Integrating TCGA LSCC RNA-seq datasets with DNA methylation datasets by MethylMix tools, we identified 88 DNA MDGs. On the basis of these DNA MDGs, we developed an epigenetic signature, which could accurately categorize patients into high-risk status and low-risk status. Stratification analysis verified that the epigenetic signature remained a significant statistical prognostic model in subsets of patients with different clinical variables. Multivariate Cox regression analysis uncovered the efficacy of epigenetic signature appears independent of other clinicopathological characteristics. With respect to predictive capacity and clinical usefulness, the epigenetic signature was superior to traditional TNM stage. Additionally, the epigenetic signature was validated in external LSCC cohorts from GEO. Finally, CMap matched the 10 most significant small molecules as promising therapeutic drugs to reverse the LSCC gene expression.
An increasing number of researches recognize that epigenetic changes such as the hypermethylation of tumor suppressor genes and hypomethylation of oncogenes in the diagnosis, progression and prognosis of LSCC played a critical role [11–13]. By quantitative methylation-specific polymerase (qMSP) chain reaction assays in 96 LSCC patients, Shen et al [11] revealed that LZTS2 promoter hypermethylation is linked to risk, progression, and prognosis of LSCC,which can serve as a diagnostic and prognostic biomarker for LSCC. Analyzing the LSCC tissues and Hep-2 cells to investigate the methylation status of the CpG islands of MYCT1 and mRNA levels, Yang et al [12] identified that hypermethylation contributed to the transcriptional downregulation of MYCT1 and could inhibit cancer cell differentiation in LSCC. By interfering with its binding to c-Myc, DNA methylation of the CGCG site (− 695 to − 692) of MYCT1 altered the promoter activity in LSCC. Recently, Weigel et al [13] concluded that a regulatory role of TREX2 DNA methylation for gene expression which might be an indicative of incidence and survival in LSCC.TREX2 DNA methylation could be a potential predictor of treatment response and as a biomarker to understand carcinogenesis in stratified epithelia. Additionally, protein levels of altered TREX2 could influence drug-induced DNA damage repair and offer novel tailored therapies in LSCC. These studies suggested the potential clinical implications of DNA methylation in providing novel biomarkers for valuable molecular targeted therapy and establish predictive models to optimize therapeutic strategies in LSCC patients. Nevertheless, almost all of the studies focused on the methylation status of one gene with limited statistical power in predictive values. Considering the heterogeneity of LSCC, entire molecular signatures derived from high-content genome screens seem to offer better prognostic value.
To our knowledge, this is the first study carried out a genome-wide integrated analysis of methylation and the transcriptome from TCGA database to seek novel biomarker as potential molecular targeted therapy and to create an epigenetic signature for LSCC patients to optimize therapeutic strategies. When applying high-throughput methodology with 450,000 probes, it is necessary to distinguish the epigenetic changes (“driver”) that act as effectors of the malignant phenotype from alterations of “passenger” without any biologic function [14]. Hence, a model-based tool (MethylMix) was applied to identify those genes with aberrant methylation and related these data to RNA-seq data affecting gene expression. Consequently, we identified a cohort of 88 MDGs in LSCC. The functional analysis indicated MDGs have mainly attached oneself to gene expression (transcription), RNA polymerase II transcription, transcription factor activity, sequence-specific DNA binding and so on. It was a hint that DNA methylation is involved in the dysregulation of genes with distinct functions and is functionally linked to outcomes in LSCC patients.
On the basis of univariate and multivariate Cox regression analyses, we selected five MDGs (MAGEB2, SUSD1, ZNF382, ZNF418 and ZNF732) to develop an epigenetic signature. It could effectively classify patients into high-risk group with shorter OS and low-risk group with longer OS in TCGA and GEO database. Remarkably, according to stratified analysis, the epigenetic signature was a statistically and clinically prognostic model in T3-T4, N0 or N1-N3, III-IV stage, G1-G2 or G3-G4, negative margin status, lymphovascular invasion or no lymphovascular invasion subgroup. Yet, in T1-T2, I-II stage and positive margin status subgroup did not reach significant statistics. One possible explanation that a small sample size of T1-T2 or I-II stage or positive margin status subgroup, consisting of less than 20 patients in the different subgroup, are not enough to generate an effect of significant statistics. Besides, univariate and multivariate Cox analysis affirmed that epigenetic signature was an independent predictor of unfavorable OS, regardless of other clinicopathologic variables in TCGA and GEO data set. To explore potential biological pathways for 5 prognostic related MDGs, we carried out the GSEA and indicated that 5 MDGs of epigenetic signature was mainly scattered in cancer-related pathways (JAK-STAT signaling pathway, P53 signaling pathway) and immune-related biological processes (T cell receptor signaling pathway, B cell receptor signaling pathway). It implied that 5 MDGs of epigenetic signature maybe involved in initiation,
maintenance, development of LSCC and associated with outcomes in LSCC patients.
Currently, in the context of the global trend toward organ-preserving treatment for LSCC patients (though total laryngectomy helps disease control and improve survival outcome, it has observed adverse effects on patients’ quality of life because of permanent tracheostomy, the loss of voice and issues with swallowing), pursuant to the newest National Comprehensive Cancer Network (NCCN) Guidelines, decision making about larynx-preserving are highly relied on tumor TNM staging. In our research, via ROC curve analysis, epigenetic signature shown more precise predictive ability compared with TNM stage model in TCGA and GEO database, which could effectively identify low-risk patients prone to larynx-preserving treatment and avoid needless total laryngectomy. Interestingly, DCA results indicated that LSCC survival-related treatment decisions based on the epigenetic signature led to more net benefit than treatment decisions based on TNM stage, or treating either all patients or none in TCGA and GEO database. Taken together, the current epigenetic signature would be clinically useful for the clinicians in tailoring survival-associated treatment decisions.
One prominent finding in our study was that combining methylation and RNA expression data with survival analysis, we identified 4 MDGs (MAGEB2, SUSD1, ZNF382, ZNF418), which may be serve as potential biomarkers or drug targets for early diagnosis and prognostic assessment. MAGEB2, as a member of the MAGEB family, which belongs to the cancer testicular antigens, is located in the last exon on chromosome X. The MAGEB2 gene has been reported to be overexpression in several types of tumors, such as esophageal cancer [15] and lung cancer [16], which has been implicated in carcinogenesis and considered as a potential cancer biomarker [17]. Pattini et al. [18] indicated that MAGEB2 is activated by promoter demethylation in HNSCC, which has growth-promoting effects on a minimally transformed oral keratinocyte cell line. SUSD1, which encodes the sushi domain-containing protein 1 precursor, acts as a common motif in protein-protein interaction [19]. SUSD1 presented subthreshold associations with venous thromboembolism was detailed in a report by Tang et al [20]. Nevertheless, information regarding the role of SUSD1 in cancer is lacking.
ZNF382(Zinc finger protein 382), a member of the Krüppel-associated box zinc finger protein family, plays critical role in regulating certain downstream gene expression as a transcription inhibitor. Due to the hypermethylation of promoter, ZNF382 is downregulated, thereby leading to inhibition of gene expression as found in multiple tumors [21, 22]. Functional researchers uncovered that ZNF382 was potential a tumor suppressor, via colony formation, inhibition of cell proliferation, invasion, migration, and acceleration of cell apoptosis as well as tumorigenic potential, in various cancers such as hepatocellular carcinoma (HCC), gastric cancer, esophagus squamous cell carcinoma (ESCC) [23–25]. ZNF418 (zinc finger protein 418), a novel
Krüppel-associated box/Cys2His2 zinc finger protein, which was involved in critical regulators for the initiation, maintenance, and development of tumor via cell differentiation, tumor suppression, proliferation and apoptosis [26, 27]. Wang et al [28] revealed that ZNF418 was identified as a direct target of miR-1204 and mediated the function of miR-1204 in HCC cells, which inhibited HCC progression through MAPK and c-Jun signaling pathways and potentially serves as a new prognostic biomarker and therapeutic target for HCC. Thus, further characterization of molecules such as MAGEB2, SUSD1, ZNF382 and ZNF418 will provide a new perspective for the development and progress of LSCC, and aided to find potential therapeutic targets for LSCC patients.
Another prominent finding in our study was that we identify a set of potential small molecule drugs that mimic the expression of normal cellular genes, analyzing the MDGs in CMap database. Small molecule drugs with a highly significant negative enrichment value possessed the potential to alter the gene expression of LSCC, and thus inhibiting the progression of tumors. Simvastatin, as a type of statin, specific inhibitors of 3-hydroxy-3-methylglutaryl coenzyme A (HMG-CoA) reductase, is frequently utilized to treat hypercholesterolemia and prevent cardiovascular diseases. A growing number of researches have reported simvastatin-related potentially beneficial effects in cancers, including prostate cancer, breast cancer, ovarian cancer, and adenocarcinoma [29–32]. Recently, Ma et al [33] revealed that simvastatin induced cell cycle arrest in the G1 phase in C666-1 cells via decreasing the expression of cyclin-dependent kinase 4 (CKD4) and cyclin D1, and increasing p27 expression in C666-1 cells. Simvastatin treatment inhibited protein kinase B and extracellular signal-regulated kinase 1/2 activation, which is a potential chemotherapy agent for nasopharyngeal carcinoma. Thiocolchicoside, a semi-synthetic colchicine derived from plant honeysuckle, is a muscle relaxant and utilized to treat orthopedic disorders and rheumatologic on account of its anti-inflammatory and analgesic mechanisms. Reuter et al [34] demonstrate that thiocolchicoside exerts an effect on anticancer via the NF-κB pathway resulting in inhibition of cyclooxygenase-2 promoter activity and NF-κB reporter activity. However, efficacy and safety of those small molecule drugs (including simvastatin and thiocolchicoside) on LSCC are still not investigated. Hence, it is urgently demanded to verify the effect of these candidate small molecule drugs on treating LSCC in further studies.
Despite the remarkable sense, it is inevitable that limitations also existed in our study. First, we merely extract retrospectively target data (TCGA and GEO datasets) through biological algorithm approaches the absence of fresh clinical samples to screen and verify our results. So, the application of our model remains needed to validate in external and multicenter prospective cohorts with large sample sizes. Second, the MethylMix is an attractive investigative tool to integrate DNA methylation with RNA expression to identify MDGs in cancers. However, the MethylMix focuses on identifying cis-regulatory effects of DNA methylation on gene expression and does not currently model trans-regulatory effects. Further studies are needed to solve the multiple testing challenges on identifying trans-regulatory effects of DNA methylation on gene expression. Third, the 5 MDGs should be further studied and verified to investigate its specific regulatory function and mechanisms in LSCC.