Could the methylation of RASSF1A be the potential epigenetic biomarker for nasopharyngeal carcinoma? – Systematic review and meta-analysis

Background: RASSF1A is a tumor suppressor gene. The methylation of RASSF1A has been reported to be associated with the nasopharyngeal tumorigenesis. Aiming to evaluate the association between the RASSF1A gene methylation and nasopharyngeal cancer, and its correlation could be used as an epigenetic biomarker for NPC cancer risk based on meta-analysis. Methods: Relevant articles were identi ﬁ ed by searching MEDLINE database. The frequency and Odds ratio (OR) were applied to estimate the effect of CDH-1 methylation based on random-/x-effects models. Furthermore, subgroup analyses were performed by test-method, ethnicity, source of NPC samples. Results: Total of 16 studies, included 1,766 samples: 1,178 samples from NPC samples, and 588 samples from non-cancerous samples, were enrolled in the meta-analysis. The overall frequency of RASSF1A methylation were 55.98% and 1.70% in case-group and control-group, respectively. By removing the poor relative studies, the heterogeneity was not observed among included studies. The association between the RASSF1A gene methylation and risk of NPC was also conrmed by calculating OR value of 51.43 (95%OR = 28.12-94.08) in x-effects model (Q = 10.63, p = 0.99, I 2 = 0.00, 95% CI = 0.00-0.00). Additionally, the signicant association was also found between the methylation of RASSF1A gene and subgroups. Conclusion: this was the rst meta-analysis provided scientic evidences to suggest the RASSF1A methylation was the potential biomarker for risk of NPC.


Background
Nasopharyngeal carcinoma (NPC), a prevalent nasopharyngeal malignant tumor with the remarkable differences in distribution, gravitating toward Southern Asia [1][2][3][4][5][6][7]. According to the data from Global Cancer Observatory, updated to 2018, 129,079 new nasopharyngeal cases were recorded in the world, of which 72,987 nasopharyngeal death cases were occurred. Even though improvements in nasopharyngeal cancer treatment have been achieved, the diagnosis at an advanced stage led to reduce the success rate of treatment as well as the survival of patients. Thus, the early screening and diagnosis represents the bene cial opportunities to increase the survival of patients as well as the effects of nasopharyngeal cancer treatment. Because of the non-speci c symptoms related to the early stage of NPC as well as the deeply seated location of nasopharynx, it leads to the major obstacle to early screening of NPC [8].
Therefore, effective biomarkers are truly needed [9]. Today, much efforts have been made for identi cation of early biomarkers based on focusing the etiological factors led to the nasopharyngeal tumorigenesis. The progression of NPC is a multiple steps associated with multiple factors, including the infection of Epstein-Barr Virus (EBV), environmental factors, as well as genetic and epigenetics alterations [10,11]. Among them, DNA hypermethylation has been postulated as the best-studied epigenetic alterations in tumorigenesis [12]. The phenomenon of tumor suppressor gene's (TSG) promoter cause the transcriptional silencing, led to gene inactivation, which inhibit the functions of those genes, resulting in the cancer development [13,14].
Many TSGs have been recorded to be involved in the critical cancer-related cellular pathways. Thus, the patterns of TSG promoters' methylation have been studied, and reported to be acted as the potential epigenetic biomarkers for NPC [9,12]. RASSF1A (RAS-association domain family 1 isoform A), located 3p21.31, belonged to the family of RAS effectors, is a tumor suppressor gene coding protein. It functioned as the protein that modulates multiple apoptotic, cell cycle pathways and DNA repair process [15,16]. Increasing evidences demonstrated that the inactivation of RASSF1A due to the hypermethylation, has been reported to associated with the pathogenesis of NPC. However, due to the different sensitivities and intra/inter-assay coe cients of variation of methods, the reported frequency of RASSF1A hypermethylation and its prognostic value is highly variable. Therefore, in current study, we performed the meta-analysis is to summarize the previously studies and identify the diagnosis and early screening value of RASSF1A for NPC.

Search strategy and inclusion/exclusion criteria
We conducted a comprehensive search strategy towards the guidelines of Preferred Reporting Items for Systematics Reviews and Meta-Analyses [17]. By using separation or combination of following keywords: "Nasopharyngeal carcinoma", "methylation", "RASSF1A", "diagnosis", "prognosis", "epigenetic", "hypermethylation", were applied to search reveal published articles in MEDLINE database (Updated on December, 2019). Additional studies were also identified via the references listed in the articles.
Studies were deeply considered eligible only when they met all of the following inclusion criteria: i) The articles were limited to studies written in English; ii) case-control study designed; iii) provided that data about the frequency of RASSF1A methylation as well as the sample size in both case and control group. Exclusion criteria were as follows: i) The articles were written in other languages; ii) abstracts, case reports, letter to editor or unpublished articles were eliminated; iii) studies were related to other tumors and not speci c for NPC; iv) studies lacked vital information for analysis.

Data extraction
The eligibility of each study, the relevant data from the eligible studies were independently retrieved by two authors. Disagreements were resolved through discussion within the third author or our research team. The relevant data were extracted from each study according to the data form, including rst Author's last name, year of publication, country where the study was performed, sample type, experimental methods to assess the methylation of RASSF1A, and number of cases and controls subjects.
Statistical analysis, publication bias and sensitivity analysis MedCalc® software, by MedCalc Software Ltd, was applied to statistically analyze extracted data (https://www.medcalc.org/). The frequency of RASSF1A methylation was calculated in both case and control group. The strength of association between RASSF1A methylation and NPC was evaluated by Odds ratio (OR) with 95% con dence intervals (95%CI). In current study, the heterogeneity among the included studies was estimated by the Cochran Q test and I 2 statistics [18]. The cut-off point: p = 0.05 for the Q test and I 2 were used to test the heterogeneity between studies. The scale of I 2 value is classi ed as following: I 2 < 25%: no heterogeneity, 25% ≤ I 2 ≤ 50%: moderate heterogeneity, and I 2 > 50%: strong heterogeneity [18][19][20]. The random-effects model was applied if the heterogeneity among studies existed (p < 0.05 for Q test, I 2 > 50%). In the case of no between-study heterogeneity, a xed-effects model was applied to compute the pooled ORs. In order to determine the presence of publication bias, the symmetry of the funnel plots in which ORs were plotted against their corresponding standard errors were assessed by the Begg's funnel plot and Egger's test (p < 0.05) indicates statistically significant [21,22]. Additionally, the sensitivity analysis was also performed by sequential omission of individual studies to evaluate stability of the results.

Study characteristics
After exclusion of studies that not met the inclusion criteria, nally, 16 studies, published from 2001 to 2015, included 1,766 samples: 1,178 samples from NPC samples, and 588 samples from non-cancerous samples, were enrolled in the meta-analysis. The characteristics of included studies of RASSF1A methylation and risk of NPC were summarized in Table 1.
The number of NPC samples included studies ranged from 6 to 220 (mean: 42.07 ± 38.90), and number of control samples included studies ranged from 2 to 50 (mean: 21.78 ± 17.38). The patients' ethnicity in 16 studies, comprised of 14 studies from Asian countries: China, Hong Kong, Singapore, Taiwan (counting for 87.50%), and 2 studies from African country: Tunisia (counting for 12.5%). Regarding to the test-method of RASSF1A methylation, twelve studies used MSP (counting for 75.00%), one study used COBRA (counting for 6.25%), one study used RT-qPCR (counting for 6.25%), one study used MMSP (counting for 6.25%), one study used MS-HRM counting for 6.25%).
Meta-analysis: The frequency of RASSF1A promoter methylation, and the association between RASSF1A gene methylation and NPC The differences of RASSF1A gene methylation between the case-group and control-group in 16 studies, including 1,178 samples from NPC samples, and 588 samples from non-cancerous samples, were accessed. Concerning to the hetero heterogeneity between studies, as there was heterogeneity across studies in the case-group (Q = 443.5, p < 0.0001, I 2 = 93.91%, 95%CI for I 2 = 92.23-95.23), and no heterogeneity across studies in the control-group (Q = 9.67, p = 1.00, I 2 = 0.00%, 95%CI for I 2 = 0.00-0.00), thus, the random-effects model and x-effects model were applied to calculate the frequency of RASSF1A gene in case-group and control-group, respectively (Fig. 1, Fig. 2). As the results, the frequency of RASSF1A gene methylation in case-group and control-group were 55.98% (ranged from 0.00% to 100.00%; 95% CI = 42.28-67.35) and 1.70% (ranged from 0.00 to 4.88%; 95% CI = 0.83-3.06), respectively. The metaanalysis result also indicated that the frequency of RASSF1A gene methylation in case-group was signi cant higher than control group.
The methylation of RASSF1A gene was signi cantly associated with an increased NPC risk with a pooled of Odds ratio (OR) of 33.40 (95%CI = 20.22-55.16) based on the x-effect model (Q = 29.56, p = 0.29, I 2 = 12.05%, 95%CI = 0.00-44.45) (Fig. 3). The funnel plot of pooled analysis, which was small asymmetric, indicated that there was still small bias among included studies, therefore publication bias cannot be completely excluded as a factor of in uence on the present meta-analysis. In additionally, the sensitivity analysis found that study by Chang  14, 95%CI = 0.80-21.29) were the relative poorquality studies. Therefore, those studies were omitted to evaluate again the association between the methylation of RASSF1A gene and NPC risk through OR. As the result, the between heterogeneity decreased to I 2 = 0.00% (p = 0.99), indicated that no heterogeneity among enrolled studies was observed. Additionally, the association between RASSF1A methylation and risk of NPC increased, which was indicated by the increased OR of 51.43 (95%OR = 28.12-94.08) in x-effects model (Q = 10.63, p = 0.99, I 2 = 0.00, 95% CI = 0.00-0.00) (Fig. 5) (compared to previously calculated OR of 33.40). Moreover, the funnel plot of pooled analysis, which was quite symmetric, indicated that there was no signi cant bias among the included studies (Fig. 6).
The subgroup analysis by ethnicity, test-method, as well as source of sample were performed. The results of subgroup analysis showed that the heterogeneity totally disappeared in all subgroup (Asian vs non-Asian countries, MPS vs other test-methods, tissue vs other source of samples, I 2 = 0.00, p > 0.05) ( Table  2). With respect to the subgroups categorized by ethnicity, the signi cantly association between RASSF1A methylation and risk of NPC was observed among the Asian country and non-Asian countries in the x-

Sensitivity analysis and publication bias
After removing the relative poor-quality study, the quite symmetric funnel plot, suggested there was no signi cant bias among included studies, was observed (Fig. 6). Additionally, the sensitivity analysis was performed to evaluate the stability and reliability of the conclusions according to the leave-one-out method by excluding one study. As the results, the pooled OR was ranged from 41.95 (95%CI = 22.13-79.49) to 55.06 (95%CI = 29.63-102.32) under the x-effects model within the I 2 = 0.00 (p > 0.05) ( Table  3), which con rmed the stability and reliability of the results. Therefore, the results, and conclusion of present meta-analysis, which was to evaluate the association between methylation of RASSF1A and NPC risk, were stable and reliable.

Discussion
It is important to nd out the effective biomarker for early screening of NPC. Based on the etiological factors of NPC, the methylation of TSGs promoter has been recognized as the common mechanism of the inactivation of TSG, leading to the tumorigenesis of NPC [13,39]. According to Xiong et al. (2004), they identi ed the susceptibility locus at the chromosome 3p21 linked to the NPC tumorigenesis based on the genome-wide linkage analyses of high-risk Chinese NPC patients [40]. RASSF1A, belonged this region, is reported as the tumor suppressor gene. The inactivation of RASSF1A via the methylation has been recognized as highly associated with the progression of NPC [41,42]. Notably that the methylation of TSGs promoter is reported as one of the earlier molecular modi cation during the human epithelial cells transformed to malignancy, and often occurred earlier than the changes of morphology of cancer [43]. Hence, the analysis of TSG hypermethylation, including RASSF1A, may be served as the potential epigenetic biomarker for early screening of NPC. As previous reports, the frequency of RASSF1A methylation in NPC was is highly variable, to determine whether RASSF1A methylation could be served as potential biomarkers for NPC, the meta-analysis and the subgroup analysis by ethnicity, source of samples, and test-methods were performed.
To our knowledge, this was the rst meta-analysis of previous published studies, revealed that methylation of RASSF1A does the increasing of NPC, to evaluate the relationship between RASSF1A promoter and NPC tumorigenesis. In this meta-analysis, sixteen published studies included 1,766 samples: 1,178 samples from NPC samples, and 588 samples from non-cancerous samples, were enrolled. As the results, the overall frequency of RASSF1A methylation in the NPC and control-group were 55.98% (ranged from 0.00% to 100.00%; 95% CI = 42.28-67.35) and 1.70% (ranged from 0.00 to 4.88%; 95% CI = 0.83-3.06), respectively. It could be observed that the individuals with RASSF1A gene methylation was signi cantly associated with NPC based on the calculation of OR (OR = 33.40; 95%CI = 20.22-55.16) based on the x-effect model (Fig. 3). However, the funnel plot as well as the sensitivity analysis found out that small bias among included studies still appeared. Therefore, the relative poorquality studies, including study by Chang et  The heterogeneity between studies decreased to I2 of 0.00 in the x-effects model. The resulting OR is 51.43 with a 95%CI ranged from 28.12 to 94.08, indicating a 51.43-fold increased odds of a positive outcome, and this increase was statistically signi cant at the 5% level. Therefore, the methylation of RASSF1A was signi cantly associated with the NPC tumorigenesis. It suggested the methylation of RASSF1A might play a crucial role in the pathogenesis of NPC. By subgroup analysis, the signi cant association between RASSF1A gene methylation and NPC risk were found among in all subgroups, including ethnicity, source of samples and test-method. Regarding to the ethnicity, a signi cant association between methylation of RASSF1A and NPC was found among the Asian region and the Non-Asia region. Additionally, most studies was performed in Asian region (14 of 16 studies), only two studies was done in Africa, thus, one again con rmed the nasopharyngeal cancer is native and posed to Asian region. The subgroup analysis by source of cancer samples revealed a signi cant correlation in both subgroup: biopsy samples and non-biopsy samples, and no heterogeneity was observed. It indicated that the type of biopsy was more suitable to apply to evaluate the methylation of RASSF1A gene. However, the source of non-biopsy samples, including nasopharyngeal swab, mouth and rinse, plasma -type of non/less-invasive source of sample, could be re ect alterations in the NPC and facility of collecting NPC samples led it a potential biomarker for early screening of NPC. Therefore, it should be focused on in the future to nd out the non/less-invasive biomarker for NPC. The MSP method was used in twelve studies used MSP (counting for 75.00%). It could be explained MSP is the "gold standard method" of evaluation of methylation. The MSP shows the useful tool for the qualitative DNA methylation analysis within the ease of design and execution, sensitivity in the ability to detect small quantities of methylated DNA [44]. Moreover, in which MSP products are run on a gel, and the results are reported as methylated or unmethylated at the target DNA sequence [45]. These results are in accordance with that documented RASSF1A gene methylation to be the common and early epigenetic event in the progression of NPC tumorigenesis. So its epigenetic event might could be used as the potential epigenetic biomarker for risk of NPC. However, the current meta-analysis exhibited some limitations due to the number of current enrolled studies of 16, the data of non-English language studies may contribute to some bias, as well as the evaluation of the correlation between methylation of RASSF1A gene and clinicopathological features, age and the TNM stage which are differences in RASSF1A methylation between cases and controls.

Conclusion
This meta-analysis highlighted the signi cant association between the methylation of RASSF1A gene and risk of NPC based on the evaluation of frequency of both case-group, counting for 55.98% and controlgroup, counting for 1.70%, as well as OR value of 51.43. Additionally, our findings underscore the correlation among RASSF1A gene methylation and all subgroups, including region, source of samples, and test-method. The nding in present meta-analysis emphasized that the methylation of RASSF1A gene was recorded as the early epigenetic event in the progression of NPC tumorigenesis based on the literature-based meta-analysis.

Author contribution
Lao DT conceived, designed, performed, analyzed data, wrote the draft manuscript and acted as primary author for the project, Thieu HH, Nguyen HD contributed to perform and analyze data. Le HAT designed, performed, analyzed data, oversaw the research and edited the article.

Conflicts of Interest
The authors report no con icts of interest. The authors alone are responsible for the content and writing of this article Figure 1 Forest plot of frequency of RASSF1A gene methylation detected in NPC samples  Forest plot of RASSF1A methylation and NPC risk using x-effects model.

Figure 4
Page 17/18 Funnel plot of RASSF1A methylation and NPC risk based on the x-effects model Figure 5 Forest plot of RASSF1A methylation and NPC risk using x-effects model after removed the relative poorquality studies Figure 6