With its extremely poor prognosis, SKCM is the most aggressive malignancy of all skin cancers and a significant health burden all over the world. Among a wide range of prognosis-related factors, variations in methylation profiles are attached to great importance for the onset and progression of melanoma 30. The current work was committed to identify potential predictors contributing to SKCM overall survival via an integrative analysis of DNA methylation and gene expression. Data of DNA methylation, gene expression and clinical presentation of SKCM were downloaded from TCGA database. A total of 83 significantly differential methylation genes were screened out. And among them, 21 were down - regulated and 62 were up - regulated (supplementary material 1), which laid a basis for the following analysis to elucidate the possible mechanism of the initiation or development of SKCM.
Previous studies have suggested that DNA hypermethylation was more common in comparison with DNA hypomethylation. It is widely accepted that hypermethylation is involved in the development of melanoma by inactivating the expression of tumor suppressor genes 11, 31. Jesper Worm et al. validated that reduced expression of hypermethylated APC (Adenomatous Polyposis Coli) could promote cell proliferation in melanoma 32. Mario Venza et al. reported that loss of hypermethylated CDH1 (E-Cadherin) correlated definitely with high melanoma grade and undesirable outcomes 33. Kathleen Conway and Elin J Ekström et al. verified that hypermethylated FRZB (Frizzled-Related Protein) was a metastasis suppressor in malignant melanoma 34, 35. Therefore, the correlation between the methylation levels of genes and the corresponding mRNA expressions is an underlying mechanism for the outcome of SKCM. In this paper, among 62 differentially hypermethylated genes, 15 were related to the corresponding mRNA expressions. And there were 6 methylated genes, namely HHEX, LINC00506, NUP62CL, TMEM255A, CH25H and C2CD4A (Fig. 3, Table 1), were identified as driven genes. With the Kaplan - Meier curves for prognostic analysis, highly-expressed and hypomethylated HHEX (p = 0.02) and LINC00506 (p = 0.005) showed better outcome than lowly-expressed and hypermethylated HHEX and LINC00506 (Fig. 5), which revealed that HHEX and LINC00506 were of significantly prognostic value in SKCM. To be more specific, the five-year survival rate of HHEX with high-expression and hypomethylation (0.629; 95% CI, 0.553 to 0.716) was significantly higher than that of HHEX with low-expression and hypermethylation (0.540; 95% CI, 0.457 to 0.639). LINC00506 with high-expression and hypomethylation (0.665; 95% CI, 0.589 to 0.752) seemed to have a better survival rate than with low-expression and hypermethylation (0.488; 95% CI, 0.406 to 0.586) (Table 3). HHEX has been extensively demonstrated to be associated with the migration and invasion of multiple malignancies, such as breast cancer, hematological malignancy, liver cancer, prostate cancer, colorectal cancer and thyroid cancer 36, 37. It has been proved to serve as a prognostic marker for various malignant tumors 38. Thus it was chosen as a possible prognosis-relevant methylation gene of interest for the subsequent analysis.
HHEX (hematopoietically expressed homeobox), also named as HEX, HOX11L-PEN, and PRHX, encodes several transcription factors of the homeobox family, which are implicated in many kinds of tumor development processes, including cell proliferation, invasion, metastasis and survival39, 40. Surprisingly, in SKCM, while extensive researches have explored its epigenetic alterations, the involvement the methylation levels and sites of HHEX in the outcome of tumor have yet to be fully elucidated. DNA methylation regulate the corresponding mRNA expression and has an impact on the molecular function of translated proteins. Aberrant HHEX methylation may be a crucial factor to pathophysiologic aggression and worse survival of cancer, which is supported in plenty of researches 36, 37, 41. Notably, the hypermethylation of HHEX has been reported to be associated with poor overall and disease-specific survival among patients with acral melanoma (AM) 42. However, AM is just a subtype of cutaneous melanoma localize in the glabrous skin of the palms, soles, or nail apparatus 43. In our study, evaluated by Kaplan-Meier survival curves, in contrast to hypermethylation, hypomethylation of HHEX displays a significant better outcome in cutaneous melanoma, which is consistent with previous studies on other cancers 41. Therefore, we hypothesize that a hypermethylated driven gene of HHEX may be a potential predictor for impaired survival in patients with SKCM.
However, according to our findings, the expression of HHEX varies in different pathological stages of SKCM (Fig. 6). It is at a relatively low level in stage zero, in which tumors are confined to the epidermis or mucosal epithelium without invasion or distant metastasis, while in stage Ⅰ-Ⅳ, tumors have infiltrated into the dermis and deeper layers, and the expression of HHEX shows a downward trend with advanced stages. As we have known, invasive melanomas are among the most immunogenic tumors 44, and some immunocytes are deemed to be involved in SKCM, which includes macrophages, neutrophils and dendritic cells etc. 45–48. Here, with multiple immune deconvolution approaches, we observed a positive correlation between HHEX expressions and the immune infiltration of the above mentioned immunocytes in the SKCM (Fig. 7). And the enrichment analysis of HHEX and its related partners further displays that a group of HHEX genes are mainly enriched in a set of immune - related processes, such as differentiation regulation of myeloid cells, lymphocytes and leukocytes (Fig. 8, Table 4). A possible explanation is that activated immune response, together with increasing hypermethylation, co-regulate the expression of HHEX. And the expression of HHEX and the immune response may be interactive in the development of SKCM. Therefore, when SKCM is in situ, the immune response may not be fully activated, so the hypermethylation status of HHEX may not be interfered by the immune response. However, when SKCM develops into invasive cancer, the immune response is correspondingly activated to some extent, resulting in gradually decreasing methylation levels of HHEX in stage Ⅰ - Ⅳ. In addition, with the deterioration of the tumor and the gradual weakening of the body’s immune response, the methylation level of HHEX increased and the corresponding mRNA level decreased. The result seems to be in line with the joint survival analysis of different pathologic stages of SKCM. Therefore, HHEX is reasonable enough to act as a potential candidate for prognostic biomarker for patients with aggressive SKCM.
Identifying the sites of methylation is a requisite for decoding the roles of methylation in gene expression regulation, which has been implicated in the pathological processes of tumor occurrence and development. Numerous studies have corroborated that methylation at specific sites of some crucial genes affects the prognosis of various cancers 49–51. Based on the prognostic value of HHEX in SKCM, the correlation between specific methylation sites of HHEX and the corresponding mRNA expression were further analyzed. In our work, a total of 15 methylation sites (cg14689537, cg24787755, cg23009123, cg18599843, cg11055493, cg26979504, cg00487187, cg16508068, cg03330490, cg02185052, cg00876273, cg07616394, cg14283380, cg09721427 and cg04413153) have been identified (Fig. 4, Table 2). Since their methylation levels are closely related to the corresponding mRNA contents, we consider that these methylation sites may indirectly play a role in disturbing certain gene expression, then make an impact on the outcome of melanoma patients. Epigenetic changes are potentially reversible in melanoma 52, 53, which seems to provide a target for the prevention and treatment of the tumor. Accordingly, we speculate that not only survival - methylated HHEX, but also its methylation loci may become potential targets for SKCM therapies.
Although bioinformatics can provide researchers with vast amounts of data and a variety of ways to study disease, there are still some limitations that cannot be ignored. In our study, nearly all related raw data were downloaded from TCGA. However, in the majority of cases, the normal sample size was relatively too small to the tumor sample sizes. And the clinical information on individuals was insufficient, which appeared to be an underlying restriction. Accordingly, patient’s material derived from TCGA database was not absolutely representative of any real SKCM population. The results of our analysis were virtual and should be confirmed in future experiments.
Taken together, our multilayer data mining and comprehensive bioinformatics analysis revealed that the methylation of HHEX may be a novel molecular tool for the prediction of survival for SKCM patients. For the limitations of the current research, additional experiments, both in vivo and in vitro, are urgently needed to further validate the potential prognostic value of methylated HHEX and its methylation sites in SKCM.