Bioinformatics analysis identifies ULK3 as a novel prognostic and immune-related biomarker in esophageal cancer

Introduction

Esophageal cancer (EC) ranks as the seventh most prevalent cancer worldwide. In 2022, approximately 511,054 new EC cases and 445,391 deaths were reported globally, making it the sixth leading cause of cancer-related mortality. China bears the heaviest burden of EC, accounting for 43.8% of all cases and 42.1% of all deaths related to this disease (1). This underscores the urgent need to identify effective therapeutic targets for EC. The tumor, node, metastasis (TNM) staging system and pathological classification serve as conventional prognostic markers for EC, which guide clinical decision-making and treatment selection.

As research on the mechanisms of esophageal carcinogenesis advances, numerous tumor biomarkers have demonstrated clinical potential or entered routine practice. Prognostic markers for EC encompass diverse genetic and epigenetic regulatory processes, including TP53 mutations (2), impaired DNA repair (3), histone lactylation (4), MT-1A promoter methylation (5), and post-transcriptional modifications involving RNA-binding proteins (6) or m5C RNA methylation (7). Non-coding RNAs like miR-21 (8), protein markers such as programmed death ligand 1 (PD-L1) (9), and post-translational modifications, including phosphorylation (10), further contribute to prognostic assessment. These biomarkers appear not only in tumor tissue but also in peripheral blood components like circulating tumor DNA. Circulating tumor cells, serum proteins [interleukin 8 (IL-8)/neutrophil-to-lymphocyte ratio], and carbohydrate antigens (CA199, CA125) similarly correlate with disease outcomes. The tumor microenvironment, particularly CD8⁺ T cell infiltration and immune checkpoint expression (1), significantly influences prognosis, while emerging evidence implicates metabolic byproducts and gut microbiota composition in disease progression (11).

Current molecular markers guiding EC diagnosis and treatment fall into three categories: targeted therapy targets such as human epidermal growth factor receptor 2 (HER2) overexpression, NTRK mutations, BRAF V600E mutations, RET mutations, CLDN18.2, epidermal growth factor receptor (EGFR) and VEGFR2 inhibition; immunotherapy markers including PD-L1 expression, microsatellite instability-high (MSI-H)/deficient mismatch repair (dMMR) status, and tumor mutational burden ; and carbohydrate-based markers for recurrence monitoring and treatment response assessment (12). Despite recent diagnostic and therapeutic advances, the 5-year survival rate remains 10–30% in most countries (13), underscoring the need for more reliable therapeutic targets in EC management.

Autophagy, a lysosomal-dependent process, serves to degrade non-essential cellular components. In mammalian cells, this mechanism manifests in various forms, including macroautophagy​, microautophagy, and chaperone-mediated autophagy. Under physiological conditions, autophagy contributes significantly to maintaining cellular homeostasis. Its disruption can trigger the transition of cells from a healthy to a diseased state, often resulting in tumor development (14). In the context of tumorigenesis, autophagy can exhibit both inhibitory and promotional effects, depending on the disease stage and mutational background (15). Tumors such as liver, melanoma, gastrointestinal, breast, and pancreatic cancers frequently arise alongside mutations and inactivation of autophagy-related genes (16-20). Interestingly, autophagy has also been observed to positively regulate tumor growth, ensuring nutrient supply in poorly vascularized regions (21) and maintaining mitochondrial function through mitophagy (22), thereby facilitating tumor progression. Moreover, autophagy plays a crucial role in tumor immunology (21). Current studies on autophagy-related genes as prognostic markers for EC remain limited.

ULK3, a serine/threonine protein kinase involved in sonic hedgehog (SHH) signal transduction, is also implicated in autophagy initiation (23). Studies have revealed associations between ULK3 and tumorigenesis. For instance, overexpression of the ULK3 can promote the malignant phenotype of bladder cancer cell line T24 (24). Conversely, increased expression of ULK3 in breast tissue is linked to a reduced risk of breast cancer (25).

The role of ULK3 in EC remains unclear. Based on these findings regarding the role of ULK3 in specific cancers and autophagy processes, we hypothesized that the ULK3 influences the development of EC. To investigate this hypothesis, we analyzed public databases, including The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO), to examine ULK3 expression and mutations in EC and their impact on patient survival, tumor immunity, and single-cell behavior. We further employed Mendelian randomization (MR) analysis to assess its potential as an EC biomarker. We present this article in accordance with the REMARK reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-989/rc).

Methods

Data download

We obtained TCGA EC datasets and six GEO datasets, downloaded TCGA EC expression profiles and clinical data from the XENA database (https://xenabrowser.net/datapages/). The TCGA EC cohort included 185 samples [96 esophageal squamous cell carcinoma (ESCC), 88 esophageal adenocarcinoma (EAC), 1 mucinous cystic neoplasm] comprising 114 Europeans, 46 Asians, 5 Africans, and 20 cases of unknown ethnicity, with clinical variables covering race, tumor location, pathological classification, smoking status, alcohol consumption, gender, age, stage, and overall survival (OS). The GEO datasets (https://www.ncbi.nlm.nih.gov/geo/) included GSE13898 (64 EAC, 15 Barrett’s esophagus, 28 normal samples from the USA), GSE53625 (179 ESCC and 179 normal samples from China), GSE77861 (7 ESCC and 7 normal samples from African Americans), GSE213565 (10 ESCC and 10 normal samples from China), and the single-cell dataset GSE196756 (3 ESCC and 3 normal samples from China). The GSE53625 dataset provided patient OS information. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Investigating ULK3 expression and mutation profiles

We analyzed ULK3 RNA expression across various cancers using the Gene Expression Profiling Interactive Analysis (GEPIA2) (http://gepia2.cancer-pku.cn/) database, focusing specifically on EC and its different stages. Protein-level expression of ULK3 in pan-cancer was examined using the University of ALabama at Birmingham CANcer data analysis Portal (UALCAN) database (https://ualcan.path.uab.edu/). Single-nucleotide variations (SNVs) of the ULK3 gene in common tumors were investigated using the gene set cancer analysis (GSCA) database, with a separate analysis for copy number variations (CNVs) in EC. We downloaded ESCC and EAC datasets such as GSE13898, GSE53625, GSE77861, GSE92396, and GSE213565 from the GEO database and conducted a two-independent samples t-test to compare ULK3 expression differences between esophageal tumor and normal groups. One-way ANOVA with Tukey’s post hoc test was used for comparisons within the GSE13898 dataset.

Exploring the relationship between ULK3 expression and survival

EC patients were stratified into high and low ULK3 expression groups based on median RNA expression. We compared OS and disease-free interval (DFI) between these groups using the GEPIA2 database. OS was further analyzed in ESCC, EAC, and 19 other tumors using the KM-PLOTTER database (https://kmplot.com/analysis/), including patients treated with immunotherapy. The impact of different CNV mutations on EC patient survival was examined in the GSCA database (http://bioinfo.life.hust.edu.cn/GSCA). The GSE53625 dataset was used to validate the effect of ULK3 expression on EC patient OS. We downloaded TCGA EC dataset from the XENA database and conducted univariate Cox regression tests in SPSS to assess the influence of ULK3 expression on EC patient survival, considering factors such as race, age, gender, tumor pathological classification, smoking status, alcohol consumption, TNM stage, tumor location, biomarkers (CD274, CEACAM5, ERBB2, EGFR, CLDN18, VEGFA), and ULK3 expression levels. Multivariate Cox analysis further examined the independence between these prognostic factors. Meanwhile, we examined the relationship between ULK3 and the aforementioned clinical factors and Gene expression of relevant biomarkers.

Investigating the link between ULK3 expression and immune and therapeutic responses

We analyzed the relationship between ULK3 expression and immune cell infiltration in EC using the TIMER database (https://cistrome.shinyapps.io/timer/). The association between ULK3 and T cell Exhaustion markers was also examined. Utilizing TCGA EC data, we compared the infiltration of 22 immune cells between ULK3-high and ULK3-low expression groups using the Cell-Type Identification By Estimating Relative Subsets Of RNA Transcripts (CIBERSORT) 0.1.0 package in R. The Tumor Immune Dysfunction and Exclusion (TIDE) method was employed to compare immunotherapy responses between the groups. The GSCA database was used to assess the influence of ULK3 expression on the sensitivity of tumor cell lines to anticancer drugs.

Pathway enrichment analysis in ULK3 high and low expression groups

In R, we used the limma package with log₂fold change (FC) >1 and false discovery rate (FDR) <0.05 as thresholds, correcting the FDR with the Benjamini-Hochberg method. We compared RNA sequencing data between ULK3-high and ULK3-low expression groups in TCGA EC samples to identify differentially expressed genes. Genes upregulated in the ULK3 high-expression group were selected for Gene Ontology (GO)-Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis. The top enriched pathways from GO and KEGG analyses were visualized. Additionally, gene set enrichment analysis (GSEA) was conducted on all differentially expressed genes, focusing on autophagy-related pathways and the top ten enriched pathways.

Single-cell analysis of ULK3 high and low expression groups

The scRNA-seq sequencing dataset GSE196756 for EC was obtained from the GEO database (26). From this dataset, three tumor samples were chosen for scRNA-seq analysis using Seurat v5.2.1. Cells with fewer than 1,000 detected genes or with mitochondrial gene content exceeding 20% were excluded during quality control. The Harmony method was employed to correct for batch effects, and nonlinear dimensionality reduction was achieved using uniform manifold approximation and projection (UMAP). Marker genes were then used for manual annotation of cell types, including epithelial cells (KRT5, KRT13, TP63), endothelial cells (ENG, VWF), fibroblasts (COL1A1, COL14A1, DCN), B cells (MS4A1, CD79A), T cells (CD2, CD3D, CD3E), myeloid cells (CD14, CD68), and mast cells (CPA3, TPSAB1, TPSB2). Subsets of T cells and epithelial cells were selected and further stratified into high- and low-expression groups based on ULK3 expression. Utilizing gene sets reported in the literature (27), the AddModuleScore method was was applied to compare T cell activation, tissue residence, cytokine, and interferon scores between these groups. Additionally, gene sets related to tumor proliferation, invasion, metastasis, and epithelial-mesenchymal transition were retrieved from the CNACERSEA database (http://biocc.hrbmu.edu.cn/CancerSEA), and epithelial cell phenotypic scores were compared between the ULK3 high- and low-expression groups using the same method.

MR analysis

We obtained ULK3 eQTL data (ID: eqtl-a-ENSG00000140474; sample size: 31,684; European population) and EC Genome-Wide Association Study (GWAS) data (ID: ieu-b-4960; 740 cases vs. 372,016 controls) from the IEU Open GWAS database (https://gwas.mrcieu.ac.uk/​). Using a significance threshold of P<5×10⁻⁶, we selected SNPs strongly associated with ULK3 and removed linkage disequilibrium (r²<0.01 within 100 kb windows) based on the 1000 Genomes European reference panel. We determined instrument strength by calculating F-statistics (F=R²×[N−2]/[1−R²]), confirming that the average F-statistic exceeded 10. After extracting exposure-associated Single Nucleotide Polymorphisms (SNPs) from outcome GWAS data, we removed potential confounders (P<5×10⁻⁸ for direct outcome associations). Primary analysis employed inverse variance weighting (IVW) (28), supplemented by MR-Egger regression, weighted median, simple mode, and weighted mode methods for robustness assessment (29). We identified outliers using MR-PRESSO and evaluated heterogeneity (Q_pval <0.05) and horizontal pleiotropy (intercept pval <0.05) using the mr_heterogeneity and mr_pleiotropy_test functions. Leave-one-out sensitivity analyses determined whether results were driven by individual SNPs (30). We validated findings using FinnGen database GWAS data (ID: finngen_R11_C3_OESOPHAGUS_EXALLC; 763 cases vs. 345,118 controls), repeating all analytical procedures.

Statistical analysis

For the statistical analysis, we adopted a two-way test criterion, considering a P value less than 0.05 as statistically significant. Cox regression analysis was performed using SPSS 25.0. The comparison of ULK3 expression between the tumor and normal groups in the GEO datasets was visualized with GraphPad Prism 8.0.1. All other analyses were conducted using R 4.4.1, and data visualization was achieved through ggplot2 3.5.1.

Results

Expression patterns and mutational landscape of ULK3

The GEPIA2 analysis revealed statistically significant differential ULK3 RNA expression in eight tumor types, including EC, vs. corresponding normal tissues. Specifically, ULK3 expression was elevated in diffuse large B-cell lymphoma and thyroid cancer tumors relative to normal tissues, whereas in other tumor types, expression was elevated in normal tissues (Figure 1A). The UALCAN database demonstrated ULK3 protein expression disparities in ten tumor types, among these, seven showed statistically significant expression differences between tumor and normal groups. Notably, liver and endometrial cancers exhibited higher ULK3 protein levels in tumor tissues, contrasting with other tumor types, where normal tissues had elevated expression (Figure 1B). Analysis specific to ULK3 expression in EC indicated higher gene expression levels in normal tissues compared to tumor tissues (Figure 1C). Although ULK3 expression did not differ significantly across different stages of EC, stage I EC appeared to have the highest expression level (Figure 1D). In terms of mutations, EC predominantly exhibited heterozygous deletions in the CNV of ULK3 (Figure 1E). Pan-cancer analysis identified missense mutations as the predominant ULK3 SNV type (Figure 1F). Furthermore, integrated analysis of five GEO datasets (GSE13898, GSE53625, GSE77861, GSE92396, and GSE213565) demonstrated significantly elevated ULK3 RNA expression in normal tissues vs. EAC, ESCC, and Barrett’s esophagus tumor tissues (Figure 1G-1K).

Figure 1 The expression and mutation profile of ULK3 in pan-cancer and esophageal cancer. (A) The RNA levels of ULK3 between tumor and normal tissue samples across pan-cancers. (B) The differences in ULK3 protein expression between tumor and normal tissues in pan-cancers, including BCA, CC, OC, ccRCC, UCEC, lung cancer, PAAD, HNSC, GBM, and liver cancer. (C) The ULK3 expression in tumor and normal tissue samples of EC. (D) The ULK3 expression in each stage of EC. (E) The CNV mutation status of ULK3 in EC. (F) The SNV mutation status of ULK3 across pan-cancer datasets. (G-K) The ULK3 expression patterns in five different GEO datasets. *, P<0.05; **, P<0.01; ****, ​P<0.0001. BCA, breast cancer; CC, colon cancer; ccRCC, clear cell renal cell carcinoma; CNV, copy number variation; EC, esophageal cancer; GBM, glioblastoma; HNSC, head and neck squamous cell carcinoma; INS, insertion; OC, ovarian cancer; PAAD, pancreatic cancer; SNV, single nucleotide variation; TPM, transcripts per million; UCEC, endometrial cancer.

The relationship between ULK3 expression and survival

Survival analyses revealed significantly better OS in patients with high vs. low ULK3 expression in EC (Figure 2A). In the GSE53625 cohort, although high-expression patients had better prognosis, the difference remained non-significant (Figure 2B). Patients in the high-expression group also exhibited a longer DFI, though this difference was non-significant (Figure 2C). High ULK3 expression was still linked to better outcomes in both esophageal adenocarcinoma (Figure 2D) and squamous cell carcinoma (Figure 2E). Among tumor patients undergoing immunotherapy, those with high ULK3 expression had better survival rates (Figure 2F). CNV analysis revealed that amplified-variation patients had the longest DFI, suggesting superior disease control, while deletion-variation patients had the shortest DFI; wild-type patients exhibited intermediate outcomes (Figure 2G). Univariate and multivariate Cox regression of TCGA EC data identified both TNM stage and ULK3 expression as significant factors: high stage inhibited long-term survival, whereas high ULK3 expression promoted it (Figure 2H,2I). The relationship between ULK3 expression and survival was also investigated in 19 other tumor types besides EC, EAC, and ESCC, significant differences in survival were observed between the high and low ULK3 expression groups in ten tumor types: bladder cancer, cervical squamous cell carcinoma, head and neck squamous cell carcinoma, renal papillary cell carcinoma, ovarian cancer, pancreatic cancer, sarcoma, gastric cancer, thymic carcinoma, and clear cell renal cell carcinoma. Excepting for clear cell renal cell carcinoma, the prognosis was more favorable in the high ULK3 expression group for the other nine tumors (Figure 3). No statistically significant correlation emerged between ULK3 expression and other clinical factors or molecular markers (Figure 4).

Figure 2 The prognostic significance of ULK3 expression and mutation status in esophageal cancer. (A) The OS between groups with high and low ULK3 expression in EC. (B) The OS variations in the GSE53625 dataset, considering ULK3 expression. (C) The DFI between groups with high and low ULK3 expression in EC. (D,E) The OS differences in EAC and ESCC based on ULK3 expression levels. (F) The OS disparities among tumor patients undergoing immunotherapy stratified by ULK3 expression. (G) The differences in DFI among EC patients with varying CNV mutation statuses. (H,I) The results of univariate and multivariate Cox regression analyses for EC. Low: ULK3 low expression group, High: ULK3 high expression group, with the division based on median expression levels. Amp, amplification; CNV, copy number variation; Del, deletion; DFI, disease-free interval; EAC, esophageal adenocarcinoma; EC, esophageal cancer; ESCC, esophageal squamous cell carcinoma; OS, overall survival; Wt, wild type.

Figure 3 The relationship between the expression of ULK3 and survival in 19 types of tumors except esophageal cancer. Low: ULK3 low expression group; High: ULK3 high expression group, with the division based on median expression levels. HR, hazard ratio.

Figure 4 Correlation analysis between ULK3 and common clinical features as well as selected gene expression profiles. (A) Differences in ULK3 gene expression levels across racial groups (White, Asian, Black, unknown race). (B) Differences in ULK3 gene expression levels across anatomical sites (upper, middle, lower, unknown site). (C) Correlation heatmap of pairwise associations between clinical variables and gene expression markers. *, P<0.05; **, P<0.01; ***, <0.001. HR, hazard ratio.

Exploring the connection between ULK3 expression and responses to immunity and chemotherapy

The CIBERSORT revealed elevated infiltration levels of naive B cells and plasma cells in the group with high ULK3 expression, whereas the levels of monocytes, M1 macrophages, and M2 macrophages were significantly reduced (Figure 5A). Analyses from the Cancer Therapeutics Response Portal (CTRP) and Genomics of Drug Sensitivity in Cancer (GDSC) databases suggested that higher ULK3 expression is linked to increased sensitivity to multiple chemotherapeutic agents (Figure 5B,5C).

Figure 5 The immune infiltration and therapeutic response in the ULK3 high and low expression groups. (A) The difference in immune infiltration of CIBERSORT in the high and low expression groups of ULK3 in TCGA EC data. (B,C) The correlations of ULK3 expression to the drug sensitivity of tumor cell lines in the CTRP and GDSC databases. A positive correlation indicates that higher gene expression may lead to drug resistance, and a negative correlation indicates that higher gene expression may cause drug sensitivity. (D) Differences in immune infiltration between the ULK3 high and low expression groups in the TIMER database. (E-H) Analysis results of TIDE in the high and low expression groups of ULK3. (I-M) The correlations between ULK3 and five markers of T cell exhaustion. *, P<0.05; **, P<0.01. Low: ULK3 low expression group; High: ULK3 high expression group, with the division based on median expression levels. CIBERSORT, Cell-Type Identification By Estimating Relative Subsets Of RNA Transcripts; CTRP, Cancer Therapeutics Response Portal; EC, esophageal cancer; GDSC, Genomics of Drug Sensitivity in Cancer; TCGA, The Cancer Genome Atlas; TIDE, Tumor Immune Dysfunction and Exclusion; TIMER, Tumor Immune Estimation Resource; TPM, transcripts per million.

Tumor Immune Estimation Resource (TIMER) database analysis revealed a statistically significant negative correlation between ULK3 expression and macrophage infiltration in EC. However, no significant correlations were observed with other immune cell infiltrates (Figure 5D). TIDE analysis indicated elevated TIDE, EXCLUSION, IFNG, and MERCK18 scores in the low- vs. high-expression group (Figure 5E-5H). Furthermore, ULK3 expression negatively correlated with the gene expressions of five key T cell exhaustion markers (Figure 5I-5M).

Pathway enrichment in the ULK3 high and low expression groups

Enrichment analysis of pathways in the high and low ULK3 expression groups revealed notable findings. The GO enrichment analysis highlighted differential gene expression enriched in processes like the development, differentiation, connection, and peptidase activity of epidermal cells (Figure 6A). The KEGG bubble plot further emphasized metabolic pathways involving linoleic acid, retinol, and arachidonic acid, as well as signal transduction via the serotonin synapse, IL-17 signaling, and estrogen signaling. Additionally, pathways related to Staphylococcus aureus infection and various other aspects were identified (Figure 6B). GSEA demonstrated positive enrichment of the GOBP-MICROAUTOPHAGY gene pathway in the high ULK3 expression group (Figure 6C). The top ten enriched items from the GSEA were related to immune response, cell activation, adhesion, and epidermal development (Figure 6D).

Figure 6 The pathway enrichment of ULK3 high and low expression groups in TCGA EC data. (A,B) The GO and KEGG enrichment results between the ULK3 high and low expression groups. (C) GSEA results of the GOBP-MICROAUTOPHAGY pathway in the ULK3 high and low expression groups. (D) The top ten GSEA results between the ULK3 high and low expression groups. EC, esophageal cancer; GO, Gene Ontology; GSEA, gene set enrichment analysis; IL-17, interleukin 17; KEGG, Kyoto Encyclopedia of Genes and Genomes; TCGA, The Cancer Genome Atlas.

Single-cell level analysis of ULK3 high and low expression groups

First, we annotated the single cells, and the annotation results are presented in Figure 7A. Subsequently, T cell subsets (Figure 7B) and epithelial cell subsets (Figure 7C) were extracted from the data. These subsets were then stratified into high and low ULK3 expression groups based on the expression level. Single-cell analysis comparing ULK3-high and ULK3-low expression groups revealed distinct differences. In GSE196756 tumor samples, T cells in the ULK3-high group exhibited elevated scores for cytokine secretion and tissue residence (Figure 7D-7G). Conversely, high-ULK3 epithelial cells demonstrated reduced scores for proliferation, invasion, metastasis, and epithelial-mesenchymal transition (Figure 7H-7K).

Figure 7 The impact of ULK3 on the esophageal cancer phenotype at the single-cell level. (A) The Annotation results of tumor samples in the GSE196756 dataset. (B,C) T cells and epithelial cells were stratified into ULK3 high (ULK3 + H) and ULK3 low (ULK3 + L) expression groups based on quantitative ULK3 mRNA levels1. (D-G) Comparison of AddModuleScore (activation, residence, cytokines, and IFN) between ULK3 + H and ULK3 + L T cell groups. (H-K) Comparison of AddModuleScore (proliferation, invasion, metastasis, and EMT) between ULK3 + H and ULK3 + L epithelial cell groups. ****, P<0.0001; ns, not significant. EML, epithelial-mesenchymal transition; IFN, interferon; UMAP, uniform manifold approximation and projection.

MR analysis

Using ULK3 as the exposure variable and the EC data from the IEU and FinnGen databases as outcome variables, we performed two MR Analyses. The mean F-statistic for the instrumental SNPs exceeded 10. The IVW indicated a statistically significant association (P<0.05) with an odds ratio (OR) less than 1 in both analyses. For outcome ieu-b-4960, P values from weighted median, simple mode, and weighted mode methods were all <0.05 (Figure 8A). For outcome finngen_R11_C3_OESOPHAGUS_EXALLC, simple mode results had P<0.05 (Figure 8B). Importantly, all analysis results successfully passed the heterogeneity, pleiotropy, and sensitivity tests (Figure 8C-8F).

Figure 8 The MR analysis results for ULK3-EC. (A,B) The MR analysis results of ieu-b-4960 and finngen-R11-C3-OESOPHAGUS-EXALLC as outcome traits. (C,D) Heterogeneity assessment and Leave-one-out method sensitivity analysis in MR using ieu-b-4960 as the outcome. (E,F) Heterogeneity assessment and Leave-one-out method sensitivity analysis in MR using finngen-R11-C3-OESOPHAGUS-EXALLC as the outcome. SE_IV, standard error of the instrumental variable; CI, confidence interval; EC, esophageal cancer; ESCA, esophageal cancer; MR, Mendelian randomization; OR, odds ratio.

Discussion

It is widely recognized that surgical resection offers an effective and reliable treatment for early-stage EC, with a 5-year survival rate of 95% (31). However, most patients present with advanced EC, where classic radiotherapy and chemotherapy yield a 5-year survival rate of only 10–30% (32,33). Immunotherapy has emerged as a treatment option, using tumor PD-L1 expression as a biomarker for predicting anti-programmed cell death protein 1 (PD-1) treatment efficacy. Yet, the ≥1% threshold for defining PD-L1 positivity fails to precisely identify immunotherapy responders. Targeted drugs for advanced EC include anti-angiogenic drugs, EGFR inhibitors, and human epidermal growth factor receptor 2 inhibitors. Despite these options, the difficulty in precisely selecting responders and the issue of resistance to immunotherapy or targeted therapy remain significant challenges. Thus, identifying biomarkers for precision therapy or reliable therapeutic targets is urgently needed

Although ULK3’s specific role in EC remains unclear from prior studies, our research reveals several key findings. We observed that ULK3 is downregulated at both RNA and protein levels in most tumors, positively correlates with OS in EC and other tumors, and serves as an independent prognostic factor for EC patients. This suggests a potential inhibitory role of ULK3 in EC. Furthermore, high ULK3 expression enhances drug sensitivity, lowers the TIDE scores, and is associated with longer OS in immunotherapy-treated patients, supporting its potential as a stratified treatment marker for EC. Notably, patients with ULK3 copy number amplification have the longest DFI, while those with deletion have the poorest prognosis, implying a gene dose effect on its functional performance. The association between ULK3 and improved prognosis has been confirmed in head and neck squamous cell carcinoma (34), cervical cancer (35), and breast cancer (25), though conflicting findings exist for rectal (36) and bladder cancers (24). MR analysis demonstrated a significant causal link (P<0.05) between ULK3 expression and lower EC risk (OR <1). These genetic findings align with ULK3’s survival-associated expression patterns in public databases, supporting its potential inhibitory role in EC.

Given the ULK family’s role in autophagy, we investigated autophagy-related pathway enrichment in high and low ULK3 expression groups. GSEA results indicated that the high expression group was enriched in the microautophagy pathway, aligning with the ULK family’s role in initiating autophagy. This suggests a potential role of ULK3 in microautophagy, a subtype closely linked to tumorigenesis (37). Additionally, ULK3, a serine/threonine protein kinase, plays a crucial role in the SHH signaling pathway, whose dysregulation is linked to various diseases, including cancer (38). ULK3 not only enhances GLI protein’s transcriptional activity but also participates in SHH signal transduction independently of its kinase activity (39).

Tumor development is not only related to tumor cell phenotypes but also to the quantity and function of infiltrating immune cells. The SHH and autophagy pathways involved in ULK3 play key roles in both tumor development and immunity (40,41). In our study, high ULK3 expression reduced the malignant phenotypic score of epithelial cells and increased the anti-tumor phenotypic score of T cells. GO and GSEA enrichment results further indicated that ULK3 might exert an inhibitory effect on EC through dual effects on tumor cells and tumor immunity. The negative correlation between ULK3 and T-cell exhaustion markers, combined with a lower EXCLUSION score, suggests that patients with high expression may have more active T-cell function and a better response to immunotherapy.

ULK3 is generally upregulated in senescent cells and can induce autophagy and premature senescence (42), indirectly proving its tumor suppressor potential. However, autophagy’s effect on tumors can vary due to different oncogene mutations (15). Some studies report ULK3 promoting the malignant phenotype of bladder cancer cells (24), while others align with our findings, showing that ULK3 downregulation weakens the proliferation and tumorigenic potential of squamous cell carcinoma (43). These differences can be attributed to varying gene mutation backgrounds among tumor types.

In conclusion, our study provides valuable insights into the role of ULK3 in EC. However, further experimental validation at the cellular, animal, and clinical tissue levels is needed to verify the feasibility of ULK3 as a prognostic biomarker and to elucidate the specific mechanisms by which ULK3 regulates immune infiltration or alters tumor cell phenotypes. Additionally, prospective clinical cohort studies are required to confirm ULK3’s impact on patient survival and its predictive value for response to chemotherapy or immunotherapy, providing a foundation for selecting therapeutic approaches.

Conclusions

This study demonstrated that reduced ULK3 expression in EC correlates with a less favorable prognosis. The impact of ULK3 on EC occurred through its regulation of tumor cell proliferation and macrophage infiltration within the tumor. When considered alongside the findings from our MR analysis, it becomes evident that individuals with high ULK3 expression face a decreased risk of developing EC. Based on these observations, we postulate that ULK3 holds promise as a prognostic marker for EC, with its expression level offering a degree of predictability regarding patient prognosis.

Acknowledgments

We offer our heartfelt thanks to all the contributors for their priceless contributions to this study. Our appreciation also reaches out to the publicly accessible databases and resources that aided our research, along with the freely available educational materials that deepened our comprehension.

Footnote

Reporting Checklist: The authors have completed the REMARK reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-989/rc

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-989/prf

Funding: This work was supported by a grant from the National Natural Science Foundation of China (No. 82060433).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-989/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. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

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. Zheng YJ, Teng Y, He SY, et al. Epidemiological Characteristics of Esophageal Cancer Worldwide and in China, 2022. China Cancer 2025;34:165-70.
2. Jiang W, Zhang B, Xu J, et al. Current status and perspectives of esophageal cancer: a comprehensive review. Cancer Commun (Lond) 2025;45:281-331. [Crossref] [PubMed]
3. Du H, Wang X, Xie S, et al. Identification of a prognostic DNA repair gene signature in esophageal cancer. J Gastrointest Oncol 2024;15:829-40. [Crossref] [PubMed]
4. Huang T, You Q, Liu J, et al. WTAP Mediated m6A Modification Stabilizes PDIA3P1 and Promotes Tumor Progression Driven by Histone Lactylation in Esophageal Squamous Cell Carcinoma. Adv Sci (Weinh) 2025;12:e06529. [Crossref] [PubMed]
5. Zhang R, Nie Y, Chen X, et al. A multicenter prospective clinical trial reveals cell-free DNA methylation markers for early esophageal cancer. J Clin Invest 2025;135:e186816. [Crossref] [PubMed]
6. Sun S, Wang J, Zhang Y, et al. Genome-wide profiling of a prognostic RNA-binding protein signature in esophageal cancer. Transl Cancer Res 2025;14:1428-46. [Crossref] [PubMed]
7. Huang Z, Peng Y, Cai S, et al. Identification and validation of a prognostic signature of m(5)C-related genes for esophageal cancer. J Thorac Dis 2025;17:4117-35. [Crossref] [PubMed]
8. He Z, Ji Y, Yuan Y, et al. Uncovering the role of microRNAs in esophageal cancer: from pathogenesis to clinical applications. Front Pharmacol 2025;16:1532558. [Crossref] [PubMed]
9. Formica V, Morelli C, Fornaro L, et al. PD-L1 thresholds predict efficacy of immune checkpoint inhibition in first-line treatment of advanced gastroesophageal adenocarcinoma. A systematic review and meta-analysis of seven phase III randomized trials. ESMO Open 2024;9:103967. [Crossref] [PubMed]
10. Qin Z, Jiang D, Yu Z, et al. Proteogenomic characterization reveals that lipid droplet formation promotes esophageal squamous cell cancer progression. Sci Transl Med 2025;17:eadt0214. [Crossref] [PubMed]
11. Cao YQ, Cheng YM, Li TC, et al. Review of metabolomics and microbiomics in esophageal cancer: From pathogenesis to prognosis. LabMed Discovery 2025;2:100045.
12. Wang Y, Chen SY, Yin WT, et al. Interpretation of the updates of NCCN esophageal and esophagogastric junction cancers clinical practice guidelines in oncology (version 1.2025). Chinese Journal of Clinical Thoracic and Cardiovascular Surgery 2025;32:1225-32.
13. Ma W, Baran N. Expanding horizons in esophageal squamous cell carcinoma: The promise of induction chemoimmunotherapy with radiotherapy. World J Clin Oncol 2025;16:104959. [Crossref] [PubMed]
14. Pandey A, Goswami A, Jithin B, et al. Autophagy: The convergence point of aging and cancer. Biochem Biophys Rep 2025;42:101986. [Crossref] [PubMed]
15. Niu X, You Q, Hou K, et al. Autophagy in cancer development, immune evasion, and drug resistance. Drug Resist Updat 2025;78:101170. [Crossref] [PubMed]
16. Barthet VJA, Brucoli M, Ladds MJGW, et al. Autophagy suppresses the formation of hepatocyte-derived cancer-initiating ductular progenitor cells in the liver. Sci Adv 2021;7:eabf9141. [Crossref] [PubMed]
17. Frangež Ž, Gérard D, He Z, et al. ATG5 and ATG7 Expression Levels Are Reduced in Cutaneous Melanoma and Regulated by NRF1. Front Oncol 2021;11:721624. [Crossref] [PubMed]
18. Kang MR, Kim MS, Oh JE, et al. Frameshift mutations of autophagy-related genes ATG2B, ATG5, ATG9B and ATG12 in gastric and colorectal cancers with microsatellite instability. J Pathol 2009;217:702-6. [Crossref] [PubMed]
19. Chourasia AH, Tracy K, Frankenberger C, et al. Mitophagy defects arising from BNip3 loss promote mammary tumor progression to metastasis. EMBO Rep 2015;16:1145-63. [Crossref] [PubMed]
20. Humpton TJ, Alagesan B, DeNicola GM, et al. Oncogenic KRAS Induces NIX-Mediated Mitophagy to Promote Pancreatic Cancer. Cancer Discov 2019;9:1268-87. [Crossref] [PubMed]
21. Debnath J, Gammoh N, Ryan KM. Autophagy and autophagy-related pathways in cancer. Nat Rev Mol Cell Biol 2023;24:560-75. [Crossref] [PubMed]
22. Du H, Xu T, Yu S, et al. Mitochondrial metabolism and cancer therapeutic innovation. Signal Transduct Target Ther 2025;10:245. [Crossref] [PubMed]
23. Jin Y, Zhao L, Zhang Y, et al. BIN1 deficiency enhances ULK3-dependent autophagic flux and reduces dendritic size in mouse hippocampal neurons. Autophagy 2025;21:223-42. [Crossref] [PubMed]
24. Zheng X, Wang ZQ, Sun JL, et al. Overexpression of ULK3 mediates the effects of Gli1 on the proliferation, migration, and invasion of bladder cancer cells and its mechanisms. Chinese Journal of Anatomy and Clinics 2025;28:812-8.
25. Zhang N, Li Y, Sundquist J, et al. Identifying actionable druggable targets for breast cancer: Mendelian randomization and population-based analyses. EBioMedicine 2023;98:104859. [Crossref] [PubMed]
26. Shi K, Li Y, Yang L, et al. Profiling transcriptional heterogeneity of epithelium, fibroblasts, and immune cells in esophageal squamous cell carcinoma by single-cell RNA sequencing. FASEB J 2022;36:e22620. [Crossref] [PubMed]
27. Zhu A, Chen Z, Yan Q, et al. Robust mucosal SARS-CoV-2-specific T cells effectively combat COVID-19 and establish polyfunctional resident memory in patient lungs. Nat Immunol 2025;26:459-72. [Crossref] [PubMed]
28. Burgess S, Davey Smith G, Davies NM, et al. Guidelines for performing Mendelian randomization investigations: update for summer 2023. Wellcome Open Res 2019;4:186. [Crossref] [PubMed]
29. Huang Z, Chen J, Shi L, et al. Causal associations between smoking and ocular diseases: a Mendelian randomization study. Adv Ophthalmol Pract Res 2025;5:220-5. [Crossref] [PubMed]
30. Xu W, Zhang T, Zhu Z, et al. The association between immune cells and breast cancer: insights from Mendelian randomization and meta-analysis. Int J Surg 2025;111:230-41. [Crossref] [PubMed]
31. Li GR, Dai JH. Research progress in screening of early esophageal cancer in China. Chinese Journal of Thoracic and Cardiovascular Surgery 2021;37:52-8.
32. Committee of Esophageal Cancer Integrated Nursing of China Anti-Cancer Association. Expert consensus on whole-process management of esophageal cancer (2025 edition). Chinese Journal of Digestive Surgery 2025;24:438-51.
33. Ni W, Yu S, Zhang W, et al. A phase-II/III randomized controlled trial of adjuvant radiotherapy or concurrent chemoradiotherapy after surgery versus surgery alone in patients with stage-IIB/III esophageal squamous cell carcinoma. BMC Cancer 2020;20:130. [Crossref] [PubMed]
34. Patel D, Dabhi AM, Dmello C, et al. FKBP1A upregulation correlates with poor prognosis and increased metastatic potential of HNSCC. Cell Biol Int 2022;46:443-53. [Crossref] [PubMed]
35. Li S, Gao K, Yao D. Comprehensive analysis of autophagy associated genes and immune infiltrates in cervical cancer. Iran J Basic Med Sci 2024;27:813-24. [Crossref] [PubMed]
36. Liu L, Zhang J, Liu H, et al. Correlation of autophagy-related genes for predicting clinical prognosis in colorectal cancer. Biomark Med 2021;15:715-29. [Crossref] [PubMed]
37. Wang L, Klionsky DJ, Shen HM. The emerging mechanisms and functions of microautophagy. Nat Rev Mol Cell Biol 2023;24:186-203. [Crossref] [PubMed]
38. Berrino C, Omar A. Unravelling the Mysteries of the Sonic Hedgehog Pathway in Cancer Stem Cells: Activity, Crosstalk and Regulation. Curr Issues Mol Biol 2024;46:5397-419. [Crossref] [PubMed]
39. Yoshida S, Yoshida K. Regulatory mechanisms governing GLI proteins in hedgehog signaling. Anat Sci Int 2025;100:143-54. [Crossref] [PubMed]
40. Wang J, Cui B, Li X, et al. The emerging roles of Hedgehog signaling in tumor immune microenvironment. Front Oncol 2023;13:1171418. [Crossref] [PubMed]
41. Wang H, Sun P, Yuan X, et al. Autophagy in tumor immune escape and immunotherapy. Mol Cancer 2025;24:85. [Crossref] [PubMed]
42. Qiu X, Guo R, Wang Y, et al. Mendelian randomization reveals potential causal relationships between cellular senescence-related genes and multiple cancer risks. Commun Biol 2024;7:1069. [Crossref] [PubMed]
43. Goruppi S, Clocchiatti A, Bottoni G, et al. The ULK3 kinase is a determinant of keratinocyte self-renewal and tumorigenesis targeting the arginine methylome. Nat Commun 2023;14:887. [Crossref] [PubMed]