FOXM1-associated melanoma stratification: integrated multi-omics and experimental validation reveal prognostic significance and therapeutic potential

Abstract

Background

Cutaneous melanoma is a highly aggressive malignancy characterized by marked biological heterogeneity and variable clinical outcomes. Reliable biomarkers for prognostic stratification and biological characterization remain limited. This study aimed to investigate the expression pattern, prognostic significance, immune associations, and functional relevance of Forkhead box M1 (FOXM1) in melanoma.

Methods

Multi-omics analyses were performed using TCGA, GTEx, GEO, and Human Protein Atlas datasets. Survival analysis, Cox regression, nomogram construction, functional enrichment, immune infiltration, genomic alteration, and drug sensitivity analyses were conducted. FOXM1 expression was further validated by immunohistochemistry. Functional experiments, including qRT-PCR, Western blotting, CCK-8 proliferation assays, and Transwell migration and invasion assays, were performed following FOXM1 knockdown in A375 melanoma cells.

Results

FOXM1 was significantly overexpressed in melanoma and independently associated with poor overall survival. A FOXM1-based prognostic model demonstrated satisfactory predictive performance. Functional analyses indicated that FOXM1 was primarily involved in cell-cycle progression and mitotic regulation. High FOXM1 expression was associated with an immunosuppressive microenvironment characterized by reduced cytotoxic immune infiltration and altered immune scores. Genomic analysis suggested copy-number amplification as a major contributor to FOXM1 overexpression. In vitro experiments confirmed efficient FOXM1 silencing and demonstrated that FOXM1 knockdown significantly inhibited melanoma cell proliferation, migration, and invasion.

Conclusions

FOXM1 is a prognostically relevant and immune-associated biomarker in melanoma. Its overexpression is associated with aggressive tumor behavior, and functional inhibition suppresses malignant phenotypes, highlighting its potential value for risk stratification and therapeutic targeting.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12885-026-16465-6.

Background

Cutaneous melanoma is a malignant tumor originating from pigment-producing melanocytes located in the basal layer of the epidermis. Although melanoma accounts for a relatively small proportion of all skin cancers, it is responsible for the majority of skin cancer–related deaths. As a disease with a high potential for cure when detected early, timely and accurate diagnosis remains critical for achieving optimal therapeutic outcomes. Globally, the burden of melanoma remains substantial. According to the GLOBOCAN 2022 estimates, melanoma of the skin accounted for 331,647 new cases and 58,645 deaths in 2022, ranking 17th among all cancers in incidence [1]. In the United States, the incidence of melanoma has increased steadily over the past decades, rising from 8.8 per 100,000 in 1975 to 28.42 per 100,000 in 2022, and an estimated 104,960 new cases and 8,430 deaths are projected for 2025 [2].

The development of melanoma results from the interplay between environmental factors and genetic susceptibility. Approximately 90% of cutaneous melanomas are attributable to ultraviolet radiation exposure-the most significant modifiable risk factor. Other established risks include a fair skin phenotype, history of sunburn, presence of numerous or atypical nevi, and family history [3]. Clinically, melanoma is categorized into four major subtypes: superficial spreading melanoma (≈ 70% of cases, linked to intermittent sun exposure), lentigo maligna, nodular, and acral melanoma. Notably, while acral melanoma accounts for only ~ 1% of total cases and is rare in White populations, it occurs with disproportionate frequency in non-White populations [3, 4].

Taking the United States as an example, melanoma is the fifth most common cancer in the country [3], melanoma has benefited substantially from advances in targeted therapy and immune checkpoint inhibitors over the past decades, leading to an annual reduction in mortality of approximately 1.4% to 6.1%. However, its incidence has not been effectively controlled [2, 5, 6]. This dissociation highlights persistent unmet needs in early detection, prevention, and mechanistic elucidation of therapeutic resistance in advanced melanoma.

With respect to treatment strategies, the management of melanoma has evolved from reliance on surgical excision alone to a multidisciplinary, integrated approach. For localized cutaneous melanoma, surgical excision remains the cornerstone of disease control and potential cure. Randomized controlled trials have demonstrated that, in high-risk melanoma with a Breslow thickness greater than 2 mm, a 2-cm excision margin is not inferior to wider margins with respect to overall survival and melanoma-specific survival [7, 8]. Although local recurrence is generally uncommon, recurrence correlates with the completeness of primary excision and sentinel lymph node status, underscoring the importance of sentinel lymph node biopsy (SLNB) in staging and prognostic assessment [9, 10]. Furthermore, the clinical application of SLNB, its relationship with different surgical approaches, and the potential role of de-escalated strategies in desmoplastic melanoma and melanoma in situ remain key issues requiring further investigation [11–13].

In terms of systemic therapy, immune checkpoint inhibitors and molecular targeted therapies have markedly improved long-term outcomes in patients with melanoma and have extended treatment strategies into the perioperative setting and longitudinal disease management. Anti–PD-1 monoclonal antibodies constitute the backbone of immunotherapy and are characterized by durable treatment responses, whereas combined BRAF/MEK inhibitor regimens targeting BRAFV600 mutations offer rapid onset of action and high objective response rates, although the development of resistance limits their long-term efficacy [14–17]. With the forward shift of treatment strategies, neoadjuvant and adjuvant therapies have shown the potential to reduce postoperative recurrence in selected high-risk patients; however, heterogeneity in treatment benefit and concerns regarding toxicity burden warrant careful evaluation [14, 15, 18–20]. In contrast, the role of conventional chemotherapy has markedly diminished in the current therapeutic landscape and is primarily reserved for salvage treatment following failure of immunotherapy and targeted therapy or for investigational purposes [16, 21]. Meanwhile, combined or sequential strategies integrating immunotherapy and targeted therapy have emerged as active areas of research, yet the optimal timing, treatment sequence, and appropriate patient selection remain to be defined by further clinical evidence [17, 20, 22].

Recent studies have demonstrated that intrinsic transcriptional regulatory networks within tumor cells play critical roles in melanoma proliferation, invasion, and resistance to therapy, with aberrantly activated transcription factors serving as key regulatory nodes linking genetic alterations to malignant phenotypic progression. Epithelial–mesenchymal transition (EMT), a central biological process underlying tumor invasion and metastasis, can markedly enhance tumor cell migratory and invasive capacities as well as therapeutic resistance, and has been demonstrated to be closely associated with multiple types of cancers.

At the molecular level, Forkhead box M1 (FOXM1) is an important member of the Forkhead transcription factor family. As a core transcription factor, FOXM1 regulates cell cycle progression and mitotic processes through transcriptional control of G2/M phase–related genes [23, 24]. In tumor biology, FOXM1 has been implicated in the progression of multiple solid malignancies. Aberrant activation of FOXM1 not only promotes sustained tumor cell proliferation but also contributes to malignant phenotypes such as invasion, metastasis, therapeutic resistance, and remodeling of the tumor microenvironment. In addition, recent studies suggest that FOXM1 may participate in tumor progression through diverse transcriptional regulatory mechanisms, including the formation of transcriptional condensates [25]. FOXM1 has also been reported to form positive feedback regulatory loops with multiple key signaling pathways, such as Nuclear factor kappa B (NF-κB), thereby contributing to cellular senescence escape, metabolic reprogramming, and tumor progression [26–28].

Notably, FOXM1 has been recognized as an important node within EMT regulatory networks. Previous studies have demonstrated that FOXM1 can promote the transition of tumor cells toward more invasive phenotypes by regulating EMT-related transcriptional programs in multiple solid tumors [29], suggesting a potential role in tumor invasion and metastasis. However, its specific function and clinical significance in melanoma remain to be systematically investigated.

In the context of melanoma management entering an era of precision stratification and individualized treatment, the present study aimed to investigate the biological functions of FOXM1 in melanoma progression and to evaluate its potential clinical value as a diagnostic and prognostic biomarker.

Materials and methods

Gene screening and identification

RNA sequencing data and clinical information for SKCM were obtained from The Cancer Genome Atlas (TCGA; https://portal.gdc.cancer.gov). Univariate Cox proportional hazards regression analysis was performed using the survival [30] and survminer [31] R packages to identify prognosis-related genes associated with overall survival (OS). Genes meeting the criteria of Hazard ratio (HR) > 1 and Padj < 0.05 were retained, resulting in 358 candidate prognostic genes. A curated set of 2,164 cancer-related genes was retrieved from the Cancer Genetics Web database, and an EMT-associated gene set comprising 1,184 genes was obtained from the dbEMT 2.0 database. Venn diagram analysis was then conducted to determine the overlapping and unique gene components between the prognostic gene set and the aforementioned cancer- and EMT-related gene sets. Visualization of Venn diagram results was performed using the ggplot2 [32] and Venn Diagram R packages.

Intersection genes were extracted from a unified TCGA–GTEx RNA-seq dataset processed through the Toil pipeline in UCSC XENA (https://xenabrowser.net/datapages), including GTEx normal samples (n = 812), TCGA adjacent normal samples (n = 1), and TCGA tumor samples (n = 469), all represented in TPM format. SKCM tumor samples from TCGA and corresponding normal tissues from GTEx were selected for downstream analyses. Heatmaps were generated using the ComplexHeatmap R package [33].

Expression analysis

Using the GEPIA2.0 database (http://gepia2.cancer-pku.cn/#index), NES-format RNA-seq data from TCGA and GTEx were used to analyze the mRNA expression levels of the intersecting genes and FOXM1 in SKCM, as well as to perform pan-cancer expression analyses. Immunohistochemical staining results of these genes in normal and tumor tissues were obtained from The Human Protein Atlas (HPA; https://www.proteinatlas.org/). HPA immunofluorescence staining images were used to display the expression and localization of FOXM1 in human cancer cell lines. Based on TCGA pan-cancer RNA-seq data and the melanoma-related GEO dataset GSE46517, FOXM1 expression differences between SKCM tissues and normal tissues were further analyzed. Chromosomal localization analysis was conducted using the circlize R package (version 0.4.15) with genome build v2023.6 [Homo sapiens; T2T-CHM13v2.0].

Clinical analysis

Gene mRNA expression data and clinical information for melanoma tumor tissues and paired normal tissues were obtained from the TCGA database. Receiver operating characteristic (ROC) curves were generated using the ‘pROC’ R package (version 1.17.0.1) in R (version 4.2.1), and visualization was performed using the ggplot2 package. The area under the curve (AUC), sensitivity, specificity (1-specificity), and confidence interval (CI) were recorded as key indicators. The survival package was used to perform proportional hazards assumption testing and to fit survival regression models, and the results were visualized using the survminer and ggplot2 packages. To further investigate the differences in FOXM1 expression across melanoma patients with different survival events (OS, DSS, and PFI), statistical analysis was performed using the Mann–Whitney U test (Wilcoxon rank sum test) based on TCGA data, and visualization was conducted with the ggplot2 package.

To construct an OS prognostic model for melanoma, a nomogram was developed using FOXM1 expression, pathologic stage, melanoma ulceration, melanoma Clark level, and Breslow depth as variables, and calibration curves for 1-, 3-, and 5-year survival were used to evaluate OS prediction accuracy. Based on clinical data from TCGA [31], proportional hazards assumption testing and Cox regression analysis were performed using the survival package. The predictive performance of the SKCM prognostic model was evaluated using ROC curves. Construction of the model was completed using the survival, regplot, and rms packages in R.

Functional enrichment analysis and protein interaction analysis

To investigate the function of the FOXM1 gene, its differentially expressed gene set was screened and subjected to Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), and GSEA enrichment analyses. Single-gene differential expression analysis was performed using the DESeq2 package based on the raw counts matrix, and normalization was conducted using the variance stabilizing transformations (VST) method provided by DESeq2. Differential expression results meeting P < 0.05 and |logFC| > 1 were visualized using volcano plots generated with the ggplot2 package. Based on the single-gene differential expression dataset, GO and KEGG enrichment analyses were performed with a screening threshold of P < 0.01, and genes with their logFC values were extracted for GSEA analysis. GSEA was performed using thresholds of P < 0.05 and q < 0.25, and the top five most relevant pathways were selected according to normalized enrichment score (NES) values. All analyses were implemented using the “cluster Profiler” package in R, and results were visualized using the ggplot2 package [34]. Functional Network (FunMap website) Gene Neighborhood module analysis was used to examine the top 50 genes associated with FOXM1, and Dense Module analysis of FOXM1 cliques was performed using the clique enumeration (ICE) algorithm. Edge weights in the protein interaction networks were quantified based on the STRING combined_score, which reflects the confidence of protein–protein interactions.

Epigenetic modification, gene mutation, and drug sensitivity analysis

SangerBox3.0 was used to analyze m¹A, m⁵C, and m⁶A mRNA modifications in TCGA pan-cancer tissues. Gene mutation profiles of FOXM1 in pan-cancer were analyzed based on TCGA data via the cBioPortal platform (https://www.cbioportal.org/). Drug sensitivity analysis of FOXM1 was performed using the GSCA database (https://guolab.wchscu.cn/GSCA/#/), integrating data from GDSC (https://www.cancerrxgene.org/) and CTRP (http://portals.broadinstitute.org/ctrp/).

Immune cell infiltration analysis

Based on the 22 immune cell signature gene sets provided by the CIBERSORTx website (https://cibersortx.stanford.edu/ ) [35], the CIBERSORT.R script algorithm was applied to transcriptome data to obtain sample-specific immune cell infiltration profiles. During data processing, visualization was performed using the ggplot2 (version 3.4.4) and ggalluvial (version 0.12.3) packages in R. Immune cell proportion distributions of FOXM1-associated subgroups were calculated, and stacked bar plots were generated using ggplot2 to display dynamic changes in immune cell composition. Based on TCGA-SKCM data, immune infiltration was further quantified using the core CIBERSORT algorithm (CIBERSORT.R script analysis) [35, 36] and the ssGSEA algorithm [37, 38] provided in the GSVA package (version 1.46.0), utilizing immune cell marker sets from the CIBERSORTx website and the Immunity publication (Bindea, Gabriela, et al., 2013). Spearman correlation analysis was performed between FOXM1 expression and the immune infiltration matrix, and visualization was conducted using lollipop plots generated with ggplot2. Samples were stratified into high- and low-FOXM1 expression groups, and Wilcoxon rank sum tests were applied. Stromal and immune scores were calculated using the estimate package (version 1.0.13) in R (Yoshihara, Kosuke, et al., 2013) [39]. Statistical analyses were performed using the stats and car packages, and visualization was completed using ggplot2.

Hematoxylin and eosin (H&E) staining

Clinical melanoma tissue specimens diagnosed by the Department of Pathology at Shandong Provincial Hospital were used for histopathological evaluation. H&E staining was performed as follows: paraffin-embedded tissue sections were deparaffinized in xylene and rehydrated through a graded ethanol series to distilled water. After washing with 1× PBS, sections were stained with hematoxylin for 5 min, differentiated briefly, blued, and counterstained with eosin for 2 min. Subsequently, sections were dehydrated through graded ethanol, cleared in xylene, and mounted with neutral resin. All stained slides were air-dried and prepared for histological evaluation.

Immunohistochemical (IHC)

Clinical melanoma tissue specimens used for both H&E staining and IHC were obtained from the Department of Pathology at Shandong Provincial Hospital. The IHC cohort included four melanoma cases, comprising one acral lentiginous melanoma, one nodular melanoma, and two melanoma cases without documented histological subtype information in the original pathology reports. Corresponding adjacent normal skin tissues were used as controls. Paraffin-embedded tissue sections were deparaffinized, rehydrated, and subjected to antigen retrieval. Endogenous peroxidase activity was blocked with 3% hydrogen peroxide, followed by blocking of non-specific binding with 3% BSA or appropriate serum. Sections were incubated overnight at 4 °C with an anti-FOXM1 antibody (rabbit monoclonal; Cell Signaling Technology, #20459; 1:600), followed by incubation with HRP-conjugated secondary antibodies. Immunoreactivity was visualized using DAB substrate, and nuclei were counterstained with hematoxylin. Slides were then dehydrated, cleared, mounted, and examined under a light microscope. FOXM1 protein expression was semi-quantitatively evaluated using the H-score method. Staining intensity was graded as 0 (negative), 1 (weak), 2 (moderate), or 3 (strong), and H-scores were calculated as follows: H-score = (% weakly stained cells × 1) + (% moderately stained cells × 2) + (% strongly stained cells × 3), yielding a final score ranging from 0 to 300. All slides were independently evaluated by two experienced pathologists who were blinded to the patients’ clinical information and outcome data during the scoring process.

Cell culture and siRNA transfection

The human melanoma cell line A375 was purchased from Procell Life Science & Technology Co., Ltd. (Wuhan, China; CL-0014) and authenticated by short tandem repeat (STR) profiling. Cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin–streptomycin at 37 °C in a humidified incubator containing 5% CO₂.

For FOXM1 knockdown, small interfering RNA (siRNA)-mediated transient transfection was performed. A375 cells were divided into three groups: Control, si-NC, and si-FOXM1. Transfection was carried out using Lipofectamine 2000 (KEL Biotech, China; KC-112) according to the manufacturer’s instructions. The final siRNA concentration was 50 nM. After 48 h of transfection, cells were harvested for subsequent experiments. The specific siRNA sequences and qRT-PCR primer sequences used in this study are listed in Table 1. Three candidate FOXM1-targeting siRNAs were initially evaluated by qRT-PCR, and the siRNA exhibiting the highest silencing efficiency was selected for subsequent experiments.

Table 1.

Sequences of siRNAs and qRT-PCR primers used in this study

Category	Target	Sequence Type	Sequence (5′→3′)
siRNA	si-NC	Sense	UUCUCCGAACGUGUCACGUTT
		Antisense	ACGUGACACGUUCGGAGAATT
	si-FOXM1	Sense	GCCCAACAGGAGUCUAAUCAA
		Antisense	UUGAUUAGACUCCUGUUGGG
qRT-PCR Primer	FOXM1	Forward	GCTTGCCAGAGTCCTTTTTGC
		Reverse	CCACCTGAGTTCTCGTCAATGC
	GAPDH	Forward	AGAAGGCTGGGGCTCATTTG
		Reverse	AGGGGCCATCCACAGTCTTC

The table summarizes the sequences of small interfering RNAs (siRNAs) used for FOXM1 silencing and the primer sequences used for quantitative real-time PCR (qRT-PCR) validation of FOXM1 expression. GAPDH served as the endogenous control for gene expression normalization

Quantitative real-time PCR (qRT-PCR)

Total RNA was extracted from cells of each group and reverse-transcribed into cDNA according to the manufacturers’ instructions. Quantitative real-time PCR (qRT-PCR) was performed using a SYBR Green-based detection system to evaluate FOXM1 mRNA expression and verify knockdown efficiency. The PCR amplification conditions were as follows: initial denaturation at 95 °C for 30 s, followed by 40 cycles of 95 °C for 5 s and 60 °C for 30 s. GAPDH was used as the internal reference gene, and relative expression levels were calculated using the 2^^−ΔΔCt method.

Western blot analysis

Western blotting was performed to verify FOXM1 knockdown efficiency. Forty-eight hours post-transfection, A375 cells were harvested and lysed in RIPA buffer for total protein extraction. Protein concentrations were determined using a BCA protein assay kit. Equal amounts of protein were separated on 10% SDS–PAGE gels and transferred onto PVDF membranes. After blocking with a rapid blocking buffer for 10–15 min at room temperature, membranes were incubated overnight at 4 °C with primary antibodies against FOXM1 (rabbit monoclonal; Cell Signaling Technology, #20459; 1:1000) and GAPDH as the loading control. Subsequently, the membranes were washed three times with TBST and incubated with HRP-conjugated secondary antibodies for 1 h at room temperature. Protein bands were visualized using an enhanced chemiluminescence (ECL) detection system and quantified using ImageJ software.

Cell proliferation assay

Cell proliferation was evaluated using a Cell Counting Kit-8 (CCK-8) assay kit (Beyotime Biotechnology, Shanghai, China). Transfected cells were seeded into 96-well plates at a density of 5 × 10³ cells per well. At the indicated time points (24, 48, 72, and 96 h), 10 µL of CCK-8 reagent was added to each well, followed by incubation at 37 °C for 1 h. Optical density (OD) values were measured at 450 nm using a microplate reader (Multiskan Sky, Thermo Fisher Scientific, USA).

Transwell migration and invasion assays

Cell migration and invasion were assessed using Falcon Transwell chambers with 8-µm pore-size membranes (Falcon, USA). For the invasion assay, the upper chambers were pre-coated with Matrigel matrix (BD Biocoat, USA), whereas uncoated chambers were used for the migration assay. Transfected cells (6 × 10⁴ cells/well) suspended in serum-free medium were seeded into the upper chamber, while 700 µL of medium containing 10% FBS was added to the lower chamber as a chemoattractant. After 24 h of incubation at 37 °C, non-migrated or non-invaded cells on the upper surface of the membrane were carefully removed. Cells on the lower surface were fixed with 4% paraformaldehyde for 30 min, washed, and stained with 0.1% crystal violet for 30 min. After rinsing three times with PBS, migrated or invaded cells were photographed and counted in randomly selected microscopic fields.

Results

Identification and prioritization of biologically relevant biomarkers in melanoma

To identify melanoma-associated biomarkers of prognostic relevance, we first screened 358 prognosis-related genes from The Cancer Genome Atlas (TCGA) dataset, retrieved 2,164 cancer-related genes from the Cancer Genetics Web database, and obtained 1,184 EMT-associated genes from the dbEMT2.0 resource. A Venn diagram integrating the three gene sets revealed nine intersecting candidates (Fig. 1a). Differential expression patterns of these genes in the skin cutaneous melanoma (SKCM) cohort were visualized using a co-expression heatmap (Fig. 1b). Based on GEPIA2.0 analyses (Fig. 1c-k), FOXM1 (P < 0.05), UHRF1 (P < 0.05), NME1 (P < 0.05), and GAB2 (P > 0.05) were upregulated in melanoma tissues, whereas MMP2 (P < 0.05), KIT (P < 0.05), SLC2A1 (P < 0.05), JAG2 (P < 0.05), and CREBBP (P > 0.05) were downregulated. Among these, FOXM1, UHRF1, and NME1 were prioritized for subsequent analyses.

Fig. 1.

Identification and differential expression analysis of EMT- and prognosis-associated candidate genes in melanoma. a Venn diagram showing the intersection of three gene sets derived from TCGA melanoma prognosis-associated genes (n = 358), Cancer Genetics Web cancer-related genes (n = 2,164), and dbEMT2.0 EMT-associated genes (n = 1,184). Nine overlapping genes (MMP2, FOXM1, KIT, UHRF1, GAB2, SLC2A1, NME1, JAG2, and CREBBP) were identified at the three-way intersection. b Heatmap illustrating differential expression patterns of the nine candidate genes in melanoma tumor versus normal samples within the SKCM cohort. c-k GEPIA2 analysis of the TCGA+GTEX cohort (tumor n = 461; normal n = 558) showing differential expression of nine intersecting EMT- and cancer-related candidate genes. FOXM1, UHRF1, and NME1 were significantly upregulated in melanoma, whereas MMP2, KIT, SLC2A1, and JAG2 were significantly downregulated. GAB2 and CREBBP showed no significant differences. FOXM1, UHRF1, and NME1 were prioritized for downstream analyses. *=p < 0.05

Protein-level expression patterns were further assessed using immunohistochemical (IHC) staining images from the Human Protein Atlas (HPA) database (Fig. 2a). Compared with normal skin tissues, melanoma tissues exhibited strong FOXM1 and NME1 positivity, indicating elevated protein expression levels. By contrast, UHRF1 showed weaker staining intensity and less pronounced differential expression. Based on these results, we subsequently evaluated the prognostic and diagnostic significance of FOXM1, UHRF1, and NME1 using mRNA expression profiles and associated clinical data from the TCGA-SKCM cohort.

Fig. 2.

Protein expression, prognostic significance, and diagnostic performance of FOXM1, UHRF1, and NME1 in melanoma. a Immunohistochemical staining images from the Human Protein Atlas (HPA) database showing protein expression of FOXM1, UHRF1, and NME1 in melanoma and normal skin tissues. FOXM1 and NME1 exhibited strong positive staining in melanoma tissues compared with normal skin, whereas UHRF1 showed relatively weak staining intensity. Kaplan–Meier survival analyses demonstrated that high expression of FOXM1, UHRF1, and NME1 was significantly associated with poorer overall survival (OS) (b–d), disease-specific survival (DSS) (e–g), and progression-free interval (PFI) (h–j) in the TCGA-SKCM cohort. Survival risk estimates indicated increased mortality and progression risks in high-expression groups (log-rank P < 0.05). Receiver operating characteristic (ROC) curve analyses showed the diagnostic performance of FOXM1 (k), UHRF1 (l), and NME1 (m) in distinguishing melanoma tissues from normal tissues. FOXM1 exhibited the highest diagnostic accuracy (AUC = 0.925), followed by NME1 (AUC = 0.918) and UHRF1 (AUC = 0.841). Number-at-risk tables are provided beneath each Kaplan–Meier survival curve

Kaplan–Meier survival analyses revealed that high expression of FOXM1, UHRF1, and NME1 was significantly associated with poorer overall survival (OS) (Fig. 2b–d). Specifically, patients with elevated FOXM1 expression exhibited a markedly increased risk of mortality compared with those with low expression (HR = 1.90, 95% CI: 1.45–2.51; P < 0.001; Fig. 2b). Similarly, high expression of UHRF1 (HR = 1.51, 95% CI: 1.14–1.98; P = 0.003; Fig. 2c) and NME1 (HR = 1.52, 95% CI: 1.16–2.00; P = 0.002; Fig. 2d) was also associated with reduced OS. Consistent with these findings, elevated expression of FOXM1, UHRF1, and NME1 was significantly associated with unfavorable disease-specific survival (DSS). High FOXM1 expression conferred an increased risk of disease-specific mortality (HR = 1.95, 95% CI: 1.45–2.61; P < 0.001; Fig. 2e), while similar associations were observed for UHRF1 (HR = 1.47, 95% CI: 1.10–1.97; P = 0.009; Fig. 2f) and NME1 (HR = 1.70, 95% CI: 1.27–2.26; P < 0.001; Fig. 2g). Moreover, high expression of FOXM1 was significantly correlated with an increased risk of disease progression, as reflected by shorter progression-free interval (PFI) (HR = 1.28, 95% CI: 1.02–1.60; P = 0.030; Fig. 2h). Similar associations were observed for UHRF1 (HR = 1.31, 95% CI: 1.05–1.64; P = 0.018; Fig. 2i) and NME1 (HR = 1.27, 95% CI: 1.02–1.59; P = 0.036; Fig. 2j). Collectively, these results indicate that elevated expression of FOXM1, UHRF1, and NME1 is closely associated with increased risks of disease progression and tumor-specific mortality in this cohort. ROC curve analyses were performed to evaluate the discriminatory ability of FOXM1, UHRF1, and NME1 between tumor and normal samples. FOXM1 exhibited strong discriminatory performance, with an AUC of 0.925 (95% CI: 0.909–0.940) (Fig. 2k). Similarly, NME1 also demonstrated high diagnostic accuracy, yielding an AUC of 0.918 (95% CI: 0.903–0.933) (Fig. 2m). UHRF1 showed moderate-to-high discriminatory capability, with an AUC of 0.841 (95% CI: 0.819–0.863) (Fig. 2l). These findings suggest that FOXM1, UHRF1, and NME1 may serve as potential biomarkers for distinguishing melanoma tissues from normal tissues. Univariate and multivariate Cox proportional hazards regression analyses (Table 2) were performed to evaluate the prognostic significance of FOXM1, UHRF1, and NME1 expressions. In univariate analysis, high expression of FOXM1 was significantly associated with increased risk of mortality (HR = 1.904, 95% CI: 1.445–2.508; P < 0.001), and similar associations were observed for UHRF1 (HR = 1.506, 95% CI: 1.145–1.981; P = 0.003) and NME1 (HR = 1.524, 95% CI: 1.164–1.996; P = 0.002). After adjustment for potential confounding clinical variables in multivariate analysis, FOXM1 remained an independent prognostic factor for overall survival (HR = 1.722, 95% CI: 1.243–2.386; P = 0.001). In contrast, the associations of UHRF1 (HR = 1.064, 95% CI: 0.770–1.471; P = 0.707) and NME1 (HR = 1.317, 95% CI: 0.996–1.743; P = 0.054) with survival were no longer statistically significant. Given that FOXM1 consistently exhibited significant prognostic relevance across multiple survival endpoints and remained an independent prognostic factor in multivariate analysis, subsequent analyses and experimental validation in this study were primarily focused on FOXM1.

Table 2.

Univariate and multivariate Cox regression analyses of three candidate genes

Characteristics	Total (N)	Univariate analysis Hazard ratio P value (95% CI)		Multivariate analysis Hazard ratio P value (95% CI)
FOXM1	457
Low	228	Reference	—	Reference	—
High	229	1.904(1.445–2.508)	< 0.001	1.722 (1.243–2.386)	0.001
UHRF1	457
Low	228	Reference	—	Reference	—
High	229	1.506(1.145–1.981)	0.003	1.064 (0.770–1.471)	0.707
NME1	457
Low	226	Reference	—	Reference	—
High	231	1.524(1.164–1.996)	0.002	1.317 (0.996–1.743)	0.054

HRs and 95% CIs were calculated using Cox proportional hazards models. High and low expression levels were stratified by cutoff as described in the Methods. HR > 1 indicates an increased risk

FOXM1 exhibits widespread upregulation across human cancers and consistent overexpression in melanoma

To systematically characterize the expression profile of FOXM1 across human malignancies, we first performed a pan-cancer analysis using transcriptomic data from TCGA. As shown in Fig. 3a, FOXM1 expression was significantly elevated in the majority of malignant tumors compared with matched normal tissues, including glioblastoma, breast cancer, colorectal cancer, lung cancer, hepatocellular carcinoma, and melanoma, among others. To clarify the genomic context of FOXM1, chromosomal localization analysis was conducted, revealing that FOXM1 is encoded by a single gene locus on chromosome 12 (Fig. 3b). This finding indicates a conserved genomic configuration and argues against the possibility that its widespread overexpression arises from multi-locus amplification or annotation artifacts.

Fig. 3.

Pan-cancer expression patterns, genomic localization, tissue-level validation, and subcellular distribution of FOXM1. Pan-cancer analysis of TCGA transcriptomes revealed that FOXM1 is broadly upregulated across multiple tumor types compared with matched normal tissues (a). Chromosomal mapping showed that FOXM1 is located on chromosome 12 (b). Cross-cohort expression analyses demonstrated significantly elevated FOXM1 levels in melanoma compared with normal skin in both TCGA+GTEx (c) and GEO (GSE46517) datasets (d). Immunofluorescence data from the HPA database showed that FOXM1 protein predominantly localizes to the nucleus (e), consistent with its function as a transcription factor. *p < 0.05, **p < 0.01, ***p < 0.001

Next, FOXM1 expression in melanoma was further evaluated using TCGA combined with GTEx datasets. As shown in Fig. 3c, FOXM1 transcript levels (log₂^{[TPM + 1]}) were markedly higher in tumor samples than in normal tissues (P < 0.001). Notably, FOXM1 exhibited a broader expression range in melanoma samples than in normal controls, indicating greater variability of FOXM1 expression among tumor samples. Consistent results were obtained in an independent dataset (GSE46517) from the Gene Expression Omnibus (GEO), in which FOXM1 expression was significantly increased in melanoma tissues relative to normal skin (Fig. 3d). These findings collectively demonstrate the reproducibility of FOXM1 overexpression across independent cohorts and sequencing platforms. Finally, subcellular localization of FOXM1 protein was examined using immunofluorescence images from the Human Protein Atlas (HPA). As shown in Fig. 3e, FOXM1 protein was predominantly localized to the nucleus, which is consistent with its known function as a transcription factor.

FOXM1-based nomogram integrating clinicopathological features predicts survival outcomes in melanoma

To evaluate the prognostic value of FOXM1 in melanoma, we constructed a multivariable survival prediction model incorporating FOXM1 expression and established clinicopathological parameters, including pathological stage, ulceration status, Clark level, and Breslow thickness. A nomogram was generated based on the Cox proportional hazards model to estimate 1-, 3-, and 5-year survival probabilities (Fig. 4a). In this model, advanced pathological stage (stage III–IV), presence of ulceration, higher Clark level (III–V), increased Breslow thickness (≥ 3 mm), and elevated FOXM1 expression collectively contributed to increased risk scores. Notably, FOXM1 contributed substantially to the nomogram and remained an independent prognostic factor in multivariable analysis, supporting its potential prognostic relevance in melanoma (Fig. 4a).

Fig. 4.

Nomogram-based prognostic modeling highlights the prognostic value of FOXM1 in melanoma. a A multivariable Cox nomogram integrating FOXM1 expression with pathological stage, ulceration, Clark level, and Breslow thickness predicted 1-, 3-, and 5-year survival. FOXM1 contributed substantially to the nomogram together with established clinicopathological variables. b Calibration plots showed high agreement between predicted and observed outcomes. c Time-dependent Receiver operating characteristic (ROC) curves demonstrated stable discrimination at 1, 3, and 5 years (AUC = 0.771, 0.777, 0.731). d–f Patients experiencing OS, DSS, or PFI events exhibited significantly higher FOXM1 expression, indicating its association with adverse clinical outcomes. *p < 0.05, **p < 0.01, ***p < 0.001

The model exhibited acceptable calibration performance across multiple time points. Calibration plots for 1-, 3-, and 5-year survival showed close agreement between predicted and observed outcomes (Fig. 4b), indicating good concordance and accurate survival estimation. Time-dependent ROC analyses further demonstrated reliable discriminatory performance of the FOXM1-based model. The AUC values for predicting 1-, 3-, and 5-year survival were 0.771, 0.777, and 0.731, respectively (Fig. 4c). These results indicate predictive acceptable performance of the model within the TCGA-SKCM cohort and suggest that it may serve as a useful tool for risk stratification of melanoma patients during both short- and long-term follow-up.

To further examine the clinical relevance of FOXM1 expression, patients were stratified according to survival events. FOXM1 expression was significantly higher in patients who experienced OS events compared with those who remained alive (P < 0.001; Fig. 4d). Similarly, FOXM1 expression was elevated in patients with DSS events compared with event-free patients (P < 0.001; Fig. 4e). In patients experiencing progression-free interval (PFI) events, FOXM1 expression was significantly higher (P < 0.05; Fig. 4f). Collectively, these findings indicate that high FOXM1 expression is consistently associated with unfavorable clinical outcomes across multiple survival endpoints in melanoma.

Functional enrichment analysis reveals FOXM1-associated transcriptional programs

To elucidate the biological processes associated with FOXM1-related transcriptional alterations in melanoma, GO and KEGG enrichment analyses were performed based on FOXM1-associated differentially expressed genes. GO Biological Process (BP) analysis (Fig. 5a) revealed significant enrichment in terms related to epidermal development, skin development, epidermal cell differentiation, keratinocyte differentiation, and keratinization, which represent core processes involved in epidermal homeostasis and tissue construction. In addition, immune-related processes such as humoral immune response, defense response to bacterium, leukocyte-mediated immunity, regulation of peptidase activity, and regulation of endopeptidase activity were also enriched, suggesting that FOXM1-associated transcriptional programs may concurrently participate in epithelial differentiation and immune regulation.

Fig. 5.

Functional enrichment analysis. a–c Results of Gene Ontology (GO) enrichment analysis for FOXM1-associated differentially expressed genes (DEGs). d Results of Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis. e Volcano plot showing differentially expressed genes between FOXM1-high and FOXM1-low melanoma samples. f–h negatively enriched pathways identified by GSEA. i–j positively enriched pathways identified by GSEA

Consistent with the BP findings, GO Cellular Component (CC) terms (Fig. 5b) were mainly enriched in the intermediate filament cytoskeleton, intermediate filaments, and keratin filaments, aligning with the enrichment of epidermal differentiation-related biological processes. Enrichment was also observed in cell–cell junctions, collagen-containing extracellular matrix, cornified envelope, and the external side of the plasma membrane, indicating potential involvement in cellular structural organization and cell–cell interactions. GO Molecular Function (MF) analysis (Fig. 5c) revealed significant enrichment in passive transmembrane transporter activity, channel activity, ion channel activity, and gated channel activity. Furthermore, receptor ligand activity, signaling receptor activator activity, and enzyme/peptidase inhibitor activity were enriched, indicating participation in signaling transduction and enzymatic regulation.

KEGG pathway analysis (Fig. 5d) demonstrated enrichment in Neuroactive ligand–receptor interaction, Cytokine–cytokine receptor interaction, Staphylococcus aureus infection, Estrogen signaling pathway, Viral protein interaction with cytokine and cytokine receptor, Salivary secretion, Hematopoietic cell lineage, Complement and coagulation cascades, Arachidonic acid metabolism, and Primary immunodeficiency, suggesting that FOXM1-related transcriptional alterations may be involved in immune signaling and inflammatory processes.

Differential gene expression analysis between the FOXM1 high- and low-expression groups was visualized by a volcano plot (Fig. 5e). To further clarify the biological programs underlying these alterations, Gene Set Enrichment Analysis (GSEA) was performed using the screened differentially expressed gene set. The top five enriched pathways showed notable negative enrichment of epidermal differentiation–associated programs (Fig. 5f–h). Specifically, the Reactome Keratinization pathway exhibited significant negative enrichment (NES = − 4.209, Padj < 0.001, False discovery rate (FDR) < 0.001; Fig. 5f), as did Reactome Formation of the Cornified Envelope (NES = − 4.162, Padj < 0.001, FDR < 0.001; Fig. 5g). The Reactome Developmental Biology pathway also showed marked negative enrichment (NES = − 3.586, Padj < 0.001, FDR < 0.001; Fig. 5h). Conversely, cell cycle–related pathways were strongly positively enriched (Fig. 5i–j), including Reactome Cell Cycle Mitotic (NES = 3.112, Padj < 0.001, FDR < 0.001; Fig. 5i) and Reactome Cell Cycle (NES = 2.869, Padj < 0.001, FDR < 0.001; Fig. 5j), indicating coordinated upregulation of genes associated with mitotic progression and cell cycle regulation. Collectively, these findings indicate a reciprocal transcriptional pattern associated with FOXM1 expression, characterized by activation of cell cycle–related programs accompanied by suppression of keratinization and epidermal differentiation signatures.

Functional network analysis revealed a FOXM1-centered gene neighborhood enriched for cell cycle–associated genes. A highly interconnected gene neighborhood positioned around FOXM1 exhibited coordinated expression in tumor samples (Supplementary Fig. 1a). This neighborhood contained numerous canonical regulators of cell cycle progression, mitotic control, and DNA replication, including CDK1, CCNB1, CCNB2, AURKB, TOP2A, RRM2, BUB1, CENPF, and MKI67. Visualization of differential gene expression showed that a substantial fraction of FOXM1-associated genes were upregulated in tumor tissues relative to normal controls, indicating coordinated transcriptional activation within this neighborhood. In addition, interaction edges were weighted according to STRING combined scores, enabling direct visualization of interaction confidence and network connectivity strength.

To further dissect the internal structure of the FOXM1-associated network, dense submodules were identified using clique enumeration. Clique ID 22 represented a highly interconnected subnetwork containing FOXM1 together with genes involved in mitotic entry, chromosome segregation, and spindle dynamics (Supplementary Fig. 1b), suggesting a functionally coherent unit associated with mitotic regulation. Clique ID 58 comprised FOXM1 and genes implicated in DNA replication and replication stress responses, including members of the MCM complex, CHEK1, TOP2A, and AURKB (Supplementary Fig. 1c), highlighting functional associations between FOXM1 and genome duplication pathways. In addition, Clique ID 139 constituted a distinct FOXM1-containing module enriched for genes related to cell cycle control and checkpoint regulation, such as CDC25C, CCNB2, HASPIN, and multiple kinesin family members (Supplementary Fig. 1d). Although partially overlapping with other FOXM1-associated submodules, this cluster maintained a compact and dense organizational structure, indicative of functional coherence.

GO enrichment analysis of FOXM1-associated cliques identified by FunMap demonstrated consistent enrichment for cell cycle–related functions across BP, MF, and CC categories (Supplementary Table 1). In BP, Clique ID 22, 58, and 139 were significantly enriched for nuclear division, chromosome segregation, and mitotic nuclear division, respectively (all FDR < 10⁻⁸). MF analysis showed enrichment for microtubule binding (Clique 22), chromatin binding (Clique 139), and DNA polymerase binding (Clique 58). CC terms localized these cliques to centromeric regions, condensed chromosomes, and the replication fork protection complex. Together, these findings demonstrate that FOXM1 is embedded within multiple tightly connected functional gene modules predominantly enriched for mitotic progression, chromosome dynamics, and DNA replication–related processes, reflecting a coordinated transcriptional network associated with FOXM1 in melanoma.

Association of FOXM1 expression with immune cell composition and infiltration in the tumor microenvironment

To characterize the relationship between FOXM1 expression and the tumor immune microenvironment (TIME), immune cell composition was first inferred using CIBERSORTx, and samples were stratified into FOXM1-low and FOXM1-high groups. The relative proportions of 22 immune cell subsets across individual samples revealed marked inter-sample heterogeneity in both groups (Fig. 6a). Nevertheless, distinct differences in overall immune infiltration patterns were observed. While macrophages, T-cell subsets, and dendritic cells constituted the major components of the immune infiltrate in both subgroups, FOXM1-high tumors exhibited altered distributions of multiple immune cell populations compared with FOXM1-low tumors. In particular, differences were observed in the relative proportions of macrophage subsets and T-lymphocyte populations, as well as in the balance between activated and resting immune cell states.

Fig. 6.

Immune infiltration patterns associated with FOXM1 expression in melanoma. a Relative proportions of 22 immune cell subsets estimated by CIBERSORTx in FOXM1-low and FOXM1-high melanoma samples. b Correlations between FOXM1 expression and CIBERSORT-inferred immune cell proportions. c Correlations between FOXM1 expression and ssGSEA-derived immune cell enrichment scores. d–f Correlations between FOXM1 expression and ESTIMATE score, immune score, and stromal score, respectively. g–i Comparison of ESTIMATE score, immune score, and stromal score between FOXM1-low and FOXM1-high groups

To further evaluate these alterations, two complementary analytical approaches were employed: CIBERSORT for immune cell proportion estimation and single-sample gene set enrichment analysis (ssGSEA) for immune cell enrichment assessment. Using CIBERSORT, FOXM1 expression showed significant correlations with multiple immune cell subsets (Fig. 6b). Specifically, FOXM1 expression was positively correlated with resting NK cells and M0 macrophages, whereas significant negative correlations were observed with key T-cell populations, including CD8⁺ T cells and regulatory T cells (Tregs) (all P < 0.05). Consistent with these findings, ssGSEA-based immune profiling demonstrated a strong positive correlation between FOXM1 expression and Th2 cell signatures (P < 0.001), but significant negative correlations with multiple effector immune populations, including cytotoxic T cells, CD8⁺ T cells, Th1 cells, Th17 cells, dendritic cells, and NK cell subsets (all P < 0.05; Fig. 6c). The concordance between CIBERSORT- and ssGSEA-based analyses supports the robustness of the observed associations between FOXM1 expression and immune infiltration patterns.

To assess the overall immune and stromal context, ESTIMATE analysis was performed. FOXM1 expression was significantly and negatively correlated with the ESTIMATE score, immune score, and stromal score (all P < 0.001; Fig. 6d–f). Consistently, stratified analysis demonstrated that FOXM1-high tumors exhibited significantly lower ESTIMATE, immune, and stromal scores than FOXM1-low tumors (Fig. 6g–i).

To further characterize the association between FOXM1 expression and cytotoxic immune activity, additional analyses were performed focusing on CD8⁺ T-cell enrichment and representative immune-related markers (Supplementary Figure S2). FOXM1 expression was significantly negatively correlated with CD8⁺ T-cell enrichment estimated by both ssGSEA and CIBERSORT algorithms (all P < 0.001), and CD8⁺ T-cell enrichment scores were significantly lower in the FOXM1-high subgroup than in the FOXM1-low subgroup. Furthermore, FOXM1 expression showed significant negative correlations with CD8A, CD8B, PDCD1, LAG3, and TIGIT expression levels, whereas weaker or non-significant associations were observed for GZMB, CTLA4, and CD274. These findings further support an association between elevated FOXM1 expression and reduced cytotoxic T-cell–related immune features in melanoma. (Supplementary Figure S2).

Collectively, these multi-dimensional analyses indicate that elevated FOXM1 expression is associated with reduced immune and stromal infiltration, lower enrichment of cytotoxic immune populations, and distinct immune-related transcriptional features within the melanoma tumor microenvironment. These findings suggest that FOXM1 expression is closely associated with the immune landscape of melanoma; however, the underlying biological mechanisms require further experimental validation.

To investigate the association between FOXM1 expression and RNA modification regulators, correlation analysis was performed across a panel of genes involved in mRNA modification, including writers, readers, and erasers of N1-methyladenosine (m¹A), 5-methylcytosine (m⁵C), and N6-methyladenosine (m⁶A) modifications. The correlation matrix, visualized by heatmap (Fig. 7a), revealed complex correlation patterns between FOXM1 and multiple RNA modification regulators. Notably, FOXM1 expression showed significant correlations with multiple m⁶A-, m⁵C-, and m¹A-related regulators. Collectively, these findings suggest that FOXM1 may be associated with post-transcriptional regulatory processes involving RNA modifications such as m⁶A, m⁵C, and m¹A (Fig. 7a).

Fig. 7.

Molecular characteristics and pharmacogenomic associations of FOXM1 in melanoma. a Correlation heatmap showing associations between FOXM1 expression and mRNA modification regulators, including m¹A, m⁵C, and m⁶A writers, readers, and erasers. b Integrated analysis of FOXM1 genomic alterations in TCGA, including mutations, structural variants, and copy-number changes, revealed a low overall alteration frequency (~ 3%). c Distribution and frequency of FOXM1 genomic alterations in melanoma, showing amplification as the predominant alteration type. d CTRP-based pharmacogenomic analysis showing associations between FOXM1 expression and responses to multiple agents. e GDSC-based pharmacogenomic analysis showing associations between FOXM1 expression and responses to multiple agents

To further characterize the genomic features of FOXM1 in melanoma, somatic mutations, structural variants, and copy number alterations (CNA) were analyzed using TCGA data. As shown in Fig. 7b, FOXM1 exhibited a relatively low somatic mutation frequency, with approximately 3% of melanoma samples harboring detectable genetic alterations. Copy number analysis indicated that FOXM1 amplification represented the predominant form of genomic alteration in melanoma (Fig. 7c). Tumors with FOXM1 amplification frequently exhibited additional co-occurring oncogenic alterations, indicating potential genomic co-occurrence patterns. These findings suggest that copy-number amplification may contribute, at least in part, to the elevated FOXM1 expression observed in melanoma.

Drug sensitivity analysis of FOXM1

To assess the therapeutic relevance of FOXM1, we analyzed its association with drug sensitivity using the CTRP and GDSC pharmacogenomic datasets. In CTRP, FOXM1 expression was significantly associated with responses to multiple agents, including bortezomib, selumetinib, and zebularine (Fig. 7d), based on FDR-adjusted correlations, whereas the association with trametinib was weaker, suggesting compound-specific heterogeneity within MAPK pathway dependency. These findings were independently supported by GDSC analyses (Fig. 7e), in which FOXM1 expression showed significant associations with responses to selumetinib and trametinib, as well as additional cell cycle– and apoptosis-related agents such as RDEA119 and Navitoclax. Together, these results indicate that FOXM1 expression is associated with distinct pharmacogenomic response patterns in melanoma. These findings may provide clues for future studies investigating the relationship between FOXM1 expression and therapeutic response.

Histopathological features distinguish tumor tissues from adjacent normal counterparts

H&E staining revealed marked histopathological differences between tumor tissues and matched normal samples. Tumor sections exhibited increased cellular density (Fig. 8a–b), disrupted tissue architecture, and irregular cellular arrangements compared with adjacent normal tissues. In several tumor samples, reduced stromal integrity was also observed. In contrast, normal tissues displayed preserved tissue architecture with well-organized stromal and epithelial components. Normal samples showed relatively low cellular density, intact extracellular matrix structure, and clear delineation of tissue layers, indicative of non-neoplastic histology. Notably, inter-sample heterogeneity was evident among tumor specimens, with variations in cellular density and stromal composition, reflecting intrinsic morphological diversity within tumor tissues. Collectively, these histopathological findings confirm clear structural and cellular differences between tumor and normal tissues, providing a morphological basis for subsequent molecular and functional analyses.

Fig. 8.

Histopathological and immunohistochemical validation of FOXM1 expression in clinical melanoma specimens. (a) Representative H&E staining images from patients with melanoma. (b) Representative H&E staining images showing additional histopathological fields from the same melanoma patients in (a). c–f Representative immunohistochemical (IHC) staining of FOXM1 in melanoma tissues, showing strong nuclear and cytoplasmic FOXM1 immunoreactivity in tumor cells. g–j Representative IHC staining of FOXM1 in the corresponding adjacent normal tissues, demonstrating weak or absent FOXM1 expression. For (c–j), low-magnification overviews are presented on the left, with corresponding high-magnification cropped images (20×) shown on the right to illustrate FOXM1 staining distribution and intensity

FOXM1 protein is highly expressed in clinical melanoma tissue specimens

IHC analysis revealed markedly elevated FOXM1 protein levels in melanoma tissues compared with matched normal samples. Across independent patients (Fig. 8c-f), melanoma specimens exhibited strong nuclear and cytoplasmic FOXM1 staining with dense accumulation in tumor cells. In contrast, FOXM1 staining was largely absent or weakly detectable in normal skin tissues from the same patients (Fig. 8g-j), demonstrating minimal background expression in non-neoplastic cells. Semi-quantitative H-score evaluation further supported these findings. FOXM1 H-scores ranged from 196.5 to 285.9 in melanoma tissues, whereas corresponding adjacent normal tissues exhibited substantially lower H-scores ranging from 0.5 to 8.7. Notably, inter-patient variability was observed among melanoma lesions, with differences in staining intensity and distribution patterns, reflecting histopathological heterogeneity within melanoma. Nevertheless, the consistent contrast between tumor and normal tissues indicates that FOXM1 is preferentially expressed in malignant cells at the protein level. These findings provide protein-level validation of the transcriptomic analyses and support the association between FOXM1 overexpression and melanoma progression.

FOXM1 knockdown suppresses proliferation, migration, and invasion of melanoma cells

To investigate the biological role of FOXM1 in melanoma, A375 cells were transfected with three independent FOXM1-targeting siRNAs. qRT-PCR analysis demonstrated that all three siRNAs significantly reduced FOXM1 mRNA expression compared with the si-NC group (all P < 0.001), with si-FOXM1-1 exhibiting the highest silencing efficiency and therefore being selected for subsequent functional experiments (Fig. 9a). Consistently, Western blot analysis confirmed a marked reduction in FOXM1 protein levels in the si-FOXM1 group compared with both the Control and si-NC groups, with GAPDH serving as the loading control (Fig. 9b).

Fig. 9.

FOXM1 knockdown inhibits proliferation, migration, and invasion of A375 melanoma cells. a Relative FOXM1 mRNA expression levels in A375 cells following transfection with three independent FOXM1-targeting siRNAs (si-FOXM1-1, si-FOXM1-2, and si-FOXM1-3), as determined by RT-qPCR. b Western blot analysis confirming FOXM1 protein knockdown in A375 cells transfected with si-FOXM1. GAPDH was used as the loading control. c CCK-8 assay showing the effects of FOXM1 silencing on A375 cell proliferation at 24, 48, 72, and 96 h. d Representative images of Transwell migration assays in the Control, si-NC, and si-FOXM1 groups. e Quantification of migrated cells shown in (d). f Representative images of Transwell Matrigel invasion assays in the Control, si-NC, and si-FOXM1 groups. g Quantification of invaded cells shown in (f). Ns, not significant. Statistical significance is indicated as *P < 0.05, **P < 0.01, and ***P < 0.001 for the comparisons shown in the figures

The effects of FOXM1 depletion on melanoma cell proliferation were evaluated using CCK-8 assays. As shown in Fig. 9c, no significant difference in cell viability was observed between the Control and si-NC groups throughout the experimental period, indicating that transfection alone had minimal effects on cell growth. In contrast, FOXM1 knockdown significantly inhibited A375 cell proliferation at 48 h (P < 0.01), and the inhibitory effect became more pronounced at 72 h and 96 h (both P < 0.001).

To further assess the impact of FOXM1 on melanoma cell motility, Transwell migration and Matrigel invasion assays were performed. Representative images showed a marked reduction in the number of migrated and invaded cells in the si-FOXM1 group compared with the Control and si-NC groups (Fig. 9d, f). Quantitative analyses further confirmed that FOXM1 silencing significantly reduced the migratory capacity (P < 0.05; Fig. 9e) and invasive ability (P < 0.01; Fig. 9g) of A375 cells. Collectively, these findings indicate that FOXM1 promotes the proliferation, migration, and invasion of melanoma cells, supporting its potential role in melanoma progression.

Discussion

In oncology, biomarkers represent a central pillar: they serve not only as molecular windows into the biological mechanisms of cancer, but also as essential tools supporting precision medicine and improving patient outcomes. Their value spans the entire clinical continuum, including elucidation of molecular mechanisms, auxiliary diagnosis, therapeutic response prediction, and prognostic assessment [40]. In recent years, cancer treatment has fully entered the era of precision medicine. Although melanoma has achieved breakthrough therapeutic progress, it remains a growing global public health challenge [41, 42]. In this highly invasive malignancy, biomarker research and clinical application are advancing at an unprecedented pace [43]. These biomarkers not only reveal the biological essence of tumors, but also demonstrate substantial value in predicting patient prognosis, guiding individualized treatment strategies, and evaluating responses to immunotherapy, representing a core driving force in advancing melanoma precision medicine [43]. Today, melanoma management has shifted toward a diversified and precision-driven therapeutic paradigm. Targeted therapies represented by BRAF/MEK inhibitors [44, 45], together with immunotherapies represented by PD-1/PD-L1 blockade, constitute the therapeutic foundation.

Accumulating evidence has demonstrated that FOXM1 is abnormally overexpressed in multiple human cancers and broadly participates in the entire sequence of tumor development, including tumor initiation, proliferation, invasion and metastasis, therapeutic resistance, and recurrence [46, 47]. Across diverse tumor types, FOXM1 upregulation is associated with increased proliferative capacity, enhanced mitotic activity, and genomic instability. From a clinical perspective, these biological characteristics are frequently linked to aggressive tumor behavior, therapeutic resistance, and poor patient outcomes across multiple cancer types. In melanoma, these features are directly relevant to risk stratification, as highly invasive and proliferative tumor states are closely associated with metastatic dissemination and unfavorable clinical outcomes. The association of FOXM1 expression with transcriptional programs related to chromosome segregation and cell cycle progression aligns with biomarkers currently employed to evaluate proliferative malignant phenotypes. These data imply that FOXM1 may represent an adjunct biomarker for assessing the aggressive phenotype of melanoma, particularly at early stages where prognostic evaluation is uncertain.

In this study, FOXM1 emerged as a promising biomarker candidate in melanoma. Pan-cancer transcriptomic analyses demonstrated that FOXM1 is broadly upregulated across multiple cancer types, including melanoma, a finding that was reproduced in independent public datasets and validated at the protein level. The tumor-restricted upregulation of FOXM1, together with its nuclear localization consistent with transcription factor function, supports the relevance of FOXM1 to melanoma pathobiology rather than being a secondary consequence of non-specific tissue alterations. FOXM1 overexpression was further associated with adverse clinicopathological features and unfavorable survival endpoints, including overall survival, disease-specific survival, and progression-free interval. Crucially, multivariable Cox regression incorporating established melanoma prognostic parameters—such as stage, ulceration, Clark level, and Breslow thickness—identified FOXM1 as a variable contributing prognostic information comparable to traditional risk factors. Integration of FOXM1 expression into a nomogram showed potential incremental prognostic value, with calibration and time-dependent ROC analyses demonstrating stable discriminatory performance at 1-, 3-, and 5-year follow-up. These findings indicate that FOXM1 may provide complementary prognostic information alongside conventional pathological assessment.

The distribution of FOXM1 expression further revealed marked intra-tumoral heterogeneity, consistent with melanoma’s transcriptional diversity and lineage plasticity. This heterogeneity may have biological implications, as FOXM1 has been linked to proliferative fitness, mitotic regulation, and invasive transcriptional programs in other tumor types. Thus, FOXM1 expression may reflect the presence of transcriptional states associated with tumor progression and metastatic potential. Whether FOXM1 marks or drives such states remains to be determined. FOXM1-related transcriptional alterations were mainly enriched in cell cycle, mitosis, and DNA replication pathways, whereas epidermal differentiation and keratinization pathways were negatively enriched. This reciprocal regulatory pattern is consistent with phenotypic switching in melanoma, in which proliferative and invasive transcriptional states can dynamically interconvert. Functional network and clustering analyses further showed that FOXM1 is positioned within highly interconnected gene modules involved in mitosis, chromosome segregation, and replication stress responses, suggesting the existence of a coherent regulatory network with biological relevance. Concurrent enrichment of immune-related and cytokine-related pathways indicates that FOXM1 may intersect with microenvironmental signaling processes influencing melanoma progression, although the causal nature of this association remains to be clarified. Studies have demonstrated that FOXM1 plays a critical role in regulating cell proliferation and cell cycle control [48], and is frequently overexpressed in tumor cells, closely associated with chromatin remodeling and replication stress through epigenetic mechanisms, supporting tumor proliferation and invasion [28]. These findings provide a functional annotation framework for understanding FOXM1-associated transcriptional alterations within the biological context of melanoma.

Our integrative immune analyses indicated that high FOXM1 expression in melanoma was associated with a distinct tumor immune microenvironment characterized by reduced cytotoxic immune activity and diminished stromal infiltration. Tumors with high FOXM1 expression exhibited systematic alterations in immune cell composition, including relative increases in macrophage subsets and resting NK cells, accompanied by significant reductions in CD8 T cells, Th1 cells, Th17 cells, and dendritic cells. These findings were validated across two orthogonal immune quantification frameworks (CIBERSORT and ssGSEA), confirming the robustness of the observed immune remodeling phenomenon. FOXM1 expression was negatively correlated with ESTIMATE score, immune score, and stromal score, further suggesting that FOXM1-high tumors exhibit an immune-excluded microenvironment with reduced infiltration of cytotoxic and antigen-presenting immune cells. Given that melanoma development is highly dependent on T cell–mediated immune surveillance [49] and demonstrates high sensitivity to immune checkpoint blockade, reductions in CD8⁺ T cells and dendritic cells suggest the possibility of downstream immune evasion programs linked to FOXM1 [50]. Numerous studies have reported that FOXM1 can remodel the tumor immune microenvironment and regulate T cells, contributing to immune escape [25, 28]. Genetic mouse models have shown that FOXM1 suppresses CD8⁺ T cell tumor infiltration and inhibits CD8⁺ T cell chemotaxis and antigen-dependent cytotoxic activity in vitro [51], while in gastric cancer, FOXM1 suppresses CD8⁺ T cell activation and induces immune escape [52]. Therefore, the present analyses indicate that FOXM1 may represent a potential regulator of proliferation-associated immune evasion in melanoma. Recent mechanistic studies further provide biological context for the potential immunoregulatory role of FOXM1. Timilsina et al. demonstrated that tumor-intrinsic FOXM1 can epigenetically suppress DNA-sensing pathways through the DNMT1–UHRF1 complex, thereby reducing ULBP1 expression and attenuating NK cell- and CD8⁺ T cell-mediated antitumor immunity [28]. Importantly, this study employed immunocompetent mouse models, single-cell transcriptomic analyses, flow cytometric immune profiling, and immune-cell depletion experiments, providing functional evidence that FOXM1 can actively shape antitumor immunity rather than merely serving as a correlative immune marker. In addition, Tang et al. reported that the AKT–FOXO3a–FOXM1 signaling axis contributes to PD-L1 regulation and tumor immune escape [53]. Although these studies were not conducted in melanoma models, they provide functional and mechanistic evidence supporting the biological plausibility of the association between elevated FOXM1 expression and reduced cytotoxic immune activity observed in the present study. Future mechanistic investigations are required to determine whether FOXM1 directly modulates immune reprogramming or merely marks a proliferative state inherently insensitive to immune surveillance. These advances will help refine the role of FOXM1 as a prognostic biomarker and candidate therapeutic target within immuno-oncology.

FOXM1 is not only a transcription factor, but is also associated with m⁶A, m⁵C, and m¹A modifications, suggesting that FOXM1 may participate in post-transcriptional regulation of gene expression, particularly in proliferative, metastatic, and drug-resistant tumor processes [54]. Upstream regulation of FOXM1 mediated by m⁶A-, m⁵C-, and m¹A-related RNA modification factors, together with genomic amplification events, may enhance FOXM1 transcriptional output and stabilize oncogenic programs [55, 56]. In addition, our pharmacogenomic analyses revealed significant associations between FOXM1 expression and the response patterns of multiple therapeutic agents, including MAPK pathway–related inhibitors such as selumetinib and trametinib. Given the established role of FOXM1 as a downstream component of MAPK-driven transcriptional programs in melanoma [57, 58], Bonet et al. demonstrated that pharmacologic inhibition of MAPK/ERK signaling reduced FOXM1 expression together with its downstream mitotic target Aurora B, whereas FOXM1 silencing similarly suppressed Aurora B expression, suggesting that FOXM1 functions as an important downstream effector of MAPK/ERK signaling in melanoma [57]. Therefore, FOXM1 expression may partially reflect the activity of MAPK-dependent proliferative signaling, providing a potential biological explanation for its association with MAPK inhibitor response. Furthermore, Liu et al. reported that modulation of the FOXO3a/FOXM1 axis was associated with trametinib activity in a non-small cell lung cancer model [59], suggesting that FOXM1-related transcriptional programs may contribute to treatment-response heterogeneity. Collectively, these findings provide biological plausibility for the observed association between FOXM1 expression and MAPK inhibitor sensitivity in the present study. In addition, Puig-Butille et al. demonstrated that FOXM1 participates in a MAPK-driven transcriptional network in BRAF- and NRAS-mutant melanoma and regulates the expression of cell-cycle–related genes, including AURKA [58]. These observations further suggest that the relationship between FOXM1 activity and MAPK signaling may be particularly relevant in specific molecular subgroups of melanoma and warrant further investigation in future studies incorporating molecular stratification. However, the present analyses were based primarily on pharmacogenomic datasets and were not adjusted for BRAF mutational status. Moreover, independent validation in external melanoma cohorts or drug-treated experimental models was not performed. Therefore, these observations should be regarded as hypothesis-generating rather than definitive evidence of predictive utility. Future studies integrating molecular stratification and experimental validation will be required to determine whether FOXM1 can independently predict sensitivity to MAPK inhibitors in melanoma.

Histopathological assessment provided morphological support for the molecular findings. H&E staining revealed pronounced structural disruption in tumor specimens, characterized by loss of tissue stratification, increased cellular density, and stromal remodeling, features consistent with tumor progression and invasive growth. These tumor-associated histological alterations formed a clear contrast with the preserved epithelial and stromal architecture observed in adjacent normal tissues, delineating pathological boundaries between malignant and non-malignant tissues and supporting the suitability of tissue specimens for biomarker validation. On this morphological basis, immunohistochemical analysis demonstrated a tumor-specific enrichment pattern of FOXM1 protein. In all melanoma samples, FOXM1 exhibited prominent nuclear and cytoplasmic staining, whereas corresponding normal tissues displayed weak or absent signals, indicating that FOXM1 activation is largely restricted to malignant cells. Collectively, these results indicate that FOXM1 is upregulated at the protein level in melanoma and can distinguish malignant from non-malignant tissues with high specificity. The pronounced differential expression of FOXM1 between tumor and normal tissues supports its potential utility as a pathological biomarker for melanoma, while also suggesting possible involvement of FOXM1 in melanoma-specific phenotypes such as proliferation, invasive spread, and stromal remodeling. In addition to histopathological and protein-level validation, functional experiments were performed to further evaluate the biological role of FOXM1 in melanoma. Following siRNA-mediated FOXM1 silencing, both qRT-PCR and Western blot analyses confirmed efficient FOXM1 knockdown. Subsequent functional assays demonstrated that FOXM1 depletion significantly suppressed the proliferation, migration, and invasion of A375 melanoma cells. These findings are highly consistent with the enrichment of cell cycle-, mitosis-, and DNA replication–related programs identified in the transcriptomic analyses, further supporting a role for FOXM1 in maintaining malignant phenotypes in melanoma. These newly added functional experiments provide direct biological evidence that FOXM1 contributes to the maintenance of proliferative, migratory, and invasive phenotypes in melanoma cells. Although the underlying molecular mechanisms remain to be elucidated, these findings extend the present study beyond descriptive bioinformatics observations and support a functional role of FOXM1 in melanoma progression.

Despite comprehensive multi-omic characterization and experimental validation, this study has several limitations. First, most analyses were based on retrospective public datasets, including TCGA, GTEx, GEO, CTRP, and GDSC; therefore, the observed associations should be interpreted cautiously and do not establish causal relationships. Although FOXM1 showed potential incremental prognostic value when integrated with conventional clinicopathological variables, the current analysis lacked several clinically important prognostic factors, such as serum lactate dehydrogenase (LDH) levels, sentinel lymph node status, detailed AJCC TNM staging information, treatment history, and prospective follow-up data. Therefore, future studies using large independent cohorts should evaluate whether FOXM1 provides additional prognostic information beyond established melanoma risk factors through C-index, net reclassification improvement (NRI), and decision curve analysis (DCA).

Second, although immunohistochemical validation supported increased FOXM1 protein expression in melanoma tissues, the clinical tissue cohort was relatively small and retrospective. Larger multicenter cohorts with standardized pathological scoring and detailed clinical annotation are required to confirm the diagnostic and prognostic utility of FOXM1 at the protein level.

Third, the functional experiments were performed in A375 melanoma cells. Although FOXM1 knockdown suppressed cell proliferation, migration, and invasion, validation in additional melanoma cell lines with different molecular backgrounds, as well as in vivo models, will be necessary to further confirm the biological role of FOXM1 in melanoma progression.

Finally, the immune infiltration and pharmacogenomic analyses were exploratory. The association between FOXM1 expression and MAPK inhibitor response was not adjusted for BRAF mutational status and was not validated in independent drug-treated melanoma cohorts or experimental models. Similarly, the mechanisms linking FOXM1 to immune remodeling, RNA modification regulators, and treatment-response heterogeneity remain to be clarified. Future mechanistic studies incorporating molecular stratification, immune co-culture systems, in vivo models, and experimental drug-response assays are warranted.

Authors' contributions

Ran Huo: Conceptualization, funding acquisition. Luying Wang: Supervision, funding acquisition. Zhichen Li and Zhenpeng Zhu: Data curation, Writing-Original draft preparation, Methodology, Software. Ruixue Zhao: Writing—reviewing and Editing. All authors have reviewed the results and approved the final version of the manuscript.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by the National Natural Science Foundation of China (Grant No. 82172227, 82372537).

Data availability

Data availability statementThe datasets presented in this study can be found in the online repositories, including The Cancer Genome Atlas (TCGA; https://portal.gdc.cancer.gov), UCSC XENA (https://xenabrowser.net/datapages), the GEPIA2.0 database (http://gepia2.cancer-pku.cn/#index), The Human Protein Atlas (HPA; https://www.proteinatlas.org/), the cBioPortal platform (https://www.cbioportal.org/), the GSCA database (https://guolab.wchscu.cn/GSCA/#/), GDSC (https://www.cancerrxgene.org/), CTRP (http://portals.broadinstitute.org/ctrp/), the CIBERSORTx website (https://cibersortx.stanford.edu/). The datasets generated during the current study, including immunohistochemical scoring data and in vitro experimental results, are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This study was conducted in accordance with the principles of the Declaration of Helsinki. Ethical approval for this study was granted by the Ethics Committee of the Provincial Hospital affiliated to Shandong First Medical University for human biomedical research (Approval No. SWYX:2025 − 935). Written informed consent was obtained from all participants prior to inclusion in the study.

Consent for publication

Not applicable. No identifiable individual participant data are included in this study. All tissue specimens were anonymized prior to analysis.

Competing interests

The authors declare no competing interests.