Lipid metabolism-related gene expression predicts prognostic outcomes in lung adenocarcinoma

Introduction

According to the 2022 Global Cancer Statistics, lung cancer remains the most perilous malignancy, with an estimated 2.5 million new cases and 1.8 million associated deaths, and poses a serious threat to human health (1). Lung adenocarcinoma (LUAD), the most prevalent histological subtype of lung cancer, accounts for approximately 40–60% of all newly diagnosed cases (2) and is associated with aggressive progression, early metastatic spread, and high recurrence rates (3). The early stages of LUAD are often asymptomatic or present with non-specific symptoms, when clinical manifestations such as cough, hemoptysis, or chest pain emerge, many patients have already progressed to advanced or metastatic disease (4). Furthermore, lung cancer frequently metastasizes to sites including the brain, bones, adrenal glands, and liver (5), which renders the treatment even more challenging. Over recent decades, considerable advancements have been made in treatment modalities, including surgery, chemotherapy, radiotherapy, immunotherapy, and molecularly targeted therapy. Despite these efforts, the prognosis for LUAD patients remains poor, with a five-year survival rate of averaging less than 20% (6). Although a subset of patients harbors targetable driver mutations in genes such as EGFR, KRAS, ALK, ROS1, and RET, the inevitable development of acquired resistance often leads to treatment failure and disease relapse (7,8). Therefore, the identification of novel biomarkers and key signaling pathways that regulate tumor progression is imperative for developing innovative therapeutic strategies and improving prognostic assessment in LUAD.

Lipids comprise a diverse group of molecules, including phospholipids, fatty acids (FAs), triglycerides, sphingolipids, and cholesterol and its esters (9). Lipid metabolic reprogramming—encompassing the synthesis, storage, and degradation of lipids—is essential for maintaining cellular energy homeostasis, facilitating signal transduction, and supporting membrane biogenesis (10). This reprogramming represents a hallmark of cancer metabolism. Tumor cells exploit lipid metabolism to fulfill their biosynthetic and bioenergetic demands, thereby promoting proliferation, invasion, and metastasis. Furthermore, altered lipid metabolism has been shown to significantly influence the response to various cancer therapies (11). Distinct lipidomic profiles have been identified in LUAD tissues compared to normal lung tissues (12). For instance, key enzymes involved in de novo FA synthesis, such as fatty acid synthase (FASN), acetyl-CoA carboxylase (ACC), and ATP citrate lyase (ACLY), are frequently overexpressed in LUAD and are strongly correlated with enhanced proliferative capacity (13,14). It has also been observed that concurrent loss-of-function mutations in serine/threonine kinase 11 (STK11) and KEAP1 are linked to aggressive tumor growth, therapy resistance, and poor survival in LUAD in upregulate the levels of stearoyl-CoA desaturase 1 (SCD1) (15,16). Metabolic reprogramming acts as a hallmark of LUAD progression and therapy resistance, its clinical practice for treatment and prognosis improvement remains limited (17). Consequently, there is a pressing need to discover novel lipid metabolism-related therapeutic targets and to elucidate their pathogenic mechanisms to advance LUAD treatment.

In this study, a prognostic signature for LUAD was developed to facilitate future clinical application. Lipid metabolism-related genes associated with overall survival (OS) were identified through univariate Cox, followed by least absolute shrinkage and selection operator (LASSO), and multivariate Cox regression analyses. A robust 12-gene prognostic model was constructed using the GSE68465 dataset from the Gene Expression Omnibus (GEO) as a training cohort, and its predictive performance was validated in an independent cohort from The Cancer Genome Atlas (TCGA)-LUAD. The levels of these 12 genes were further examined in a normal lung epithelial cell line (BEAS-2B) and three LUAD cell lines (A549, H1975, and H157) utilizing quantitative reverse transcription polymerase chain reaction (qRT-PCR). Among these, MBTPS2 was selected for in-depth functional characterization. For the first time, this study demonstrates that MBTPS2 upregulation promotes proliferative and aggressive phenotypes in LUAD via a series of in vivo and in vitro functional assays. This project provides novel insights into individualized therapy for LUAD by establishing a lipid metabolism-related risk signature and identifying MBTPS2 as a potential tumor promoter. We present this article in accordance with the TRIPOD reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1-2665/rc).

Methods

Data acquisition and preprocessing

The mRNA expression profiles along with corresponding clinical records of human LUAD were acquired from the GEO repository (http://www.ncbi.nlm.nih.gov/geo/). Dataset GSE68465 was analyzed on the GPL96-57554 platform [Affymetrix Human Genome U133A Array (HG-U133A)] and was used as the training set. Transcription mRNA expression profiles and clinical prognostic data of LUAD patients were obtained from TCGA data portal (https://gdc-portal.nci.nih.gov/) and was used as the test set. Lipid metabolism-related gene set was downloaded from Gene Set Enrichment Analysis (GSEA) online database (http://software.broadinstitute.org/gsea/). Next, we preprocessed the data. Firstly, we processed the raw expression data via Robust Multi-array Average (RMA) background correction, and employed for processed signals through log₂ transformation and normalization. Then, we processed the probe sets utilizing the “affy” R package with median-polish summarization, followed by gene annotation with corresponding Affymetrix annotation files.

Development and validation of the prognostic signature

In the training set (GSE68465), samples from normal lung tissue and those lacking precise OS information were excluded. Transcriptional profiles of lipid metabolism-related genes were extracted. To preliminarily screen genes whose expression levels were linked to OS in LUAD patients, a univariate Cox regression analysis was employed. Here, genes with P<0.01 were considered statistically significant and were subsequently subjected to LASSO regression analysis. Those genes chosen by LASSO regression were then included in a multivariate Cox regression proportional analysis to recognize independent prognostic predictors. The risk score was calculated as a weighted sum of mRNA expression levels, with each weight corresponding to the regression coefficient (beta) derived from the multivariate Cox regression analysis. Individual patients’ risk score was computed as follows: risk score = Σ (βgene_i × exprgene_i), where i ranges from 1 to N (total number of signature gene), expr denotes gene expression levels, and β corresponds to the regression coefficient originated from multivariate Cox regression analysis for each gene. Using the median risk score as the splint point, patients were stratified into high- and low-risk subgroups. The predictive performance of the signature was assessed by Kaplan-Meier survival analysis, and the time-dependent receiver operating characteristic (ROC) curve was plotted to evaluate the specificity and sensitivity of the index using the R packages survival and survivalROC, respectively. Patient survival status distribution and a heatmap of signature gene expression were visualized utilizing pheatmap package. The prognostic robustness of the signature was further validated in the independent TCGA-LUAD cohort. Furthermore, univariate and multivariate Cox regression analyses, which incorporated key clinical parameters (including gender, age, adjuvant chemotherapy, adjuvant radiotherapy, tumor stage, and lymph node stage), were performed to determine whether the risk score acted as an independent prognosis predictor.

Differentially expressed genes (DEGs) identification in high- and low-risk cohorts and functional enrichment investigation

DEGs between high- and low-risk subgroups in the training cohort were recognized through the ‘limma’ R package. The significance analysis of microarrays (SAM) method was employed, with a false discovery rate (FDR) <0.05 and an absolute log₂ fold change (|log₂FC|) >1 set as the significance thresholds. Genes meeting these criteria were selected for further analysis. To elucidate the potential biological functions of these DEGs, Gene Ontology (GO) annotation and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were conducted using the R package ‘clusterProfiler’, with an adjusted P<0.05 as the significance threshold. For GSEA (version 4.0.3), the Hallmarks gene set collection (h.all.v2022.1.Hs.symbols.gmt) from the Molecular Signatures Database (MSigDB) was used as the reference. Gene sets enriched in the high-risk group were selected and visualized, with P<0.05 was set as cut-off criteria.

Cell culture

Human bronchial epithelial cell line (BEAS-2B) and LUAD cell lines (H1975, A549 and H157) were purchased from the Cell Bank of the Chinese Academy of Sciences. The cells were maintained in RPMI-1640 medium (HyClone, USA) supplemented with 10% fetal bovine serum (FBS, Gemini Bio-Products, West Sacramento, CA, USA) and 1% penicillin/streptomycin (Sigma-Aldrich, St. Louis, MO, USA) and incubated at 37 °C with 5% CO₂ in a humidified atmosphere. The culture medium was refreshed every 2–3 days. In each experiment, cells were passaged a minimum of three times after thawing. Cells were trypsinized with 0.25% trypsin-ethylenediaminetetraacetic acid (EDTA) [Gibco (Thermo Fisher Scientific), Waltham, MA, USA] for subculture and cryopreservation. For freezing, cells were suspended in freezing medium composed of 50% DMEM, 40% FBS and 10% DMSO (Sigma-Aldrich, St. Louis, MO, USA).

qRT-PCR

Total RNA was purified from the normal bronchial cell line (BEAS-2B) and three LUAD cell lines (H1975, A549, H157) using Trizol reagent [Invitrogen (Thermo Fisher Scientific), Waltham, MA, USA]. Subsequently, complementary DNA (cDNA) was synthesized from the extracted RNA using the HiScript 1st Strand cDNA Synthesis Kit according to the manufacturer’s protocol. qRT-PCR was then executed utilizing AceQ qPCR SYBR Green Master Mix (Vazyme Biotech Co., Ltd., Nanjing, China). The expression levels of the target genes were normalized to the endogenous control GAPDH, and relative gene expression was calculated using the 2^–ΔΔCt method. The primer sequences are listed in Table S1. To further validate the expression of the 12 signature genes, their expression levels in normal lung tissues versus LUAD tissues were analyzed using the publicly accessible online tool, Gene Expression Profiling Interactive Analysis 2 (GEPIA2) (http://gepia2.cancer-pku.cn/). For this analysis, an FDR < 0.05 and |log₂ FC| >0.5 were set as significance thresholds.

Western blotting

Protein was extracted from cells and tissues using RIPA lysis buffer (Beyotime Biotechnology, Shanghai, China) containing a protease inhibitor cocktail. The lysates were sonicated on ice (amplitude 20%, 20 s, 10 s off) and centrifuged to collect the supernatant. Protein concentration was measured using BCA protein assay kit (Absin Bioscience Inc., Shanghai, China) and 20 µg of protein per lane was separated on 12.5% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) gels (Omni-Easy^TM kit, Epizyme Biomedical Technology, Shanghai, China) at 150 V for 50 minutes. The separated proteins were then electrophoretically transferred onto polyvinylidene difluoride (PVDF) membranes at 280 mA for 90 minutes. The membrane was incubated overnight at 4 °C with primary antibodies: MBTPS2 (1:1,000, abs112944, Absin), SREBP1 (1:1,000, 14088-1-AP, Proteintech Group, Inc., Rosemont, IL, USA), and GAPDH (1:5,000, AF7021, Affinity Biosciences, Cincinnati, OH, USA). Next, the membranes were incubated with HRP-conjugated secondary antibodies (1:5,000, ABclonal Technology, Wuhan, China) for 1 hour at room temperature. Signal detection was conducted using an enhanced chemiluminescence (ECL) kit (Feien Biotechnology, Guangzhou, China), and band intensity was quantified through ImageJ software.

Stable cell lines establishment

To overexpress MBTPS2, the full-length cDNA was amplified and inserted into the pCDH lentiviral vector (Obio Technology Corp., Ltd., Shanghai, China). For the knockdown of MBTPS2 in H1975 cells, short hairpin RNA (shRNA) targeting the gene was cloned into a pLKO.1 vector (Obio Technology). Simultaneously, non-targeting control sequences (scrambled shRNA) were cloned into the respective vectors to generate control cell lines. Lentiviruses were packaged and transduced into A549 (for overexpression) and H1975 (for knockdown) cells. Following transduction, cells were selected with puromycin [2 µg/mL, Amersco (now part of VWR International), Solon, OH, USA] for two weeks to establish stable polyclonal populations. The efficiency of overexpression and knockdown was confirmed by qPCR and Western blot. For SREBP inhibition experiments, A549 cells stably overexpressing MBTPS2 was treated with 10 µM Fatostatin (CM06532, Proteintech) for 24 h.

Nile red staining

Intracellular lipid content was assessed using Nile red staining. Cells were incubated with 1 µM Nile red (HY-D0718, MCE) in pre-warmed phosphate-buffered saline (PBS) for 20 minutes at 37 °C in the dark. Subsequently, nuclei were counterstained with DAPI (1:2,000 dilution, Beyotime) for 10 minutes. After washing with Hank’s Balanced Salt Solution (HBSS) containing calcium and magnesium, images were acquired using a fluorescence microscope (IX73, Olympus Corporation, Tokyo, Japan). Nile red fluorescence was detected in the PE channel. The threshold for Nile red positivity was established using unstained cells and dimethylsulphoxide (DMSO)-treated control cells. The “Count%” represents the percentage of single cells falling within the Nile red-positive gate. Data were analyzed using FlowJo software and plotted using GraphPad Prism.

Cell Counting Kit-8 (CCK-8) assay

Cell viability was examined through the CCK-8 (C0038, Beyotime). Cells under their logarithmic growth phase were collected and plated in 96-well plates at 2×10³ cells per well in 100 µL medium, followed by 24-hour incubation. At designated time points (0, 1, 2, 3, and 4 days), 10 µL of CCK-8 reagent was added to each well, followed by incubation for 2 hours at 37 °C in the dark. Absorbance was measured at 450 nm with a reference wavelength at 650 nm by microplate reader (Spark 10M, Shenyang, China). Cell viability (%) was calculated as follows: cell viability (%) = [optical density (OD)_treatment − OD_blank]/(OD_control − OD_blank) ×100%. All experiments were conducted in triplicate.

Cell proliferation assay

For colony formation assay, cells were trypsinized and seeded into 6-well plates at a density of 400 or 1,000 cells/well. Cell medium was changed every 2–3 days until visible colonies formed. Colonies were fixed with formaldehyde for 15 minutes and stained with Giemsa for 10–30 minutes. Clones containing >50 cells were counted either visually or under a microscope with low magnification. Cloning efficiency (%) was calculated as follows: cloning efficiency (%) = (the number of cell colonies/the number of seeded cells) ×100%.

Wound healing assay

Cell migration was assessed using a wound healing assay. Cells were seeded into 6-well plates and cultured until they reached over 90% confluency. A uniform wound was created in the cell monolayer using a sterile 200 µL pipette tip. Following gently washing with PBS to remove cellular debris, serum-free medium was added to plate. After 24 hours, Images were taken by a microscope and cell migration capacity was quantified using ImageJ software.

Clinical sample collection

Fresh clinical surgical specimens, including cancerous tissues and matched normal adjacent tissues, were acquired from eight LUAD patients during resection procedures at Shangrao Affiliated Hospital. All collected samples were immediately snap-frozen in liquid nitrogen and preserved at –80 °C until protein extraction. This research was approved by the Ethics Committee of Shangrao People’s Hospital [approval No. (2025) Physician Review (109)]. All procedures were performed in accordance with the Declaration of Helsinki and its subsequent amendments. Written informed consent was acquired from all participants following a comprehensive explanation of the study. All participants data were de-identified to ensure privacy.

Statistical analysis

Bioinformatic analysis was conducted utilizing R language (v4.1.2) and graphs were generated with GraphPad Prism (v10.0). Experimental data are presented as the mean ± standard deviation (SD) from at least three independent replicates. Prognostic predictors were identified, and a risk signature was constructed using univariate Cox, LASSO, and multivariate Cox regression analyses. Survival differences were evaluated by Kaplan-Meier analysis with the log-rank test, and the predictive accuracy of the signature was assessed using ROC curves. For in vitro experiments, comparisons between two groups were performed using a two-tailed Student’s t-test, and comparisons among multiple groups were conducted by one-way analysis of variance (ANOVA) (SPSS v25.0). A P value <0.05 was considered statistically significant.

Results

Recognition of prognostic genes correlated with OS in the training set

The overall workflow for the establishment and validation of the twelve-gene signature and the subsequent functional investigation of MBTPS2 was illustrated in Figure 1. To identify genes associated with OS, a univariate Cox regression analysis was performed on the training cohort (GSE68465). This analysis identified 42 lipid metabolism-related genes that were significantly correlated with prognosis (P<0.01), which were subsequently subjected to LASSO regression analysis to mitigate overfitting (Figure 2A,2B). The 22 genes selected by the LASSO regression are listed in the caption for Figure 2B. Among these, ten genes (HILPDA, PPARD, SEC23A, ETNK1, HACD1, SRD5A3, MBTPS2, TSPO, UGT8, and AKR1C4) were identified as protective factors, as their elevated expression was associated with longer survival times (hazard ratio <1). In contrast, the remaining twelve genes (CERS4, ACHE, PIK3R1, PIK3CD, LDLRAP1, ST3GAL2, NCOA2, MAPKAPK2, LPIN2, CYP11B2, SPHK2, and SMPD3) were risk factors, with higher expression correlated with poorer survival (hazard ratio >1).

Figure 1 Workflow depicting the procedure for developing a prognostic index and assessing the function of MBTPS2 in vitro laboratory studies. CCK-8, Cell Counting Kit-8; EdU, 5-ethynyl-2'-deoxyuridine; GESA, Gene Expression Set Analysis; LASSO, least absolute shrinkage and selection operator; LMGs, lower-grade gliomas; LUAD, lung adenocarcinoma; qPCR, quantitative polymerase chain reaction; RNA-Seq, RNA sequencing; TCGA, The Cancer Genome Atlas.

Figure 2 Independent lipid metabolism-related prognostic genes identification through LASSO and multivariate Cox hazard regression. (A) Plots illustrating the error rates derived from ten-fold cross-validation. (B) Profiles of the LASSO coefficient profiles of 22 lipid metabolism-related prognostic genes. (C) The multivariate hazard Cox proportional regression analysis revealed 12 independent prognostic genes linked to lipid metabolism. CI, confidence interval; HR, hazard ratio; LASSO, least absolute shrinkage and selection operator.

Construction and validation of twelve-gene signature for OS prediction in the training dataset

The 22 prognostic genes were further analyzed through multivariate Cox regression to determine independent prognostic predictors. This process yielded a final signature of 12 genes (HILPDA, ACHE, PPARD, ETNK1, ST3GAL2, NCOA2, HACD1, MBTPS2, UGT8, LPIN2, SPHK2, and AKR1C4) that were used to develop a prognostic risk model (Figure 2C). A risk score for each patient was calculated using a calculation formula derived from the multivariate Cox regression coefficients: risk score = (−0.231803703746239) × HILPAD expression + (−0.188682081140626) × ACHE expression + (0.477392486429451) × PPARD expression + (0.192044305543993) × ETNK1 expression + (−0.322881396996877) × ST3GAL2 expression + (−0.47393130412863) × NCOA2 expression + (0.088555689157376) × HACD1 expression + (0.288948136662179) × MBTPS2 expression + (0.166616371531222) × UGT8 expression + (−0.221400735405029) × LPIN2 expression + (−0.318490234445077) × SPHK2 expression + (0.0888826320668024) × AKR1C4 expression. In the training dataset, patients were stratified into high-risk (n=221) and low-risk (n=221) groups based on the median risk score (0.970941). A highly significant difference (P=1.71E−13) in OS was discovered between two groups. Patients in the high-risk group exhibited a markedly shorter median survival (2.63 years) compared to the low-risk group (4.93 years) (Figure 3A). The predictive power of the signature was robust, as demonstrated by ROC curve analysis, which showed an area under the curve (AUC) of 0.761 for OS prediction (Figure 3B). The distribution of risk scores, patient survival status, and the expression patterns of the 12 signature genes were visualized in Figure 3C. The corresponding heatmap revealed that five genes (ACHE, ST3GAL2, NCOA2, LPIN2, and SPHK2) were more highly expressed in the low-risk group, while the remaining seven genes (HILPDA, PPARD, ETNK1, HACD1, MBTPS2, UGT8, and AKR1C4) were upregulated in the high-risk group.

Figure 3 Prognostic performance of the twelve-gene signature for OS in LUAD patients (GSE68465 training set and TCGA test set). (A,D) The Kaplan-Meier survival curves comparing high- and low-risk patients in the training and test sets. (B,E) The ROC curves evaluating the predictive power of the twelve-gene risk model in both the training and test dataset. (C,F) The twelve-gene risk score distribution, survival of patients and heatmap of the twelve-gene expression profiles in the training set and test set. In the heatmaps, the color scale represents the z-score normalized expression levels, ranging from low expression (blue) to high expression (red). AUC, area under the curve; LUAD, lung adenocarcinoma; OS, overall survival; ROC, receiver operating characteristic; TCGA, The Cancer Genome Atlas.

Validation of the twelve-gene signature in the test cohort

In order to validate the prognostic efficacy of the twelve-gene index, an independent dataset from the TCGA-LUAD were utilized. According to the risk score formula originated from the training cohort, 88 patients in the TCGA test set were classified into high-risk (n=44) and low-risk (n=44) groups. A statistically significant difference in OS was found between two categories (P=0.009; median survival: 2.007 vs. 1.508 years) (Figure 3D). ROC curves demonstrated favorable prognostic precision, achieving an AUC of 0.741 for the 12-gene signature in the TCGA test dataset (Figure 3E). In line with the training cohort, the distribution of risk scores, patient survival status, and the expression patterns of the twelve signature genes in the test set are visualized in Figure 3F. Collectively, these results substantiate the robustness and generalizability of the twelve-gene signature for prognostic stratification in LUAD.

Independent prognostic evaluation of the prognostic signature

To determine whether the signature serves as an independent prognosis predictor of OS, univariate and multivariate Cox proportional regression analyses were executed, incorporating multiple key clinicopathological features such as age, gender, adjuvant chemotherapy, adjuvant radiotherapy, tumor stage, and lymph node stage. Analysis suggested that the twelve-prognostic signature was substantially associated with the OS in both univariate (P<0.001, Figure 4A) and multivariate Cox regression analysis (P<0.001, Figure 4B), after adjusting for additional clinical variables. It was concluded that the twelve-gene prognostic model represents a standalone prognostic marker of OS in LUAD.

Figure 4 Univariate Cox and multivariate Cox hazard regression of the prognostic model incorporating with clinical variables, including age, gender, adjuvant chemotherapy, adjuvant radiotherapy, lymph nodal stage and tumor grade. (A) Univariate Cox regression analysis for signature model and clinical characteristics. (B) Multivariate Cox regression analysis for signature and clinical features. Adjuvant_chemo, adjuvant chemotherapy; Adjuvant_rt, adjuvant radiotherapy; CI, confidence interval; HR, hazard ratio; N, lymph node; T, tumor.

Development and calibration of nomogram

To facilitate clinical translation, a comprehensive prognostic nomogram was developed by integrating the patient risk scores from the twelve-gene signature with clinical characteristics, including age, gender, tumor stage, and lymph node stage (Figure 5A). In this nomogram, individual points were assigned in accordance with patient’s risk score and other clinical features, and the cumulative total score was utilized to gauge their odds of 1-, 3- and 5-year OS. For example, the profile of a 75-year-old female patient in stage I with T1, N0 and a low-risk score yielded a total of 194 points, corresponding to estimated 1-, 3- and 5-years survival probabilities of 95.5%, 84.0% and 74.9%, respectively. The predictive performance of the signature was evaluated using time-dependent ROC analysis in both the training and test datasets (Figure 5B,5C). The model achieved AUCs of 0.788, 0.794, and 0.782 for predicting 1-, 3-, and 5-year survival in the training set, and 0.823, 0.852, and 0.648 in the test set, respectively. Moreover, the calibration curves showed strong agreement between the nomogram-predicted and actual observed survival probabilities at 1, 3, and 5 years in both datasets (Figure 5D,5E). Decision curve analysis also demonstrated the clinical utility of the nomogram (Figure 5F,5G). In conclusion, the nomogram was validated as an effective and reliable tool for predicting individual prognosis in LUAD patients.

Figure 5 Nomogram development in the training set (GSE68465) and validate in test set TCGA. (A) The developed nomogram incorporating risk score values and clinical variants based on training dataset. (B,C) Nomogram calibration via ROC curves in training and test sets at 1, 3 and 5 years. (D,E) Calibration curves for predicting 1-, 3- and 5-year OS in training and test sets. (F,G) Decision curve analysis was performed in training and test sets at 1, 3 and 5 years. ***, P<0.001. AUC, area under the curve; CI, confidence interval; N, node; OS, Overall survival; ROC, receiver operating characteristic; T, tumor; TCGA, The Cancer Genome Atlas.

Stratified evaluation of the twelve-gene signature for prognosticating other clinical variables

To further determine the prognostic performance of the twelve-gene signature, subgroup assessments were employed based on clinical characteristics including gender, age, adjuvant chemotherapy, adjuvant radiotherapy, tumor stage, node lymph stage, histologic grade and tumor progression. Results were visualized in a forest plot (Figure 6). The risk score remained a significant prognostic factor in gender-stratified datasets (female and male), age-divided cohorts (≤65 and >65 years), and among patients with or without adjuvant chemotherapy or radiotherapy. Significant prognostic value was also maintained across lymph node stages (N0, N1, N2), T-stage subgroups (T1, T2), histologic stage II/III subgroups, and in patients classified by relapse status. However, the prognostic significance was not observed in the T3/T4 and histologic stage I subgroups, which was likely attributable to the limited number of patients in these categories. A significant difference was also observed in the subgroup of patients with an unknown progression status.

Figure 6 Survival analyses of LUAD patients categorized by gender, age, adjuvant chemotherapy, adjuvant radiotherapy, tumor stage, nodal lymph stage, progression and histologic grade with the twelve-gene signature in GSE68465. CI, confidence interval; HR, hazard ratio; LUAD, Lung adenocarcinoma; N, node; RT, radiotherapy; T, tumor.

Functional annotation

To elucidate the biological significance of twelve-gene prognostic index, samples from the GSE68465 training cohort were classified into high- and low-risk subgroups. Under the threshold of a FDR <0.05 and |log₂ FC| >1, a total of 372 DEGs were detected (170 down-regulated and 202 up-regulated). The distribution of these DEGs is visualized in the volcano plot (Figure 7A). Subsequent protein-protein interaction (PPI) network analysis revealed central roles for the CCNB and CXCL gene families, which were known to be crucial involved in cell cycle regulation and immune-inflammatory responses, respectively (Figure 7B). These findings underscore the interconnection between lipid metabolism, cell cycle progression, and immune modulation in LUAD pathogenesis.

Figure 7 Functional annotation of DEGs. (A) The volcano plot illustrated the DEGs based on a comparison between the low- and high-risk groups. (B) Construction and analysis of the PPI network for DEGs. (C) GO functional profiling of the DEGs. (D) KEGG pathway enrichment analysis for DEGs. (E) The ten enriched pathways in the high-risk group were analyzed by GSEA. BP, biological process; CC, cellular component; DEGs, differentially expressed genes; ECM, extracellular matrix; FDR, false discovery rate; GO, Gene Ontology; GSEA, Gene Set Enrichment Analysis; KEGG, Kyoto Encyclopedia of Genes and Genomes; MF, molecular function; PPI, protein-protein interaction.

Next, GO and KEGG pathway enrichment analyses were subsequently conducted. GO analysis revealed that biological processes were primarily enriched in the regulation of peptidase activity, chromosome segregation, myeloid leukocyte migration, humoral immune response, and extracellular matrix organization (Figure 7C). Cellular components were enriched for the collagen-containing extracellular matrix, cornified envelope, and vesicle lumen. Molecular functions were significantly associated with peptidase regulator activity, CXCR chemokine receptor binding, chemokine activity, and endopeptidase activity. KEGG pathway analysis indicated that the DEGs were enriched in pathways including transcriptional misregulation in cancer, the cell cycle, amoebiasis, cellular senescence, the IL-17 signaling pathway, the p53 signaling pathway, rheumatoid arthritis, small cell lung cancer, and viral protein interaction with cytokine and cytokine receptors (Figure 7D). Moreover, GSEA suggested that high-risk cases showed marked enrichment in mTORC1 signaling, E2F targets, G2/M checkpoint, MYC targets V1, PI3K/AKT/mTOR signaling, glycolysis and several lipid-related pathways, including hypoxia, oxidative phosphorylation, FA metabolism and cholesterol homeostasis (Figure 7E).

Validation of 12 genes relative expression

Through the GEPIA online database, expression levels of 12 genes between LUAD tissues and match normal tissues from TCGA normal and Genotype-Tissue Expression (GTEx) data were contrasted. Subsequently, the relative mRNA expression of the 12 signature genes was explored in multiple LUAD cell lines (H1975, A549, H157) and a normal lung bronchial epithelial cell line (BEAS-2B) using qRT-PCR. As shown in Figure 8A, the transcript levels of MBTPS2 were markedly elevated across all tested LUAD cell lines compared to the BEAS-2B control. This finding was validated using the GEPIA online database, which confirmed that MBTPS2 expression was significantly higher in LUAD tissues than in normal lung tissues (Figure S1). Consistent with the mRNA data, the protein expression of MBTPS2 was also found to be significantly upregulated in LUAD cell lines compared to BEAS-2B cells (Figure 8B).

Figure 8 Validation of genes relative expression involved in lipid metabolism-related prognostic signature. (A) The mRNA expression of MBTPS2, HILPDA, PPARD, ST3GAL2, ETNK1, NCOA2, HACD1, ACHE, LPIN2, UGT8, SPHK2 and AKR1C4 in bronchial cell line (BEAS-2B) and LUAD cell lines (A549, H1975 and H157). (B) Protein expression level of MBTPS2 in lung cell line (BEAS-2B) and LUAD cell lines (A549, H1975 and H157). *, P<0.05; **, P<0.01; ***, P<0.001; ****, P<0.0001 vs. BEAS-2B.

MBTPS2 promotes proliferation and suppresses apoptosis of LUAD cells in vitro

To investigate the underlying biological function of MBTPS2 in LUAD, isogenic cell models were generated. MBTPS2 was stably overexpressed in A549 cells and knocked down in H1975 cells using shRNA. The successful construction of these models was confirmed by qRT-PCR and western blot assays, which showed a significant decrease in MBTPS2 expression in knockdown cells (Figure 9A,9B) and a significant increase in overexpressing cells (Figure 9C,9D). Functional assays demonstrated that MBTPS2 knockdown strongly suppressed clonogenic survival (Figure 9E) and cellular proliferation (Figure 9F), whereas its overexpression enhanced these capacities (Figure 9G,9H). Wound healing assays confirmed that cell migration may been inhibited upon MBTPS2 knockdown (Figure 9I) and promoted by its overexpression (Figure 9J). After treating with the specific SREBP inhibitor Fatostatin (18), the enhanced carcinogenic activity was significantly reversed (Figure 9H,9J). Flow cytometry analysis revealed a significantly improved apoptosis rate in MBTPS2-knockdown cells (Figure 9K) and a decreased rate in overexpressing cells (Figure 9L). Western blot analysis indicated that MMP9 expression and the Bcl-2/Bax ratio were significantly reduced following MBTPS2 knockdown (Figure 9M) and elevated upon its overexpression (Figure 9N). Collectively, these findings demonstrated that MBTPS2 promotes proliferation, and invasion while suppressing apoptosis in LUAD cells, potentially through the modulation of MMP9, Bcl-2, and Bax.

Figure 9 MBTPS2 promotes the capacity of proliferation, motility and invasion and inhibited the apoptosis ability of LUAD cells in vitro. (A,B) The decreased MBTPS2 expression at mRNA (A) and protein (B) levels in H1975 cells by MBTPS2 knockdown. (C,D) The increased MBTPS2 expression at mRNA (C) and protein (D) levels in A549 cells after MBTPS2 overexpression. (E,G) MBTPS2 knockdown suppressed colony formation (E), whereas its overexpression enhanced it (G) (crystal violet staining). (F,H) MBTPS2 knockdown resulted in a suppression of cell proliferation (F), while upregulation of MBTPS2 stimulated proliferation activity (H) in CCK-8 assay. (I,J) MBTPS2 knockdown reduced cell migratory activity and its overexpression shown the opposite result in wound healing assay. (H,J) The enhanced carcinogenic activity was significantly reversed after treating with the specific SREBP inhibitor Fatostatin. (K,L) MBTPS2 knockdown promoted cell apoptosis (K) and its overexpression exhibited opposite effect (L). (M,N) MBTPS2 regulated the protein expressions related to invasion, metastasis and apoptosis in H1975 (M) and A549 (N) cells. (A,C,E,G,I-L) **, P<0.01; ***, P<0.001; ****, P<0.0001. (F,H) **, P<0.01; ****, P<0.0001, compared with A549-phage; ^#, P<0.05; ^###, P<0.001; ^####, P<0.0001, compared with A549-MBTPS2 + Fatostatin. CCK-8, Cell Counting Kit-8; FITC, fluorescein isothiocyanate; LUAD, lung adenocarcinoma; mRNA, messenger RNA; OD, optical density; PI, propidium iodide.

Potential impact of MBTPS2 in lipid metabolism for LUAD

The potential role of MBTPS2 in lipid metabolism was explored by analyzing its relationship with key lipid metabolism regulators. A positive correlation was observed between MBTPS2 and SREBP1 expression; SREBP1 levels were decreased upon MBTPS2 knockdown and markedly upregulated with MBTPS2 overexpression at both the mRNA and protein levels (Figure 10A,10B). Intracellular lipid content, as measured by flow cytometry, was suppressed in MBTPS2-knockdown cells (Figure 10C) and increased in MBTPS2-overexpressing cells (Figure 10D). Subsequently, the mRNA levels of key lipogenic enzymes FASN and SCD1 were significantly downregulated in MBTPS2-knockdown cells (Figure 10E) and upregulated in MBTPS2-overexpressing cells (Figure 10F). Furthermore, western blot analysis of eight paired clinical LUAD and adjacent normal tissue samples revealed that MBTPS2 protein levels were highly elevated in tumor tissues, and the positive correlation between MBTPS2 and SREBP1 expression has been confirmed in clinical specimens (Figure 10G). It was concluded that MBTPS2 may promote malignant features in LUAD by positively regulating key genes involved in lipid metabolism, including SREBP1, FASN, and SCD1.

Figure 10 The effects of MBTPS2 in regulating lipid metabolism in LUAD. (A,B) The SREBP1 expression was decreased when MBTPS2 was knocked down (A), while MBTPS2 overexpression upregulated the expression of SREBP1 (B) by western blot analysis. (C,D) MBTPS2 knockdown reduced lipid content (C), while its overexpression exerted the opposite effect (D) in flow cytometry. Count (%) represents the percentage of cells with positive Nile red staining (lipid droplets) as determined by flow cytometry. (E,F) MBTPS2 knockdown downregulated the mRNA expression levels of the key lipogenic enzymes FASN and SCD1 (E); on the contrary, overexpression of MBTPS2 increased their levels (F). (G) 8 pairs of clinical tissue samples suggested concurrent upregulation of MBTPS2 and SREBP1 in LUAD tissues compared to matched normal adjacent tissues. Notes: taking into account the compactness of the content, the parts of Figure 10A,10B and Figure 9B,9D are shared. Statistical difference was denoted as follows: **, P<0.01; ***, P<0.001; ****, P<0.0001. FASN, fatty acid synthase; LUAD, lung adenocarcinoma; mRNA, messenger RNA; SCD1, stearoyl-CoA desaturase 1.

Discussion

In present study, a robust prognostic signature comprising twelve lipid metabolism-related genes (HILPDA, ACHE, PPARD, ETNK1, ST3GAL2, NCOA2, HACD1, MBTPS2, UGT8, LPIN2, SPHK2 and AKR1C4) was developed and validated. This model effectively stratified LUAD patients into distinct risk groups with significant differences in OS. The predictive strength and accuracy of this signature were confirmed in an independent external cohort from TCGA. Furthermore, both univariate and multivariate Cox regression analyses established the risk score as an independent prognostic factor. Collectively, this work underscores the critical role of lipid metabolism in LUAD pathogenesis and provides a reliable tool for prognostic assessment, which may inform personalized therapeutic strategies for LUAD patients.

The genes comprising this prognostic signature have established roles in oncogenesis. For instance, ETNK1 (ethanolamine-kinase-1) involved in the de novo synthesis of membrane phospholipids, and recurrent ETNK1 mutations have been identified in a variety of myeloid neoplasms (19). HILPDA (hypoxia-inducible lipid droplet associated protein), also known as HIG2, mediates lipidomic remodelling to enhance the progression of non-alcoholic steatohepatitis-driven hepatocellular carcinoma and nasopharyngeal carcinoma (20,21). High levels of LPIN2 have been associated with lymph node metastasis and unfavourable prognosis in cervical cancer (22). Also, upregulation of LPIN2 in lung cancer stem cells (CSCs) drives triacylglycerol synthesis to enable lipid droplet-mediated reactive oxygen species (ROS) scavenging, thereby bolstering CSC resistance and tumor initiation (23). ACHE has been shown to induce platinum resistance in gastric cancer via regulating glycerophospholipid metabolism and has been implicated in the loss of innervation observed in inclusion body myositis (24). AKR1C4 (Aldo-Keto reductase family 1 member c4), while primarily hepatic and involved in regulating circulating steroid hormone levels, is upregulated and correlates with a poor prognosis in breast and nasopharyngeal carcinomas (25,26). PPARD (peroxisome proliferator-activated receptor delta), also known as NR1C2, promotes gastric carcinogenesis in mice through regulating FA oxidation (27). SPHK2 improves tumor expansion and metastasis in non-small cell lung cancer via activation of the Wnt/β-catenin molecular pathway (28). In addition, Wang et al. reported that SPHK2 (sphingosine kinase 2) was overexpressed in regorafenib-resistant hepatocellular carcinoma cells, and its inhibition reserved this acquired resistance (29). ST3GAL2 (ST3 Beta-galactoside alpha-2,3-sialyltransferase 2), in conjunction with ST3GAL1, promotes tumor survival in melanoma and its overexpression in docetaxel-resistant prostate cancer (30,31).

Among 12 genes, UGT8, MBTPS2, ST3GAL2 and HACD1 expression exhibited statistical significance between LUAD tissues and match normal tissues from TCGA normal and GTEx data.

To evaluate their prognostic value, we analyzed the four genes in LUAD using the Kaplan-Meier Plotter database. The results indicated that UGT8 and MBTPS2 were negatively correlated with patient survival, in both OS and first progression (FP). This finding is consistent with the GEO analysis. Considering that previous studies have reported that UGT8 is closely correlated with progression in LUAD (32,33), MBTPS2 was selected further functional characterization. Quantitative RT-PCR analysis revealed that MBTPS2 was significantly upregulated in multiple LUAD cell lines compared to a normal bronchial epithelial cell line, a finding consistent with our bioinformatic predictions and subsequent western blot validation. MBTPS2, also known as Site-2 Protease (S2P), is an integral membrane protein and plays crucial function in cholesterol and FA metabolism (34,35). While one study reported that high MBTPS2 expression predicted a favorable prognosis and that its overexpression may suppressed migration and invasion in osteosarcoma (36), bioinformatic analyses have consistently identified it as an unfavorable factor in LUAD, though functional validation was lacking (37-39). Our experimental data unequivocally demonstrate that MBTPS2 acts as a tumor promoterin LUAD. We found that MBTPS2 overexpression significantly enhanced the proliferative, migratory, and invasive capacities of LUAD cells while inhibiting apoptosis, whereas MBTPS2 knockdown produced the opposite effects. Mechanistically, MBTPS2 knockdown in H1975 cells led to reduced MMP9 protein expression and a decreased Bcl-2/Bax ratio, while its overexpression in A549 cells had the opposite effect. These findings propose that MBTPS2 may promote invasion and migration while suppressing apoptosis, potentially through the regulation of MMP-9, Bcl-2, and Bax.

Dysregulated lipid metabolism is a recognized hallmark of cancer, including LUAD, and is frequently characterized by the coordinated overexpression of key lipogenic enzymes such as FASN, ACC, ACLY, and SCD1 (40,41). High expression of FASN and SCD1 is correlated with tumor progression in LUAD (42). The sterol regulatory element-binding protein (SREBP) family of transcription factors, particularly SREBP1, are master regulators of lipogenesis. SREBP1 activates the transcription of key genes, including FASN and SCD1 (43,44). Given that MBTPS2 is known to process and activate SREBP transcription factors (34), we investigated its relationship with this pathway. Consistently, our data confirmed a positive correlation between MBTPS2 expression and the mRNA levels of SREBP1, FASN, and SCD1 in LUAD cells. Furthermore, both MBTPS2 and SREBP1 protein levels were elevated in clinical LUAD specimens compared to matched adjacent normal tissues. In addition, after using the specific inhibitor of SREBP in MBTPS2 overexpressing cells, carcinogenic activity of MBTPS2 was significantly suppressed. Hence, we concluded that MBTPS2 may be partially dependent on SREBP signaling. Finally, Nile red staining demonstrated that intracellular lipid content was reduced upon MBTPS2 knockdown and increased upon its overexpression. Based on its established role in SREBP activation, we propose a model wherein MBTPS2 promotes LUAD progression by upregulating SREBP1 signaling to drive a lipogenic program that fuels malignant growth.

In summary, this study developed and validated a lipid metabolism-related 12-gene signature which effectively predicts prognosis in LUAD patients. Furthermore, we established that MBTPS2 functions as a tumor promoter, promoting proliferation and lipogenesis while inhibiting apoptosis. Our findings provide novel insights into the role of lipid metabolism in LUAD pathogenesis and highlight MBTPS2 as a potential therapeutic target. Nevertheless, this study has certain limitations. First, the prognostic model was validated in only one independent cohort; future validation using multi-center prospective cohorts is warranted. Second, the oncogenic role of MBTPS2 was demonstrated primarily through in vitro models; its function and therapeutic potential should be further investigated in vivo.

Conclusions

In summary, a prognostic signature based on twelve lipid metabolism-related genes was developed and demonstrated to be a robust and independent predictor of OS in patients with LUAD. Among these genes, MBTPS2 was found to be significantly upregulated in LUAD and was functionally validated as a promoter of oncogenic phenotypes, enhancing proliferation while suppressing apoptosis. Furthermore, MBTPS2 expression was positively correlated with key lipid metabolism regulators, including SREBP1, FASN, and SCD1, indicating its role in regulating lipid metabolism in LUAD. This study sheds light on the molecular mechanisms linking lipid metabolism to LUAD pathogenesis and offers a foundation for future clinical applications, potentially contributing to improved risk stratification and personalized therapeutic strategies.

Acknowledgments

We would like to appreciate the contributors of data from databases for free use involved in the present study. We also thank the technical assistance and helpful discussions from our laboratory members.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1-2665/rc

Data Sharing Statement: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1-2665/dss

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1-2665/prf

Funding: This work was funded by the Key Laboratory of Artificial Blood Vessels and Valves, Health Commission of Jiangxi Province (No. G/Y3534), and the Wu Jieping Medical Foundation (No. J1998).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1-2665/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This research was approved by the Ethics Committee of Shangrao People’s Hospital [approval No. (2025) Physician Review (109)]. All procedures were performed in accordance with the Declaration of Helsinki and its subsequent amendments. Written informed consent was acquired from all participants following a comprehensive explanation of the study. All participants data were de-identified to ensure privacy.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Bray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2024;74:229-63. [Crossref] [PubMed]
2. Qing B, Li X, He X, et al. Characterizing Stage-Specific Cellular Dynamics and Microenvironmental Remodeling in Lung Adenocarcinoma by Single-Cell RNA Sequencing. Adv Sci (Weinh) 2026;13:e10847. [Crossref] [PubMed]
3. Alipour M, Moghanibashi M, Naeimi S, et al. LINC00894, YEATS2-AS1, and SUGP2 genes as novel biomarkers for N0 status of lung adenocarcinoma. Sci Rep 2025;15:10628. [Crossref] [PubMed]
4. Setlhare TC, Mpolokang AG, Flahaut E, et al. Determination of lung cancer exhaled breath biomarkers using machine learning-a new analysis framework. Sci Rep 2025;15:26085. [Crossref] [PubMed]
5. Fan Y, Wu Q, Lou M. Reevaluating the role of surgery in lung cancer with liver metastases: a propensity score-matched survival analysis. Discov Oncol 2025;16:521. [Crossref] [PubMed]
6. Chen L, Qi T, Zhang B, et al. NT5E (CD73) as a prognostic biomarker and therapeutic target associated with immune infiltration in lung adenocarcinoma. Sci Rep 2025;15:4340. [Crossref] [PubMed]
7. Yokochi Y, Miura M, Hamakawa M. Retrospective database study of druggable mutation detection among patients with non-small cell lung cancer in Japan. Sci Rep 2025;15:17692. [Crossref] [PubMed]
8. Terashima Y, Park S, Ikeuchi H, et al. Discovery of Novel RASGRF2 Fusions as a Therapeutic Target in Lung Adenocarcinoma of Never or Light Smokers. Cancer Sci 2025;116:2868-81. [Crossref] [PubMed]
9. Oliviero B, Caretti A, Mondelli MU, et al. Lipid Metabolism Reprogramming in Tumor-Associated Macrophages Modulates Their Function in Primary Liver Cancers. Cancers (Basel) 2025;17:1858. [Crossref] [PubMed]
10. An Y, Song H, Qiu H, et al. Lipid Metabolism in Gastrointestinal Malignancies: Exploring Dysregulation, Biomarkers, and Treatment Strategies. Cancer Med 2025;14:e70975. [Crossref] [PubMed]
11. He T, Wang ZY, Xu B, et al. CXCL6 Reshapes Lipid Metabolism and Induces Neutrophil Extracellular Trap Formation in Cholangiocarcinoma Progression and Immunotherapy Resistance. Adv Sci (Weinh) 2025;12:e2503009. [Crossref] [PubMed]
12. Cífková E, Brumarová R, Ovčačíková M, et al. Lipidomic and metabolomic analysis reveals changes in biochemical pathways for non-small cell lung cancer tissues. Biochim Biophys Acta Mol Cell Biol Lipids 2022;1867:159082. [Crossref] [PubMed]
13. Fu D, Zhang B, Fan W, et al. Fatty acid metabolism prognostic signature predicts tumor immune microenvironment and immunotherapy, and identifies tumorigenic role of MOGAT2 in lung adenocarcinoma. Front Immunol 2024;15:1456719. [Crossref] [PubMed]
14. Nuzzo PV, Rodrigues S, Ribeiro CF, et al. Targeting cancer metabolism: Therapeutic potential of the fatty acid synthase (FASN) inhibitors. Crit Rev Oncol Hematol 2025;214:104910. [Crossref] [PubMed]
15. Wu T, Ji M, Li T, et al. The molecular and metabolic landscape of ferroptosis in respiratory diseases: Pharmacological aspects. J Pharm Anal 2025;15:101050. [Crossref] [PubMed]
16. Shen Y, Chen JQ, Li XP. Differences between lung adenocarcinoma and lung squamous cell carcinoma: Driver genes, therapeutic targets, and clinical efficacy. Genes Dis 2025;12:101374. [Crossref] [PubMed]
17. Jia J, Feng S, Wang R, et al. Metabolic reprogramming-driven stratification and therapeutic targeting in lung adenocarcinoma: implications for prognosis and personalized treatment. Front Oncol 2025;15:1696117. [Crossref] [PubMed]
18. Yin F, Yang J, Luo H, et al. Targeting oncogenic activation of FLT3/SREBP/FASN promotes the therapeutic effect of quizartinib involving disruption of mitochondrial phospholipids. Cell Death Dis 2025;16:327. [Crossref] [PubMed]
19. Tessier S, He R, Greipp P, et al. Molecular and clinical characterization of ETNK1-mutated myeloid neoplasms: the Mayo Clinic experience. Blood Adv 2024;8:4807-11. [Crossref] [PubMed]
20. Povero D, Chen Y, Johnson SM, et al. HILPDA promotes NASH-driven HCC development by restraining intracellular fatty acid flux in hypoxia. J Hepatol 2023;79:378-93. [Crossref] [PubMed]
21. Zhang Y, Pang C, Zhang C, et al. HILPDA-mediated lipidomic remodelling promotes radiotherapy resistance in nasopharyngeal carcinoma by accelerating mitophagy. Cell Mol Life Sci 2023;80:242. [Crossref] [PubMed]
22. Yang P, Ruan Y, Yan Z, et al. Comprehensive analysis of lymph nodes metastasis associated genes in cervical cancer and its significance in treatment and prognosis. BMC Cancer 2021;21:1230. [Crossref] [PubMed]
23. Lin JH, Hsu TW, Cheng WC, et al. YAP/TEAD-activated TAG synthesis and peroxidation in lipid droplets confer ROS resistance in cancer stem cells. Redox Biol 2026;89:103968. [Crossref] [PubMed]
24. Chen M, Zhang C, Li H, et al. PLA2G4A and ACHE modulate lipid profiles via glycerophospholipid metabolism in platinum-resistant gastric cancer. J Transl Med 2024;22:249. [Crossref] [PubMed]
25. Shaon MA, Ansari F, Mahmud Z, et al. Genetic and computational analysis of AKR1C4 gene rs17134592 polymorphism in breast cancer among the Bangladeshi population. Sci Rep 2025;15:27526. [Crossref] [PubMed]
26. Guo SS, Chen YZ, Liu LT, et al. Prognostic significance of AKR1C4 and the advantage of combining EBV DNA to stratify patients at high risk of locoregional recurrence of nasopharyngeal carcinoma. BMC Cancer 2022;22:880. [Crossref] [PubMed]
27. Pudakalakatti S, Titus M, Enriquez JS, et al. Identifying the Metabolic Signatures of PPARD-Overexpressing Gastric Tumors. Int J Mol Sci 2022;23:1645. [Crossref] [PubMed]
28. Wang X, Liu C, Wang J, et al. Resveratrol suppresses NSCLC cell growth, invasion and migration by mediating Wnt/β-catenin pathway via downregulating SIX4 and SPHK2. J Chemother 2024;36:411-21. [Crossref] [PubMed]
29. Shi W, Zhang S, Ma D, et al. Targeting SphK2 Reverses Acquired Resistance of Regorafenib in Hepatocellular Carcinoma. Front Oncol 2020;10:694. [Crossref] [PubMed]
30. Kar S, Yanke AF, Berico P, et al. Integrated in vivo functional screens and multiomics analyses identify α-2,3-sialylation as essential for melanoma maintenance. Sci Adv 2025;11:eadg3481.
31. Drápela S, Kvokačková B, Slabáková E, et al. Pre-existing cell subpopulations in primary prostate cancer tumors display surface fingerprints of docetaxel-resistant cells. Cell Oncol (Dordr) 2025;48:205-18. [Crossref] [PubMed]
32. Ji J, Xie M, Qian Q, et al. SOX9-mediated UGT8 expression promotes glycolysis and maintains the malignancy of non-small cell lung cancer. Biochem Biophys Res Commun 2022;587:139-45. [Crossref] [PubMed]
33. Lemay AM, Courtemanche O, Couttas TA, et al. High FA2H and UGT8 transcript levels predict hydroxylated hexosylceramide accumulation in lung adenocarcinoma. J Lipid Res 2019;60:1776-86. [Crossref] [PubMed]
34. Tibbo AJ, Hartley A, Vasan R, et al. MBTPS2 acts as a regulator of lipogenesis and cholesterol synthesis through SREBP signalling in prostate cancer. Br J Cancer 2023;128:1991-9. [Crossref] [PubMed]
35. Bienvenue-Pariseault J, Sagrillo-Fagundes L, Wong-Yen P, et al. Melatonin Induces PERK-ATF4 Unfolded Protein Response and Apoptosis in Human Choriocarcinoma Cells. J Pineal Res 2025;77:e70072. [Crossref] [PubMed]
36. Ning B, Liu Y, Xu T, et al. Construction and validation of a prognostic model for osteosarcoma patients based on autophagy-related genes. Discov Oncol 2022;13:146. [Crossref] [PubMed]
37. Zhang MY, Huo C, Liu JY, et al. Identification of a Five Autophagy Subtype-Related Gene Expression Pattern for Improving the Prognosis of Lung Adenocarcinoma. Front Cell Dev Biol 2021;9:756911. [Crossref] [PubMed]
38. Zhang F, Xie S, Zhang Z, et al. A Novel Risk Model Based on Autophagy Pathway Related Genes for Survival Prediction in Lung Adenocarcinoma. Med Sci Monit 2020;26:e924710. [Crossref] [PubMed]
39. Liu Y, Lin W, Qian H, et al. Integrated multi-omic analysis and experiment reveals the role of endoplasmic reticulum stress in lung adenocarcinoma. BMC Med Genomics 2024;17:12. [Crossref] [PubMed]
40. Wu J, Chen Y, Zou H, et al. 6-Phosphogluconate dehydrogenase promotes glycolysis and fatty acid synthesis by inhibiting the AMPK pathway in lung adenocarcinoma cells. Cancer Lett 2024;601:217177. [Crossref] [PubMed]
41. Deng B, Kong W, Shen X, et al. The role of DGAT1 and DGAT2 in regulating tumor cell growth and their potential clinical implications. J Transl Med 2024;22:290. [Crossref] [PubMed]
42. Yang X, Zhang Q, Wei L, et al. HIF1A/PCDH7 axis mediates fatty acid synthesis and metabolism to inhibit lung adenocarcinoma anoikis. J Biochem Mol Toxicol 2024;38:e70001. [Crossref] [PubMed]
43. Fei X, Huang J, Li F, et al. The Scap-SREBP1-S1P/S2P lipogenesis signal orchestrates the homeostasis and spatiotemporal activation of NF-κB. Cell Rep 2023;42:112586. [Crossref] [PubMed]
44. Li S, Yang H, Zhou W, et al. Selenium Nanoparticles Decorated With Stevioside Potentially Attenuate Fructose Palmitate Induced Lipid Accumulation in HepG2 Cells. Mediators Inflamm 2025;2025:7942947. [Crossref] [PubMed]