Methylation-Based Signatures for Gastroesophageal Tumor Classification

Abstract

Contention exists within the field of oncology with regards to gastroesophageal junction (GEJ) tumors, as in the past, they have been classified as gastric cancer, esophageal cancer, or a combination of both. Misclassifications of GEJ tumors ultimately influence treatment options, which may be rendered ineffective if treating for the wrong cancer attributes. It has been suggested that misclassification rates were as high as 45%, which is greater than reported for junctional cancer occurrences. Here, we aimed to use the methylation profiles of GEJ tumors to improve classifications of GEJ tumors. Four cohorts of DNA methylation profiles, containing ~27,000 (27k) methylation sites per sample, were collected from the Gene Expression Omnibus and The Cancer Genome Atlas. Tumor samples were assigned into discovery (nEC = 185, nGC = 395; EC, esophageal cancer; GC gastric cancer) and validation (nEC = 179, nGC = 369) sets. The optimized Multi-Survival Screening (MSS) algorithm was used to identify methylation biomarkers capable of distinguishing GEJ tumors. Three methylation signatures were identified: They were associated with protein binding, gene expression, and cellular component organization cellular processes, and achieved precision and recall rates of 94.7% and 99.2%, 97.6% and 96.8%, and 96.8% and 97.6%, respectively, in the validation dataset. Interestingly, the methylation sites of the signatures were very close (i.e., 170–270 base pairs) to their downstream transcription start sites (TSSs), suggesting that the methylations near TSSs play much more important roles in tumorigenesis. Here we presented the first set of methylation signatures with a higher predictive power for characterizing gastroesophageal tumors. Thus, they could improve the diagnosis and treatment of gastroesophageal tumors.

Identification of the subtypes of gastric cancer based on DNA methylation and the prediction of prognosis

“…Furtherly we analyzed the corresponding genes of the independent prognosis-related CpGs by ClueGo and found most genes were enriched in pathways in cancer, hepatocellular carcinoma, and gastric cancer (Fig. 7 a, b), while in the research of Nikolay et al [ 37 ], the identified methylation signatures were associated with protein binding, gene expression, and cellular component organization cellular processes. In the study of Hu et al [ 38 ], they found DNA methylation gene signatures consisting of five genes (SERPINA3, AP00357.4, GZMA, AC004702.2, and GREB1L) as prognosis predictors with AUC of 0.72.…”

“…Although the TCGA research network identified two different CpG island methylator phenotypes (CIMP) of gastric cancer: Epstein-Barr virus-CIMP (EBV-CIMP) and Gastric-CIMP, it did not reveal the relationship between the prognosis and the two methylator phenotypes; furthermore, the TCGA network did not do a systematical analysis to find the key CpGs or establish an efficient model to identify patients with poor prognosis [ 34 – 36 ]. In the research by Nikolay et al [ 37 ], they aimed to find methylation biomarkers to solve the problem of the misclassifications of gastroesophageal junction tumors. For the data preprocessing, they excluded the probes that were not part of the IIumina Human Methylation 27k array probes and used multiple survival screening methodology to find the targeted CpGs such as cg26117023, cg0402816 and cg21475255, which were very close to their downstream transcription start sites (TSSs) [ 37 ].…”

“…In the research by Nikolay et al [ 37 ], they aimed to find methylation biomarkers to solve the problem of the misclassifications of gastroesophageal junction tumors. For the data preprocessing, they excluded the probes that were not part of the IIumina Human Methylation 27k array probes and used multiple survival screening methodology to find the targeted CpGs such as cg26117023, cg0402816 and cg21475255, which were very close to their downstream transcription start sites (TSSs) [ 37 ]. In our study, we aimed to find the methylation sites that could predict the prognosis of patients diagnosed with gastric cancer.…”

“…In our study, we aimed to find the methylation sites that could predict the prognosis of patients diagnosed with gastric cancer. To achieve it, we merged the patients’ prognosis information such as the survival time or survival state with the methylation files and specially examined CpGs in promoter regions (2 kb upstream to 0.5 kb downstream); then, we used the multivariate Cox proportional risk regression models to find the independent prognosis methylation sites and divided the gastric cancer into 7 subtypes; these identified independent prognosis methylation sites were different from the sites in the study of Nikolay’s team [ 37 ]. Furtherly we analyzed the corresponding genes of the independent prognosis-related CpGs by ClueGo and found most genes were enriched in pathways in cancer, hepatocellular carcinoma, and gastric cancer (Fig.…”

Comparison of neoadjuvant regimens for resectable gastroesophageal junction cancer: a systematic review of randomized clinical trials across three decades

“…While the incidence of squamous cell histology has declined in recent years, especially in the United States, the incidence of adenocarcinoma histology continues to increase, perhaps as a result of the increased prevalence of obesity, gastroesophageal reflux disease, and Western diet and lifestyle factors (3). Recently, there has been also better molecular and genetic characterization of gastroesophageal junction (GEJ) tumors over traditional anatomical classification, namely unique DNA methylation signatures, mRNA and microRNA expression patterns (4,5). However, despite improvements in diagnostics and curative and life-prolonging treatments for esophageal and gastric adenocarcinoma, the optimal perioperative treatment remains uncertain for operable adenocarcinoma of the GEJ.…”