Development of a prognostic prediction model incorporating KDM1B for esophageal squamous cell carcinoma: an integrated transcriptomic and functional analysis

Introduction

Background

Esophageal cancer is the eighth most common malignancy and the sixth leading cause of cancer-related death worldwide (1-3). Because early-stage disease is often asymptomatic, many patients present with locally advanced or metastatic disease and are ineligible for surgery. For these patients, chemoradiotherapy remains a cornerstone of treatment, yet 5-year overall survival (OS) is only 15–20% (4). These data highlight the need to identify therapeutic targets and prognostic biomarkers.

Rationale and knowledge gap

Increasing evidence indicates that epigenetic alterations play a critical role in cancer initiation and progression (5,6). Among epigenetic modifications, histone methylation is a key chromatin-based regulator of transcription and is dynamically controlled by histone lysine methyltransferases and demethylases (KDMs) (7). Aberrant histone methylation can drive tumor progression and treatment resistance (7,8).

Histone demethylases comprise the JmjC-domain-containing enzymes and the flavin-dependent lysine-specific demethylase (LSD) family; the latter includes LSD1 (KDM1A) and LSD2 (KDM1B) (9). KDM1A and KDM1B, although belonging to the same LSD family, exhibit distinct substrate specificities and genomic profiles (7,10). KDM1A demethylates mono- and dimethylated H3K4 and, in certain transcriptional complexes such as those involving the androgen receptor, can also demethylate H3K9me1/2 (10). In contrast, KDM1B primarily demethylates H3K4me1/2 (7), suggesting a more restricted but potentially distinct role in chromatin regulation and transcriptional control. Beyond histone substrates, members of the LSD family have also been reported to act on non-histone proteins, including MYPT1, p53, and SOX2, thereby contributing to the regulation of proliferation, differentiation, and tumor progression (11-13).

Extensive studies have established the oncogenic roles of LSD1 in multiple malignancies (14-20). In contrast, the biological role of KDM1B in cancer remains less well understood. Emerging evidence suggests that KDM1B contributes to multiple processes relevant to tumorigenesis, including gene silencing, transcriptional regulation, and cell-cycle control, although its precise mechanisms remain incompletely understood. In breast cancer, KDM1B has been shown to promote cell proliferation and cancer stem cell properties while attenuating cell motility and invasiveness (21).

Evidence regarding KDM1B in gastrointestinal malignancies remains limited but is gradually increasing. Previous studies have implicated dysregulated KDM1B in gastric, hepatocellular, pancreatic, and colorectal cancers. For instance, ADPGKAS1 has been reported to promote gastric cancer progression through LSD2-mediated activation of the PI3K/AKT/mTOR pathway (22). Elevated KDM1B expression has also been associated with poor prognosis in hepatocellular carcinoma (23). In pancreatic cancer, KDM1B is overexpressed in tumor tissues and cell lines, and its knockdown inhibits proliferation and induces apoptosis, accompanied by alterations in ERK1/2, Smad2, p53, and apoptosis-related signaling pathways (24). Similarly, in colorectal cancer, KDM1B has been shown to promote cell proliferation and suppress apoptosis, at least in part by repressing p53 expression through histone demethylation and inhibiting the p53-p21 signaling pathway (25).

Objective

Despite these findings, the biological and clinical significance of KDM1B in esophageal squamous cell carcinoma (ESCC) remains unclear. Therefore, this study aimed to investigate KDM1B expression and genomic alterations in ESCC tissues, examine its biological effects in ESCC cells in vitro, and develop a prognostic prediction model incorporating KDM1B. This work characterizes the expression landscape, clinical relevance, and functional impact of KDM1B in ESCC while exploring its potential value in prognostic stratification. We present this article in accordance with the TRIPOD reporting checklist (available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0230/rc).

Methods

Patients and public transcriptomic datasets

RNA-sequencing and clinicopathologic data across TCGA cancer types were downloaded from The Cancer Genome Atlas (TCGA) via the Genomic Data Commons (GDC) portal in January 2025. ESCC cases were selected from the TCGA-esophageal carcinoma (ESCA) cohort. To externally validate the differential expression of KDM1B, two Gene Expression Omnibus (GEO) datasets were analyzed: GSE26886 (9 ESCC and 19 non-tumor esophageal tissues) and GSE161533 (28 ESCC tumors and 28 matched adjacent non-tumor tissues).

Pan-cancer and ESCC-specific expression pattern of KDM1B

TCGA KDM1B expression {log₂[transcripts per million (TPM) +1]} was profiled across cancer types, and tumor-normal differences were examined within each tumor type. For ESCC, KDM1B expression was analyzed using both unpaired comparisons of all available tumor and normal samples and paired comparisons restricted to cases with matched adjacent non-tumor tissues. Differential expression was independently validated in GSE26886 and GSE161533.

Survival analysis and prognostic model construction

Pan-cancer associations between KDM1B expression and OS were explored using KMplotter. For ESCC-specific prognostic analyses, only patients in the TCGA-ESCA cohort with histologically confirmed ESCC and available clinicopathologic and survival data were included. A total of 82 eligible patients were divided into high- and low-KDM1B expression groups according to the median messenger RNA (mRNA) expression level. OS, defined as the time from diagnosis to death from any cause or last follow-up, was the primary prognostic outcome. Disease-specific survival (DSS) and progression-free interval (PFI) were analyzed as secondary survival outcomes. Survival differences among groups were compared using Kaplan-Meier analysis with log-rank testing. Cases with missing clinicopathologic or survival data were excluded from the corresponding prognostic analyses, and complete-case analysis was performed.

Cox proportional hazards regression was then performed to evaluate the prognostic value of KDM1B expression and clinicopathologic variables obtained from the corresponding TCGA clinical annotations. Variables with P<0.10 in univariable analyses were entered into a multivariable Cox model to identify prognostic factors for OS. In addition, histologic grade was included in the multivariable model based on its established clinical relevance in ESCC. A nomogram incorporating the selected variables was constructed to estimate 1-, 3-, and 5-year OS, with variable weights determined according to the regression coefficients derived from the multivariable Cox model. The diagnostic performance of KDM1B expression for discriminating ESCC tumors from adjacent normal tissues was assessed using receiver operating characteristic (ROC) analysis and the area under the curve (AUC).

Correlation and functional enrichment analyses

Using normalized TCGA-ESCC expression data, Pearson correlation was computed between KDM1B and all genes to derive a co-expression profile. Genes with |r| ≥ 0.4 were defined as KDM1B-correlated genes and subjected to Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses. To link KDM1B-associated programs to clinical outcome, KDM1B-correlated genes were intersected with prognosis-associated genes identified by univariable Cox regression, and the overlapping gene set was further analyzed by GO/KEGG enrichment.

Protein-protein interaction (PPI) network analysis

Overlapping genes that were both KDM1B-related and prognosis-associated in ESCC were imported into the STRING database to construct a PPI network. Interactions meeting the default confidence score were retained and the network was examined to identify functional modules and hub genes. Pairwise expression correlations among hub genes were then calculated in TCGA-ESCC.

Immune infiltration analysis

To investigate the association between KDM1B expression and the tumor immune microenvironment, we estimated infiltration scores for multiple immune cell subsets across TCGA cancer types. Immune cell-specific gene signatures from published studies were used to calculate single-sample enrichment scores with the gene set variation analysis (GSVA) package in R. In ESCC, Spearman’s rank correlation coefficients were calculated between KDM1B expression and infiltration scores for selected immune cell subsets.

Human esophageal cancer tissues

A total of 52 primary esophageal cancer tissue samples were collected from Peking University Cancer Hospital, with paired normal samples consisting of 9 normal esophageal tissues from surgical resections and 43 peripheral blood samples. All specimens were reviewed by a pathologist to confirm the absence of tumor in normal samples and to verify squamous cell carcinoma histology with an estimated tumor purity of 60–70% in tumor tissues.

The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by Peking University Cancer Hospital & Institute Research Ethics Committee (No. 2024YJZ48), and individual consent for this retrospective analysis was waived.

Sequencing of KDM1B and data analyses

Genomic DNA was assessed for purity using a NanoDrop^® spectrophotometer (Thermo Fisher, Waltham, MA, USA), and DNA concentration was measured with a Qubit^® DNA Assay Kit on a Qubit^® 3.0 Fluorometer (Life Technologies, Carlsbad, CA, USA). Whole-exome sequencing libraries were prepared using the NimbleGen SeqCap EZ Human Exome V3 kit (Roche, Basel, Switzerland). Indexed libraries were clustered on an Illumina cBot system and sequenced on an Illumina HiSeq X Ten platform to generate 150-bp paired-end reads. Sequencing reads were aligned to the human reference genome (hg19) using the Burrows-Wheeler Aligner (BWA) (26). Variant calling followed the Genome Analysis Toolkit (GATK) best-practice workflow (27). Single-nucleotide variants (SNVs) and small insertions/deletions (indels) were called with GATK HaplotypeCaller and filtered using VariantFiltration. Variants were annotated with ANNOVAR using hg19 gene model annotation files (28). Copy-number alterations of KDM1B were inferred from exome read-depth profiles using a standard copy-number variation (CNV)-calling pipeline and categorized according to predefined thresholds.

Cell lines and reagents

Human ESCC cell lines KYSE150 (RRID: CVCL_1348) and TE1 (RRID: CVCL_C6K3), as well as human embryonic kidney 293T cells, were obtained from the National Infrastructure of Cell Line Resource (Beijing, China). According to the provider’s quality control documentation, these cell lines were authenticated by short tandem repeat (STR) profiling in December 2019 and tested for Mycoplasma contamination. Cells were used within 6 months of resuscitation. No additional authentication was performed in our laboratory. Cells were cultured in Dulbecco’s modified Eagle medium (DMEM; Hyclone™, Cat. SH30022.01B; GE Healthcare Bio-Sciences, Beijing, China) supplemented with 10% fetal bovine serum (FBS; Cat. A11-102; Shanghai Weike Biotechnology Co., Ltd., Shanghai, China) at 37 ℃ in a humidified incubator with 5% CO₂.

For western blotting, primary antibodies against KDM1B (SAB1302905; Sigma-Aldrich Ltd., Shanghai, China) and GAPDH (SC-32233, RRID: AB_627679; Santa Cruz Biotechnology, Santa Cruz, CA, USA), together with an HRP-conjugated secondary antibody (SC-2004, RRID: AB_631746; Santa Cruz Biotechnology, CA, USA), were used. For real-time quantitative polymerase chain reaction (RT-qPCR), M-MLV reverse transcriptase and RNase inhibitor (both from Promega Biotech Co., Ltd., Beijing, China) and SYBR Master Mix (Takara Bio, Dalian, China) were used. Transwell inserts (Corning Ltd., Shanghai, China) were used for migration assays, and methyl thiazolyl tetrazolium (MTT; Gen-View Scientific Inc., CA, USA) was used for cell viability assays.

Lentiviral short hairpin RNA (shRNA)-mediated knockdown of KDM1B

A shRNA targeting human KDM1B (sequence: GCAAGCAAGATTGCAGCATTT) was designed and cloned into the pGV115 lentiviral expression vector using the Lentivirus Expression System (Genchem Co., Ltd., Shanghai, China). 293T cells were cultured in DMEM supplemented with 10% heat-inactivated FBS and 1% penicillin-streptomycin and used for lentiviral packaging. Lentiviral particles were generated by co-transfecting 293T cells with pGV115-KDM1B or empty pGV115 vector (shCtrl), together with the Helper1.0 and Helper2.0 packaging plasmids.

Viral supernatants were harvested 48 h after transfection, and viral titers were determined using a fluorescence-based method. KYSE150 and TE1 cells were transduced with shKDM1B or shCtrl lentivirus, followed by selection with puromycin (1 µg/mL) to establish stable knockdown and control cell lines.

RT-qPCR

Total RNA was extracted from KYSE150 and TE1 cells after lentiviral transduction and reverse-transcribed into cDNA using the reagents from Promega Biotech Co., Ltd. (Beijing, China). KDM1B expression was measured by RT-qPCR. The primer sequences for KDM1B were: forward 5'-CTCTCCTGTGGGGAACATTTC-3' and reverse 5'-GACTAGGTTCGGTTTTGCCATT-3'. GAPDH was used as an internal control (forward 5'-TGACTTCAACAGCGACACCCA-3' and reverse 5'-CACCCTGTTGCTGTAGCCAAA-3'). Relative KDM1B expression levels were calculated using the 2^−ΔΔCt method.

Western blotting

KYSE150 and TE1 cells with or without KDM1B knockdown were grown to near confluence, harvested, and washed twice with cold PBS. Cells were lysed in radioimmunoprecipitation assay (RIPA) buffer (Beyotime Biotechnology Ltd., Shanghai, China) supplemented with 1 mM phenylmethylsulfonyl fluoride (PMSF) on ice for 15 min. Lysates were briefly sonicated and centrifuged at 10,000 × g for 15 min at 4 ℃, and the supernatants were collected. Protein concentrations were determined using a bicinchoninic acid (BCA) protein assay kit (Beyotime Biotechnology Ltd.), and samples were adjusted to equal protein concentrations. Equal amounts of protein were separated by sodium dodecyl-sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and transferred onto polyvinylidene difluoride (PVDF) membranes (Millipore, Burlington, MA, USA). Membranes were blocked and then incubated with primary antibodies against KDM1B (1:250) and GAPDH (1:2,000), followed by horseradish peroxidase-conjugated secondary antibodies (1:2,000). Signals were detected using LumiGLO chemiluminescent substrate [Cell Signaling Technology (China) Ltd., Shanghai, China] and visualized on X-ray film.

Cell proliferation assays in vitro

For cell proliferation assays, KYSE150 and TE1 cells were seeded into 96-well plates at 2,000 cells per well, and cell numbers were recorded daily using a Celigo automated cell counter (Nexcelom Bioscience Instruments Co., Ltd., Shanghai, China).

In parallel, cell viability was assessed by MTT assay. Briefly, 20 µL of MTT solution (5 mg/mL) was added to each well and cells were incubated for 4 h. Then 100 µL of DMSO was added to dissolve the formazan crystals. Absorbance was measured at 490 and 570 nm using a Tecan Infinite M200 PRO microplate reader (Tecan Group, Mannedorf, Switzerland).

Flow cytometric analysis of apoptotic cells

Apoptosis was assessed using an eBioscience™ Annexin V Apoptosis Detection Kit (Thermo Fisher Scientific, Waltham, MA, USA). Both adherent cells and cells in the culture medium were collected, combined, and washed twice with cold D-Hanks buffer (pH 7.2–7.4). After centrifugation, the cell pellet was resuspended in 1× annexin-binding buffer and adjusted to approximately 1×10⁶ cells/mL. Cells were then incubated with fluorescein isothiocyanate (FITC)-conjugated Annexin V and analyzed on a BD Accuri C6 Plus flow cytometer (BD Biosciences, Franklin Lakes, NJ, USA).

Wound-healing and Transwell migration assays

Migration after KDM1B knockdown was assessed by wound-healing and Transwell assays. For wound healing, KYSE150 and TE1 cells were seeded in 96-well plates (1×10⁴ cells/well) and grown to ~90% confluence; scratches were introduced and cells were maintained in 1% FBS medium. Images were acquired at 0, 8, and 24 h, and migration was quantified as wound closure relative to baseline.

For Transwell migration, 7×10⁴ cells in serum-free medium were plated in the upper chamber, with 30% FBS medium in the lower chamber as a chemoattractant. After 24 h, migrated cells were fixed, crystal-violet stained, and quantified.

Statistical analysis

Differential expression between tumor and normal tissues was assessed using Wilcoxon tests (rank-sum for unpaired and signed-rank for paired comparisons). Survival outcomes (OS/DSS/PFI) were analyzed using Kaplan-Meier estimates with log-rank testing. For prognostic analyses, KDM1B expression was dichotomized into high- and low-expression groups according to the median mRNA expression level in the TCGA-ESCC cohort. Clinicopathologic variables were analyzed according to their original categorical definitions in the TCGA clinical annotations. Univariable and multivariable Cox proportional hazards models were used to estimate hazard ratios (HRs) with 95% confidence intervals (CIs). Variables meeting the prespecified criterion in univariable analyses (P<0.10) were entered into a multivariable Cox model, and a nomogram incorporating the selected variables was constructed to estimate 1-, 3-, and 5-year OS. Model performance was assessed using the concordance index (C-index), time-dependent ROC curves, and calibration plots. Internal validation was performed by bootstrap resampling with 1,000 iterations.

Associations between KDM1B expression and clinicopathologic variables were evaluated by logistic regression, and tumor-normal discrimination was assessed by ROC curve analysis with the AUC. Functional enrichment (GO/KEGG) was performed for KDM1B-correlated and overlapping gene sets. Gene co-expression and immune infiltration associations were evaluated using Pearson and Spearman correlation, respectively.

In vitro data are presented as mean ± standard deviation (SD) from ≥3 independent experiments performed in triplicate and were analyzed using Student’s t-test or one-way analysis of variance (ANOVA), as indicated. Analyses were performed in R (version 3.5.3) and SPSS (version 17.0). All tests were two-sided, and P<0.05 was considered statistically significant.

Results

Expression of KDM1B across various cancers

We first evaluated KDM1B expression across multiple tumor types using TCGA pan-cancer data (Figure 1). KDM1B was significantly upregulated in several cancers, including breast invasive carcinoma (BRCA), cholangiocarcinoma (CHOL), colon adenocarcinoma (COAD), ESCA, head and neck squamous cell carcinoma (HNSC), liver hepatocellular carcinoma (LIHC), lung squamous cell carcinoma (LUSC), and stomach adenocarcinoma (STAD), compared with corresponding normal tissues. By contrast, KDM1B expression was lower in tumors than in normal tissues for kidney chromophobe (KICH) and thyroid carcinoma (THCA), and in many other cancer types no significant difference was observed between tumor and normal samples (Figure 1A).

Figure 1 Pan-cancer expression pattern of KDM1B and its clinical relevance in ESCC. (A) KDM1B mRNA expression in tumor vs. normal tissues across TCGA cancer types. (B,C) KDM1B expression in ESCC and normal esophageal tissues from the TCGA-ESCA cohort: (B) unpaired comparison between all ESCC and normal samples; (C) paired comparison between ESCC tissues and matched adjacent normal epithelium. (D,E) Validation of KDM1B upregulation in GEO ESCC datasets GSE26886 (D) and GSE161533 (E) comparing tumor and para-carcinoma tissues. (F-H) Kaplan-Meier curves for OS (F), DSS (G) and PFI (H) in ESCC patients from the TCGA-ESCA cohort stratified into high- and low-KDM1B expression groups. *, P<0.05; **, P<0.01; ***, P<0.001. CI, confidence interval; DSS, disease-specific survival; ESCA, esophageal carcinoma; ESCC, esophageal squamous cell carcinoma; GEO, Gene Expression Omnibus; HR, hazard ratio; mRNA, messenger RNA; OS, overall survival; PFI, progression-free interval; TCGA, The Cancer Genome Atlas; TPM, transcripts per million.

We then focused on ESCC. In the TCGA-ESCC cohort, KDM1B expression was significantly higher in tumors than in normal esophageal tissues in both unpaired analyses of all available samples and paired analyses restricted to matched tumor-normal pairs (Figure 1B,1C). The upregulation of KDM1B in ESCC was independently validated in two GEO datasets, GSE26886 and GSE161533, in which KDM1B expression was consistently elevated in ESCC compared with adjacent non-tumor esophageal tissues (Figure 1D,1E).

KDM1B expression and clinicopathologic features

We next examined associations between KDM1B expression and clinicopathologic characteristics in ESCC. Higher KDM1B expression was significantly associated with better primary therapy outcome, as patients in the high KDM1B group exhibited a higher response rate to systemic treatment than those in the low-expression group. In addition, KDM1B expression differed by race, with a higher proportion of Asian patients classified into the high-expression group. KDM1B expression was also related to body weight and body mass index (BMI): patients with weight ≤70 kg or BMI ≤25 kg/m² were more likely to have high KDM1B expression (Table 1). In multivariable logistic regression analysis, race, BMI, and weight remained independently associated with KDM1B expression (Table 2).

Prognostic impact of KDM1B in esophageal cancer

Pan-cancer survival analysis using KMplotter showed that high KDM1B expression was associated with favorable OS in ESCC, kidney renal clear cell carcinoma, rectum adenocarcinoma, and STAD, but predicted worse OS in LIHC, sarcoma, and uterine corpus endometrial carcinoma; in other tumor types, KDM1B expression was not significantly associated with OS (all P>0.05) (Figure 2), indicating that the prognostic impact of KDM1B is highly context- and tissue-dependent.

Figure 2 Pan-cancer prognostic significance of KDM1B expression for OS in TCGA cohorts. (A-I) Kaplan-Meier OS curves for patients with ESCC (A), EAC (B), HNSC (C), KIRC (D), SARC (E), READ (F), STAD (G), UCEC (H) and LIHC (I) from TCGA. CI, confidence interval; EAC, esophageal adenocarcinoma; ESCC, esophageal squamous cell carcinoma; HNSC, head and neck squamous cell carcinoma; HR, hazard ratio; KIRC, kidney renal clear cell carcinoma; LIHC, liver hepatocellular carcinoma; OS, overall survival; READ, rectum adenocarcinoma; SARC, sarcoma; STAD, stomach adenocarcinoma; TCGA, The Cancer Genome Atlas; UCEC, uterine corpus endometrial carcinoma.

Within the TCGA-ESCA cohort, 82 patients with histologically confirmed ESCC and available clinicopathologic and survival data were included in the prognostic analyses, among whom 56 OS events were observed during follow-up. Patients with high KDM1B expression exhibited significantly longer OS compared with those with low expression (P=0.02) (Figure 1F). Consistent patterns were observed for DSS and PFI (Figure 1G,1H).

To further assess the prognostic value of KDM1B in ESCC, we performed Cox regression analyses. In univariable models, pathologic N category, pathologic M category, pathologic stage, gender, and KDM1B expression were associated with OS (P<0.10). These variables, together with histologic grade based on clinical relevance, were subsequently entered into a multivariable Cox model. After adjustment, high KDM1B expression remained an independent favorable prognostic factor for OS (Table 3, Figure 3). A prognostic nomogram incorporating KDM1B expression and selected clinicopathologic variables was then constructed to estimate 1-, 3-, and 5-year OS in ESCC (Figure 4A). For individual prediction, points assigned to each predictor are summed to generate a total score, which corresponds to the estimated probabilities of 1-, 3-, and 5-year OS. The prognostic nomogram showed acceptable discriminative ability, with a C-index of 0.711 (95% CI: 0.640–0.782). After bootstrap internal validation, calibration curves showed good agreement between the predicted and observed probabilities of 1-, 3-, and 5-year OS.

Figure 3 Forest plot of multivariable Cox regression evaluating the prognostic impact of KDM1B and clinicopathologic variables in ESCC. CI, confidence interval; ESCC, esophageal squamous cell carcinoma; HR, hazard ratio; Inf, infinity; M, metastasis; N, node.

Figure 4 Diagnostic performance of KDM1B expression and nomogram-based prediction of OS in ESCC. (A) Nomogram constructed from a multivariable Cox proportional hazards model integrating pathologic N stage, pathologic M stage, overall pathologic stage, gender, histologic grade and KDM1B expression to estimate individual 1-, 3-, and 5-year OS probabilities in ESCC patients. (B) ROC curve evaluating the ability of KDM1B mRNA expression to distinguish ESCC tissues from adjacent normal esophageal mucosa in the TCGA-ESCC cohort. AUC, area under the curve; CI, confidence interval; ESCC, esophageal squamous cell carcinoma; FPR, false positive rate; mRNA, messenger RNA; M, metastasis; N, node; OS, overall survival; ROC, receiver operating characteristic; TCGA, The Cancer Genome Atlas; TPR, true positive rate.

In addition to its prognostic relevance, ROC analysis showed that KDM1B expression could discriminate ESCC tumors from adjacent normal tissues, with an AUC of 0.805 (Figure 4B). These findings support the potential value of KDM1B in both diagnostic discrimination and prognostic assessment in ESCC.

KDM1B-related genes and functional enrichment in ESCC

To explore pathways associated with KDM1B in ESCC, we constructed a co-expression profile of KDM1B using TCGA transcriptomic data (Figure 5). Pearson correlation coefficients between KDM1B and all other genes were calculated, and the 25 most positively and 25 most negatively correlated genes were shown in heatmaps (Figure 5A,5B). Genes with an absolute correlation coefficient (|r|) ≥0.4 (n=688) were defined as KDM1B-related genes and subjected to GO and KEGG enrichment analysis. These genes were significantly enriched in pathways involved in nucleocytoplasmic transport, cell-cycle regulation, ubiquitin-mediated proteolysis, the Fanconi anemia pathway, and ribosome biogenesis in eukaryotes (Table 4, Figure 5C), indicating that KDM1B expression is strongly associated with proliferation- and genome maintenance-related processes in ESCC.

Figure 5 Correlation landscape and functional annotation of KDM1B-related genes in ESCC. (A,B) Heatmaps showing the expression patterns of the 25 genes most negatively (A) and most positively (B) correlated with KDM1B in the TCGA-ESCC cohort. (C) GO enrichment bubble plot for genes with an absolute correlation coefficient |r|>0.4 with KDM1B. (D) GO enrichment analysis of the overlapping genes that are both KDM1B-related (|r|>0.4) and survival-related, displayed as a bubble plot. (E) Venn diagram showing the intersection between KDM1B-related genes (|r|>0.4) and survival-related genes in ESCC. ESCC, esophageal squamous cell carcinoma; GO, Gene Ontology; LSU, large subunit; rRNA, ribosomal RNA; SSU, small subunit; TCGA, The Cancer Genome Atlas.

To connect these KDM1B-related genes with clinical outcome, we intersected the 688 KDM1B-related genes (|r|≥0.4) with 813 prognosis-associated genes identified by univariable Cox regression in ESCC, and obtained 154 overlapping genes (Figure 5E). GO and KEGG enrichment analyses of this overlapping gene set again highlighted nucleocytoplasmic transport and cell-cycle pathways as the most significantly enriched categories (Table 5, Figure 5D). These results suggest that KDM1B-related cell-cycle and transport pathways may underlie the prognostic association of KDM1B in ESCC.

PPI network analysis

To further characterize KDM1B-related genes with potential functional importance, we constructed a PPI network for the 154 overlapping genes using the STRING database. The network contained several densely connected regions, and one prominent interaction cluster included 11 hub genes (ABCE1, MTREX, UTP20, POLR1A, WDR43, HEATR1, TSR1, DHX33, NOL10, UTP25, and DHX37) (Figure 6A). Their pairwise expression relationships in ESCC were evaluated using TCGA transcriptomic data, and the heatmap showed strong positive correlations among most of these hub genes (Figure 6B), supporting their participation in a coordinated module.

Figure 6 PPI network and co-expression patterns of prognosis-related KDM1B-associated genes in ESCC. (A) PPI network constructed in STRING for the intersecting genes that were both correlated with KDM1B (|r| >0.4) and associated with ESCC patient prognosis. (B) Heatmap showing pairwise expression correlations among the key hub genes extracted from the PPI network in TCGA-ESCC cohort. **, P<0.01; ***, P<0.001. Cor, correlation coefficient; ESCC, esophageal squamous cell carcinoma; PPI, protein-protein interaction; TCGA, The Cancer Genome Atlas.

Correlation between KDM1B and immune infiltration in ESCC

We next examined the relationship between KDM1B expression and immune cell infiltration using immune cell-specific gene signatures. In the pan-cancer analysis, higher KDM1B expression was generally positively correlated with T helper cells and central memory T cells (Tcm), but negatively correlated with natural killer (NK) CD56^bright cells, NK CD56^dim cells, and plasmacytoid dendritic cells (pDC) (Figure 7A). In ESCC, KDM1B expression showed a distinct immune pattern. Higher KDM1B levels were significantly negatively associated with cytotoxic cells, CD8⁺ T cells, total T cells, and NK CD56^dim cells (Figure 7B-7F).

Figure 7 Association between KDM1B expression and immune cell infiltration in ESCC and other cancer types. (A) Heatmap depicting the correlations between KDM1B mRNA expression and the estimated infiltration levels of multiple immune cell subsets across TCGA cancer types. (B) Lollipop plot summarizing the correlations between KDM1B expression and immune cell infiltration in ESCC cohort. (C-F) Scatter plots illustrating representative correlations between KDM1B expression and enrichment scores of total T cells (C), CD8 T cells (D), cytotoxic cells (E) and NK CD56dim cells (F) in ESCC tumor samples. *, P<0.05; **, P<0.01; ***, P<0.001; ns, not significant. Cor, correlation coefficient; ESCC, esophageal squamous cell carcinoma; mRNA, messenger RNA; NK, natural killer; TCGA, The Cancer Genome Atlas; TPM, transcripts per million.

KDM1B genomic alterations in ESCC tissues

Targeted sequencing of KDM1B was performed in 52 primary ESCC tumors. Intronic KDM1B SNVs were identified in 6 cases (11.5%) (Table 6), whereas no SNVs or small indels were detected in the coding exons. Copy-number analysis showed low-level copy-number gain at the KDM1B locus in 4 of 52 tumors (7.7%) with an AAB genotype (Table 7), while the remaining tumors were copy-number neutral. Kaplan-Meier analysis showed no significant association between OS and either KDM1B SNVs or copy-number gain in this cohort (SNVs: P=0.88; copy-number gain: P=0.37).

To further assess the clinical relevance of KDM1B genomic alterations, we analyzed mutation data from TCGA-ESCC cases. Consistently, no significant difference in OS was observed between patients with KDM1B-mutant and KDM1B-wild-type tumors (P=0.57). Moreover, KDM1B mutation status was not significantly associated with KDM1B expression levels in TCGA-ESCC samples (P=0.51).

KDM1B knockdown in ESCC cell lines

To investigate the functional role of KDM1B in ESCC, we used lentiviral shRNA to silence KDM1B in KYSE150 and TE1 cells. Knockdown efficiency was confirmed by RT-qPCR and western blotting. Compared with control cells, KDM1B expression was reduced by approximately 50% in KYSE150 cells and 65% in TE1 cells (Figure 8).

Figure 8 KDM1B knockdown suppresses the proliferation of ESCC cells in vitro. (A,B) Knockdown efficiency of KDM1B in TE1 (A) and KYSE150 (B) cells, confirmed by RT-qPCR and western blotting. (C,D) Proliferation curves of TE1 (C) and KYSE150 (D) cells with or without KDM1B knockdown assessed by MTT assay. (E,F) Celigo-based cell counting assays showing the effects of KDM1B knockdown on the proliferation of TE1 (E) and KYSE150 (F) cells. *, P<0.05; **, P<0.01. ESCC, esophageal squamous cell carcinoma; MTT, methyl thiazolyl tetrazolium; RT-qPCR, real-time quantitative polymerase chain reaction; sh, short hairpin RNA.

KDM1B knockdown suppresses ESCC cell proliferation

We next evaluated the effect of KDM1B knockdown on ESCC cell proliferation. Celigo-based cell counting and MTT assays were used to assess cell growth in vitro. In the Celigo assay, KYSE150 cells transduced with shKDM1B showed a markedly lower growth rate at day 5 than control cells (2.14±0.07 vs. 3.84±0.02, P < 0.01) (Figure 8). TE1 cells with KDM1B knockdown also exhibited a reduced growth rate at day 5 compared with control cells (2.63±0.12 vs. 4.04±0.08, P < 0.01) (Figure 8E,8F).

MTT assays yielded similar results. Five days after seeding, the absorbance of KYSE150 cells in the shKDM1B group was 2.59±0.02, compared with 4.11±0.19 in the control group (P<0.01). In TE1 cells, the corresponding values were 2.18±0.09 in the shKDM1B group and 3.60±0.01 in the control group (P<0.01) (Figure 8C,8D). These findings indicate that KDM1B promotes the proliferative capacity of ESCC cells.

KDM1B knockdown increases apoptosis in ESCC cells

To determine whether the reduced cell growth after KDM1B knockdown was associated with apoptosis, we performed flow cytometric analysis using Annexin V staining. The proportion of apoptotic cells was higher in shKDM1B-transduced KYSE150 and TE1 cells than in their respective controls. Under serum-free conditions, the apoptotic rates of KYSE150 and TE1 cells with KDM1B knockdown were 7.29%±0.16% and 8.43%±0.13%, respectively, compared with 2.49%±0.06% and 3.51%±0.08% in control cells cultured in normal medium. KDM1B knockdown therefore significantly increased apoptosis in both KYSE150 (P<0.01) and TE1 cells (P<0.01) (Figure 9A).

Figure 9 KDM1B regulates apoptosis and migration of esophageal cancer cells. Flow cytometry was used to analyze the impact of KDM1B on the apoptosis of esophageal cancer cells (A). Wound-healing assay (B) and Transwell based migration assay (C) was performed to evaluate the impact of KDM1B on the migration of esophageal cancer cells. *, P<0.05; **, P<0.01. FACS, fluorescence-activated cell sorting; sh, short hairpin RNA.

KDM1B knockdown reduces migration of ESCC cells

To assess the impact of KDM1B knockdown on ESCC cell migration, we performed wound-healing and Transwell migration assays. In the wound-healing assay, KYSE150 cells transduced with shKDM1B showed a significantly lower wound closure rate at 8 h than control cells (75.45%±5.79% vs. 96.52%±3.45%, P<0.01). In TE1 cells, KDM1B knockdown also significantly reduced wound closure at 8 h compared with control cells (65.42%±8.54% vs. 89.01%±9.74%, P<0.05) (Figure 9B).

In the Transwell migration assay, the number of KYSE150 cells that migrated to the lower surface of the membrane within 24 h was markedly decreased in the shKDM1B group compared with the control group (34.04±5.30 vs. 84.15±5.30, P<0.01). In TE1 cells, KDM1B knockdown also led to a pronounced reduction in the number of migrated cells (5.37±2.00 vs. 92.00±8.58, P<0.01) (Figure 9C). These data indicate that KDM1B contributes to the migratory capacity of ESCC cells.

Discussion

Epigenetic dysregulation is a well-established driver of tumor initiation and progression through its effects on oncogene activation and tumor suppressor silencing (29). LSDs (KDMs), including KDM1A and KDM1B, play critical roles in regulating histone methylation (7,30,31). While the oncogenic functions of LSD1 have been extensively characterized across multiple malignancies (16-18,20), the biological significance of LSD2 is less understood, particularly in esophageal cancer. In the present study, we performed a comprehensive, multi-level investigation of KDM1B in ESCC, integrating transcriptomic, genomic, immunological, and functional analyses. To our knowledge, this is the first systematic exploration of KDM1B’s clinical and biological role in ESCC.

Our analysis showed that KDM1B expression varies across tumor types, indicating that its biological function is context-dependent. In ESCC, KDM1B expression was consistently elevated across datasets and linked to improved OS and better response to systemic therapy, suggesting its potential role as a prognostic and predictive biomarker. Notably, multivariate Cox regression analysis identified KDM1B expression and histologic grade as independent prognostic factors in ESCC, underscoring the clinical relevance of KDM1B beyond conventional clinicopathologic parameters.

Histologic grade is a known prognostic factor in ESCC, with tumors typically exhibiting more aggressive behavior and worse outcomes (32-36). Therefore, histologic grade was included in the multivariate Cox regression to adjust for potential confounders. The identification of both KDM1B expression and histologic grade as independent prognostic factors supports the robustness of our findings and suggests that KDM1B provides prognostic information beyond traditional pathological parameters.

In recent years, prognostic models integrating molecular biomarkers with clinicopathologic variables have been increasingly applied in oncology to facilitate individualized risk stratification and clinical decision-making (37-41). However, prognostic models specifically developed for ESCC remain limited, and most existing models primarily focus on patients undergoing curative surgery (42,43). In this study, we constructed a prognostic model incorporating KDM1B expression and key clinical variables to predict OS in ESCC. This model may provide complementary prognostic information and has potential utility in guiding risk assessment and treatment strategies.

From a clinical perspective, the model may serve as a practical tool for risk stratification, enabling estimation of individual survival risk. Patients predicted by this model to have a relatively favorable prognosis may be considered for more active and comprehensive treatment approaches when clinically appropriate, including appropriately intensive systemic therapy and consideration of additional local treatment modalities, such as surgery or radiotherapy. In contrast, for patients with a predicted poor prognosis, treatment strategies may need to be more cautiously balanced with expected benefit and treatment-related toxicity. In this way, the model may contribute to more individualized and risk-adapted clinical decision-making, helping to optimize treatment selection and patient management. Nevertheless, these potential applications require further validation in independent and prospective cohorts before routine clinical implementation.

To gain insight into the potential mechanisms underlying the role of KDM1B in ESCC, we performed correlation analysis followed by functional enrichment of genes most strongly associated with KDM1B expression. These analyses revealed that KDM1B-correlated genes were enriched in pathways related to nucleocytoplasmic transport, cell cycle regulation, ubiquitin-mediated proteolysis, the Fanconi anemia pathway, and ribosome biogenesis, suggesting a close association between KDM1B expression and fundamental cellular processes. Notably, among the genes that were both highly correlated with KDM1B expression and significantly associated with patient prognosis, functional enrichment again highlighted cell cycle-related pathways, reinforcing the potential link between KDM1B and cell cycle regulation in ESCC. Furthermore, PPI network analysis identified 11 hub genes (ABCE1, MTREX, UTP20, POLR1A, WDR43, HEATR1, TSR1, DHX33, NOL10, UTP25, and DHX37), most of which have been implicated in cell cycle or apoptosis in different cancers (44-48). Collectively, these integrative analyses suggest that KDM1B may influence ESCC biology by modulating transcriptional programs related to cell cycle control and apoptosis, thereby supporting tumor cell proliferation at the cellular level.

Given the increasing importance of immunotherapy in ESCC (49-52), we further explored the relationship between KDM1B expression and immune infiltration. KDM1B expression was negatively correlated with several immune cell populations, including cytotoxic cells, CD8⁺ T cells, and NK cells. Although the precise mechanisms remain unclear, these findings suggest that KDM1B may also participate in shaping the tumor immune microenvironment and could influence treatment response. Further studies are needed to clarify the interplay between KDM1B-mediated epigenetic regulation and antitumor immunity in ESCC.

Intronic variants and copy-number alterations of KDM1B were detected only in a small subset of ESCC tumors. Their functional and clinical significance remains uncertain and will require further mechanistic and larger-cohort studies to clarify.

Consistent with our bioinformatic findings, functional assays showed that KDM1B knockdown significantly suppressed proliferation and increased apoptosis in ESCC cell lines. These results are in line with previous reports in pancreatic (24) and colorectal cancers (25). Previous work has suggested that KDM1B can enhance tumor growth by repressing the p53-p21 axis through H3K4me2 demethylation at the p53 promoter (25). Whether a similar mechanism operates in ESCC remains to be determined.

In addition to its effects on proliferation and apoptosis, KDM1B knockdown also significantly reduced the migration of ESCC cells, suggesting a role in regulating tumor cell motility. Notably, a previous study in gastric cancer suggested that elevated KDM1B expression may be associated with activation of the PI3K/AKT/mTOR signaling pathway (53). Whether this pathway contributes to KDM1B-driven migration in ESCC requires further investigation.

Interestingly, the findings of our study were not entirely consistent with our initial expectations. Although KDM1B expression was significantly elevated in ESCC tissues and in vitro assays suggested that KDM1B may promote tumor cell proliferation and inhibit apoptosis, survival analyses demonstrated that patients with higher KDM1B expression had better survival. One plausible explanation is that ESCC tumors with high KDM1B expression may be more sensitive to cytotoxic chemotherapy. In this context, enhanced proliferative activity could increase susceptibility to treatment, thereby offsetting aggressive biological behavior and ultimately translating into improved survival. In addition, the observed association between high KDM1B expression and altered immune cell infiltration raises the possibility that KDM1B-related epigenetic programs may also modulate the tumor immune microenvironment in a way that interacts with systemic therapy. Nevertheless, the molecular mechanisms underlying this observation remain unclear and warrant further investigation.

There are several limitations in this study. KDM1B protein expression was not evaluated in ESCC tissues, and the consistency between mRNA and protein levels, as well as the association between protein expression and patient prognosis, remain to be further validated in future studies. In addition, in vivo validation was not performed; therefore, the role of KDM1B in tumor growth and tumor-immune interactions remains to be further investigated using appropriate animal models. Furthermore, the prognostic model was developed using a relatively small TCGA-ESCC cohort and still requires independent external validation despite internal validation. These limitations indicate important directions for future research.

Conclusions

In summary, our study demonstrates that KDM1B is significantly upregulated in ESCC compared with normal esophageal epithelium and is functionally associated with tumor cell proliferation, apoptosis, and migration. Clinically, elevated KDM1B expression was correlated with improved treatment response and independently predicted favorable prognosis in ESCC. Furthermore, we developed a prognostic model incorporating KDM1B expression and clinicopathologic variables to estimate 1-, 3-, and 5-year OS, which may help refine risk stratification, although external validation is still needed. Collectively, these findings highlight the potential clinical relevance of KDM1B in ESCC and lay the groundwork for future studies aimed at elucidating its molecular mechanisms and therapeutic implications.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0230/rc

Data Sharing Statement: Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0230/dss

Peer Review File: Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0230/prf

Funding: This study was partially funded by the “Beijing Weiai Foundation” (No. JYKY2023-0101220005) and the “Capital’s Funds for Health Improvement and Research” project (No. 2016-2-2152).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0230/coif). All authors report the funding from the “Beijing Weiai Foundation” (No. JYKY2023-0101220005) and the “Capital’s Funds for Health Improvement and Research” project (No. 2016-2-2152). The authors have no other 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. The study was approved by Peking University Cancer Hospital & Institute Research Ethics Committee (No. 2024YJZ48), and individual consent for this retrospective analysis was waived.

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. Aoyama T, Kazama K, Maezawa Y, et al. Usefulness of Nutrition and Inflammation Assessment Tools in Esophageal Cancer Treatment. In Vivo 2023;37:22-35. [Crossref] [PubMed]
2. Dong K, Tian Z, Zhang Y, et al. Long non-coding RNA LINC00092 inhibits esophageal squamous cell carcinoma progression by promoting ferroptosis through the MAZ/NFE2L2 axis. J Thorac Dis 2026;18:39. [Crossref] [PubMed]
3. He S, Xu J, Liu X, et al. Advances and challenges in the treatment of esophageal cancer. Acta Pharm Sin B 2021;11:3379-92. [Crossref] [PubMed]
4. Wang X, Hobbs B, Gandhi SJ, et al. Current status and application of proton therapy for esophageal cancer. Radiother Oncol 2021;164:27-36. [Crossref] [PubMed]
5. Liu Z, Wang Y, Dou C, et al. Hypoxia-induced up-regulation of VASP promotes invasiveness and metastasis of hepatocellular carcinoma. Theranostics 2018;8:4649-63. [Crossref] [PubMed]
6. Ye C, Yan X, Gao Y. Advances in epigenetic therapy for esophageal cancer. Clin Epigenetics 2026;18:42. [Crossref] [PubMed]
7. Sterling J, Menezes SV, Abbassi RH, et al. Histone lysine demethylases and their functions in cancer. Int J Cancer 2021;148:2375-88. [Crossref] [PubMed]
8. Eckschlager T, Vicha A, Frolikova D. Lysine demethylases and cancer. Pathol Res Pract 2025;271:156011. [Crossref] [PubMed]
9. Maiques-Diaz A, Somervaille TC. LSD1: biologic roles and therapeutic targeting. Epigenomics 2016;8:1103-16. [Crossref] [PubMed]
10. Hosseini A, Minucci S. A comprehensive review of lysine-specific demethylase 1 and its roles in cancer. Epigenomics 2017;9:1123-42. [Crossref] [PubMed]
11. Cho HS, Suzuki T, Dohmae N, et al. Demethylation of RB regulator MYPT1 by histone demethylase LSD1 promotes cell cycle progression in cancer cells. Cancer Res 2011;71:655-60. [Crossref] [PubMed]
12. Santarelli R, Di Dio C, Di Crosta M, et al. Modulatory Effect of Curcumin on Expression of Methyltransferase/Demethylase in Colon Cancer Cells: Impact on wt p53, mutp53 and c-Myc. Molecules 2025;30:3054. [Crossref] [PubMed]
13. Zhang C, Hoang N, Leng F, et al. LSD1 demethylase and the methyl-binding protein PHF20L1 prevent SET7 methyltransferase-dependent proteolysis of the stem-cell protein SOX2. J Biol Chem 2018;293:3663-74. [Crossref] [PubMed]
14. Malagraba G, Yarmohammadi M, Javed A, et al. The Role of LSD1 and LSD2 in Cancers of the Gastrointestinal System: An Update. Biomolecules 2022;12:462. [Crossref] [PubMed]
15. Liu YW, Xia R, Lu K, et al. LincRNAFEZF1-AS1 represses p21 expression to promote gastric cancer proliferation through LSD1-Mediated H3K4me2 demethylation. Mol Cancer 2017;16:39. [Crossref] [PubMed]
16. Majello B, Gorini F, Saccà CD, et al. Expanding the Role of the Histone Lysine-Specific Demethylase LSD1 in Cancer. Cancers (Basel) 2019;11:324. [Crossref] [PubMed]
17. Miller SA, Policastro RA, Sriramkumar S, et al. LSD1 and Aberrant DNA Methylation Mediate Persistence of Enteroendocrine Progenitors That Support BRAF-Mutant Colorectal Cancer. Cancer Res 2021;81:3791-805. [Crossref] [PubMed]
18. Sheng W, LaFleur MW, Nguyen TH, et al. LSD1 Ablation Stimulates Anti-tumor Immunity and Enables Checkpoint Blockade. Cell 2018;174:549-563.e19. [Crossref] [PubMed]
19. Qin Y, Vasilatos SN, Chen L, et al. Inhibition of histone lysine-specific demethylase 1 elicits breast tumor immunity and enhances antitumor efficacy of immune checkpoint blockade. Oncogene 2019;38:390-405. [Crossref] [PubMed]
20. Boulding T, McCuaig RD, Tan A, et al. LSD1 activation promotes inducible EMT programs and modulates the tumour microenvironment in breast cancer. Sci Rep 2018;8:73. [Crossref] [PubMed]
21. Chen L, Vasilatos SN, Qin Y, et al. Functional characterization of lysine-specific demethylase 2 (LSD2/KDM1B) in breast cancer progression. Oncotarget 2017;8:81737-53. [Crossref] [PubMed]
22. Hu ZQ, Li HC, Teng F, et al. Long noncoding RNA MAFG-AS1 facilitates the progression of hepatocellular carcinoma via targeting miR-3196/OTX1 axis. Eur Rev Med Pharmacol Sci 2020;24:12131-43. [Crossref] [PubMed]
23. Bayo J, Fiore EJ, Dominguez LM, et al. A comprehensive study of epigenetic alterations in hepatocellular carcinoma identifies potential therapeutic targets. J Hepatol 2019;71:78-90. [Crossref] [PubMed]
24. Wang Y, Sun L, Luo Y, et al. Knockdown of KDM1B inhibits cell proliferation and induces apoptosis of pancreatic cancer cells. Pathol Res Pract 2019;215:1054-60. [Crossref] [PubMed]
25. Cai S, Wang J, Zeng W, et al. Lysine-specific histone demethylase 1B (LSD2/KDM1B) represses p53 expression to promote proliferation and inhibit apoptosis in colorectal cancer through LSD2-mediated H3K4me2 demethylation. Aging (Albany NY) 2020;12:14990-5001. [Crossref] [PubMed]
26. Li H, Durbin R. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 2009;25:1754-60. [Crossref] [PubMed]
27. McKenna A, Hanna M, Banks E, et al. The Genome Analysis Toolkit: a MapReduce framework for analyzing next-generation DNA sequencing data. Genome Res 2010;20:1297-303. [Crossref] [PubMed]
28. Wang K, Li M, Hakonarson H. ANNOVAR: functional annotation of genetic variants from high-throughput sequencing data. Nucleic Acids Res 2010;38:e164. [Crossref] [PubMed]
29. Ilango S, Paital B, Jayachandran P, et al. Epigenetic alterations in cancer. Front Biosci (Landmark Ed) 2020;25:1058-109. [Crossref] [PubMed]
30. Bonnici J, Schofield CJ, Kawamura A, Jmj C. Histone Demethylases: Beyond Histone Lysine Demethylation. Chimia (Aarau) 2026;80:138-44. [Crossref] [PubMed]
31. Wang A, Qi D, Ma Y, et al. Histone lysine demethylases in breast cancer: molecular mechanisms, biological functions, and therapeutic intervention. Mol Cancer 2025;25:23. [Crossref] [PubMed]
32. McKay SC, Louie BE, Molena D, et al. Tumor differentiation impacts response to neoadjuvant therapy and survival in patients with esophageal adenocarcinoma. J Thorac Cardiovasc Surg 2024;167:1943-50. [Crossref] [PubMed]
33. Ryu DG, Kim K, Liu H, et al. Clinical Features and Prognosis of Cervical Esophageal Cancer. J Clin Med 2025;14:3803. [Crossref] [PubMed]
34. Huang A, Jiang X, Qi Y, et al. PRDM1: a useful indicator of differentiation and prognosis in esophageal squamous cell carcinoma. Diagn Pathol 2026;21:33. [Crossref] [PubMed]
35. Hayashi M, Abe M, Fujita T, et al. Prognostic difference of esophageal squamous cell carcinoma based on tumor differentiation: a single center retrospective study. Scand J Gastroenterol 2025;60:983-91. [Crossref] [PubMed]
36. Guo X, Qin L, Yang R, et al. Survival outcomes and prognostic factors of early-onset and late-onset metastatic esophageal cancer: a population-based study. Sci Rep 2025;15:23588. [Crossref] [PubMed]
37. Tong D, Tian Y, Zhou T, et al. Improving prediction performance of colon cancer prognosis based on the integration of clinical and multi-omics data. BMC Med Inform Decis Mak 2020;20:22. [Crossref] [PubMed]
38. Alanni R, Hou J, Azzawi H, et al. Cancer adjuvant chemotherapy prediction model for non-small cell lung cancer. IET Syst Biol 2019;13:129-35. [Crossref] [PubMed]
39. Sridharan P, Ghosh M. Gene expression and agent-based modeling improve precision prognosis in breast cancer. Sci Rep 2025;15:17059. [Crossref] [PubMed]
40. Yuan L, Wen B, Li X, et al. Development and validation of a risk prediction model for overall survival in cervical cancer patients under 50: a prospective cohort study in southwest China. PeerJ 2026;14:e20509. [Crossref] [PubMed]
41. Shi Y, Zhou J, Jia K, et al. Molecular clustering and prognostic features based on integrated databases predict survival and immune status in patients with gastric cancer. Front Oncol 2025;15:1642911. [Crossref] [PubMed]
42. van den Boorn HG, Engelhardt EG, van Kleef J, et al. Prediction models for patients with esophageal or gastric cancer: A systematic review and meta-analysis. PLoS One 2018;13:e0192310. [Crossref] [PubMed]
43. Chen H, Huang X, Huang C, et al. The neoadjuvant esophageal score: a prognostic tool for predicting survival and postoperative complications in esophageal squamous cell carcinoma. Front Immunol 2025;16:1706548. [Crossref] [PubMed]
44. Natarajan SR, Krishnamoorthy R, Alshuniaber MA, et al. ABCE1 facilitates tumour progression via aerobic glycolysis and inhibits cell death in human colorectal cancer cells through the p53 signalling pathway. Sci Rep 2025;15:24674. [Crossref] [PubMed]
45. Di Y, Jing X, Hu K, et al. The c-MYC-WDR43 signalling axis promotes chemoresistance and tumour growth in colorectal cancer by inhibiting p53 activity. Drug Resist Updat 2023;66:100909. [Crossref] [PubMed]
46. Fu J, Liu Y, Wang X, et al. Role of DHX33 in c-Myc-induced cancers. Carcinogenesis 2017;38:649-60. [Crossref] [PubMed]
47. Yuan B, Wang X, Fan C, et al. DHX33 Transcriptionally Controls Genes Involved in the Cell Cycle. Mol Cell Biol 2016;36:2903-17. [Crossref] [PubMed]
48. Liu Z, Ye Y, Liu Y, et al. RNA Helicase DHX37 Facilitates Liver Cancer Progression by Cooperating with PLRG1 to Drive Superenhancer-Mediated Transcription of Cyclin D1. Cancer Res 2022;82:1937-52. [Crossref] [PubMed]
49. Wang Y, Li H, Zhao P, et al. Advances and challenges in immunotherapy for advanced esophageal squamous cell carcinoma. Front Immunol 2026;17:1739762. [Crossref] [PubMed]
50. Chuang CH, Guo JC, Kato K, et al. Exploring novel immunotherapy in advanced esophageal squamous cell carcinoma: Is targeting TIGIT an answer? Esophagus 2025;22:139-47. [Crossref] [PubMed]
51. Wang G, Li Z, Tan S, et al. Research progress on the compositional characteristics of the tumor immune microenvironment and immunopredictive models in esophageal squamous cell carcinoma. Cancer Biol Ther 2026;27:2641263. [Crossref] [PubMed]
52. Yu C, Zhang D, Chen C, et al. Neoadjuvant immunotherapy for resectable esophageal cancer: Research progress and clinical perspectives. Crit Rev Oncol Hematol 2026;221:105191. [Crossref] [PubMed]
53. Huang Z, Yang H. Upregulation of the long noncoding RNA ADPGK-AS1 promotes carcinogenesis and predicts poor prognosis in gastric cancer. Biochem Biophys Res Commun 2019;513:127-34. [Crossref] [PubMed]