Associations of five polymorphisms in the CD44 gene with cancer susceptibility in Asians

CD44 polymorphisms have been previously associated with cancer risk. However, the results between independent studies were inconsistent. Here, a meta-analysis was performed to systematically evaluate associations between CD44 polymorphisms and cancer susceptibility. A comprehensive literature search conducted in PubMed, Embase, and Web of Science databases through August 10, 2016 yielded 11 eligible publications consisting of 5,788 cancer patients and 5,852 controls. Overall, odds ratios (OR) calculated with 95% confidence intervals (CI) identified a significant association between CD44 polymorphism rs13347 and cancer susceptibility under all genetic models. Additionally, the minor allele of polymorphism rs11821102 was associated with a decreased susceptibility to cancer in allele contrast, dominant, and heterozygous models, while no significant association was identified for polymorphisms rs10836347, rs713330, or rs1425802. Subgroup analysis by ethnicity revealed rs13347 was significantly associated with cancer susceptibility for Chinese but not for Indians. Linkage disequilibrium (LD) between different polymorphisms varied across diverse ethnic populations. In conclusion, the results indicate that CD44 polymorphism rs13347 acts as a risk factor for cancer, especially in Chinese, while the minor allele of polymorphism rs11821102 may be associated with a decreased susceptibility to cancer. Nevertheless, further studies on a larger population covering different ethnicities are warranted.

performed on Chinese also revealed a similar positive correlation with statistical power in acute myeloid leukemia, nasopharyngeal carcinoma, and colorectal cancer [14][15][16][17] . However, three other studies on Chinese, failed to demonstrate any statistically significant associations [18][19][20][21] . Furthermore, two studies conducted on Indians demonstrated that no significant association existed between CD44 rs13347 polymorphism and risks for gallbladder or breast cancers [22][23][24] . Here, a comprehensive meta-analysis was performed to derive a more precise estimation of the relationship between CD44 polymorphisms and the susceptibility to cancer. Eligible studies included diverse forms of cancer and two ethnicities, Chinese and Indian. These results have potential implications for the basis of cancer development as well as for functional consequences of allele specific expression of CD44.

Materials and Methods
Search strategy. PubMed, Embase, and Web of Science databases were searched for studies reporting association of CD44 polymorphisms with cancer risk in any type of cancer performed up to the date of August 10, 2016. Gene-specific terms (CD44) were combined with polymorphism-(polymorphism or polymorphisms or variation or variations or variant or variants or mutation or mutations or genotype or genotypes) and disease-specific terms (cancer or cancers or tumor or tumors or neoplasm or neoplasms). The search was performed by the method of free-text word combined with the Medical Subject Headings (MeSH) and included the following search terms: "Neoplasms", "Antigens, CD44" and "Polymorphism, Single Nucleotide". A thorough review of all referenced materials within retrieved studies was also performed in order to identify additional potentially eligible studies.
Inclusion criteria and exclusion criteria. Inclusion criteria for eligible studies were the following: (1) case-control or cohort study design; (2) sufficient data for the evaluation of CD44 SNPs in cancer risk; (3) articles published in English; (4) studies performed on humans; and (5) in the case of multiple publications from the same study group, the most complete and recent results were used.    The exclusion criteria were defined as the following: (1) abstracts, reviews and animal studies; (2) useless reported data such as genotype number or frequency was not included; and (3) articles published in languages other than English.

TagSNPs selection. Previous studies performed bioinformatics analysis with Haploview software 4.2 to
analyze the haplotype block based on the Chinese Han Beijing population data of HapMap (HapMap Data Rel 27 Phase II + III, February 2009, on NCBI B36 assembly, dbSNP b126). Six tagSNPs were found to cover all the potential functional common SNPs (MAF > 0.05) in the CD44 gene: rs8193, rs11821102, rs10836347, and rs13347 in the 3′ UTR, rs1425802 in the promoter and rs9666607 in exon region. CD44 rs8193 and rs13347 were in high linkage disequilibrium (LD) (D' = 1.0, r 2 = 0.527), so the selection of rs13347 was sufficient to represent the two SNPs. In addition, due to the difficulty in genotyping rs9666607 using MALDI-TOF, the polymorphism rs713330, which is in complete LD with rs9666607 (D' = 1.0, r 2 = 1.0) was chosen to replace it 13,15,17 . Data extraction. Two reviewers (Q.C.Q. and J.W.W.) independently selected studies and extracted the following data from each study: first author's surname, publication year, ethnicity, cancer types, numbers of cases and controls, and genotype distributions of cases and controls. The results were compared and disagreement was resolved through discussion until consensus was reached. Study design was stratified into population-based studies and hospital-based studies.
Quality assessment. The Newcastle-Ottawa Scale and Agency for Healthcare Research and Quality (http:// www.ohri.ca/programs/clinical_epidemiology/oxford.asp, maximum score = 9 points) was used to evaluate the methodological quality, which scored studies by the selection of patients, the comparability of the groups, and the quality of the sampling process. Studies awarded a score of 0-3, 4-6, or 7-9 were considered as low-, moderate-, or high-quality studies, respectively.  Statistical analysis. The meta-analysis assessed associations between polymorphisms and cancer risk under allele contrast, dominant, recessive, homozygous, and heterozygous models. Odds ratios (ORs) with 95% confidence intervals (CIs) were used to estimate the strength of associations. Heterogeneity among the included studies was evaluated using the Chi square-based Q statistic. The fixed-effect (Mantel-Haenszel method) or random-effects model (DerSimonian-Laird method) was used to calculate pooled effect estimates in the presence (P < 0.05) or absence (P > 0.05) of heterogeneity. Sensitivity analysis was conducted by sequentially excluding one study at a time. The Begg's test and the Egger's test were performed to assess publication bias. If any possible bias was observed, the trim and fill method was used to identify and adjust for those studies. Data analysis was carried out using Stata software, version 11.0 (Stata Corporation; College Station, TX, USA). P-values < 0.05 were considered to be statistically significant.

Results
Study characteristics. The literature search yielded 15 studies that were carefully reviewed based on the inclusion criteria 13-24,26-28 . Among these articles, the following studies were excluded: 2 studies that did not investigate the association between cancer risk and the polymorphism of interest 27,28 ; 1 study that did not offer sufficient raw data 26 ; and 1 study in which the data was repeated in a second publication with a larger population conducted by the same study group 22,24 . Thus, 11 publications were eligible for the meta-analysis for a total of 5,788 cancer patients and 5,852 controls and 9 different cancer types. The main characteristics of the eligible studies are listed in Table 1. All studies scored a value of 7 or above (high-quality) as determined in the Newcastle-Ottawa Scale, and individuals therein were of Chinese or Indian descent. Two publications that contained the same population of controls from the same research group were considered as one case-control study 18,19 . The following is the breakdown of the number of studies, including cancer types, cases and controls, that met our eligibility criteria for each polymorphism evaluated: rs13347 polymorphism, 10 case-control studies including 9 different cancer types with 5,788 cases and 5,852 controls; rs11821102 polymorphism, 7 case-control studies including 7 different cancer types with 4,686 cases and 4,893 controls; rs10836347 polymorphism, 6 case-control studies including 6 different cancer types with 4,411 cases and 4,618 controls; rs713330 polymorphism, 5 case-control studies including 6 different cancer types with 3,453 cases and 3,397 controls; and rs1425802 polymorphism, 5 case-control studies including 6 different cancer types with 3,453 cases and 3,397 controls (Table 1).

Quantitative synthesis.
The results of association of CD44 SNPs with a general risk for cancer including data for all individuals are shown in Table 2. Significant association between CD44 polymorphism rs13347 and cancer susceptibility was observed under all genetic models (  (Fig. 2). However, no significant association was identified for polymorphisms rs10836347, rs713330, or rs1425802 and overall cancer susceptibility ( Table 2).
For rs13347, significant between-study heterogeneity was present in all genetic models; therefore, a random-effects model (Der-Simonian Laird) was used. For all other polymorphisms, no significant heterogeneity was found in any genetic model. A fixed-effect model (the Mantel-Haenszel method) was thus applied.

Subgroup analysis.
For the rs13347 polymorphism, the studies consisted of individuals of Chinese or Indian descent. However, for other polymorphisms, individuals from all eligible studies were only of Chinese descent, and the number of studies for each cancer type was less than two. Therefore, stratified analysis on the basis of ethnicity, source of controls, and cancer type was only performed on rs13347 (Table 3).
In an ethnicity subgroup analysis, polymorphism rs13347 was significantly associated with cancer susceptibility for Chinese: allele contrast (T vs. the stratification analysis was conducted based on cancer type, we uncovered that the T allele was significantly associated with an increased susceptibility to colorectal cancer, nasopharyngeal carcinoma, and acute myeloid leukemia in all genetic models, and an increased risk for breast cancer only in recessive and homozygous models. Sensitivity analysis. In order to examine the influence of individual data sets on the pooled ORs, single studies were sequentially excluded from analysis. Pooled ORs were persistent, indicating that the results were statistically stable and robust (data not shown). Sensitivity analysis of the rs13347 polymorphism in an allelic comparison and the rs11821102 polymorphism in the heterozygous model are presented in Fig. 3.
Publication bias. The Begg's and the Egger's tests were performed to quantitatively evaluate publication bias of the included studies. Asymmetry was not observed in the funnel plots of any of the polymorphisms. The funnel plots for the rs13347 polymorphism in an allelic comparison and the rs11821102 polymorphism in the heterozygous model are presented in Fig. 4, and all P-values from the two tests are listed in Table 2. These results were consistent with the absence of significant publication bias in the meta-analysis except in the case of polymorphism rs10836347. For this polymorphism, publication bias was apparent in the recessive and homozygous models. However, analysis with the trim and fill method demonstrated that the results of our study did not significantly change even after adjusting for the publication bias. LD analysis across populations. In order to investigate the relationship between polymorphisms in the different populations, LD analysis was performed to test for the existence of bins in the region containing the seven CD44 polymorphisms evaluated, including rs10836347, rs11821102, rs13347, rs1425803, rs713330, rs8193, and rs9666607 (Fig. 5). Relatively low LD values between rs13347 and rs9666607 polymorphisms (D' = 0.709, r 2 = 0.07) were calculated for ASW populations, while complete LD (D' > 0.90) was observed for the other ethnic populations examined. The rs713330 polymorphism was in complete LD with rs9666607 (D' = 1.0, r 2 > 0.80) for all populations. LD plots for the ASW population yielded a high LD value between rs13347 and rs11821102 polymorphisms, moderate LD values for CEU, FIN, GIH, ITU, and TSI populations, and lower LD values for CHB, CHS, GBR, and JPT populations.

Discussion
The genetic basis for the development of cancer involves inherited as well as somatic components. While familial predispositions to cancer have been more easily identified, those genetic factors more generally affecting  populations in conjunction with environmental factors have been more difficult to identify. Thus, the focus has become polymorphisms in well-known cancer associated genes, such as CD44. The meta-analysis performed here with data pooled from 5,788 cases and 5,852 controls supports previous findings that a significant association between CD44 polymorphism rs13347 and cancer risk exists under all genetic models. Thus, the T allele emerged as a risk marker for cancers, especially in Chinese, which is a finding consistent with the results of most of the included studies.
The full-length CD44 gene is located on chromosome 11 and consists of 20 exons and 19 introns 1,5 . Transcripts for the CD44 gene undergo complicated alternative splicing, which results in many functionally distinct protein isoforms, such as CD44 standard (CD44s) and CD44 variant isoforms (CD44v), which account for the heterogeneity in this protein family 1,4,5 . It is widely accepted that CD44 is expressed as multiple transcriptional variants, but the 3′ UTR, which is ~3000 nucleotides long, is in fact relatively conserved 2,14 . The CD44 polymorphism rs13347 has been the subject of intense investigation in cancer susceptibility studies primarily due to its location in the 3′ UTR of the CD44 gene/transcripts 10,13,14,16,17 . The polymorphism rs13347 may lead to alterations in mRNA and/or protein levels of CD44 due to the change of the C and the T allele. In transient transfections using a construct in which the reporter gene is modulated by the CD44 3′ UTR, the presence of the rs13347 T allele led to increased transcriptional activity relative to the C allele 13,[15][16][17] . Similarly, by immunohistochemistry and Western blotting, carriers of the rs13347 variant genotypes (CT and TT) were shown to exhibit dramatically increased CD44 than CC carriers in breast cancer, acute myeloid leukemia, nasopharyngeal carcinoma, and colorectal cancer 13,[15][16][17] . Additionally, functional analyses demonstrated that hsa-mir-509-3p binds and negatively regulates the transcription of CD44 more strongly in C allele carriers than in T allele carriers 13,[15][16][17] . These results support previous studies that have implicated hsa-mir-509-3p as a tumor suppressor gene in the development of human cancer [29][30][31] .
In addition, when the stratification analyses were conducted based on ethnicity and source of control for the rs13347 polymorphism, we identified a significant association in all the genetic models for Chinese but not Indians, only in population-based studies. Moreover, we performed LD analyses to determine the LD between the related polymorphisms. The rs713330 polymorphism was in complete LD with rs9666607 for all populations. Similarly, the high LD values between rs13347 and rs9666607 polymorphisms were observed in all populations except the ASW population. LD plots for the ASW population yielded a high LD value between rs13347 and Scientific RepoRts | 6:39485 | DOI: 10.1038/srep39485 rs11821102 polymorphisms, for CEU, FIN, GIH, ITU, and TSI populations, moderate LD values, and for CHB, CHS, GBR, and JPT populations, lower LD values. The limited number of studies and sample size may account for this discrepancy; therefore, future studies are necessary to verify these results.
None of the eligible studies for CD44 polymorphism rs11821102 indicate a positive correlation with statistical power associated with cancer; yet pooled analysis demonstrated that the minor allele of rs11821102 polymorphism might be related to a decreased cancer risk in Chinese. The polymorphism rs11821102 is also located in 3′ UTR of the CD44 gene 13,17 ; however, no functional assay has been performed to explore how this polymorphism influences CD44 expression or function to date. The limited number of studies and sample size may account for this discrepancy; therefore, larger, preferably population-based case-control studies, as well as mechanistic studies involving CD44 polymorphism rs11821102, are warranted to validate our findings and to further investigate the role of CD44 in the development of cancer.
Although we have conducted a comprehensive retrieval for all eligible studies and polymorphisms, which markedly expanded the sample size and revealed some associations not observed in previous work, some limitations in this study exist. First, statistically significant heterogeneity was confirmed in rs13347 but not in other polymorphisms within all genotype models, and stratified analysis was used to explore the source of heterogeneity. Subgroup analysis by source of controls demonstrated that the between-study heterogeneity was still significant. When subgroup analysis was performed by ethnicity, the heterogeneity was dramatically reduced in Indians but not in Chinese. However, stratified analysis based on cancer type demonstrated that the heterogeneity was not significant in studies performed on nasopharyngeal carcinoma under all genetic models or on breast cancer under recessive and homozygous models. Therefore, the between-study heterogeneity might be mainly attributed to various cancer types. Second, our analysis was limited to individuals of Asian descent, mainly Chinese. LD between different polymorphisms varied across diverse ethnic populations. It, therefore, remains uncertain as to whether the result can be generalized to other ethnic populations.
To the best of our knowledge, this is the first meta-analysis to date to investigate the associations of CD44 polymorphisms with cancer risk. In summary, our study suggests CD44 polymorphism rs13347 can serve as a risk factor for cancer, particularly in Chinese. Our findings also indicate that the minor allele of rs11821102 polymorphism may be associated with a decreased susceptibility to cancer in Chinese.