Ataxia-Telangiectasia Mutated Loss of Heterozygosity in Melanoma

ATM germline pathogenic variants were recently found enriched in high-risk melanoma patients. However, ATM loss of heterozygosity (LOH) has never been investigated in melanoma and, therefore, a causal association with melanoma development has not been established yet. The purpose of this study was to functionally characterize 13 germline ATM variants found in high-risk melanoma patients—and classified by in silico tools as pathogenic, uncertain significance, or benign—using multiple assays evaluating ATM/pATM expression and/or LOH in melanoma tissues and cell lines. We assessed ATM status by Immunohistochemistry (IHC), Western Blot, Whole-Exome Sequencing/Copy Number Variation analysis, and RNA sequencing, supported by Sanger sequencing and microsatellite analyses. For most variants, IHC results matched those obtained with in silico classification and LOH analysis. Two pathogenic variants (p.Ser1135_Lys1192del and p.Ser1993ArgfsTer23) showed LOH and complete loss of ATM activation in melanoma. Two variants of unknown significance (p.Asn358Ile and p.Asn796His) showed reduced expression and LOH, suggestive of a deleterious effect. This study, showing a classic two-hit scenario in a well-known tumor suppressor gene, supports the inclusion of melanoma in the ATM-related cancer spectrum.


Introduction
The ataxia-telangiectasia mutated (ATM) gene encodes a serine/threonine kinase protein that plays a critical role in the activation of cellular responses to DNA doublestrand breaks through downstream phosphorylation of central players in the DNA damageresponse pathways, including BRCA1, p53, and Chk2 [1].
Germline inactivation of both copies of the ATM gene causes Ataxia-Telangiectasia (A-T), an autosomal recessive disorder characterized by progressive cerebellar degeneration, oculocutaneous telangiectasias, immunodeficiency, and susceptibility to cancer, mainly leukemias and lymphomas [2]. Studies of A-T families showed that heterozygous ATM pathogenic variant (PV) carriers are at increased risk of breast cancer [3]. Since then, other malignancies, such as pancreatic cancer, have been linked to ATM through studies Int. J. Mol. Sci. 2022, 23, 16027 2 of 13 conducted on cancer cohorts. Indeed, ATM PV carriers are primarily identified outside of an A-T familial context through multigene panel testing. Therefore, cancer risks associated with ATM variants are of relevance to genetic counselling and cancer risk management of a much larger population [4]. However, the hypervariability due to the large size of this gene, and the fact that several of the identified variants are variants of uncertain significance (VUSs), pose a challenge to the inclusion of additional tumors in the ATM-related cancer spectrum.
In a recent multicentric international study on 2105 melanoma cases, we associated germline heterozygous variants in ATM with melanoma risk, therefore proposing ATM, which was previously established as a melanoma genome-wide association study (GWAS) hit [5,6], as a melanoma intermediate-risk gene [7].
However, loss of heterozygosity of the ATM gene as described, for example, for breast cancer [8,9] has never been investigated in melanoma. Therefore, a causal association with melanoma development has not been established yet. No histopathological and molecular features have been studied in melanomas developed by carriers of germline ATM PV or VUS. Therefore, the ATM impact on melanoma predisposition is unclear.
Somatic ATM mutations rarely occur in melanoma (2.6%), and the literature on ATM protein expression and its clinical significance in melanoma is scant, although Bhandaru et Al. suggested that both loss and gain in phosphorylated ATM (pATM) expression were unfavorably associated with survival in melanoma patients [10].
Based on these premises, the purpose of this study was to functionally characterize ATM variants classified as PV or VUS by in silico tools found in the germline of melanoma patients by determining ATM and pATM expression and assessing loss of heterozygosity (LOH) in patients' derived melanoma tissues and cell lines.

Classification and Distribution of Germline ATM Variants
Three germline pathogenic variants (PVs), eight variants of unknown significance (VUSs), and one benign (B) variant (Figure 1), classified according to the ACMG criteria [11], were studied by comparing them with an ATM wild-type control. The p.Met1484fs PV, which we previously described in patients diagnosed with multiple primary melanoma and pancreatic cancer [7,12], has never been detected in patients with AT, but was found in a patient with Metastatic Castration-resistant Prostate Cancer [13]. described in breast cancer [18,19]. Conversely, the p.Asp639Gly variant has only been reported once (https://www.ncbi.nlm.nih.gov/clinvar/RCV001013687/, accessed on 16 June 2022) and classified as VUS, but no associations with cancers or cancer-related syndromes have been observed, to our knowledge. The p.His 1436Tyr has been described in breast cancer [20] and prostate cancer [8], and the p.Val1570Ala has been described in breast and pancreatic cancer and in chronic lymphocytic leukemia [21][22][23]. For additional information on personal/family history of cancer and co-segregation analysis, see Table 1. Figure one shows the distribution of germline ATM variants included in this study. Of all the analyzed variants, two VUS, p.Val2873Ile and p.Arg2912Gly, are located in the kinase domain. The latter was described in a breast cancer patient [18] and has been found to result in a partial impairment of ATM kinase activity in a previous publication [24]. The remaining variants are distributed outside the gene's functional domains, with no particular clustering.  The p.Ser1135_Lys1192del and p.Ser1993ArgfsTer23 PV, as well as the p.Gly2023Arg likely benign (LB) variant, were previously described in either homozygous or compound heterozygous A-T patients [14,15], as well as in breast cancer patients [16][17][18]. Both the patient carrying the p.Ser1135_Lys1192del PV in this study and two other carriers not included in this study were found to share a common recurrent haplotype at the ATM locus already identified in Italian AT patients [14], suggesting a possible founder effect in this ATM variant. Eleven variants were found in high-risk melanoma patients reported in our previous studies [7,12], whereas one variant, the p.Asn796His, has never been described in melanoma but has been described in breast cancer [18,19]. Conversely, the p.Asp639Gly variant has only been reported once (https://www.ncbi.nlm.nih.gov/clinvar/RCV0010136 87/, accessed on 16 June 2022) and classified as VUS, but no associations with cancers or cancer-related syndromes have been observed, to our knowledge. The p.His 1436Tyr has been described in breast cancer [20] and prostate cancer [8], and the p.Val1570Ala has been described in breast and pancreatic cancer and in chronic lymphocytic leukemia [21][22][23]. For additional information on personal/family history of cancer and co-segregation analysis, see Table 1. Figure one shows the distribution of germline ATM variants included in this study. Of all the analyzed variants, two VUS, p.Val2873Ile and p.Arg2912Gly, are located in the kinase domain. The latter was described in a breast cancer patient [18] and has been found to result in a partial impairment of ATM kinase activity in a previous publication [24]. The remaining variants are distributed outside the gene's functional domains, with no particular clustering.

ATM Immunohistochemistry and LOH Assessment
IHC expression showed high concordance with the in silico classification of germline variants (Table 2).
Indeed, in 2 out of 3 patients carrying a germline PV (Mel_1 and Mel_3), loss of ATM and pATM expression was observed (Mel_3, Figure 2A) and correlated with the presence of the mutated allele at a high frequency as detected by DNA sequencing in the melanoma tissue, as well as in breast cancer tissue from one of the two patients (Table 2).
Mel_1, carrying the p.Ser1135_Lys1192del PV, showed loss of ATM and pATM expression in both breast cancer tissue and melanoma. According to previous studies conducted on non-melanoma cancers, this variant results in the skipping of ATM exons 26-27, with impaired/absent protein function [25,26]. LOH was confirmed by microsatellite analysis in both cases (Table 2). Therefore, expression and LOH analysis support the in silico classification of two out of three PVs as pathogenic for melanoma (p.Ser1993ArgfsTer23, p.Ser1135_Lys1192del).
Moreover, in Mel_3, whole-exome sequencing (WES) was performed on DNA extracted from fresh-frozen melanoma tissue (cutaneous metastasis), and the p.Ser1993ArgfsTer23 variant was found at an allele frequency of 74%. Considering that the aberrant cell fraction was 80% in this sample, we could deduce that the pathogenetic variant is present in all tumor cells ( Figure 3A). The same results were obtained by analyzing an additional fresh-frozen metastatic sample obtained from the same patient (muscle metastasis).
Conversely, Mel_2, bearing the p.Met1484fs PV, retained ATM and pATM expression by immunohistochemistry. However, no further tissues were available for LOH validation. Among the ten samples carrying VUSs, 8 retained ATM and pATM expression, and in one of them (Mel_12) in which microsatellite analysis was performed, the IHC results were consistent with the absence of LOH. Conversely, in two samples (Mel_4 and Mel_6), ATM and pATM expressions were equal to or lower than 50% ( Figure 2B) and LOH was confirmed in one of them (Mel_6) by microsatellite analysis. Sanger sequencing, when performed, confirmed LOH analysis data (see Table 2). As expected, the samples carrying the p.Pro604Ser benign and the p.Gly2023Arg likely benign variants retained both ATM and pATM expression in immunohistochemistry, in line with the in silico classification ( Figure 2C).  IHC expression showed high concordance with the in silico classification of germline variants (Table 2).
Indeed, in 2 out of 3 patients carrying a germline PV (Mel_1 and Mel_3), loss of ATM and pATM expression was observed (Mel_3, Figure 2A) and correlated with the presence of the mutated allele at a high frequency as detected by DNA sequencing in the melanoma tissue, as well as in breast cancer tissue from one of the two patients ( Table 2).   In the sample with the p.Ser1993ArgfsTer23 PV (Mel_3), LOH was confirmed by RNA sequencing conducted on two different metastatic melanoma samples, in which the mutant allele frequency was 94% ( Figure 3B) and 82%. Moreover, CNV analysis of WES data revealed a whole-gene deletion on the WT ATM allele in both samples, which was confirmed by Sanger sequencing and microsatellite analysis, showing loss of the WT allele in both tissues and in the melanoma cell line ( Figure 3C-E). Loss of ATM and pATM expression was also revealed by WB performed on the cell line established from the patient's fresh tissue ( Figure 3F).   In addition, we analyzed a melanoma tissue sample from one patient (Mel_15) with a somatic ATM variant classified as tier 1 [27] and no germline variants. We found the variant in a heterozygous status by both WES and Sanger sequencing in multiple lesions from the same patient, and used it as a heterozygous control. In this sample, IHC showed high expression of ATM and pATM ( Figure 3F). WB confirmed this result as cells expressed ATM and upmodulated the expression of its phosphorylated form upon irradiation at 4 Gray ( Figure 3F).

Discussion
The increasing availability of novel therapies that exploit vulnerabilities in the DNA Damage Response (DDR) machinery has prompted the investigation of DDR alterations in multiple cancers, both at the germline and the somatic level. Indeed, as for other cancers, loss of mechanisms maintaining genomic integrity could be one of the critical events in the pathogenesis of melanoma. Recently, germline PVs in the ATM gene have been found enriched in high-risk melanoma patients, and have been reported with a frequency of 1.06% in high-risk Italian melanoma patients [28]. However, the association of ATM with melanoma needs functional assessment starting from LOH analysis to definitively include this tumor in the ATM-related cancer spectrum.
With this study, we wanted to determine the role of ATM disruption in melanoma development by integrating multiple in vitro assays. In 2 out of 3 patients carrying PVs, the ATM protein was absent in melanoma tissue, hence supporting the hypothesis that loss of ATM tumor suppressor function is involved in melanoma development. For one of these variants-the founder p.Ser1135_Lys1192del-our results confirm, in melanoma, the negative effect on the ATM protein reported in previous studies [25,26]. In the third patient with a PV, however, ATM expression and kinase activity were apparently unimpaired according to IHC. Unfortunately, however, we could not confirm or contradict this result with a second technique. As expected, the second allele was retained in the analyzed tumor samples when the germline variant was classified as benign. VUS analysis allowed us to explore not only the effect of ATM on melanoma, but also the effect of the variant itself on ATM function. Indeed, a growing number of in silico tools are available to classify genetic variants. However, although those tools can predict pathogenicity through several sophisticated algorithms, in vitro/in vivo experiments are ultimately required to obtain conclusive proof of the nature of a variant. For the large majority of the VUSs analyzed in this study, we did not observe loss of the ATM protein in tumor tissue. Therefore, we cannot provide conclusive information on the pathogenicity of these variants. In the tumor samples of two patients with VUS, however, ATM expression was ≤50%, suggesting that these two germline variants represented the first of two hits that contributed to ATM deficiency and melanoma development. These data suggest that these two missense variants should probably be included in Class 4/5 of pathogenicity. However, further research is needed in order to reclassify them. In fact, our study had some limitations. For instance, although we employed multiple techniques to study the abundance of ATM in all tumor samples, the paucity of available material prevented us from applying all the techniques to each sample. However, for the p.Ser1993ArgfsTer23 PV, we were able to integrate concordant results obtained from seven different assays (Figure 3) in multiple tumor tissues and in an established melanoma cell line from the same patient.
Taken together, our results provide, for the first time, experimental data in support of the role of ATM impairment in melanoma development, showing a classic two-hit scenario in a well-known tumor suppressor gene. This is of particular relevance, as the identification of ATM as a melanoma-causative gene has potentially multiple clinical implications. Indeed, the inclusion of ATM in melanoma multi-gene diagnostic panels would increase the number of high-risk individuals identified through genetic testing who could benefit from dermatologic surveillance.
Moreover, this work provides the rationale for further studies, some of which are already ongoing, with the aim of exploring the potential actionability of DDR genes in selected melanoma patients. Specifically, in germline carriers, the deficiency of ATM and its related pathways could be exploited for personalized anticancer treatment approaches.

Selection of ATM Variant Carriers and Their Tumors
Eighteen tissue samples were retrieved for the present study from fourteen individuals belonging to the Genoa (N = 16) and Padua (N = 2) cohort of high-risk individuals carrying germline heterozygous ATM variants for whom melanoma tissue (and/or breast cancer and dysplastic naevus in two cases) was available at the Institutional pathology departments. In addition, one junctional nevus from a healthy individual carrying a heterozygous germline ATM variant (CTRL_1) and one melanoma tissue from a sporadic melanoma patient, a carrier of a heterozygous somatic ATM variant from which we established a melanoma cell line, were also included as controls (Mel_15). Heterozygous germline variants were classified according to the ACMG guidelines [11] and a subset had been previously described in the germline of melanoma patients [7,12] (Table 1).

Establishment of Melanoma Cell Lines
Human melanoma samples were obtained from patients selected at the IRCCS Ospedale Policlinico San Martino under local Institutional-Review-Board-approved protocols (CER Liguria: 046REG2017). Patient biopsies were placed in a sterile conical tube containing NaCl 0.9% and 1% antibiotic and antimycotic on wet ice and handled no later than 4 h after resection. Upon arrival, biopsies were manually minced using a sterile scalpel and dissociated using 70 µm cell strainers (Miltenyi Biotec) at room temperature. Cells were cultured in RPMI 10%FBS, 1%HEPES, 1%Na Pyruvate, 2%Glutamine, 1%Pen/Strep, and 1% antimycotic and stabilized as culture after the 8th passage. Positivity for melanoma antigens was assessed by flow cytometry (FACSCalibur; Becton Dickinson, East Rutherford, NJ, USA) in Mel_3 and Mel_15 cells using anti-CD146 (BD Pharmingen, cod 550315) and anti-MCSP (Miltenyi Biotec, cod 130-091-225) monoclonal antibodies.
A549, a NSCLC cell line expressing functional ATM, was purchased from the ICLC cell bank of IRCCS Ospedale Policlinico San Martino and cultured in RPMI 10%FBS, 2%Glutamine, and 1%Pen/Strep.

Irradiation Conditions
Cells (25 × 10 4 /mL in 4 mL) were irradiated (4 Gy) using IBL 437C (CIS Bio International). Cells were centrifuged and resuspended in fresh medium and kept at 37 • C for 1 h, then centrifuged and cell pellets lysed.

Sanger Sequencing
Sanger Sequencing (SS) was used to confirm the presence of germline variants in the selected tissues/cell lines as previously described [29][30][31]. The sequencing products were separated by capillary electrophoresis (ABI 3130XL Genetic Analyzer, Applied Biosystems, Waltham, MA, USA) according to the manufacturer's instructions.

RNA Sequencing
RNA from fresh tissue was extracted by the Tissue Lyser plus Maxwell ® RSC sim-plyRNA Tissue Kit (AS1340 Promega, Southampton, UK) in accordance with the manufacturer. Total RNA concentration was quantified by the Qubit ® 2.0 Fluorometer (Invitrogen) using Qubit™ RNA High-Sensitivity (HS) Assay Kits (ThermoFisher, Waltham, MA, USA). The RNA library was prepared using the Illumina Stranded Total RNA Prep with Ribo-Zero Plus (Illumina Inc., San Diego, CA, USA). Briefly, after ribosomal and globin RNA depletion, RNA was fragmented and denatured, and cDNA synthesized. Then, the 3 ends were adenylated, and anchors ligated, followed by a PCR amplification step to add the indexadapter sequences. After clean-up and final amplification of the dual indexed libraries, the quality of the libraries was assessed by the 2100 Bioanalyzer with a High-Sensitivity DNA chip (Agilent), and libraries concentration was assessed with a High-Sensitivity Assay on a Qubit 4 fluorometer (ThermoFisher). Libraries were then pooled at 1.4 nM and pairedend-sequenced (2 × 150) on a NovaSeq 6000 Sequencing System instrument (Illumina Inc., San Diego, CA, USA) with the addition of 1% PhiX. Run metrics were 85.13% PF (clusters Passing Filter) and 91.02 Q30. Sequencing reads were aligned to the reference genome (GENCODE GRCh38 version 33) using STAR v.2.7.3a in two-pass basic mode preventing multimappings.

LOH Status by Microsatellite Analysis
Tumor tissue (FT and/or FFPE and/or melanoma cell line) and matched blood DNA were evaluated in a subset of patients for which good-quality and -quantity somatic DNA was available by using a PCR-based LOH assay with five fluorescence-labeled microsatellite markers (namely D11S1816, D11S1819, D11S2179, D11S1778, and D11S1294). Capillary electrophoresis was performed on the ABI 3100xL DNA analyzer. Raw electrophoretic data were analyzed with GeneMapper software to assess allele ratios. We considered LOH at the ATM locus when the allele ratio fell below 50% in the tumor DNA sample compared to the germline.

Institutional Review Board Statement:
The research protocol was approved by the local Institutional Reviewer Board (CER Liguria; approval #534/2020).

Informed Consent Statement:
Written informed consent was obtained from all subjects involved in the study. Data Availability Statement: Data available upon request.