The contribution of germline BRCA1 and BRCA2 mutations to familial ovarian cancer: no evidence for other ovarian cancer-susceptibility genes.

Abstract

To establish the contribution of germline BRCA1 and BRCA2 mutations to familial ovarian cancer, we have analyzed both genes in DNA samples obtained from an affected individual in each of 112 families containing at least two cases of epithelial ovarian cancer. Germline mutations were found in 43% of the families; BRCA1 mutations were approximately four times more common than BRCA2 mutations. The extent of family history of ovarian and breast cancers was strongly predictive of BRCA1-mutation status. Segregation analysis suggests that a combination of chance clustering of sporadic cases and insensitivity of mutation detection may account for the remaining families; however, the contribution of other genes cannot be excluded. We discuss the implications for genetic testing and clinical management of familial ovarian cancer arising from the data presented in these studies.

Full text

Page 1

Am. J. Hum. Genet. 65:1021–1029, 1999
1021
The Contribution of Germline BRCA1 and BRCA2 Mutations to Familial
Ovarian Cancer: No Evidence for Other Ovarian Cancer–Susceptibility
Genes
Simon A. Gayther,1Paul Russell,1Patricia Harrington,1Antonis C. Antoniou,2
Douglas F. Easton,2and Bruce A. J. Ponder1
1Department of Oncology and
2Genetic Epidemiology Unit, Cancer Research Campaign (CRC), Strangeways Research Laboratory, Cambridge
Summary
To establish the contribution of germline BRCA1 and
BRCA2 mutations to familial ovarian cancer, we have
analyzed both genes in DNA samples obtained from an
affected individual in each of 112 families containing at
least two cases of epithelial ovarian cancer. Germline
mutations were found in 43% of the families; BRCA1
mutations were approximately four times morecommon
than BRCA2 mutations. The extent of family history of
ovarian and breast cancers was strongly predictive of
BRCA1-mutation status. Segregation analysis suggests
that a combination of chance clustering ofsporadiccases
and insensitivity of mutation detection may account for
the remaining families; however, the contribution of
other genes cannot be excluded. We discuss the impli-
cations for genetic testing and clinical management of
familial ovarian cancer arising from the data presented
in these studies.
Introduction
Familial ovarian cancer occurs as part of two clinically
distinct syndromes. The most common of these syn-
dromes is ovarian cancer, which is either apparently site
specific or occurs in association with early-onset breast
cancer (Easton et al. 1993). Predisposition to ovarian
cancer also occurs as part of Lynch type II or hereditary
nonpolyposis colorectal cancer (HNPCC) syndrome
(Lynch et al. 1982). The results of genetic-linkagestudies
have suggested that, in 190% of families with multiple
Received March 31, 1999; accepted for publication July 29, 1999;
electronically published September 1, 1999.
Address for correspondence and reprints: Dr. Simon A. Gayther,
Department of Oncology, CRC, Strangeways Research Laboratory,
Wort’s Causeway, Cambridge CB1 8RN, United Kingdom. E-mail:
simon.gayther@srl.cam.ac.uk
? 1999 by The American Society of Human Genetics. All rights reserved.
0002-9297/1999/6504-0011$02.00
cases of breast and ovarian cancer, the cancers are the
result of mutations in either the BRCA1 gene on chro-
mosome 17q12-21 or the BRCA2 gene on chromosome
13q12-13 (Narod et al. 1995; Ford et al. 1998). Linkage
data from a small series of families suggest that BRCA1
is also responsible for the majority of “site-specific ovar-
ian cancer” families with threeormorecasesofepithelial
ovarian cancer (Steichen-Gersdorf et al. 1994).
Germline mutations in the BRCA1 and BRCA2 genes
have been identified in a large number of families with
multiple cases of early-onset breast and/or ovarian can-
cer (Miki et al. 1994; Shattuck-Eidens et al. 1995; Woo-
ster et al. 1995; Tavtigian et al. 1996; Gayther et al.
1997b). The mutation spectrum is similar in both genes:
most germline mutations are predicted to result in pro-
tein truncation caused by frameshift, nonsense,orsplice-
site alterations, and the mutations are spread along the
length of the coding region (Breast Cancer Information
Core). A small number of founder mutations that are
common in specific populations have been described. In
particular, these include the 185delAG mutation in
BRCA1 and the 6174delT mutation in BRCA2, which
are common in Ashkenazi Jews (Struewing et al. 1995;
Neuhausen et al. 1996a, 1996b); the 5382insC mutation
in BRCA1, which is common in the eastern European
population (Gayther et al. 1997a; Ramus et al. 1997);
and the 999del5 mutation, which is common in the Ice-
landic population (Thorlacius et al. 1996). The precise
functions of the BRCA1 and BRCA2 proteins are un-
known; they have recently been shown to interact with
each other, which suggests that they may function in the
same pathway (Chen et al. 1998). Both proteins interact
in vitro with RAD51, a protein that has been implicated
in the repair of double-strand DNA breaks (Scully 1997;
Wong et al. 1997). This finding and the results of an
analysis of homozygous mutant mice suggest a possible
functional role for BRCA1 and BRCA2 in DNA repair
and genome integrity (Gayther and Ponder 1998; Zhang
et al. 1998).
An immediate consequence of the identification of the
BRCA1 and BRCA2 genes is that DNA-based predictive
testing has become possible in families when the mu-

Page 2

1022
Am. J. Hum. Genet. 65:1021–1029, 1999
tation is known. Since the search for the mutation is
laborious, it is helpful to know the prevalence of BRCA1
and BRCA2 mutations in families with different cancer
histories. In this study, we set out to obtain this infor-
mation in relation to ovarian cancer. One hundred and
twelve families were ascertained to have a family history
in which at least two first- or second-degree relatives
had been given a diagnosis of epithelial ovarian cancer.
To search for disease-associated mutations, the entire
BRCA1 and BRCA2 coding sequence was analyzed in
one affected individual from each family. This enabled
us to estimate the proportion of families with ovarian
cancer that resulted from these two genes, and it also
enabled us to predict whether there are likely to be other
highly penetrant susceptibility genes that are responsible
for a proportion of cases of familial ovarian cancer.
Material and Methods
Patient Material
One hundred and three families in which at least two
first-degree relatives had epithelial ovarian cancer di-
agnosed at any age and nine families in which two cases
of epithelial ovarian cancer were diagnosed in second-
degree relatives were identified through the Familial
Ovarian Cancer Register of the United Kingdom Co-
ordinating Committee on Cancer Research (UKCCCR).
To be included in the register, families must have at least
two cases of ovarian cancer confirmed by pathology re-
ports or by a death certificate. Borderline cases of ovar-
ian cancer were not included in the study. Fifty-three
families had ovarian cancer only, whereas the remaining
59 families had at least one case of breast cancer, in
addition to ovarian cancer, diagnosed in a family mem-
ber before the age of 60 years. These families included
a total of 330 cases of ovarian cancer and 133 cases of
breast cancer diagnosed before the age of 60 years. Con-
firmation of diagnosis was obtained from either pa-
thology reports or a death certificate, for a total of 281
cases of ovarian cancer and 70 cases of breast cancer.
Initial DNA analysis was performed with genomic DNA
obtained from one affected individual from each family,
and, when material was available, mutations were con-
firmed in other affected individuals from the family.
Mutation Analysis
BRCA1 and BRCA2 were both screened for germline
mutations, by use of a combination of the protein-trun-
cation test (PTT) and nonradioactive single-strand con-
formation analysis/heteroduplex analysis (SSCA/HA).
PTT was performed, as described elsewhere (Friedman
et al. 1997), for the two largest exons of BRCA2 and
for the largest exon of BRCA1. For SSCA/HA, coding
exons 2, 3, 5–10, and 12–24 of BRCA1 and coding
exons 2–9 and 12–27 of BRCA2 were amplified from
genomic DNA. The 5?and 3?splice boundaries of exon
11 of BRCA1 and exons 10 and 11 of BRCA2 were also
amplified from genomic DNA. Nonradioactive SSCA/
HA was performed as described elsewhere,andgelswere
visualized by silver staining (Gayther et al. 1995). Direct
sequence analysis was used to characterizethenucleotide
alteration associated with PTT and/or SSCA/HA vari-
ants, as described elsewhere (Gayther et al. 1995).
We also screened for a recently described 6-kb dupli-
cation near exon 13 of BRCA1 (Puget et al. 1999) in
families in which no BRCA1 or BRCA2 codingmutation
was identified. Duplication-specific primers (forward
primer, 5?-GATTATTTCCCCCCAGGCTA-3?; reverse
primer, 5?-AGATCATTAGCAAGGACCTGTG-3?) will
only produce a PCR-amplification product if the dupli-
cation is present in the DNA sample. Two positive ge-
nomic DNA controls for the duplication were included
in each experiment. In addition, for each sample, a sec-
ond, independent PCR reaction was simultaneously per-
formed with the use of forward primer 5?-TCACA-
ATTCCGAGACATC-3?and reverse primer 5?-AAC-
GGCTACTGCACAGTTCT-3?. This second reaction
amplified a product with a size similar to that of the
duplication and was used to control both for the effi-
cacy of each genomic DNA sample and for PCR
amplification.
Statistical Analysis
The effects, on mutation prevalence, of the numbers
of individuals affected with breast and ovarian cancer
were assessed with the use of x2tests for trend. To eval-
uate the evidence for existence of a third susceptibility
gene, in addition to BRCA1 and BRCA2, we performed
segregation analysis using the MENDEL program
(Lange et al. 1988). To allow for the ascertainment of
families on the basis of multiple affected individuals,
analyses were based on the conditional likelihood
Lik(M,D/D), where M was the observed mutation status
of the index case and D was the observed disease phe-
notype in the family. BRCA1 and BRCA2 were initially
assumed to confer the age-specific risks for breast and
ovarian cancer given in previous analyses (Ford et al.
1995). Carriersof a susceptibilityallele,atahypothetical
third gene locus (OVCA) and unlinked to BRCA1 and
BRCA2, had an increased relative risk (r) of ovarian
cancer compared with noncarriers, but they had no in-
creased risk for breast cancer. Dominant, recessive, and
codominant models for OVCA were explored. The ab-
solute risks by genotype were determined by the con-
straint that the overall incidence of the disease must
agree with the incidence rates for England and Wales.
We then estimated r and the frequencies p1, p2, and p3
of the susceptibility alleles of BRCA1, BRCA2, and

Page 3

Gayther et al.: BRCA1 and BRCA2 Mutations in Families with Ovarian Cancer
1023
OVCA, respectively, according to maximum likelihood.
The sumwas constrained to be no higher than
p ? p
12
.003, which is the combined frequency, given by the
model of Claus et al. (1991), of breast cancer–suscep-
tibility alleles and the highest plausible frequency for
BRCA1 and BRCA2 mutations combined. We also al-
lowed for a sensitivity parameter, g, which was used to
determine the probability of detecting a mutation, if one
was present (the probability was assumed to be the same
for BRCA1 and BRCA2).
We restricted consideration to those genetic models
that gave familial risks of ovarian cancer that were con-
sistent with those observed in epidemiological studies.
Stratton et al. (1998) have previously estimated that,
among first-degree relatives of patients withovariancan-
cer, the relative risk for ovarian cancer is 2.4, with a
95% confidence interval of 1.9–3.2, on the basis of a
meta-analysis of published cohort studies. We incorpo-
rated this estimate into our analysis by adding a “pen-
alty” to the log-likelihood, to reflect deviation of risk
from 2.4:
P = [log(2.4) ? log(l )] /2v
predicted familial relative risk from the model andwhere
is the variance of the log (relative risk) from
v = .0177
the meta-analysis.
, where lois the
2
o
Results
Germline mutations, which are thought to be disease
associated, were detected in 48 (43%) of the 112 fam-
ilies. These data are listed in table 1. Twenty-nine dif-
ferent BRCA1 mutations were identified in 40 (36%) of
the 112 families. The spectrum of BRCA1 mutations in
this data set is illustrated in figure 1a. Mutations are
spread quite evenly along the length of the gene. Six
mutations were detected on more than one occasion;
these six mutations represent 43% of all BRCA1 mu-
tations identified in this study. Haplotype analysis does
not indicate a common founder mutation in the British
population (Gayther et al. 1995; Neuhausen et al.
1996b). Thirty-five (88%) of the 40 mutations are either
frameshift or nonsense mutations that would be pre-
dicted to result in premature truncation of the BRCA1
protein. Four mutationsoccurat theboundariesbetween
exons and introns and are expected to affect splicing.
The results of analysis of reverse-transcribed RNA sam-
ples obtained from two affected individuals with splice
variants (4304GrA in family OV161 and 4476?6TrC
in family OV003) have previously shown that these al-
terations cause aberrant splicing (Gayther et al. 1995).
There was no available RNA with which to testtheeffect
of the remaining two splice-site mutations, although
both are predicted to abolish highly conserved splice-
site consensus sequences. Pro1749Arg, the putative mis-
sense mutation in family OV073, has not been described
elsewhere and was not detected in 346 normal chro-
mosomes. The functional significance of this missense
mutation has been assessed elsewhere (Chapman and
Verma 1996). The ability of the C terminus of BRCA1
to cause transcriptional activation in vitro was abolished
by a GAL4-BRCA1 fusion protein containing the
Pro1749Arg missense alteration, a finding that suggests
this change is functionally significant.
Seventy-two families without identifiable BRCA1 mu-
tations were analyzed for mutations throughout the
BRCA2 coding sequence. Seven different mutations that
were considered to be pathogenic were identified in eight
(7%) of these families. The spectrum of these mutations
throughout the gene is illustrated in figure 1b. In seven
families, the mutation was a frameshift deletion, and, in
one family, it was a nonsense mutation. One mutation,
6503delTT, was detected in two different families. Two
distinct mutations seen in families OV085 and OV027
occur at nucleotide 5573, which is the beginning of a
repeat of seven adenines; this finding possibly suggests
this region is a mutation hotspot. Three putative mis-
sense mutations were also identified: A75P in exon 3
(family OV120), R2502H in exon 15 (family OV222),
and Y3098H in exon 25 (family OV045). These mu-
tations were not detected in 344, 326, and 310 normal
chromosomes, respectively. However, because these mu-
tations have an unclear disease association and because
there is no available functional assay with which to fur-
ther investigate the significance of the mutations, they
have not been included in the statistical analyses.
Subsequent to our analysis of BRCA1 and BRCA2 for
coding mutations, large genomic alterations involving
the BRCA1 gene have been reported. These alterations
would not have been detected by the PCR-based ap-
proaches to mutation screening used in our study. One
of these genomic alterations, a duplication and insertion
involving 6 kb around exon 13 of BRCA1, is predicted
to lead to protein truncation and may be of Britishorigin
(Puget et al. 1999). A PCR-based assay has been spe-
cifically designed to detect this mutation. We were able
to screen for the exon 13 duplication in 55 families that
did not have a coding mutation in BRCA1 or BRCA2;
the mutation was not detected in any of the families
(data not shown).
Mutation Prevalence, by Family Type
Table 2 gives the proportion of families with BRCA1
and BRCA2 mutations, according to the numberofcases
of breast and ovarian cancer in the family. The numbers
of ovarian cancers and breast cancers in the family were
strongly predictive of BRCA1-mutation status (
, and
18.2 P = .0001
x = 15.29 P ! .0001
BRCA2-mutation status was significantly associated
with the number of cases of breast cancer in the family
(, ), but there was no significant asso-
2
1 x =
,, respectively).
2
1
2
1 x = 3.80 P = .05

Page 4

1024
Am. J. Hum. Genet. 65:1021–1029, 1999
Table 1
Germline Mutations in BRCA1 and BRCA2
FAMILY
NO. OF CONFIRMED
(UNCONFIRMED)
CASES OF CANCER
DESCRIPTION OF MUTATION
Ovarian BreastGene
Exon or
IntronCodon NucleotideAlteration Type
OV010
OV022
OV131
OV047
OV008
OV002
OV110
OV150
OV132
OV014
OV069
OV001
OV020
OV175
OV252
OV181
OV133
OV171
OV013
OV029
OV160
OV189
OV012
OV033
OV007
OV206
OV282
OV025
OV019
OV042
OV044
OV245
OV161
OV226
OV003
OV148
OV073
OV139
OV074
OV224
OV092
OV072
OV140
OV027
OV085
OV254
OV016
OV162
3 (1)
3
2
3 (2)
3 (1)
6
2
4
3
3 (3)
2
3
3
2
3
2
2 (1)
2
2
6
2
3
7 (1)
2
4 (3)
3 (4)
3
3 (3)
3
2 (2)
3 (7)
3
6
2 (1)
2 (1)
2 (1)
3
2
2
2 (1)
2 (1)
3
2
2
2
2 (1)
3
3
1 (3)
1
1 (2)
2 (2)
2 (2)
3 (1)
1
2
0 (1)
0
0
1 (1)
1 (1)
2 (1)
2 (2)
2 (1)
1 (1)
0
0
1 (1)
0
0
1
0
0 (1)
0
1
2
1
1 (5)
0 (7)
0 (1)
1
0 (1)
2
2
0 (1)
0 (1)
0 (1)
1 (7)
2 (1)
0
1
1 (1)
0
0 (2)
4
1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA1
BRCA2
BRCA2
BRCA2
BRCA2
BRCA2
BRCA2
BRCA2
BRCA2
Exon 2
Intron 5
Intron 6
Exon 8
Exon 11
Exon 11
Exon 11
Exon 11
Exon 11
Exon 11
Exon 11
Exon 11
Exon 11
Exon 11
Exon 11
Exon 11
Exon 11
Exon 11
Exon 11
Exon 11
Exon 11
Exon 11
Exon 11
Exon 11
Exon 11
Exon 11
Exon 11
Exon 11
Exon 11
Exon 11
Exon 11
Exon 11
Exon 12
Exon 13
Intron 13
Exon 16
Exon 20
Exon 20
Exon 24
Exon 24
Exon 11
Exon 11
Exon 11
Exon 11
Exon 11
Exon 11
Exon 11
Exon 11
23
—
—
169
367
392
392
392
607
608
627
652
652
655
690
871
885
916
1001
1001
1001
1056
1112
1112
1252
1252
1252
1289
1355
1355
1355
1355
1395
1443
—
1648
1749
1755
1835
1837
1366
1491
1602
1782
1782
1970
2092
2092
185
331
421
185delAG
331?2TrC
421-2delA
Gln169ter
1220insC
1294del40
1294del40
1294del40
lys607ter
1942del4
2000del4
2073delA
2073delA
2082insG
2190delA
2731insT
2774delCT
2867insA
3121delA
3121delA
3121delA
3286delC
3452del4
3452del4
3875del4
3875del4
3875del4
3986delAA
4184del4
4184del4
4184del4
4184del4
4304GrA
Arg1443ter
4476?6TrC
5061delA
Pro1749Arg
5382insC
Arg1835ter
5629delG
4326insCATC
4701delGAAA
5032insA
5573delAA
5573delA
Ser1970ter
6503delTT
6503delTT
Frameshift
Splice site
Splice site
Nonsense
Frameshift
Frameshift
Frameshift
Frameshift
Nonsense
Frameshift
Frameshift
Frameshift
Frameshift
Frameshift
Frameshift
Frameshift
Frameshift
Frameshift
Frameshift
Frameshift
Frameshift
Frameshift
Frameshift
Frameshift
Frameshift
Frameshift
Frameshift
Frameshift
Frameshift
Frameshift
Frameshift
Frameshift
Splice site
Nonsense
Splice site
Frameshift
Missense
Frameshift
Nonsense
Frameshift
Frameshift
Frameshift
Frameshift
Frameshift
Frameshift
Nonsense
Frameshift
Frameshift
624TrG
1220
1294
1294
1294
1938ArT
1942
2000
2073
2073
2082
2187
2731
2774
2867
3121
3121
3121
3286
3452
3452
3875
3875
3875
3986
4184
4184
4184
4184
4304
4446CrT
4476
5061
5365CrG
5382
5622CrT
5629
4326
4701
5032
5573
5573
6137CrA
6503
6503
ciation with the number of cases of ovarian cancer.
BRCA1 mutations were more frequent than BRCA2mu-
tations in familieswithlargenumbersofovariancancers;
in particular, 14 (35%) of 40 families with BRCA1 mu-
tations had four or more ovarian cancers, whereas none
of the families with BRCA2 mutations fell into this cat-
egory. However, this difference did not reach formal sta-
tistical significance (
As shown in table 2, mutations were found in 66%
of families with either three or more reported cases of
, ).
2
1 x = 2.89 P = .08

Page 5

Gayther et al.: BRCA1 and BRCA2 Mutations in Families with Ovarian Cancer
1025
Figure 1
or more confirmed cases of epithelial ovarian cancer.
Location and type of germline mutations detected in (a) the BRCA1 gene and (b) the BRCA2 gene, in families containing two
ovarian cancer or four or more reported cases of breast
or ovarian cancer; however, they were found in only
20% of families with two cases of ovarian cancer alone.
If families are classified only on the basis of confirmed
cases, then these frequencies are altered to 71% and
22%, respectively. In the families with a BRCA1 mu-
tation, the average patient age at diagnosis of ovarian
cancer (49.9 years) was significantly lower than that in
the families without the BRCA1 mutation (53.9 years)
( ). In the families with the BRCA2 mutation,
P = .003
the average age at diagnosis of ovarian cancer (55.0
years) did not differ significantly from that in either the
families with the BRCA1 mutation or the families with
no mutation.
Model Fitting
To evaluate the evidence for further ovarian can-
cer–susceptibility genes, in addition to BRCA1 and
BRCA2, we performed a segregation analysis. We tested
several models that allowed for a third gene (OVCA),
which was unlinked to BRCA1 or BRCA2 and which
had two alleles; in these models, carriers of the suscep-
tibility allele had an increased risk for ovarian cancer,
compared with noncarriers, but they had no increased
risk for breast cancer. Models were constrained for plau-
sible BRCA-gene frequencies and by the requirement
that the predicted overall incidence of breastandovarian
cancer must agree with known incidence. The frequen-
cies of BRCA1 and BRCA2 were estimated in allmodels.
We did not assume any values for these parameters. The
only assumption was that their combined frequency
should not be 1.003. To assess the goodness of fit of
different genetic models to the data, we computed, for
each model, the predicted probability of observing a
BRCA1 or BRCA2 mutation in families, in different cat-
egories of family history. We then compared these pre-
dictions with what we actually observed.
Table 3summarizes the estimated parameters for some
of these models. In all but one of the models (model 4),
the relative risk conferred by the OVCA susceptibility
allele converged to 1 (i.e., there was no effect). This was
true whether OVCA was assumed to be dominant, re-
cessive, or codominant. In the case of model 4, the like-
lihood differed only slightly from that of the correspond-
ing model showing no effect of OVCA (model 4a). We
also fitted models in which the penetrance estimates for
BRCA1 and BRCA2 were reduced so that the incidence
rates for both breast and ovarian cancer were 50% of
the Breast Cancer Linkage Consortium (BCLC) esti-
mates. However, these models also converged to a rel-
ative risk of 1 for OVCA (data not shown). Models in
which the risk for breast cancer conferred by BRCA1
was given by the BCLC values (models 1 and 2) had a
poor fit, with the observed number of BRCA1 and
BRCA2 mutations being greater than the predictednum-
ber of such mutations. The models with the reduced
incidence rates, in which BRCA1 and BRCA2 were at
50% of the BCLC rates, showed a better fit than did
the models in which the actual BCLC rates were used;
however, they did not show a better fit than did the
average heterogeneity rates. Better fits were obtained in
the models in which the average of the two penetrance
estimates given by Easton et al. (1995) were assumed.
In each case, models assuming 90% sensitivity for mu-
tation detection showed better fit than did models as-
suming 64% sensitivity. Table 2 shows that, when model