PEG1/MEST and IGF2 DNA methylation in CIN and in cervical cancer

Invasive cervical cancer (ICC) is the second most common cancer in women worldwide, with the heaviest burden borne by women in developing countries [1]. In the USA, more than 12,000 new ICC cases and 4,290 ICC-related deaths were reported in 2010 [2]. Incidence was 60 % higher and mortality was more than two times higher in Hispanics and African Americans compared with whites [1], despite comparable screening prevalence across racial/ethnic subgroups and comparable prevalence of infection with high-risk human papillomavirus (HR-HPV) [3]. Current screening guidelines recommend liquid-based cytology screening methods followed by testing for HR-HPV genotypes, since most ICCs evolve slowly and harbor HR-HPV genotypes [4]. However, the sensitivity and specificity of existing screening tests remains suboptimal, leading to false positives and overtreatment which have been linked to subfertility and cervical stenosis [5]. Molecular features that together with known environmental factors can improve prediction of progression of cervical intraepithelial neoplasia (CIN) to ICC, are required.

DNA methylation has been extensively investigated in human cancers and other disorders with the aim of improving present diagnostic tools [6], but appropriate target sequences are unknown. Predictions based on genome-wide approaches have been proposed as reviewed by Wentzensen et al. [7]; however, findings have been difficult to replicate or interpret, because DNA methylation patterns are cell-type specific and may be transient [7]. DNA sequences at differentially methylated regions (DMRs) that regulate imprinted genes (a group of genes that are defined by expression from only a single parental allele) offer key advantages. DNA methylation profiles that regulate these genes are normally stably maintained throughout somatic cell division [8]. In cancer, imprinted DMRs are often deregulated, showing gains or losses in methylation that affect changes in expression and sometimes imprint status [9]. DNA methylation changes can result in either silencing of the only active allele or, alternatively, aberrant activation of the normally silenced allele, which may double the gene dosage [10]. Given that about 80 % of known imprinted genes are found in clusters [8], altered methylation at a DMR can have wide-reaching effects within a cluster of genes.

The most intensively studied genomically imprinted gene, Insulin-like Growth Factor 2 (IGF2), encodes for IGF2, a potent mitogenic growth factor, the expression of which is controlled by DMRs located in the IGF2/H19 locus at chromosome 11p15.5. IGF2 is maternally imprinted and expressed from the paternal allele, while the H19 gene is maternally expressed. At the IGF2/H19 locus, aberrant methylation at DMRs that regulate the expression of IGF2 have been associated with IGF2 deregulation in many cancers, including colorectal, pancreatic and ovarian cancers and Wilms’ tumors [11, 13]. We have also identified a novel consensus binding site for the zinc finger protein, CTCF, located in IGF2 intron 3, and confirmed binding of CTCF to this region. The consensus sequence contains a single CpG dinucleotide whose methylation is strongly positively correlated with expression levels of IGF2 in ovarian cancer (Huang and Murphy, in press). In addition, Paternally Expressed Gene 1/Mesoderm-Specific Transcript (PEG1/MEST) is an imprinted gene located at chromosome 7q32.2 and is expressed from the paternal allele. PEG1/MEST encodes for an enzyme belonging to the α/β-hydrolase fold family [14]. The gene promoter of the active paternal allele of PEG1/MEST is hypomethylated while the transcriptionally silent maternal allele is hypermethylated, regulating the monoalleleic expression of PEG1/MEST. Deregulated methylation and/or expression of PEG1/MEST have been associated with invasive breast cancer [15] and uterine leiomyoma [16] in humans, and growth retardation, defective nurturing and increased lethality in mice [17].

We examine whether aberrant DNA methylation of the PEG1/MEST and IGF2/H19 regulatory regions and of three other imprinted gene DMRs belonging to this imprinted gene cluster: Pleomorphic Adenoma Gene-Like 1 (PLAGL1), Neuronatin (NNAT), Delta-like homolog 1 and noncoding Maternally Expressed Gene 3 (DLK1/MEG3), is associated with CIN and ICC, taking into account HPV infection.

Table 1 summarizes the characteristics of the 234 study participants. In all, 148 women (63 %) were classified as controls (women with neither CIN nor ICC). The remaining 86 women (37 %) were classified as cases; of these, 48 women (21 %) had ICC, 17 (7 %) had either CIN2 or CIN3, and 21 (9 %) had CIN1. The median age was 45 years; women with no cervical abnormalities and those with CIN1 were younger [mean 40.3 years (SD = 9.9) and 35.7 years (SD = 12.2), respectively] than women with cervical carcinoma or high-grade CIN [mean 55.2 years (SD = 12.3) and 44.7 (SD = 9.8) years, respectively]. All women with ICC and CIN2/3 were parous versus 86 % of women with CIN1 and 95 % of controls. Hormonal contraceptive use was more frequently reported by CIN cases (59–76 %) and controls (68 %) than ICC cases (40 %). Almost all (89 %) women with cancer were positive for at least one HPV genotype whereas only a small proportion of controls (14 %) had detectable HPV.

Table 2 summarizes odds ratios (ORs) and 95 % confidence intervals for the associations between IGF2, IGF2 intron 3, H19, PEG1/MEST, MEG3, PLAGL1 and NNAT methylation and CIN or ICC. When compared with controls, a 10 % decrease in methylation at the IGF2 intron 3 region (methylation mean = 71.97, SD = 8.5) was associated with a two-fold increase in ICC risk (OR 2.00, 95 % CI 1.14–3.44), after adjusting for age, gravidity, HIV-1 status, HPV infection, and oral contraceptive use. The association between lower methylation at the IGF2 intron 3 region and ICC risk remained essentially unchanged (OR 1.88, 95 % CI 1.10–3.33) when restricting analysis to women aged 30+ years. A similar pattern of association, albeit weaker, was also observed between a 10 % methylation decrease for the IGF2 intron 3 region and CIN (OR 1.51, 95 % CI 1.0–2.50).

Higher DNA methylation levels at the PEG1/MEST DMR were also associated with ICC (OR 1.44, 95 % CI 0.90–2.35), and restricting analysis to women 30+ years revealed a somewhat stronger association (OR 2.13, 95 % CI 1.10–4.16). Associations between DNA methylation at the IGF2 or H19 DMRs and risk of ICC or its precursor lesions were in a similar direction albeit non-significant. Restricting the analyses to women older than 30 years did not alter these findings. No associations between DNA methylation and ICC, or CIN, were found at PLAGL1, NNAT and MEG3 DMRs (Table 2). The direction and magnitude of these associations remained unchanged when we excluded CIN1 lesions.

To examine whether DNA methylation at the IGF2 intron 3 region, and at the IGF2, H19 and PEG1/MEST DMRs may mediate the association between HPV infection and ICC, we ran mixed models—to allow for unconstrained model entry of individual CpGs at each DMR—to evaluate associations between DNA methylation levels at each of these four DMRs and HPV infection, in which no HPV infection served as the baseline category (Table 3). We found that both HR-HPV and LR-HPV infections were statistically significantly associated with decreased DNA methylation at the IGF2 intron 3 region (β = −8.7, SE = 1.75, p < 0.0001; β = −17.74, SE = 4.70, p = 0.0002, respectively) and at the IGF2 DMR (β = −3.38, SE = 0.85, p < 0.0001; β = −4.88, SE = 2.17, p = 0.02, respectively). HR-HPV infection was associated with higher methylation at the PEG1/MEST DMR (β = 1.75, SE = 0.78, p = 0.02). No associations were found for the H19 DMR.

To determine the functional significance of the association between DMR methylation at regulatory sequences of IGF2, we computed correlation coefficients for DNA methylation and log-IGF2 gene expression from cervical cancer biopsies. We found higher IGF2 expression in samples where aberrant DNA methylation at IGF2 regulatory regions was observed, compared to normal IGF2 methylation at these same regions (r = 0.6, p = 0.001) (data not shown).

We examined the association between DNA methylation at sequences regulating the PEG1/MEST, IGF2/H19, PLAGL1, MEG3 and NNAT imprinted domains and the risk of ICC and its precancerous lesions. After adjusting for age, lifetime gravidity, HIV-1 status, HPV infection, and oral contraceptive use, we found significant associations between altered DNA methylation levels at the PEG1/MEST and IGF2/H19 imprinted domains, and risk of CIN and ICC. Our findings are consistent with the hypothesis that DNA methylation at regulatory sequences of imprinted genes represents susceptibility loci for ICC and its precursor lesions [7]. If confirmed in a larger number of these regulatory regions, these findings also suggest that DNA methylation, at these and perhaps other loci, may be developed for inclusion in risk stratification to identify individuals likely to progress from CIN to ICC.

The age at diagnosis in our Tanzania-based study is comparable to that of women in the USA and other developed countries, where ICC diagnosis tends to be in the fourth or fifth decades of life (SEER Stat Fact Sheets: Cervix Uteri), although this differs by racial/ethnic subgroup. In the USA, the age of peak incidence is 35–44 years in whites but 65+ in African Americans and Hispanics, despite minimal differences in Pap smear screening rates (whites 91.4 %, African Americans 89.7 %) [25]. Age has been shown to influence methylation marks and other regions such as gene promoters but not in DMRs regulating imprinted genes [26], regions previously thought to be resistant to the effects of aging [27]. Our finding of a stronger association in women aged 30+ are consistent with the findings of others [28] and suggests that DNA methylation marks in regulatory sequences of imprinted genes such as PEG1/MEST and IGF2/H19 can be harnessed to improve prediction of CIN cases likely to progress in older women.

Although imprinted gene dysregulation has been shown to contribute to the development and progression of many human cancers [29], DNA methylation in DMRs regulating imprinted genes causally associated with altered expression of multiple genes within imprint gene clusters remains understudied. The most intensively studied imprinted gene, IGF2, has not yet been examined in relation to HPV infection and CIN progression. In the present study, we measured IGF2 transcript levels in women with ICC and found elevated IGF2 levels in samples where decreased methylation at the IGF2 regulatory region was observed. Other studies have reported elevated IGF2 levels among women with cervical cancer and high- but not low-grade squamous intraepithelial lesions when compared with women with no history of abnormal Pap tests. Bulky tumors (>3 cm) are more common in women with higher IGF2 protein levels than in those with smaller tumors (<3 cm) [30]. It is possible that aberrant methylation in some CpGs in the IGF2/H19 imprinted domain such as the intron 3 CTCF site may be influencing loss of imprinting and the exclusive usage of IGF2 promoter 1, inducing IGF2 overexpression, as has been previously shown in cervical carcinomas [30]. Albeit IGF2 DMR aberrant methylation has been linked with other cancers [12, 13], this is the first time that IGF2 intron 3 methylation has been examined in relation to cervical cancer.

Our findings of a strong association between HR-HPV infection and methylation levels support the hypothesis that aberrant DNA methylation may mediate the association between HPV infection and risk of CIN and ICC. If aberrant methylation occurs early in development, it would increase susceptibility to HPV infection and subsequently lesion progression, via elevated expression of these growth effectors—the IGF2 and PEG1/MEST genes.

The study’s main limitation was its small sample size, which required combining CIN1-3 cases even though a sizable proportion of lower-grade CIN cases may not progress. A second limitation is that only a small fraction (<4 %) of Tanzanian women over 30 years of age are nulliparous (National Bureau of Statistics [Tanzania], ICF Macro. 2011), making parity-related sub-analyses difficult. Third, a larger sample size would enable testing the hypothesis of age-dependent DNA methylation in ICC. Finally, PEG1/MEST transcripts were not compared between normal and aberrant methylation as samples were limited; thus, the functional significance of aberrant DNA methylation can be inferred only from previous molecular data. However, the present data suggest that decreased methylation at the IGF2 regulatory region upregulates IGF2 expression in cervical cancer tissue. Our study provides early empirical evidence for associations between DNA methylation patterns in regulatory sequences of imprinted gene and cervical cancer and its precursors. Despite these limitations, we found moderately strong associations between aberrant DNA methylation and ICC and its precursors, independent of known risk factors.

In summary, altered methylation at the PEG1/MEST DMR and at the IGF2 intron 3 CTCF site was associated with increased ICC risk and HR-HPV infection. Larger studies are required to confirm our findings.