Abstract
Background
Environmental exposure to non-methane volatile organic compounds (NMVOCs) may contribute to chronic kidney disease (CKD), but the underlying molecular pathways remain unclear. We integrated epidemiological, transcriptomic, single-cell, and experimental analyses to identify candidate mechanisms linking NMVOCs to kidney injury.
Methods
Province-level NMVOC emission burden was linked to residential province for 3,472 participants in the China Health and Retirement Longitudinal Study (CHARLS). CKD was defined as an eGFR <60 mL/min/1.73 m2. Associations were evaluated using multivariable logistic regression. Putative NMVOC-CKD targets, kidney transcriptomic datasets, 113 machine-learning models, and SHapley Additive exPlanations (SHAP) were used to prioritize biomarkers. Single-cell transcriptomics, cell-communication analysis, virtual knockout, and hydroquinone-treated cell experiments were used to investigate candidate pathways.
Results
Among 3,472 participants, 417 had CKD. In the fully adjusted model, higher province-level NMVOC emission burden was associated with greater odds of prevalent CKD (OR = 1.05). The best transcriptomic classifier achieved areas under the curve of 0.964 and 0.836 in the training and validation sets. Integrated analyses prioritized SLC9A3 and localized its expression to proximal tubular cells. Single-cell analysis predicted SPP1-ITGAV/ITGB1 signaling between proximal tubular cells and fibroblasts. In a benzene-metabolite model, hydroquinone increased tubular SLC9A3 and SPP1 expression; conditioned medium increased fibroblast ITGB1 and α-smooth muscle actin expression and migration.
Conclusions
Province-level NMVOC emission burden was associated with CKD. The findings nominate proximal-tubule SLC9A3/SPP1-integrin signaling as a candidate link between NMVOC emission burden and fibroblast activation, warranting validation through individual-level exposure assessment and targeted perturbation studies.
Keywords: Chronic kidney disease, non-methane volatile organic compounds, SLC9A3, single-cell RNA sequencing, renal fibrosis
1. Introduction
Chronic kidney disease (CKD) is a global non-communicable public health issue, and its incidence and mortality rates have continued to rise over the past few decades [1]. Traditionally, CKD has been primarily viewed as organ damage caused by metabolic and hemodynamic disorders such as diabetes and hypertension [2]. However, with the advancement of epidemiological research, a disease model based solely on metabolic and lifestyle factors is no longer sufficient to fully explain the complex global burden of CKD.
In the field of environmental exposure, previous studies have mostly focused on heavy metals and fine particulate matter [3,4]. With global industrialization and urbanization, non-methane volatile organic compounds (NMVOCs) are gradually emerging as a new frontier in environmental nephrology [5]. NMVOCs refer to organic carbon compounds, excluding methane, that have high vapor pressure and readily evaporate at room temperature [6]. Their emission sources are extremely broad, involving fossil fuel combustion, industrial production, building materials, and daily consumer goods [7]. NMVOCs can enter the systemic circulation through inhalation, skin contact, or the digestive tract. After being metabolized in the liver by the cytochrome P450 enzyme system, their highly active metabolites induce excessive production of reactive oxygen species (ROS) by interfering with the mitochondrial electron transport chain or activating NADPH oxidase, thereby triggering oxidative stress and lipid peroxidation [8,9]. Renal proximal tubular epithelial cells highly express various solute carrier (SLC) transporters, through which NMVOC metabolites enter and accumulate locally in tubular epithelial cells, subsequently inducing mitochondrial damage, cell apoptosis, and interstitial fibrosis [10]. Previous population studies have shown that NMVOC emission burden is significantly associated with chronic kidney function decline [11].
To fill the knowledge gap between epidemiological observations and molecular toxicological mechanisms, integrated epidemiological exposure assessment, network toxicology, public bulk and single-cell transcriptomic analyses, and in vitro experiments, and in vitro experiments, aiming to explore the heterogeneity of renal microenvironmental responses to environmental toxicants and to provide scientific evidence and potential targets for early intervention in environmental toxicant-induced kidney injury.
2. Methods
2.1. Study population and environmental exposure assessment
The epidemiological data for this study were sourced from the China Health and Retirement Longitudinal Study (CHARLS) cohort, we conducted a cross-sectional analysis using participants with available kidney-function and residential-province data from the 2015 CHARLS wave [12]. The inclusion criteria for this study were subjects with complete residential province information, clear CKD outcome status, and complete covariate data. The primary outcome was CKD. The determination of the outcome was mainly based on the subject’s estimated glomerular filtration rate (eGFR). Subjects with an eGFR less than 60 mL/min/1.73m2 were clearly defined as having CKD [13,14]. Participants with missing data or aged under 45 were excluded from the analysis. Province-level NMVOC emission data for 2015 were obtained from the Multi-resolution Emission Inventory for China (MEIC; http://meicmodel.org.cn/). Each participant was assigned the emission value corresponding to their recorded province of residence, such that all participants residing in the same province received the same value. This metric was interpreted as an area-level proxy for regional NMVOC emission burden rather than a measure of personal exposure, ambient concentration, absorbed dose, or average exposure during 2011–2015. CHARLS has been approved by the Biomedical Ethics Review Committee of Peking University. All participants signed informed consent forms. The ethical approval number for CHARLS is IRB00001052-11015. Since the current study is a secondary analysis of publicly available data, additional ethical approval was not required. The authors accessed the CHARLS dataset for research purposes on 1.2026. The authors did not have access to information that could identify individual participants during or after data collection.
2.2. Epidemiological cohort association analysis
Continuous variables for the baseline characteristics of the population are presented as the median and interquartile range (Median [IQR]). Categorical variables are presented as frequencies and percentages (n, %). We compared the differences between groups using the Mann Whitney U test or the chi square test. We applied a multivariate logistic regression model to estimate the association between the province-level NMVOC emission proxy and the odds of prevalent CKD, calculating the odds ratios (OR) and 95% confidence intervals (CI). Three progressively adjusted models were created. Model 1 is the unadjusted model. Model 2 adjusted for age, gender, BMI, place of residence, education level, and marital status. Model 3 made further comprehensive adjustments for smoking, alcohol consumption, and multiple biochemical indicators. CKD was defined as an eGFR <60 mL/min/1.73 m2. In a sensitivity analysis, participants were classified as G1 (≥90), G2 (60–89), or G3+ (<60 mL/min/1.73 m2), and multinomial logistic regression was performed using G1 as the reference. These categories represented kidney-function levels rather than CKD stages. Furthermore, we carried out stratified subgroup analyses based on obesity, abnormal blood lipids, smoking, and drinking status. All epidemiological statistical analyses were performed using R version 4.2.2 under a two-sided test threshold of p < 0.05.
2.3. Target gene acquisition and functional enrichment analysis
The chemical structures and standard simplified molecular structures of the main NMVOC components (formaldehyde, benzene, isoprene, and trichloroethylene) were obtained from PubChem (https://pubchem.ncbi.nlm.nih.gov/). We identified potential targets by screening the CTD (https://ctdbase.org/), TargNet (http://targetnet.scbdd.com), and SwissTargetPrediction (http://swisstargetprediction.ch/) databases. At the same time, we used OMIM (https://omim.org/) and GeneCards (https://www.genecards.org/) to search for and filter genes related to CKD. We then found the intersection of these two gene sets and used a Venn diagram to confirm the core potential targets linking NMVOC emission burden to CKD. Finally, we used the clusterProfiler R package to conduct Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis on these intersecting genes.
2.4. PPI network construction and network topology analysis
The selected overlapping genes were submitted to the STRING database, with the species limited to Homo sapiens, to obtain data on protein interactions (PPI). This PPI data was then imported into the Cytoscape software to visualize the network. Next, a tool within Cytoscape was used to calculate the network properties of each node. The genes were sorted based on their degree values to identify the top ten hub genes. Finally, information on the compound components, shared targets, and the disease was combined to create a comprehensive regulatory network linking the compounds, targets, and the disease.
2.5. Building the machine learning diagnostic model and analyzing feature interpretability
Independent transcriptome datasets GSE37171 (training set) and GSE66494 (validation set) were obtained from the GEO database. We built and evaluated 113 combinations of machine learning algorithms in the training set. The best diagnostic model was then selected based on the overall area under the receiver operating characteristic curve (AUC) score. Next, we used the SHAP (SHapley Additive exPlanations) method to analyze the interpretability of the best model. By calculating the average absolute SHAP values, we measured the contribution of feature genes to pinpoint the core diagnostic markers. We plotted single gene ROC curves to evaluate their diagnostic performance. Finally, we verified the significant expression differences of these core genes across the two independent cohorts.
2.6. Single cell RNA sequencing and virtual knockout analysis
Single-cell transcriptome data of normal controls and CKD samples (GSE199711) were obtained. This dataset contains the transcriptome information of 2 normal kidney cell samples and 3 CKD kidney cell samples. The Seurat R package was used for data quality control, normalization, dimensionality reduction, and non-linear clustering. Quality control excluded cells with gene counts <200 or >6,000, or a mitochondrial transcript proportion > 20%. Cell types were annotated using SingleR canonical markers. The expression specificity of core genes in specific subpopulations was evaluated using FeaturePlot and violin plots. The CellChat R package was utilized to construct a global cell communication network of the microenvironment, calculate the signal sending and receiving intensity among cell subpopulations, and precisely locate the key receptor-ligand interaction axes between damaged cells (Sender) and interstitial cells (Receiver). Finally, the single-cell virtual knockout (Virtual KO) algorithm scTenifoldKN was used to simulate the loss of function of key genes. By comparing the transcriptional profiles before and after the knockout, the compensatory expression changes of downstream reabsorption receptors and injury markers were evaluated, and pathway enrichment analysis was performed on their differentially expressed genes.
2.7. Cell culture and conditioned medium preparation
Rat renal tubular epithelial NRK-52E cells and rat renal interstitial fibroblast NRK-49F cells (Wuhan Pricella Biotechnology, Wuhan, China) were cultured in DMEM supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin at 37 °C in a humidified atmosphere containing 5% CO2. To examine the predicted tubular response in a tractable in vitro system, hydroquinone (HQ; HY-B0951, MedChemExpress), a bioactive oxidative metabolite of benzene, was selected for the proof-of-concept experiments. Benzene was one of the four NMVOC components included in the upstream target-screening analysis. HQ was used as a benzene-derived model metabolite and was not intended to reproduce the composition or combined effects of real-world NMVOC mixtures. NRK-52E cells were treated with 0, 10, or 20 μmol/L HQ for 24 h. The HQ-containing medium was then completely removed, and the cells were washed three times with phosphate-buffered saline. Fresh HQ-free complete medium was added, and the culture supernatant was collected after an additional 24 h and centrifuged to remove cells and cellular debris. Conditioned medium from untreated cells was designated CM-Control, whereas conditioned medium from cells previously exposed to 10 or 20 μmol/L HQ was designated CM-HQ10 and CM-HQ20, respectively. NRK-49F cells were subsequently assigned to six groups: fresh medium containing 10% FBS without HQ or conditioned medium (FBS-Only), CM-Control, CM-HQ10, CM-HQ20, fresh medium containing 10% FBS and 10 μmol/L HQ without conditioned medium (Direct-HQ10), and fresh medium containing 10% FBS and 20 μmol/L HQ without conditioned medium (Direct-HQ20). This design distinguished fibroblast responses to epithelial-derived paracrine factors from the direct effects of residual HQ and serum-induced responses.
2.8. Cell viability assay
The cytotoxicity of HQ in NRK-52E cells was evaluated using a Cell Counting Kit-8 (CCK-8, Beyotime, China) assay. NRK-52E cells were seeded into 96-well plates at a density of 5 × 103 cells per well and treated with 0, 10, or 20 μmol/L HQ. Cell viability was assessed at 0, 24, 48, 72, and 96 h. At each time point, 10 μL of CCK-8 reagent was added to each well and incubated at 37 °C for 2 h. Absorbance was measured at 450 nm using a microplate reader.
2.9. Quantitative real-time PCR (qRT-PCR)
Total RNA was extracted from cultured NRK-52E cells using TRIzol reagent (Vazyme, China) according to the manufacturer’s instructions. After evaluating the concentration and purity of the isolated RNA, it was reverse-transcribed into cDNA using a cDNA synthesis kit (R223-01, Vazyme, China). Quantitative real-time PCR was performed using the SYBR Green master mix system. The relative mRNA expression levels of the target genes were calculated by normalizing to the endogenous control GAPDH. Primer sequences were as follows: SLC9A3: forward primer: 5′- CAGACGATGCGACTCACAG −3′, reverse primer: 5′- GGTAGTGCTCCATAAACACGC −3′; SPP1: forward primer: 5′- AGCGAGCGAGGAGTTGAATG −3′, reverse primer: 5′- GTTGTCCTCAGCTGCCATT −3′; GAPDH: forward primer 5′-CCCATGTTCGTCATGGGTGT-3′, reverse primer: 5′-TGGTCATGAGTCCTTCCACG-3′.
2.10. Western blotting
Total protein was extracted from NRK-52E and NRK-49F cells using RIPA lysis buffer supplemented with protease and phosphatase inhibitors. Protein concentrations were determined using a BCA protein assay kit. Equal amounts of protein samples were separated by SDS-PAGE and subsequently transferred onto PVDF membranes. After blocking with 5% nonfat milk for 1 h at room temperature, the membranes were incubated overnight at 4 °C with specific primary antibodies: SLC9A3 (1:1000, 33559-1-AP, Proteintech, China), SPP1 (1:1000, 22952-1-AP, Proteintech, China), ITGB1 (1:5000, 12594-1-AP, Proteintech, China), α-SMA (1:2000, 14395-1-AP, Proteintech, China), and GAPDH (1:10000, 60004-1-Ig, Proteintech, China). After washing, the membranes were incubated with HRP-conjugated secondary antibodies for 1 h. Protein bands were visualized using an ECL kit (Biosharp, China), and the relative protein expression was semi-quantified using ImageJ software, normalized to GAPDH as the loading control.
2.11. Transwell
The directed migration capacity of fibroblasts was assessed using chambers with an 8.0 μm pore size (Corning, China). NRK-49F cells were resuspended in serum-free medium and seeded into the upper chambers. The lower chambers were filled with FBS-Only medium, CM-Control, CM-HQ10, CM-HQ20, Direct-HQ10, or Direct-HQ20 to serve as chemoattractants. After incubation at 37 °C for 24 h, the non-migrated cells on the upper surface of the membrane were gently removed with a cotton swab. The cells that had migrated to the lower surface of the membrane were fixed with 4% paraformaldehyde for 20 min and stained with 0.1% crystal violet for 15 min. The migrated cells were observed and photographed under an inverted microscope (Olympus), and quantified by counting the number of cells in randomly selected fields.
3. Results
3.1. Association between province-level NMVOC emission burden prevalent and CKD
This study enrolled 3,472 participants, including 3,055 non-CKD and 417 CKD individuals (Table 1). CKD disease burden exhibited obvious geographical disparities across regions (Figure 1A, Supplementary Table S1). Baseline analyses revealed that the CKD group had higher proportions of female and less educated individuals, alongside significantly elevated regional NMVOC emission burden. Additionally, the CKD group presented poorer cardiovascular and metabolic profiles, with higher rates of obesity and dyslipidemia, as well as increased BMI, blood lipid, blood pressure, and CRP levels (p < 0.05). Multivariable logistic regression showed that higher province-level NMVOC emission burden was associated with greater odds of prevalent CKD (Figure 1B). Higher province-level NMVOC emission burden was associated with greater odds of prevalent CKD in the fully adjusted model (OR = 1.05, 95% CI: 1.03–1.08). In exploratory subgroup analyses, the association was observed among participants with obesity (Figure 1C). The significant association persisted in participants with and without dyslipidemia (Figure 1D). By contrast, smoking and drinking status did not modify the correlation, with no statistical significance observed in these subgroups (Figure 1E-F). In the fully adjusted model, higher province-level NMVOC emission burden was associated with greater odds of G2 and G3+ relative to G1 (Supplementary Table S2).
Table 1.
Baseline characteristics of the study population.
| Characteristics | Overall (N = 3472) | Non-CKD (N = 3055) | CKD (N = 417) | P |
|---|---|---|---|---|
| Age (year) | 62.0 [55.0, 68.0] | 61.0 [55.0, 68.0] | 62.0 [56.0, 68.0] | 0.088 |
| Sex | ||||
| Female | 1904 (54.8) | 1650 (54.0) | 254 (60.9) | 0.009 |
| Male | 1568 (45.2) | 1405 (46.0) | 163 (39.1) | |
| Residence | ||||
| Rural | 2322 (66.9) | 2045 (66.9) | 277 (66.4) | 0.878 |
| Urban | 1150 (33.1) | 1010 (33.1) | 140 (33.6) | |
| Marital status | ||||
| Married and living with a spouse | 2882 (83.1) | 2552 (83.6) | 330 (79.1) | 0.074 |
| Married but living without a spouse | 125 (3.6) | 106 (3.5) | 19 (4.6) | |
| Single, divorced, and widowed | 463 (13.3) | 395 (12.9) | 68 (16.3) | |
| Education status | ||||
| Elementary school or below | 2472 (71.2) | 2154 (70.5) | 318 (76.3) | 0.018 |
| Middle school or above | 1000 (28.8) | 901 (29.5) | 99 (23.7) | |
| Smoking status | ||||
| No | 1969 (56.7) | 1706 (55.8) | 263 (63.1) | 0.006 |
| Yes | 1503 (43.3) | 1349 (44.2) | 154 (36.9) | |
| Drinking status | ||||
| No | 2283 (65.9) | 1980 (65.0) | 303 (72.8) | 0.002 |
| Yes | 1179 (34.1) | 1066 (35.0) | 113 (27.2) | |
| Regional category | ||||
| East | 1164 (33.5) | 1004 (32.9) | 160 (38.4) | 0.08 |
| Midland | 1303 (37.5) | 1156 (37.8) | 147 (35.3) | |
| West | 1005 (28.9) | 895 (29.3) | 110 (26.4) | |
| Cooking fuel use | ||||
| Clean fuel | 1077 (52.2) | 943 (52.0) | 134 (53.6) | 0.693 |
| Solid fuel | 985 (47.8) | 869 (48.0) | 116 (46.4) | |
| Dyslipidemia | ||||
| No | 2599 (76.3) | 2339 (78.0) | 260 (63.4) | <0.001 |
| Yes | 809 (23.7) | 659 (22.0) | 150 (36.6) | |
| Obesity | ||||
| No | 1818 (59.3) | 1664 (61.4) | 154 (42.9) | <0.001 |
| Yes | 1250 (40.7) | 1045 (38.6) | 205 (57.1) | |
| BMI (kg/m2) | 23.6 [21.1, 26.2] | 23.4 [21.0, 25.9] | 24.9 [22.4, 27.4] | <0.001 |
| Waist (cm) | 85.8 [78.8, 93.0] | 85.2 [78.4, 92.5] | 89.4 [82.2, 96.1] | <0.001 |
| TG (mg/dl) | 115.0 [83.2, 169.9] | 112.4 [81.4, 165.5] | 142.5 [98.5, 207.7] | <0.001 |
| LDL (mg/dl) | 101.2 [82.6, 119.7] | 100.4 [82.6, 119.3] | 105.2 [83.8, 123.8] | 0.008 |
| HDL (mg/dl) | 50.2 [43.2, 57.9] | 50.6 [43.6, 58.3] | 47.9 [41.7, 55.6] | <0.001 |
| FBG (mg/dl) | 93.7 [86.5, 100.9] | 93.7 [86.5, 99.1] | 104.5 [93.7, 122.5] | <0.001 |
| SBP (mmHg) | 126.5 [114.0, 141.0] | 126.0 [113.5, 140.5] | 130.5 [117.5, 144.2] | 0.001 |
| DBP (mmHg) | 74.5 [67.5, 82.5] | 74.5 [67.0, 82.5] | 75.5 [69.5, 83.5] | 0.007 |
| CRP (mg/dl) | 1.4 [0.8, 2.6] | 1.4 [0.8, 2.6] | 1.7 [1.0, 3.0] | <0.001 |
| NMVOC (mg/m3) | 6.1 [2.9, 10.1] | 6.1 [2.9, 10.1] | 7.3 [2.8, 13.0] | 0.006 |
Note: CKD: chronic kidney disease; BMI: body mass index; TG: triglycerides; LDL: low density lipoprotein; FBG: fasting blood glucose; SBP: systolic blood pressure; DBP: diastolic blood pressure; CRP: C reactive protein; NMVOC: non-methane volatile organic compound.
Figure 1.
Spatial distribution of province-level NMVOC emission burden and its association with the odds of prevalent CKD. (A) Spatial geographic distribution map of CKD prevalence across regions. The color intensity represents the level of prevalence or exposure. (B) Multivariate logistic regression model evaluating the association between province-level NMVOC emission burden and CKD risk. (C-F) Stratified subgroup analyses of the association between province-level NMVOC emission burden and CKD according to (C) obesity status, (D) lipid status, (E) smoking status, and (F) alcohol consumption status. Model 1: Unadjusted; Model 2: Adjusted for age, gender, BMI, place of residence, educational level, and marital status; Model 3: Further adjusted for smoking, alcohol consumption, TG, HDL, LDL, CRP, SBP, and DBP based on Model 2.
3.2. Identification of NMVOC related targets in CKD
Next, to further explore candidate molecular pathways potentially linking NMVOC-related compounds to CKD, we collected the targets of the main NMVOC components (formaldehyde, benzene, isoprene, and trichloroethylene) and CKD-related genes. By intersection analysis, a total of 165 common target genes shared between NMVOCs and CKD were identified (Figure 2A). Subsequently, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were performed on these 165 intersecting targets. The results of the GO clustering analysis (Figure 2B) showed that the targets were mainly enriched in the response to oxygen levels and the response to hypoxia, and were primarily located in membrane microdomains, membrane rafts, and caveolae. At the molecular function level, they were mainly involved in protein serine/threonine/tyrosine kinase activity. Furthermore, KEGG pathway enrichment analysis revealed the key signaling pathways of these targets within the disease network (Figure 2C). The results indicated that the intersecting genes were highly enriched in apoptosis, lipid and atherosclerosis, the PI3K-Akt signaling pathway, the AGE-RAGE signaling pathway in diabetic complications, and the MAPK signaling pathway.
Figure 2.
Identification of core targets and functional enrichment analysis of province-level NMVOC emission burden driving CKD development. (A) Venn diagram illustrating the intersection between the targets of main NMVOC components (formaldehyde, benzene, isoprene, and trichloroethylene) and CKD-related genes. (B) Bar chart of Gene Ontology (GO) functional enrichment analysis for the intersecting targets. (C) Bar chart of KEGG pathway enrichment analysis for the intersecting targets.
3.3. Core target screening and regulatory network construction
To further identify the key hub genes involved in how NMVOCs affect CKD, the 165 shared genes obtained earlier were submitted to the STRING database to gather data on protein interactions (PPI). The Cytoscape software was then used to successfully build a comprehensive regulatory network linking the compounds, targets, and the disease (Figure 3A). This network diagram clearly shows how the database-derived associations linked the selected NMVOC components to candidate targets. Together, they form a complex and coordinated regulatory network. Subsequently, visualization and network analysis were performed. By calculating the degree value of each node in the network, the importance of each gene in regulating the network was evaluated, and core genes with high connectivity were selected. The top ten key hub genes, in order, are SLC9A3, MAPK1, SIRT1, PTGS2, CYP19A1, ESR1, TNF, MAOA, APP, and CDKN1A (Figure 3B). These highly connected genes likely play the most critical central role in the network that may be involved in cellular stress and kidney-disease-related pathways.
Figure 3.
Comprehensive regulatory network and screening of core hub genes. (A) Network diagram constructed using Cytoscape, illustrating the regulation of CKD by the main exposure components of NMVOCs via 165 intersecting targets. Node size and color represent their topological importance (Degree value) within the network. (B) The top 10 key hub genes screened through network topological analysis.
3.4. Construction of the machine learning diagnostic model and verification of core feature genes based on SHAP analysis
To further pinpoint the core feature genes with high clinical diagnostic value from the previously screened gene set, 113 machine learning algorithm combinations were introduced for modeling and evaluation. The GSE37171 dataset was used as the training set, and the GSE66494 dataset served as the independent external validation set. Based on the overall area under the receiver operating characteristic curve (AUC) scores, all models were evaluated and ranked (Supplementary Table S3), and a heatmap showing the overall performance of the top ten models was created (Figure 4A). The results showed that the Stepglm[both] + Lasso algorithm combination performed the best. Its AUC in the training set reached 0.964, and its AUC in the external validation set reached 0.836, achieving the highest overall score among all models (AUC = 0.900). After establishing the best black box model, the SHAP method was applied to analyze feature importance. This was done to deeply explain the decision making process of the model and to measure the diagnostic contribution of each feature gene. By combining SHAP bar plots, swarm plots, force plots, and waterfall plots (Figure 4B-E), the impact of core genes on predicting individual disease risk was clearly mapped out. Based on the ranking of the absolute average SHAP values, the top candidate feature genes were CLK4 (0.276), CDC25B (0.184), SLC9A3 (0.111), METAP2 (0.089), FDPS (0.022), and APP (0.014). These genes contributed to model discrimination in the model’s ability to distinguish between healthy and diseased samples. To evaluate the individual diagnostic performance of these core feature genes selected by the best model, an independent validation analysis was conducted using ROC curves (Figure 4F). The results indicated that all these key genes demonstrated excellent single gene diagnostic ability. Among them, CLK4 had the most outstanding AUC performance (AUC = 0.870). The AUC values for METAP2, SLC9A3, and FDPS also reached 0.791, 0.754, and 0.735.
Figure 4.
Construction of machine learning diagnostic models and evaluation of core feature genes based on SHAP analysis. (A) Heatmap of the top 10 machine learning algorithm combinations based on their comprehensive performance (AUC values) in the training set (GSE37171) and the external validation set (GSE66494). (B-E) SHAP visualization analysis of the optimal model, including a bar plot (B), beeswarm plot (C), force plot (D), and waterfall plot (E), demonstrating the contributions and weights of the core genes to the model predictions. (F) Single-gene ROC curves of the core feature genes in the independent validation, evaluating their individual diagnostic efficacy.
3.5. Validation of differential expression of core feature genes
We validated five machine learning-identified core genes in two independent CKD transcriptomic datasets (GSE37171 and GSE66494). SLC9A3 was significantly upregulated, while CLK4 and CDC25B were downregulated in CKD samples with consistent results. In contrast, METAP2 and FDPS showed inconsistent expression trends (Figure 5). These findings verify that elevated SLC9A3 expression facilitates CKD progression.
Figure 5.
Validation of the differential expression of the 5 core feature genes in external transcriptome cohorts.
3.6. Single-Cell analysis suggests the involvement of SLC9A3/SPP1-ITGB1 signaling
We analyzed a publicly available single-cell RNA-seq dataset comprising healthy control and CKD kidney samples. Through non-linear dimensionality reduction clustering (UMAP), we successfully identified 10 major cell subpopulations (Figure 6A). Feature distribution plots, indicating the specific expression of SLC9A3 in proximal tubules and its significant upregulation in CKD, showed that the expression of SLC9A3 was highly and specifically enriched in proximal tubular cells (Figure 6B). CellChat analysis identified SPP1 as a prominent inferred signaling pathway (Figure 6C). Ligand-receptor analysis further predicted SPP1-ITGAV/ITGB1 signaling from proximal tubular cells to fibroblasts (Figure 6D). Virtual-knockout simulation predicted altered receptor expression following simulated SLC9A3 loss, with ITGB1 showing the strongest predicted compensatory increase (Z-score = 6.24; Figure 6E). Other examined receptors showed smaller predicted changes, and directional analysis indicated a consistent pattern in the predicted pathway activities (Figure 6F).
Figure 6.
Single-cell analysis and virtual knockout-based investigation of SLC9A3-associated microenvironmental signaling in CKD. (A) UMAP dimensionality reduction clustering plot of the global cell landscape in human healthy control and CKD kidney samples, identifying a total of 10 major cell subpopulations. (B) Highly specific expression enrichment of the SLC9A3 gene in proximal tubular cells. (C) Signaling pathway analysis within the CKD microenvironment, showing significant enrichment of the SPP1 pathway. (D) Cell communication analysis demonstrating the targeted interaction between the SPP1 ligand secreted by proximal tubular cells and the integrin receptor complex (ITGAV+ITGB1) on the surface of fibroblasts. (E, F) Directional analysis of predicted pathway activities and compensatory expression changes following targeted perturbation of the SPP1 signaling axis receptors.
3.7. In vitro experimental validation
To examine whether a benzene-derived model metabolite could induce cellular responses consistent with the predicted tubular-fibroblast signaling pathway, NRK-52E cells were treated with HQ. Treatment with 10 or 20 μmol/L HQ did not significantly alter NRK-52E cell viability during the 96-h observation period (Figure 7A). After 24 h of exposure, both HQ concentrations significantly increased SLC9A3 and SPP1 mRNA expression compared with the untreated control (Figure 7B). Western blotting showed corresponding increases in SLC9A3 and SPP1 protein expression, with the highest levels observed following treatment with 20 μmol/L HQ (Figure 7C-D). Conditioned medium collected after HQ removal and washout increased fibroblast activation markers in NRK-49F cells. Compared with the FBS-only and CM-Control groups, CM-HQ10 and CM-HQ20 significantly increased ITGB1 and α-SMA protein expression, with a greater response in the CM-HQ20 group (Figure 7E-F). No significant difference was observed between the FBS-only and CM-Control groups. Direct exposure to 10 or 20 μmol/L HQ did not reproduce the increases induced by conditioned medium. Consistent with the protein-expression results, CM-HQ10 and CM-HQ20 significantly increased NRK-49F migration, with the greatest number of migrated cells observed in the CM-HQ20 group (Figure 7G-H).
Figure 7.
In vitro proof-of-concept assessment of tubular and fibroblast responses in a hydroquinone model. (A) Viability of NRK-52E cells treated with 0, 10, or 20 μmol/L hydroquinone (HQ) for the indicated periods. (B) Relative SLC9A3 and SPP1 mRNA expression in NRK-52E cells after 24 h of HQ treatment. (C-D) Representative Western blots and densitometric quantification of SLC9A3 and SPP1 protein expression in NRK-52E cells. GAPDH served as the loading control. (E-F) Representative Western blots and densitometric quantification of ITGB1 and α-SMA expression in NRK-49F cells assigned to the FBS-only, CM-Control, CM-HQ10, CM-HQ20, Direct-HQ10, and Direct-HQ20 groups. (G-H) Representative Transwell images and quantitative analysis of migrated NRK-49F cells in the six experimental groups. Scale bars, 200μm. FBS-only denotes fresh complete medium containing 10% FBS without HQ or conditioned medium. CM-Control denotes conditioned medium collected from control NRK-52E cells. CM-HQ10 and CM-HQ20 denote conditioned medium collected after HQ removal and washout from NRK-52E cells previously exposed to 10 or 20 μmol/L HQ, respectively. Direct-HQ10 and Direct-HQ20 denote fresh complete medium containing the corresponding HQ concentration without conditioned medium. ns, P ≥ 0.05; ***P < 0.001.
4. Discussion
CKD is a major challenge in global public health, and its incidence rate has continued to rise over the past few decades [7]. For a long time, the discussion on the causes of CKD has mainly focused on traditional risk factors such as diabetes, high blood pressure, autoimmune abnormalities, and genetic variations. The latest environmental health assessment models indicate that in high risk populations, up to 26% of the risk for developing kidney disease may be directly linked to complex environmental toxin exposure [15]. Within this large matrix of environmental pollution, NMVOCs are gradually being recognized as a potential environmental contributor to kidney damage. This is because they are everywhere, from industrial emissions and indoor building materials to solvent evaporation and vehicle exhaust [16]. In the present study, higher province-level NMVOC emission burden was associated with greater odds of prevalent CKD. SLC9A3 was prioritized as a candidate gene, while single-cell analyses suggested potential SPP1-ITGB1 signaling between proximal tubular cells and fibroblasts.
Previous health assessments of volatile organic compounds (VOCs) have shown a bias regarding target organs, mostly focusing on their role in causing blood cancers or respiratory and cardiovascular diseases [17,18]. However, the kidney receives up to 20%-25% of the cardiac output and serves as a core excretory organ. Because of its role in secreting and reabsorbing toxins, it can also become a primary target organ for environmental exposure. Recent studies based on large population cohorts show that increased concentrations of VOC metabolites in urine are significantly and positively correlated with a higher risk of CKD and a progressive decline in eGFR [19]. The findings of our study are broadly consistent with this pattern. They also suggest that, unlike the acute tubular necrosis caused by high dose poisoning, exposure to background environmental concentrations of NMVOCs presents a hidden and progressive pattern of chronic tubulointerstitial damage and fibrosis. Combined with the stratified analysis in this study, we observed a potential subgroup difference. The association was observed among participants with obesity [20]. This phenomenon may partly relate to the highly fat soluble nature of NMVOCs and their specific toxicokinetic characteristics [21]. The large amount of white adipose tissue in obese patients may influence the distribution of lipophilic compounds for these fat soluble toxins [22]. This not only greatly slows down the detoxification and excretion of toxins through the liver and kidneys [23]. More critically, when daily energy fluctuations trigger lipolysis, the fat tissue turns into an internal source of toxin release [24]. This low dose and long term pattern of internal release keeps the kidney parenchyma cells under prolonged toxicological stress, even after external pollution has stopped, which accelerates the chronic depletion of renal tubules [25]. On the other hand, systemic metabolic inflammation induced by obesity is another potential contributor that works together to worsen kidney damage [26]. Extremely enlarged fat cells and the proinflammatory macrophages they attract keep the microenvironment constantly filled with proinflammatory cytokines and ROS [27]. When NMVOCs are converted into highly active intermediates through Phase I metabolism and cause a local burst of ROS [28], the background inflammation of obesity and the acute stress caused by environmental exposure may produce overlapping effects in the renal tubulointerstitium. These combined stresses may affect and depletes the internal antioxidant system of the kidney. This then triggers large scale lipid peroxidation and DNA oxidative damage. Biomarkers (such as 8-OHdG and F2 isoprostanes) confirm the worsening of this synergistic damage [29]. The combined stress mentioned above may impair the integrity of the renal tubular epithelial cells. This becomes the key pathological foundation potentially contributing to cell injury and fibrosis.
The sodium hydrogen exchanger 3 (NHE3), encoded by SLC9A3, is a key transmembrane cation transporter in the renal proximal tubule [30]. Under normal physiological conditions, NHE3 relies on the sodium concentration gradient inside and outside the cell to reabsorb about 70% of the filtered sodium in the primary urine and secrete hydrogen ions. It is a key molecular hub for regulating the volume of human extracellular fluid, maintaining blood pressure, and keeping acid base balance [31]. As the first contact station for toxins after filtration and concentration, the proximal tubular epithelial cells show early pathological responsiveness based on markers of changes in their microenvironment. When the proximal tubule is stimulated by early oxidative stress, lipid peroxidation, or microenvironmental hypoxia induced by NMVOCs, even if the cells have not yet reached irreversible necrosis, the actin cytoskeleton at their apical brush border will rapidly undergo pathological disassembly [32]. This sublethal stress may contribute to a large number of NHE3 protein complexes, which are firmly anchored to the apical membrane, to shed. These complexes can then be released into the urine in the form of exosomes [33]. Independent clinical studies have confirmed that during the subclinical damage stage of chronic toxin exposure, the adjusted abundance of SLC9A3 protein in the urine can show a significant increase. Moreover, this upward trend shows a highly positive correlation with the expression levels of local early proinflammatory factors in the kidney, such as IL 6 and TNF alpha [34]. Furthermore, through the SPP1 and ITGB1 axis of the microenvironmental fibrosis cascade reaction, damaged proximal tubular (PT) cells and specific macrophage subpopulations recruited by microenvironmental inflammation begin to largely express and secrete the profibrotic mediator SPP1 [35]. The free SPP1 ligands are released into the interstitial fluid. Using the conserved binding domains in their structure, they specifically bind with high affinity to the transmembrane integrin ITGB1 subunit receptors on the surface of resting fibroblasts and some pericytes [36]. The docking of the ligand and the ITGB1 receptor rapidly activates transmembrane biochemical signal transduction. This leads to the phosphorylation of focal adhesion kinase inside the cell, which in turn fully triggers inflammatory and proliferative transcription factor networks such as STAT3 and ERK [37]. This signaling cascade prompts resting fibroblasts to undergo a significant phenotypic transformation. They evolve into myofibroblasts, which have high contractility and matrix synthesis activity, and are marked by high expression of alpha SMA [38]. Ultimately, overactivated myofibroblasts continuously synthesize and release large amounts of extracellular matrix proteins into the microenvironment, such as type COL1A1, COL3A1, and FN1 [39]. The dense and abnormally crosslinked matrix excessively deposits in the tubulointerstitium. This not only compresses the surviving nephrons but also causes the degradation of the peritubular microvascular network and impaired blood flow, plunging the kidney into deeper hypoxia. At this point, the metabolic abnormalities of the proximal tubules triggered by tiny NMVOC emission burden in the environment, magnified by the cascade of the single cell communication network, finally develop into the irreversible end stage of organic renal interstitial fibrosis.
This study analyzes the evidence of association in environmental epidemiology from a new perspective. First, these complementary lines of evidence provide tentative support for the SLC9A3/SPP1–ITGB1 pathway. Second, the integration of machine learning-based feature selection with microscopic mechanism analysis under single-cell resolution provides a certain basis for the clinical translation potential of the core target SLC9A3. Finally, in vitro experiments provided proof-of-concept evidence that hydroquinone treatment was associated with higher SLC9A3 and SPP1 expression in renal tubular epithelial cells, while conditioned medium from hydroquinone-treated cells increased fibroblast migration and α-SMA expression. This finding provides supportive in vitro evidence that the SLC9A3/SPP1 signaling axis may be involved in renal microenvironmental communication, a key cytological event. However, this study also has certain limitations. First, exposure was assigned using a 2015 province-level emission inventory, with all participants in the same province receiving the same estimate. This ecological proxy did not capture temporal, spatial, or individual-level exposure variability, and emission quantities do not directly represent ambient concentrations or absorbed doses. Therefore, the findings reflect an ecological association between province-level NMVOC emission burden and kidney-function status rather than an individual-level or causal relationship. Second, the eGFR <60 mL/min/1.73 m2 definition may have missed early kidney injury. Future studies should incorporate multi-year, finer-resolution exposure estimates and urinary kidney-damage biomarkers. Third, HQ represents only a benzene-related metabolic pathway and cannot reproduce the chemical diversity or combined effects of real-world NMVOC mixtures. Accordingly, these experiments provide proof-of-concept evidence specific to the HQ model, and further validation using multiple NMVOC parent compounds and metabolites is needed. Finally, immortalized rat cell lines cannot fully reproduce the cellular heterogeneity or biological responses of the human kidney. Further validation using primary human proximal tubular epithelial cells and primary human renal fibroblasts is therefore warranted.
5. Conclusion
In conclusion, higher province-level NMVOC emission burden was associated with greater odds of CKD. Integrated analyses nominate SLC9A3/SPP1-ITGB1 signaling as a candidate pathway, with proof-of-concept support from a benzene-metabolite model. However, individual-level exposure assessment, multi-compound models, and targeted in vivo studies are needed for confirmation.
Supplementary Material
Acknowledgment
The authors thank the volunteers who participated in the research.
Funding Statement
This work was supported by the National Natural Science Foundation of China (No. 81900754).
Ethics approval and consent to participate
This study was conducted in accordance with the principles of the Declaration of Helsinki. The CHARLS protocol was approved by the Biomedical Ethics Review Board of Peking University (IRB00001052-11015).
Consent form
CHARLS data have been approved by the Biomedical Ethics Review Board of Peking University, with informed consent from participants.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Data availability statement
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.
References
- 1.Cockwell P, Fisher LA.. The global burden of chronic kidney disease. Lancet. 2020;395(10225):662–664. doi: 10.1016/S0140-6736(19)32977-0. [DOI] [PubMed] [Google Scholar]
- 2.Luyckx VA, Tuttle KR, Garcia-Garcia G, et al. Reducing major risk factors for chronic kidney disease. Kidney Int Suppl. 2017;7(2):71–87. doi: 10.1016/j.kisu.2017.07.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Danziger J, Dodge LE, Hu H, et al. Susceptibility to environmental heavy metal toxicity among Americans with kidney disease. Kidney360. 2022;3(7):1191–1196. doi: 10.34067/KID.0006782021. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Bowe B, Artimovich E, Xie Y, et al. The global and national burden of chronic kidney disease attributable to ambient fine particulate matter air pollution: a modelling study. BMJ Glob Health. 2020;5(3):e002063. doi: 10.1136/bmjgh-2019-002063. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Baek SU, Yoon JH.. Association between exposure to VOCs mixture and impaired renal function in Korean adults. Am J Epidemiol. 2025;194(12):3520–3528. doi: 10.1093/aje/kwaf215. [DOI] [PubMed] [Google Scholar]
- 6.Mangotra A, Singh SK.. Volatile organic compounds: a threat to the environment and health hazards to living organisms – A review. J Biotechnol. 2024;382:51–69. doi: 10.1016/j.jbiotec.2023.12.013. [DOI] [PubMed] [Google Scholar]
- 7.Jiang L, Wang H, Xiao Y, et al. Exploring the association between volatile organic compound exposure and chronic kidney disease: evidence from explainable machine learning methods. Ren Fail. 2025;47(1):2520906. doi: 10.1080/0886022X.2025.2520906. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Liu H, Xiang X, Shi C, et al. Oxidative stress and inflammation in renal fibrosis: novel molecular mechanisms and therapeutic targets. Chem Biol Interact. 2025;421:111784. doi: 10.1016/j.cbi.2025.111784. [DOI] [PubMed] [Google Scholar]
- 9.Vervaet BA, D’Haese PC, Verhulst A.. Environmental toxin-induced acute kidney injury. Clin Kidney J. 2017;10(6):747–758. doi: 10.1093/ckj/sfx062. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Lewis S, Chen L, Raghuram V, et al. “SLC-omics” of the kidney: solute transporters along the nephron. Am J Physiol Cell Physiol. 2021;321(3):C507–c518. doi: 10.1152/ajpcell.00197.2021. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Lin YL, Yang YC.. Association of urinary volatile organic compounds and chronic kidney disease in patients with diabetes: real-world evidence from the NHANES. Metabolomics. 2024;20(6):121. doi: 10.1007/s11306-024-02188-x. [DOI] [PubMed] [Google Scholar]
- 12.Zhao Y, Hu Y, Smith JP, et al. Cohort profile: the China Health and Retirement Longitudinal Study (CHARLS). Int J Epidemiol. 2014;43(1):61–68. doi: 10.1093/ije/dys203. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Inker LA, Schmid CH, Tighiouart H, et al. Estimating glomerular filtration rate from serum creatinine and cystatin C. N Engl J Med. 2012;367(1):20–29. doi: 10.1056/NEJMoa1114248. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Webster AC, Nagler EV, Morton RL, et al. Chronic kidney disease. Lancet. 2017;389(10075):1238–1252. doi: 10.1016/S0140-6736(16)32064-5. [DOI] [PubMed] [Google Scholar]
- 15.Sanders AP, Saland JM, Wright RO, et al. Perinatal and childhood exposure to environmental chemicals and blood pressure in children: a review of literature 2007–2017. Pediatr Res. 2018;84(2):165–180. doi: 10.1038/s41390-018-0055-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Xie S, Huang J, Chen G, et al. Urine volatile organic compounds in predicting chronic obstructive pulmonary disease risk in a national observational study. Environ Sci Process Impacts. 2025;27(8):2588–2600. doi: 10.1039/d5em00181a. [DOI] [PubMed] [Google Scholar]
- 17.Li X, Li Z, Shen H, et al. Effects of long-term and low-concentration exposures of benzene and formaldehyde on mortality of Drosophila melanogaster. Environ Pollut. 2022;300:118924. doi: 10.1016/j.envpol.2022.118924. [DOI] [PubMed] [Google Scholar]
- 18.Zheng LY, Sanders AP, Saland JM, et al. Environmental exposures and pediatric kidney function and disease: a systematic review. Environ Res. 2017;158:625–648. doi: 10.1016/j.envres.2017.06.029. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Zhang Y, Qiu X, Wu Z, et al. Comprehensive analysis of the association between volatile organic compound pollutants and chronic kidney disease in hypertensive populations: insights from multi-omics approaches and identification of potential therapeutic targets. Environ Res. 2025;282:121966. doi: 10.1016/j.envres.2025.121966. [DOI] [PubMed] [Google Scholar]
- 20.Čolak E, Pap D.. The role of oxidative stress in the development of obesity and obesity-related metabolic disorders. J Med Biochem. 2021;40(1):1–9. doi: 10.5937/jomb0-24652. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Gracia FJ, Silla I, Peiró JM, et al. [The state of the psychological contract and its relation with employees’ psychological health]. Psicothema. 2006;18(2):256–262. [PubMed] [Google Scholar]
- 22.Wang C, Wang P, Yang L-P, et al. Association of CYP2C9, CYP2A6, ACSM2A, and CPT1A gene polymorphisms with adverse effects of valproic acid in Chinese patients with epilepsy. Epilepsy Res. 2017;132:64–69. doi: 10.1016/j.eplepsyres.2017.02.015. [DOI] [PubMed] [Google Scholar]
- 23.Qiang YW, Hao MH, Yang J.. Urinary cadmium was linearly and positively associated with cardiac infarction/injury score and subclinical myocardial injury in the general population without cardiovascular diseases and chronic kidney disease. Int Urol Nephrol. 2024;56(5):1677–1685. doi: 10.1007/s11255-023-03853-1. [DOI] [PubMed] [Google Scholar]
- 24.Manna P, Jain SK.. Obesity, oxidative stress, adipose tissue dysfunction, and the associated health risks: causes and therapeutic strategies. Metab Syndr Relat Disord. 2015;13(10):423–444. doi: 10.1089/met.2015.0095. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Obrador GT, Schultheiss UT, Kretzler M, et al. Genetic and environmental risk factors for chronic kidney disease. Kidney Int Suppl. 2017;7(2):88–106. doi: 10.1016/j.kisu.2017.07.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Woroniecka KI, Park ASD, Mohtat D, et al. Transcriptome analysis of human diabetic kidney disease. Diabetes. 2011;60(9):2354–2369. doi: 10.2337/db10-1181. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Shi J-S, Qiu D-D, Le W-B, et al. Identification of transcription regulatory relationships in diabetic nephropathy. Chin Med J (Engl). 2018;131(23):2886–2890. doi: 10.4103/0366-6999.246063. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Sabath E, Robles-Osorio ML.. Renal health and the environment: heavy metal nephrotoxicity. Nefrologia (Engl Ed). 2012;32(3):279–286. [DOI] [PubMed] [Google Scholar]
- 29.Haruna I, Obeng-Gyasi E.. Combined effect of per- and polyfluoroalkyl substances, toxic metals, and essential elements on chronic kidney disease. Pollutants. 2025;5(2):12. doi: 10.3390/pollutants5020012. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Wilson D, Jackson T, Sapey E, et al. Frailty and sarcopenia: the potential role of an aged immune system. Ageing Res Rev. 2017;36:1–10. doi: 10.1016/j.arr.2017.01.006. [DOI] [PubMed] [Google Scholar]
- 31.Du Cheyron D, Daubin C, Poggioli J, et al. Urinary measurement of Na+/H + exchanger isoform 3 (NHE3) protein as new marker of tubule injury in critically ill patients with ARF. Am J Kidney Dis. 2003;42(3):497–506. doi: 10.1016/s0272-6386(03)00744-3. [DOI] [PubMed] [Google Scholar]
- 32.Jana S, Mitra P, Roy S.. Proficient novel biomarkers guide early detection of acute kidney injury: a review. Diseases. 2022;11(1):8. doi: 10.3390/diseases11010008. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Pike LS, Smift AL, Croteau NJ, et al. Inhibition of fatty acid oxidation by etomoxir impairs NADPH production and increases reactive oxygen species resulting in ATP depletion and cell death in human glioblastoma cells. Biochim Biophys Acta. 2011;1807(6):726–734. doi: 10.1016/j.bbabio.2010.10.022. [DOI] [PubMed] [Google Scholar]
- 34.Morita M, Shinbo S, Asahi A, et al. Very long chain fatty acid β-oxidation in astrocytes: contribution of the ABCD1-dependent and -independent pathways. Biol Pharm Bull. 2012;35(11):1972–1979. doi: 10.1248/bpb.b12-00411. [DOI] [PubMed] [Google Scholar]
- 35.Zhou Z, Yang L, Liao H, et al. Analysis of a child with carnitine palmitoyl transferase 1A deficiency due to variant of CPT1A gene. Zhonghua Yi Xue Yi Chuan Xue Za Zhi. 2021;38(2):184–187. doi: 10.3760/cma.j.cn511374-20200309-00146. [DOI] [PubMed] [Google Scholar]
- 36.Shintani S, Shiigai T.. Repeat MRI in acute rhabdomyolysis: correlation with clinicopathological findings. J Comput Assist Tomogr. 1993;17(5):786–791. doi: 10.1097/00004728-199309000-00023. [DOI] [PubMed] [Google Scholar]
- 37.Shin SM, Baek EJ, Kim KH, et al. Polydeoxyribonucleotide exerts opposing effects on ERK activity in human skin keratinocytes and fibroblasts. Mol Med Rep. 2023;28(2):148. doi: 10.3892/mmr.2023.13035. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Meng X-M, Wang S, Huang X-R, et al. Inflammatory macrophages can transdifferentiate into myofibroblasts during renal fibrosis. Cell Death Dis. 2016;7(12):e2495. doi: 10.1038/cddis.2016.402. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Montgomery TA, Xu L, Mason S, et al. Breast regression protein-39/chitinase 3-like 1 promotes renal fibrosis after kidney injury via activation of myofibroblasts. J Am Soc Nephrol. 2017;28(11):3218–3226. doi: 10.1681/ASN.2017010110. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.