The association of FOXP3 gene polymorphisms with cancer susceptibility: a comprehensive systemic review and meta-analysis

The role of forkhead box P3 (FOXP3) protein in tumorigenesis has long been controversial and existing data on the association between FOXP3 gene polymorphisms and cancer susceptibility were inconsistent. Here, we conducted a meta-analysis to better clarify the relationship. A comprehensive search of studies published from July 2008 to June 2018 was conducted. The statistical analyses of the pooled odds ratios (ORs) and the corresponding 95% confidence intervals (95% CIs) were performed using the Revman 5.2 software. A total of 12 articles with 19 case–control studies and 10389 participants were included. Three FOXP3 polymorphisms and six cancer types were evaluated. While no significant results were observed in overall and breast cancer groups for rs3761548 (A/C) polymorphisms, the pooled data showed an elevated risk of cancer in variant AA genotypes and A allele for Chinese population (AA vs. AC+CC: OR = 1.61, 95% CI = 1.09, 2.39; AA vs. CC: OR = 1.74, 95% CI = 1.05, 2.89; A vs. C: OR = 1.34, 95% CI = 1.00, 1.78). Neither the overall group analyses nor the subgroup analyses stratified by cancer type and ethnicity proposed any significant association of rs2280883 (C/T) and rs3761549 (T/C) polymorphisms with cancer susceptibility. This meta-analysis suggested that FOXP3 rs3761548 (A/C) polymorphisms were associated with increased cancer risk in Chinese population while rs2280883 (C/T) and rs3761549 (T/C) polymorphisms were not. More large-sample researches with diverse ethnicities and cancer types are needed to draw a concrete conclusion.


Introduction
Tumor microenvironment (TME) plays an important role in cancer suppression and promotion [1]. As a crucial component of TME, regulatory T cells (Tregs) are responsible for down-regulating chronic inflammation, hindering autoimmune reactions, and maintaining peripheral immunological tolerance [2][3][4].
Recently published data demonstrated that Tregs-mediated immunosuppression was a pivotal tumor immune evasion mechanism, and might contribute to the failure of tumor immunotherapy [5,6]. A notable characteristic of Tregs is their expression of the transcription factor forkhead box P3 (FOXP3). The FOXP3 protein, which is encoded from FOXP3 gene located on the X chromosome at Xp11.23, belongs to the forkhead/winged-helix transcription factor family and functions as a transcriptional repressor to down-regulate cytokine production of Tregs [7]. Several studies reported that FOXP3+ Tregs could infiltrate tumors at higher ratios than other T cells. The accumulation of FOXP3+Tregs in tumors and local lymph nodes could inhibit immune responses and thus result in tumorigenesis with a less favorable prognosis [8][9][10]. While Tregs are the major cell type expressing FOXP3 under physiological conditions, it has recently been found that FOXP3 was also expressed in a variety of cancers, such as ovarian, hepatocellular, pancreatic, and thyroid [11][12][13][14]. However, the role of FOXP3 as a tumor suppressor has also been documented. Zuo et al. [15] found that FOXP3 could be expressed in breast epithelial cells but was down-regulated in mammary cancer tissues. Li et al. [16] reported several mutations of this gene in prostate cancer patients and explored the tumor suppressor relationship between the FOXP3 and the Hippo pathways. This reminded us the complex role of FOXP3 and raised the possibility that mutations of FOXP3 gene might cause immune dysregulation and further cancer development.
Genetic variants, mainly composed of single-nucleotide polymorphisms (SNPs), have been proved to cause alterations in protein function in multiple diseases [17]. Several FOXP3 SNPs have been unveiled and their role in cancer susceptibility was explored. For example, Fazelzadeh Haghighi et al. [18] enrolled 312 Iranian participants and reported that T allele in rs3761549 (T/C) was correlated with susceptibility to lung cancer. You et al. [19] studied rs3761548 (A/C) polymorphisms in Chinese population and found the frequency of the A allele was significantly lower in endometrial cancer women than that in healthy controls. Another widely studied polymorphism was rs2280883 (C/T). Zheng et al. [20] recruited 1049 breast cancer patients from multiple centers in China but failed to report a significant correlation between allele C mutation and breast cancer risk.
To solve the controversy, a meta-analysis in 2014 was published and suggested that FOXP3 rs3761549 (T/C) and rs3761548 (A/C) polymorphisms were not associated with the risk of breast cancer, but with the risk of lung cancer and hepatocellular cancer [21]. Unfortunately, it only included five articles with two types of polymorphisms. Since new case-control studies and more polymorphisms were published in recent years, we conducted a comprehensive search of relevant studies with the aim to better clarify the association between FOXP3 polymorphisms and cancer susceptibility.

Literature search
A comprehensive search of studies published from July 2008 to June 2018 was conducted in online databases of PubMed, Medline (Ovid), Embase, CNKI, Weipu, and Wanfang. The following search query was used: 'FOXP3' , 'Forkhead box protein' , 'polymorphism' , 'mutation' , 'variant' , 'cancer' , and 'malignancy' . The search was updated twice a week until 30 June 2018. Language restrictions were not set and references of identified articles were also assessed for inclusion.

Inclusion and exclusion criteria
An eligible study was included if it was consistent with the following criteria: (i) studied FOXP3 polymorphisms in cancer risk; (ii) analyzed the polymorphisms that appeared in at least two independent articles for potential meta-analysis; (iii) provided sufficient data for extraction and calculation; and (iv) case-control studies based on human patients. When duplicated data appeared in different publications, only the most recent one was included. Meanwhile, studies that did not fulfill the above criteria were excluded.

Data extraction and quality assessment
Potential studies were independently reviewed by two investigators (Y.C. and X.Q.). The following information was extracted from both cases and control groups: first author, year of publication, ethnicity, cancer type, SNPs, control type, genotyping method, adjusted parameters, and genotype distributions. Any discrepancies were resolved through a panel discussion until a consensus was reached.
The Newcastle-Ottawa Scale (NOS) was used to investigate the methodological quality of included studies. Three aspects of selection, comparability, and exposure were carefully evaluated. A score of 0-9 was determined and studies of moderate or high quality were included (score above 5). (http://www.ohri.ca/programs/clinical epidemiology/ oxford.asp) [22].

Statistical analysis
To estimate the strength of the association between different FOXP3 polymorphisms and cancer susceptibility, pooled odds ratios (ORs) and corresponding 95% confidence intervals (C95% Is) were calculated. Since SNPs were considered as binary variables of wild (W) and variant (V) alleles, five comparative models were used as follows: recessive model (VV vs. WW), dominant model (VV+ VW vs. WW), heterozygous model (VW vs. VV+WW), co-dominant model (VV+ VW vs. WW), and allelic model (V vs. W) [23]. Heterogeneity in each study was evaluated based on Higgins I 2 test. The random-effects model was applied when I 2 > 50%, indicating the presence of heterogeneity. Otherwise, if the I 2 was less than 50%, the fixed-effects model was used [24]. The Z-test was then performed to determine the significance of the pooled ORs where P<0.05 was illustrated as statistically significant. Eventually, the presence of publication bias was evaluated by visually inspecting the funnel plots. When asymmetry was suspected, Egger's test was performed and P Egger above 0.05 indicating the absence of bias. All statistical analyses were performed using Revman 5.2 software (Cochrane Collaboration, Copenhagen, Denmark) except the Egger's test, which was conducted using STATA 14.0 (StataCorp LP, College Station, TX, U.S.A.).

Search results
As shown in Figure 1, a total of 232 publications were retrieved after the initial research. In our further screening, 82 articles were excluded for duplicity while 104 articles were removed since they were irrelevant to FOXP3 polymorphisms or cancer risk according to titles and abstracts. Amongst the remaining 46 articles for full-text evaluation, 24 articles were biochemical studies, reviews, or meta-analysis; 6 articles studied non-cancer diseases such as autoimmune diseases; 2 articles analyzed several FOXP3 polymorphisms that were not studied in other independent researches, resulting in the impossibility of data pooling; 2 articles failed to offer sufficient data for calculation. Therefore, we enrolled 12 articles in this meta-analysis [18][19][20][25][26][27][28][29][30][31][32][33].

Characteristics of included studies
The 12 enrolled articles consisted of 19 case-control studies with three FOXP3 polymorphisms (rs2280883 in four studies, rs3761548 in ten studies, and rs3761549 in five studies) and six cancer types (breast cancer in ten studies, colorectal cancer in one study, endometrial cancer in one study, hepatocellular in two studies, lung cancer in three studies, and thyroid cancer in two studies). Ethnicities included Brazilian, Chinese, Indian, Iranian, and Israeli. The control sources were population-based in two studies and hospital-based in eight studies (two studies failed to mention details on control type). Different genotyping methods were utilized including PCR-PAGE, PCR-restriction fragment length polymorphism (PCR-RFLP), and MS. The NOS showed that four articles were of high quality (NOS score of 8 or 9) and eight were of moderate quality (NOS score of 6 or 7). Adjusted parameters that might affect the cancer susceptibility were also listed ( Table 1). All studies reported the numbers of corresponding genotypes for both case and control groups as to recessive, heterogeneous, and wild genotypes ( Table 2).

Quantitative analysis
Pooled ORs and corresponding 95% CIs were shown in Table 3 Figure 2A). Thus it was impossible to draw a definite conclusion. Since six out of ten studies were focussed on breast cancer and Chinese  . Amongst the five studies that focussed on rs3761549 (T/C) polymorphisms, three reported increased cancer risk for mutated genotypes while two reported insignificant results. By pooling the 4424 participants together, no significant correlation was found between T/C mutation and cancer risk. The results of subgroup analysis on breast cancer and Chinese population were consistent with the overall group analysis.

Publication bias
The publication bias was visually examined on the funnel plots generated by Revman 5.2 software. No obvious asymmetry could be observed (Figure 3). We further performed Egger's tests in the three analyses that proposed significant association between rs3761548 (A/C) polymorphisms and cancer susceptibility in Chinese population. The results demonstrated no significant publication bias (P>0.05, data not shown).

Discussion
Since alteration of the human immune system can contribute to the development of human cancer, FOXP3 has attracted attention in recent decades as one of the main transcription factors for Tregs, an important participant of immune evasion and surveillance in TME [34][35][36]. While imbalance of FOXP3+Tregs has been widely reported in autoimmune diseases such as allergic rhinitis and Graves' disease, the role of FOXP3 in tumorigenesis has long been controversial [37,38]. FOXP3 is able to repress oncogenes while activating additional tumor suppressor genes. Evidences of this dual role include the down-regulation of MYC and HER2 by FOXP3+ Tregs, and the up-regulation of FOXP3 protein in both Tregs and tumor cells in patients with lung cancer and hepatocellular carcinoma [39][40][41][42].
Polymorphisms of the FOXP3 gene may change FOXP3 protein quantitatively or functionally, thus contributing to predisposition and progression of cancer. To date, several polymorphisms of FOXP3 have been found including rs2280883, rs3761548, rs3761549, rs2294020, rs2294021, rs5906761, rs5902434 etc [19,20,32]. Their roles in cancer susceptibility remain undetermined due to sample size, ethnicity, and other confounding factors. Here, we conducted a thorough meta-analysis with the aim to address the inconsistencies of existing publications and to draw a more concrete conclusion. We enrolled 19 case-control studies with 10389 participants in this meta-analysis. Three types of FOXP3 polymorphisms and six types of cancers were analyzed. The results showed that rs3761548 (A/C) were correlated with cancer susceptibility in Chinese population. The variant AA genotypes and A allele imposed a significant higher cancer risk compared with their counterparts ( [19], both the variant A allele and mutated genotypes (AA plus AC) in rs3761548 showed a statistically significant protective effect on endometrial cancer, which is contrary to the rest of included studies [19]. If we excluded it from the meta-analysis, an elevated risk was again concluded for A/C mutation in overall cancer risk (AA vs. AC+CC: OR = 1.47, 95% CI = 1.07, 2.00). The intriguing results reminded us the controversial role of FOXP3 in cancer immunity, especially in different cancer types. As an X-linked gene, the mutated FOXP3 in females depends on X-chromosome inactivation [43,44]. Whether this rs3761548 position was related to gender-specific cancers or hormone-related cancers such as breast and endometrial cancers remained to be solved. Another relevant point lies in the location of mutated FOXP3 protein in different tumor cells. According to Lopes et al. [31], most TNBC patients had cytoplasmic expression of FOXP3 protein and only some had concomitant perinuclear and/or nuclear expression. The re-localization of FOXP3 protein due to polymorphisms like rs3761548 in certain types of cancer might affect transcription functions and thus cytokine production of Tregs [45].
The results of rs3761549 (T/C) polymorphisms in cancer susceptibility were consistent with the previous meta-analysis [21]. With 4424 participants enrolled, the meta-analysis revealed no significant relationship between rs3761549 (T/C) polymorphisms and cancer risk. However, whether T/C mutations were correlated with increased risk of hepatocellular cancer like previously described remained questionable due to non-repeated researches and limited sample sizes. Specifically, our subgroup analysis stratified by Chinese population indicated a lack of significant association, which was never reported before. Like rs3761548 (A/C), rs3761549 (T/C) polymorphisms were also located in the promoter region of the FOXP3 gene, which is considered to cause mRNA instability thus affecting FOXP3 production and activity [46,47]. The specific reason why the two types of polymorphisms acted differently in case-control studies is not clear, which is a promising subject for future studies. We also explored the role of rs2280883 (C/T) polymorphisms in cancer risk. The meta-analysis failed to draw a significant conclusion, neither in general sample nor in different subtypes. The less aggressive role of rs2280883 (C/T) polymorphisms might be due to the location, which was in intron 9 very near a conserved transcription region of FOXP3 gene. This could cause splicing downstream, resulting in a less functional gene. Further researches are needed to consolidate the exact mechanism [30,48]. Despite our efforts to include all the existing publications, some disadvantages of the present meta-analysis should be notified. First, insufficient published studies were enrolled in this meta-analysis. Although the number of participants was so far the largest, more individual studies were still required to determine a precise conclusion, especially for rs2280883 (C/T) polymorphisms. Second, amongst the 19 studies, only six types of cancer were included. It is known that Tregs play a dual role in tumorigenesis. Whether the tumor promotion effect takes place or the contrary is largely due to the biological characteristics of primary cancer [49]. Thus, caution must be paid when explaining the results to other cancers such as the sex-related and hormone-related cancers. Third, the populations of included studies were Chinese, Iranian, Brazilian, Israeli, and Indian, and many other ethnicities like blacks and Caucasians were ignored. This might affect the overall group analysis since we noticed a different result when Chinese population was stratified for rs3761548 (A/C) polymorphisms. Fourth, our evaluation was based on unadjusted results. Risk factors like body mass index, smoking habit, and menstruation status are also known to be important to tumorigenesis in several types of cancer [50,51]. These confounding factors might cause distorted results. Therefore, studies on more types of cancer are needed to help draw a concrete conclusion.
Taken together, based on current articles in databases, our meta-analysis suggested that FOXP3 rs3761548 (A/C) polymorphisms were associated with cancer risk in Chinese population while no significant correlation was confirmed in rs2280883 (C/T) and rs3761549 (T/C) polymorphisms. To the best of our knowledge, the present study was the most comprehensive one to explore the relationship between FOXP3 polymorphisms and cancer risk. It is also the first to pool the results of rs2280883 (C/T) polymorphisms and conduct subgroup analysis stratified by ethnicity. Considering the limitations mentioned above, more large-sample researches with diverse ethnicities and cancer types are needed to help reach a consensus.