The relationship of 3′UTR HLA‐G14‐bp insertion/deletion and +3142 C/G polymorphisms and soluble HLA‐G expression with gynecological cancers: An updated meta‐analysis

Abstract Objectives Human leukocyte antigen‐G (HLA‐G) is implicated in several cancers and is considered to be an immune checkpoint regulator. We determined the association between polymorphisms in the 3′ untranslated region of HLA‐G and soluble HLA‐G (sHLA‐G) expression with gynecological cancers (GCs). Methods A meta‐analysis was conducted to examine the association between HLA‐G14‐bp insertion (I)/deletion (D) and +3142C/G polymorphism in GC and to evaluate sHLA‐G expression Results We revealed a significant association between the +3142C/G polymorphism and invasive cervical cancer (ICC) based on the allelic model G versus C (odds ratio [OR] = 0.738, 95% confidence interval [CI] = 0.563–0.966, p = 0.027), dominant GG+GC versus CC (OR = 0.584, 95% CI = 0.395–0.862, p = 0.007), and codominant GG versus CC (OR = 0.527, 95% CI = 0.312–0.891, p = 0.017) models, suggesting that the G allele and GG genotype are protective against ICC. In gynecological precancerous patients with human papillomavirus (HPV) infection, we found that the 14‐bp I/D under the codominant DD versus DI model (OR = 0.492, 95% CI = 0.241–1.004, p = 0.051) was of borderline significance. Soluble HLA‐G levels were significantly higher in patients compared with healthy controls (standardized mean differences [SMD] = 1.434, 95% CI = 0.442–2.526, p = 0.005). Stratification by cancer type revealed that the sHLA‐G levels were significantly increased in cervical cancer (SMD = 4.889, 95% CI = 0.468–9.310, p = 0.030) and in subjects of Asian ethnicity (SMD = 4.889, 95% CI = 0.467–9.309, p = 0.030). Conclusions HLA‐G14‐bp I/D and +3142 C/G polymorphisms are associated with GC and HPV‐associated cervical cancer. In addition, we found significantly increased sHLA‐G levels in cancer patients. These results provide a basis for further studies in diagnostics and immunotherapy of GC.


| INTRODUCTION
Human leukocyte antigen-G (HLA-G) is a nonclassical major histocompatibility complex class I antigen 1 encoded by a gene on chromosome 6 at region 6p21.3. 2 Various physiological factors modulates HLA-G secretion in fetal tissues, 3 adult immune-privileged organs, and cells of hematopoietic lineage. 1 Furthermore, it is expressed under pathological conditions, including cancer, viral infection, inflammatory diseases, autoimmune diseases, and transplantation. [4][5][6] In cancer, the expression of HLA-G is heterogeneous and shows high association with an immunosuppressive microenvironment, advanced tumor stage, poor response to treatment, and prognosis. 7,8 The expression levels of HLA-G and its isoforms profiles vary among tumor types, metastasis status, and disease outcome. [9][10][11] HLA-G is significantly expressed in ovarian cancer and may directly inhibit the lysis of NK-92 cells in in vitro experiments. 12 It helps cancerous ovarian cells in their evasion of host immune-surveillance. 12 Thus, it is considered as a potential candidate marker for disease progression. 13 Cervical carcinoma is a common gynecological malignancy and is the fourth most common cause of mortality from cancer among women worldwide. 14 Several studies indicate that the infection by the human papillomavirus (HPV) is a risk factor for the development of cervical cancer. A progressive increase in HLA-G protein expression in the HPV-infected cervix and cervical carcinoma has been reported. 15 The progressive upregulation of HLA-G may be an important factor in the maintenance of HPV conducive to cervical cancer. 15 Both tumor cells and viruses use the similar strategy to evade from the immune response. Increased HLA-G expression was revealed in the immune cells of patients infected with cytomegalovirus 16 and human immunodeficiency virus. 17 HIV infection appears to increase the HLA-G production by naïve T CD8+ cells and also the levels of effector and memory cells. 17 Because of the induction of sHLA-G by the virus, it is intuitive that antiretroviral therapy would downregulate the HLA-G production. 18 Genetic variations in the HLA-G gene, including 14-bp insertion/deletion (I/D; rs371194629) and +3142 C/G (rs1063320), showed significant association with cervical cancer risk 19 and have been linked to the soluble HLA-G (sHLA-G) levels [20][21][22][23][24] and HLA-G messenger RNA (mRNA) regulation. 22,23,25 Interestingly, HLA-G 3′ untranslated region (UTR) alleles were linked to HIV infection in adults 21,26 and in perinatal HIV transmission. 27 HIV-positive women have a higher risk for HPV coinfection, which increases the risk of human cervical cancer. 28 Both HPV infection 29,30 and HPV-associated cervical cancer 31 showed association withHLA-G gene polymorphisms. Although aberrant HLA-G expression has been reported to be associated with advanced tumor stage, metastasis status, and poor disease outcome, some discrepancies remain in various cancer types. 32 The current evidence supports that HLA-G is involved in cancer; however, discrepancies related to cancer heterogeneity and differences in methods remain. This study aimed to further investigate the role of HLA-G in gynecological pathologies by analyzing data from existing published reports.

| Inclusion and exclusion criteria and data extraction
We searched for published studies involving an association between HLA-G polymorphisms and gynecological cancers in the MEDLINE, Embase, and Cochrane databases (up to December 2021) using Medical Subject Headings and keyword combinations, including "HLA-G," "polymorphism," "gynecological cancer," and "neoplasm". Similarly, we searched for published studies involving the association between sHLA-G levels and gynecological cancers by using "sHLA-G," "level," "neoplasm," and "gynecological cancers" as keyword combinations. Furthermore, we identified by a manual search and reviewed additional studies that were not indexed in the MEDLINE, Embase, and Cochrane databases. We considered that studies are eligible if they respond to the following inclusion criteria 1 : evaluation of HLA-G polymorphisms in patients with gynecological cancers and in healthy controls and 2 the availability of mean and standard deviation (SD) data for sHLA-G levels in both patients and controls. When only the median and range were reported, we calculated the mean and SD according to the method of Hozo et al. 33 We excluded studies that (1) included irrelevant, incomplete, and redundant data or (2) were systematic reviews, meta-analyses, or case reports. For each included study, we extracted the following information: first author, publication year, country, cohort ethnicity, allele, and genetic frequencies of the studied HLA-G gene polymorphisms, and mean and SD values of sHLA-G. Two independent reviewers (K. T. and I. Z.) evaluated original studies and extracted the needed data. The metaanalysis was conducted according to PRISMA guidelines. 34

| Statistical analyses
We conducted a comprehensive meta-analysis to test the allelic, recessive, dominant, homozygous, and codominant models for HLA-G gene polymorphisms (14-bp I/D and +3142 C/G). In case of dichotomous data, the odds ratios (ORs) and 95% confidence intervals (CIs) were calculated. In case of continuous data, standardized mean differences (SMDs) were presented. The heterogeneity and variance between studies were evaluated using the I 2 and Tau 2 (τ 2 ). The I 2 values were interpreted according to the Cochrane guidelines. 35 Heterogeneity was quantified using I 2 , which varied from 0% to 100%, and reflecting the proportion of variation among studies. I 2 values of 25%, 50%, and 75% were considered low, moderate, and high, respectively. The random-effects model, testing both sampling errors within the study and variances among studies, is used when heterogeneity among studies is significant (I 2 > 50%). 36 The τ 2 test measures the variance of the true effect sizes and may be used to test the variance of the effect size parameters across studies. 37 Egger's test of the intercept was employed to estimate the sample size effect, and the two-tailed p Egger value was reported. For the Begg and Mazumdar rank correlation test, we calculated a twotailed p-value without continuity correction. The funnel plot is used to measure the study size. 38 Meta-analysis and statistical analysis are conducted using the comprehensive meta-analysis software program (Biostat). A result is considered statistically significant if p < 0.05.

| Studies included in the metaanalysis
We identified 83 eligible studies concerning HLA-G polymorphisms in gynecological cancers. Based on the title and abstract, we selected 42 studies for full-text analysis. Some studies were excluded because they were reviews or meta-analyses 10 and because of unavailable texts 2 or a lack of relevant data. 25 Therefore, a total of five articles met our inclusion criteria. 29,31,[39][40][41] The results are summarized in Table 1 and Figure 1.
For sHLA-G and gynecological cancers, 101 studies were identified. Of these, 31 were selected based on the title and abstract for full-text analysis. Studies were excluded as they were reviews 3 and because of unavailable text, 2 irrelevant data, 12 and lack of data. 9 This resulted in a total of six articles that met our inclusion criteria. [42][43][44][45][46][47] Selected study details are summarized in Table 2 and  indicating that theG allele andGG genotype may represent a protective allele and genotype against ICC. The +3142 C/G polymorphism did not exhibit an association with low-grade squamous intraepithelial lesions (LSIL) or highgrade squamous intraepithelial lesions (HSIL). Both conditions were classified as squamous intraepithelial lesions of the uterine cervix (Table 3). Heterogeneity was moderate to high in the ICC and LSIL groups (I 2 > 50%). In the HSIL group, only two studies were included, and therefore, heterogeneity was not detected (I 2 = 0%). The variance observed in the forest plot reflects a difference in the true effect sizes rather than sampling errors, suggesting calculations in the random-effects model. Egger and Begg'sp-values were not significant (p-value twotailed > 0.05), suggesting the absence of the sample size effect for all genetic models of ICC. A funnel plot was symmetric, revealing the absence of publication bias T A B L E 1 Characteristics of studies included in the meta-analysis: HLA-G 14-bp I/D and +3142 C/G polymorphisms and cervical pathologies.  ( Figure 3B). These results confirm the robustness of the results revealed by this meta-analysis. The 14-bp I/D did not show a significant association in the study models ( Figure 4A, Table 4). We detected substantial heterogeneity, particularly for ICC; however, LSIL and HSIL showed low to moderate heterogeneity. We did not detect a publication bias using Egger and Begg's tests as well as a funnel plot ( Figure 4B).

Study
Haplotype analysis, including 14-bp I/D and +3142 C/G polymorphisms, revealed borderline associations for the Del/G haplotypes in ICC (OR = 0.492, 95% CI = 0.240-1.009, p = 0.053) ( Figure 5A). However, only three studies were included, and more studies and stratification analyses should reveal new and strong associations (Table 5). Heterogeneity was absent for the I/G haplotype but moderate to high for the D/G, D/C, and I/C haplotypes. We did not detect any publication bias using a funnel plot ( Figure 5B).
Stratification by ethnicity revealed that Asian patients have significantly increased sHLA-G levels (SMD = 4.889, 95% CI = 0.467-9.309, p = 0.030; Table 6); however, Caucasians did not exhibit increased sHLA. The plasma/serum samples showed significantly higher sHLA-G levels (SMD = 1.428, 95% CI = 0.142-2.715, p = 0.030; Table 6), whereas peritoneal fluids (ascites) did not show an increase. Because the number of included studies was small (N = 2), more larger cohorts are required to confirm the conclusion of the present study. All calculations were performed with the random-effects model, because heterogeneity was high (I 2 > 75%). The high heterogeneity detected among the studies in the overall analysis (I 2 = 96.5%) could be explained by the heterogeneity of cancers in addition to differences in the sHLA-G detection methods (Table 6). No bias of publication was detected ( Figure 6B).

| DISCUSSION
Gynecological pathologies, particularly gynecological cancers, are multifactorial disorders in which genetic factors constitute strong determinants. HLA-G, a potent inhibitory checkpoint of immune cell function, is highly expressed in gynecological cancers. The membranous HLA-G is significantly expressed in cancers and plays a significant role in their diagnosis and progression. [48][49][50] Moreover, the vesicular-bound HLA-G was associated with a high risk of ovarian cancer progression. 51 To the best of our knowledge, this is the first meta-analysis F I G U R E 1 Flow diagram of the systematic review and meta-analysis literature search results concerning HLA-G polymorphisms in gynecological cancers. HLA-G, human leukocyte antigen-G.
T A B L E 2 Characteristics of studies included in the meta-analysis: sHLA-G and gynecological cancers.  investigating the association of 3′UTR HLA-G gene polymorphisms (14-bp I/D and +3142 C/G) and sHLA-G with gynecological pathologies. By pooling several eligible studies, the meta-analysis provided accurate data by increasing the statistical power and analysis resolution.

Study
In the present study, we evaluated eligible studies that included 1243 patients for HLA-G 14-bp I/D, 1240 patients and HLA-G +3142 C/G, along with 4069 controls. Our data revealed a link between ICC and the +3142 C/G polymorphism. However, squamous intraepithelial lesions classified as precancerous lesions did not exhibit an association. We demonstrated that the G allele and GG T A B L E 3 Meta-analysis of the association between HLA-G +3142 C/G polymorphism and cervical pathologies under random effects model. genotype are likely protective against ICC under different contrast models (allele, dominant, and codominant models). The presence of the G allele was proposed to enhance the affinity of HLA-G mRNA for microRNAs, thus increasing the instability of the HLA-G transcripts, 52 which may explain this protection. Indeed, ICC patients with one or two exemplaryG alleles may exhibit low HLA-G expression and reduced immune tolerance associated with the HLA-G molecule. The meta-analysis conducted for the HLA-G 14-bp I/D did not reveal an association with precancerous lesions or with ICC (Table 4). However, we reported that the Del/G haplotypes exhibited a borderline association with ICC. Del/G is proposed as a potential protective haplotype; however, more larger studies are required to corroborate these findings.
The current meta-analysis conducted on 299 cases and 375 controls revealed increased levels of sHLA-G in all gynecological cancers (SMD = 1.434, 95% CI = 0.442-2.426, p = 0.005) and, particularly, in cervical cancer (SMD = 4.889, 95% CI = 0.468-9.310, p = 0.030). After stratification by ethnicity, the association of increased sHLA-G expression with gynecological cancers was maintained only for the Asian population (p = 0.030). Our findings support sHLA-G as a key biomarker for gynecological cancers. It plays an important role as a powerful immune regulator implicated in immune tolerance and the inhibition of immune cell function.
Although limited publication bias proves the robustness of our analyses, this meta-analysis is the first to investigate HLA-G polymorphisms and sHLA-G expression in gynecological cancers. Our results provide two parallel lines of evidence exploring both the sHLA-G levels and certain HLA-G 3′UTR polymorphisms in patients with gynecological pathologies. HLA-G, which is highly expressed during gynecological carcinogenesis, might be an excellent molecular target for immunotherapy. More studies are needed to decipher the pathways associated with cervix infections and cervical cancer. A recent meta-analysis by Moossavi et al. 53 has investigated HLA-G +3142 C/G genetic variant and the risk of human papillomavirus infection, but they did not found any significant association.
T A B L E 4 Meta-analysis of the association between HLA-G 14-bp I/D polymorphism and cervical pathologies under random effects model. F I G U R E 5 Association between HLA-G Del/G haplotype and invasive cervical cancer (A) forest plot, (B) funnel plot. Forest plot shows the odds ratio and respective 95% confidence intervals for the different studies included in the meta-analysis. For each study in the forest plot, the area of the black square is proportional to study weight and the horizontal bar represents the 95% confidence interval. Z-score, the standardized expression of a value in terms of its relative position in the full distribution of values. CI, confidence interval. Funnel plot assesses the presence or absence of publication bias. HLA-G, human leukocyte antigen-G.
T A B L E 5 Meta-analysis of the Association between HLA-G haplotypes and cervical pathologies under random effects models.   Abbreviations: CI, confidence interval; HLA-G, soluble human leukocyte antigen-G; N, number of studies; NA, not applicable; OR, odds ratio; p Begg , p-value associated to Begg and Mazumdar rank correlation test (two-tailed) without continuity correction; p Egger , p-value associated to Egger's test (two-tailed); p H , p-value associated to heterogeneity; p SMD , p-value associated to SMD; R, random effects model; SEM, standard errors of the mean; SMD, standardized mean differences.
However, more studies still needed to confirm this finding since cancer is a multifactorial disease, and persistent viral infections would increase considerably cancer susceptibility and/or progression. Nevertheless, there are some limitations associated with the present study that require consideration. The first is the limited number of eligible studies available for analysis. Second, there was a poor representation of some ethnicities, such as Africans, Middle Eastern populations, and Americans. Third, we detected significant heterogeneity among the studies and subgroups. This is not surprising since discrepancies among and within different tumor types with respect to sHLA-G expression profiles have been observed in various cancers. 54 The discrepancies may reflect differences between assessment methods, which include a wide variation of experimental protocols.

| CONCLUSION
This meta-analysis of the association between HLA-G 3′UTR polymorphisms and sHLA-G expression with gynecological cancers revealed (i) a significant association of HLA-G +3142 C/G with reduced susceptibility to cervical pathologies and (ii) high sHLA-G levels in patients with gynecological cancers, which support an important role for HLA-G polymorphisms and sHLA-G expression in the pathogenesis of gynecological cancers.