Integrated network toxicology, transcriptomic, and Mendelian randomization analyses explore potential molecular links between acrylamide exposure and clear cell renal cell carcinoma

Abstract

Acrylamide is a potential carcinogen widely encountered in occupational settings and heat-processed foods. However, the potential molecular relationships between acrylamide-associated targets and clear cell renal cell carcinoma remain poorly characterized. The toxicological characteristics of acrylamide were predicted using ProTox-3.0 and ADMETlab 3.0. Potential acrylamide-associated targets were identified using SwissTargetPrediction, the Similarity Ensemble Approach, and TargetNet and were intersected with clear cell renal cell carcinoma (ccRCC)-related genes obtained from GeneCards. Protein–protein interaction networks and functional enrichment analyses were performed using Search Tool for the Retrieval of Interacting Genes/Proteins, Cytoscape, and bioinformatics approaches to characterize the biological functions of the shared targets. Using The Cancer Genome Atlas Kidney Renal Clear Cell Carcinoma transcriptomic data, least absolute shrinkage and selection operator-Cox regression was applied to identify prognostic genes and construct a risk-prediction model. Two-sample Mendelian randomization analysis was subsequently conducted to evaluate associations between genetically predicted candidate-gene expression and kidney cancer risk. Representative immunohistochemical images from the Human Protein Atlas were examined as qualitative supporting evidence. Acrylamide was predicted to have potential carcinogenic properties. Among 161 putative acrylamide-associated targets and 1235 ccRCC-related genes, 50 overlapping targets were identified and were mainly enriched in chemical stress, oxidative stress, hypoxia-related processes, phosphoinositide 3-kinase–protein kinase B, hypoxia-inducible factor 1 (HIF-1), FoxO, and central carbon metabolism pathways. Least absolute shrinkage and selection operator identified 8 prognostic genes: TP53, EGFR, PTEN, IL6, MTOR, SLC2A1, SCARB1, and IKBKB. The internally assessed 1-, 3-, and 5-year area under the curves were 0.688, 0.647, and 0.698, indicating moderate apparent discrimination. SLC2A1 was significantly upregulated in ccRCC, and genetically predicted higher SLC2A1 expression was associated with increased kidney cancer risk (odds ratio = 1.762, 95% confidence interval: 1.247–2.491, raw P = .001, false discovery rate-adjusted P = .006). TP53 showed a less robust inverse association (odds ratio = 0.506, 95% confidence interval: 0.291–0.881, raw P = .016, false discovery rate-adjusted P = .048), with evidence of directional horizontal pleiotropy. Representative Human Protein Atlas images provided qualitative support for the transcriptomic findings. This integrative analysis identified SLC2A1 as a prioritized candidate at the intersection of predicted acrylamide-associated targets and ccRCC-related biological processes. The convergent network, transcriptomic, and genetic findings provide a hypothesis for further investigation but do not establish that SLC2A1 or TP53 mediates an effect of acrylamide on renal carcinogenesis.

1. Introduction

Renal cell carcinoma (RCC) is the most common malignant tumor of the adult kidney, and clear cell renal cell carcinoma (ccRCC) represents the predominant histological subtype, accounting for approximately 75% to 80% of RCC cases.^[1] In addition to established risk factors such as cigarette smoking, obesity, and hypertension, occupational and environmental chemical exposures may contribute to renal carcinogenesis, although the available epidemiological and mechanistic evidence remains incomplete.^[2,3]

The therapeutic landscape of advanced RCC has changed substantially with the introduction of immune checkpoint inhibitors, tyrosine kinase inhibitors, and their combinations. Immune-based combinations have become important 1st-line treatment options, although their safety profiles differ from those of targeted-agent monotherapy.^[4] Recent real-world investigations have further demonstrated the clinical activity of immune-oncology combinations and pembrolizumab plus lenvatinib in patients with metastatic RCC.^[5,6] Nevertheless, therapeutic effectiveness remains heterogeneous across prognostic subgroups, and both intrinsic and acquired resistance to immune checkpoint inhibitors and tyrosine kinase inhibitors continue to restrict the duration of clinical benefit.^[7,8] In addition, immune-based treatments may be accompanied by clinically relevant adverse events, including infusion-related reactions and electrolyte disturbances such as hyponatremia, emphasizing the need to balance treatment efficacy with long-term safety.^[4,9]

Acrylamide is a low-molecular-weight organic compound primarily used in the production of polyacrylamide and is extensively applied in various industrial processes, including paper manufacturing, mineral processing, petroleum extraction, textile production, and dye manufacturing. Occupational populations may be exposed to acrylamide through inhalation, dermal contact, or accidental ingestion.^[10] Moreover, acrylamide is generated during high-temperature processing of foods rich in asparagine and reducing sugars through Maillard reactions, making fried potatoes, potato chips, baked cereals, and coffee major dietary sources of exposure in the general population.^[11] Epidemiological investigations have suggested a possible association between acrylamide exposure and cancer risk. For example, a large prospective cohort study from the Netherlands reported that higher dietary acrylamide intake was associated with altered cancer risk; however, the biological mechanisms underlying these observations remain unclear.^[12] Furthermore, long-term carcinogenicity studies in rodents have demonstrated that chronic acrylamide exposure is associated with increased tumor incidence in multiple tissues, supporting its potential carcinogenic properties.^[13]

Network toxicology can integrate predicted compound-associated targets, disease-related genes, protein–protein interactions (PPIs), and enriched pathways to generate experimentally testable hypotheses.^[14] When integrated with transcriptomic profiles and clinical outcomes from patients with ccRCC, this approach enables further evaluation of candidate molecules regarding their expression patterns and prognostic relevance in tumor tissues. In addition, expression quantitative trait locus (eQTL)-based Mendelian randomization (MR) analysis can be applied to investigate the genetic associations between candidate-gene expression and disease susceptibility, potentially reducing the influence of confounding factors and reverse causation under specific assumptions.^[15]

Therefore, this study integrated network toxicology, The Cancer Genome Atlas Kidney Renal Clear Cell Carcinoma (TCGA-KIRC) transcriptomic and survival data, eQTL-based MR analysis, and qualitative assessment of representative Human Protein Atlas (HPA) staining patterns to identify candidate molecular targets potentially linking acrylamide-associated molecular alterations with ccRCC-related biological processes. Our findings provide a systematic framework for exploring the potential relationship between acrylamide exposure and ccRCC and may contribute to the identification of candidate biomarkers for further experimental investigation.

2. Materials and methods

2.1. Prediction of the toxicological characteristics of acrylamide

The Simplified Molecular Input Line Entry System structure and 2-dimensional chemical structure of acrylamide were obtained from the PubChem database^[16] and submitted to ProTox-3.0^[17] and ADMETlab 3.0^[18] (accessed on May 24, 2026) for toxicity prediction. Predicted toxicological endpoints, including organ toxicity, toxicity classification, carcinogenicity, genotoxicity, and neurotoxicity, were collected. For categorical prediction models, the predicted classification and corresponding probability scores were recorded according to the thresholds defined by each platform. For continuous probability-based models, potential toxicity risks were determined based on the recommended cutoff values provided by the respective databases. The overall toxicological profile of acrylamide was evaluated by integrating the prediction results from multiple platforms.

2.2. Identification of potential acrylamide-associated targets

The chemical structure of acrylamide was submitted to SwissTargetPrediction,^[19] Similarity Ensemble Approach,^[20] and TargetNet^[21] (accessed on May 24, 2026) to predict potential molecular targets. The analysis was restricted to Homo sapiens. Candidate targets were retained according to the recommended criteria of each platform and subsequently standardized into official gene symbols using the UniProt database.^[22] After merging and removing duplicate entries, the final set of potential acrylamide-associated targets was obtained. These target-prediction platforms primarily infer potential protein interactions based on the structure of the parent compound and existing ligand–target information. Therefore, the resulting targets were considered putative acrylamide-associated targets rather than experimentally confirmed binding partners.

2.3. Identification of RCC-associated genes

RCC-associated genes were retrieved from the GeneCards database using the keywords “renal cell carcinoma” and “kidney renal clear cell carcinoma” (accessed on May 24, 2026).^[23] Genes were selected according to the relevance score provided by GeneCards, and the top 50% of ranked genes were retained for subsequent analyses. Duplicate genes were removed to generate the RCC-related gene set. The intersection between acrylamide-associated targets and RCC-related genes was defined as acrylamide–RCC overlapping targets. The top 50% cutoff was used as a pragmatic filtering strategy to reduce low-relevance database entries rather than as a biologically validated threshold.

2.4. Construction of the PPI network

The overlapping targets were uploaded to the Search Tool for the Retrieval of Interacting Genes/Proteins database to construct a PPI network. The organism was restricted to H sapiens, and the minimum interaction confidence score was set to 0.4. Isolated nodes were excluded from further analysis.^[24] The resulting interaction network was imported into Cytoscape software for visualization, and NetworkAnalyzer was applied to calculate topological parameters, including degree centrality. Node size and color were mapped according to network connectivity, with higher centrality values indicating stronger interactions with other nodes.

2.5. Functional enrichment analyses

Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were performed using the clusterProfiler package in R software (version 4.4.0; R Foundation for Statistical Computing). Gene identifiers were converted using the org.Hs.eg.db annotation package. Multiple-testing correction was performed using the Benjamini–Hochberg method. Enriched terms with an adjusted P value <.05 were considered statistically significant. The top 10 significantly enriched GO terms and KEGG pathways were selected for visualization.

2.6. Acquisition and processing of TCGA-KIRC data

RNA sequencing expression profiles and corresponding clinical information of patients with kidney renal clear cell carcinoma were obtained from the TCGA-KIRC database. Samples lacking complete overall survival time, survival status, or essential clinical information were excluded. The processed expression matrix was used for subsequent prognostic analyses.

2.7. Identification of prognostic genes and construction of the risk score model

Least absolute shrinkage and selection operator (LASSO)-Cox regression analysis was performed using the glmnet package in R to identify prognostic genes among the candidate targets. The optimal penalty parameter (λ) was determined by 10-fold cross-validation.^[25] The risk score for each patient was calculated using the following formula: Risk score = Σ(βi × Expi), where βi represents the LASSO-Cox regression coefficient of gene i, and Expi represents the corresponding gene expression level. Patients were stratified into high-risk and low-risk groups according to the median risk score. Kaplan–Meier survival analysis was performed to evaluate differences in overall survival between groups, and statistical significance was assessed using the log-rank test. Time-dependent receiver operating characteristic (ROC) curves were generated using the timeROC package to evaluate predictive performance at 1-, 3-, and 5-year intervals. Calibration curves were further generated to assess the agreement between predicted survival probabilities and observed outcomes.^[26] The prognostic signature was developed and evaluated using the same TCGA-KIRC cohort. No independent training and validation cohorts were available. Therefore, the time-dependent ROC and calibration analyses were considered internal and exploratory assessments of apparent model performance rather than evidence of external validity or clinical utility. The incremental prognostic value of the signature beyond established clinical factors was not evaluated.

2.8. Differential expression analysis of candidate genes

The expression levels of genes included in the prognostic model were compared between ccRCC tumor tissues and normal kidney tissues from the TCGA-KIRC dataset. Differences between groups were evaluated using the Wilcoxon rank-sum test, with P < .05 considered statistically significant.

2.9. Two-sample MR analysis

To evaluate the potential causal associations between genetically predicted expression levels of the candidate genes and kidney cancer risk, a cis-eQTL-based 2-sample MR analysis was performed. Genetically predicted gene expression, rather than acrylamide exposure itself, was considered the exposure; therefore, the MR analysis was used to support candidate-gene prioritization rather than to estimate the causal effect of acrylamide exposure.

Cis-eQTL summary statistics for the candidate genes were obtained from the eQTLGen Consortium.^[27] Single-nucleotide polymorphisms (SNPs) associated with gene expression at genome-wide significance (P < 5 × 10⁻⁸) and located within ±500 kb of the corresponding gene were selected as candidate instrumental variables. To ensure independence among instruments, linkage disequilibrium clumping was performed using a 10,000 kb window and an R² threshold of 0.1, based on the European reference panel. Instrument strength was assessed using the F statistic, calculated as \(F = [R²/(1 − R²)](N − 2)\), and only SNPs with F > 10 were retained to minimize weak-instrument bias. Detailed information for all included SNPs, including their corresponding F statistics, is provided in Table S1, Supplemental Digital Content 1.

Summary-level genome-wide association study data for kidney cancer were obtained from the FinnGen Consortium (OpenGWAS ID: finn-b-C3_KIDNEY_NOTRENALPELVIS_EXALLC). Exposure and outcome datasets were harmonized by aligning effect alleles using the harmonise_data function in the TwoSampleMR package. Allele-frequency information was used to resolve palindromic variants where possible, whereas ambiguous variants were excluded. Because both the eQTLGen and FinnGen datasets were derived predominantly from individuals of European ancestry, substantial ancestry mismatch was minimized. Based on publicly available cohort descriptions, substantial sample overlap between the exposure and outcome datasets was considered unlikely.

The inverse-variance weighted (IVW) method was used as the primary MR approach. MR-Egger regression, weighted median, weighted mode, and simple mode methods were additionally applied, where appropriate, to assess the robustness of the estimates. Cochran Q test was used to assess heterogeneity, whereas the MR-Egger intercept and MR Pleiotropy RESidual Sum and Outlier (MR-PRESSO) global test were used to evaluate potential horizontal pleiotropy. Leave-one-out and single-SNP analyses were further performed to assess whether the overall estimates were disproportionately driven by individual variants. All analyses were conducted using the TwoSampleMR and MR-PRESSO packages in R, with a 2-sided P < .05 considered statistically significant.

Because 6 candidate genes were evaluated in the primary MR analysis, multiple-testing correction was applied to the P values derived from the IVW estimates. The Benjamini–Hochberg false discovery rate (FDR) method was used as the primary correction approach to control for multiple comparisons across candidate genes. In addition, Bonferroni correction was performed as a more conservative sensitivity analysis. An FDR-adjusted P value <.05 was considered statistically significant after multiple-testing correction.

2.10. Qualitative assessment using the HPA database

Representative immunohistochemical images of the prioritized candidate genes were obtained from the HPA database.^[28] Staining patterns in representative normal kidney and KIRC tissues were compared descriptively to assess their qualitative consistency with the transcriptomic findings. Because the displayed samples were limited, nonpaired, and not subjected to standardized quantitative analysis in the present study, the HPA images were considered qualitative supporting evidence rather than independent protein-level validation.

3. Results

3.1. Predicted toxicological characteristics of acrylamide

ProTox-3.0 analysis predicted acrylamide as an active carcinogenic compound with a confidence score of 0.89. In addition, potential neurotoxicity and blood–brain barrier permeability signals were predicted (Fig. 1A and B). Consistently, ADMETlab 3.0 predicted a carcinogenicity score of 0.714 and indicated potential risks related to neurotoxicity, respiratory toxicity, and skin sensitization (Fig. 1C). These computational predictions suggest that acrylamide possesses multiple potential toxicological properties requiring further investigation.

Figure 1.

Predicted toxicological characteristics of acrylamide. (A) Radar plot showing the predicted toxicological profiles of acrylamide obtained from ProTox-3.0, including potential toxicity-related endpoints. (B) Detailed prediction results from ProTox-3.0, including organ toxicity, toxicity endpoints, carcinogenicity, genotoxicity, neurotoxicity, and blood–brain barrier permeability, together with the corresponding prediction probabilities. (C) Toxicological prediction profile of acrylamide generated by ADMETlab 3.0, including carcinogenicity, skin sensitization, respiratory toxicity, drug-induced neurotoxicity, and other toxicity-related parameters.

3.2. Identification of acrylamide–RCC overlapping targets

After removal of duplicate entries, 161 potential acrylamide-associated targets and 1235 RCC-related genes were identified. Intersection analysis revealed 50 overlapping targets between acrylamide-associated targets and RCC-related genes (Fig. 2A). These shared targets were subsequently subjected to PPI network construction and functional enrichment analyses.

Figure 2.

Identification and functional characterization of acrylamide–RCC overlapping targets. (A) Venn diagram showing the intersection between potential acrylamide-associated targets and RCC-related genes. A total of 50 overlapping targets were identified. (B) Protein–protein interaction (PPI) network of the overlapping targets constructed using the STRING database and visualized with Cytoscape. Node size and color represent the degree of connectivity, with larger and darker nodes indicating higher network centrality. (C) Gene Ontology (GO) enrichment analysis of the overlapping targets, including biological process (BP), cellular component (CC), and molecular function (MF) categories. (D) Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of the overlapping targets. The enriched pathways mainly included cancer-related signaling pathways, PI3K–Akt, HIF-1, FoxO signaling, and central carbon metabolism in cancer. HIF-1 = hypoxia-inducible factor 1, PI3K–Akt = phosphoinositide 3-kinase–protein kinase B, RCC = renal cell carcinoma, STRING = Search Tool for the Retrieval of Interacting Genes/Proteins.

3.3. Construction of the PPI network

A PPI network was constructed using the 50 overlapping targets through the Search Tool for the Retrieval of Interacting Genes/Proteins database. Topological analysis demonstrated that TP53 exhibited the highest connectivity, while EGFR, PTEN, HSP90AA1, IL6, and MTOR were also located in central positions within the network. In addition, PIK3CA, PIK3CB, RELA, IKBKB, BRAF, PARP1, PTGS2, and HDAC1 showed extensive interactions with multiple network nodes (Fig. 2B).

3.4. Functional enrichment analysis of overlapping targets

To explore the potential biological functions of acrylamide–RCC overlapping targets, GO and KEGG enrichment analyses were performed. GO biological process analysis revealed that these targets were primarily enriched in processes related to chemical stress responses, oxidative stress responses, responses to external stimuli, and hypoxia-related processes. Cellular component analysis indicated enrichment in structures including the basement membrane, membrane raft, membrane microdomain, cell basal region, and midbody. Molecular function analysis showed enrichment in transcription factor binding, ubiquitin protein ligase binding, protein serine/threonine kinase activity, histone deacetylase binding, and p53 binding (Fig. 2C).

KEGG pathway analysis demonstrated significant enrichment in cancer-associated pathways, including phosphoinositide 3-kinase–protein kinase B, HIF-1, and FoxO signaling pathways, as well as renal cell carcinoma, central carbon metabolism in cancer, and EGFR tyrosine kinase inhibitor resistance pathways (Fig. 2D).

3.5. Construction and evaluation of the prognostic risk model

LASSO-Cox regression identified 8 prognostic genes, including TP53, EGFR, PTEN, IL6, MTOR, SLC2A1, SCARB1, and IKBKB (Fig. 3A and B). According to the median risk score, 529 patients were stratified into high-risk (n = 264) and low-risk groups (n = 265). Kaplan–Meier survival analysis demonstrated that patients in the high-risk group exhibited significantly poorer overall survival compared with those in the low-risk group (P < .001; Fig. 3C).

Figure 3.

Construction and evaluation of the prognostic risk model based on acrylamide-associated candidate genes. (A) LASSO coefficient profiles of candidate genes used for prognostic model construction. (B) Selection of the optimal penalty parameter (λ) by 10-fold cross-validation in the LASSO-Cox regression model. (C) Kaplan–Meier survival analysis comparing overall survival between high-risk and low-risk groups stratified according to the median risk score. (D) Time-dependent receiver operating characteristic (ROC) curves evaluating the predictive performance of the prognostic model for 1-, 3-, and 5-year overall survival. (E) Calibration curves assessing the consistency between predicted and observed survival probabilities at 1-, 3-, and 5-year time points. (F) Differential expression analysis of the 8 prognostic genes between ccRCC tumor tissues and normal kidney tissues from the TCGA-KIRC dataset. Statistical significance was determined using the Wilcoxon rank-sum test. ccRCC = clear cell renal cell carcinoma, LASSO = least absolute shrinkage and selection operator, OS = overall survival, TCGA-KIRC = The Cancer Genome Atlas Kidney Renal Clear Cell Carcinoma.

The time-dependent ROC analysis showed that the area under the curve values for predicting 1-, 3-, and 5-year overall survival were 0.688, 0.647, and 0.698, respectively (Fig. 3D). Calibration curves indicated reasonable consistency between predicted survival probabilities and observed outcomes (Fig. 3E). Collectively, the risk model demonstrated moderate prognostic discrimination for stratifying ccRCC patients according to survival risk.

3.6. Expression patterns of risk model genes in ccRCC tissues

The expression patterns of the 8 risk model genes were further compared between ccRCC tissues and normal kidney tissues. TP53, EGFR, SLC2A1, SCARB1, and IKBKB were significantly upregulated in tumor tissues, whereas MTOR expression was significantly reduced. No significant differences were observed for PTEN and IL6 (Fig. 3F).

Among these genes, SLC2A1 and SCARB1 exhibited relatively prominent expression differences, suggesting their potential involvement in the metabolic characteristics of ccRCC.

3.7. MR analysis of candidate genes and kidney cancer risk

Two-sample MR analysis was performed to investigate associations between genetically predicted candidate-gene expression and kidney cancer risk. The IVW analysis showed that genetically predicted higher TP53 expression was associated with a lower risk of kidney cancer (odds ratio = 0.506, 95% confidence interval: 0.291–0.881, P = .016), whereas genetically predicted higher SLC2A1 expression was associated with an increased risk of kidney cancer (odds ratio = 1.762, 95% confidence interval: 1.247–2.491, P = .001). To account for multiple testing across the 6 candidate genes, the IVW-derived P values were further adjusted using the Benjamini–Hochberg FDR method. After FDR correction, the association for SLC2A1 remained statistically significant (FDR-adjusted P = .006), and the inverse association for TP53 also remained significant (FDR-adjusted P = .048). Under the more conservative Bonferroni correction, only SLC2A1 remained significant (adjusted P = .006), whereas the association for TP53 no longer reached statistical significance (adjusted P = .096). For TP53, the MR-Egger intercept test indicated potential directional horizontal pleiotropy (P = .034), although neither Cochran Q test (P = .061) nor the MR-PRESSO global test (P = .089) provided significant evidence of heterogeneity or overall pleiotropic outliers. Therefore, the inverse association observed for TP53 should be interpreted cautiously. No statistically significant associations were observed between genetically predicted expression of SCARB1, PTEN, MTOR, or IKBKB and kidney cancer risk after multiple-testing correction (Fig. 4 and Table 1).

Figure 4.

Mendelian randomization analysis of candidate genes associated with kidney cancer risk. Forest plot showing the results of 2-sample Mendelian randomization (MR) analyses evaluating associations between genetically predicted candidate-gene expression and kidney cancer risk. The inverse-variance weighted (IVW) method was used as the primary analysis approach. Odds ratios (ORs) and corresponding 95% confidence intervals (CIs) are presented for TP53, SLC2A1, SCARB1, PTEN, MTOR, and IKBKB. The FinnGen outcome represented kidney cancer and was not restricted to histologically confirmed ccRCC. ccRCC = clear cell renal cell carcinoma, FDR = false discovery rate, nSNP = number of single-nucleotide polymorphism.

Table 1.

Mendelian randomization analysis of candidate genes associated with kidney cancer risk.

Exposure	Method	nSNPs	β	SE	Pval	OR	LCI95	UCI95	Q_Pval	Presso	Ple	BH-FDR	Bonferroni
TP53	MR-Egger	7	−1.591	0.372	.008	0.204	0.098	0.422	.602	.089	.034
	Weighted median	7	−0.481	0.295	.103	0.618	0.347	1.102
	Inverse-variance weighted	7	−0.681	0.283	.016	0.506	0.291	0.881	.061			0.048	0.096
	Simple mode	7	−0.136	0.488	.790	0.873	0.335	2.273
	Weighted mode	7	−0.348	0.458	.476	0.706	0.288	1.733
SLC2A1	MR-Egger	7	−0.190	0.863	.835	0.827	0.152	4.494	.724	.754	.412
	Weighted median	7	0.669	0.228	.003	1.953	1.249	3.053
	Inverse-variance weighted	7	0.567	0.177	.001	1.762	1.247	2.491	.724			0.006	0.006
	Simple mode	7	0.681	0.330	.085	1.975	1.034	3.773
	Weighted mode	7	0.698	0.300	.059	2.009	1.117	3.616
SCARB1	MR-Egger	55	−0.156	0.087	.077	0.855	0.722	1.013	.229	.120	.024
	Weighted median	55	−0.037	0.059	.527	0.964	0.859	1.081
	Inverse-variance weighted	55	0.019	0.043	.659	1.019	0.936	1.110	.120			0.791	1.000
	Simple mode	55	−0.044	0.091	.627	0.957	0.800	1.143
	Weighted mode	55	−0.044	0.060	.463	0.957	0.850	1.076
PTEN	MR-Egger	8	−0.306	1.308	.823	0.736	0.057	9.566	.956	.975	.655
	Weighted median	8	0.284	0.284	.317	1.329	0.762	2.318
	Inverse-variance weighted	8	0.299	0.234	.200	1.349	0.853	2.133	.971			0.400	1.000
	Simple mode	8	0.297	0.460	.539	1.345	0.547	3.311
	Weighted mode	8	0.305	0.434	.506	1.356	0.579	3.179
MTOR	MR-Egger	24	−0.037	0.154	.811	0.963	0.712	1.303	.836	.895	.787
	Weighted median	24	−0.037	0.092	.687	0.964	0.805	1.153
	Inverse-variance weighted	24	0.001	0.068	.994	1.001	0.876	1.142	.869			0.994	1.000
	Simple mode	24	0.029	0.143	.841	1.029	0.778	1.363
	Weighted mode	24	−0.036	0.096	.709	0.964	0.799	1.164
IKBKB	MR-Egger	24	0.095	0.214	.663	1.099	0.723	1.672	.712	.797	.888
	Weighted median	24	0.048	0.105	.648	1.049	0.855	1.287
	Inverse-variance weighted	24	0.066	0.079	.403	1.068	0.915	1.248	.762			0.605	1.000
	Simple mode	24	0.037	0.167	.827	1.038	0.748	1.439
	Weighted mode	24	0.045	0.138	.748	1.046	0.797	1.372

The table summarizes the results of 2-sample Mendelian randomization (MR) analyses evaluating the associations between genetically predicted expression of candidate genes and kidney cancer risk. The inverse-variance weighted (IVW) method was considered the primary MR analysis, whereas MR-Egger, weighted median, simple mode, and weighted mode methods were used as complementary sensitivity analyses. Multiple-testing correction was applied to the IVW-derived P values across the 6 candidate genes using both the Benjamini–Hochberg false discovery rate (BH-FDR) method and Bonferroni correction. An FDR-adjusted P value <.05 was considered statistically significant. Cochran Q test was used to assess heterogeneity among instrumental variables, the MR-Egger intercept test was used to evaluate directional horizontal pleiotropy, and the MR-PRESSO global test was used to detect overall horizontal pleiotropy and potential outlier variants.

BH-FDR = Benjamini–Hochberg false discovery rate-adjusted P value, LCI95 = lower limit of the 95% confidence interval, MR = Mendelian randomization, nSNPs = number of single-nucleotide polymorphisms, OR = odds ratio, Ple = P value of the MR-Egger intercept test, Presso = P value of the MR-PRESSO global test, Pval = P value, Q_Pval = P value of Cochran Q test, SE = standard error, UCI95 = upper limit of the 95% confidence interval, β = beta coefficient.

3.8. Qualitative assessment of TP53 and SLC2A1 staining patterns

Representative immunohistochemical images from the HPA showed visually stronger nuclear TP53 staining and stronger SLC2A1 staining in the selected ccRCC tissues than in the displayed normal kidney tissues (Fig. 5). Because the images were obtained from a limited number of nonpaired samples and were not quantitatively analyzed, they provide qualitative support for the transcriptomic findings rather than independent protein-level validation.

Figure 5.

Qualitative assessment of TP53 and SLC2A1 staining patterns in representative normal kidney and ccRCC tissues. Representative immunohistochemical images of TP53 and SLC2A1 obtained from the Human Protein Atlas (HPA). Staining patterns in the displayed normal kidney and kidney renal clear cell carcinoma (KIRC) tissues were compared descriptively for qualitative consistency with the transcriptomic findings. The limited, nonpaired images do not constitute quantitative or independent protein-level validation. ccRCC = clear cell renal cell carcinoma.

4. Discussion

In this study, we integrated toxicological prediction, network toxicology, TCGA-KIRC transcriptomic analysis, MR, and qualitative assessment of representative HPA staining patterns to explore potential molecular relationships between acrylamide-associated targets and ccRCC biology. Fifty overlapping targets were identified and were mainly enriched in chemical stress responses, oxidative stress, hypoxia-related processes, and metabolic regulation. Among the prioritized genes, SLC2A1 showed increased expression in ccRCC tissues and an association between genetically predicted expression and kidney cancer risk. Collectively, these findings support SLC2A1 as a candidate for further investigation but do not demonstrate that it mediates an effect of acrylamide exposure on renal carcinogenesis.

The potential carcinogenicity of acrylamide has been primarily attributed to its reactive metabolite, glycidamide, which can induce genetic damage through DNA adduct formation. Following metabolic activation, glycidamide forms DNA adducts in experimental animal tissues, among which N7-glycidamide-guanine adducts are frequently detected and may contribute to mutagenic events.^[29,30] In addition to direct genotoxic effects, previous studies have demonstrated that acrylamide exposure may induce redox imbalance and secondary oxidative DNA damage.^[31] Consistently, the enrichment of overlapping targets in chemical stress responses, oxidative stress, and responses to exogenous stimuli observed in our study aligns with these previously reported toxicological mechanisms. Nevertheless, conventional target-prediction approaches may not fully represent the toxicological effects of acrylamide. Because SwissTargetPrediction, Similarity Ensemble Approach, and TargetNet primarily infer potential protein interactions from the structure of the parent compound and known ligand–target relationships, they may not adequately capture glycidamide-mediated covalent damage, metabolite-specific effects, exposure duration, or tissue toxicokinetics. Therefore, the overlapping targets identified in this study should be regarded as a computationally prioritized subset of potentially relevant molecular features rather than a comprehensive map of acrylamide toxicity.

SLC2A1 encodes glucose transporter 1 (GLUT1), and its potential involvement in ccRCC may be related to metabolic reprogramming under the von Hippel–Lindau (VHL)/HIF dysregulation background. VHL inactivation and subsequent HIF pathway activation represent fundamental molecular characteristics of ccRCC and contribute substantially to altered tumor metabolism.^[32] In VHL-deficient RCC models, GLUT1 expression and glucose utilization capacity are enhanced, supporting the dependence of these tumors on increased glucose uptake and glycolytic metabolism.^[33] Moreover, HIF-2α can promote tumor progression in certain VHL-deficient RCC models through regulation of c-Myc activity, although the biological functions of HIF-1α and HIF-2α are not completely identical in renal cancer.^[34] Functional studies have further shown that miR-1291 is downregulated in RCC tissues and that restoration of miR-1291 expression suppresses tumor cell proliferation through direct inhibition of SLC2A1.^[35] However, although early pharmacological studies suggested that GLUT1 inhibition could impair glucose metabolism in VHL-deficient RCC cells, subsequent studies indicated that compounds such as STF-31 may also target NAMPT, suggesting that their antitumor effects cannot be attributed solely to GLUT1 inhibition.^[33,36] Clinically, elevated GLUT1 expression has been associated with higher Fuhrman grade and poorer survival in some cohorts,^[37] whereas other studies failed to demonstrate significant associations with tumor stage, grade, or prognosis.^[38] These discrepancies indicate that the clinical significance of SLC2A1 may be influenced by cohort characteristics and tumor heterogeneity. In the present study, SLC2A1 showed increased transcriptomic expression in ccRCC tissues, and representative HPA images provided qualitatively consistent staining patterns, while genetically predicted higher SLC2A1 expression was associated with increased kidney cancer risk. These convergent findings support the prioritization of SLC2A1 as a candidate potentially relevant to acrylamide-associated molecular perturbations and RCC biology. However, because acrylamide exposure was not included in the MR analysis, whether acrylamide directly regulates SLC2A1 and whether such regulation contributes to renal carcinogenesis remain to be experimentally determined.

The role of TP53 in ccRCC is likely related to impaired p53 pathway activity rather than simple changes in expression levels. Following DNA damage, p53 induces cell cycle arrest, DNA repair, or apoptosis to prevent the expansion of genetically damaged cells.^[39] Although TP53 mutations are not among the most frequent genomic alterations in ccRCC,^[40] VHL loss-mediated HIF-2α accumulation can suppress p53 activity through HDM2 activation, whereas inhibition of HIF-2α or disruption of the HDM2–p53 interaction may restore p53 function and increase tumor cell sensitivity to stress-induced cell death.^[41] Genetic models have further demonstrated that combined Vhl and Trp53 deficiency promotes renal epithelial abnormalities, chromosomal instability, and tumorigenic transformation,^[40] while combined deletion of Vhl, Trp53, and Rb1 can induce ccRCC-like tumors in mice.^[42] Clinically, increased p53 immunostaining has been associated with unfavorable outcomes in ccRCC; however, such accumulation may reflect abnormal stabilization or functional impairment rather than enhanced tumor suppressive activity.^[43] The TP53 association should be interpreted cautiously because it was only borderline significant after FDR correction, did not survive Bonferroni correction, and showed evidence of directional horizontal pleiotropy.

Although SLC2A1 emerged as the most consistently prioritized candidate, its identification should not be equated with biomarker validation. Translation of a computationally identified candidate into a clinically useful biomarker requires several additional levels of evidence, including analytical validity, assay standardization, reproducibility across laboratories and platforms, biological specificity, validation in independent and clinically representative cohorts, prespecified thresholds, and demonstration of prognostic value beyond established clinical factors. Methodological variation in antibodies, scoring systems, cutoffs, specimen selection, and pre-analytical procedures can substantially affect biomarker interpretation.^[44] Accordingly, the transcriptomic expression difference, internally assessed prognostic model, MR association, and representative HPA images reported here support the prioritization of SLC2A1 for further study but do not establish it as a validated diagnostic or prognostic biomarker.

Several limitations should be acknowledged. First, acrylamide-associated targets were primarily derived from computational prediction and database integration, which may be affected by algorithmic limitations, database bias, and incomplete representation of glycidamide-mediated covalent mechanisms. Although multiple acrylamide target-prediction platforms were used, no individual-level data on dietary or occupational acrylamide exposure or exposure biomarkers were available. Second, the MR analysis evaluated genetically predicted candidate-gene expression rather than acrylamide exposure and may also be influenced by the tissue specificity of the eQTL resource, instrumental variable strength, horizontal pleiotropy, population ancestry, and multiple testing. In addition, the FinnGen outcome represented kidney cancer and was not restricted to histologically confirmed ccRCC; therefore, subtype heterogeneity cannot be excluded, and the MR findings should not be interpreted as ccRCC-specific evidence. Accordingly, the MR findings support candidate-gene prioritization but cannot establish that SLC2A1 or TP53 mediates an effect of acrylamide on renal carcinogenesis. Third, the prognostic model lacked independent external validation and adjustment for established clinical variables, while the HPA analysis provided only representative protein-staining patterns rather than quantitative experimental validation. Finally, no in vitro or in vivo exposure experiments were performed to directly confirm acrylamide-mediated regulation of SLC2A1, TP53, or related metabolic pathways. Future studies incorporating individual-level exposure assessment, long-term low-dose exposure models, kidney-specific genetic analyses, and SLC2A1 functional manipulation are warranted to validate these candidate molecular relationships.

5. Conclusion

This integrative study explored potential molecular relationships between acrylamide-associated targets and ccRCC-related biological processes. SLC2A1 showed relatively consistent evidence across network, transcriptomic, and genetic analyses and was, therefore, prioritized as a candidate for further investigation. However, the present findings do not establish that acrylamide causes RCC or that SLC2A1 or TP53 mediates an acrylamide-associated carcinogenic effect. Direct exposure and functional experiments are required to test this hypothesis.

Acknowledgments

We would like to express our sincere gratitude to all the staff members of the databases utilized in this study for their valuable contributions.

Author contributions

Conceptualization: Shuwen Deng.

Data curation: Lunhong Zou, Yubo Zhou.

Formal analysis: Zhiqiang Zeng, Tao Li.

Funding acquisition: Tao Zhou.

Investigation: Shuwen Deng.

Methodology: Shuwen Deng, Yubo Zhou, Lei Lei, Tao Zhou.

Software: Shuang Wang.

Writing – original draft: Shuwen Deng, Shuang Wang, Zhiqiang Zeng, Tao Li, Lunhong Zou, Yubo Zhou, Lei Lei, Tao Zhou.

Writing – review & editing: Shuwen Deng, Shuang Wang, Yubo Zhou, Lei Lei, Tao Zhou.