AGPAT family members act as potential predictive biomarkers and therapeutic targets in cervical cancer Authors: Nannan Gui, Hanyi Pan, Weijuan Lu, Guiqiong Pan, Xiaoyu Zhou, Yuzhen Chen, Mingyang Jin, Changyong Yang, Mingyou Dong, Yuexiu Liang Published: 2026 ABSTRACT AbstractCervical squamous cell carcinoma and endocervical adenocarcinoma (CESC) are the primary histological subtypes of cervical cancer. Although the AGPAT family of enzymes is implicated in various cancer types, the specific roles of its members in cervical cancer remain unclear. In the present study, we investigated the value of AGPAT1–5 as potential biomarkers and therapeutic targets in cervical cancer by assessing their impact on disease development and outcomes. AGPAT gene expression data and clinical information from 306 patients with CESC and three control subjects were collected from The Cancer Genome Atlas. Given the limited number of normal cervical samples in TCGA (n = 3), we utilized the GEPIA database to integrate GTEx normal cervical samples (n = 10) as controls for differential expression analysis. These data were analyzed for differential gene expression, gene-gene and protein-protein interactions, prognostic and diagnostic value, clinical correlations, functional enrichment, and various tumor-infiltrating immune cell types. Validation was performed using two independent Gene Expression Omnibus (GEO) datasets (GSE6791 and GSE63514). The study revealed that AGPAT4 mRNA expression was significantly downregulated in cervical cancer tissues compared to normal tissues in the GEPIA analysis, while AGPAT1, AGPAT2, AGPAT3, and AGPAT5 showed no significant differences. Validation in GEO datasets demonstrated that AGPAT1, AGPAT2, and AGPAT3 were consistently downregulated in tumor tissues, whereas AGPAT5 was upregulated, and AGPAT4 showed no significant change. High levels of AGPAT3 and AGPAT4 expression, in particular, were associated with a worse prognosis in CESC patients. Immune infiltration analysis restricted to CESC samples revealed that AGPAT3 expression was significantly correlated with multiple immune cell types, including positive correlations with Macrophages M1, T cells CD4 memory resting, and Monocytes. Furthermore, guided by functional enrichment analysis implicating immune-related pathways, we examined the correlation between AGPAT3 expression and key T cell regulatory molecules, including FOXP3, IL2RA, IKZF2, and CCR8, revealing significant positive associations. In vitro assays demonstrated that knocking out AGPAT3 expression significantly decreased the proliferation and migration of HeLa and C-33 A cervical cancer cell line. These results suggest that AGPAT3 could be a valuable biomarker and a promising therapeutic target in cervical cancer.Supplementary InformationThe online version contains supplementary material available at 10.1007/s12672-026-05322-4.Keywords: Cervical cancer, Prognosis, Immune infiltration, TCGA, AGPAT3 Introduction Cervical squamous cell carcinoma and endocervical adenocarcinoma (CESC) are the predominant histological forms of cervical cancer [1]. Despite widespread cervical cancer screening [2, 3] and human papillomavirus vaccination programs [4], cervical cancer remains a significant global health issue in terms of both incidence and mortality [5, 6]. Cervical cancer is currently the fourth most common cancer and a leading cause of death among women worldwide, with 670,000 cases diagnosed and 350,000 deaths reported in 2024 [7]. This alarming statistic underscores the urgent need for more effective predictive biomarkers. Currently, surgery and radiotherapy are the most effective treatment modes for cervical cancer [8]; however, these interventions are not equally accessible across the globe, resulting in disproportionately high mortality rates [9], mainly in low-middle income countries. Patients in the advanced stages of disease, or those experiencing recurrence and metastasis, have a particularly poor prognosis [10, 11]. Several studies have employed bioinformatics methods to identify key genes associated with cervical cancer, yielding preliminary results and promising research directions. However, many of these studies are hindered by a lack of clinical samples and cell experiments for validation [12], or they fail to investigate the specific mechanisms through which these genes contribute to the development of cervical cancer [13, 14]. This limitation restricts the capacity of these studies to advance further and make significant progress in clinical treatment. Therefore, understanding the molecular mechanisms underlying cervical cancer and identifying effective diagnostic/prognostic biomarkers are essential for improving patient survival.1-acylglycerol-3-phosphate-O-acyltransferases (AGPATs), also referred to as lysophosphatidyl acyltransferases (LPAATs), are a group of enzymes essential for converting lysophosphatidic acid (LPA) to phosphatidic acid (PA) in the triacylglycerides (TAG) biosynthesis pathway [15]. Currently, 11 acylglycerol-3-phosphate acyltransferases (AGPATs), including AGPAT1-5, have been identified, and their molecular structures have been elucidated, with additional members being discovered [16]. In this study, AGPATs 1–5 were selected as the research subjects based on a comprehensive review of the literature [17, 18]. AGPATs 1–5 specifically utilize acyl-CoA as the acyl donor and lysophospholipid as the acyl acceptor. They play a role in cancer cell survival and growth [19] by participating in the production of TAGs and storing lipids in lipid droplets (LDs), structures which are crucial for storing energy and limiting lipotoxicity. Additionally, AGPATs are involved in the synthesis of phospholipids, which are required for new membrane formation in rapidly dividing cells and in the regulation of cell signaling pathways. Recent research indicates that AGPATs are also implicated in membrane fission, vesicular transport, and communication between cancer cells and the tumor microenvironment (TME); collectively, these factors contribute to cancer metastasis. Multiple research studies of various cancer types have emphasized the importance of AGPATs in cancer cell proliferation, tumor growth, metastasis, patient prognosis, and cancer type categorization. For example, Ren et al. identified elevated AGPAT1 expression as an unfavorable indicator in colorectal cancer, which was associated with an increased risk of recurrence and reduced survival [20]. Similarly, Song et al. showed that inhibiting AGPAT2 prototed the death of cisplatin-resistant osteosarcoma cell lines [21]. Moreover, variations in AGPAT expression levels between cancerous and healthy tissues, combined with computational analyses, have resulted in the incorporation of AGPAT isoforms into predictive metabolic gene profiles. AGPATs plays a vital role in the development [22] of tumors; as such, they can be used as indicators of prognosis [23] and response to immunotherapy. However, the specific role of AGPATs in CESC and their potential impact on cancer prognosis require further investigation. In the present study, The Cancer Genome Atlas (TCGA, https://portal.gdc.cancer.gov/) and the Genotype-Tissue Expression (GTEx, https://commonfund.nih.gov/gtex) database were used as sources of transcriptomic data derived from cervical cancer and normal tissue samples. Given the limited number of normal cervical samples in TCGA (n = 3), we utilized the GEPIA database to integrate GTEx normal cervical samples (n = 10) as controls for more robust differential expression analysis. The differential expression of AGPAT genes in cervical cancer versus normal tissues was analyzed using the GEPIA database and validated in two independent GEO datasets. Genetic alterations affecting AGPAT family members and molecular mechanisms underlying AGPAT signaling were investigated in the context of immunotherapy. We also performed focused immune infiltration analyses restricted to CESC samples using ssGSEA and CIBERSORTx algorithms, and examined correlations between AGPAT3 and T cell key regulatory molecules. In vitro assays were conducted to validate the functional role of AGPAT3 in cervical cancer cell lines. We envision that the implementation of AGPATs as biomarkers and potential therapeutic targets will help improve the prognosis and management of patients with CESC. Materials and methods Patients data The present study collected tissue samples from nine patients with CESC who had undergone surgery at the Affiliated Hospital of Youjiang Medical University for Nationalities in Guangxi, China. All participants provided written informed consent. Gene expression data (HTSeq-FPKM) of 306 patients with CESC were obtained from TCGA database (https//portal.gdc.cancer.gov/). Normal cervical tissue data (n = 10) were obtained from the GTEx database. Given the limited number of normal cervical samples in TCGA (n = 3), differential expression analysis was performed using the GEPIA database (http://gepia.cancer-pku.cn/), which integrates TCGA tumor samples (n = 306) with GTEx normal samples (n = 10) as controls. The validation was conducted using two independent datasets from the Gene Expression Omnibus (GEO; accessions GSE6791 and GSE63514). GSE6791 comprises 20 CESC tumors and 8 normal cervical tissues, and GSE63514 includes 28 CESC tumors and 24 normal cervical tissues. CESC patient clinical information, including pathologic stage, TNM stage, histologic stage, age, and OS are presented in the following figures. Experimental validation was conducted using clinical specimens and in vitro cellular experiments. mRNA expression AGPAT1-5 mRNA expression was detected in normal human tissues of healthy individuals using the ‘GTEx Expression’ module within the GSCA database (http://bioinfo.life.hust.edu.cn/GSCA/) [24]. Additionally, interactive body maps of AGPAT1-5 were created using the TIMER database (https://cistrome.shinyapps.io/timer/) [25] to visualize the median expression in normal tissues and cancers. Differential expression of AGPAT family members between CESC and normal cervical tissues was analyzed using the GEPIA database (http://gepia.cancer-pku.cn/), which integrates TCGA tumor samples (n = 306) with GTEx normal samples (n = 10) to overcome the limitation of limited normal samples in TCGA. A significance level of p-value < 0.05 was applied. Validation was performed using the GEO dataset GSE6791 and GSE63514. AGPAT genetic alterations and interaction networks Genetic changes [26] affecting AGPAT family members were examined in the 306 CESC samples obtained from TCGA Firehose Legacy project through the cBioPortal database (https://www.cbioportal.org/) [27]. Mutation frequencies were compared to the average mutation rate in CESC (3.5 mutations/Mb) for context. GGI (gene-gene interaction) among AGPAT gene family members were predicted using the GeneMANIA database [28]. The STRING database (https://cn.string-db.org/cgi/input.pl), was then used to contruct an AGPAT PPI (protein-protein interactions) network, by specifying parameters such as organism (“Homo sapiens”), network type (“full STRING network”), and minimum interaction score (“medium confidence 0.400”). Spearman correlation analysis was used to evaluate the relationship between the expression of AGPAT family members and gene expression data obtained from CESC samples in TCGA database. Prognostic and diagnostic value analysis The Xiantao Academic database (https://www.xiantao.love/) was accessed to perform univariate and multivariate Cox regression analyses on AGPAT family members in CESC. Patients were stratified into high and low expression groups using the median expression level as the cutoff, which is a common approach in cancer biomarker studies and allows for balanced group sizes. The survival rates of patients with cervical squamous cell carcinoma who had high levels of AGPAT3/4 were evaluated using the Kaplan-Meier method. Logistic regression was employed to investigate the correlation between AGPAT3 expression and clinical variables. A nomogram was created to forecast the 1-, 3-, and 5-year OS rates of patients with CESC by combining AGPAT3/4 expression data with four clinical indicators (T, N stages, histologic stage and clinical stage) of CESC patients from TCGA database. The prognostic value of the nomogram in CESC was assessed using the ‘pROC’ R package. Furthermore, the diagnostic efficacy of AGPAT members in detecting CESC in the Xiantao Academic database was assessed using the ‘pROC’ R package. A p-value < 0.05 was deemed as a measure of statistical significance. The AUC of the ROC curve was used to summarize the diagnostic efficacy of the biomarkers. Clinical correlation analysis The “Expression” module within the GSCA database (https://guolab.wchscu.cn/GSCA/#/) was employed to examine the relationship between AGPAT1–5 and different survival measures (i.e., DSS, OS, PFS, DSS) and clinical stage among CESC patients. Furthermore, the Xiantao Academic database (https://www.xiantao.love/) was used to explore the correlation between AGPAT3 gene expression and various clinical characteristics including TNM stage, tumor grade, patient age, tumor histological type, and OS. Immune cell infiltration analysis The relationship between AGPAT3 and 24 types of infiltrating immune cells was explored specifically in CESC samples using two complementary algorithms. First, single-sample Gene Set Enrichment Analysis (ssGSEA) was performed using the “GSVA” R package to estimate the enrichment scores of 24 immune cell types based on their specific gene signatures. Second, the CIBERSORTx algorithm was applied to deconvolute the relative abundance of 22 immune cell types in each tumor sample using the LM22 signature matrix with 100 permutations. The ESTIMATE method [29] was employed to characterize the TME of each tumor specimen by calculating the ImmuneScore, StromalScore, and ESTIMATEScore. Correlations between AGPAT3 expression and immune cell infiltration were assessed using Spearman correlation analysis with FDR correction. Furthermore, to validate the association between AGPAT3 and T cell regulatory, we analyzed the correlation between AGPAT3 expression and key T cell regulatory molecules, including CXCL9, CXCL10, PDCD1, and HAVCR2. Spearman correlation coefficients were calculated and visualized using scatter plots with fitted regression lines and 95% confidence intervals. Functional enrichment analysis Patients with CESC in the CESC dataset of TCGA were divided into the AGPAT3-high and AGPAT3-low groups, using the median expression level of AGPAT3 as the cutoff value. The ‘DESeq2’ R package was employed to detect DEGs between the two groups, whereby upregulated DEGs were defined by log2 fold change values ≥ 1.5 and downregulated DEGs by log2 fold change values ≤ − 1.5. A significance threshold of FDR < 0.05 was established after Benjamini-Hochberg correction for multiple testing. Next, an analysis of KEGG and GO enrichments was carried out with the ‘ClusterProfiler’ R package to explore the functions of the DEGs related to AGPAT3 levels in CESC. Immune cell infiltration analysis The relationship between AGPAT3 and 24 types of infiltrating immune cells was explored using the “GSVA” software package. The ESTIMATE method, which examines stromal and immune cells in 33 tumors using gene expression information, was employed to characterize the TME of each tumor specimen by calculating the ImmuneScore, StromalScore, and ESTIMATEScore. Correlations were adjusted for tumor purity using ESTIMATE scores. The relationship between AGPAT3 expression and the infiltrating immune cells in the TME was examined across 33 cancer types using the ‘”Limma” R package. A Spearman correlation heat map was generated to visualize the relationship between AGPAT3 and various immune cell types. Furthermore, the connection between AGPAT3 and immunoregulatory genes, including those encoding the major histocompatibility complex (MHC), immune activators, chemokines, and chemokine receptors, was investigated by analyzing TCGA data. Additionally, the correlation between AGPAT3 expression and that of immune checkpoint molecules was examined through Spearman correlation analysis with FDR correction. Patient tissue specimens The present study collected tissue samples from twenty-four patients with CESC who had undergone surgery at the Affiliated Hospital of Youjiang Medical University for Nationalities in Guangxi, China between 2020 and 2023. Tumor and paracancerous tissue sample pairs were collected; six pairs were paraffin-embedded and eighteen pairs were kept fresh for qRT-PCR analysis. Ethical approval for the research was granted by the hospital’s Ethics Committee (IRB number YYFY-LL-2023HY2-06). The study was performed in accordance with the Declaration of Helsinki and its amendments. All participants provided written informed consent. CESC was graded according to the International Federation of Gynecology and Obstetrics staging criteria. Clinicopathological findings were derived from surgical records and pathology reports. Immunohistochemistry The six pairs of cancerous and paracancerous tissues collected from CESC patients were fixed for 1 week in 10% formalin, prior to being embedded in paraffin. The tissue specimens were then sectioned to 4-µm sections, deparaffinized, and subjected to antigen retrieval by microwaving. Immunostaining was performed using a polyclonal antibody against AGPAT3 (Thermo Fisher Scientific, PA5-98855, USA, dilution 1:200). Meanwhile, surgical samples were preserved in 4% formaldehyde, cut into 4-µm sections, and then slides were heated at 60 °C for 1 h to avoid peeling. Paraffin was removed using xylene, rehydrated in an alcohol gradient, and then repaired with citric acid for 20 min. Next, the slides were treated with 3% hydrogen peroxide for 15 min, blocked with 5% bovine serum albumin (BSA) in PBS for 30 min, and then stained with primary antibodies in PBS containing 5% BSA overnight at 4 °C. The standard EnVision technique (Dako, Denmark) was used to detect protein binding based on the extent of color development. After counterstaining with hematoxylin and sealing with neutral resin, the stained sections were observed under a light microscope. The immunohistochemistry staining results were analyzed in a double-blinded manner using clinicopathological criteria. AGPAT3 staining was classified as either negative (0), weak (1), moderate (2), or strong (3). The h-score was determined by multiplying the various levels of staining (0, 1, 2, 3) by the proportion of cells positive for each level of staining using the equation: h-score = (1 × % weakly-stained cells) + (2 × [% moderately-stained cells]) + (3 × [% strongly-stained cells]) [30]. qRT-PCR Total RNA was extracted from eighteen pairs of fresh frozen CESC tumor and adjacent normal tissues using TRIzol reagent (Invitrogen, 15596026) according to the manufacturer’s instructions. Add 1 mL of Trizol and three grinding beads, then grind thoroughly and lyse on ice for 5 min. Next, add 200 µL of chloroform, mix well, and allow the mixture to stand for 5 min. Centrifuge at 12,000 rpm for 15 min, and then take 300 µL of the supernatant. Add an equal volume of pre-chilled isopropanol, mix thoroughly, and let it sit for an additional 5 min. Centrifuge again at 12,000 rpm for 10 min, discarding the supernatant while retaining the precipitate. Add 1 mL of pre-chilled 75% ethanol, gently vortex the solution, then centrifuge and discard the supernatant. Open the EP tube cap to allow the ethanol to evaporate completely. Add 20 µL of sterile enzyme-free water to the EP tube to obtain the RNA. Use a UV spectrophotometer to assess the purity and concentration of the RNA, ensuring it is suitable for the experiment. Reverse transcription was performed using the ToloScript ALL-in-one RT EasyMix for qPCR kit (TOLOBIO, Guangzhou, China) according to the manufacturer’s protocol. Briefly, 1 µg of DNase-treated RNA was reverse transcribed in a 20 µL reaction containing 4 µL 5× All-in-one RT Buffer and 1 µL All-in-one Enzyme Mix. The reaction was incubated at 50 °C for 15 min, followed by 85 °C for 5 s to inactivate the reverse transcriptase. No-RT controls were included for each sample to verify the absence of genomic DNA amplification. Follow the reverse transcription kit instructions to prepare the reverse transcription reaction system as outlined in Table 1, and reverse transcribe the RNA into cDNA. Prior to reverse transcription, dilute the RNA concentration to 1000 ng/µL with enzyme-free water. Set the reaction program to 50 °C for 15 min, followed by 85 °C for 5 s. Table 1Reverse transcription reaction systemRegentVolume5× All-in-one RT Buffer4µLAll-in-one Enzyme Mix1µLTemplate RNA(1 pg ~ 1 µg)RNase-free ddH2OUp to 20 µL Reverse transcription reaction system Prepare the qRT-PCR reaction system according to Table 2, add the reaction system to eight rows of connected tubes in sequence, shake to mix evenly, and centrifuge. Table 2PCR amplification and loading systemRegentVolumeDNA template1 µLForward Primer0.4 µLReverse Primer0.4 µLSYBR Green Fast qPCR Mix5 µLRNase-free WaterTo 20 µL PCR amplification and loading system Set up the qRT-PCR reaction program according to the manufacturer’s instructions. The RNA expression level of AGPAT3 was then detected using qRT-PCR (ToloScript ALL-in-one RT EasyMix for qPCR, TOLOBIO) with specific primers (Human-AGPAT3-F: CGCCTGTCGGACTACCCC; Human-AGPAT3-R: CTTAGCAGCCGCCACCTC). The reference gene β-actin was used with primers (F: CATGTACGTTGCTATCCAGGC; R: CTCCTTAATGTCACGCACGAT). Cell culture and AGPAT3 gene knockout The HeLa and C-33 A cervical cancer cell lines was acquired from the Shanghai Institute of Biochemistry and Cell Biology (SIBCB) and grown in DMEM (MeilunBio, MA0212) supplemented with 10% fetal bovine serum and 100 IU/mL penicillin-streptomycin. All cell lines were authenticated by short tandem repeat profiling and tested negative for mycoplasma contamination. The AGPAT3 knockout (KO) cell line was generated using the CRISPR-Cas9 system. Briefly, two independent sgRNAs (GTGCACGTGAGAAGCGTAAT, GCTAGGGTGAGCACCTTCCA) were designed using the CRISPOR website (http://crispor.tefor.net/). The annealed gRNA sequences were cloned into the BsmBI (Thermo Fisher Scientific, ER0451) site of the pSpCas9(BB)−2 A-Puro (PX459) V2.0 vector (Addgene, 62988). Subsequently, the targeting plasmids were transferred to HeLa cells along with PEI-MAX (Polysciences, 49553-93-7) at a ratio of 1:2, following the manufacturer’s instructions. After 48 h, cells were screened in culture using medium containing 1 µg/mL puromycin (Thermo Fisher Scientific, A1113802). Once the untransfected control cells were completely dead, the knockout cell lines were established through monoclonal selection and validated via Sanger sequencing, PCR and Western blotting analysis. Western blotting Cellular proteins were extracted and lysed with RIPA buffer. After being mixed with 1× SDS loading buffer, heat-denatured, and separated on an 10% SDS-PAGE gel, the proteins transferred onto a PVDF membrane (Immobilon, Millipore-p, IPVH00010, CA, USA) and subsequently blocked with 5% non-fat milk at room temperature. Next, the membrane was incubated overnight at 4 °C with a primary anti-AGPAT3 antibody (PA5-98855, Thermo, USA) and then a secondary antibody (Jackson ImmunoResearch, 115-035-004, USA) at room temperature. Protein bands were detected using enhanced chemiluminescence (ECL) after membrane washing. Vinculin (Abcam, ab129002, dilution 1:2000) was used as a loading control. Validation of gene expression levels in knockout cell lines WT and AGPAT3-KO HeLa cells were seeded into PCR tubes at a 1:4 ratio (v/v) with proteinase K (diluted 1:200; Biosharp, BL628A). After a 1-hour incubation in the PCR machine at 55 °C, genome extraction was peformed at 95 °C for 10 min. The extracted DNA was then combined with two primers (CATCTTCTGCACCTCTCACG and CACTGCTCCACCCAGTACCT) and the PCR Mix (Yeasen, 10157ES03), and subjected to PCR amplification as per the manufacturer’s instructions. The changes in DNA length were confirmed by analyzing the product using agarose gel electrophoresis. Cell proliferation assay Cell proliferation was assessed using the CCK-8 and colony formation assays. For the CCK-8 test, WT and AGPAT3-KO HeLa cells were placed in 96-well dishes and allowed to attach for 1–7 days, prior to a 1-hour incubation with CCK-8 solution (10 µL CCK-8 and 90 µL McCoy’s 5 A per well). Subsequently, the OD450 was assessed on a microplate reader. Statistical analysis was performed using two-way ANOVA with Tukey’s multiple comparisons test to account for both time and genotype effects. In the colony formation assay, WT and AGPAT3-KO HeLa cells were grown in 6-well dishes (at 400 cells/well) to promote colony formation. After a 2-week culture period, the cells were treated with anhydrous methanol, stained with Crystal Violet, dried, and finally imaged under a microscope. The quantity of cell colonies on each plate was measured using ImageJ software. Colonies were defined as clusters of ≥ 50 cells. Transwell migration assay Transwell chambers with an 8-µm pore size (Corning, Lowell, MA, USA) were placed in 24-well cell culture plates (BIOFIL, TCP010024, Guangzhou, China). The chambers were utilized for the cell migration assays in the absence of a gel matrix. The upper compartment of the Transwell received 500 µL of cell suspension in serum-free medium containing 5 × 104 cells, while the lower compartment was filled with 600 µL of medium containing serum. After a 24-hour incubation period, cells that had migrated and adhered to the underside of the chamber were fixed using anhydrous methanol. Any non-migratory cells were gently removed with a damp cotton swab, following which the chamber was allowed to dry. Subsequently, images were captured using a Nikon Eclipse TE2000-U inverted microscope at a total magnification of 100 × (10× objective and 10× eyepiece) in nine randomly selected regions and the cells were counted. Cells were counted in three independent experiments (n = 3 biological replicates). Wound healing assay WT and AGPAT3-KO HeLa cells were seeded at a 50% density in a six-well plate and incubated for 12 h until fully adhered to the surface. Next, a 200 µL yellow pipette tip was used to generate a ‘scratch’ by scraping the layer of adherent cells in a linear fashion following. After washing with PBS to eliminate debris, the cells were cultured in a serum-free medium. Photographs were taken at 0, 24, and 48 h with a Nikon Eclipse TE2000-U inverted microscope at 40× magnification (10× objective and 4× eyepiece) in five separate scratch locations to measure the size of the area that had not healed. The rate of wound closure was then calculated. Statistical analysis SPSS 25.0 software was used for statistical analysis, while Graphpad Prism 8 software was employed to generate graphs. The difference between groups was assessed through independent sample t-tests or paired sample tests. In addition, univariate and multivariate analyses regression analyses were performed using the Cox proportional hazards regression model. OS was assessed using the Kaplan-Meier method and Log-rank test. For immune infiltration analyses, Spearman correlation coefficients were calculated, and p-values were adjusted using the false discovery rate (FDR) method. Statistical significance was defined as FDR < 0.05. Statistically significant differences were determined on the basis of a p-value < 0.05. Results Gene expression of AGPAT family members First, we assessed the expression levels of AGPATs in various tissues such as adipose tissue, breast, liver, small intestine, and cervix (Fig. 1A). Interestingly, AGPAT5 expression was significantly upregulated in the pancreas, salivary gland, and blood vessels in relation to other tissues. Notably, the expression of AGPAT3–5 was lower than that of other AGPAT family members in the healthy cervix. Furthermore, AGPAT1–5 had different expression patterns in the 33 types of human tumors examined, characterized by tumor-type-specific upregulation or downregulation (Fig. 1B). These findings indicate that AGPAT family members exhibit tissue-specific expression. The distribution of AGPAT family in human normal and tumor tissues. (A) Heatmap of the expression profiles of AGPAT members in diverse tissues in the GTEx dataset. (B) The FPKM values of AGPAT genes in Multiple types of cancer. Boxplots were used to display the FPKM values for AGPAT1, AGPAT2, AGPAT3, AGPAT4, and AGPAT5 across various cancer types using data from the TCGA database Differential expression of AGPAT family members in cervical cancer The TIMER database was use to assess the expression of AGPATs in both pan-cancer and normal tissues. The results showed that AGPAT1–5 expression was elevated in various types of cancer (including BLCA, BRCA, CESC, CHOL, COAD, ESCA, GBM, HNSC, KIRC, KIRP, LIHC, LUAD, LUSC, READ, STAD, and UCEC), compared to that in normal tissues (Supplementary Fig. 1 and 2). Given the limited number of normal cervical samples in TCGA (n = 3), we utilized the GEPIA database to integrate GTEx normal cervical samples (n = 10) as controls for differential expression analysis. As shown in Supplementary Fig. 3, AGPAT4 mRNA expression was significantly downregulated in CESC tumor tissues (n = 306) compared to normal cervical tissues (n = 13) (p < 0.05). However, AGPAT1, AGPAT2, AGPAT3, and AGPAT5 showed no significant differences between tumor and normal tissues in this analysis (Supplementary Fig. 3). Specifically, the expression of AGPAT2, AGPAT3, and AGPAT5 mRNA was elevated, while that of AGPAT1 and AGPAT4 was reduced in CESC tissues. These findings were further validated in the independent GEO datasets GSE6791 and GSE63514, which showed that the expression levels of AGPAT1, AGPAT2, and AGPAT3 were significantly lower in tumor tissues compared with normal tissues. Conversely, AGPAT5 expression was significantly higher in cervical cancer tissues, while AGPAT4 exhibited no significant difference between the two groups (p < 0.001, t-test with FDR correction) (Supplementary Fig. 4). These findings suggest distinct that the expression levels of the AGPAT family members vary between CESC and normal tissues. Genetic alterations affecting AGPATs and the corresponding gene and protein interaction networks The frequencies of alterations in members of the AGPAT family were assessed in 306 CESC samples using the cBioportal tool. The result showed that AGPAT family members were rarely mutated (AGPAT1, 1.8%; AGPAT2, 1.1%; AGPAT3, 1.8%; AGPAT4, 2.2%; AGPAT5, 1.1%) compared to the average mutation rate in CESC (3.5 mutations/Mb), suggesting they are highly conserved (Fig. 2A). Analysis of the gene-gene interaction (GGI) network using the GeneMANIA database identified connections between AGPATs and 20 possible target genes (Fig. 2B). We then generated a protein-protein interactions (PPI) network and used the STRING platform to examine the relationship between AGPAT family members, revealing a significant (PPI enrichment p-value < 1.0 × 10− 16) association among them (Fig. 2C). In addition, the expression of AGPAT family genes was analyzed using TCGA data, revealing a strong positive relationship between most AGPAT members, particularly between AGPAT3 and AGPAT1/AGPAT5 (Fig. 2D). Genomic changes in AGPAT family members and the network of interactions between genes and proteins of the AGPAT family. (A) The cBioPortal database provides an overview of genomic alterations of the AGPAT family members in CESC patients. (B)The GeneMANIA database was used to analyze the gene network linked to the AGPAT family. (C) A diagram illustrating the connections among proteins produced by genes belonging to the AGPAT family. (D) Analyzing the correlation between members of the AGPAT family. Significance levels: * indicates p-value less than 0.05, ** indicates p-value less than 0.01, *** indicates p-value less than 0.001 Association between AGPAT expression and survival in CESC Univariate Cox regression analysis was used to show that AGPAT3 and 4 overexpression was linked to the poor prognosis of CESC patients. Moreover, multivariate Cox regression analysis revealed that elevated levels of AGPAT3 and 4 were independent predictors of overall CESC patient survival (Fig. 3A). These results were supported by the outcome of the Kaplan-Meier analysis, which showed that CESC patients with high AGPAT3 levels had reduced OS, disease-specific survival (DSS), and progression-free survival (PFS) (Fig. 3B–D). Similarly, AGPAT4 expression also correlated inversely with survival times (Fig. 3E–G). The findings indicate that AGPAT3 and 4 could serve as predictors of unfavorable outcomes in CESC. Analysis of the survival rate of the AGPAT gene family in CESC. (A) In CESC, AGPAT1-5 were analyzed using both univariate and multivariate Cox regression methods. Hazard ratios are shown with 95% confidence intervals. (B-D) Kaplan-Meier survival analysis was conducted to compare the survival rates of CESC patients with high and low levels of AGPAT3 expression, including overall survival, disease-specific survival, and progression-free interval survival. (E-G) Kaplan-Meier survival analysis was conducted to compare the survival rates of CESC patients with high and low levels of AGPAT4 expression, including overall survival, disease-specific survival, and progression-free interval survival Prognostic and diagnostic significance of AGPATs in CESC Having identified AGPAT3 and AGPAT4 expression as potential prognostic biomarkers of CESC, we combined them with four clinical indicators in a nomogram designed to predict the OS of CESC patients, whereby the size of the nomogram score correlated inversely with CESC patient survival time (Fig. 4A). The calibration graph was used to show a strong correlation between the predicted and observed outcomes (Fig. 4B). Thus, we demonstrated that the combination of AGPAT3/4 mRNA expression levels and four clinical indicators in a nomogram functioned as a promising prognostic tool for predicting OS in CESC, surpassing the performance of each of the individual prognostic indicators. Next, the diagnostic accuracy of AGPAT family members was evaluated through a receiver operating characteristic (ROC) curve analysis. AGPAT1–4 displayed AUC values of 0.839, 0.688, 0.836, and 0.893, respectively (Fig. 4C). Given that all these AUC values exceeded 0.6, AGPAT1–4 (and especially AGPAT3 and 4) may serve as potential diagnostic biomarkers for CESC patients. The AGPAT family has prognostic and diagnostic significance in CESC. (A) A nomogram combines the levels of AGPAT3/4 expression with T, N stages, histologic stage and clinical stage to forecast the likelihood of 1-year survival for patients with CESC. (B) The nomogram model’s calibration curve. (C) ROC analysis of AGPAT family members. AGPAT1-4 exhibits strong discriminatory ability between CESC and healthy tissues, with AUC values above 0.6, suggesting they could serve as valuable cancer biomarkers for clinical use. Confidence intervals are shown for AUC values Clinical correlation analysis Interrogation of the GSCA database revealed a significant association between AGPAT1–5 expression and survival outcomes (DSS, OS, PFS) as well as clinical stage in CESC patients. The expression of AGPAT3 and AGPAT1 in particular was strongly associated with reduced survival and late disease stage (Supplementary Fig. 5A–B). Additionally, AGPAT3 mRNA expression was notably higher in patients aged > 50 years compared to those aged ≤ 50 years (Supplementary Fig. 5H). Moreover, patients with adenocarcinoma and adenosquamous exhibited higher levels of AGPAT3 expression than those with squamous cell carcinoma (Supplementary Fig. 5I), however, the difference between them was not statistically significant after FDR correction. These results suggest that increased levels of AGPAT3 mRNA expression are associated with negative prognostic factors such as reduced survival time and an increased risk of progressing to advance stages of disease. Functional enrichment analysis of AGPAT3 in CESC The high levels of AGPAT3 expression detected in CESC tissues were associated with poor survival outcomes. However, the impact of this elevated AGPAT3 expression on signaling pathways implicated in CESC remained ambiguous. We therefore next conducted a functional enrichment analysis of gene sets related to AGPAT3 in CESC. Using a CESC dataset of TCGA, we identified 286 genes that were differently expressed between CESC patients with high or low AGPAT3 expression levels after FDR < 0.05. Of these genes, 253 were upregulated and 33 were downregulated in the high versus the low AGPAT3 expression groups (Fig. 5A, B). Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis showed that the upregulated genes were linked to tuberculosis and the phagosome signaling pathways (Fig. 5C). Meanwhile, the analysis of Gene Ontology (GO) terms showed that these upregulated genes were notably concentrated in pathways related to the maintenance of stem cell populations and the regulation of cell numbers (Fig. 5D). Conversely, KEGG analysis showed that genes that were downregulated in the high versus the low AGPAT3 expression groups were implicated in the IL-17 signaling pathway (Fig. 5E). GO term analysis also indicated that these downregulated genes were enriched in pathways related to skin and epidermis development (Fig. 5F). Given the well-documented involvement of these pathways in immunity [31–33], these findings suggest that AGPAT3 may impact CESC outcomes by its potential involvement in immune regulation. Function enrichment analysis of AGPAT3-related gene sets in CESC. (A) In the volcano plot, DEGs were displayed for high AGPAT3 expression compared to low AGPAT3 expression in CESC. Orange dots indicated significantly up-regulated genes, while blue dots indicated significantly down-regulated genes in CESC with high AGPAT3 expression. (B) The heatmap displays the levels of expression of 286 genes that are differentially expressed in AGPAT3-high and AGPAT3-low groups. (C)Conducting enrichment analysis on the KEGG pathway for genes that are up-regulated. (D) Conducting enrichment analysis for Gene Ontology terms associated with genes that are up-regulated. (E) Conducting enrichment analysis on the KEGG pathway for genes that are down-regulated. (F) Performing enrichment analysis for Gene Ontology terms associated with genes that are down-regulated Association between AGPAT3 expression and tumor immunity in pan-cancer tissues We first examined the relationship between AGPAT3 expression and tumor purity-adjusted immune scores in CESC. AGPAT3 expression showed a significant negative correlation with StromalScore, ImmuneScore, and ESTIMATEScore in several tumors including PRAD, BLCA, UCEC, LIHC, and BRCA after adjusting for tumor purity; meanwhile, contrasting results were observed in LAML, UVM, and HNSC (Supplementary Fig. 6A). We next used a Spearman analysis to investigate the correlation between AGPAT3 levels and the presence of 24 immune cell types in tumor tissues. AGPAT3 expression in different types of tumors was positively associated with the numbers of resting CD4+ memory T cells and M2 macrophages (Supplementary Fig. 6B), and negatively correlated with those of plasma cells, memory B cells, and gamma delta T cells (Supplementary Fig. 6B). We also investigated the relationship between the levels of AGPAT3 and the expression of immunomodulatory genes, including those encoding immune checkpoint proteins, immune stimulators, immune inhibitors, chemokines, and chemokine receptors [34], in 33 tumors. In most tumors, the expression of AGPAT3 was significantly positively correlated with the expression of genes encoding chemokines, chemokine receptors, and immune stimulators (Supplementary Fig. 7 A - C). Furthermore, our research revealed a positive association between AGPAT3 expression and that of immune checkpoint protins such as PD-1, PD-L1, CTLA4, TIGIT, LAG3, HAVCR2, and PDCD1LG2 across various cancer types (Supplementary Fig. 7D). AGPAT3 expression was also strongly positively correlated with the expression of immuinhibitors in the majority of tumors (Supplementary Fig. 7E). However, these correlations were moderate and should be interpreted as associations rather than evidence of direct regulation. Thus, AGPAT3 may be a candidate for further investigation in immunotherapy. Association between AGPAT3 expression and immune cell infiltration in CESC To investigate the potential immunomodulatory role of AGPAT3 in cervical cancer, we performed focused immune infiltration analyses restricted to CESC samples using two complementary algorithms: ssGSEA and CIBERSORTx. As shown in Fig. 6A, ssGSEA analysis revealed that AGPAT3 expression was significantly positively correlated with the infiltration levels of Th17 cells (ρ = 0.127, p < 0.05), NK cells (Cor = 0.128, p < 0.05), CD8 T cells (ρ = 0.138, p < 0.05), Tem (ρ = 0.182, p < 0.01), and T helper cells (ρ = 0.244, p < 0.001). Conversely, AGPAT3 expression was significantly negatively correlated with NK CD56 bright cells (ρ = −0.153, p < 0.01) and Th1 cells (ρ = 0.129, p < 0.05). Correlation of AGPAT3 Expression with Immune Cell Infiltration and T Cell Regulatory Molecules in Cervical Cancer. (A, B) Heatmaps displaying Spearman correlation coefficients between AGPAT3 mRNA expression and the abundance of 24 immune cell types in the TCGA-CESC cohort (n = 306). Immune infiltration levels were quantified using (A) single-sample Gene Set Enrichment Analysis (ssGSEA) and (B) the CIBERSORTx deconvolution algorithm. Red indicates a positive correlation, and blue indicates a negative correlation. *p < 0.05 (FDR-corrected) was considered statistically significant. (C–F) Spearman correlation analyses of AGPAT3 expression with key chemokines (C, CXCL9; D, CXCL10) and co-receptor markers (E, PDCD1; F, HAVCR2) Using the CIBERSORTx algorithm Fig. 6B, we found that AGPAT3 expression was significantly positively correlated with Macrophages M1 (ρ = 0.118, p < 0.05), T cells CD4 memory resting (ρ = 0.122, p < 0.05), and Monocytes (ρ = 0.139, p < 0.05). In contrast, AGPAT3 expression was negatively correlated with Dendritic cells activated (ρ = −0.144, p < 0.05) and Eosinophils (ρ = −0.121, p < 0.05). Given the enrichment of immune-related pathways in the functional analysis of AGPAT3-associated genes, we further investigated the relationship between AGPAT3 expression and key T cell regulatory molecules. As shown in Fig. 6C–F, AGPAT3 expression was positively correlated with the expression of key chemokines (CXCL9: ρ = 0.134, p = 0.019; CXCL10: ρ = 0.129, p = 0.024) and T cell regulatory markers (PDCD1: ρ = 0.153, p = 0.007; HAVCR2: ρ = 0.145, p = 0.011) in the CESC cohort. These findings suggest a potential association between AGPAT3 and immune modulation within the tumor microenvironment, warranting further mechanistic investigation. AGPAT3 is highly expressed in CESC tissues AGPAT3 expression was next analyzed in six CESC specimens and their corresponding paraneoplastic tissues using immunohistochemistry staining. We found that high AGPAT3 expression was confined to the cytoplasm of CESC cells (Fig. 7A, B). RNA, isolated from three sets of frozen CESC tumor tissue samples and their corresponding adjacent tissues, was then subjected to reverse transcription. The level of AGPAT3 RNA expression was then measured using quantitative real-time PCR (qRT-PCR), with the β-actin gene serving as an internal control. The results showed that AGPAT3 mRNA expression was elevated in tumor tissue compared to paraneoplastic tissues (Fig. 7C). The expression of AGPAT3 in CESC specimens and their paraneoplastic tissues. (A) Typical images of HE and AGPAT3 immunohistochemistry staining of cancerous and paracancerous tissues in CECS patient samples. (B) The h-scores for AGPAT3 immunohistochemistry staining were calculated in cancerous and paracancerous tissues from 6 CESC patients by using the formula h-score = 1 x (% weak cells) + 2 x (% moderate cells) + 3 x (% strong cells). (C) The RNA levels of AGPAT3 were measured in 18 sets of frozen CESC tumor tissue samples and their corresponding adjacent tissues AGPAT3 promotes the proliferation and migration of cervical cancer cells To further validate the effects of AGPAT3 on the proliferation and migration capacities of cervical squamous cell carcinoma (CESC) cells, we used CRISPR-Cas9 technology to generate AGPAT3-knockout (KO) cell lines in the cervical cancer cell lines HeLa and C-33 A. The efficiency of the AGPAT3 knockout in HeLa cells was confirmed by PCR (Fig. 8A) and western blotting (Fig. 8B). Cell Counting Kit-8 (CCK-8) assays revealed a significant reduction in the proliferation of AGPAT3-KO HeLa cells compared to that of their wild-type (WT) HeLa cells (Fig. 9A). Moreover, the colony-forming ability of HeLa cells markedly diminished following AGPAT3 deletion (Fig. 9B, C). Additionally, both Transwell (Fig. 9D, E) and wound healing (Fig. 9F, G) assays demonstrated that AGPAT3 knockout attenuated the migratory potential of HeLa cells. To confirm the broad-spectrum pro-proliferative and pro-migratory effects of AGPAT3 across cervical cancer cell lines, we validated the efficiency of AGPAT3 knockout in C-33 A cells (Supplementary Fig. 8A) and assessed alterations in their proliferation and migration capacities (Supplementary Fig. 8B - H). Consistent with the observations in HeLa cells, AGPAT3 knockout led to impaired proliferation and migration of C-33 A cells. Collectively, these findings conclusively demonstrated that AGPAT3 promotes the proliferation and migration of CESC. The knockout effectiveness of AGPAT3 in HeLa cells was confirmed through PCR and Western blotting analysis. (A) An image displaying PCR amplification products of AGPAT3 gene fragments from HeLa WT and HeLa AGPAT3-KO cells on an agarose gel electrophoresis. (B) Western blotting assay showing the expression level of AGPAT3 in HeLa WT and HeLa AGPAT3-KO cells, with Vinculin used as an internal control. Representative of three independent experiments The cell proliferation and migration ability of HeLa cells were inhibited after AGPAT3 gene was knocked out. (A) CCK-8 assay was used to detect growth curves of HeLa WT and HeLa AGPAT3-KO cells. Data are mean ± SEM, n = 3 biological replicates. Statistical analysis: two-way ANOVA with Tukey’s multiple comparisons test. (B) Colony formation assay of HeLa WT and HeLa AGPAT3-KO cells. (C) Quantitative statistical results of Colony formation units of HeLa WT and HeLa AGPAT3-KO cells, from three independent experiments. Data are mean ± SEM, n = 3, unpaired t-test p < 0.01. (D) Transwell migration assay of WT and AGPAT3-KO cells. After seeding cells in Transwell, they were allowed to migrate for 24 h. Following this, photographs were taken and the number of cells was counted; scale bar was set at 100 μm. (E) Quantitation of the transwell migration assay results. Data are mean ± SEM, n = 3, unpaired t-test p < 0.001. (F) Scratch assay of WT and AGPAT3-KO cells. Microscopy images were captured at 0, 24, and 48 h with a scale bar of 100 μm. (G) Quantitation of the scratch assay results. A t-test was used for statistical analysis. A t-test was used for statistical analysis. Not significant, P greater than 0.05; *, P less than 0.05; **, P less than 0.01; and ***, P less than 0.001. Data are mean ± SEM, n = 3, two-way ANOVA p < 0.001. All statistical tests used FDR correction where applicable Discussion Cervical cancer ranks fourth in terms of morbidity and mortality among all cancers affecting women worldwide. Patients with advanced disease, recurrence, or metastasis have an especially poor prognosis [35]. Thus, there is an urgent need for novel biomarkers and promising therapeutic targets in cervical cancer. The present study highlights the AGPAT family of enzymes as noteworthy biomarkers and potential therapeutic targets in cervical cancer, given their significant role in the incidence and prognosis of this disease. We initially used TCGA data to investigate AGPAT family expression, but recognized the limitation of only three normal cervical samples in TCGA. To address this, we utilized the GEPIA database to integrate GTEx normal samples (n = 10) as controls and validated our findings in two independent GEO datasets (GSE6791 and GSE63514). The GEPIA analysis revealed that only AGPAT4 was significantly downregulated in CESC, while AGPAT1-3 and AGPAT5 showed no significant differences. However, validation in GEO datasets demonstrated consistent downregulation of AGPAT1, AGPAT2, and AGPAT3 in tumor tissues, and consistent upregulation of AGPAT5. AGPAT4 showed no significant change in either GEO dataset. These findings highlight the importance of using multiple independent cohorts for validation and suggest that AGPAT family members exhibit distinct expression patterns that may be context-dependent. The discrepancy between TCGA/GEPIA and GEO results may be attributed to differences in platform technology, sample size, and patient population characteristics. Despite these variations, the consistent downregulation of AGPAT1-3 in both GEO datasets supports their potential relevance in cervical cancer pathogenesis. Next, we performed an analysis of GGI and PPI networks and found that AGPAT3 formed associations with AGPAT1 and AGPAT5. By conducting a survival analysis, we identified AGPAT3 as the most promising potential prognostic biomarker in CESC. Among the AGPAT family members, although differential expression patterns varied across datasets—with AGPAT1, AGPAT2, and AGPAT3 showing downregulation in GEO validation datasets while AGPAT4 exhibited no significant difference—AGPAT3 emerged as the most promising candidate due to its significant association with overall survival, disease-specific survival, and progression-free survival in CESC patients, as well as its consistent functional validation in cellular assays. AGPAT4 also showed prognostic significance but lacked consistent differential expression across datasets. To explore the biological function of AGPAT3 in CESC, we performed focused immune infiltration analyses restricted to CESC samples only, rather than extrapolating from pan-cancer data. Using both ssGSEA and CIBERSORTx algorithms, we found that AGPAT3 expression was significantly correlated with multiple immune cell types. Notably, AGPAT3 showed positive correlations with Th17 cells, NK cells, CD8 + T cells, and macrophages M1, suggesting a potential role in modulating anti-tumor immunity. The positive correlation with T cells CD4 memory resting and negative correlation with dendritic cells activated further support the immunomodulatory role of AGPAT3. Notably, AGPAT3 showed positive correlations with CD8⁺ T cells, T helper cells, and Th17 cells in ssGSEA analysis, as well as with resting CD4⁺ memory T cells in CIBERSORTx analysis, suggesting a potential role in modulating T cell-mediated anti-tumor immunity. The positive correlation with key chemokines CXCL9 and CXCL10, which are critical for T cell recruitment into the tumor microenvironment, further supports this notion. Additionally, AGPAT3 expression was positively correlated with T cell regulatory markers such as PDCD1 and HAVCR2, which are involved in T cell activation and exhaustion dynamics. These findings collectively suggest that AGPAT3 may influence T cell recruitment and functional states within the cervical cancer microenvironment. It is important to note that while our correlation analyses revealed significant associations between AGPAT3 expression and various T cell populations and related markers, these findings are correlative in nature and do not establish direct causal relationships. The observed positive correlations with multiple T cell subsets—including CD8⁺ T cells, T helper cells, Th17 cells, and resting CD4⁺ memory T cells—indicate a broad association with T cell-mediated immunity rather than a specific link to a single T cell lineage. Future studies employing co-culture systems, T cell functional assays, and in vivo models are warranted to elucidate the mechanistic basis of these associations. Immunohistochemical and qRT-PCR analyses confirmed that AGPAT3 expression was higher in tumor tissues than in paraneoplastic ones. Furthermore, we constructed the HeLa AGPAT3-KO cell line to show that AGPAT3 depletion significantly reduced the proliferation and migration of HeLa cells. Collectively, these findings suggest that AGPAT3 could serve as a valuable biomarker and a promising therapeutic target for cervical cancer. Many patients with cervical cancer, particularly those diagnosed at advanced stages of the disease, face a grim prognosis. Consequently, extensive research is currently underway to identify effective biomarkers and therapeutic targets that would facilitate early detection and targeted treatment of cervical cancer. Yan et al. discovered that genes encoding members of the tumor necrosis factor (TNF) family could serve as prognostic biomarkers for cervical cancer; these genes are primarily associated with the TGF-β pathway and may influence the response to immunotherapy [36]. However, this study lacks cellular experiments to substantiate these hypotheses and verify the underlying mechanisms. In a separate study, Zhang et al. demonstrated that METTL11A, a member of the methyltransferase-like gene family, enhances the motility of cervical cancer cells through an ELK3-related mechanism [37]. Nevertheless, this study does not analyze the expression levels of METTL11A in relation to the survival rates of cervical cancer patients, nor does it conduct a comprehensive investigation into the specific mechanisms by which METTL11A influences the migration of cervical cancer cells via ELK3.AGPATs are essential in converting LPA to PA during TAG biosynthesis. AGPATs play a crucial role in the survival and growth of cancer cells by facilitating TAG biosynthesis and promoting lipid storage within LDs [38]. Emerging evidence underscores that lipid metabolic reprogramming constitutes a critical hallmark of cervical cancer and is intricately linked to the oncogenic activities of Human Papillomavirus (HPV). Specifically, HPV E6 and E7 oncoproteins have been demonstrated to drive the activation of key lipogenic pathways—such as those involving fatty acid synthase (FASN) and stearoyl-CoA desaturase (SCD1)—thereby supporting essential biosynthetic processes like membrane biogenesis and redox homeostasis [39]. Notably, phosphatidic acid (PA), the enzymatic product of AGPAT family members including AGPAT3, functions not only as a central metabolic intermediate in phospholipid synthesis but also as a bioactive signaling molecule capable of activating the mTOR pathway [40]. The mTOR cascade serves as a master regulator of cell growth and proliferation and is frequently hyperactivated in HPV-associated cervical carcinogenesis, largely through direct modulation by viral E6/E7 proteins via the PI3K/AKT/mTOR axis [41]. However, the specific functional role of this methyltransferase-like gene family in cervical cancer remains unclear. This study explored the roles of AGPAT family members in CESC, examining their expression level, mutational status, diagnostic/prognosis significance, relationship with immune cell infiltration, and immune checkpoint involvement. The AGPAT family expression was assessed in pan-cancer and normal tissues using The TIMER database. We found that AGPAT1–5 expression was elevated in 15, 10, 18, 16, and 15 different types of cancers, respectively versus normal tissues. We also examined the genetic alterations affecting AGPAT family members in CESC. The results revealed that AGPATs were not prone to mutation (~ 7 mutations were detected in 306 samples), and the few genetic alterations detected did not impact the survival of CESC patients. The findings indicated that the AGPAT gene family is highly conserved, and the dysregulation of AGPAT members observed in CESC is not caused by genetic mutations. Various studies have recognized AGPATs as potential biomarkers for the diagnosis, prognosis, and progression of cancer. For instance, increased AGPAT1 expression has been associated with a greater likelihood of recurrence and reduced survival in individuals with colorectal cancer [20, 28, 42]. AGPAT1 was also shown to function as a new tumor suppressor and prognostic indicator for ovarian cancer [43, 44]. These findings prompted us to evaluate the diagnostic and prognostic potential of AGPAT family members in CESC. Our ROC analysis showed that AGPAT1/2/3/4 exhibited excellent levels of accuracy (AUC > 0.6) in differentiating between CESC patients and healthy individuals, suggesting their utility as diagnostic indicators. Furthermore, the results of the survival analysis indicated that elevated levels of AGPAT3/4 were associated with a worse outcome in patients with CESC, indicating their potential as prognostic markers. Due to the possible cancer-causing function of AGPAT3 in CESC, its expression was further examined in correlation with clinical factors of CESC patients. Acknowledgement is made of the limitation that the control group included only three normal cervical tissue samples, which may affect the accuracy of the results, highlighting the need for further validation with a larger sample size. We also note that while the current study identified AGPAT3 as a promising biomarker, the expression patterns of other AGPAT family members across different datasets showed heterogeneity, suggesting that their functional relevance may be context-dependent. To explore the biological function of AGPAT3 in CESC, differentially expressed genes (DEGs) were identified in samples grouped according to high or low AGPAT3 expression levels, before being subjected to functional enrichment analysis. Upregulated genes associated with high AGPAT3 expression were connected to immune cell maturation and activation, suggesting that AGPAT3 may modulate the immune response in CESC. We subsequently, evaluated how AGPAT3 expression influenced the infiltration of immune cells into CESC tumors. The high AGPAT3 expression group exhibited significantly increased proportions of CD8⁺ T cells and T helper cells, while correlation analysis demonstrated a significant association between AGPAT3 expression and the abundance of these T cell populations. These findings align with the broader T cell-related associations observed in our immune infiltration analyses. In addition, we used tissue samples derived from CESC patients to examine previous theories, confirming the elevated AGPAT3 mRNA and protein expression in tumor tissue versus paraneoplastic tissue. Additionally, we showed that AGPAT3 knockout in HeLa and C-33 A cells inhibited cell proliferation and migration in vitro. These results indicate that increased AGPAT3 levels could promote the onset, progression, and spread of cervical cancer cells, further supporting the potential of AGPAT3 to act as valuable biomarker and therapeutic target in CESC. Our study has several limitations that should be acknowledged. First, the expression patterns of AGPAT family members varied across different datasets; notably, AGPAT1, AGPAT2, and AGPAT3 were consistently downregulated in GEO validation datasets, while AGPAT4 showed no significant difference and AGPAT5 was upregulated. This heterogeneity highlights the importance of validation in multiple independent cohorts and suggests that the functional relevance of individual AGPAT members may be context-dependent. Although AGPAT3 was selected as the primary candidate based on its significant prognostic value and functional validation in vitro, future studies should consider comprehensive in vivo validation of multiple AGPAT family members to delineate their respective roles in cervical cancer. Second, the immune analyses are correlative and require mechanistic validation. Specifically, while we observed significant positive correlations between AGPAT3 expression and multiple T cell populations (including CD8⁺ T cells, T helper cells, and resting CD4⁺ memory T cells), as well as with T cell regulatory markers such as PDCD1 and HAVCR2, the correlational nature of these findings precludes causal inference. Further experiments, including T cell co-culture assays, cytokine profiling, and in vivo models, are needed to establish whether AGPAT3 directly influences T cell recruitment, activation, or exhaustion within the tumor microenvironment. Third, while we validated AGPAT3’s functional role in two cervical cancer cell lines, additional cell lines representing different HPV genotypes would strengthen our conclusions. Fourth, animal studies would be needed to confirm AGPAT3’s role in vivo. Fifth, we acknowledge the limitation that the control group in the TCGA database included only three normal cervical tissue samples, which may affect the accuracy of the initial differential expression analysis; however, we mitigated this by validating in two independent GEO datasets. Despite these limitations, our multi-omics approach combining bioinformatics, clinical samples, and experimental validation provides robust evidence for AGPAT3’s importance in cervical cancer. In the present study, we investigated the expression of the AGPAT family members in CESC and examined its correlation with the prognosis and immune function of patients with this cancer. Our findings suggest that AGPAT3 is significantly associated with the onset and progression of CESC, potentially through its influence on T cell-mediated immunity within the tumor microenvironment. The observed correlations with CD8⁺ T cells, T helper cells, and key T cell-related chemokines and regulatory markers highlight the potential immunomodulatory role of AGPAT3. In summary, we have identified AGPAT3 as a promising biomarker and potential therapeutic target for cervical cancer. However, our study has certain limitations. While functional enrichment and immune infiltration analyses imply that AGPAT3 may promote cervical cancer via its involvement in immune regulation, this hypothesis requires further experimental validation. Furthermore, pan-cancer analysis revealed varied AGPAT family member expression across different cancers and its impact on immunity, prompting the need for further exploration of the role of AGPATs in other cancer types. Supplementary Information Supplementary Material 1. Acknowledgements We gratefully acknowledge the assistance of X.D. Zhang in preparing the gene knockout cells. We also thank Jingsheng Ao and Feng Shi for helpful discussion on experimental design and data analysis. Author contributions Conceptualization: Yuexiu Liang, Mingyou Dong.Data curation: Nannan Gui.Formal analysis: Yun Cheng.Funding acquisition: Weijuan LuInvestigation: Hanyi Pan, Weijuan Lu, Xiaoyu Zhou, Mingyang JinSupervision: Yuexiu Liang.Validation: Nannan Gui, Guiqiong Pan.Writing – original draft: Hanyi Pan, Weijuan Lu, Xiaoyu Zhou, Changyong YangWriting – review & editing: Yuexiu Liang, Mingyou Dong, Nannan Gui. Funding This work was supported by the Guangxi Zhuang Autonomous Region Health Commission (No. Z-L20240823, Z-L20240829 and Z-L20250732) and Innovation and Entrepreneurship Training Program for College Students of Youjiang Medical University for Nationalities in 2025 (No. 1120220314). Data availability All data supporting the findings of this study are publicly available. The transcriptomic and clinical data for cervical cancer were obtained from The Cancer Genome Atlas (TCGA-CESC) project (https://portal.gdc.cancer.gov/projects/TCGA-CESC). Additional analyses utilized data from the following databases: GTEx (https://gtexportal.org/), GSCA (http://bioinfo.life.hust.edu.cn/GSCA/), TIMER (https://cistrome.shinyapps.io/timer/), cBioPortal (https://www.cbioportal.org/), GeneMANIA (https://genemania.org/), STRING (https://string-db.org/), Xiantao Academic (https://www.xiantao.love/), UALCAN (http://ualcan.path.uab.edu/), and MethSurv (https://biit.cs.ut.ee/methsurv/). The GEO dataset GSE6791 is available at https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi? acc=GSE6791. No new datasets were generated in this study. Declarations Ethics approval and consent to participate The study protocol was reviewed and approved by the Ethics Committee of the Affiliated Hospital of Youjiang Medical University for Nationalities (Approval No. YYFY-LL-2023HY2-06; dated November 8, 2023), in accordance with the ethical principles of the Declaration of Helsinki. All participants provided written informed consent prior to participation. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Nannan Gui, Hanyi Pan and Weijuan Lu have contributed equally to this work. Contributor Information Mingyou Dong, Email: mydong@ymun.edu.cn. Yuexiu Liang, Email: 1262@ymun.edu.cn. References Associated Data Supplementary Materials Supplementary Material 1. Data Availability Statement All data supporting the findings of this study are publicly available. The transcriptomic and clinical data for cervical cancer were obtained from The Cancer Genome Atlas (TCGA-CESC) project (https://portal.gdc.cancer.gov/projects/TCGA-CESC). Additional analyses utilized data from the following databases: GTEx (https://gtexportal.org/), GSCA (http://bioinfo.life.hust.edu.cn/GSCA/), TIMER (https://cistrome.shinyapps.io/timer/), cBioPortal (https://www.cbioportal.org/), GeneMANIA (https://genemania.org/), STRING (https://string-db.org/), Xiantao Academic (https://www.xiantao.love/), UALCAN (http://ualcan.path.uab.edu/), and MethSurv (https://biit.cs.ut.ee/methsurv/). The GEO dataset GSE6791 is available at https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi? acc=GSE6791. No new datasets were generated in this study.