Development and Validation of a Combined Ferroptosis and Immune Prognostic Model for Melanoma

Background Melanoma development and progression are significantly influenced by ferroptosis and the immune microenvironment. However, there are no reliable biomarkers for melanoma prognosis prediction based on ferroptosis and immunological response. Methods Ferroptosis-related genes (FRGs) were retrieved from the FerrDb website. Immune-related genes (IRGs) were collected in the ImmPort dataset. The TCGA (The Cancer Genome Atlas) and GSE65904 datasets both contained prognostic FRGs and IRGs. The model was created using multivariate Cox regression, the least absolute shrinkage and selection operator (LASSO) Cox regression analysis, and the analysis and comparison between the expression patterns of ferroptosis and immune cell infiltration were done. Last but not least, research was conducted to assess the expression and involvement of the genes in the comprehensive index of ferroptosis and immune (CIFI). Results Two prognostic ferroptosis- and immune-related markers (PDGFRB and FOXM1) were utilized to develop a CIFI. In various datasets and patient subgroups, CIFI exhibits consistent predictive performance. The fact that CIFI is an independent prognostic factor for melanoma patients was revealed. Patients in the CIFI-high group further exhibited immune-suppressive characteristics and had elevated ferroptosis gene expression levels. The results of in vitro research point to the possibility that the PDGFRB and FOXM1 genes function as oncogenes in melanoma. Conclusion In this study, a novel prognostic classifier for melanoma patients was developed and validated using ferroptosis and immune expression profiles.


Introduction
Melanoma is the deadliest type of skin cancer, with a yearly rise in incidence [1,2]. Te limited treatment of choice for advanced melanoma is immune checkpoint blockade (ICB) and molecularly targeted therapies, such as CTLA-4, PD-1/ PD-L1 inhibitors, and BRAF inhibitors due to melanoma's high level of heterogeneity and aggressiveness [3,4]. However, primary or secondary drug resistance afects more than 50% of melanoma patients, which presents a signifcant clinical treatment challenge [5,6]. In addition to the timing of treatment being delayed once patients develop drug resistance, there will be limited options for subsequent therapeutic approaches [7]. Developing precise, individualized treatment plans for various patients and accurately predicting the risk and drug treatment response for diferent individuals is an efective approach to addressing the drug resistance problem. Obtaining reliable biomarkers for a melanoma diagnosis is therefore crucial to initiate clinical treatment. Ferroptosis is an excessive lipid peroxidationinduced regulated cell death mode that is iron-dependent and associated with the progression and treatment response of multiple types of tumors [8,9]. Recent research has demonstrated that ferroptosis may result in immunosuppression brought by infammation in the tumor microenvironment [10,11]. Additionally, there are interactions between immune and tumor cells that are connected to ferroptosis [12,13]. Ferroptosis is regulated by a variety of molecular components in the tumor microenvironment. Increased iron accumulation, free radical generation, fatty acid supply, and lipid peroxidation are critical for the development of ferroptosis [14]. Numerous studies have demonstrated that the immunosuppressive microenvironment can be impacted by ferroptosis intervention. Ferroptosis may expose tumor antigens, increasing the tumor microenvironment's immunogenicity and enhancing the efectiveness of immunotherapy [15][16][17].However, the JAK-STAT1 pathway is activated, and SLC7A11 and SLC3A2 expression are downregulated, which causes ferroptosis in tumor cells when interferon gamma is released by cytotoxic T cells [18]. Te long-term efects of ferroptosis on tumor immunity depend on the interactions among cancer cells and other immune cell subsets. For instance, the lymphatic system inhibits melanoma cells from ferroptosis via the mechanism of enhancing the synthesis of ACSL3-dependent MUFAs, which promote tumor spread [19]. Despite a strong link between ferroptosis efects and the immune microenvironment, their role in melanoma is yet unknown. In this work, a comprehensive index of ferroptosis and immune (CIFI) model was created and validated using ferroptosisrelated genes (FRGs) and immune-related genes (IRGs). Te CIFI model demonstrated consistent prognostic predictive performance in patients with various clinical features and across various datasets. Te outcomes of in vitro experiments and clinical validation were used to validate the expression and function of the PDGFRB gene and the FOXM1 gene in CIFI.

Data
Preparation. RNA-seq information and follow-up data for melanoma patients were acquired from TCGA (472 samples). Data from the Gene Expression Omnibus (GEO) were obtained, and the GSE65904 dataset (214 samples) was chosen since it had the biggest sample set in the GEO database and detailed follow-up information. Te FerrDb website includes FRGs. Te ImmPort dataset's IRGs were downloaded.

Construction and Validation of the CIFI.
We selected the independent prognostic genes across FRGs and IRGs using the "survival" program and univariate Cox analysis. Both LASSO analysis and sequential Cox proportional hazards regression were employed to develop CIFI. Te risk score's optimal cutof value was used to establish the CIFI-high and CIFI-low groups (categories). To examine the diferences (variations) in overall survival between the CIFI-high and CIFI-low groups, the Kaplan-Meier survival analysis was utilized. Te time-dependent ROC analysis was also used to assess CIFI's predictive ability. To assess the CIFI's independent prognostic signifcance, both univariate and multivariate Cox regression analyses were utilized.

Potentially Regulatory Pathways
Analysis. Te score of each pathway per sample was determined using a singlesample gene set enrichment analysis (ssGSEA) with the aid of the "GSVA" package. Te analysis of the correlation between the CIFI and ssGSEA scores of each sample was then conducted using the potential regulatory pathways.

Immunohistochemistry (IHC) Analysis.
Te First Hospital of China Medical University provided clinical samples (paired nontumor skin tissue and melanoma). According to a prior study [20], IHC staining and scoring were conducted. Te following antibodies were used: PDGFRB (1 : 200; ab69506; Abcam) and FOXM1 (1 : 1000; ab207298; Abcam). Te Ethics Committee of the First Hospital of China Medical University approved this study.

CCK8 Assay and Colony-Forming Experiments.
Te negative control siRNA (NC-siRNA), the PDGFRB-specifc siRNA, and the FOXM1-specifc siRNA were all transfected into cell cultures in 96-well plates. Cells were cultured with CCK8 solution (C0038, Beyotime, Shanghai, China) for another 2 hours after 0, 24, 48, and 72 hours. To determine cell vitality, an optical density (OD) value at 450 nm was recorded. Cells (500/well) treated with siRNAs for colonyforming experiments were added to 12-well plates. Te colonies were counted after two weeks.

qPCR and Western
Blot. qPCR and western blot procedures were conducted as previously mentioned [21].

Statistical
Analysis. Te statistical tool SPSS 21.0 was used to examine the data (IBM Corporation, Armonk, NY, USA). Software called GraphPad Prism 8.0 was used to generate the graphs (GraphPad Software, Inc., San Diego, CA). Student's t-tests were applied. With regard to t-tests, a two-tailed p < 0.05 indicated a signifcant value.

TCGA Dataset Prognostic Analysis of CIFI.
Te risk score distribution from the TCGA dataset, which was initially computed for each sample in Figure 2(a), was shown. Patients in the CIFI-high group showed remarkably worse overall survival rates than those in the CIFI-low group, as per a Kaplan-Meier survival analysis (Figure 2(b); P < 0.0001). Te AUC values over 1, 3, and 5 years of survival were shown by ROC curve analysis to be 0.614, 0.587, and 0.619, respectively ( Figure 2(c)).

Verifcation of CIFI in GSE65904.
Te stability and dependability of CIFI are then further verifed. Figure 3     Journal of Oncology considerably worse than that of the CIFI-low group. However, CIFI was able to diferentiate the prognosis of several patient subgroups.

Cox Analysis of CIFI and Nomogram Construction.
We then performed univariate and multivariate Cox regression analyses. Age, CIFI, M, N, T, and tumor stage were     Journal of Oncology all linked to patients' prognosis, according to a univariate Cox analysis (Figure 5(a)). According to a multivariate Cox analysis, the patient's prognosis was determined independently by CIFI, age, M stage, N stage, and T stage ( Figure 5(b)). Tese fndings showed that a high CIFI was an independent predictor of outcomes in melanoma patients. Te development of a quantitative technique in clinical practice may also aid doctors in assessing melanoma patients' prognosis. Based on the outcomes of multivariate Cox regression analysis, a nomogram incorporating clinicopathological traits and CIFI was constructed ( Figure 5(c)). Te model's potent ability to predict patient outcomes over 5 years was confrmed by the calibration curves, which revealed considerable overlap between the calibration points and the standard curve ( Figure 5(d)). Additionally, the decision curve analysis demonstrated that the nomogram model accurately predicted overall survival (OS) compared to a single clinicopathological characteristic ( Figure 5(e)). Tese fndings revealed the therapeutic use of the CIFI-based nomogram to determine the prognosis of melanoma patients.

Immune Profle in the CIFI.
We then investigated if CIFI may indicate a melanoma immunological state. First, we investigated the link between CIFI and immune-invading cells. ssGSEA was conducted to determine the level of infltration of 28 immune cells. Some immune cells were expressed aberrantly in both the CIFI-high and CIFI-low groups (Figures 6(a) and 6(b)). Following that, we sought to investigate the link between CIFI and the tumor immune microenvironment. Figure 6(c) showed that CIFI was positively correlated with immune score (R = 0.1, P < 0.043) and stromal score (R = 0.38, P < 0.001).

Identifying Pathways Related to CIFI.
Te link between CIFI and biological function was the subject of our next analysis efort. Using a cutof of P < 0.05, it was observed that the TCGA cohort samples with CIFI-high had 237 substantially upregulated genes and 38 considerably downregulated genes (Figure 7(a)). Te expression levels of these diferentially expressed genes (DEGs) were then imported into Metascape. Te extracellular matrix    ns ns ns ** ns ** ************ ns * ** ns ns ******** ** ns * ** * *** ******** *** *** **** ns CIFI high low Activated CD4 T cell Activated CD8 T cell Activated dendritic cell  Journal of Oncology organization, collagen synthesis, blood vessel development, etc. were the primary areas where the upregulated genes were enriched. Te majority of the downregulated genes were specialized for pigmentation, inner ear development, etc. (Figure 7(b)). Te scores of all patients in various routes were then obtained by calculating the ssGSEA scores by GSVA. As the risk score grew, ECM receptor interaction, focal adhesion, and other activities increased, whereas ribosome, basal transcription factors, and other activities decreased (Figure 7(c)).

Validation of the CIFI Genes at the mRNA and Protein
Levels, and Its Functional Analysis. As shown in Figures 8(a) and 8(b), melanoma cells (A375, A875, and MeWo) had considerably higher PDGFRB and FOXM1 mRNA and protein levels than melanocyte PIG1 cells. Te outcomes of immunohistochemistry experiments demonstrated that PDGFRB and FOXM1 were overexpressed in melanoma samples in comparison to normal tissues (Figure 8(c)). Furthermore, we transfected the A375 cells, and transfection efciency was shown in Figure 8(d). Based on in vitro experiments, we found that silencing PDGFRB and FOXM1 inhibited the proliferative capacity of melanoma cells (Figures 8(e) and 8(f )). Tese fndings collectively indicated that PDGFRB and FOXM1 may act as carcinogens in melanoma.

Discussion
Melanoma is highly aggressive, yet the patient's prognosis remains poor due to limited conventional treatment options [22,23]. Terefore, there is a need to establish prognostic characteristics for melanoma patients. Several studies have provided potential prognostic assessment models for melanoma patients [24][25][26]. However, most of the studies were genomic or transcriptomic-based and did not validate biological functions. Melanoma onset and progression are signifcantly infuenced by ferroptosis and the immune microenvironment [27,28]. We developed CIFI in this study utilizing FRGs and IRGs based on open-source datasets. In many datasets and patient subsets, CIFI exhibits consistent predictive accuracy. Signifcantly, CIFI is an independent prognostic factor for melanoma patients. In conclusion, CIFI has the potential to be useful in clinical settings and is quite efective in predicting the prognosis of patients with melanoma.
Currently, several studies have reported immune-related molecular markers as prognostic markers for melanoma [29][30][31], but these prognostic markers had limitations. First, molecular markers in previous studies contained various genes, which increased the workload and cost in clinical practice and constrained the clinical utility of these molecular markers to some extent [29,32,33]. Second, these studies have not further investigated the underlying mechanisms or clinical signifcance of prognostic markers [28,34], and therefore, the clinical reliability of these prognostic markers was unclarifed. Terefore, it is imperative to fnd a reliable and practical prognostic marker. Our study yields a predictive model for melanoma patients based on ferroptosis and immunity, two essential tumor characteristics. Tis model may simultaneously refect changes in melanoma's ferroptosis and immunological state. Our methodology is also practical for clinical use. Most importantly, the outcomes of clinical samples and in vitro tests have validated the expression level and probable functions of genes in our model. Te CIFI we constructed contains two genes, PDGFRB and FOXM1. Among these, PDGFRB and FOXM1 are genes involved in ferroptosis and the immune system, respectively. Embryonic development, cell proliferation, survival, diferentiation, chemotaxis, and migration are all regulated by the protein that the PDGFRB gene produces [35,36]. Previous studies have established the link between PDGFRB overexpression and a poor prognosis in patients with renal cell carcinoma [37], oral squamous cell carcinoma [38], ovarian cancer [39], and colorectal cancer [40]. Involved in cell proliferation, the transcriptional  activator FOXM1 may have an impact on the expression of several cell cycle genes, including cyclin B1 and cyclin D. Numerous studies have revealed the association of elevated FOXM1 expression levels with a poor prognosis in patients with ovarian cancer [41], pancreatic and esophageal cancers [42], malignant rhabdoid tumors [41], and small-cell lung cancer [43]. FOXM1 expression level is elevated and activated in malignant melanoma [44]. FOXM1 inhibition might be a promising treatment strategy for metastatic melanoma [45].
We also compared our model with others. Only two models based on immune or ferroptosis-related genes have been discovered so far [46,47]. However, in our analysis, for the frst time, the incorporated ferroptosis and immune gene set were utilized to develop a melanoma-related prognostic model, which could play a consistent prognostic performance in diverse data sets that could be utilized as an independent prognosisrelated predictive indicator for melanoma patients.
Although CIFI has the potential to be an excellent model for predicting prognosis in individuals with melanoma, it has certain drawbacks. To begin, all samples in this study were obtained retrospectively, and potential samples are currently being validated. We, therefore, examined CIFI's prognostic signifcance in clinic settings. Te roles of PDGFRB and FOXM1 in CIFI require further in vivo and in vitro investigations.

Conclusions
In conclusion, a novel prognostic classifer based on ferroptosis and immune expression profles in patients with melanoma was developed and validated.

Data Availability
Te data that support the fndings of this study are available from the corresponding author upon reasonable request.