A case–control study was set up in a Brazilian female population to investigate the relationships between HPV infection, prevalence of HPV types, methylation status in the gene promoter of the tolerogenic HLA-G protein and high-grade cervical lesions.
We explored the occurrence of spontaneous de-methylation in the HLA-G promoter as a surrogate of re-expression of the HLA-G protein in HPV-infected cells, as the HLA-G protein is a recognized inducer of a tolerogenic effect and tumor escape from immunosurveillance. By exploring this in a case–control study, our goal was to try and highlight any association of decreased methylation with the carcinogenic process. Indeed, according to the hypothesis of the association with, and role of de-methylation of the HLA-G protein on oncogenic progression, our controls were expected to show high HLA-G methylation, and our CIN2/3 cases were expected to show low HLA-G methylation. We did not include CIN1 in our study, as it is less informative given its high rate of spontaneous regression [29
]. Similarly, a recent publication exploring the association between HLA-G expression and cervical cancer progression also focused on high-grade lesions only [34
We did not consider invasive cervical cancer in the present study, since the occurrence of HLA-G expression in cervical cancer cells is still controversial in the scientific literature. Some authors reported variable HLA-G expression in cervical cancer cells [15
], others very low or no expression [37
], suggesting that if methylation status plays a role in promoting carcinogenesis, it probably acts in the early phases, rather than in the advanced phases of the process. For these reasons we focused our investigation on comparing normal cervical cells with high-grade cervical lesions, in which the carcinogenic process, if it has started, is more frequently active.
The HLA-G gene is silenced under physiological conditions independently from proliferative or differentiative status of normal cells [10
]. Therefore the collection of cervical cells by cytobrush should not have biased the results of our methylation analyses even if many dead epithelial cells were present. As expected, we did find high HLA-G methylation levels in normal cervical cells, which in this context can be considered appropriate controls.
The percentage of HPV-positive CIN2/3 cases was very high, as expected. The low proportion of HPV-negative samples among CIN2/3 cases is consistent with previous reports of HPV DNA-negative CIN2 and CIN3, even though an incorrect histological diagnosis was suspected [41
]. Among controls, HPV positivity was about 20%, which is in agreement with the mean prevalence described in Brazilian populations (10.4%-24.5%) [44
]. Frequency analysis of population characteristics and HPV infection were conducted in controls only, as almost all CIN2/3 cases were HPV-positive. This analysis confirmed the risk factors for HPV infection already described in the literature, including young age, low education level, smoking and a higher lifetime number of sexual partners. Brazilian Mixed, was the ethnic group that showed the highest prevalence of HPV infection, both in CIN2/3 cases and controls.
The analyses of HPV type distribution in our study population showed a slight increase in the prevalence of some types compared to the distribution that has been previously described in Brazil [24
], but were in agreement with the reported prevalence for South America in a worldwide analysis of HPV type distribution [45
To our knowledge, this is the first study on HLA-G methylation and its association with high-grade cervical lesions. We found a high mean percentage of methylation in both CIN2/3 cases and controls, without substantial differences. This is not in line with data reported for some other cancer types. In ovarian cancer, malignant cells were reported to show higher levels of methylation than normal control cells in some CpG sites, even though expression of the protein did not properly correlate with the methylation status [17
]. In renal cancer cells, HLA-G expression via partial de-methylation of its promoter was counted among the strategies used by malignant cells to escape immune response [46
]. Indeed HLA-G expression is widely documented in renal cancer cells, while no expression has been reported in normal renal cells [47
]. Although HLA-G expression has been documented in several other cancer sites, i.e. cervical cancer [15
], melanoma [49
], breast [49
], colorectal [55
], gastric [57
], esophageal [57
], lung [57
], and other cancers [49
], the implications of HLA-G methylation on the expression of the protein have not been described.
It has been suggested that even a single CpG dinucleotide could represent a regulatory sequence highly predictive of the explored outcome [65
]. Thus we also restricted our association analyses to two specific CpGs that might play a regulatory role, one located in the binding site of transcription factor Sp1, and one in the enhancer
region. However, no significant differences were found in methylation between CIN2/3 cases and controls. Evidence of de-methylation events in CIN2/3 cases with respect to controls, which could suggest a re-expression of the HLA-G protein in the cells of cervical high-grade lesions, was not observed. If anything, we found a slight increase in the overall methylation percentage in CIN2/3 cases. This increase was more evident when the analysis was restricted to the CpG located in the enhancer
region, specifically when a threshold was set for the methylation level. Although it seems paradoxical, this could be explained by the concomitant global DNA methylation induced by persistent HPV infection [66
]. Recent studies have suggested that HPV can modulate DNA methylation patterns in order to control cell proliferation. The oncogenic HPV E7 protein can bind DNA methyltransferases, stimulating their activity [67
]. Indeed, many genes have been shown to be hypermethylated in neoplastic cervical lesions [68
]. We cannot exclude that the overlap of these hypermethylation events could overshadow low levels of de-methylation in the HLA-G promoter that may be present, or that may occur earlier in the carcinogenesis process. However the impact of low levels of de-methylation is unlikely to be functionally relevant.
Our findings of low HLA-G hypermethylation in CIN2/3 cases also suggested that alterations in methylation can be detected in the cervical samples of subjects with disease despite contamination of the sample by normal cells. This suggests that the results we obtained in our pilot analysis on HLA-G methylation are sufficiently suggestive of the absence of detectable de-methylation events in the HLA-G promoter, without requiring an extension of the analyses to the entire study population. Realistically, as has been previously reported, other mechanisms like histone modifications [69
], polymorphisms [70
] or miRNA [71
] may modify HLA-G expression.
We compared population characteristics by HPV status and HLA-G promoter methylation, as some characteristics, including ethnicity, have been reported to affect either one or both of them [72
]. If we had found an association, it would have been appropriate to evaluate our results in relation to the demographic and lifestyle characteristics of both cases and controls. However, while we could confirm known associations of some characteristics with HPV infection, we did not find any association with HLA-G methylation; we did not observe significant differences between CIN2/3 cases and controls, nor between HPV-positive and HPV-negative control women.