Mutation analysis and characterization of ATR sequence variants in breast cancer cases from high-risk French Canadian breast/ovarian cancer families

All common cancers show some degree of familial clustering [1]. Most of the familial aggregation, especially in breast cancer [2], results predominantly from inherited susceptibility [3]. Linkage studies in the 1990s led to the discovery of several predisposition genes associated with many rare familial cancer syndromes, thus providing fundamental insights into various pathways of carcinogenesis [4]. Nevertheless, this approach has mainly been limited to genes with relatively rare, highly penetrant alleles, for several reasons, such as a lack of power to detect alleles conferring modest or moderate risks that are believed to be involved in common cancers [1, 5‐7]. Analyses of risk attributable to such alleles in the known breast cancer susceptibility genes (e.g. BRCA1, BRCA2, TP53, PTEN, ATM) suggest they are responsible for ~25% of the familial component of breast cancer risk [6, 8, 9]. The number and properties of genetic variants that account for the remaining 75% of inherited risk are largely unknown. It has been proposed that a complex polygenic model is the best explanation for this missing genetic risk [10, 11] and perhaps the majority of breast cancers arise in a susceptible minority of women [2, 12].

Under the Common Variant/Common Disease (CV/CD) model, disease susceptibility is suggested to result from the joint action of several common variants, with unrelated affected individuals sharing a substantial proportion of disease alleles [13‐15]. The alternative is the heterogeneity hypothesis, which maintains that genetic susceptibility to common disease is caused by many different rare genetic variants, with a relatively large effect produced by each allele [16‐19]. If most cancer susceptibility is related to fundamental processes of cellular control, rare alleles might turn out to be the more important component and should be detectable by linkage analysis and/or the candidate gene re-sequencing approach [5, 6].

The central role of BRCA1 and BRCA2 genes in DNA repair, recombination, cell cycle control and transcription [20, 21] has led to the investigation of the implication of several similarly acting genes in breast and/or ovarian cancer predisposition, including ATM (Ataxia telangiectasia-mutated) [22‐27], CHEK2 [28, 29], TP53 [30], PTEN [31], STK11 [32] and a few other genes involved in DNA repair [33]. Ataxia-telangiectasia-mutated and Rad3-related (ATR) is a member of the phosphatidyl inositol-kinase (PIK)-related family which plays, along with ATM, a central role in cell-cycle regulation, by transducing DNA damage signals to downstream effectors of cell-cycle progression [34]. In response to double-strand breakage, stalled replication forks or DNA adducts, ATR complexed with ATR-interacting protein (ATRIP) is recruited and then phosphorylates a number of proteins involved in DNA damage, including H2AX, 53BP, TP53, NBS1 and CHEK1 [35‐38], thereby activating cell checkpoints, DNA repair or apoptosis. ATR is also able to bind to Rad17 and BRCA1 and to associate with components of the nucleosome remodeling and deacetylating complex [39‐41]. Furthermore, ATR has recently been shown to interact with the Fanconi Anemia complex [42], which growing number of evidences link to the two BRCA genes [[21], for review see [43]]. A recent study has also demonstrated that the Mre11/Rad50/NBS1 (MRN) complex, a central component in the cellular response to ionizing radiations and other causes of double-strand breaks, is required for ATR-dependant phosphorylation mechanisms of the protein Smc1 (Structural maintenance of chromosomes 1) [44]. ATR knockout studies showed that ATR is essential for somatic cell growth and genomic integrity in the embryo and that its deletion leads to genomic disruption and early embryonic lethality in mice [45, 46]. Moreover, it has been reported that disruption of the ATR gene leads to an increase in the incidence of large benign tumors in heterozygotes, possibly indicating that deficiency in ATR affects the rate of tumor initiation [45].

Based on the major role of ATR in cellular response to DNA damage and its multiple interactions with several proteins such as BRCA1 [40, 47], ATR represents an attractive candidate gene to potentially explain a fraction of the remaining breast cancer susceptibility. The current study was designed to assess the possible involvement of ATR germline mutations in breast cancer susceptibility. For this purpose, the complete sequence of the 47 exons and flanking intronic sequences of the ATR gene were analyzed in DNA samples from individuals affected with breast cancer from non-related BRCA1- and BRCA2-negative high-risk French Canadian breast/ovarian families.

Since it is well established that the residual familial risk of breast cancer, not caused by BRCA1 or BRCA2 genes, could be explained by a polygenic or high-risk genes heterogeneity model [72, 73], we selected individuals affected with breast cancer without mutations in BRCA1/2 genes from high-risk families (one individual per family), in order to increase the power of the study to find genetic variants involved in breast cancer susceptibility. So far, several genes have been investigated based on their interaction with BRCA1/2 or their involvement in DNA repair mechanisms. Since BRCA1/2 genes are intimately linked to genomic stability, other genes involved in this pathway are very good candidates to be BRCA3, and this is especially true of ATM and ATR which play a central role in genome stability maintenance. The ATM gene has been suspected to be a breast cancer susceptibility locus, due to the presence of breast cancer in A-T families, particularly among ATM heterozygotes [74]. ATM mutations have already been reported to increase breast cancer susceptibility [9, 27, 75], while some other sequence variants located in this gene do not seem to be linked to breast cancer [24].

Based on the similar roles played by ATM and ATR as sensors of DNA damages, ATR may be considered a putative candidate gene that could possibly explain a fraction of the remaining familial breast cancer risk. Association of ATR germline mutation with breast cancer susceptibility has been previously analyzed in Finnish 126 families [70], and no germline mutation was identified in this founder population. The current study, performed in a French Canadian cohort, also being a founder population, was designed to assess the possible involvement of ATR germline deleterious mutations in breast cancer predisposition.

No deleterious germline mutation leading to a premature termination of the protein were identified in the coding region. However, 41 sequence variants were identified, among which 16 were coding variants while 21 were novel changes. In addition we find it unlikely that neither of the common missense substitutions located in the FAT and kinase domains (c.6394T>G and c.7274G>A) have a significant effect on protein function because: (i) their frequencies are similar in cases and controls, especially for c.7274G>A whose MAF is greater than 20% in controls and (ii) these residues are not well conserved in other species (Table 4). Indeed, the polymorphisms displaying a significant deviation from HWE are composed of a group of 14 uncommon polymorphisms identified in the same 3 breast cancer cases (2 homozygotes and 1 heterozygote), and therefore this most likely constitutes a single relatively rare allele. It has to be stated that no particular characteristics seem to emerge for the families bearing any of these rare variants, as both the French Canadian and the validation families have been recruited on the basis of high-risk breast cancer families.

Comparison of polymorphism frequencies between our cohort and the Finnish cohort [70] is not fully informative since the latter does not distinguish the number of heterozygotes and homozygotes found in their cohort but only the number of carriers of a given polymorphism. However, if we also use this method to calculate polymorphism frequencies observed in our cohort, only SNP40 displayed a notably lower frequency than that found in the Finnish cohort. As stated earlier, both studies (Heikkinen et al. and the present study) have been designed to identify ATR deleterious germline mutations in breast cancer cases. No such mutation was found in either study, therefore ATR is unlikely to play a major role as a high penetrance gene in breast cancer predisposition. Even though novel variants have been identified, the possible involvement of polymorphisms or haplotypes observed in cases compared to those found in controls would need a lot more individuals to obtain a significant value of association to breast cancer susceptibility [76, 77]. We thus sought to identify tSNPs that could be useful to other studies and populations.

Our pairwise linkage disequilibrium analysis (Figure 2) did not seem to identify any distinct LD blocks within ATR. This observation is supported by the fact that SNP1 is in perfect LD with most other SNPs, including the most distal SNP41, and is also in accordance with what is seen in the French Canadian founder population which displays large conserved haplotypes as reported at the BRCA1 locus [49]. However, using the Haploview software, three distinct LD blocks were identified at the ATR locus when using SNPs showing a MAF >5% in healthy French Canadian individuals (Figure 3). The breakage of strong LD seems to be located in the region of exon 31, and between exon 43 and exon 47.

Based on the same algorithm (Haploview), and using the SNPs genotyped in HapMap database showing a MAF higher than 5%, two LD blocks could be identified; the first block comprising the SNPs located from intron 1 to exon 43, while the second block included all the remaining SNPs until exon 47. However, it should be noted that the majority of the SNPs used to determine haplotype blocks have a MAF higher than 0.4, which represent common SNPs found in many different populations and therefore probably exclude the SNPs specifically observed in our French Canadian founder population.

We were able to demonstrate that 8 tSNPs are sufficient to represent the majority of ATR haplotypes in our French Canadian individuals, which will greatly facilitate subsequent studies. Our results of 8 tSNPs at the ATR locus in our population is consistent with previously reported number of tSNPs required at other gene loci in other populations [78, 79]. We can therefore be quite confident that these tSNPs will be useful in subsequent analyses. Moreover, out of 72 SNPs genotyped in the HapMap database (HapMap data rel#20 on NCBI B35 assembly, dbSNP b125) at the ATR locus, only 40 displayed a MAF >5%. Among them, 7 tSNPs were identified, 3 of which have been identified as tSNPs in our analyses (rs10804682, rs2229032, rs1802904). Of the remaining four tSNPs identified in HapMap database (rs11920625, rs9856772, rs6805118 and rs9816736), three have not been genotyped in our cohort as they were located in intronic regions (>150–200 bp) and one (rs11920625) was not observed in our individuals.

Surprisingly, sequence analysis of this cDNA region in our immortalized cell lines revealed a deletion of the last 121 nucleotides of exon 33 instead of a skipping of exon 34 (or a portion of exon 34), as expected. This deletion of 121 nucleotides alters the ORF and results in a putative truncated protein of 1889 amino acids. Although interesting, this deletion is observed at similar levels in all tested individuals and is therefore unlikely related to the c.5739-4del9+T polymorphism. This Δ33 splice form may be explained by the weak wild-type donor site score of exon 33 (0.11) and the presence of an additional donor site located within exon 33, which exhibits a score of 0.63. No splice form involving the skipping of exon 34 has been identified when using specific primers located on the putative Δ33-35 or 33–35 exon junctions.

Splicing score analyses of exon 41 flanking intronic sequences were also analyzed since c.7041+8G/A could potentially affect the splicing in this region. While the exon 41 donor splice site showed a relative low splicing score in all species, the putative intronic donor site (splicing score of 0.90) located 441 nucleotides downstream of exon 41 became of interest, given its potential effect on splicing in this region (Figure 4). The insΔInt41 splice form could not be detected using standard procedures. However, this splice form has been amplified and subcloned by using specific primers located on this putative exon junction, demonstrating its very low mRNA expression (at the limit of detection). Due to this low expression, it was impossible to conclude whether or not this insΔInt41 splice form is associated with the c.7041+8G/A variant.

The ratio of Δ33 splice form/WT form being a potentially important factor regarding DNA repair and other related functions in genome stability, we performed QRT-PCR to estimate the relative abundance of WT and Δ33 splice form mRNAs, using TAQMAN probes to allow discrimination between both forms. No correlation was found between the presence of c.5739-4del9+T in either the heterozygous or homozygous state and the expression levels of the Δ33 splice form. However, it is very interesting that significant relative expression of the Δ33 splice mRNA is observed in breast and ovarian tissues, as well as in MCF7 and HaCat (human skin keratinocytes) cells (Figure 6A), especially since expression levels of the WT ATR form in these tissues (Figure 6B) seem to be relatively similar to other tissues and cell lines. The ratio of expression levels between both mRNAs could therefore be of primary issue regarding the effect of the balance of these transcript levels on cell integrity in different human tissues. However it should be noted that only one sample per tissue was analyzed, which by no means represents a mean expression in these tissues or cell lines. While alternative splicing within the non-catalytic domain of ATR mRNA transcript causing skipping of exon 6 had already been observed [80], in 2003 O'Driscoll and coll. [81] identified a founder mutation (2101A→G) in ATR that affects exon 9 splicing in two related Pakistani families affected with Seckel syndrome. This study also shows an impaired response to DNA damage in a cell line from an affected parent who carried the mutation. Further characterization of ATR-Seckel cells showed impaired phosphorylation of ATR-dependent substrates, impaired G2/M checkpoint arrest and supernumery centrosomes in mitotic cells, clearly demonstrating a role for ATR in the maintenance of centrosome stability [82]. More recently, two other splicing alterations of ATR have been reported in clinical samples with pyothorax-associated lymphoma [83].

No deleterious germline mutations have been identified in French Canadian breast cancer cases. However, we have conducted the first detailed haplotype tagging analysis of the ATR gene within control individuals from the French Canadian population. The data presented here clearly identified 8 ATR tSNPs, which will be useful for other large-scale association studies. We did not find any germline mutations in the ATR gene potentially involved in breast cancer predisposition. However, given that different splicing alterations of ATR have been associated with impaired response to DNA damage, the notably significant expression of the novel Δ33 splice form observed in breast and ovarian tissues could have a potential effect on DNA repair mechanisms in these cells, although exhaustive analyses should be required to verify this hypothesis. Further analyses in other populations and larger cohorts will be required to define the possible association of ATR gene polymorphisms with breast cancer susceptibility.