Hereditary diffuse gastric cancer: association with lobular breast cancer

CDH1 (OMIM *192090), located on chromosome 16q22.1 encodes, E-cadherin, an epithelial transmembrane cell–cell adhesion molecule and member of the cadherin superfamily of glycoproteins. In a zipper-like fashion, its extracellular domain forms calcium-dependent homodimers between the E-cadherin molecules of adjacent epithelial cells, to act as the primary mediator of epithelial cell adhesion at the adherens junction complex [11]. Through interactions of its cytoplasmic tail with multiple signalling and structural molecules, such as the catenins, E-cadherin, maintains cellular adhesion and epithelial architecture with this link to the cytoskeleton. The cytoplasmic tail of E-cadherin directly associates with β-catenin and γ-catenin which in turn binds to the f-actin microfilaments of the cytoskeleton, directly or through α-catenin [12]. p120-catenin also associates with E-cadherin’s cytoplasmic tail at a different site, the juxtamembrane domain, and acts to both strengthen the adhesion between cells and regulate cadherin membrane trafficking and degradation [13, 14]. E-cadherin is considered to have an invasion suppressor role, where decreased expression permits cells to dissociate from each other in order to migrate and invade [15]. In cancers, this manifests as increased infiltrative and metastatic potential [16]. E-cadherin is also thought to act as a tumour suppressor, potentially through its interaction with the multipurpose β -catenin molecule which is an effector of the WNT signalling pathway [17]. Loss of E-cadherin can result in β-catenin release from the membrane and translocation to the nucleus where it complexes with Tcf/Lef-1 transcription factors to initiate transcription of WNT responsive genes [18]. Activation of these genes have been implicated in tumourigenesis through the WNT signalling pathway as seen in adenomatous polyposis coli (APC) [19]. In support of the role of E-cadherin as a tumour suppressor is the observation of abnormal or absent E-cadherin expression in precursor lesions of DGC and LBC, where the phenomenon is seen in in situ signet ring cell carcinomas found in prophylactic gastrectomy specimens from germline CDH1 mutation carriers [20] and the lobular carcinoma in situ lesions seen adjacent to invasive lobular breast cancers [21]. These examples suggest that loss of E-cadherin is an early or even tumour-initiating event, however the actual molecular basis of such a potential role of E-cadherin in such cases is unknown.

Inactivating CDH1 mutations are found in 50% of sporadic DGCs [22, 23] and cluster between exons seven and nine [11], in contrast with the low percentage of mutations seen in sporadic intestinal type GCs [23]. Decreased expression of E-cadherin in DGCs may account for morphologic differences between intestinal and DGC variants [24]. Unlike somatic CDH1 mutations, germline mutations associated with DGC are distributed throughout the gene [7] (Fig. 2). In the cancers from individuals with CDH1 mutations, CDH1 acts as a classic tumour suppressor gene with loss of expression of the wildtype allele [25, 26]. In a single study of 6 hereditary DGC cancers, inactivation of the wild-type allele could be attributed to promoter hypermethylation in 5 (83%) of cases [26]. This finding warrants verification in a larger cohort as abnormal promoter methylation in early cancers could potentially form the basis of a screening test.

Currently germline mutations in single genes account for approximately 5–10% of breast cancer [27]. High penetrance genes such as BRCA1 and 2 account for 3–8%, and TP53 and PTEN as seen in Li-Fraumeni and Cowden syndrome together only account for <0.1% of breast cancer diagnoses [28]. Other medium and low penetrance genes such as CHK2, BRIP1, PALB2 and ATM [29‐32] have been identified, however, there still remains a proportion of hereditary breast cancer not yet determined. LBC accounts for approximately 10% of all breast cancers compared to the other major histologic subtype, invasive ductal carcinoma (IDC) [33]. Several factors suggest that LBC has a stronger hereditary basis relative to IDC, such as the higher frequency of bilateral disease [33], and also where excess familiality of LBC has been observed in population studies [34]. LBCs compose only 3% and 9% of the breast cancer tumour types seen in germline BRCA 1 and 2 mutation carriers, respectively [35], illustrating that the genetic risk factors for the majority of cases are unaccounted for by these genes.

The histology of LBC is characterized by infiltrative cancer cells which are isolated, highly dispersive and demonstrate a growth pattern with scattered and single files of tumor cells dispersed in stromal tissue [36]. This pathologic appearance is remarkably similar to DGCs and both LBC and DGC demonstrate characteristic mucinous, signet ring cells. This is not unexpected as E-cadherin staining is absent in 85% of sporadic invasive LBC [37] and somatic CDH1 mutations have been identified in 56% of sporadic LBCs [38]. Furthermore, in IDC, somatic CDH1 mutations are not found [38] and complete loss of E-cadherin expression is an uncommon feature. As loss of E-cadherin expression is a distinctive trait of both LBCs and DGCs, it likely contributes to the unique histopathologic features shared by the two cancers.

There are some differences with regard to the nature of the mutations seen in LBC and DGC. Generally mutations associated with sporadic LBC have been found to be nonsense or frameshift mutations [39] which encode truncated, non-functional proteins, whereas in sporadic DGC, mutations have generally been found to be splice site and in-frame mutations [11]. In sporadic LBC, mutations in CDH1 are spread throughout the gene [11] compared with the mutations seen in sporadic DGC which tend to cluster. Germline CDH1 mutations associated with DGC and/or LBC occur throughout the gene (Fig. 2). However, when the DGC and LBC associated CDH1 mutations are tabulated and compared based on their 3′ or 5′ positions relative to the halfway point of the CDH1 coding sequence (1324 or the start of exon 10), LBC associated mutations show a statistically significant trend towards clustering at the 3′ end (Fisher’s exact test, two-tailed P-value equals 0.0467) (Fig. 2). As this association is of weak statistical significance, it is unlikely to impact clinical testing strategies. Future analyses of novel germline LBC-associated CDH1 mutations should help to confirm this observation. Another difference between the molecular genetics of the two types of cancers, is that in sporadic LBC, silencing of E-cadherin expression is generally accomplished by a mutation in one allele in combination with loss of heterozygosity (LOH) or promoter hypermethylation in the remaining allele [40]. This is in contrast to sporadic DGC, where biallellic inactivation is achieved by mutations in one allele in concert with promoter hypermethylation in the other [41].

We have recently identified a truncating germline CDH1 mutation in an LBC family where analysis of the tumour was suggestive of partial LOH in the WT allele [9]. Our current case demonstrates a germline CDH1 mutation (1565 + 1G > A) in a predominantly breast cancer family, which is predicted to disrupt splicing. The mutation is in the same conserved position as a previously reported mutation (1565 + 1G > T) which was found in an Arabian HDGC family with no recorded history of breast cancer [42]. Moreover, a previous study reported a germline missense mutation in a proband with LBC but did not detail family history, or functionally characterize the missense mutation [43]. These examples demonstrate the need for further studies of germline mutations in LBC families in order to determine the mutation frequency and potential genotype-phenotype correlations.

Breast cancer has been observed in HDGC kindreds to the extent where clustering of LBCs within HDGC families has led to the misclassification of families as breast cancer kindreds who test negative for BRCA1/2 mutations [4]. In 1998, Keller described the first case of histologically defined LBC in association with HDGC [5]. Since then, several more HDGC families with associated breast cancer were reported where it was observed, that these cases were LBCs when pathology was available [4, 6‐8].

Prior to establishment of the association between HDGC and LBC, several efforts to determine whether CDH1 was a breast cancer susceptibility gene were attempted in view of the well-recognised phenotype of loss of E-cadherin expression displayed by the breast cancer subtype. For various reasons these studies failed to demonstrate the link. Rahman et al. examined 65 cases of lobular carcinoma in situ, however did not pre-screen the cases based on family history and included a wide age range, from 26 to 71 years, not necessarily in keeping with the usual age of onset seen in hereditary cancer syndromes [44]. Salashor examined 19 breast cancer tumours exhibiting LOH at the CDH1 locus, however of those, only 3 were confirmed to be pure LBC or mixed LBC/IDC pathology [45]. Lei examined 13 familial LBC cases and found no mutations, however did not define the extent of the family history [46].

Penetrance data based on 11 HDGC families, estimated the cumulative risk for LBC for female mutation carriers to be 39% (95% CI, 12–84%) by 80 years of age [47]. More recently we have published an estimated cumulative risk for breast cancer for females by the age of 75 years as being 52% (95% CI, 29–94%) from analysis of 4 predominantly gastric cancer pedigrees from Newfoundland with the 2398delC CDH1 founder mutation [4]. This is with the caveat that LBC risk for CDH1 mutation carriers has been assessed within high risk HDGC families, leading to a potential ascertainment bias and underestimation of the role of CDH1 mutations in LBC development. To accommodate for this we have begun analysis of CDH1 mutations within familial lobular breast cancer families or those families ascertained through a relatively young index case with confirmed LBC and have found germline CDH1 mutations in these kindreds [9].

At this time, it seems reasonable to conclude that at least four groups of women are at increased risk for LBC: women with LBC and a family history of breast cancer, women with a known CDH1 mutation, women from families with diffuse gastric cancer in whom no CDH1 mutation has yet been identified; and women with a germline BRCA2 mutation. Since there has not yet been a large population based study of the prevalence of CDH1 mutations among women with lobular breast cancer, it is premature to recommend genetic evaluation to women with a family history of breast cancer unless, at the very least, one of the breast cancers can be shown to have been lobular. Additional research can be expected to provide better guidance for these families.

Although there are not yet definitive data available on surveillance or risk reduction programs for women with known CDH1 mutations or untested women from CDH1-positive families, the high lifetime risk of LBC (39–52%) [4, 47] that these women face mandates their careful management. We suggest that they follow the recommendations for other high-risk women with hereditary breast cancer predisposition. This subgroup should be advised to practice breast self-examination; and to have annual mammograms, and semiannual clinical breast examination, beginning at least by age 30. There is certainly interest in regular bilateral breast MRI, as lobular breast cancer are known to frequently elude mammographic detection because they do not form masses or develop calcifications. These women can also be counseled to consider hormonal chemoprevention, since most LBCs are estrogen receptor positive [33], and both tamoxifen and raloxifene reduce the risk of estrogen receptor positive [48, 49] breast cancers in randomized trials. In addition, the risk reduction was greatest with both agents in women with lobular carcinoma in situ [50].

Prophylactic mastectomy may also be considered an option by some CDH1-positive women, particularly those who have been previously diagnosed with breast cancer in one breast or those who have had to undergo multiple biopsies for abnormal clinical findings. Several studies have reported a 90% reduction in breast cancer incidence with prophylactic mastectomy among women with a strong family history or with a germline BRCA1 or BRCA2 mutation [51, 52]. The published series include some lobular breast cancers, but not at numbers sufficient to permit meaningful subset analysis at this time.

Penetrance studies examining data from HDGC families, have estimated the lifetime risk of developing gastric cancer by age 75 and 80 respectively, to be from 40–67% in men, to 63–83% in women [4, 47]. Although identification of germline CDH1 mutations has enabled a significant proportion of HDGC families to utilise predictive testing to determine their individual risks of GC within CDH1 mutation positive pedigrees, unfortunately screening for DGC is ineffective and the current recommendation is for consideration of prophylactic gastrectomy in mutation positive individuals. Positron emission tomography [53] and chromoendoscopic-directed biopsies [54] have been proposed over basic endoscopy as more sensitive means of screening carriers, however screening methods have been consistently undermined by the recurrent discovery of multifocal DGC lesions underlying normal mucosa in prophylactic gastrectomy specimens of individuals with recent negative screening [4, 55, 56]. Regardless of the current limitations of screening, it is currently recommended that consideration for genetic testing and screening begin in at risk individuals in the late teens or early twenties [4] and that prophylactic total gastrectomy be considered in the early twenties for mutation carriers. Female mutation carriers will need specialized counseling to the potential nutritional effects on pregnancy following gastrectomy [57]. Further studies are currently underway to examine the quality of life impact of prophylactic gastrectomies. In the case report herein, although there was a GC in the maternal grandfather, the family history was more striking for the large number of breast cancer cases. This highlights the particular challenges we currently face with regard to counseling these families which appear to be mainly breast cancer, as it is unknown if the penetrance of DGC in this family is as high as it is in other HDGC pedigrees.