Polymorphisms in endoplasmic reticulum aminopeptidase genes are associated with cervical cancer risk in a Chinese Han population

Antigen-processing machinery molecules play crucial roles in infectious diseases and cancers. Studies have shown that polymorphisms in endoplasmic reticulum aminopeptidase (ERAP) genes can influence the enzymatic activity of ERAP proteins and are associated with the risk of diseases. In the current study, we evaluated the influence of ERAP gene (ERAP1 and ERAP2) polymorphisms on susceptibility to cervical intraepithelial neoplasia (CIN) and cervical cancer. Six single nucleotide polymorphisms (SNPs) in ERAP1 and 5 SNPs in ERAP2 were selected and genotyped in 556 CIN patients, 1072 cervical cancer patients, and 1262 healthy control individuals. Candidate SNPs were genotyped using SNaPshot assay. And the association of these SNPs with CIN and cervical cancer was analysed. The results showed that allelic and genotypic frequencies of rs26653 in ERAP1 were significantly different between cervical cancer and control groups (P = 0.001 and 0.004). The allelic frequencies of rs27044 in ERAP1 and rs2287988 in ERAP2 were significantly different between control and cervical cancer groups (P = 0.003 and 0.004). Inheritance model analysis showed that genotypes of rs27044, rs26618, rs26653 and rs2287988 SNPs may be associated with the risk of cervical cancer (P = 0.003, 0.004, 0.001 and 0.002). Additionally, haplotype analysis results showed that the ERAP1 haplotype, rs27044C-rs30187T-rs26618T-rs26653G-rs3734016C, was associated with a lower risk of cervical cancer (P = 0.001). The ERAP2 haplotypes rs2549782G- rs2548538A-rs2248374A-rs2287988G-rs1056893T (P = 0.009 and 0.006) and rs2549782T-rs2548538T-rs2248374G-rs2287988A-rs1056893T (P = 0.003 and 0.009) might be associated with cervical cancer and the development from CIN to cervical cancer. Our results indicated that rs27044, rs26618 and rs26653 in ERAP1 and rs2287988 in ERAP2 influenced susceptibility to cervical cancer.


Background
The antigen-processing machinery (APM) is composed of the proteasome, where exogenous and tumour antigens are degraded into peptides; transporters associated with antigen presentation (TAPs), which are responsible for the translocation of peptide precursors; endoplasmic reticulum aminopeptidases (ERAPs), which trim the peptides to fit major histocompatibility complex (MHC) molecules; and MHC proteins, which present antigen peptides on the cell surface [1,2]. Human ERAPs, which belong to the oxytocinase subfamily of M1 metalloproteases, are crucial molecules of the APM. In the endoplasmic reticulum lumen, ERAP1 and ERAP2 trim peptides into their final length to render them suitable for loading onto HLA class I molecules [3,4]. Recently, several studies have shown that ERAP proteins play crucial roles in autoimmune diseases [5,6], infectious diseases [7,8], and cancers [9,10].
Cervical cancer is the fourth most common malignancy in women globally [11]. Persistent human papillomavirus (HPV) infection confers a high risk of cervical cancer [12,13]. Since the HLA class I antigen-presenting system is responsible for the presentation of foreign and cancerous antigens to the immune system [14,15], and ERAPs downregulation was observed in cervical cancer [16,17], therefore, ERAP proteins may play crucial roles in the initiation and development of cervical cancer [18].

Study population
In the current study, a total of 556 patients with CIN and 1072 patients with cervical cancer were enrolled at the Third Affiliated Hospital of Kunming Medical University from May 2014 to August 2018. The inclusion criteria were as follows: 1) diagnosis of CIN or cervical cancer according to Current Diagnosis and Treatment: Obstetrics and Gynaecology and International Federation of Gynaecology and Obstetrics (2009) guidelines; 2) no other malignancy in patients and no history of cancer or other chronic diseases in control individuals; and 3) no preoperative neoadjuvant therapies (including chemotherapy and radiotherapy). The exclusion criteria for patients were as follows: 1) a prior history of primary cancer other than cervical cancer; 2) malignant tumours other than cervical cancer; 3) currently receiving radiotherapy or chemotherapy; and 4) an unclear diagnosis. Over the same period, 1262 healthy women from a health screening project at the same hospital were enrolled as controls.

SNP selection and genotyping
Six SNPs located in ERAP1 and 5 SNPs located in ERAP2 were selected in the current study. The minor allele frequency should be over 0.05 in East Asian population (http://asia.ensembl.org/index.html). The details of the selected SNPs are displayed in Supplementary  Table 1. Venous blood samples were collected for the extraction of genomic DNA, using the QIAamp Blood Mini Kit (Qiagen NV, Venlo, Netherlands). Genotyping of the 11 SNPs was performed using the SNaPshot SNP assay (Thermo Fisher Scientific, Waltham, MA, USA), and results were analysed using GeneMapper TM 4.0 software (Applied Biosystems, Foster City, CA, USA). For quality control, 5% of samples from the case and control groups were genotyped twice with unique analysis serial numbers and the reproducibility was found to be 100%.

Statistical analysis
Hardy-Weinberg equilibrium (HWE) was evaluated to determine the representativeness of the study population. The differences in age among the CIN, cervical cancer, and control groups were analysed using a oneway ANOVA, with a least significant difference test for multiple comparison correction. Allelic and genotypic frequencies of these SNPs were compared between different groups using a Chi-square test and odds ratios (ORs) with associated 95% confidence intervals (CIs) were calculated. Additionally, linkage disequilibrium (LD) was calculated and a D' value greater than 0.80 was considered to indicate LD. The haplotypes among these SNPs were analysed using SHEsis software [35,36]. Subsequently, the distribution of the haplotypes between different groups was compared using a Chi-square test. In addition, inheritance analysis adjusted by age was performed using SNPstats software to identify the relationship between genotypes at these SNPs and cervical cancer [37]. In the inheritance analysis, four inheritance models (codominant, dominant, recessive, and log-additive) were analysed. Simultaneously, Akaike information criterion (AIC) and Bayesian information criterion (BIC) values were calculated to determine the inheritance model with the best fit, i.e. the model with the smallest AIC and BIC values [37]. The statistical power was calculated using Power and Sample Size software (V3.1.2) [38]. Bonferroni correction was performed for multiple comparisons, after which the statistical significance threshold was set at P < 0.0045 (0.05/11). Table 1 shows the clinical data of the subjects in the present study. There was no significant difference in age among the control, CIN, and cervical cancer groups (P > 0.05, F = 1.438), as evaluated by one-way ANOVA. In the CIN group, there were 65 patients with low-grade CIN (I/II) and 491 patients with high-grade CIN (III). In the cervical cancer group, there were 151 patients with adenocarcinoma, 903 patients with squamous cell carcinoma, and 18 patients with other pathological types.

Association of the eleven SNPs with CIN and cervical cancer
All 11 SNPs were in HWE in the control group (P > 0.05) (Supplementary Table 1). The allelic and genotypic frequencies of these SNPs are presented in Tables 2 and 3. The results showed that the allelic and genotypic frequencies of rs26618 (P = 0.021 and 0.016, respectively), rs26653 (P = 0.001 and 0.004), rs27044 (P = 0.003 and 0.012) and rs30187 (P = 0.008 and 0.020) in ERAP1 (Table 2) and rs2248374 (P = 0.014 and 0.020) and rs2287988 (P = 0.004 and 0.007) in ERAP2 (Table 3) were significantly different between cervical cancer and control groups. Additionally, the allelic and genotypic distributions of rs2248374 (P = 0.015 and 0.041, respectively) and rs2287988 (P = 0.014 and 0.039) in ERAP2 were significantly different between CIN and cervical cancer groups (Table 3). However, after Bonferroni correction, only rs26653, rs27044, and rs2287988 were associated with cervical cancer risk (P < 0.0045). The results indicated that, in ERAP1, the G allele of rs26653 may be associated with a lower risk of cervical cancer compared with C allele (OR = 0.829; 95% CI: 0.738-0.930) and the C allele of rs27044 may be a protective factor for cervical cancer (OR = 0.838, 95% CI: 0.746-0.941). Moreover, the G allele of rs2287988 in ERAP2 may be associated with a higher risk of cervical cancer (OR = 1.187, 95% CI: 1.057-1.332). There were no SNPs in ERAP1 (Table 2) or ERAP2 ( Table 3) that exhibited a significantly different distribution between the CIN and control groups or between the CIN and cervical cancer groups after Bonferroni correction (P > 0.0045).

Inheritance model analysis
To evaluate the genotypic association of the 11 SNPs with CIN and cervical cancer, inheritance analysis was performed among cervical cancer, CIN, and control groups (Table 4, Table 5, and Supplementary Tables 2-5). The CC genotype of rs26618 was a risk factor for cervical cancer, compared with TT-CT genotype (P = 0.004; OR = 1.53, 95%CI: 1.14-2.05) in the recessive model (the best-fit inheritance model for the comparison between control and cervical cancer groups) ( Table 4). The 2GG + CG genotype of rs26653 was associated with a lower risk of cervical cancer compared with the CC genotype (P = 0.001, OR = 0.82; 95% CI: 0.73-0.93) in the log-additive model (the best-fit inheritance model for the comparison between control and cervical cancer groups) ( Table 4). The 2CC + CG genotype of rs27044 may be a protective factor against cervical cancer compared with the GG genotype (P = 0.003, OR = 0.84; 95% CI: 0.75-0.94) in the log-additive model (the best-fit inheritance model for the comparison between control and cervical cancer groups) ( Table 4) and the GG-GA genotype of rs2287988 may be a risk factor for cervical cancer compared with the AA genotype (P = 0.002, OR = 1.33; 95% CI: 1.11-1.60) in the dominant model (the best fit inheritance model for the comparison between control and cervical cancer groups) ( Table 5).

Discussion
The immune system is activated by MHC-peptide complexes, after which it eliminates infected and cancerous cells in various ways. The APM plays crucial roles in the initiation and development of various human diseases. As components of the APM, ERAP1 and ERAP2 are important determinants of the repertoire of peptides ultimately presented by HLA class I molecules [39][40][41][42]. Moreover, the SNPs in ERAP genes have been shown to affect the function of ERAPs by changing their peptidome or enzymatic activity [29,30,43]. In cervical cancer, ERAP1 and ERAP2 proteins have been reported to be highly variable, ranging from low to high expression   Note: The statistical significant threshold was set at P < 0.0045 after Bonferroni correction levels [44][45][46]. Although there are inconsistencies among these studies, it is clear that the dysregulated expression of ERAP proteins, which may be induced by ERAP gene SNPs [47,48], is associated with cervical cancer risk.
In 2007, Mehta et al. found that rs27044 in ERAP1 was associated with cervical cancer risk. In the current study, rs27044 was found to be associated with cervical cancer risk (P = 0.003). The C allele of rs27044 (Q730) was found to be a protective factor for cervical cancer (OR = 0.838, 95% CI: 0.746-0.941) ( Table 2), which was consistent with the results of Mehta's study [24]. The SNP, rs27044, a non-synonymous polymorphism, leads to a Q730E substitution in the IV catalysis domain of ERAP1 [33] and may change the substrate length preferences of ERAP1 [49]. Therefore, rs27044 may play a role in cervical cancer by affecting ERAP1 function.
The SNP, rs26618, in ERAP1 leads to an amino acid substitution (I276M) and the current study showed that the CC genotype of this SNP may be associated with an increased risk of cervical cancer (OR = 1.53; 95% CI: 1.14-2.05) compared with TT-CT genotypes (Table 4). In 2016, Guasp et al. reported that I276M (rs26618) may affect the peptidome of ERAP1 by destroying peptides with p2 Ala, unless the p1 amino acid was resistant to ERAP1 trimming [43], which indicated that rs26618 may be associated with cervical cancer. However, in a Netherlands population, Mehta et al.  In 2007, Mehta et al. reported that the C allele of rs26653 in ERAP1 was associated with a higher cervical cancer risk in a Netherlands population [24]. In the current study, the G allele (OR = 0.829; 95% CI: 0.738-0.930) ( Table 2), compared to the C allele, and the 2GG + CG genotype, compared to the CC genotype (OR = 0.82; 95% CI: 0.73-0.93) of rs26653, were associated with lower cervical cancer risk (Table 4). In 2014, Stratikos et al. and Alvarez-Navarro et al. reported that rs26653, which is a non-synonymous polymorphism resulting in a P127R substitution, may be associated with ERAP expression [18,49], and this substitution may also affect the enzymatic activity of ERAP1 in the editing of tumour antigen peptides. This finding may explain the association between rs26653 and cervical cancer risk; however, the mechanisms need to be determined in functional studies.
In the current study, we found an association between rs2287988 in ERAP2, which is responsible for a   (Table 3). Moreover, the GG-GA genotype was associated with an increased risk of cervical cancer (P = 0.002; OR = 1.33, 95% CI: 1.11-1.60) (Table 5). However, association studies of this SNP are rare. Previous studies have found that ERAP2 haplotypes containing rs2287988 affect ERAP2 splicing and expression [50,51]. Thus, additional association studies in different populations are necessary to investigate the role of this polymorphism during the initiation and development of cervical cancer. ERAPs are markedly polymorphic and ERAP haplotypes whose protein products differ at multiple amino acids may affect peptide editing by ERAPs [29,30,52,53]. In the current study, we also analysed haplotypes of ERAP SNPs in LD. The results showed that the ERAP1 haplotype, rs27044C-rs30187T-rs26618T-rs26653G-rs3734016C and the ERAP2 haplotypes, rs2549782T-rs2548538T-rs2248374G -rs2287988A-rs1056893T and rs2549782G-rs2548538A-rs2248374A-rs2287988G-rs1056893T may be associated with cervical cancer risk. These results indicated that SNPs in polymorphic genes may have combinatorial effects on disease susceptibility.

Conclusion
Studies indicated that genetic factors might be correlated with cervical cancer risk [54][55][56], the clinical parameters of cervical cancer [57,58] and the clinical outcome of cervical cancer [59,60]. In the current study, we found that genetic polymorphisms in ERAP1 and ERAP2 genes might be associated with CIN and cervical cancer, and suggested that polymorphisms in key antigen-processing genes could affect susceptibility of cervical cancer. The strength of our study could be we investigated the association of ERAP SNPs with different stages of cervical cancer (healthy individuals, CIN and cervical cancer patients). By contrary, the limitations of our study are that we could not collect more details of the patients' clinical parameters and had no functional verification. Association studies can only provide preliminary results for the correlation between genetic factors and cervical cancer susceptibility, the determination of the SNPs' roles in cervical cancer requires functional studies to be resolved in the future.