Mutation analysis of FANCD2, BRIP1/BACH1, LMO4 and SFN in familial breast cancer

Pathogenic mutations in BRCA1, BRCA2, TP53, PTEN, ATM and CHEK2 account for approximately a third of high-risk breast cancer families, suggesting that other breast cancer susceptibility genes exist [1–5]. Given the number of candidate breast cancer susceptibility genes, any approach to their identification needs to be focussed. Genes whose products are known to interact with BRCA1 and/or BRCA2, or are down-regulated in breast tumours, are particularly attractive candidates, and can be prioritised for investigation.

FANCD2 is one of eight genes known to cause the fatal human autosomal recessive disorder Fanconi anaemia (FA) [6, 7]. FA is a heterogenous condition characterised by progressive bone marrow failure, congenital abnormalities, hypersensitivity to DNA damaging agents and, most importantly, an increased risk of developing cancer [8]. There are currently eight cloned FA genes (FANCA, FANCC, FANCD1/BRCA2, FANCD2, FANCE, FANCF, FANCG and FANCL), all of which interact with each other in a common cellular pathway [6, 7]. Five of the FA proteins (FANCA, C, E, F, and G) form a constitutive complex in the nucleus of normal cells [9]. With the help of the recently identified ubiquitin ligase protein PHF9 (or FANCL), this multisubunit nuclear complex mediates the monoubiquitination of the FANCD2 protein at lysine 561 in response to the S-phase of the cell cycle or DNA damage [6]. The activated FANCD2 protein is then translocated to chromatin and DNA-repair foci, where it co-localises with other DNA repair proteins such as BRCA1, BRCA2, ATM, NBS1 and RAD51 [9]. Interestingly, this translocation has been recently identified to be BRCA1 dependent, suggesting that FANCD2 and BRCA1 interact in this process [6]. In response to ionising radiation, FANCD2 is also phosphorylated by ATM on serine 222, which leads to the activation of an S-phase checkpoint of the cell cycle [9]. FANCD2 is located at 3p25.3 and consists of 44 exons, encoding a protein of 1,451 amino acids. Houghtaling et al. [10] showed that FANCD2 homozygous and heterozygous mice display a high incidence of epithelial tumours, including mammary and ovarian carcinomas. These mice display other features found in BRCA2 mutant mice, including germ-cell defects, small size, and perinatal lethality [11]. FANCD2, like BRCA2, may, therefore, play an important role in the recombination DNA repair pathways [10]. The FA pathway has also been implicated in ovarian cancer, as the FANC-BRCA pathway was shown to be disrupted in a subset of ovarian tumour lines [12]. Furthermore, the 3p25-26 region of the human genome has been shown to have a high incidence of loss of heterozygosity in ovarian tumours [13]. Analysis of the FA genes (FANCA, B, C, D1, D2, E, F, G) in 88 non-BRCA1, non-BRCA2 breast cancer families failed to identify any penetrant mutations, but none of these families were known to share a haplotype around the relevant FANC genes, or to include cases of ovarian cancer [14].

BRIP1/BACH1 was first isolated and identified by using a glutathione S-transferase fusion protein containing the BRCT motifs and the carboxyl terminus of BRCA1. This protein was originally named BACH1 (for BRCA1-associated carboxy-terminal helicase 1), but is also known as BRIP1 (for BRCA1 interacting protein 1) [15]. The BRIP1/BACH1 gene maps to 17q22 and contains 20 exons, encoding a protein of 1,249 amino acids. Amino acid residues 888 to 1,063 of BRIP1/BACH1 interact with the BRCT domain of BRCA1 during the process of DNA repair [15]. Cantor et al. [15] screened the BRIP1/BACH1 gene for mutations in 21 sporadic breast/ovarian cancer cell lines, and 65 individuals with early onset breast cancer. Two germline heterozygous missense variants (p. P47A and M299I) were detected in the germlines of two early onset breast cancer patients but no family members were available for segregation analysis. Both variants are within the helicase domain of BACH1 (residues 1 to 888), with P47A located in the highly conserved nucleotide binding box, and M299I situated between two other conserved motifs [15]. Two other studies looking at variants in the BRIP1/BACH1 gene in breast cancer families failed to find any highly penetrant mutations, although these studies were limited in their sample size, and the number of available samples from additional family members, and none of the families were known to share a haplotype around BRIP1/BACH1 [16, 17].

LMO4 is a member of the LIM-only (LMO) family of transcription regulators. The four known members of this group (LMO1 to LMO4) are composed of two LIM domains and are thought to function as transcriptional cofactors via protein-protein interactions (reviewed in [18]). LMO1 and LMO2 overexpression is linked to T-cell tumourigenesis and LMO4 has been associated with breast oncogenesis, where overexpression is observed in approximately 50% of breast cancer cell lines and primary breast cancers [19]. Furthermore, overexpression of LMO4 induces mammary hyperplasia in transgenic mice and may be a predictor of poor outcome in breast cancer [20]. The presence of LMO4 in a complex containing the binding partners Ldb1, CtIP and the familial breast cancer tumour suppressor BRCA1 provides further compelling evidence for LMO4 playing a significant role in breast cancer pathogenesis [21], and activating mutations might be predicted to occur in some tumours and even in the germline of some patients. Although no activation mutations have been found, one somatic truncation mutation of LMO4 has been reported in a sporadic breast tumour [22]. This finding, as well as the deregulation of LMO4 expression in breast cancer and the interaction between LMO4 and the tumour suppressor BRCA1, prompted us to screen non-BRCA1/2 familial breast cancer cases for genetic alterations in LMO4 that may contribute to pathogenesis.

Stratifin (SFN; 14-3-3 σ; HME1) was first identified by serial analysis of gene expression (SAGE) analysis as an epithelial specific marker that was expressed at seven-fold lower levels in breast cancer cells compared to normal breast epithelium [23]. Recently, hypermethylation of SFN was detected in more than 90% of invasive breast cancers and was specifically associated with lack of expression [24]. In addition, methylation of this gene was detected in 83% of ductal carcinoma in situ and 38% of atypical hyperplasias but was unmethylated in all hyperplasias without atypia and normal breast epithelium obtained from patients without breast cancer [25]. Of most interest was the fact that SFN hypermethylation was also detected in the histologically normal adjacent breast epithelium in patients with breast cancer, suggesting that methylation of this gene may be an early event in breast cancer development. SFN is a negative regulator of cell cycle progression and is suggested to have an important function in preventing breast tumour cell growth, particularly at the G2 cell cycle checkpoint [26]. BRCA1 is a co-activator of SFN, and the expression of SFN is modulated by the BRCA1 status of the cell and requires intact BRCA1 and p53 to synergistically induce the optimal level of stratifin required for DNA damage response [27]. Interestingly, there is a nine-fold decreased expression of SFN in BRCA1- and BRCA2-related tumours compared to sporadic breast tumours [28]. SFN is located on 1p36.11 and is encoded by a single 747 base pair (bp) exon; 1p36 is a target of loss of heterozygosity in 16% to 37% of sporadic breast tumours [29, 30] and in 32% to 35% of familial tumours [31]. To our knowledge there has been no report of mutation analysis of SFN in familial breast cancer.

We sought to carry out mutation analysis of FANCD2, BRIP1/BACH1, LMO4 and SFN in a large number of non-BRCA1/2 breast cancer families. For the biggest genes, FANCD2 and BRIP1/BACH1, we screened a smaller number of families, but included those in which the affected family members shared a haplotype around the gene of interest. We also screened additional index cases for mutations in the FANCD2 exons that contain the ATM phosphorylation (S222) and the FANCD2 monoubiquitination regions (K561), and the BRIP1/BACH1 exons that contained the previously reported breast cancer-association variants, p. P47A and p. M299I.

Previous analyses of FANCD2 and BRIP1/BACH1 in non-BRCA1/2 families failed to identify any pathogenic mutations [14, 16, 17]; however, these studies did not choose the families to be screened on the basis of haplotype sharing, or the occurrence of other cancers (e.g. ovarian cancer in the case of FANCD2) in the family. Furthermore, some of these studies were limited by the fact that DNA from additional family members was not available for genotyping. To our knowledge, LMO4 and SFN have not been previously examined as BRCAx candidate genes. We therefore hypothesized that germline mutations in the BRCA1-interacting genes, FANCD2, BRIP1/BACH1, LMO4 and SFN, may account for some non-BRCA1/2 multiple-case breast cancer families. LMO4 and SFN are both small genes and so 247 and 267 index cases, respectively, were screened for these two genes. Because of the large size of the FANCD2 and BRIP1/BACH1 genes, however, we screened a smaller number of non-BRCA1/2 breast cancer families for mutations in these genes (30 and 75 families, respectively), but they were selected on the basis of all the available affected individuals sharing a haplotype around FANCD2 or BRIP1/BACH1, having at least one case of ovarian cancer (FANCD2), or having had full sequence analysis of BRCA1 and BRCA2 (BRIP1/BACH1). In addition, we screened a further 399 and 253 cases, respectively, for specific regions of the FANCD2 and BRIP1/BACH1 genes that contained functionally important domains, or variants previously found in the germline of breast cancer cases.

DHPLC analysis the FANCD2 gene indicated a high level of conservation in the coding sequence, indicated by a paucity of missense changes. Only six coding variants (c. 633 C>T, 1122 A>G, 1440 T>C, 1509 C>T, 2148 C>G and 3558C>G) were identified in the FANCD2 gene, all of which were silent. c.633 C>T and 2148 C>G did not appear to dramatically affect the predicted RNA splicing or folding by in silico analyses and were, therefore, assumed to be neutral SNPs. 3558 C>G (L1186L) was suggested by the BDGP Splice site predictor program (but not SpliceSiteFinder) to result in the complete loss of the donor site, possibly resulting in the misplicing of exon 35. RT-PCR analysis of a lymphoblastoid cell line from the carrier failed, however, to identify any aberrant transcripts, suggesting that this variant is unlikely to be pathogenic. The L1186L variant was not identified in any of 93 controls, but it did not segregate with breast cancer in the single family in which it was found. Therefore, this variant was also assumed to be a rare, neutral SNP.

Two common haplotypes of the FANCD2 gene were identified, one of which (haplotype B) was identical to the reference sequence obtained from the NCBI database. All of the rare variants were found to occur on haplotype B. Even though individually these variants were classified as neutral SNPs, an association study designed to test whether the two haplotypes confer different breast cancer risks would be worthwhile.

In the analysis of the BRIP1/BACH1 gene, we did not observe the two previously reported variants, p. P47A and p. M299I, in the 253 non-BRCA1/2 breast cancer cases [15]. However, we did identify three non-conservative missense variants (p. A144T, p. L195P, and p. G1155E) and one novel frameshift mutation (c. 3401delC) in the 75 selected non-BRCA1/BRCA2 breast cancer index cases. None of these variants were found in 93 controls. Additional genotyping of a total of 68 family members indicated, however, that these variants are not the underlying cause of breast cancer in these families, as none of the other affected relatives carried the variants. Nevertheless, it is possible that these variants are low-risk breast cancer susceptibility alleles, in which case further investigation may be warranted. The 3782 T>C 3' UTR variant of BRIP1/BACH1 is predicted to alter the folding of the transcript; however, the biological significance and frequency of this change in the normal population has yet to be determined.

Mutation analysis of 82 sporadic tumours previously revealed one somatic frameshift mutation of LMO4 [22]. No activating or inactivating coding or splice site mutations of LMO4 were found by DHPLC analysis of 247 index cases from non-BRCA1/2 families. Two intronic variants were found, each in two index cases. Their recurrent nature in two families and apparent lack of effect on splicing suggests that they are rare SNPs.

SFN is markedly down-regulated in breast cancer tissue compared to normal mammary epithelium but to our knowledge has not been evaluated for germline mutations in familial breast cancer. We screened the majority of the coding region of the gene in 267 index cases and found one silent change, T207T, in 23 index cases. This silent change has been previously reported (rs11542704) [43]. We also found some variants in the 3' UTR, and three different missense changes in one individual (F198L, L218I and Q244K) that were considered unlikely to be pathogenic because of the multiple occurrences in one individual. None of these variants have been previously reported.

Mutation analysis of the BRCA1-interacting genes FANCD2, BRIP1/BACH, LMO4 and SFN in a large number of non-BRCA1/2 breast cancer families did not identify any highly penetrant, pathogenic mutations. Given that DHPLC is a robust and sensitive screening technique, we consider it unlikely that we missed any coding or splice site pathogenic mutations among the index cases analysed. In particular, we analysed each PCR fragment at all the temperatures recommended by the DHPLC melt algorithm and under these conditions DHPLC has been reported to have a sensitivity of 99.4% [40]. It appears unlikely, therefore, that FANCD2, BRIP1/BACH, LMO4 and SFN account for more than a small proportion of inherited forms of breast cancer. Many novel SNPs were identified in these genes, however, and large association studies of breast cancer cases and controls is warranted to determine whether any of these variants confer small risks of breast cancer.