MV1035

ALKBH5 gene is a novel biomarker that predicts the prognosis of pancreatic cancer: A retrospective multicohort study

Backgrounds/Aims: Discovery of new prognostic factors for cases in which the pancreatic cancer scoring and staging system does not result in a clear definition is imperative. We examined the role of Human AlkB homolog H5 (ALKBH5) as a prognostic marker for pancreatic cancer. Methods: Patient data were extracted from the International Cancer Genome Consortium (ICGC) and The Cancer Genome Atlas (TCGA). The prognostic value of ALKBH5 was confirmed via analysis of ALKBH5 and other clinical factors, such as age, sex, and stage, using the time-dependent area under the curve (AUC) of Uno’s C-index, the AUC value of the receiver operating characteristics (ROC) at three years, the Kaplan-Meier survival curve, and multivariate analysis. Results: ALKBH5 showed excellent prognosis prediction in com- parison with existing markers in the two independent cohorts (n=262). Kaplan-Meier survival analysis showed that ALKBH5 expression was positively associated with overall survival (log-rank test, ICGC, p=0.001; TCGA, p=0.01). Notably, comparison of C-index and AUC values in ROC analysis showed that ALKBH5 was associated with high C-index and AUC values compared with other clinical variables (C-index: ICGC, 0.621; TCGA, 0.614 and AUC at three years: ICGC, 0.609; TCGA, 0.558). Multivariate analysis demonstrated that ALKBH5 is an independent prognostic factor (ICGC, p=0.0123; TCGA, p<0.001). Conclusions: These findings contribute to the study of RNA methylation in pancre- atic cancer. We believe that ALKBH5 is a new prognostic marker for pancreatic cancer. (Ann Hepatobiliary Pancreat Surg 2018;22:305-309)

INTRODUCTION
Pancreatic neoplasms are one of the few cancers with a mortality rate approaching 100%.1 Furthermore, only 15-20% of patients are diagnosed during early stages be- cause of nonspecific symptoms and approximately 50% present with distant metastasis at the time of diagnosis.2,3 As the etiology and screening test for this highly lethal disease have yet to be well defined, it is important to identify genetic factors that contribute to the development of this cancer.4N6-methyladenosine (m6A) is the most prevalent epi- transcriptomic modification of the mRNA of higher eukaryotes.5,6 Human AlkB homolog H5 (ALKBH5) is ademethylase that oxidatively reverses several mRNAs, and significantly affects mRNA export and RNA metabolism, as well as the assembly of mRNA processing factors in nuclear speckles. This modification plays an essential role in the regulation of mRNA translation and RNA metabolism. Evidence strongly suggests that epigenetic al- terations, especially methylation induce tumorigenesis- as- sociated cellular changes, suggesting the role of oncogenic mechanisms beyond DNA mutations in cancer.6-9The methylation of related genes and mRNA has been strongly linked to the development and progression of pancreatic cancer. The regulation of the expression of on- cogenic and tumor suppressor genes by demethylation or methylation mechanisms plays a significant role in identi-fied tumorigenesis.

Thus, studies investigating epigenetic changes associated with pancreatic cancer are very im- portant for the development of new diagnostic and ther- apeutic methods.Based on the function of ALKBH5 as an mRNA meth- ylation regulator, we hypothesized that ALKBH5 influen- ces the survival of patients with pancreatic cancer. Therefore, we examined the role of ALKBH5 as a novel prognostic marker complementing the current deficient staging and scoring systems.Primary and processed data, including mRNA ex- pression and clinical information, were downloaded from The Cancer Genome Atlas (TCGA) and International Cancer Genome Consortium (ICGC) in October 2017.10-12 These processes were performed using the cgdsr package of R software version 3.5.0 (The R Foundation for Statistical Computing, 2018). The flow of this study de- scribed in Fig. 1.Survival analyses were performed to predict the overall survival (OS) using three methods: [1] Kaplan-Meier sur- vival curves to evaluate the accuracy of the discrim- ination, [2] Uno’s C-index for the time-dependent area un- der the curve (AUC) analysis, and [3] AUC values in re- ceiver operating characteristics (ROC) at three years. These analyses were performed using R packages survivaland survAUC. C-indices and AUC values ≥0.6 were con-sidered acceptable for survival predictions. Using Kaplan-Meier analyses, we determined the optimal cutoff value with the maximal Uno’s C-index by 5-fold cross-validation. We used univariate and multivariate Cox regression to compare the effect of ALKBH5 on prognosis along with other clinical variables. Statistical analyses were performed using R software.

RESULTS
In total, 172 TCGA and 90 ICGC patients were in- cluded in this study. Of the 172 patients in the TCGAcohort, 112 showed high ALKBH5 expression values and 60 had low ALKBH5 expression values. The average age at diagnosis of TCGA and ICGC was 64.7 years and 65.9 years, respectively. Of the 90 patients in the ICGC cohort, 61 showed high and 29 low ALKBH5 expression. Patient information used in the current study is listed in Table 1.To identify the discriminatory power of ALKBH5 as acategorical variable, we analyzed Kaplan-Meier curves for ALKBH5 gene expression and survival. High expression of ALKBH5 was a significant predictor of overall survival in the TCGA and ICGC cohorts (p=0.01 and p=0.001, Fig. 2A, D). The prognostic value was further confirmed using multivariate analysis (ICGC, p=0.0123; TCGA, p<0.001, Table 2).ALKBH5 C-index and AUC valuesTo evaluate the prognostic value of ALKBH5, we com- pared this gene with other clinical factors, such as age, stage, and sex, in both cohorts. We analyzed gene ex- pression values as continuous variables using Uno’s C-in- dices and AUC values at three years. ALKBH5 showedHigh expression (event)Low expression (event)high C-index values in the two independent cohorts com- pared with the other factors (TCGA: 0.614, ICGC: 0.621, Fig. 2B, E). The three-year AUC value was also sig-nificantly higher than that of the other factors across the two cohorts (ICGC, 0.609; TCGA, 0.558, Fig. 2C, F).

DISCUSSION
In this study, we investigated the survival rate of pa- tients diagnosed with pancreatic cancer based on the ALKBH5 gene expression using TCGA and ICGC databases. Patients in the two cohorts were comparable in age and sex but not significantly. As ICGC is an ongoing project, additional patient information will be released in the future. Survival analysis demonstrated that high ALKBH5 expression showed good discriminatory power compared with age, sex, and stage.Less than 10% of patients diagnosed with pancreatic cancer present with resectable status and the overall5-year survival is <4%.3,13 Patients with small, surgicallyresectable cancers have a realistic chance of a cure and exhibit 5-year survival rates of approximately 20-30%.3,13-16 Although many studies tried to predict the prognosis of patients with pancreatic cancer based on clinical and ge- netic variables, CA-19-9 is the only clinical biomarker ap- proved by the United States Food and Drug Administration.17,18 The poor prognosis of pancreatic can- cer is attributed to its late presentation and lack of accu- rate predictive biomarkers for early diagnosis and prognosis.19 In the current study, ALKBH5 facilitated the classification of patients into two risk groups in two in- dependent cohorts. In addition, it also scored higher C-in- dex compared with stage and age that are important prog- nostic factors. In particular, the AUC value of ALKBH5 remained constant regardless of the follow-up time, which provides a stable prediction of prognosis of the patient.

The past several years have witnessed a vast increase in our knowledge regarding epigenetic features in humancancers.20,21 M6A RNA methylation is regulated by sev- eral molecules such as methylases (METTL3, METTL14, WTAP) and demethylases (FTO, ALKBH5)20-23 that act as oncogenes and/or tumor suppressor genes in various cancers. High expression of ALKBH5, a well-known m6A demethylase, leads to oncogenic progression in glio- blastoma and breast cancer by stabilizing FOXM1, NANOG, and KLF4 mRNA. In breast cancer, hypoxia in the cancer progenitor cells induces high expression of ALKBH5 via activation of hypoxia-inducible factors. Thus, the m6A modification level is reduced, which in- creases the stability and expression of NANOG and KLF4, which are involved in the development of breast cancer stem cells. In the case of glioblastoma, m6A mod- ification modulates methylases such as METTL3 or METTL14, which increases the expression of oncogenes such as KLF4. ALKBH5 regulates FOXM1 gene ex- pression and the altered expression affects GSC tumori- genesis via the FOXM1 axis.20-23 The discovery of new biomarkers via elucidation of these RNA methylation mechanisms is expected to facilitate pancreatic cancer di- agnosis and prognosis by complementing the current chro- mosomal and nucleotide methods.With increasing evidence suggesting the role of epi- genetic alterations especially methylation in triggering tu- morigenesis-associated cellular changes, the field of can- cer research has evolved to incorporate oncogenic mecha- nisms beyond DNA mutations.In particular, the im- portance of m6A in mRNA has attracted attention due to its role in epitranscriptomic modification of cellular differ- entiation and pluripotency. However, the effect of mod- ification of m6A mRNA on the complex process of tu-morigenesis is unclear.

In conclusion, we found that the survival rate of pa- tients with pancreatic cancer was enhanced under high ALKBH5 expression. Despite the unclear mechanism of action of ALKBH5 in tumorigenesis, it has been suggested that ALKBH5 may have a positive effect on the develop- ment and progression of pancreatic cancer. We believe that these findings will MV1035 contribute to the study of RNA methylation in pancreatic cancer.