E-cadherin (CDH1) is expressed on the cell surface in most epithelial tissues [1,2], and evidence is rapidly accumulating that the mammalian E-cadherin gene product plays a role in epithelial tumourigenesis [3]. Germline mutations in E-cadherin have been described in families with early-onset diffuse gastric cancer, and loss of function of this gene has been implicated in the pathogenesis of early-onset colorectal and breast cancers. Despite the fact that E-cadherin expression is decreased in 50% of invasive ductal carcinoma [4], mutations in the coding sequence of E-cadherin have not been observed in this type of breast cancer [5,6]. The role of E-cadherin mutation in development of hereditary gastric cancer has been shown, but its role in predisposing to breast cancer is still unproved. In order to further explore the involvement of E-cadherin in breast cancer, we examined all 16 exons of E-cadherin in 31 familial breast cancer patients, in whom BRCA1 and BRCA2 had been excluded as predisposing genes [7,8,9].

In a previous study of a family with diffuse gastric cancer (Fig. 1a) (Salahshor S, et al, manuscript submitted), we identified an E-cadherin germline mutation that cosegregated with the disease. This missense mutation in exon 12 (Ala592Thr) was also detected in the index patient's mother, who had ductal breast cancer. In an attempt to clarify a possible role for the Ala592Thr alteration in predisposing to breast cancer, we screened for this specific alteration in different series of breast cancer patients and control individuals. In total, 1328 patients with sporadic or familial breast cancer and 497 control individuals were analyzed for this specific alteration.

Nineteen probands with familial breast cancer who demonstrated LOH at chromosome 16q, and 12 patients from 10 families with breast, gastric and colorectal cancers were screened for germline mutations in the coding sequences of E-cadherin. Nine tumours from the 19 breast cancer cases were also screened for somatic mutations. The frequencies of the 1774G→A variant (Ala592Thr) were determined in DNA extracted from the blood of 358 unrelated probands with familial breast cancer, 214 unrelated early-onset breast cancer patients (age of onset <41 years), 126 unrelated BRCA1 or BRCA2 carriers, and 604 unselected breast cancer patients and 497 control individuals, who were considered to represent the general population in Sweden and Norway.

Two markers D16S7/p79-2-23 and APRT/HUAP15, which both map to 16q24.3, were used to identify LOH, as in a previous study of allelic loss at 16q in familial tumours [10].

Single-stranded conformation polymorphism (SSCP) and denaturing high-performance liquid chromotography (DHPLC; Transgenomic, Santa Clara, CA, USA) were used to screen 16 exons of the E-cadherin gene for the presence of alterations (Table 1). We used the same primers as described by Berx et al [5] and Salahshor et al (manuscript submitted), but with some modification. Samples that exhibited aberrant bands on the SSCP gel or an altered DHPLC pattern were sequenced by reamplifying the corresponding exon. Sequencing was performed either manually using the ThermoSequensase (Amersham Pharmacia Biotech, Uppsala, Sweden) kit, or by 377 ABI automated sequencer using the Taq-Cycle sequencing BigDye Terminator kit (Applied Biosystems Inc ABI, Stockholm, Sweden), according to conditions recommended by the manufacturers. The Ala592Thr variant frequency was detected using specific polymerase chain reaction (PCR)-restriction fragment length polymorphism tests (Salahshor S, et al, manuscript submitted).

The entire coding region of the E-cadherin gene was studied, including splice sites, using PCR/SSCP and sequencing in 10 families with histories of breast, gastric and colon cancer. No pathogenic germline mutation was identified in any of the individuals screened.

Negative results were also obtained by screening for germline alterations in 19 individuals with familial breast cancer who exhibited LOH at the E-cadherin locus in their tumours. We also searched for somatic alterations in nine of these 19 cases in whom tumours were available for analysis using Wave^® (Transgenomic), an HLPC-based nucleic acid fragment analysis system. One somatic mutation in an exon 2 splicing site (49-2A→C) was detected in the tumour that was of both ductal and lobular histology (Fig. 2). Clinical outcomes of these 19 patients were available, with 5-12 years of follow up. The two lobular breast cancer patients had both developed distant metastases, whereas distant metastases had developed in five out of 10 patients with ductal breast cancer. The patient with ductal-lobular histology and the mutation did not have distant metastasis 8 years after surgery. One previously reported common polymorphism in E-cadherin [11] was found in exon13 at codon 692 (GCC→GCT).

In a previous study of one family (family 26) with familial colon and gastric cancer (Fig. 1a) (Salahshor S, et al, manuscript submitted), we identified a germline missense alteration in exon12 (Ala592Thr) of E-cadherin. This alteration cosegregated with diffuse gastric cancer and colorectal cancer in this family, although the penetrance was not complete. The alteration was present in the index patient's mother, who had ductal breast cancer of comedo type at age 65 years (Figure 1a, II:12). In order to investigate whether this alteration was associated with an increased breast cancer risk we used a specific restriction enzyme digestion/PCR test to detect the variant, in an allelic association study of familial and sporadic breast cancer patients as well as control individuals who were considered to represent the general population.

We found the mutation in two out of 358 (0.56%) non-BRCA1/BRCA2 carrier women with familial breast cancer (families 205 and 2027). In family 205 (Fig. 1b), one woman with ductal comedo-type breast cancer at age 75 years had the Ala592Thr variant. The other available case from this family (Fig. 1b, II:2) tested negative, and the histology report was not available. In family 2027 (Fig. 1c), one woman with a breast cancer of ductal comedo type at age 37 years exhibited the Ala592Thr variant, whereas her two sisters with breast cancer at ages 45 and 51 years did not. One of the sisters had lobular breast cancer, but histopathological information on the third case was not available. Because the variant did not segregate with the disease, it is not likely that this alteration predisposes to breast cancer in either of these two families.

We also screened 126 BRCA1 or BRCA2 carriers from different families, and identified the mutation in one patient with breast cancer (family 4056) with a BRCA1 germline mutation (2594delC). It was possible to obtain a sample from her sister who also carried the BRCA1 mutation, and she was found to share the Ala592Thr variant also (Fig. 1d). Both tumours were of ductal comedo type. Because both of these tumours were caused by a germline BRCA1 mutation, the E-cadherin variant was not likely to have predisposed to breast cancer.

Among the 604 sporadic breast cancer patients, five were found to carry the Ala592Thr alteration (0.83%; Table 2). One of these was of lobular type and the other four were of ductal type. One of the ductal tumours was of comedo type, but it was not known for the other three whether they were of comedo type. Four cases with Ala592Thr alteration were also found among the 497 control patients (0.80%). The alteration was not found among any of the 214 early-onset breast cancer patients.

Thus far, E-cadherin germline mutations have been reported in 17 diffuse gastric cancer families [12,13,14,15,16]. The high rate of mutations found in the Maori families probably reflects the criteria used for selection of the families. In these families at least three first-degree relatives were affected by gastric cancer, and the age of onset of at least one case was under 40 years. In other studies, a low frequency of E-cadherin germline mutations in familial gastric cancer has been reported [37,38]. The frequency of E-cadherin germline mutations in breast cancer reported thus far is low [17,18]. Mutations in E-cadherin reported in lobular breast cancer cases are in most cases tumour restricted, and are not germline alterations [5,19]. In the 17 gastric cancer families with germline mutations, five members from three families were also affected by breast cancer [13,17]. Three of these patients from three different families had both breast and gastric cancer. One patient had developed both lobular breast cancer and diffuse-type gastric cancer [17]. In the other cases the gastric cancers were of diffuse type, and the histology of the breast tumours was unknown [13]. In one of these families, two patients with breast cancer only were also found; these cases were early-onset breast cancer.

In the present study, we did not find any pathogenic alterations in 10 families with cases of breast, gastric or colon cancer. We conclude that, although germline E-cadherin mutations are sometimes found in familial gastric and colon cancer, they are not frequently involved in families in which breast cancer appears to segregate as a part of an inherited predisposition for gastric and colon cancer.

The frequency of Ala592Thr alteration, which was first identified in one family with familial diffuse gastric, colon and breast cancer (Fig. 1a), was almost the same in the different groups studied here (Table 2). In familial breast cancer cases the frequency was 0.56% in non-BRCA1/BRCA2 cases, and 0.79% in BRCA1 or BRCA2 carriers. In the sporadic cases and in the control population, the frequencies were 0.83 and 0.80%, respectively. This finding does not support an effect of this alteration in predisposing to breast cancer in general. Thus, available data thus far indicate that germline E-cadherin mutations do not constitute a major risk factor for breast cancer. However, it is quite clear that somatic alterations can have an impact on the phenotype in lobular breast cancer. The fact that many breast cancer cases with the alteration (Ala592Thr) in the present study were of the ductal comedo type may indicate a genetic basis for the phenotypic divergence caused by this germline E-cadherin alteration. Although lobular breast carcinoma and ductal breast carcinoma of comedo type differ in many respects histologically, both lack formation of more differentiated mammary gland structures such as tubules. In lobular cancer, the cell characteristically grows in thin single-cell files (so-called Indian files). The comedo cancer is a poorly differentiated ductal cancer, which is rich in mitoses, often with large intraductal masses of cells with central necroses (comedons). Neither of these cancers carry a good prognosis, but individual prognosis with breast cancer is assessed mainly by grading rather than by histological type [20,21].

If E-cadherin functions as a classical tumour suppressor gene, then loss or inactivation of the remaining normal E-cadherin allele would be an important event in tumourigenesis in individuals who carry a germline mutation. In tumours with somatic E-cadherin mutations, there is often evidence of deletion of the wild-type E-cadherin allele [19], whereas in cases with germline mutations in the E-cadherin gene this is not usually seen [13,15,17]. We did not find any germline mutations in the 19 familial breast cancer cases with LOH at 16q studied here, which indicates that E-cadherin does not act as a predisposing factor for breast cancer in these families. In order to examine whether somatic alterations in this gene are involved in the carcinogenesis of these tumours, we screened for somatic mutations in nine available tumours from these cases, and we found one somatic mutation in one of the lobular cancer cases. In cases with no mutation in E-cadherin, however, inactivation of this gene could have been achieved by a variety of other mechanisms. Genetic alterations in any component of the cadherin–catenin complex appear to induce loss of adhesion function [39,40]. Furthermore, epigenetic changes such as promoter region hypermethylation, post-transcriptional alteration and aberrant phosphorylation of members of the cadherin-catenin complex can dysregulate E-cadherin function [41,42,43,44]. It is also possible that other gene(s) located at 16q are altered, rather than E-cadherin.