ACACA modulates R-loop homeostasis to enhance lipid metabolism and microenvironmental interactions in ccRCC

Abstract

R-loops are dynamic nucleic acid structures implicated in genome regulation and instability, yet their contributions to the tumor microenvironment (TME) of clear cell renal cell carcinoma (ccRCC) remain unclear. Here, we integrated publicly available ccRCC transcriptomic and clinical datasets to quantify R-loop-associated activity using single-sample gene set enrichment analysis (ssGSEA) and develop a prognostic model using linear machine-learning algorithms. R-loop activity was elevated in advanced disease and associated with unfavorable outcomes. Among R-loop-related genes, ACACA was prioritized for its dominant contribution across multiple linear prognostic models, alongside its strong association with metabolic reprogramming. Single-cell transcriptomics showed relatively high ACACA expression in malignant cells that functioned as communication hubs with lipid metabolism-related signaling, and spatial transcriptomics confirmed preferential enrichment of ACACA within malignant regions. In vitro and xenograft experiments further demonstrated that ACACA promoted proliferation and migration and suppressed apoptosis, accompanied by reduced R-loop accumulation, enhanced fatty acid metabolism and improved mitochondrial function. Together, these findings identify ACACA as an R-loop-associated metabolic driver connecting genomic stress with lipid reprogramming and TME remodeling in ccRCC, supporting its potential as a prognostic biomarker and therapeutic target.

Introduction

Clear cell renal cell carcinoma (ccRCC), the most common histologic subtype of kidney cancer, accounts for the majority of RCC-related deaths despite recent therapeutic advances^1,2,3. Metabolic reprogramming is a defining hallmark of ccRCC, exemplified by striking lipid droplet deposition and extensive alterations in fatty acid and cholesterol pathways, underscoring the pivotal role of lipid metabolism in tumor biology^4,5. These metabolic programs integrate with mitochondrial activity and redox control, functionally linking lipid handling to tumor progression⁶. Together, these features highlight the central role of metabolic dysregulation in ccRCC and raise the need to further elucidate the molecular mechanisms driving this process.

R-loops are three-stranded RNA-DNA hybrid structures formed during transcription that exert important regulatory functions, yet their accumulation can lead to replication stress and genome instability, thereby promoting oncogenesis^7,8. Aberrant R‑loop accumulation disrupts DNA replication and repair, driving genomic instability-a hallmark of cancer development⁹. In various malignancies, dysregulated R‑loop metabolism contributes to transcriptional misregulation, DNA damage response dysfunction, and therapeutic resistance^10,11. Emerging evidence suggests that R‑loop patterns not only contribute to genomic instability but also correlate with metabolic reprogramming phenotypes in cancer. For example, an R‑loop scoring signature has been shown to associate with lipid metabolism gene expression and prognosis in hepatocellular carcinoma, highlighting a link between R‑loop dynamics and tumor metabolic pathways¹². Moreover, dysregulated R‑loop homeostasis plays a pivotal role in melanoma progression by shaping both immunosuppressive environments and metabolic reprogramming, driving tumor aggressiveness¹³. Given that ccRCC is a typical metabolic tumor with altered lipid metabolism, it is likely that R‑loop dysregulation contributes to its development and progression. However, little is known about the role of R‑loop dynamics in ccRCC, particularly regarding its involvement in metabolic reprogramming.

Recent advances in bioinformatics have facilitated the comprehensive integration of multi-omics datasets, allowing large-scale transcriptomic resources to be combined with single-cell and spatial transcriptomic analyses^14,15. These approaches enable systematic assessment of gene activity in diverse tumor contexts, while resolving cell-type heterogeneity and spatial architecture within the tumor microenvironment¹⁶. By linking molecular signatures to clinical phenotypes and tissue organization, such integrative strategies provide powerful means to identify functional drivers and uncover mechanisms underlying ccRCC progression¹⁷.

To clarify R-loop involvement in ccRCC, we combined multi-cohort transcriptomic analyses with linear modeling to identify key R-loop-associated genes. Among these, ACACA (acetyl-CoA carboxylase 1) consistently emerged as a top candidate across multiple frameworks. We then focus on ACACA, a rate-limiting enzyme in de novo fatty-acid synthesis with key roles in tumor lipid metabolism, to explore its links to cell-cycle and DNA-damage programs and lipid-mediated microenvironmental signaling¹⁸. Through this integrative approach, we aim to bridge R-loop biology with metabolic reprogramming in ccRCC, offering novel perspectives on tumor pathogenesis.

Results

R-loop activity landscape and clinical significance in ccRCC

Analysis of TCGA-KIRC samples showed that R-loop-related gene activity, quantified by ssGSEA, exhibited substantial interpatient variability but was consistently higher in ccRCC samples and tended to increase with advancing disease stage (Fig. 1A). Clinical correlation showed that ccRCC with advanced stage (T3/4) as well as those with distant metastasis (M1) displayed significantly increased R-loop activity compared with early-stage or non-metastatic tumors (Fig. 1B). Survival analyses identified multiple R-loop-associated genes that were linked with patient outcomes, and the top 20 candidates with the most significant associations were highlighted for OS, DSS, and PFI (Fig. 1C). Differential expression profiling further revealed a subset of R-loop genes dysregulated between ccRCC and normal samples (Fig. 1D). The intersection of these dysregulated genes with those showing prognostic relevance yielded a panel of 44 candidates, which were retained for subsequent functional exploration (Fig. 1E). Together, these findings suggest that enhanced R-loop activity is a characteristic feature of ccRCC, relates to aggressive clinical behavior, and provides a pool of candidate biomarkers with potential prognostic utility.

Fig. 1: Comprehensive characterization of R-loop activity in ccRCC.

A Quantification of R-loop-related gene activity by ssGSEA across TCGA-KIRC samples, B Association of R-loop activity with clinical features, highlighting significantly higher levels in advanced T stage (T3/4) and metastatic cases (M1) compared with respective controls, C Prognostic relevance of R-loop-related genes, with the top 20 candidates (ranked by P value) significantly associated with OS, DSS, and PFI, D Differential expression analysis of R-loop-related genes between ccRCC and normal samples, E Venn diagram showing the intersection of prognostic genes and differentially expressed genes, identifying 44 candidates for further analyses.

Linear machine learning models identify ACACA as a key R-loop-related gene for risk stratification in ccRCC

Using the 44 R-loop-related genes, we constructed prognostic models with multiple linear machine learning algorithms and compared their performance. Time-dependent ROC analysis identified Elastic_net_0.1 as the best-performing method, showing the highest mean AUC values across datasets (Fig. 2A). Analysis of model coefficients across algorithms highlighted ACACA as a key determinant, as it consistently exhibited strong and non-zero weights, suggesting a central role in the prognostic signature (Fig. 2B). Meta-analysis of the Elastic_net_0.1-derived risk score indicated reproducible prognostic performance, yielding a pooled hazard ratio (HR) significantly greater than 1 across independent cohorts (Fig. 2C). Kaplan-Meier survival analyses further supported the discriminative ability of the model, with high-risk patients exhibiting markedly worse OS, DSS, and PFI in the TCGA-KIRC cohort (Fig. 2D). External validation in E-MTAB-1980 and GSE167573 confirmed the robustness of the model, showing consistent survival disadvantage in the high-risk group (Fig. 2E). To improve clinical utility, a nomogram integrating ACACA expression with T stage, M stage, and histologic grade was established (Fig. 2F). Calibration curves showed close agreement between predicted and observed survival at 1, 3, 5, and 8 years (Fig. 2G). Together, these findings demonstrate that Elastic_net_0.1 represents the most effective linear machine learning framework for R-loop-based risk modeling, and underscore ACACA as a critical R-loop-related gene driving risk stratification in ccRCC.

Fig. 2: Construction and validation of an R-loop-based prognostic model using linear machine learning algorithms.

A Heatmap of mean AUC values at 1-, 3-, and 5-year intervals across multiple algorithms, identifying Elastic_net_0.1 as the best-performing model, B Coefficient matrix of 44 R-loop-related genes across algorithms; ACACA consistently displayed high non-zero coefficients, highlighting its pivotal contribution to the prognostic signature, C Forest plot of meta-analysis evaluating the pooled HR of the Elastic_net_0.1 risk score across independent cohorts, D Kaplan–Meier survival curves for TCGA-KIRC, showing significantly poorer OS, DSS, and PFI in the high-risk group, E External validation in the E-MTAB-1980 and GSE167573 cohorts confirmed inferior survival outcomes in the high-risk group, F Nomogram combining ACACA expression with T stage, M stage, and histologic grade for individualized survival prediction, G Calibration plots showing strong agreement between predicted and observed 1-, 3-, 5, and 8-year survival probabilities.

Biological functions associated with ACACA expression in ccRCC

Given the central role of ACACA in the prognostic model, we further explored its potential biological functions. Stratifying tumors by ACACA expression and performing GSEA revealed that the high-expression group was significantly enriched for cell cycle-related programs, such as G2M checkpoint and E2F targets, together with multiple metabolic pathways (Fig. 3A). To complement these findings, we correlated ACACA expression with 14 functional states derived from the CancerSEA database. In the TCGA-KIRC cohort, ACACA showed significant positive correlations with Cell cycle, DNA damage, and DNA repair, indicating that elevated ACACA expression is linked to accelerated proliferative activity and altered genomic stability (Fig. 3B). Importantly, these associations were reproduced in the E-MTAB-1980 dataset, where ACACA again correlated with heightened cell-cycle activity and DNA damage as well as DNA repair capacity (Fig. 3C). Collectively, these results indicate that ACACA expression is linked to altered cell-cycle activity and metabolic processes, highlighting its potential relevance to ccRCC biology.

Fig. 3: Functional characterization of ACACA in ccRCC.

A GSEA comparing high- and low-ACACA expression groups, B Correlations between ACACA expression and functional state scores in TCGA-KIRC, highlighting significant positive associations with Cell cycle, DNA damage, and DNA repair, C Validation of ACACA-function correlations in the E-MTAB-1980 cohort, confirming similar positive relationships with Cell cycle, DNA damage, and DNA repair.

Single-cell landscape of ACACA expression and lipid metabolism-related signaling

To further investigate the cellular context of ACACA, we analyzed single-cell RNA-seq data GSE159115. UMAP visualization delineated major cell lineages, including CD8⁺ T cells, endothelial cells, epithelial cells, erythroblasts, malignant cells, monocytes/macrophages, pericytes, and plasma cells (Fig. 4A), and ACACA expression showed heterogeneous distribution with relatively high levels in malignant cells (Fig. 4B, C). Cell-cell communication analysis indicated that ACACA-high malignant cells exhibited dense intercellular connectivity in both connection number and interaction strength, particularly with immune and endothelial compartments (Fig. 4D), and outgoing/incoming signaling patterns identified malignant cells as central hubs (Fig. 4E). Given the link between ACACA and lipid metabolism, we interrogated the ANGPTL pathway and found ACACA-high malignant cells to be prominent senders and receivers within this network (Fig. 4F). Ligand-receptor analysis highlighted ANGPTL2-PIRB, ANGPTL2-ITGA5/ITGB1, and ANGPTL4-SDC family interactions as key axes (Fig. 4G). Collectively, these data suggest that ACACA-high malignant cells are embedded in an interaction-rich microenvironment and may modulate lipid metabolism-related signaling via ANGPTL in ccRCC.

Fig. 4: Single-cell characterization of ACACA expression and lipid metabolism-related signaling in ccRCC.

A UMAP plot showing major cell lineages in the KIRC single-cell dataset (GSE159115), B UMAP visualization of ACACA expression across cell populations, C Comparison of mean ACACA expression among different cell types, indicating relatively high levels in malignant cells, D Cell-cell communication networks by interaction number (left) and interaction strength (right), E Heatmaps of outgoing (left) and incoming (right) signaling patterns across cell lineages, identifying malignant cells as communication hubs, F Analysis of the ANGPTL signaling pathway, with malignant cells acting as key senders and receivers in interactions with immune and endothelial cells, G Chord plots showing representative ANGPTL ligand-receptor pairs (ANGPTL2-PIRB, ANGPTL2-ITGA5/ITGB1, ANGPTL4-SDC family) mediating cell-cell communication.

Spatial transcriptomic distribution of ACACA in ccRCC

Spatial transcriptomic analysis of ccRCC samples from GSE179572-GSM5420752 and GSE175540 revealed heterogeneous expression of ACACA across tissue sections, with variable but consistently elevated levels observed in tumor regions (Fig. 5A). Four representative samples with relatively high ACACA expression-KIRC (GSE179572-GSM5420752), KIRC7 (GSE175540-GSM5924035), KIRC9 (GSE175540-GSM5924037), and KIRC21 (GSE175540-GSM5924049)-were selected for detailed assessment. Spatial mapping demonstrated that ACACA expression was enriched within malignant compartments, displaying distinct distribution patterns among samples (Fig. 5B, C). Correlation analyses further showed that ACACA levels were positively associated with tumor cell-related signatures and variably linked to immune and endothelial cell programs (Fig. 5D). Quantitative comparisons consistently indicated significantly higher ACACA expression in malignant versus non-malignant regions (Fig. 5E). These findings underscore the spatially restricted and tumor-associated expression pattern of ACACA in ccRCC.

Fig. 5: Spatial transcriptomic profiling of ACACA in ccRCC.

A Expression of ACACA across spatial transcriptomic slices from GSE179572-GSM5420752 and GSE175540. B, C Spatial maps of four samples (KIRC, KIRC7, KIRC9, and KIRC21) showing cell-type annotation and ACACA distribution, D Correlations between ACACA expression and deconvolved cell-type programs within each sample. E Comparison of ACACA expression between malignant and non-malignant regions, with statistical significance indicated.

ACACA promoted the proliferation and migration, and reduced apoptosis in ccRCC

ACACA expression in two renal clear cell carcinoma cell lines, Caki-1 and OS-RC-2, was measured by qRT-PCR. ACACA expression was significantly higher in Caki-1 than in OS-RC-2 cells (Fig. 6A). We then transfected shRNA or an ACACA overexpression plasmid into Caki-1 and OS-RC-2 cells, respectively (Fig. 6B). CCK-8 assays showed that ACACA enhanced ccRCC cell proliferation (Fig. 6C). Transwell assays further indicated that ACACA knockdown reduced the migration of Caki-1 cells, whereas ACACA overexpression promoted the migration of OS-RC-2 cells (Fig. 6D, E). Flow cytometry revealed that ACACA silencing increased apoptosis, while ACACA overexpression decreased apoptotic cell proportions (Fig. 6F). In addition, tube formation assays demonstrated that ACACA overexpression enhanced angiogenic capacity, whereas ACACA knockdown impaired tubule formation (Fig. 6G). Collectively, these findings indicate that ACACA promotes proliferation, migration, and angiogenesis while inhibiting apoptosis in ccRCC cells.

Fig. 6: ACACA promoted the proliferation and migration and reduced apoptosis in ccRCC.

A Expression of ACACA in Caki-1 and OS-RC-2 cells (n = 3), B Expression of ACACA in Caki-1 cells transfected with sh-ACACA and OS-RC-2 cells transfected with the ACACA overexpressing plasmid (n = 3), C CCK8 assays used to detect cell proliferation for Caki-1 and OS-RC-2 cells (n = 3). D, E Transwell experiment used to detect cell migration for Caki-1 and OS-RC-2 cells (magnification: ×200, Bar = 100 μm) (n = 3). F Flow cytometry used to detect cell apoptosis for Caki-1 and OS-RC-2 cells. G Tube formation assays used to evaluate angiogenesis for Caki-1 and OS-RC-2 cells (magnification: ×100, Bar = 200 μm) (n = 3).

ACACA reduces the accumulation of R-loops and DNA damage in ccRCC

Dysregulation of R-loop homeostasis has been implicated in the progression of various tumors. The immunofluorescence (IF) stainings of S9.6 and γH2AX were used to assess the accumulation of the R structural ring and DNA damage in ccRCC. The results showed that silencing of ACACA led to an increase in R-loops, and upregulation of ACACA reduced R-loop accumulation (Fig. 7A). Also, the fluorescence intensity of γH2AX was higher in Caki-1 cells transfected with shRNA of ACACA, and ACACA overexpression in OS-RC-2 cells decreased the γH2AX expression (Fig. 7B). These suggested that ACACA might promote proliferation and migration by reducing R-loop accumulation and its associated DNA damage.

Fig. 7: ACACA reduces the accumulation of R-loops and DNA damage in ccRCC.

A IF staining of S9.6 in Caki-1 and OS-RC-2 cells (n = 3) (magnification: ×200, Bar = 100 μm), B IF staining of γH2AX in Caki-1 and OS-RC-2 cells (n = 3) (magnification: ×200, Bar = 100 μm).

ACACA enhances fatty acid metabolism and improves mitochondrial function in ccRCC

To investigate the role of ACACA in lipid metabolism of clear cell renal cell carcinoma (ccRCC), we performed Bodipy staining to visualize cytosolic lipid droplets and measured cellular levels of fatty acids and triglycerides. Knockdown of ACACA in Caki-1 cells led to a decrease in lipid droplets compared to the negative control (NC), whereas overexpression of ACACA in OS-RC-2 cells resulted in a substantial increase in lipid droplet accumulation (Fig. 8A). Furthermore, ACACA silencing significantly reduced fatty acid and triglyceride levels in Caki-1 cells, while its overexpression elevated these levels in OS-RC-2 cells (Fig. 8B, C). Given the critical role of mitochondrial function in ccRCC metabolism and energy supply, we assessed mitochondrial membrane potential using the JC-1 assay via flow cytometry. Downregulation of ACACA in Caki-1 cells increased the proportion of depolarized mitochondria with reduced membrane potential, whereas overexpression of ACACA in OS-RC-2 cells enhanced mitochondrial membrane potential (Fig. 8D). Using the DCFH-DA fluorescent probe, we observed that ACACA knockdown induced ROS accumulation in Caki-1 cells, while its overexpression reduced ROS levels in OS-RC-2 cells (Fig. 8E). Additionally, MitoTracker staining revealed that ACACA knockdown decreased mitochondrial content in Caki-1 cells, and its overexpression increased mitochondrial number in OS-RC-2 cells (Fig. 8F). Collectively, these findings demonstrate that ACACA significantly influences fatty acid metabolism and mitochondrial function in ccRCC, suggesting its potential role in tumor energy supply.

Fig. 8: ACACA enhances fatty acid metabolism and improves mitochondrial function in ccRCC.

A Cytosolic lipid droplets were stained using BODIPY for Caki-1 and OS-RC-2 cells (magnification: ×200, Bar = 100 μm) (n = 3). B Fatty acid level of Caki-1 and OS-RC-2 cells (n = 3). C Triglyceride level of Caki-1 and OS-RC-2 cells (n = 3). D Flow cytometry of JC-1 used to detect mitochondrial membrane potential for Caki-1 and OS-RC-2 cells (n = 3). E The intracellular ROS level of Caki-1 and OS-RC-2 cells (n = 3) (magnification: ×200, Bar = 100 μm). F Mitochondrial staining with Mitotracker for Caki-1 and OS-RC-2 cells (n = 3) (magnification: ×1000, Bar = 20 μm).

ACACA promoted tumor growth by alleviating the accumulation of R-loops and DNA damage in vivo

Next, the tumor cell xenograft model was used to determine whether ACACA promotes the growth and energy metabolism of ccRCC in vivo. The results of IVIS showed that the knockdown of ACACA inhibited the growth of ccRCC in vivo (Fig. 9A). The hematoxylin and eosin (HE) staining results showed that the tumor cell density in the sh-NC group was higher compared to that of the sh-ACACA group (Fig. 9B). The immunohistochemistry (IHC) staining of Ki67 suggested that the proportion of proliferating cells was lower in the sh-ACACA group (Fig. 9C). In addition, the IF staining of S9.6 and γH2AX revealed that knockdown of ACACA induced the accumulation of R-loops and DNA damage in vivo (Fig. 9D, E). Consistent with our cellular experiments, these in vivo findings reinforce the role of ACACA in regulating tumor growth and genomic stability in ccRCC.

Fig. 9: ACACA promoted tumor growth by alleviating the accumulation of R-loops and DNA damage in vivo.

A Representative images of whole-body in vivo imaging once every five days after transplantation with Caki-1 transfected with sh-ACACA or sh-NC (n = 5), B HE staining for the tumor sample in the two groups (n = 5) (magnification: ×200, Bar = 100 μm), C IHC staining of Ki67 for the tumor sample in the two groups (n = 5) (magnification: ×200, Bar = 100 μm), D IF staining of S9.6 (n = 5) (magnification: ×200, Bar = 100 μm), E IF staining of γH2AX (n = 5) (magnification: ×200, Bar = 100 μm).

ACACA enhanced fatty acid metabolism and improved mitochondrial function in vivo

Meanwhile, the levels of mitochondrial function and lipid metabolism in the tumor samples were also evaluated. The results showed that knockdown of ACACA reduced the number of cytosolic lipid droplets and the level of fatty acid and triglycerides (Fig. 10A, B). Also, knockdown of ACACA increased the ROS level, and reduced the mitochondrial membrane potential and the number of mitochondria in vivo (Fig. 10C–E). Taken together, these findings demonstrated that knockdown of ACACA could inhibit the growth of ccRCC tumors in vivo by increasing DNA damage, reducing fatty acid metabolism, and damaging mitochondrial function.

Fig. 10: ACACA enhanced fatty acid metabolism and improved mitochondrial function in vivo.

A Cytosolic lipid droplets were stained using BODIPY for the tumor sample in the two groups (n = 5) (magnification: ×200, Bar = 100 μm). B Fatty acid and triglyceride levels for the tumor sample in the two groups (n = 5). C Flow cytometry of JC-1 used to detect mitochondrial membrane potential for the tumor sample in the two groups (n = 5). D The ROS levels for the tumor sample in the two groups (magnification: ×200, Bar = 100 μm). E Mitochondrial staining with Mitotracker for the tumor sample in the two groups (n = 5) (magnification: ×400, Bar = 50 μm).

Discussion

This study provides an integrated characterization of R-loop-associated biology in ccRCC, demonstrating that R-loop activity is elevated in advanced disease and correlates with poor clinical outcomes. Using linear Cox-based modeling across multiple cohorts, we consistently identified ACACA as the key R-loop-associated gene with strong prognostic value. Given that excessive R-loop accumulation can disrupt transcriptional homeostasis and induce replicative stress, it is plausible that R-loop dysregulation may indirectly influence metabolic rewiring, including pathways governed by ACACA, thereby linking genomic instability with lipid metabolic adaptation in ccRCC

Functional analyses showed that high ACACA expression is linked to increased cell-cycle activity, activation of DNA damage and repair pathways, and metabolic reprogramming. Single-cell and spatial transcriptomic data further revealed that ACACA is preferentially expressed in malignant cells, exhibits extensive communication with immune and endothelial populations, and is spatially enriched within tumor regions.

Placed in context, our results align with an expanding literature establishing R-loops as double-edged regulators that, when dysregulated, fuel replication stress, impede DNA repair, and foster genome instability-core engines of tumor evolution¹⁹. Recent reviews synthesize these roles across malignancies and emphasize how unresolved RNA: DNA hybrids perturb genome maintenance and oncogenic transcriptional circuits. Evidence from disease-focused studies further illustrates clinical and mechanistic relevance: in hepatocellular carcinoma, R-loop-dependent regulation by MTA2 sustains stem-like properties through chromatin-metabolic coupling²⁰. In osteosarcoma, multi-omics analyses identified PSIP1 as a key R-loop modulator whose perturbation alters DNA damage and prognosis²¹. These observations from other tumor types underscore the generality of R-loop dysregulation in cancer and support our finding that R-loop-linked programs carry prognostic and biological weight in ccRCC.

Our single-cell communication analysis revealed that malignant cells with high ACACA expression engage in extensive crosstalk with immune and endothelial compartments through lipid metabolism-related pathways. Among these, signaling routes involving the ANGPTL family emerged as representative lipid-associated mediators, consistent with their recognized functions in coordinating lipid utilization with inflammatory signaling, vascular remodeling, and tumor progression^22,23. These findings suggest that dysregulated lipid metabolism is not only a metabolic feature of malignant cells but also a means of shaping the surrounding microenvironment through paracrine communication.

The clinical implications of our findings are notable. First, R-loop-based modeling provides interpretable prognostic features, with ACACA consistently capturing a substantial component of risk across cohorts; incorporation of ACACA with clinicopathological factors such as stage, metastasis, and grade further enhances individualized survival prediction. Second, ACACA represents a druggable metabolic vulnerability. Pharmacologic ACACA inhibition, exemplified by small molecules such as ND-646, has been shown to suppress lipogenesis and exert anti-tumor effects in vivo, supporting the therapeutic potential of ACACA-targeted strategies in lipid-dependent malignancies²⁴. Given the lipid-rich phenotype of ccRCC, combinatorial approaches that integrate ACACA blockade with interventions directed at mitochondrial metabolism, redox homeostasis, or angiogenic and immune signaling may offer particular promise.

Several limitations of this study should be acknowledged. First, the bioinformatic analyses were primarily based on publicly available transcriptomic datasets, which inevitably carry heterogeneity in sample collection, processing, and clinical annotation, despite validation across multiple cohorts. Second, although we combined multi-omics profiling with in vitro and in vivo functional assays to substantiate the role of ACACA, the molecular mechanisms linking R-loop dynamics with lipid metabolism and mitochondrial pathways are still not fully understood. Third, the experimental systems employed in this work, including established cell lines and xenograft models, may not completely recapitulate the complexity of the human tumor microenvironment. Finally, the translational relevance of targeting ACACA in ccRCC has not yet been assessed in preclinical drug intervention models or clinical studies, and further work will be required to evaluate therapeutic efficacy and safety.

In conclusion, our study identifies ACACA as a central R-loop-related gene with strong prognostic relevance in ccRCC. Through integrative analyses and experimental validation, we demonstrate that ACACA links dysregulated lipid metabolism with tumor progression and microenvironmental interactions. These insights expand the current understanding of R-loop biology in kidney cancer and lay the groundwork for future mechanistic and translational research.

Methods

Data acquisition and preprocessing

A total of 1185 R-loop-related genes were obtained from the R-loopBase database (https://rloopbase.nju.edu.cn/download.jsp). Transcriptome profiles and corresponding clinical information for clear cell renal cell carcinoma (ccRCC) were obtained from the TCGA-KIRC cohort (The Cancer Genome Atlas; TCGA-KIRC, https://portal.gdc.cancer.gov/)²⁵. In addition, two independent datasets, E-MTAB-1980 (ArrayExpress; https://www.ebi.ac.uk/arrayexpress/) and GSE167573 (Gene Expression Omnibus; https://www.ncbi.nlm.nih.gov/geo/), were incorporated for external validation. Expression data were provided in TPM format. Genes with expression levels below 0.5 TPM were removed, and the remaining genes were log2-transformed. Patients lacking complete clinical or survival information were excluded from subsequent analyses.

Quantification of R-loop activity

R-loop activity was quantified using single-sample GSEA (ssGSEA) implemented in the GSVA R package (gsva function, method = “ssgsea”). Gene sets were loaded with GSEABase (getGmt), and the resulting ssGSEA scores were normalized using min-max scaling²⁶.

Clinical correlation and survival analysis

Associations between R-loop activity and clinicopathological features were assessed using the Wilcoxon rank-sum test. The prognostic significance of individual R-loop-related genes for overall survival (OS), disease-specific survival (DSS), and progression-free interval (PFI) was evaluated by univariate Cox regression. Differentially expressed genes (DEGs) were identified using the limma R package. Genes with |log2FC | > 1 and a P value < 0.05 were considered significantly differentially expressed.

Linear machine learning prognostic model development

Prognostic models were established with multiple linear machine learning approaches under the Cox proportional hazards framework, including Lasso regression, Elastic Net regression, Ridge regression, stepwise Cox regression, and CoxBoost. Hyperparameters for each algorithm were optimized by tenfold cross-validation or likelihood-based stepwise selection. Model coefficients were extracted, and non-zero weights were used to quantify the contribution of individual genes.

Risk score calculation and model evaluation

For each algorithm, an individualized risk score was computed as the linear combination of gene expression levels and their corresponding model-derived coefficients. Model performance was evaluated using time-dependent ROC curves at 1-, 3-, and 5-year intervals, and mean Area under the curve (AUC) values were compared across algorithms to identify the best-performing approach. The prognostic value of the optimal model was further assessed through Cox regression and validated across independent datasets using meta-analysis. Kaplan-Meier survival analysis with log-rank testing was conducted to compare survival outcomes between risk groups, with patients stratified into high- and low-risk cohorts according to the median risk score.

Differential expression and gene set enrichment analysis

Samples were stratified into high- and low-ACACA expression groups according to the top and bottom 30% of expression values. Differentially expressed genes were identified using the limma R package. Genes were ranked by log2 fold change for subsequent gene set enrichment analysis (GSEA). GSEA was conducted with the clusterProfiler package using the Hallmark and KEGG gene sets, and normalized enrichment scores (NES) with adjusted P values were used to assess pathway significance.

Functional state correlation analysis

To further explore functional associations, 14 cancer-related states were obtained from the CancerSEA database²⁷. Z-score-based gene set variation analysis (GSVA) was applied to calculate activity scores for each state, which were subsequently standardized. Pearson correlation analysis was performed to assess the relationship between ACACA expression and functional state scores.

Single-cell analysis

Single-cell RNA-seq data of clear cell renal cell carcinoma (ccRCC) were retrieved from the TISCH2 database, with the KIRC dataset (GSE159115) selected for analysis²⁸. Cell-type annotations, including malignant, immune, endothelial, and stromal populations, were obtained from the database. ACACA expression across cell types was visualized using UMAP and box plots, and intercellular communication was inferred with the CellChat R package to evaluate interaction number and strength, as well as outgoing and incoming signaling patterns. To explore lipid metabolism-related mechanisms, the ANGPTL signaling pathway was specifically examined, and ligand-receptor analyses were performed to identify key interactions among malignant, immune, and endothelial compartments.

Spatial transcriptomic analysis

Spatial transcriptomic data of ccRCC were retrieved from the Sparkle database (https://grswsci.top/), which provides standardized processing and curated cell-type annotations, supplemented by information from previously published studies^29,30. Representative ccRCC samples with relatively high ACACA expression from public spatial transcriptomic datasets (GSE179572 and GSE175540) were selected for downstream analyses^31,32. ACACA expression was quantified across annotated slices, with malignant and non-malignant regions defined according to Sparkle annotations. Spatial mapping was used to visualize tissue architecture and gene expression distribution. Cell-type deconvolution and correlation analyses were performed to examine the association between ACACA expression and microenvironmental components, particularly immune and endothelial cells. Differential expression testing was conducted to compare ACACA levels between malignant and non-malignant compartments.

Cell culture and transfection

The HUVEC (passage 3–5), HEK-293T (passage 10–15), Caki-1 (passage 10–20), and OS-RC-2 (passage 10–20) cell lines used in this study were obtained from Procell (Wuhan, China). All cell lines were routinely tested for mycoplasma contamination and authenticated using short tandem repeat (STR) profiling. HEK-293T cells were maintained in Dulbecco’s Modified Eagle Medium (DMEM; Gibco, USA) supplemented with 10% fetal bovine serum (FBS; Gibco, Australia) and 1% penicillin-streptomycin (P/S). Caki-1 cells, a human metastatic renal clear cell carcinoma cell line with epithelial-like morphology, were cultured in McCoy’s 5 A medium (PM150710) supplemented with 10% FBS and 1% P/S. OS-RC-2 cells, also exhibiting epithelial-like morphology and derived from human primary renal carcinoma, were maintained in RPMI-1640 medium supplemented with 10% FBS. All cell lines were incubated at 37 °C in a humidified atmosphere containing 5% CO₂, and passaged every 2–3 days upon reaching 80–90% confluence.

For RNA interference and overexpression experiments targeting ACACA, small hairpin RNAs (shRNAs) and overexpression plasmids were procured from Addgene (Beijing Zhongyuan Co.). HEK293T cells were cultured to 60-80% confluence and then transfected with the respective plasmids and packaging plasmids. After transfection with Lipofectamine 3000 (Invitrogen, USA) for 6 h, the new medium was replaced. The lentiviral supernatant was collected after 48 h. For a higher viral titer, the supernatant was concentrated by ultracentrifugation at 50,000 × g for 90 min. The sequences of shRNA and plasmids were presented in Table S1. For transduction, target cells, Caki-1 and OS-RC-2, were grown to 70% confluence and transduced with the lentiviral particles in the presence of polybrene (Shanghai Genechem, Shanghai, China) to enhance transduction efficiency. Following 48 h of transduction, cells were harvested for subsequent experiments. To establish stable cell lines, transduced cells were selected with puromycin for two weeks post-transfection. Resistant colonies were expanded, and the expression of ACACA shRNA or overexpression constructs was validated using quantitative RT-PCR (qRT-PCR) to confirm knockdown or overexpression efficiency.

RNA isolation and quantitative reverse-transcription PCR (qRT-PCR) assay

Total RNA was isolated from cultured cells using TRIzol reagent (Invitrogen). RNA concentration and purity were assessed with a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific). Complementary DNA (cDNA) was synthesized from 1 µg of total RNA using the PrimeScript RT kit (Takara, Japan) according to the manufacturer’s instructions. Quantitative real-time PCR (qRT-PCR) was performed using Power SYBR Green master mix (Yeasen, China) on a real-time thermal cycler. All reactions were run in triplicate, and mRNA expression levels were normalized to GAPDH as an internal control. Relative gene expression was calculated using the 2^−ΔΔCt method. Primer sequences are listed in Table S2.

Cell proliferation, migration, and apoptosis assays

Cell proliferation, migration, and apoptosis were assessed using Cell Counting Kit-8 (CCK-8), Transwell, and flow cytometry assays. For CCK-8 assays, cells were plated in 96-well plates at 5 × 10³ cells/well and cultured for 24–72 h. After adding CCK-8 solution (Yeasen, Shanghai, China), absorbance was measured at 450 nm. Cell migration was evaluated using 8 μm Transwell chambers (Corning, USA). 5 × 10⁴ cells in serum-free medium were placed in the upper chamber, with 10% FBS medium as a chemoattractant below. After 24 h, migrated cells were fixed, stained, and counted. Apoptosis was detected using an Annexin V-FITC/PI kit (KeyGen Biotech, Nanjing, China), followed by flow cytometry to detect the cell-apoptosis rate in each tube (BD FACSCalibur).

Tube formation assay

The tube formation assay was performed to evaluate angiogenesis in vitro using Matrigel (BD Biosciences, USA). Briefly, Matrigel was thawed overnight at 4 °C and carefully spread into 48-well plates (200 μL per well), followed by polymerization for 30 min at 37 °C. HUVEC endothelial cells were then seeded onto the gel layer and treated with conditioned medium collected from either Caki-1 or OS-RC-2 cells. After 24 h of incubation at 37 °C under 5% CO₂, tube-like structures were visualized and imaged using a SOPTOP CX40 microscope (Shanghai, China).

Immunofluorescence (IF) assays

ccRCC cells were cultured to 50–60% confluence for immunofluorescence. After transfection, ccRCC cells were washed with PBS, fixed in 4% paraformaldehyde for 15 min, and permeabilized with 0.1% Triton X-100 for 10 min. After blocking with 5% Bovine serum albumin (BSA), cells were incubated with primary S9.6 antibodies (Kerafast, USA, 1:200) or γH2AX antibodies (Sigma, USA, 1:200) overnight at 4 °C. Subsequently, cells were washed with TBST and incubated with fluorescent secondary antibodies (Cell Signaling, USA) for 2 h at room temperature, followed by DAPI counterstaining. Images were acquired using a fluorescence microscope (IX35, Olympus, Japan). The fluorescence 3D analysis was carried out using ImageJ.

Boron-dipyrromethene (BODIPY) staining of lipid droplets

To visualize neutral lipid content and cytosolic lipid droplets, live cells were labeled with BODIPY 493/503 (4,4-Difluoro-1,3,5,7,8-Pentamethyl-4-Bora-3a,4a-Diaza-s-Indacene; Invitrogen, USA) at a concentration of 1 μg/mL according to the manufacturer’s instructions. Briefly, after treatment, cells were incubated with BODIPY 493/503 in culture medium at 37 °C for 30 min, protected from light. Subsequently, cells were washed gently with PBS to remove excess dye. For imaging, samples were mounted using ProLong Gold Antifade reagent (Invitrogen) to reduce photobleaching. Images were captured using a fluorescence microscope (IX35, Olympus, Japan). The fluorescence 3D analysis was carried out using ImageJ.

Fatty acid and triglyceride levels quantification

Fatty acid and triglyceride levels of ccRCC cells or tissue were quantified using commercial assay kits (Fatty Acid Quantitation Kit and Triglyceride Quantification Kit, BioVision, USA) according to the manufacturer’s instructions. Briefly, lipids were extracted from cell or tissue samples using a chloroform: isopropanol: Triton X-100 (7:11:0.1) mixture via homogenization. Then saponify the extracted lipids with 100 μL 0.5 M KOH-methanol. For the assay, add 20 μL of the samples to a 96-well plate, mix with 180 μL working reagent, incubate at 37 °C in the dark for 30 min, and measure absorbance at 570 nm to detect fatty acid. For TG quantification, NP-40 lysates were prepared from approximately 10 million cells and heated to 80–100 °C before fluorometric analysis (Ex/Em = 535/590 nm). All measurements were normalized to total protein concentration.

ROS assay

Intracellular reactive oxygen species (ROS) levels were detected using the fluorogenic probe 2′,7′-dichlorodihydrofluorescein diacetate (DCFH-DA) (Abcam, Cambridge, UK), following the manufacturer’s protocol. Briefly, ccRCC cells were seeded into 96-well plates and allowed to adhere overnight. Cells were then incubated with 20 μM DCFH-DA in serum-free medium at 37 °C for 45 min in the dark. Following incubation, the cells were gently washed twice with PBS to remove excess probe. Fluorescence images were acquired using a fluorescence microscope (IX35, Olympus, Japan). The fluorescence 3D analysis was carried out using ImageJ.

Mitochondrial membrane potential assay and mitochondrial immunofluorescence

Mitochondrial membrane potential (ΔΨm) was assessed using the JC-1 fluorescent probe (Invitrogen, CA, USA), which exhibits potential-dependent accumulation in mitochondria. In healthy mitochondria with high membrane potential, JC-1 forms aggregates that emit red fluorescence, whereas in depolarized mitochondria, it remains in monomeric form, emitting green fluorescence. Cells were seeded in 6-well plates at a density of 3 × 10⁵ cells per well and cultured for 24 h. Subsequently, they were incubated with 5 µM JC-1 in complete medium at 37 °C for 15 min in the dark. After staining, cells were washed twice with PBS and analyzed immediately using a flow cytometer (Becton Dickinson, USA). Data were processed with FlowJo software, and the ratio of green fluorescence was calculated to evaluate mitochondrial depolarization.

In parallel, mitochondrial mass was evaluated using MitoTracker probes (Invitrogen, USA), which accumulate in mitochondria independent of membrane potential. Live cells were incubated with the probe at a recommended concentration in serum-free medium at 37 °C for 30 min, protected from light. After incubation, cells were washed gently with pre-warmed PBS to remove unbound dye. Imaging was performed using a confocal microscope (STELLARIS 5, Leica, Germany) with appropriate filter sets.

In vivo assays

Six-week-old male nude mice were obtained from Beijing Vital River Laboratory Animal Technology Co., Ltd. (Beijing, China) and randomly assigned into two groups (n = 5 per group). sh-ACACA or sh-NC Caki-1 cells (6 × 10⁶ cells/mouse, suspended in 100 µL serum-free medium) were subcutaneously injected under the armpit of the mice. Tumor growth was monitored every five days using the In Vivo Imaging System (IVIS, Xenogen Corp., USA). Prior to imaging, mice were intraperitoneally injected with D-luciferin (150 mg/kg) and anesthetized with isoflurane. Bioluminescence signals were acquired from both ventral and dorsal views with automatic exposure settings. Quantitative analysis was performed using LivingImage software (Xenogen) by measuring photon flux within standardized regions of interest (ROIs). All mice were euthanized 30 days post-inoculation, and tumor tissues were harvested for subsequent analysis.

Histological analysis

Formalin-fixed tumor samples were embedded in paraffin and sectioned at 4 μm thickness for HE staining, IHC, IF, and fluorescent probe assays. Proliferation was evaluated via Ki67 immunohistochemistry. Sections were incubated with primary antibodies followed by peroxidase-conjugated secondary antibodies (Cell Signaling, USA). Signal was developed using a DAB Peroxidase Substrate Kit (Maxin, China) and visualized under a SOPTOP CX40 microscope (Shanghai, China). Protocols for IF and fluorescent probe staining followed previously described methods.

Statistical analysis

Bioinformatic data were analyzed using R (version 4.4.1). Continuous variables were compared with the Mann–Whitney U test. Kaplan-Meier survival curves were plotted and compared via the log-rank test. For cellular and animal experimental data, statistical analysis was performed using GraphPad Prism (version 9, USA). Data are presented as mean ± standard deviation. Differences between the two groups were assessed by Student’s t test, while comparisons among multiple groups were analyzed by one-way ANOVA followed by Tukey’s post hoc test. ImageJ software was used for image quantification and for analyzing the immune fluorescence in 3D images. The p-value of less than 0.05 was considered statistically significant.