Detection of ATM germline variants by the p53 mitotic centrosomal localization test in BRCA1/2-negative patients with early-onset breast cancer

Background Variant ATM heterozygotes have an increased risk of developing cancer, cardiovascular diseases, and diabetes. Costs and time of sequencing and ATM variant complexity make large-scale, general population screenings not cost-effective yet. Recently, we developed a straightforward, rapid, and inexpensive test based on p53 mitotic centrosomal localization (p53-MCL) in peripheral blood mononuclear cells (PBMCs) that diagnoses mutant ATM zygosity and recognizes tumor-associated ATM polymorphisms. Methods Fresh PBMCs from 496 cancer patients were analyzed by p53-MCL: 90 cases with familial BRCA1/2-positive and -negative breast and/or ovarian cancer, 337 with sporadic cancers (ovarian, lung, colon, and post-menopausal breast cancers), and 69 with breast/thyroid cancer. Variants were confirmed by ATM sequencing. Results A total of seven individuals with ATM variants were identified, 5/65 (7.7 %) in breast cancer cases of familial breast and/or ovarian cancer and 2/69 (2.9 %) in breast/thyroid cancer. No variant ATM carriers were found among the other cancer cases. Excluding a single case in which both BRCA1 and ATM were mutated, no p53-MCL alterations were observed in BRCA1/2-positive cases. Conclusions These data validate p53-MCL as reliable and specific test for germline ATM variants, confirm ATM as breast cancer susceptibility gene, and highlight a possible association with breast/thyroid cancers. Electronic supplementary material The online version of this article (doi:10.1186/s13046-016-0410-3) contains supplementary material, which is available to authorized users.


(Continued from previous page)
Abbreviations: A-T, Ataxia telangiectasia; ATM, Ataxia telangiectasia mutated; ATR, Ataxia telangiectasia and Rad3 related; BRCA, Breast cancer susceptibility gene; CHK2, Checkpoint kinase 2; EBV, Epstein barr virus; FANC-A, Fanconi anemia complementation group A; HBC, Hereditary breast cancer; HBOC, Hereditary breast and ovarian cancer; LCLs, Lymphoblastoid cell lines; MRE11, Meiotic recombination 11; NBS1, Nijmegen breakage syndrome 1; p53-MCL, p53-Mitotic centrosomal localization; PALB2, Partner and localizer of BRCA2; PBMCs, Peripheral blood mononuclear cells; PHA, Phytohemagglutinin; SMC1A, Structural maintenance of chromosomes 1A; WRN, Werner syndrome Background Biallelic mutations in the ATM gene cause Ataxiatelangiectasia (A-T), a rare autosomal recessive multisystemic disorder characterized by progressive cerebellar ataxia, immune defects, insulin-resistant diabetes, radiosensitivity, and high risk for malignancy [1,2]. The ATM gene spans approximately 160 Kb of genomic DNA containing 66 exons [3] and encodes ATM protein, a serine/ threonine kinase mainly involved in DNA damage response pathways following DNA double strand breaks [4]. An enormous number of mutations (more than 600) can occur in the coding and noncoding regions of the ATM gene without hotspots [5]. In A-T patients, the large majority of ATM mutations are protein-truncations or splice-junction variants that can be easily distinguished by the numerous ATM polymorphisms [6,7].
Heterozygous carriers of variants in the ATM gene (from here on, ATM carriers) are usually asymptomatic and largely considered healthy carriers. However, they have been reported to be more sensitive to ionizing radiation and susceptible to ischemic heart disease, diabetes, and cancer, particularly of the breast, but also digestive tract and lung [2,8]. Many association studies have been performed on breast cancer susceptibility. Initially, epidemiological studies on relatives of A-T patients revealed a two to fivefold increased in the risk of breast cancer for female obligate ATM carriers [9]. The increased risk of breast cancer in ATM carriers was then confirmed by direct ATM sequencing in breast cancer cases compared to controls [10] and ATM is now considered a moderate-penetrance cancer susceptibility gene in BRCA1/2-negative patients with familial early-onset breast cancer [11]. Along with A-T associated mutations, several ATM screenings in cancer patients identified missense ATM variants, particularly amino acid substitutions that are not expected to be associated with A-T [12]. However, discrimination of these ATM variants from ATM polymorphisms and their contribution to health risks is still controversial. In addition, distinguishing between deleterious and neutral ATM alterations is required to allow the definition of standard-of-care clinical guidelines for the management of ATM carriers and their families [11].
Systematic review and meta-analysis of ATM sequencing data have been conducted to evaluate the health risks for parents and siblings of A-T patients, but similar large-scale screenings for ATM carriers in the general population by direct sequencing are not costeffective yet [13].
Recently, we have developed a rapid, reliable and nonexpensive test based on the ATM-dependent p53mitotic centrosomal localization (p53-MCL) that clearly discriminates ATM carriers of A-T mutations and at least some of the ATM cancer predisposition variants in lymphoblastoid cell lines (LCLs) and PBMCs. At variance with other diagnostic tests, the p53-MCL assay does not measure a continuous quantitative variation (e.g., radiosensitivity, ATM protein levels, phosphorylation of ATM targets) but a "binary" outcome. Indeed, at the single cell level, p53 does or does not localize at the centrosomes while, at the cell-population level, the number of cells showing one or the other phenotype allows to unambiguously diagnose A-T homozygotes and ATM carriers [14].
In a preliminary set up and validation of the p53-MCL test, we showed that it is highly sensitive, specific, and precise. In particular, we assessed the specificity by analyzing LCLs from monogenic disease carrying mutations in a series of DNA-damage related factors, such as MRE11, NBS1, SMC1A, WRN, ATR, FANC-A, and p53 [14]. In addition, p53-MCL test revealed 7 ATM carriers among 80 sporadic breast cancer patients. By direct ATM sequencing of 3 of these carriers, we identified the cancer-prone intronic c.8786 + 8A > C variant [15] in one patient and the c.2572 T > C (p.F858L) missense mutation [16] in other two patients. No ATM carriers were observed in a comparable cohort of healthy donors [14]. These data support p53-MCL as promising candidate test for cost-effective, large-scale screenings of ATM carriers.
Here, we examined validity and specificity of p53-MCL analyzing 15 LCLs from familial breast and ovarian cancer cases and fresh PBMCs from a total of 496 cases including BRCA1/2-positive and -negative familial breast and ovarian cancer and different sporadic cancers.

Patients
A total of 496 cancer patients were enrolled from 2010 to 2015 at three different Italian institutes: Policlinico Umberto I (University La Sapienza), Sant' Andrea Hospital (University La Sapienza), and Regina Elena National Cancer Institute -IRCCS. In particular, according to previously described criteria [17], we selected 90 unrelated families affected with breast and/or ovarian cancer after interview at the Hereditary Tumors Counseling Centre of the Policlinico Umberto I. Pre-test counseling was performed by an expert cancer risk counselor and the probands analyzed in this study belonged to different high-risk classes. During the genetic counseling, we calculated the a priori probability of carrying a pathogenic BRCA1/2 germline mutation by the statistical model BRCAPRO that considers: proband health state (un/affected); current age and age at diagnosis of the proband and all family members of four consecutive generations (only first/second degree relatives); typology of the existent tumors (unilateral/bilateral breast cancer, other cancers) [18]. Then, 69 patients with breast and thyroid cancer regardless of the sequence of appearance were recruited from Sant' Andrea Hospital and Regina Elena National Cancer Institute. Others 337 patients unselected cancer patients diagnosed at any age were recruited from Regina Elena National Cancer Institute.

p53-MCL test
Proliferating cells (i.e., LCLs or PHA-stimulated PBMCs) were set up for p53-MCL test as previously described [14]. Cells were examined under an Olympus BX53 microscope equipped with epifluorescence. Percentages of p53-MCL were measured by counting 100 cells in metaphase and analyzing two coverslips for each sample. The percentage of p53 mitotic localization is from 75 to 90 % for normal subjects, from 40 to 55 % for ATM carriers, and from 0 to 30 % for A-T patients.

BRCA1/2 gene sequencing
Genomic DNA was extracted from peripheral blood of all probands using commercial kit (Qiamp Blood Kit, Qiagen, Hilden, Germany). The entire coding sequence and each intron/exon boundary of BRCA1 and BRCA2 were screened by direct sequencing. All truncating and/ or novel genetic variants were confirmed by sequencing different samples on both DNA strands. Sequencing was performed using the BigDye Terminator v3.1 Cycle Sequencing Kit and a 3130xl Genetic Analyzer (Applied Biosystems, CA, USA). Reference sequence for BRCA1 was Genebank NM_007294.3, NG_005905.2 and reference sequence for BRCA2 was Genebank, NM_000059.3, NG_012772.3. BRCA1/2 genomic rearrangements were searched by the Multiple Ligation dependent Probe Amplification (MLPA) methodology. MLPA procedure was carried out according to the manufacturer's instructions. Variations in peak height were evaluated comparing each sample with a normal control and by a cumulative comparison.

ATM gene sequencing
Genomic DNA was extracted from PBMCs by Quick-gDNA MiniPrep (Zymo Research, CA, USA) according to the manufacturer's instructions. Sixty-two ATM exons were amplified using AmpliTaq Gold (Applied Biosystems, CA, USA), as described [14], and subjected to direct sequencing at the Genechron Laboratory (Rome, Italy). Reference sequence for ATM was Genebank U82828.1.

In silico analysis
To predict possible impact of amino acid substitutions on the structure and function of human proteins, the PolyPhen-2 (Polymorphism Phenotyping v2) software, that uses straightforward physical and evolutionary comparative considerations [19], was employed.
Efficient splicing of many exons requires splicingenhancers to promote splicing at unfavorable splice-sites or splicing-silencers to repress more favorable splice-sites nearby. As a consequence, single nucleotide changes in an exon or intron close to these splice-sites may be predicted to disrupt splicing. For in silico prediction of the effects of mutations on normal splicing, the interactive biosoftware Alamut v2.3 [20] was adopted.

Statistic
We determined statistical significance of differences between two groups by 2-tailed Student's t test. P values less than 0.01 were considered significant.

Results
Beyond the high penetrance BRCA1/2 genes, mutations of several cancer susceptibility genes, including ATM, CHK2 and PALB2, have been shown to associate, with a moderate penetrance, with familial breast and/or ovarian cancers [21,22]. Thus, we first verified p53-MCL test specificity compared to mutant BRCA1/2 and CHK2 proteins that are also involved in DNA damage response and centrosome amplification and localization [23][24][25]. LCLs from 15 familial breast and ovarian cancer ( Table 1) including four BRCA1-positive, 2 BRCA2positive, 1 CHK2-positive and 8 BRCA1/2/CHK2/ PALB2-negative LCLs were analyzed by the p53-MCL test. As shown in Table 1 and Fig. 1a, the percentage of p53-MCL was compatible with that of wild-type cells in 13 out of 15 LCLs, while two cases showed a p53-MCL reduction typical of ATM carriers, one in a BRCA1positive case and the other in a BRCA1/2/CHK2/ PALB2-negative case. Direct ATM sequencing confirmed the presence of ATM variants (Table 1) in both cases demonstrating p53-MCL specificity for ATM variants also in comparison with mutations in the centrosome related factors BRCA1/2 and CHK2.
Next, we analyzed 90 cases of familial breast and/or ovarian cancer ( Table 2). All 90 patients were screened for mutations in the BRCA1 and BRCA2 genes and 20 were found to carry pathogenic variants with an overall mutation rate of 22.2 % (Table 2). In particular BRCA1 pathogenic mutation recurred in about 26 % (8/30) of the Hereditary Breast and Ovarian Cancer (HBOC) families and in about 7 % (4/60) of the Hereditary Breast Cancer (HBC) families, while BRCA2 pathogenic mutation occurred in about 13 % (4/30) and 7 % (4/60) of the HBOC and HBC families, respectively. Of the 20 mutation-positive probands, 11 had breast cancer alone (5 BRCA1 and 6 BRCA2), three had ovarian cancer alone (3 BRCA1), three had bilateral breast cancer (2 BRCA1 and 1 BRCA2), two had both breast and ovarian cancer (1 BRCA1 and 1 BRCA2) and one had both bilateral breast and ovarian cancer (BRCA1). When fresh PBMCs from BRCA1/2-positive and -negative patients were analyzed by the p53-MCL test, five out of 90 cases showed aberrant p53-MCL (Fig. 1b). Interestingly p53-MCL positive cases accounted for 6.6 and 5 % of the HBOC and HBC families, respectively ( Table 2). Of note, this rate is similar to that of BRCA2 mutations [26].
As shown in Fig. 2 and Additional file 1: Table S1, 11 different ATM variants were detected among the 9 p53-MCL positive cases (i.e., 2 LCLs and 7 PBMCs). Specifically, in breast cancer cases, patient #1 presents c.4578C > T and c.1899-55 T > G variants. The c.4578C > T was previously described [27] and corresponds to a synonymous substitution (P1526P), which is predicted to create a new exonic site with increased affinity for the SRp55 splicing factor by an in silico analysis (Fig. 3a). The c.1899-55 T > G is a previously described intronic variant [28] predicted to increase the affinity for the SRp40 splicing factor (Fig. 3b). Patient #2 presents two variants not previously described in the literature nor in the ATM variation database (c.908A > C and c.5919-49C > T). Our in silico analyses predict a higher affinity for the splicing factor SC35, for both variants (Fig. 3c, d). The c.5557G > A (D1853N) variant presents in patient #3 is predicted to be "possibly damaging" by the PolyPhen-2 tool and has been intensively studied with respect to its possible association with breast cancer susceptibility [29]. Patient #4 and patient #5 present two deleterious variants, c.824delT and c.8833delCT, that induce early protein truncation at the level of exon 9 and 63, respectively, which have already been described in ATM families [30,31]. In the breast and thyroid cancer cases, patient #6 presents two already described variants, c.3161C > G and c.3576G > A. The c.3161C > G (P1054R) is a substitution predicted to be "possibly damaging" by PolyPhen-2 and implicated in breast and prostate cancer risk [32]. The c.3576G > A variant leads to exon 26 skipping and causes A-T syndrome in homozygosis [30]. Patient #7 presents the intronic variant c.4436 + 24A > G that in silico analyses predict a mild higher affinity for the splicing factors SC35 and SRp55 (Fig. 3e). LCLs BR317 presents c.146C > G (S49C) variant, a substitution predicted to be possibly damaging by PolyPhen-2 and reported as breast cancer susceptibility variant [16]. Finally, LCLs BR377 presents the same c.3161C > G (P1054R) variant found in patient #6.

Discussion
Genetic susceptibility plays an important role in several common chronic diseases including many types of cancer. Genetic testing for large-scale, general population screening can be very expensive and non-cost effective for National Health Services. Concerning the ATM gene, the complex genomic organization, the large number of polymorphisms, the absence of mutation hot-spots, and the frequent occurrence of variants of yet uncharacterized but predicted deleterious functions, make direct gene sequencing not yet a cost-effective approach. This is particularly evident for large-scale surveys of mutant/ variant ATM carriers for which gene sequencing is not sufficient to classify the rare hits. All these factors hinder both genetic counseling and clinical guidelines for risk management of ATM carriers and their families. The use of the recently developed p53-MCL functional test to detect ATM carriers might overcome at least some of these limitations. Here, we established the p53-MCL specificity for ATM in respect to the high-risk, BRCA1/2 breast/ovarian cancer susceptibility genes and confirmed p53-MCL as reliable test to detect variant ATM carriers in cancer patients. BRCA1/2 and ATM proteins share functional activities both in DNA damage response pathways and in centrosome regulation [33]. Thus, it was mandatory to establish whether a functional test, such as the p53-MCL test, would discriminate between ATM and BRCA1/2 variant carriers. Analysis of BRCA1/2-positive LCLs (n = 6) and PBMCs (n = 20), and of 1 CHK2-positive LCL established the specificity of p53-MCL test on ATM in respect to BRCA1/2 and CHK2 mutations (Table 1 and Fig. 1).
A total of seven individuals with ten ATM variants were identified by p53-MCL test in PBMCs: five breast cancer patients in HBC and HBOC families and two breast/thyroid cancer patients. The finding that all cases were among pre-menopausal breast cancer cases but not among patients with post-   [34] or by p53-MCL in sporadic breast cancer [14]. Interestingly, the a priori probability of carrying a pathogenic BRCA1/2 germline mutation measured by the BRCA-PRO score showed a mean for the entire cohort of 27.3 while the mean for the five ATM carriers was 52.2 % (Table 2). This difference is statistically significant (P =0.01) and supports the presence of a strong genetic component in the ATM carriers with earlyonset breast cancer, also considered that the BRCA-PRO score for the BRCA positive cases is very similar to ATM positive ones (50.1 % vs. 52.2 %). Evaluation of the histopathologic features of the breast cancers developed in the seven identified ATM carriers showed the presence of estrogen and progesterone receptors and the absence of HER2 receptor in four out of five cases that have been tested (Tables 4 and 5), confirming our previous results [14]. Thus, the presence of germline ATM variants recognized by p53-MCL appears to identify a subset of tumors with a more favorable biomarker asset, despite their earlier onset. In addition, a possible association between breast and thyroid cancer was highlighted by the identification of two ATM carriers among the sporadic cancer patients with breast/ thyroid cancer (2/69, 2.9 %), encouraging for further studies.
Familial breast and ovarian cancers are linked to highly penetrant mutations in the BRCA1/2 susceptibility genes that overall account for 20-25 % of hereditary breast cancers and 15 % of ovarian cancers [26]. Gene panel next generation sequencing approaches identified moderate-penetrant mutations in the ATM gene in 2.9 % [11] and 2.3 % [34] of BRCA1/2-negative cases. By the p53-MCL test, in this study we found 5.5 % (5/90) of ATM mutation  Most of the ATM mutations occurring in A-T patients are frameshift or nonsense mutations leading to protein truncation or splice junction variants [6,7] The role of A-T-causing mutations in cancer susceptibility (in particular breast cancer) is still debated and some studies have shown that a subset of rare, evolutionarily unlikely missense substitutions are important [15,35,36]. Here, the p53-MCL test identified 11 ATM variants in 9 breast and breast/thyroid patients.
Three variants (c.824delT; c.8833delCT; c.3576G > A) identified in 3 different patients are known to cause A-T in homozigosity [30,31]. Other 3 variants (c.5557G > A; c.146C > G; c.3161C > G) identified in 3 different patients have been shown to be associated with an increased cancer risk [16,29,32]. The last 3 patients present 5 different variants (c.4578C > T; c.1899-55 T > G; c.908A > C; c.5919-49C > T; c.4436 + 24A > G) that our in silico analysis predicted to increase the affinity for splicing factor and modify alternative splicing activities. Whether these predicted modifications alter ATM function/s and cancer predisposition remain to be evaluated. Since loss of p53 centrosomal localization was the functional readout through which we diagnosed the ATM carriers, we can conclude that these variants are at least able to impair the mitotic localization of p53 at the Fig. 3 In silico analysis using Alamut software. ATM variants identified in PBMCs derived from two familial breast cancer patients (BC patient#1 and #2) and one breast-thyorid cancer patient (BTC patient#7) lead to an increased affinity and/or appearance of a de novo sites for SR proteins centrosomes. It will be relevant to study whether all ATM variants induce this p53 defect or whether only functionally relevant variants are able to impair p53-MCL. Whether this impaired p53 localization has a role in tumorigenesis is presently unknown. Studying the mechanistic basis of p53 centrosomal localization will give insights on the contribution that different ATM variants with uncertain significance might have in cancer predisposition. Application of the p53-MCL test to LCLs or PBMCs with a broad spectrum of ATM variants will help to define these issues.

Conclusions
Our results show that p53-MCL test may offer the opportunity for screening of the general population and to identify the differences among deleterious, neutral and beneficial variants helping, in the future, to define the guidelines for ATM carriers not only in the A-T families.