Fine-scale mapping of chromosome 9q22.33 identifies candidate causal variant in ovarian cancer

Introduction

Ovarian cancer is one of the most common gynecological cancers and is the second most fatal gynecological malignancy in Chinese women (Sung et al., 2021). As a complex polygenic disease (Dareng et al., 2022; Stewart, Ralyea & Lockwood, 2019), genetic factors play a significant role in ovarian cancer etiology (Jones et al., 2017). So far, over 40 susceptibility loci of ovarian cancer have been identified by genome wide association studies (GWAS) (Glubb et al., 2020; Kar et al., 2018; Lawrenson et al., 2019; Manichaikul et al., 2020). However, the mechanisms through which the most of these loci contribute to the ovarian cancer carcinogenesis remain unclear. A functional study was conducted at 9p22.2 locus, which was the first and most significant risk locus for ovarian cancer identified through GWAS in European women. Buckley et al. (2019) identified multiple candidate causal variants (CCVs) at the locus and these variants may be mediated by changes in a transcriptional regulatory network of several regulatory elements that act on BNC2, potentially influencing ovarian cancer susceptibility.

In our previous GWAS, we determined that rs1413299 at 9q22.33 was significantly associated with the risk of epithelial ovarian cancer (EOC) in Chinese women (Chen et al., 2014). SNP rs1413299 is located at intron six of COL15A1, which encodes the alpha chain of collagen type XV (COL15). COL15, a member of the FACIT (fibril-associated collagens with interrupted triple helices) family, is a proteoglycan closely associated with the basement membrane, which has a broad tissue distribution (Bretaud et al., 2020; Clementz & Harris, 2013). COL15A1 gene expression levels were significantly different between ovarian cancer tumor tissues and normal tissues in The Cancer Genome Atlas (TCGA) data (Chen et al., 2014).

However, the effect of the variants in this region on the occurrence and development of ovarian cancer is still unclear. To identify the potential causal variants and additional novel variants at 9q22.33, we conducted a target sequencing followed by a validation study in a large sample size. We also performed a preliminary functional exploration to investigate the molecular mechanism of the susceptibility region.

Materials and Methods

Study subjects were recruited from the Tianjin and Guangdong provinces in China. Cases were newly diagnosed and histologically confirmed to be epithelial ovarian cancer. Controls were selected from healthy and cancer free women who underwent regular health check-ups or participants in the cancer screening program. Informed consent was obtained from each subject when recruited. This study was approved by the Medical Ethics Committee of Tianjin Medical University Cancer Institute and Hospital (Approval number: bc201810). 200 cancer free women were selected for target sequencing and a total of 1,099 epithelial ovarian cases and 1,591 age matched controls were included in the association study.

Results

Association analysis of the variants and ovarian cancer risk

Ten variants were genotyped in 1,099 EOC cases and 1,591 healthy controls (Table S5) using the Sequenom iPlex platforms. We identified that rs7027650 (P = 1.91 × 10⁻⁷, OR =1.31, 95%CI [1.21−1.41]) and rs1889268 (P = 3.71 × 10⁻², OR =1.14, 95%CI [1.02−1.26]) were significantly associated with ovarian cancer risk in additive model (Table 1 and Table S6). The association between rs7027650 and ovarian cancer risk was also significant in both the dominant (P = 5.90 × 10⁻¹⁰, OR =1.80, 95%CI [1.62−1.98]) and recessive models (P = 4.88 × 10⁻³, OR =1.26, 95%CI [1.10−1.42]). For rs1889268, a significant association was also observed in recessive model (P = 2.32 × 10⁻², OR =1.20, 95%CI [1.04−1.35]). We then performed stratification analysis by histological subtype, finding that rs7027650 was significantly associated with three major histotypes of epithelial ovarian cancer, (Serous OC: P = 1.33 × 10⁻³; Endometrioid OC: P = 5.57 × 10⁻⁶; Mucinous OC: P = 1.27 × 10⁻³). Additionally, we found that rs1889268 was significantly associated with Serous OC (P = 4.99 × 10⁻²), and rs1413298 was associated with Endometrioid OC (P = 8.72 × 10⁻³) (Fig. 1 and Table S7)

Table 1:

Association results of SNPs at 9q22.33 in validation stage.

SNP	Positiona	Alleleb	RAF	Genotypes	Case/Control	OR (95% CI)	P
rs7027650	101,741,969	T/A	0.57	AA	182/422	1.00(reference)
				TA	450/569	1.82 (1.47–2.26)	3.44E−08
				TT	419/522	1.86 (1.50–2.31)	1.69E−08
				Addictive		1.39(1.24–1.55)	1.91E−07
				Dominant		1.80(1.62–1.98)	5.90E−10
				Recessive		1.26(1.10–1.42)	4.88E−03
rs1889268	101,767,961	T/C	0.32	CC	468/755	1.00 (reference)
				TC	488/656	1.20 (1.02–1.41)	3.16E−02
				TT	115/153	1.21 (0.93–1.58)	1.61E−01
				Addictive		1.13(1.02–1.25)	3.71E−02
				Dominant		1.11(0.85–1.37)	4.23E−01
				Recessive		1.20(1.04–1.35)	2.32E−02

DOI: 10.7717/peerj.16918/table-1

Notes:

CI

confidence interval

OR

odds ratio

RAF

risk allele frequency

SNP

single-nucleotide polymorphism

aPosition is given with respect to genome build 37.

bRisk allele/other allele.

Functional annotation and eQTL analysis of candidate causal variants

We annotated these SNPs using publicly available data from the Roadmap project (DNase-Seq data and Chip-seq profiles for H3K27ac and H3K4me1 histone modification of primary ovary tissue) and in-house ATAC-seq profiles in three ovarian cancer cell lines (OVCA432, DOV13, and SKOV3). We found that rs7027650, the most significant variant, overlapped with DNase hypersensitivity peaks in normal primary ovarian tissue which represent the open chromosome region. ATAC-seq analyses showed that rs1889268 and rs1413298 resided in open chromatin regions in ovarian cancer cells (Fig. 2A).

To identify candidate functional variants, we then performed expression quantitative trait locus (eQTL) analysis in 102 ovarian tumor tissues from Chinese women. We found that the allele T of rs7027650 was significantly associated with the decreased mRNA expression of COL15A1 (P = 0.009) (Fig. 2B and Fig. S2). However, there was no significant association between the other two variants and expression of the COL15A1 gene (rs1889268: P = 0.072; rs1413298: P = 0.193). We have also explored the GTEx project data, we found that there were several variants associated with gene expression. The most significant variant, 7027650, was associated with COL15A1 expression (P <  0.05) in Thyroid and Nerve-Tibial tissues. However, the significant correlation was not observed in normal ovary tissues (Table S8).

Validation of the candidate causal variants

We performed the luciferase assay and EMSA to explore the possible function of potential variants. In order to explore whether the variant interacts with the promoter of COL15A1, we constructed luciferase reporter vectors containing the rs7027650-A allele or rs7027650-T allele and transfected these plasmids into the SKOV3 and 293T cell lines, respectively. Compared to the fragment containing allele-A, the transcription activity of COL15A1 in the fragment with allele-T was lower in both SKOV3 and 293T cells (Fig. 3A). The results suggested that the gene fragment containing rs7027650 could interact with the promoter region of COL15A1 and may reduce its activity. Next, we performed EMSAs using the probes containing allele A, allele T, and controls. The results showed differences in the allele-specific protein binding capacity of rs7027650 (Fig. 3B). These results indicate that rs7027650 may be a potential functional SNP.

Discussion

In this study, we conducted targeted sequencing of 130 kb on the COL15A1 gene at the 9q22.33 region and subsequently validated the results in 1,099 EOC cases and 1,591 controls. We identified an unreported genetic variant, rs7027650, that was significantly associated with ovarian cancer risk and significantly correlated with the expression of COL15A1 in eQTL analysis. Preliminary molecular analysis revealed that rs7027650 may regulate the expression level of COL15A1.

In the previous three-stage GWAS study of ovarian cancer in Han Chinese women, we discovered two new susceptibility loci associated with the risk of epithelial ovarian cancer (Chen et al., 2014). SNP rs1413299 at the 9q22.33 region was the most significant SNP in this GWAS study. However, the functional variant at this locus and the molecular mechanism accounting for the risk of ovarian cancer were still unknown. Therefore, we performed fine-scale mapping, validation in a large case control cohort, and eQTL analysis to explore the candidate functional variants at the susceptibility risk loci 9q22.33. We determined that rs7027650 located at intron two of the COL15A1 gene was a potential candidate causal variant in this region.

Functional annotation revealed that rs7027650 overlaps with DNase I hypersensitivity sites in normal ovarian tissue and ovarian cancer cell lines, suggesting that this variant resides in open chromatin region and may serve a transcriptional regulation function. The Luciferase assay revealed that the gene fragments with rs7027650 have lower activity in SKOV3 and 293T cells indicating that the SNP could interact with the promoter region of COL15A1. The EMSA showed that rs7027650 exhibited different protein binding capacities between allele-A and allele-T. Collectively, these molecular analyses and experiments suggested that SNP rs7027650 may be a functional variant, potentially affecting COL15A1 transcription regulation and may affect ovarian cancer carcinogenesis.

The COL15A1 gene encodes the alpha chain of type COL15 which is a member of the FACIT collagen family. COL15 is a structurally complex macromolecule with many unique features and a range of biological properties. Light microscopy experiments have reported that COL15 is localized to endothelial, muscle, nerve, adipose, and most epithelial basement membrane regions of human tissues. This wide distribution suggests COL15 has an adhesive role, connecting the basement membrane to connective tissue (Amenta et al., 2005; Clementz & Harris, 2013). In breast cancer, collagen types XV and XIX were lost early in the development of invasive tumors. The disappearance of these proteins may signal remodeling of the extracellular matrix to promote tumor cell infiltration (Amenta et al., 2003). In a transgenic MMTV-PyMT mouse mammary carcinoma model, inactivation of COL15A1 modulated the tumor extracellular matrix and increased mammary tumor growth (Martinez-Nieto et al., 2021). Type XV collagen played a role in the adherence of the basement membrane to surrounding connective tissue and it may be associated with the tumorigenesis of keratinocytes and melanocytes (Fukushige et al., 2005). In cervical carcinoma cells, collagen XV functioned as a dose-dependent suppressor of tumorigenicity (Harris, Harris & Hollingsworth, 2007), and was found to inhibit the adhesion and migration of fibrosarcoma cells when present in fibronectin-containing matrices (Hurskainen et al., 2010). Kimura et al. (2016) reported that immunohistochemical examination of collagen XV and COL15A1 mRNA both showed increased expression in tumoral regions than in non-tumoral regions in human hepatocellular carcinoma (HCC), and collagen XV was considered to be the factor that contributed to the capillarization of HCC. Collagen XV was also reported to function as a metastasis inhibitor in HCC by regulating the discoidin domain receptor 1 (DDR1)-Snail/Slug axis, thus regulating epithelial-mesenchymal transition (EMT) (Yao et al., 2022). The COL15A1 gene has been reported to be upregulated by the interaction between Clusterin (CLU) and myocyte enhancer factor 2A (MEF2A). This upregulation is associated with blocking the EMT process, inhibiting the invasion and metastasis of testicular seminoma (Cui et al., 2021). In ovarian cancer, COL15A1 gene expression was significantly lower in tumor tissues than normal tissues in TCGA data (Chen et al., 2014), suggesting COL15A1 has a potential tumor suppressor role in ovarian cancer. Further experimental investigation is warranted to reveal the underlying molecular mechanism of COL15A1 in ovarian cancer carcinogenesis.

There are still some limitations to this study. Due to insufficient samples, we did not validate our results in another larger sample. We also did not find any significant association between rs7027650 and ovarian cancer risk in the European population from the Ovarian Cancer Association Consortium (OCAC) data. The inconsistent result with European population may be due to the race- and ethnicity heterogeneity of cancer susceptibility (Henderson et al., 2012). The ethnicity specific cancer susceptibility was also observed in several other kinds of cancers, such as colorectal cancer and prostate cancer etc. (Lu et al., 2019; Rebbeck, 2017). Additionally, this study only conducted a preliminary exploration of the potential function of the causal variant, and more experiments are needed to provide additional evidence to support the regulatory function of this variant.

Conclusions

In conclusion, we identified a potential causal variant within the chromosome 9q22.33 region, rs7027650, which was associated with ovarian cancer susceptibility. We also demonstrated that rs7027650 may regulate COL15A1 gene expression, offering suggestions for further functional studies to clarify the underlying biological mechanism in ovarian cancer etiology. Nonetheless, the transcriptional regulation mechanism of rs7027650 on COL15A1 remains unclear and needs to be clarified through further functional experiments.

Supplemental Information