Identification of twelve new susceptibility loci for different
histotypes of epithelial ovarian cancer
A full list of authors and affiliations appears at the end of the article.
#
Correspondence to: PDPP, Departments of Oncology and Public Health and Primary Care, University of Cambridge, Cambridge,
England.pp10001@medschl.cam.ac.uk; Tel +44 1223 740166.
*
These authors contributed equally to this manuscript
§
These authors jointly directed this work
Author Contributions
Writing group: C.M.P., K.B.K., J.P.T, S.P.K., K.L., S.W., D.H., M.A.E., A.N.M., G.C.-T., E.L.G., S.J.R., T.A.S., S.A.G., A.C.A. and
P.D.P.P. Co-ordinated OCAC OncoArray genotyping: C.M.P., M.J. R., G.C. Coordinated CIMBA OncoArray genotyping: G.C.-T.,
L.McG, J.S., P.S. OncoArray genotyping: CIDR (M.A., T.S., K.F.D., J.Romm, E.P.), Mayo (J.M.C.), UCam (C.Luccarini). Oncoarray
genotype calling and quality control: K.B.K, D.F.E., J.D., D.B., E.D., A.Pirie, A.Lee, J.L., G.L. Performed statistical analyses for
OCAC: J.P.T., P.D.P.P. Performed statistical analyses for CIMBA: K.B.K., A.L., A.C.A. Performed the meta-analyses: K.B.K., A.C.A.
OCAC database management: M.J.R., A.Berchuck. CIMBA database management and
BRCA1/2
variant nomenclature and
classification: L.McG., G.L, A.B.S.. Supervised OCAC statistical analyses: P.D.P.P. Supervised CIMBA statistical analyses: A.C.A.
Conceived and coordinated the synthesis of the Oncoarray: P.D.P.P., D.F.E., C.A., S.Chanock, S.G., B.H., D.J.H., A.C.A., J.S..
Functional analyses: P.C.L Jr., S.Coetzee, M.A.E., S.A.G., E.L.G., D.H., S.P.K., K.L., J.M.L, G.M.-F., A.N.M., S.J.W, G.C-T., J.
Beesley.
Provided DNA samples and/or phenotypic data: C.M.A., K.K.H.A., J.Adlard, I.L.A., H.A-C., N.Antonenkova, G.A., N.Arnold,
B.K.A., B.A., J.Azzollini, J.Balmaña, S.N.B., L.Barjhoux, R.B.B., Y.B., M.W.B., A.B.-F., J.Benitez, A.Berchuck, M.Bermisheva,
M.Bernardini, M.J.Birrer, M.Bisogna, L.Bjorge, A.Black, K.Blankstein, M.J.Blok, C.Bodelon, N.B., A.Bojesen, B.Bonanni, A.Borg,
A.R.B., J.D.B, C.Brewer, L.Brinton, P.B., A.B.-W., F.B., J.Brunet, B.Buecher, R.B., S.S.B., J.Y.B., T.Caldes, M.A.C., I.C., R.C.,
M.E.C., T.Cescon, S.B.C., J.C.-C., X.Q. C., G.C-T., Y.-E.C., J.Chiquette, W.K.C., K.B.M.C., T.Conner, J.Cook, L.S.C., F.J.C., D.W.C.,
A.A.D., M.B.D., F.Damiola, S.D.D., A.D.-M., F.Dao, R.D., A.dF., C.D., O.D., Y.C.D., J.A.D., S.M.D., C.M.D., T.D., L.D., M.Duran,
M.Dürst, B.D., D.E., T.E., R.E., U.E., B.E., A.B.E., S.E., M.E., K.H.E., C.E., D.G.E., P.A.F., S.F., S.F.F., J.M.F., T.M.F., Z.C.F., R.T.F.,
F.F., W.D.F., G.F., B.L.F., E.F., D.F., P.A.G., J.Garber, M.J.G., V.G.-B., S.A.G., A.G., A.G.-M., A.M.-G., G.Giles, R.G., G.Glendon,
A.K.G., D.E.G., E.L.G., M.T.G., T.G., M.G., M.H.G., J.Gronwald, E.Hahnen, C.A.H., N.H., U.H., T.V.O.H., P.A.H., H.R.H., J.Hauke,
A.Hein, A.Henderson, M.A.T.H., P.H., S.H., C.K.H., E.Høgdall, F.B.L.H., H.H., M.J.H., K.H., R-Y.H., P.J.H., J.Hung, D.G.H., T.H.,
E.N.I., C.I., E.S.I., L.I., A.I., A.Jakubowska, P.J., R.J., A.Jensen, M.J., U.B.J., E.M.J., S.J., M.E.J., P.K., B.Y.K., A.Karzenis, K.K.,
L.E.K., C.J.K., E.K., L.A.K., J.I.K., S.-W.K., S.K.K., M.K., R.K.K., T.A.K., J.K., A.Kwong, Y.L., D.Lambrechts, N.L., M.C.L.,
C.Lazaro, N.D.L., L.LeM., J.W.L., S.B.L., A.Leminen, D.Leroux, J.Lester, F.L., D.A.L., D.Liang, C.Liebrich, J.Lilyquist, L.Lipworth,
J.Lissowska, K.H.L., J.Lubi ski, L.Lundvall, P.L.M., S. Manoukian, L.F.A.G.M., T.M., S.Mazoyer, J.McA., V.McG., J.R.McL.,
I.McN., H.E.J.M.-H., A.M., U.M., A.R.M., M.Merritt, R.L.M., G.M., F.M., J.M.-S., M.Moffitt, M.Montagna, K.B.M., A.M.M., J.M.,
S.A.N., K.L.N., L. N., R.B.N., S.L.N., H.N., D.N., R.L.N., K.Odunsi, K.Offit, E.O., O.I.O., H.O., C.O., D.M.O’M., K-R.O., N.C.O.-
M., N.O., S.O., A.O., L.O., D.P., L.Papi, S.K.P., T-W.P.-S., J.P., C.L.P., I.S.P., P.H.M.P., B.Peissel, A.Peixoto, T.Pejovic, L.M.P., J.B.P.,
P.Peterlongo, L.P., G.P., P.D.P.P., C.M.P., K.-A.P., M.P., M.C.P., A.M.P., S.R.P., T.Pocza, E.M.P., B.Poppe, M.E.P., F.P., D.P., M.A.P.,
P.Pujol, P.Radice, S.J.R., J.Rantala, C.R.-F., G.R., K.R., P.Rice, A.Richardson, H.A.R., M.R., G.C.R., C.R-A., M.A.Rookus,
M.A.Rossing, J.H.R., A.Rudolph, I.B.R., H.B.S., D.P.S., J.M.S., R.K.S., M.J.S., T.A.S., L.Senter, V.W.S., G.Severi, P.Sharma, N.S.,
L.E.Side, W.S., J.S., C.F.S., H.Sobol, H.Song, P.Soucy, M.S., A.B.S., Z.S., D.S., D.S.-L., L.E.S.-C., G. Sukiennicki, R.S., C.S., A.J.S.,
C.I.S., L.Szafron, Y.Y.T., J.A.T., M.-K.T., M.R.T., S.-H.T., K.L.T., M.Thomassen, P.J.T., L.C.V.T., D.L.T., L.T., A.V.T.,
M.Tischkowitz, S.T., A.E.T., A.Tone, B.T., R.T., A. Trichopoulou, N.T., S.S.T., A.M.V.A., D.V.D.B., A.H.V.D.H., R.B.V.D.L., M.V.H.,
E.V.N., E.J.V.R., A.Vanderstichele, R.V.-M., A.Vega, D.V.E., I.V., J.V., R.A.V., A.Vratimos, L.W., C.W., D.W., S.W.-G., B.W., P.M.W.,
C.R.W., J.N.W., N.W., A.S.W., J.T.W., L.R.W., A.W., M.W., A.H.W., X.W., H.Y., D.Y., A.Z., K.K.Z.
All authors read and approved the final manuscript.
Competing interests
The authors declare no competing financial interests related to this manuscript.
Websites
Nature Publishing Group.
Nature Genetics - iCOGS
, http://www.nature.com/icogs/
The Cancer Genome Atlas Project - http://cancergenome.nih.gov/
The cBio Cancer Genomics Portal - http://www.cbioportal.org/
Pupasuite 3.1 -http://pupasuite.bioinfo.cipf.es
OCAC - http://apps.ccge.medschl.cam.ac.uk/consortia/ocac/
CIMBA QC guidelines-http://ccge.medschl.cam.ac.uk/consortia/cimba/members/data%20management/CIMBA%20and%20BCAC
%20Quality%20Control%20November%202008%20v2.doc
HHS Public Access
Author manuscript
Nat Genet
. Author manuscript; available in PMC 2017 November 01.
Published in final edited form as:
Nat Genet
. 2017 May ; 49(5): 680–691. doi:10.1038/ng.3826.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Abstract
To identify common alleles associated with different histotypes of epithelial ovarian cancer (EOC),
we pooled data from multiple genome-wide genotyping projects totaling 25,509 EOC cases and
40,941 controls. We identified nine new susceptibility loci for different EOC histotypes: six for
serous EOC histotypes (3q28, 4q32.3, 8q21.11, 10q24.33, 18q11.2 and 22q12.1), two for
mucinous EOC (3q22.3, 9q31.1) and one for endometrioid EOC (5q12.3). We then meta-analysed
the results for high-grade serous ovarian cancer with the results from analysis of 31,448
BRCA1
and
BRCA2
mutation carriers, including 3,887 mutation carriers with EOC. This identified an
additional three loci at 2q13, 8q24.1 and 12q24.31. Integrated analyses of genes and regulatory
biofeatures at each locus predicted candidate susceptibility genes, including
OBFC1
, a novel
susceptibility gene for low grade/borderline serous EOC.
Epithelial ovarian cancer (EOC) is a heterogeneous disease commonly classified into five
major histotypes of invasive disease
1
- (high grade serous (HGSOC), low grade serous
(LGSOC), mucinous (MOC), endometrioid (ENOC) and clear cell carcinoma (CCOC)) -
and two histotypes of borderline disease – serous and mucinous. The histotypes have
differences in lifestyle and genetic risk factors, precursor lesions, patterns of spread,
molecular events during oncogenesis, response to chemotherapy, and prognosis. HGSOC are
thought to be derived from fallopian tube secretory epithelial cells through foci of
endosalpingiosis existing as inclusion cysts lined with tubal epithelium at the ovarian and
peritoneal surface
2
. In contrast, CCOC, ENOC, and sero-endometrioid carcinomas appear to
develop from endometriosis
3,4
. MOC resembles adenocarcinoma of the gastric pylorus,
intestine, or endocervix and the majority of these tumors show gastrointestinal
differentiation
5
.
Approximately 20% of the familial component of EOC risk is attributable to high-to-
intermediate risk genes
6
. An unknown fraction is due to more common, lower risk genetic
variation
7
. In European populations, genome-wide association studies (GWAS) have
identified 23 EOC susceptibility alleles including 18 common variants associated with all
histologies and/or serous EOC
8-15
, one with borderline serous tumors
13
, three with
MOC
16
and one with CCOC
12
. The majority of these loci also showed associations
(p<0.05) with EOC risk for
BRCA1
or
BRCA2
mutation carriers
15
. Five additional loci
associated with EOC and breast and/or prostate cancer have been identified
17
; three of these
were associated with susceptibility to EOC, breast and prostate cancers, and two were
associated only with breast and EOC risk. However, the common genetic variants explain
only 3.9% of the inherited component of EOC risk
15
and additional susceptibility loci are
likely to exist, particularly for the less common, non-serous histotypes.
We designed a custom Illumina array named the ‘OncoArray’, in order to identify new
cancer susceptibility loci
18
. The OncoArray includes ~533,000 variants (of which 260,660
formed a GWAS backbone) and has been used to genotype over 500,000 samples, including
EOC case-control studies of the Ovarian Cancer Association Consortium (OCAC) and
BRCA1
and
BRCA2
mutation carriers of the Consortium of Investigators of Modifiers of
BRCA1/2
(CIMBA). These data were combined with genotype data from the Collaborative
Oncological Gene-environment Study (COGS) project
14,19
and three EOC GWAS
8,9
. We
Phelan et al.
Page 2
Nat Genet
. Author manuscript; available in PMC 2017 November 01.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
present the results of these association analyses together with functional annotation of the
new genome-wide significant EOC susceptibility loci.
Results
Association analyses
Genetic association analyses were performed using genotype data from 25,509 population-
based EOC cases and 40,941 controls from OCAC and meta-analysis of these data with
19,036
BRCA1
and 12,412
BRCA2
mutation carriers from CIMBA, of whom 2,933 and
954, respectively, were affected with EOC. The numbers of participants by study for OCAC
and CIMBA are shown in Supplementary table 1 and Supplementary table 2, respectively.
We used data from the 1000 Genomes Project
20
reference panel to impute genotypes for
11,403,952 common variants (MAF>1%) and evaluated the associations of these SNPs with
EOC risk. In OCAC alone, nine histotypes were investigated (all invasive, serous invasive,
HGSOC, LGSOC, serous borderline, LGSOC and serous borderline combined, ENOC,
CCOC and MOC). Association analyses revealed six novel loci associated with serous EOC
histotypes at genome-wide significance (p<5×10
-8
): rs9870207 at 3q28, rs13113999 at
4q32.3, rs150293538 at 8q21.11, rs7902587 at 10q24.33, rs8098244 at 18q11.2 and
rs6005807 at 22q12.1. Five of these loci were associated with borderline serous EOC (3q28,
4q32.3, 8q21.11, 10q24.33 and 18q11.2) and four with LGSOC tumors (3q28, 8q21.11,
10q24.33 and 18q11.2) (Table 1). We also identified two loci associated with MOC
(rs112071820 at 3q22.3 and rs320203 at 9q31.1) and one locus associated with ENOC
(rs555025179 at 5q12.3). The meta-analysis of OCAC and CIMBA revealed three additional
serous EOC loci (rs2165109 at 2q13; rs9886651 at 8q24.21; rs7953249 at 12q24.31). The
8q24.21 SNP rs9886651 is close to two SNPs previously associated with serous EOC
9
(and
Gjyshi A, Mendoza-Fandino G, Tyrer J, Woods NT, Lawrenson K et al., personal
communication). Multi-variable analysis of OCAC data showed that this is a third
independent-associated variant in this region (unadjusted OR = 1.07, OR adjusted for
rs1400482 and rs13255292 =1.07). Variant rs6005807 at 22q12.1 was previously reported to
be associated with serous EOC at sub-genome-wide significance
21
.
The association of the top SNP in each region with the nine EOC histotypes studied with
EOC risk in
BRCA1
and
BRCA2
carriers is shown in Figure 1. Four SNPs, rs8098244
(18q11.2), rs2165109 (2q13), rs9886651 (8q24.21), rs7953249 (12q24.31) showed
associations with EOC risk for
BRCA1
mutation carriers and one SNP, rs9886651 (8q24.21)
showed an association with risk for
BRCA2
carriers (P<0.05)
Eighteen of the 23 previously published loci were associated with the same histotype at
genome-wide significance (excluding the 5 pleitropic loci published by Kar et al,
Supplementary table 3). Of these, 11 showed an association with EOC risk for
BRCA1
mutation carriers and eight showed an association with risk for
BRCA2
carriers (P<0.05).
There was significant heterogeneity of risk between the five main, non-overlapping
histotypes (high grade serous, low grade/borderline serous, endometrioid, clear cell and
invasive/borderline mucinous) for 28 of the 40 new and previously published loci
(Supplementary table 3).
Phelan et al.
Page 3
Nat Genet
. Author manuscript; available in PMC 2017 November 01.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
We carried out a competing-risks association analysis in
BRCA1
and
BRCA2
mutation
carriers in order to investigate whether the observed associations with ovarian cancer in
mutation carriers are influenced by associations with breast cancer risk. For this we used the
most significantly associated genotyped SNPs for this
22
. The EOC HR estimates were
consistent with the estimates from the main analysis for all SNPs (results not shown). Some
evidence suggested that rs7953249 at 12q24.31 was associated with reduced breast cancer
risk in
BRCA1
mutation carriers (HR=0.95, 95%CI 0.91-0.99, p=0.034) and that SNP
rs2165109 at 2q13 was associated with increased breast cancer risk in
BRCA2
mutation
carriers (HR=1.08, 95%CI 1.01-1.14, p=0.02). When these associations were analyzed by
tumor estrogen-receptor status, the associations for the two SNPs were restricted to ER-
negative breast for
BRCA1
(p=0.026 for rs7953249) and
BRCA2
(p=0.02 for rs2165109)
mutation carriers.
Association analyses adjusted for the most significant SNP in each region (including 3
independent SNPs at 8q24.21) did not reveal any additional independent association signals
in these regions. At the 12 new EOC risk regions, 571 SNPs were deemed potentially causal
(Supplementary table 4) and carried forward for functional annotation, eQTL and mQTL
analyses.
Functional and molecular analyses
Of the 571 candidate causal variants in the 12 novel loci, 562 variants are located in non-
coding DNA sequences and may influence the expression of nearby target genes
23
. We used
a variety of
in silico
approaches to identify putative, tissue-specific, regulatory biofeatures
and candidate susceptibility genes associated with risk SNPs at each locus. For the few risk-
associated, non-synonymous variants in protein coding genes, we also evaluated predicted
effects on protein function.
Functional annotation of candidate causal alleles—We mapped the set of 562 non-
protein coding candidate causal SNPs at the 12 susceptibility loci to regulatory biofeatures,
using a variety of epigenomic marks profiled in normal and cancer tissues relevant to the
cellular origins of different ovarian cancer histotypes (Supplementary table 5). The cell types
interrogated included: (1) fallopian tube (FT33; FT246) and ovarian surface epithelial cell
lines (IOSE4; IOSE11) for serous precursor tissues; (2) serous-related cancer cell lines
including HGSOC cell lines (UWB1.289; CaOV3) and a LGSOC cell line (OAW42); (3)
endometriosis epithelial cells (EEC16), as a likely precursor of ENOC; (4) cell types
relevant to MOC, including MOC cell lines (GTFR230; MCAS) and both colonic normal
(colon crypt) and cancer tissues (HCT116; HeLa-S3). The epigenomic marks annotated
were open chromatin, identified using formaldehyde assisted isolation of regulatory element
sequencing (FAIRE-seq) and DNase I hypersensitivity sequencing (DNase-seq) and
chromatin immunoprecipitation sequencing (ChIP-seq) of histone modifications, specifically
histone H3 lysine 27 acetylation (H3K27ac, which denotes active chromatin) and histone H3
lysine 4 monomethylation (H3K4me1, which marks active and poised enhancers). SNPs
were also intersected with ENCODE transcription factor ChIPseq data. All tissue types were
evaluated for all risk loci. The SNP-biofeature intersections by tissue type are illustrated in
Figure 2 and Supplementary table 6.
Phelan et al.
Page 4
Nat Genet
. Author manuscript; available in PMC 2017 November 01.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Nine (1.6%) of the 571 candidate causal SNPs lie in protein coding sequences. Five of these
are synonymous and four are non-synonymous but predicted to be benign by Polyphen-2
(Supplementary table 6). Four SNPs lie within untranslated regions of protein-coding genes
and so could affect mRNA stability: rs1051149 and rs1051150 in the 3’ UTR of
LAMA3
and rs12327412 in the 5’ UTR of
TTC39C,
all at the 18q11.2 locus; and rs1018128 in the 3’
UTR of
GMNC
at 3q28. The majority of biofeature-SNP intersections (n=166, 29% of all
candidate causal SNPs and 97% of candidate causal SNPs overlapping a biofeature) were for
SNPs lying within active chromatin, and/or open chromatin. Eleven SNPs lie in the
promoters of four genes (
PVT1
,
HNF1A
,
TTC39C
and
TTC28
) (Supplementary Table 6).
At six serous risk loci (4q32.3; 3q28; 8q21; 18q11; 8q24; 22q12) we observed extensive
SNP-biofeature overlaps, particularly in serous-related tissue types. In contrast, the two
MOC susceptibility loci (3q22.3, 9q31.1) were biofeature-poor regions and showed little or
no SNP-biofeature intersections in any of the tissue types under investigation, including
MOC and ENCODE cell lines. At the endometrioid EOC risk locus (5q12.3) we observed
enhancers in endometriosis, ovarian, fallopian and EOC cell types flanking the small number
risk associated SNPs (n=8), none of which coincided with regulatory elements.
Several studies have shown that common variant susceptibility alleles are significantly
enriched for regulatory elements detected in disease-relevant tissue types. Therefore we
tested for enrichment of SNP-H3K27ac intersections at each locus because H3K27ac was
the most comprehensively profiled regulatory feature across different tissue types
(Supplementary table 7). At the 12q24.31 locus a large region of active chromatin spanning
the
HNF1A
promoter drove a strong enrichment for risk SNP-H3K27ac intersects in the
OAW42 LGSOC cell line (
P
=4.45×10
-22
). At 10q24.33 (which is associated with LGSOC
and borderline SOC) we identified a significant enrichment of acetylated H3K27 in normal
fallopian cells (FT33
P
=1.09×10
-4
, FT246
P
=4.29×10
-3
), HGSOC ovarian cancer cells
(UWB1.289
P
=6.23×10
-3
), MOC cells (GTFR230
P
=5.16×10
-3
) as well as, somewhat
surprisingly, colorectal cancer cells (HCT116
P
=2.64×10
-4
) and cervical cancer cells (HeLa-
S3
P
=9.60×10
-12
). This locus contains several clusters of H3K27ac activity and TF binding
in ovarian and ENCODE datasets, and these highly active regions showed extensive overlap
with candidate causal alleles (Figure 3).
Identifying candidate susceptibility gene targets at risk loci—We used several
approaches to identify candidate target genes at the 12 risk loci. First, we hypothesized that
target genes underlying disease susceptibility are more likely to display prevalent copy
number alterations in ovarian tumor tissues. Amplifications were the most frequent
alteration at 6 of the 12 susceptibility loci (Supplementary figure 1). Contiguous genes were
commonly amplified in the same sample indicating segmental amplifications (data not
shown).
HNF1A
,
ORAI1, CHEK2
,
XPB1
,
BUB1,
and
FOXL2
are found inside the same
topologically associating domain (TAD) as candidate causal SNPs and have been previously
implicated in ovarian cancer development (Supplementary figure 2). Notably,
HNF1A
,
ORAI1,
and
FOXL2
are amplified in >5% of EOC samples. No TAD was identified for
8q24.21; but
MYC
and
PVT1
appear to be the targets for multiple enhancer elements
containing independent EOC risk associations for HGSOC at this locus (Gjyshi et al.,
personal communication).
Phelan et al.
Page 5
Nat Genet
. Author manuscript; available in PMC 2017 November 01.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript