Rare Germline ATM Variants Influence the Development of Chronic Lymphocytic Leukemia

PURPOSE

Germline missense variants of unknown significance in cancer-related genes are increasingly being identified with the expanding use of next-generation sequencing. The ataxia telangiectasia–mutated (ATM) gene on chromosome 11 has more than 1,000 germline missense variants of unknown significance and is a tumor suppressor. We aimed to determine if rare germline ATM variants are more frequent in chronic lymphocytic leukemia (CLL) compared with other hematologic malignancies and if they influence the clinical characteristics of CLL.

METHODS

We identified 3,128 patients (including 825 patients with CLL) in our hematologic malignancy clinic who had received clinical-grade sequencing of the entire coding region of ATM. We ascertained the comparative frequencies of germline ATM variants in categories of hematologic neoplasms, and, in patients with CLL, we determined whether these variants affected CLL-associated characteristics such as somatic 11q deletion.

RESULTS

Rare germline ATM variants are present in 24% of patients with CLL, significantly greater than that in patients with other lymphoid malignancies (16% prevalence), myeloid disease (15%), or no hematologic neoplasm (14%). Patients with CLL with germline ATM variants are younger at diagnosis and twice as likely to have 11q deletion. The ATM variant p.L2307F is present in 3% of patients with CLL, is associated with a three-fold increase in rates of somatic 11q deletion, and is a hypomorph in cell-based assays.

CONCLUSION

Germline ATM variants cluster within CLL and affect the phenotype of CLL that develops, implying that some of these variants (such as ATM p.L2307F) have functional significance and should not be ignored. Further studies are needed to determine whether these variants affect the response to therapy or account for some of the inherited risk of CLL.

INTRODUCTION

Tumors carry both somatically acquired pathogenic variants (ie, mutations that may activate oncogenes or inactivate tumor suppressors) and germline variants. Although some pathogenic germline variants contribute to oncogenesis,^1,2 the majority of germline variants are missense mutations with an unclear impact on gene function and are therefore termed variants of uncertain significance (VUSs).³ With the increasing use of next-generation sequencing (NGS) in the clinic, the incidental identification of germline VUS has become frequent.⁴ A better understanding of individual VUS as either benign or pathogenic facilitates clinical care as, for example, the identification of pathogenic germline variants often leads to more intense primary preventative efforts in both the proband and relatives and can influence the cancer phenotype and treatment.^5-7 Efforts to reclassify germline VUS (in genes such as BRCA1⁸) as benign or pathogenic are increasingly common.

CONTEXT

• Key Objective

• Inherited germline variation can predispose to cancer and affect the biology of the disease that develops. This study examined, in a consecutive clinical cohort of more than 3,000 patients with hematologic neoplasia, whether rare germline variants in ATM are associated with and influence the biology of the hematologic malignancies that develop.

• Knowledge Generated

• Patients with chronic lymphocytic leukemia (CLL) have a uniquely high rate of rare germline variants in ATM compared with patients with other hematologic malignancies. Furthermore, the CLL that develops in patients with rare germline ATM variants is more likely to have 11q deletion and somatic ATM mutations, consistent with tumor suppressor behavior. Certain ATM variants such as p.L2307F are hypofunctional and may explain these observations.

• Relevance (J.W. Friedberg)

• The observed high frequency of germline ATM variants has implications on both etiology and outcome of CLL. These results require validation in prospective data sets to determine the true clinical significance of these variants and to provide additional rationale for routine next-generation sequencing testing of patients with newly diagnosed CLL.*

  *Relevance section written by JCO Editor-in-Chief Jonathan W. Friedberg, MD.

The B-cell malignancy chronic lymphocytic leukemia (CLL) has a strong hereditary component.^9-11 Linkage studies of affected families have not yielded variants that reproducibly lead to Mendelian inheritance of the disease,^12,13 suggesting that inherited susceptibility may be due to common variants of low relative risk and low penetrance. Genome-wide association studies (GWAS) have identified more than 40 loci potentially linked to CLL development,^14-18 but all have low penetrance and many are not assessed by routine clinical testing, making it difficult to apply to individual patients.¹² Targeted association analyses of polymorphisms in genes important in B-cell biology have repeatedly identified missense variants in the gene ATM (among others) as potential risk loci,^19,20 and a recent whole-exome sequencing study identified ATM as the cancer-related gene with one of the highest number of rare germline variants in a cohort of patients with B-cell malignancies.²¹ In addition, a retrospective unbiased exome-wide comparison of rare germline variants identified rare ATM variants as enriched in CLL cases compared with controls.²² Although the repeated identification of ATM VUS is suggestive, these studies have been limited by low numbers of patients with variants, heterogeneity of sequencing methods, and heterogeneity of control populations. With rare exception,²³ in vitro validation of putative disease-associated variants identified in either GWAS or association analysis studies is infrequently pursued.

ATM encodes a protein kinase and tumor suppressor that facilitates the repair of DNA double-strand breaks. Pathogenic, inactivating somatic mutations in ATM are seen in 4%-15% of CLL cases,^24-28 and deletion of the long arm of chromosome 11 that contains ATM (del(11q)) occurs in 11%-17% of patients with untreated CLL.^29,30 Somatic inactivation of ATM predicts shorter time to first treatment (TTFT),²⁹ shorter progression-free survival after chemoimmunotherapy,^31,32 and shorter overall survival³² in patients with CLL. Somatic ATM inactivation is also seen in mantle cell lymphoma^33,34 and T-cell prolymphocytic leukemia,^35,36 but less commonly in other lymphoid malignancies^37-40 and myeloid disease.⁴¹ These findings suggest that ATM loss is particularly important in CLL disease biology, and therefore, that germline ATM variants could potentially influence clinical outcomes. Indeed, patients with ataxia telangiectasia (ie, homozygous germline pathogenic ATM mutations) have a 5,000-fold higher risk of developing non-Hodgkin lymphoma.⁴² Some missense variants, typically classified in clinical reports as VUS, may actually be significant. ATM p.F858L and p.P1054R were enriched in patients with CLL in a study of single nucleotide polymorphisms in DNA-damage response genes.¹⁹ Our whole-exome sequencing analysis in 516 patients with CLL and ethnically matched normal controls identified ATM p.L2307F as one of the most enriched variants in patients with CLL.²² On the basis of these findings, we systematically studied the prevalence of all ATM variants, as uniformly assessed by clinical-grade NGS, in a large consecutive population of patients with hematologic malignancies, including CLL. We then examined whether rare germline ATM variants affect either the phenotype of CLL that develops or the in vitro behavior of lymphocytes bearing a uniquely enriched rare variant, p.L2307F.

METHODS

Patient Population and Collection of Clinical Characteristics

The patients studied included all patients seen by 21 clinicians within the hematologic malignancies division at the Dana-Farber Cancer Institute (DFCI) over a 5-year span (August 2014 until August 2019) who had NGS performed on either peripheral blood or bone marrow. Patients who had NGS performed multiple times were counted once. Information collected from the chart included birthdate, diagnosis assigned by the treating clinician, date of NGS, and sex. Patients with T-cell prolymphocytic leukemia were excluded given the known role of ATM in this disease. For patients with CLL, additional information collected included the date of diagnosis (defined as when initial biopsy or diagnostic flow cytometry was performed), date of first treatment, date of last follow-up or death, karyotype/FISH, presence of TP53 abnormality (either TP53 mutation or deletion of chromosome 17p), and IGHV status. If the exact date of an event was not available, then the date was rounded to the first of the month or the first of the year provided for the event of interest. Ethical approval was obtained from the DFCI Institutional Review Board.

Targeted Clinical DNA Sequencing

The NGS assay used in this study was developed at Brigham and Women's Hospital and is detailed elsewhere⁴³ and in the Data Supplement (online only). The version of the assay used in this study reported all variants with ≤ 1% total population frequency in the Single Nucleotide Polymorphism Database (dbSNP) or the Exome Aggregation Consortium (ExAC) database.

Confirmatory Sequencing of DNA From Peripheral Blood and Saliva

Saliva and blood samples were obtained from patients enrolled on an Institutional Review Board–approved tissue bank. All patients signed written informed consent. Peripheral blood genomic DNA was isolated using the QIAamp DNA blood mini (Cat. No. 51104, Qiagen, Germantown, MD) or midi (Cat. No. 51183) kits, and genomic DNA from saliva was isolated using an Oragene DNA OG-600 kit (DNA Genotek, Ottawa, ON, Canada) per the manufacturer's protocols. PCR was performed using primers as shown in the Data Supplement.

Generation and Analysis of ATM^L2307F Knock-in Cell Lines

NALM-6 cells (DSMZ, Braunschweig, Germany) with ATM p.L2307F knocked into the endogenous allele, along with ATM wild-type and knockout genotypes as controls, were all generated via nucleofection as described in the Data Supplement. Immunoblots were performed on lysates from cells after 6 hours of 2-μM etoposide treatment using the antibodies as shown in in the Data Supplement. Annexin/PI staining was performed according to the manufacturer's instructions (APC Annexin V Apoptosis Detection Kit with PI, BioLegend, 640932) after 24 hours of 2-μM etoposide treatment.

Statistics

Associations were investigated among the diagnostic category and age (analysis of variance), sex and percent of germline VUS (pairwise chi-square test), and the five most common germline variants (Fisher's exact test). Within the CLL group, associations were investigated between the presence or absence of germline VUS and the continuous variable age at diagnosis (Wilcoxon rank-sum test) and the binary variables sex, IGHV status, 11q-deleted disease, ATM-aberrant disease, and TP53-aberrant disease (chi-squared test); analyses were similarly performed to compare patients with the ATM p.L2307F variant with patients without any ATM variant. Kaplan-Meier plots and log-rank tests were used to assess the significance of del(11q) and ATM variants on TTFT.

RESULTS

Identification of Four Clinically Defined Cohorts

We selected 3,128 patients seen during a 5-year period within the Hematologic Malignancies clinic at DFCI who also had clinical-grade NGS that sequenced the entire coding region of ATM. Patients were grouped into one of four categories on the basis of their diagnosis: CLL, non-CLL lymphoid, myeloid, and non-neoplastic disorders. We chose these diagnostic categories on the basis of distinct biology and the frequency of somatic ATM mutations in each disease category in cBioportal,⁴⁴ which we interpreted as an indication of the importance of ATM (and potentially ATM germline variants) in the biology of that disease category. In cBioPortal, somatic ATM mutations were present in 0.5% of patients with myeloid diseases,^45-51 4.8% of patients with non-CLL lymphoid disorders,^34,38,52-60 and 9.6% of patients with CLL.^25-28 In our patient population, 825 patients (26%) had CLL, monoclonal B lymphocytosis, or B-cell prolymphocytic leukemia; 1,101 patients (35%) had other lymphoid disorders; 1,059 (34%) had myeloid disorders; and 143 (5%) had no neoplastic process (eg, allogeneic stem-cell transplant donors and patients with autoimmune cytopenias; Fig 1A). Breakdown of individual diseases within each category is given in detail in the Data Supplement. The non-neoplastic cohort had significantly more female patients compared with the other cohorts (Fig 1B), reflecting the male predominance of most hematologic malignancies.⁶¹ The non-neoplastic cohort was also significantly younger (Fig 1C).

FIG 1.

Characteristics of patient populations enrolled in this study. (A) Pie chart demonstrating the four major cohorts of patients in this study along with the frequencies of the various diseases within each cohort. (B) Gender distribution among the four cohorts. (C) Age distribution among the four cohorts. B-PLL, B-cell prolymphocytic leukemia; CLL, chronic lymphocytic leukemia; DLBCL, diffuse large B-cell lymphoma; LPL, lymphoplasmacytic lymphoma; MBL, monoclonal B lymphocytosis; MDS, myelodysplastic syndrome; MDS/MPN, myelodysplastic syndrome/myeloproliferative neoplasm overlap syndromes; MPN, myeloproliferative neoplasm; WM, Waldenstrom macroglobulinemia.

Identification of Germline and Somatic ATM Variants in the Clinical Cohorts

Our analysis was restricted to rare ATM variants present at a total population allelic frequency of ≤ 1% in the Single Nucleotide Polymorphism Database⁶² and ExAC⁶³ database. Two hundred sixty-nine different ATM variants were identified. Since NGS was performed on hematologic tissue containing both somatic and germline variants, we established criteria to define a variant as germline or somatic (Fig 2A). Fifty variants were prioritized for germline DNA sequencing because (1) the variants were present in only one patient with CLL each; (2) the variant allele fraction (VAF, the fraction of sequencing reads carrying a particular variant, expressed as a percentage) was > 20%, making it difficult to determine germline or somatic status; and (3) banked salivary DNA was available for germline DNA sequencing (Fig 2A). For all other variants, the hierarchical flowchart was followed (Fig 2A). Of 158 germline variants, 138 (87%) were deemed germline by a method other than VAF (eg, direct sequencing, recurrent in multiple patients, or presence in gnomAD; Fig 2B). Most germline variants (73%) were predicted to be VUS, 16% were predicted to be benign/likely benign, and 10% were predicted to be pathogenic/likely pathogenic by ACMG classification (Fig 2C).³ Most somatic variants (62%) were predicted to be pathogenic/likely pathogenic. Individual details on each variant are given in the Data Supplement.

FIG 2.

Characterization of ATM variants identified in this study. (A) Diagram demonstrating the hierarchical classification system used to assign each variant as either somatic or germline. (B) Pie chart demonstrating the frequencies by which each indicated method was used to call a variant germline or somatic. Sequencing refers to paired sequencing of DNA isolated from banked peripheral blood (ie, neoplastic) and saliva (ie, germline). Nonsense/splice site/indel refers to nonsense variants, splice site mutations, and insertions and deletions. Multiple different patients indicate that the variant was present in multiple patients within the cohort. gnomAD indicates that the variant was present in the gnomAD database, and VAF refers to the fact that variant allele fraction was used to determine if a variant was germline or somatic. (C) Predicted pathogenicity of each variant according to American College of Medical Genetics and Genomics (ACMG) classification rules, broken down by germline or somatic status. (D) Bubble plot demonstrating median VAFs for germline and somatic variants. Each bubble represents a distinct variant with bubble size proportional to the number of patients carrying the variant. (E) Frequencies of germline ATM variants in patients in each clinical cohort. VAF, variant allele fraction; VUS, variant of unknown significance.

We assessed the accuracy of our classification via multiple methods. VAF is a crude method for assessing germline status, as heterozygous germline variants carry a VAF close to 50% unless there are copy number changes or copy-neutral loss of heterozygosity affecting ATM (eg, del(11q)). Germline variants did cluster around a median VAF of 50%, whereas somatic variants had a broader range of VAFs not centered around any number (Fig 2D). Next, we tested our hierarchical rules by removing individual rules (eg, germline status determined by the presence of the variant in gnomAD or the presence of the variant in multiple patients) from the algorithm and reclassifying the variants. We found that at most 9 of 269 variants (and 11 of 613 total patients with variants) had germline or somatic status changed, indicating robust agreement between the rules (Data Supplement). Finally, we identified 75 additional patients in our consecutive clinical cohort with germline tissue (most often salivary DNA) available for ATM sequencing. The concordance rate between the gold standard of germline DNA sequencing and our algorithm was 96% (72 of 75 matched; Data Supplement).

Comparison of the Frequencies of Germline ATM Variants in the Clinical Cohorts

Our cBioPortal analysis demonstrated a greater rate of somatic ATM mutations in CLL as compared with other hematologic malignancies, highlighting the importance of ATM in CLL biology and leading to our hypothesis that germline variants of ATM might also have a functional impact and be more abundant in patients with CLL than in patients with other diagnoses. Patients with CLL did possess significantly more germline ATM variants (24% of all patients) than patients with non-CLL lymphoid disorders (16%), myeloid disorders (15%), or no neoplastic process (14%; Fig 2D); none of the differences between the latter three groups were significant. In line with cBioPortal data, somatic ATM variants were also most frequent in CLL (Data Supplement). We subdivided the non-CLL lymphoid population and found that the rate of germline ATM variants in CLL was greater than the rate of variants in any single lymphoma diagnosis (eg, mantle cell lymphoma; Data Supplement). We asked whether the difference in ATM variant frequencies could be attributed to skewed disease representation in our sequenced population, but after correcting for disease representation by adjusting disease prevalence in our cohort to match the prevalence of each lymphoid/myeloid disease in SEER, we still found significantly more germline ATM variants in the CLL cohort (24% of all patients) compared with patients with non-CLL lymphoid disorders (16%) or myeloid disorders (16%; Data Supplement). Given the unique enrichment of rare germline ATM variants in patients with CLL and the known role of ATM in CLL biology, we focused our next studies specifically on the CLL cohort.

Clinical Characterization of the CLL Cohort

The median age at CLL diagnosis was 60 (range, 23-89) years, the median length of follow-up was 44 (range, 0.5-522) months, and 51% of the patients had not received treatment. The immunoglobulin heavy-chain variable region (IGHV) status was unmutated in 43%, mutated in 39%, and unknown in the remained. Eighteen percent had TP53 aberrant disease (defined as either a pathogenic TP53 mutation or del(17p)), and 16% had del(11q); Table 1). Regarding ATM mutations, 21.5% of patients (n = 177) had only germline variants, 4% (n = 33) had only somatic variants, 2.6% (n = 21) had both germline and somatic variants, and 72% (n = 594) had no ATM variants. Immortal time bias did not confound these results as rates were the same whether patients were diagnosed with CLL before 2014 or diagnosed between 2014 and 2019 (Data Supplement).

TABLE 1.

Clinical Characteristics of the Chronic Lymphocytic Leukemia Patient Cohort

Germline ATM Variants Affect the Clinical Characteristics of CLL

We hypothesized that if germline ATM variants have functional significance, then the clinical characteristics of CLL in patients with germline variants will be different from those without. Certain clinical characteristics were similar between patients with CLL with (n = 198) and without (n = 628) germline ATM variants: anatomic sex, IGHV status (Fig 3A), and TP53 aberrancy (Data Supplement). However, patients with germline ATM variants were diagnosed at a statistically significantly younger age compared with those without germline ATM variants (Fig 3B), 58 years versus 61 years, respectively. The rate of 11q-deleted CLL in patients with germline variants was 24.5% compared with a rate of 13.3% in patients without germline variants (P < .0005, Fig 3C). Somatic ATM variants were also present in 11.1% of patients with germline variants compared with 5.6% of patients without germline variants (P = .065, Data Supplement). When considering ATM-aberrant disease (defined as either del(11q) or a somatic mutation in ATM), patients with germline ATM variants developed ATM-aberrant CLL 29.8% of the time compared with 15.0% in patients without germline ATM variants (P < 1 × 10^–5, Fig 3D). This association was found regardless of the date when CLL was diagnosed, arguing against any confounding effect of immortal time bias (Data Supplement).

FIG 3.

Comparison of clinical characteristics between patients with CLL with and without germline ATM variants. (A) Frequency of IGHV-mutated and IGHV-unmutated CLL as a function of the presence or absence of germline ATM variants. (B) Relative age of patients with and without germline variants. (C) Frequency of 11q-deleted CLL in patients with germline ATM variants compared with those without. (D) Frequency of ATM-aberrant disease (defined as either an 11q deletion or a somatic ATM mutation) in patients with germline ATM variants compared with those without. (E) TTFT for four groups of patients subdivided by the presence or absence of an 11q deletion and the presence or absence of a somatic ATM mutation, restricted only to patients diagnosed with CLL between 2014 and 2019. (F) TTFT with the group that possesses an 11q deletion without a somatic ATM mutation now split into two groups of patients, those who possess a germline ATM variant and those who do not. CLL, chronic lymphocytic leukemia; IGHV, immunoglobulin heavy-chain variable region; TTFT, time to first treatment.

Next, we examined TTFT in four groups on the basis of somatic events affecting ATM: patients with del(11q) and a somatic ATM mutation, patients with del(11q) only but no somatic mutation, patients with a somatic ATM mutation only, or patients with no ATM aberrancy (Fig 3E). We restricted our analysis only to patients diagnosed between 2014 and 2019 to avoid any confounding effects from immortal time bias. Patients with loss of both alleles of ATM had a shorter TTFT compared with patients with both ATM alleles intact (median TTFT 2.3 years v 4.5 years, hazard ratio [HR], 2.3; 95% CI, 1.29 to 4.08; P = .005), whereas patients with del(11q) and no somatic mutation had an intermediate phenotype between the two (median TTFT 3.1 years, HR, 1.73; 95% CI, 1.14 to 2.63; P = .01 compared with the group with no ATM aberrancy), in line with previous work.⁶⁴ We asked if the intermediate phenotype of this patient population could be due to the fact that it combined two types of patients, one group that had del(11q) with a germline variant on the remaining allele (more closely resembling biallelic loss), and one group that had del(11q) without a germline variant (ie, with the remaining allele intact). Division of this intermediate population into these two subgroups showed that the TTFT curve of the del(11q) plus germline variant group (median TTFT 1.9 years) overlapped with that of the del(11q) plus somatic variant group (median TTFT 2.3 years), whereas the TTFT curve of the del(11q) without a germline variant group shifted rightward (median TTFT 4.3 years, HR, 0.48; 95% CI, 0.22 to 1.03; P = .058 compared with the del(11q) group with germline variants; Fig 3F). These findings were not affected by the small fraction of patients who required dates of diagnosis and treatment to be imputed because of lack of exact date documentation in the chart (Data Supplement).

Interrogation of the Relative Frequencies of Individual Germline ATM Variants in the Clinical Cohorts

The variant p.L2307F was the most common ATM variant in the CLL population, present in 2.8% of patients with CLL (Fig 4A and Data Supplement). This was numerically greater than the percentage of patients with ATM p.L2307F in the non-CLL lymphoid (1.5%, P = .1), myeloid (0.67%, P < .05), and non-neoplastic (0%, P = .1) cohorts. The other four most common ATM variants (p.D1853V, p.S707P, p.F858L, and p.S49C) were also often present in numerically greater frequencies in patients with CLL, but in no other instance, did this difference reach statistical significance (Data Supplement).

FIG 4.

Investigating the clinical and in vitro phenotypes of the ATM^L2307F variant. (A) Lollipop diagrams of individual ATM variants detected in each cohort. The height of the bars indicates the frequency of that variant in each cohort, and the most frequent variant is labeled. The amino acids of the ATM protein from the N to C terminus serve as the x-axis. Represented ATM protein domains include TAN (Tel1/ATM N-terminal) motif, FAT (FRAP-ATM-TRRAP) domain, and PI3-PI4 (phosphatidylinositol 3-/4-) kinase catalytic domain. (B) Frequency of 11q-deleted CLL in patients with a germline ATM^L2307F variant as compared with those without any germline ATM variant. (C) Frequency of ATM-aberrant disease (defined as either an 11q deletion or a somatic ATM mutation) in patients with a germline ATM^L2307F variant as compared with those without any germline ATM variant. (D) Sanger sequencing of ATM cDNA reverse transcribed from ATM mRNA derived from three different clonal populations of NALM6 cells. The heterozygous clones demonstrate the presence of a C to T knock-in mutation that creates the L2307F variant (red arrow). The blue arrows indicate silent synonymous mutations that were simultaneously introduced with L2307F to prevent sgRNA recognition and re-editing of a successfully recombined allele. The clone with a null second ATM allele also demonstrates a frameshift on Sanger sequencing. (E) Six clones of the indicated genotypes were exposed to 2-μM etoposide for 6 hours, and activation of the downstream ATM targets KAP1 and p53 was assessed by immunoblot. NALM6 cells have wild-type p53, which is not constitutively present at baseline, so both total p53 protein and its phosphorylation on serine 15 are induced by etoposide treatment. (F) Percentage of cells with the indicated apoptotic phenotype as determined by annexin/PI staining performed 24 hours after treatment with 2-μM etoposide. *P < .05 when compared with the heterozygote ATM^wt/– clone. CLL, chronic lymphocytic leukemia.

Investigation of the Clinical and Functional Consequences of the L2307F Variant

Comparing the 23 patients with CLL with a germline ATM p.L2307F variant with those without any germline variants, no significant difference was seen in age at the time of diagnosis or IGHV mutational status (Data Supplement). Patients with ATM p.L2307F had a significantly greater incidence of 11q-deleted disease (35%) compared with patients without any germline ATM variants (13.3%, P = .009; Fig 4B); similar findings were seen with ATM-aberrant disease (35% v 15%, respectively, P = .018; Fig 4C). Within the top five most common ATM variants, patients with p.S707P and p.F858L (but not p.D1853V or p.S49C) were also significantly enriched for ATM-aberrant disease (Data Supplement).

To assess the in vitro function of the p.L2307F variant, we used CRISPR/Cas9 and a homology-directed repair template to introduce p.L2307F into one endogenous ATM locus of the NALM6 cell line, a diploid lymphoid cell line with both copies of ATM intact. We kept clones that underwent both successful and unsuccessful recombination to develop an isogenic panel of ATM^+/+, ATM^+/–, ATM^–/–, ATM^L2307F/+, and ATM^L2307F/– cell lines (including two independent clones identified to have the ATM^L2307F/– genotype). We verified p.L2307F knock-in with Sanger sequencing of ATM cDNA (Fig 4D). We treated these cell lines with etoposide, a DNA-damaging agent that induces ATM-dependent activation of downstream targets such as KAP1 and p53. Clones with p.L2307F had reduced KAP1 and p53 activation compared with wild-type and heterozygous clones that did not possess p.L2307F (Fig 4E). These findings were confirmed when the cells were exposed to ionizing radiation, which also induces DNA damage (Data Supplement). Compared with cells with a wild-type allele, a greater percentage of cells with ATM p.L2307F were in apoptosis (as assessed by annexin/PI staining) after 24 hours of exposure to etoposide (Fig 4F), demonstrating that p.L2307F is a hypomorph.

DISCUSSION

We find that rare germline ATM variants, when considered in aggregate, are more common in patients with CLL compared with other lymphoid and myeloid disorders. These ATM variants have prognostic and functional significance. Patients with germline ATM variants are diagnosed at a younger age and are twice as likely to have CLL with 11q deletion, an adverse marker associated with more rapid disease progression. The variant ATM p.L2307F in particular is enriched in patients with CLL and is also significantly associated with the development of 11q-deleted disease. Lymphocytes with an endogenous ATM^L2307F mutation have an impaired DNA damage response compared with lymphocytes with wild-type ATM.

An awareness of the potential clinical significance of rare germline variants is increasing in other hematologic diseases^65,66 and CLL. With respect to ATM, heterozygous germline pathogenic nonsense mutations have been reported in patients with CLL and are associated with 11q-deleted disease.^24,67,68 Such variants are present in only 1%-3% of the general population,⁶⁹ however, and are unlikely to play a role in CLL pathogenesis in large numbers of patients. The contribution of more prevalent missense ATM variants was less clear, as most have not been characterized and are classified as VUS. A previous candidate gene association study identified ATM p.F858L and p.P1054R as more frequent in patients with CLL compared with controls.¹⁹ A previous study of nonsynonymous ATM variants in patients with CLL treated with chemoimmunotherapy found a frequency of ATM variants closely matching what we report here (23%), but the frequency of 11q-deleted disease in patients with versus without ATM VUS (29% v 13%, respectively) was not significantly different.⁷⁰ However, only 17 patients in that study had ATM missense variants (compared with 198 patients with CLL here), and thus, statistical power might have been limited. Our study suggests that the definition of what constitutes a deleterious germline ATM variant must be expanded to include certain rare germline ATM variants, which (in aggregate) are present in 24% of patients with CLL.

Multiple lines of evidence support the explanation that germline ATM variants are more frequent in patients with CLL because some variants lead to a hypofunctional protein, which over decades manifests as a predisposition to the development of CLL. CLL that occurs in a patient with a germline variant is more likely to have deletion of the other ATM allele (ie, del(11q)) than if the patient does not have a germline ATM variant. Previous work has demonstrated an association between pathogenic somatic ATM variants and del(11q),²⁶ reflective of the two-hit hypothesis in which the neoplastic cell acquires a growth advantage when both alleles of a tumor suppressor are lost.⁷¹ Our findings extend this observation to germline ATM variants. The presence of one hit already present in the germline is also consistent with our finding that patients with germline ATM variants present at a slightly younger age. Although not every VUS will affect ATM function, the implication that some VUSs are hypofunctional, in conjunction with the finding that a remarkably large fraction of patients with CLL possess an ATM germline variant, suggests that defects in ATM function may contribute to the development of a significant portion of CLL cases.

ATM p.L2307F is likely one such hypofunctional ATM variant. Occurring at an allelic frequency of 0.129% in gnomAD v2.1.1, it is the most common ATM variant in our CLL population. Previous work suggested that ATM p.L2307F may be a variant of particular interest. In silico tools predicted that the variant was not tolerated (SIFT⁷²) or probably damaging (PolyPhen-2⁷³). An exome-wide sequencing study previously identified p.L2307F as enriched in CLL.²² ATM p.L2307F was strongly associated with the development of lung adenocarcinoma and loss of the remaining ATM allele in a GWAS.⁷⁴ Here, 11q-deleted disease was three times more common in patients with CLL with p.L2307F, and lymphoid cells with p.L2307F had a defective response to DNA-damaging agents, supporting the conclusion that ATM p.L2307F is a hypomorph.

The strengths of our study include the large population size, the uniform use of a clinical-grade NGS assay to assess ATM variants across a consecutive clinical cohort, a rigorous approach to classifying variants as germline or somatic (with 96% concordance between predicted germline/somatic status and germline DNA sequencing), the availability of clinical patient-level data matched to each variant, and our functional validation of p.L2307F. The limitations of our study include the relatively short follow-up time, which precludes analysis of the effects of germline variants on metrics such as overall survival. Given the indolent nature of CLL and effectiveness of novel agents, longer follow-up will be needed to determine if ATM germline variants affect progression-free or overall survival. We were unable to sequence all variants to confirm germline status, and some variants might have been misclassified. The population reflects our North American, predominantly White, patients. Germline variants not prevalent in this demographic are not addressed in this study. Given that we examined patients who were seen in clinic and had NGS performed in a certain timeframe, rather than restricting only to patients with newly diagnosed hematologic malignancies, it remains possible that immortal time bias affected our conclusions about the rates of ATM variants in the four cohorts. However, such bias would not be expected to affect the non-neoplastic cohort, and the rates of somatic ATM variants in our cohorts closely match the previously reported rates of somatic ATM variants in other studies (eg, cBioPortal). Furthermore, within our CLL cohort, restricting our analyses only to patients diagnosed between 2014 and 2019 gave the same results. Finally, we focused on CLL because CLL had the highest rate of germline ATM variants and was the most common disease in our consecutive clinical cohort, and ATM loss is an important prognostic marker in CLL. Future work can investigate the role of rare germline variants in other lymphoid diseases with high rates of somatic ATM mutations, such as mantle cell lymphoma.

These findings have immediate clinical implications. Clinicians should be aware that patients with CLL diagnosed at younger age may possess rare germline ATM variants and should consider NGS testing; patients with an ATM VUS on NGS are at greater risk of the poor prognostic marker del(11q). Furthermore, our findings argue for a liberal threshold for reporting VUS to clinicians (including rare germline variants with population allelic frequency up to 1%) because some variants may be later reclassified as pathogenic.

In summary, we have identified a high frequency of germline ATM variants uniquely present in patients with CLL. These variants affect the clinical characteristics of the CLL that develops. These findings suggest that germline ATM variants may play a larger role in the development of CLL than previously appreciated and contribute to a growing awareness of the significance of VUS in the development of hematologic malignancies and cancer in general.⁷⁵ This work encourages future efforts to determine the clinical significance of the long list of germline variants for which clinical significance is currently unknown.

AUTHOR CONTRIBUTIONS

Conception and design: Benjamin L. Lampson, Annette S. Kim, Jennifer R. Brown

Financial support: Jennifer R. Brown

Administrative support: Annette S. Kim, Jennifer R. Brown

Provision of study materials or patients: Annette S. Kim, Jennifer R. Brown

Collection and assembly of data: Benjamin L. Lampson, Aditi Gupta, Kiyomi Mashima, Natalia Wojciechowska, Conner J. Shaughnessy, Peter O. Baker, Stacey M. Fernandes, Samantha Shupe, John-Hanson Machado, Rayan Fardoun, Jennifer R. Brown

Data analysis and interpretation: Benjamin L. Lampson, Aditi Gupta, Svitlana Tyekucheva, Kiyomi Mashima, Zixu Wang, Conner J. Shaughnessy, Rayan Fardoun, Jennifer R. Brown

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Rare Germline ATM Variants Influence the Development of Chronic Lymphocytic Leukemia

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/jco/authors/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).