, whereas CD27 was mainly confined to T-cell regions. This spatial overlap supports potential CD70-CD27 interactions between tumor and T lymphocytes. In Figure S2B, PD-L1 (CD274) and PD-1 (PDCD1) were only weakly expressed in ccRCC tumor cells, with CD274 displaying scattered signals in tumor regions but stronger expression in stromal compartments, particularly CAFs and subsets of TAMs.

Figure 8.

ST analysis. (A) Deconvolution analysis illustrating variations in cell-type composition among spatial samples. (B) Spatial expression patterns of key genes (ACAA2, CUBN, PRKAA2), indicating gene localization (bottom left), expression intensity (bottom middle), and cell-type specificity (bottom right). ccRCC, clear cell renal cell carcinoma.

Pseudotime analysis reconstructed differentiation trajectories, progressing from early (State 1) to terminal states (State 3) (Figure 9A). Expression dynamics revealed that ACAA2 was gradually upregulated along pseudotime, showing minimal expression at early stages, intermediate levels at transitional states, and maximal expression at terminal states (Figure 9B). Integration of pseudotime with spatial coordinates (samples GSM7974844, GSM7974845, etc.) further illustrated the tissue distribution of differentiation stages, with a color gradient mapping the transition from early to late states (Figure 9C). Together, integration of ST and pseudotime analyses systematically delineated differentiation trajectories, dynamic gene-expression patterns, and their spatial localization, providing deeper insight into the spatial heterogeneity of cellular differentiation in ccRCC tissues.

Figure 9.

Spatial developmental trajectory. (A) Spatial pseudotime (left: continuous value; right: discrete state). (B) Expression dynamics of key genes (ACAA2, CUBN, PRKAA2) correlated with pseudotime progression; left panels illustrate developmental states by color, while right panels display gene expression trajectories across pseudotime, indicated by black trend lines. (C) Mapping of pseudotime values onto H&E-stained images from five spatial samples; shades of blue represent pseudotime values, visually depicting spatial differentiation progression. H&E, hematoxylin and eosin.

Enrichment analysis of key genes in signaling pathways

GSEA was performed to investigate signaling pathways associated with the five key genes. YBX3 was significantly enriched in the TNF signaling, cytosolic DNA sensing, and p53 signaling pathways (Figure 10A), all of which are tightly linked to immune regulation, cellular stress responses, and tumor suppression, highlighting a pivotal role for YBX3 in ccRCC progression. CUBN was enriched in ascorbate and aldarate metabolism, histidine metabolism, and phenylalanine metabolism pathways (Figure 10B), suggesting a role in metabolic regulation. SNHG8 showed strong enrichment in ribosome, spliceosome, and HIF-1 signaling pathways (Figure 10C), implicating it in protein synthesis, RNA processing, and hypoxia responses. ACAA2 was enriched in metabolic pathways including propionate, butyrate, and histidine metabolism (Figure 10D), while PRKAA2 displayed enrichment in similar metabolic processes (Figure 10E). Collectively, these results reveal that while YBX3 is primarily involved in immune and stress-response pathways, CUBN, SNHG8, ACAA2, and PRKAA2 are more closely associated with metabolic and transcriptional regulation, underscoring their complementary but distinct roles in ccRCC biology.

Figure 10.

Functional networks of key genes. (A-E) GSEA analyses (top) and interaction networks (bottom) for key genes (YBX3, CUBN, SNHG8, ACAA2, PRKAA2). GSEA curves illustrate the ES for core pathways. Network diagrams represent interaction strength (edge width) and expression direction (color). ES, enrichment scores; GSEA, gene set enrichment analysis.

Functional validation of YBX3 knockdown and its impact on signaling protein phosphorylation in 786-O cells

Effective knockdown of YBX3 (Figure 11A) markedly suppressed the proliferation of 786-O cells. Compared with controls, proliferation was reduced by 37.78% at 24 h and 54.71% at 36 h post-transfection (both P<0.001; Figure 11B). Consistent with this, wound-healing assays demonstrated significantly impaired migration capacity. YBX3-silenced cells exhibited wound closure rates of only 17.78% at 24 h and 45.50% at 36 h, compared with 27.87% and 68.23% in control cells, respectively (both P<0.001; Figure 11C,11D). Together, these results confirm a critical role for YBX3 in sustaining RCC cell proliferation and migration.

Figure 11.

YBX3 knockdown suppresses proliferation and migration in 786-O cells. (A) WB showing YBX3 protein expression following transfection with YBX3-targeting siRNAs (siRNA1-3) compared to NC-si. GAPDH served as the internal loading control. (B) MTT assay evaluating cell proliferation at 24 and 36 h post-transfection. (C) Representative phase-contrast microscopy images (magnification 50×) of wound healing assays at 0, 24, and 36 h post-scratch. (D) Quantitative analysis of wound closure presented as the percentage recovery at 24 and 36 h. Statistical significance was assessed using Student’s t-test: ***, P<0.001 vs. control; ^###, P<0.001 vs. NC-si. GAPDH, glyceraldehyde-3-phosphate dehydrogenase; MTT, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide; NC, negative control; WB, Western blot.

Discussion

In this study, single-cell transcriptomic profiling was applied to characterize the cellular composition of the ccRCC TME. UMAP-based dimensionality reduction and clustering identified 16 distinct cell subpopulations (Figure 2A,2B). Among these, ccRCC tumor cells and exhausted CD8⁺ T cells displayed particularly distinct expression profiles (Figure 2C), consistent with previous reports highlighting the immunosuppressive landscape of ccRCC (24). Across multiple samples, ccRCC cells represented the dominant population (Figure 2D), underscoring their central role in shaping the TME. Functional enrichment analysis revealed that ccRCC cells were highly enriched in glycolysis, pyruvate metabolism, and HIF-1 signaling pathways (Figure 3E,3F). Consistent with a previous study showing that HIFs are constitutively activated in ccRCC due to von Hippel-Lindau (VHL) loss and drive vascularization through downstream targets, our findings further underscore the central role of HIF-1 signaling in linking metabolic reprogramming with the highly vascularized phenotype of these tumors (25). It is noteworthy that although angiogenesis-related genes are broadly expressed in single-cell and spatial transcriptomic datasets of ccRCC, their specificity as discriminatory markers is sometimes limited; nevertheless, this does not diminish their well-established role as canonical drivers in ccRCC biology. These metabolic alterations are linked to both hypoxic adaptation and immune modulation within the TME. Recent evidence (26) further implicates HIF-1α in driving renal fibrosis through the PKM2/mTORC1/YME1L axis and M1 macrophage polarization, with co-culture models supporting the involvement of metabolic reprogramming in promoting pro-fibrotic immune responses. These alterations reflect metabolic reprogramming reminiscent of the Warburg effect (27) and are consistent with hypoxia-driven adaptations resulting from VHL gene inactivation (25). In addition, ligand-receptor interaction analysis identified active communication among ccRCC cells, TAMs, and CAFs, primarily mediated through the CCL5-CCR1 axis (Figure 3C). Such interactions may facilitate immune evasion (28) and provide mechanistic insights into the observed resistance to immunotherapy in ccRCC.

Integration of scRNA-seq and scATAC-seq data uncovered critical epigenetic regulatory mechanisms underlying ccRCC. Notably, chromatin accessibility at the CD2 locus was elevated in immune cells but reduced in tumor cells (Figure 4E), consistent with the notion that RCC is among the most immune-infiltrated tumors (29). These accessibility differences may enhance antitumor immunity in immune subsets while facilitating immune evasion in tumor cells. Genome-wide scATAC-seq analysis further revealed that ccRCC-specific differential peaks were enriched within exonic and intronic regions (Figure 5B), highlighting these regions as potential hubs for regulatory control. Functional enrichment of these peaks indicated associations with endocytosis, ER protein processing, and autophagy pathways (Figure 5C), pointing to the importance of epigenetic regulation in metabolic adaptation and tumor progression. Consistent with recent findings on the kynurenine pathway in ccRCC, emerging evidence further suggests that activation of specific metabolic pathways contributes to the regulation of angiogenesis and inflammatory signatures (30). Moreover, TF activity analysis identified six key TFs with significant activity in ccRCC, including HNF1B and FOS-JUNB (Figure 5D,5E). These TFs may act as central regulators by reshaping chromatin accessibility to support tumor growth and survival, providing promising candidates for further mechanistic and therapeutic investigation.

Our integrated multi-omics analysis combined with RSF modeling identified YBX3 as a pivotal oncogenic driver in ccRCC progression. YBX3, a member of the YBX family, is an RNA-binding protein primarily expressed in epithelial cells, where it regulates epithelial development, differentiation, stress responses, and organogenesis (31). Increasing evidence highlights its oncogenic functions across diverse cancers. For example, YBX3 promotes breast cancer cell proliferation (32), facilitates angiogenesis and hematogenous metastasis in bladder cancer (33), and acts as a potential oncogene in ccRCC (34). In our study, YBX3 exhibited strong prognostic significance (HR =1.715), while its persistent upregulation across pseudotime trajectories (Figure 7E) suggests a central role in sustaining malignant differentiation states. Such continuous overexpression likely drives hallmark tumorigenic processes, particularly uncontrolled proliferation and resistance to apoptosis, consistent with its established oncogenic properties. Notably, precancerous clones in mutant p53 (mutp53)-driven thymic lymphoma models with extensive chromosomal alterations also displayed marked YBX3 upregulation, implicating it in early tumorigenesis (35). Specifically, in ccRCC, elevated YBX3 expression in low-grade tumors and its heterogeneous nuclear localization in high-grade lesions further support its contribution to malignant phenotypes (36). Collectively, our multi-omics findings are in strong agreement with these independent studies, reinforcing YBX3 as a key molecular driver of ccRCC pathogenesis.

We also identified several additional key genes, CUBN, PRKAA2, SNHG8, and ACAA2, that exhibited distinct expression patterns significantly correlated with favorable patient outcomes (Figure 6C-6F). Pseudotime trajectory analysis demonstrated progressive upregulation of both CUBN and PRKAA2 during tumor progression (Figure 7E), indicating increased expression at advanced stages. Interestingly, ST analysis revealed particularly low PRKAA2 expression within tumor cells (Figure 8). This spatial pattern suggests that the tumor-suppressive effects of PRKAA2 are largely mediated by non-tumor cellular components of the TME, such as immune cells, stromal cells, and vascular endothelial cells, rather than tumor-intrinsic mechanisms. Prior studies have established PRKAA2 as the catalytic subunit of AMP-activated protein kinase (AMPK), a master regulator of cellular energy balance that functions, in part, through inhibition of mechanistic target of rapamycin (mTOR) signaling (37). The restriction of PRKAA2 expression to non-tumor compartments therefore implies that its anti-tumor activity may involve metabolic crosstalk between tumor cells and stromal elements within the TME.

In the single-cell dataset, the three genes with the highest relevance scores, SLC2A5, MUC1, and NFE2L2, exhibited strong co-expression with five key genes (PRKAA2, SNHG8, CUBN, YBX3, and ACAA2) (Figures S3-S7). Notably, MUC1 showed significant co-expression and, consistent with recent studies, may contribute to the modulation of cancer cell metabolism and immunoflogosis in ccRCC (38). Furthermore, differential expression analysis confirmed that this entire set of genes was significantly upregulated in tumor tissues compared to normal samples (Figure S8A). Correlation analyses and visualizations further emphasized their expression patterns and distributions across distinct cell populations, offering insights into their potential interactions in ccRCC. Co-expression network analysis (Figure S8B) revealed a positive correlation between YBX3 and CA9 (r=0.678) and a negative correlation between PRKAA2 and LGALS1 (r=−0.649). These results suggest the presence of a coordinated regulatory network involving these genes in ccRCC metabolism, immune evasion, and tumor progression. Figure S8C further illustrates YBX3 as a regulator of cell-cycle progression and hypoxic responses. Studies have shown that genetic deletion of YBX3 attenuates renal ischemia-reperfusion injury by modulating immune cell infiltration and inflammation, achieved through enhanced mitochondrial function and suppressed ferroptosis (39). Consistent with our findings, PRKAA2—linked to the PI3K–AKT signaling cascade—has also been reported to exert methylation-mediated immunoregulatory effects in autoimmune thyroiditis (40). Importantly, PRKAA2 demonstrated enhanced activity in oxidative phosphorylation, highlighting its contribution to the activation of multiple metabolic pathways that shape the biological properties of cancer stem cells (41,42). Cancer stem cells exhibit increased dependence on alternative metabolic pathways such as oxidative phosphorylation and fatty acid metabolism, which provide the necessary energy and building blocks for the self-renewal and drug resistance of cancer cells (43). GSEA (Figure 10) and gene set variation analysis (GSVA) (Figure S9) analyses revealed that YBX3 promotes tumor progression primarily through the TNF and p53 signaling pathways (Figure 10A and Figure S9A) (34). A prior study in diabetic nephropathy have shown that TNF signaling can enhance leukocyte infiltration and sustain pro-inflammatory cytokine networks, while p53 modulates immune responses through stress-induced transcriptional programs and regulation of inflammatory mediators (44). Knockdown of YBX3 (Figure 11A) significantly suppressed 786-O cell proliferation, reducing growth by 37.78% at 24 h and 54.71% at 36 h compared with control cells (P<0.001; Figure 11B). Cell migration was also markedly impaired, with wound-healing rates declining to 17.78% at 24 h and 45.50% at 36 h (P<0.001; Figure 11C,11D). These broad molecular alterations underscore the central role of YBX3 in orchestrating tumor-promoting networks in RCC. Moreover, the pleiotropic effects observed imply that YBX3 may function as an epigenetic regulator or RNA-binding protein, exerting influence through post-translational modifications. Such coordinated regulation of diverse signaling pathways likely explains its profound impact on cell proliferation and migration, establishing YBX3 as a compelling therapeutic target. Future work should clarify whether YBX3 regulates these pathways through direct interactions with upstream kinases or phosphatases.

As shown in Figure S10, analysis of YBX3 expression using the universal analysis of cancer (UALCAN) database (Figure S10A) and TCGA datasets (Figure S10B) revealed significantly elevated protein and mRNA levels in ccRCC tissues compared with normal counterparts (Figure S10A,S10B), implicating YBX3 in ccRCC pathogenesis. Expression was markedly higher in advanced T stages (T3 and T4) relative to early stages (T1 and T2), underscoring its potential role in tumor progression (Figure S10C). Elevated YBX3 levels also correlated with poorer therapeutic responses, as reflected by increased expression in tumors classified with progressive disease (PD) and stable disease (SD) (Figure S10D). Furthermore, YBX3 expression was highest in advanced clinical stages (stage III and IV) compared with early stages (stage I and II), supporting its association with heightened tumor aggressiveness (Figure S9E). Kaplan-Meier survival analysis of TCGA patients demonstrated that high YBX3 expression significantly correlated (P<0.05) with reduced OS, disease-specific survival (DSS), and progression-free interval (PFI) (Figure S10G-S10I). Collectively, these findings emphasize the pivotal role of YBX3 in driving ccRCC progression and its strong association with poor patient prognosis.

Several limitations should be acknowledged. First, the relatively small sample size may restrict the generalizability of our findings. Second, while the use of public datasets enhances reproducibility, potential sampling biases inherent to these resources must be taken into account, and the lack of validation in an independent patient cohort remains a major limitation. Although the function of YBX3 has been preliminarily validated in 786-O cells, further confirmation in a broader panel of RCC cell lines (45)—such as CAKI-1, CAKI-2, and RCC-4—with distinct VHL and HIF expression profiles (VHL⁺/⁻, HIF⁺/⁻) would strengthen the findings. Additionally, its clinical applicability requires further validation in independent patient cohorts and in vivo experimental models. Collectively, these methodological constraints underscore important considerations for future research aimed at confirming and extending the present results.

Conclusions

This study comprehensively characterized the cellular heterogeneity, epigenetic regulatory mechanisms, and functional roles of key genes driving ccRCC progression. Through the integration of multi-omics data, we uncovered novel insights into ccRCC biology, with particular emphasis on the pivotal role of YBX3. Our findings provide a deeper understanding of disease mechanisms and may inform future efforts toward the identification of biomarkers and potential therapeutic strategies.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Data Sharing Statement

Available at https://tau.amegroups.com/article/view/10.21037/tau-2025-328/dss

tau-14-10-2953-dss.pdf ^{(70.1KB, pdf)}

DOI: 10.21037/tau-2025-328