Introduction

The protein kinase ataxia telangiectasia mutated (ATM) functions as a tumor suppressor, is activated in response to double-stranded DNA breaks, and plays a crucial role in DNA damage response [1]. In chronic lymphocytic leukemia (CLL), somatic mutations in ATM have previously been reported in 7–16% of cases [2,3,4,5,6,7], while the ATM gene, located in the q-arm of chromosome 11, is recurrently deleted in 10–24% of patients [3, 5, 6, 8,9,10]. Although del(11q) has long been recognized as a marker of poor prognosis in CLL and patients carrying this chromosomal abnormality frequently present with lymphadenopathy, face a higher risk of progression, and generally have an inferior prognosis [8, 9, 11,12,13,14,15,16], there are currently conflicting data on whether mutations in ATM with or without del(11q) are indicative of a poor clinical outcome.

In separate studies from the UK Leukemia Research Fund CLL 4 trial, patients with ATM mutations treated with chemotherapy exhibited shorter progression-free survival (PFS) and overall survival (OS) compared to ATM-wildtype patients, though the differences did not reach statistical significance [7, 17]. Of note, reduced survival was observed only in cases with biallelic ATM inactivation, defined as an ATM mutation combined with del(11q) [7, 17]. In contrast, Austen et al. demonstrated inferior OS and treatment-free survival in patients with ATM mutations [18], while Nadeu et al. identified ATM mutations, irrespective of del(11q), as a biomarker of shorter time-to-first-treatment (TTFT) but not OS in patients with CLL [4, 19]. More recently, Nguyen-Khac et al. observed that, although ATM mutations were associated with shorter PFS in univariable analysis, this association was lost in multivariable models when accounting for other recurrent CLL mutations [20]. Conversely, Hu et al. reported that ATM mutations remained a significant biomarker for shorter TTFT in both univariable and multivariable analyses in treatment-naive patients with CLL [21].

In a recent study, we investigated the prognostic impact of nine recurrently mutated genes, ATM not included, in a large well-annotated series of pre-treatment samples from 4580 patients with CLL [22]. We evaluated the relative clinical importance of mutations within each gene in the poor-prognostic immunoglobulin heavy variable (IGHV)-unmutated CLL (U-CLL) and the more favorable-prognostic IGHV-mutated CLL (M-CLL) subgroups separately, focusing on early-stage patients and TTFT. Notably, we reported that these recurrently mutated genes carry different prognostic impact depending on IGHV gene somatic hypermutation (SHM) status, indicating the need for a more compartmentalized strategy to enable tailored management and care for patients belonging to the M-CLL and U-CLL subgroups.

In the present investigation, which includes 3631 patients with CLL, we aimed to investigate the clinical impact of somatic mutations and/or deletions of ATM, with particular consideration of IGHV gene SHM status and other gene mutations associated with CLL prognosis, focusing primarily on early-stage patients. To our knowledge, this represents the largest study to date that investigates the prognostic implications of ATM dysfunction in this patient population, aiming to provide a comprehensive evaluation of its role in disease progression in relation to other genetic features.

Materials and methods

Patient cohort and clinicobiological characteristics

The present study included pre-treatment samples from 3631 patients with CLL collected from 22 European centers (Supplementary Table S1). This is a subset from the previously described cohort of 4580 patients for whom ATM mutational data was available [22]. The sample collection period spanned from 1996 to 2020 and the median time from diagnosis to sample collection was 3 months (IQR 0-34, data available for 3572 cases, 98.7%). The clinicobiological features of the analyzed cohort are outlined in Table 1 and are highly similar to those of the larger cohort, with no statistically significant differences [22]. In summary, the patients had a median age of 64.7 years at diagnosis, with a male-to-female ratio of approximately 2:1 (63% male). Moreover, 2604 patients (72%) were in Binet stage A, 1502 (44%) were categorized as U-CLL, and 2122 patients (58%) required treatment during follow-up of which the majority were treated with either chemotherapy (41%) or chemoimmunotherapy (38%) at first line, while only 1% received targeted therapy. All cases were diagnosed according to the iwCLL guidelines [23]. Informed consent was obtained according to the Helsinki declaration and the study was approved by the local Ethics Review Committees.

Table 1 Clinicobiological characteristics of the studied cohort (n = 3631).
Full size table

Mutational analysis and variant classification

All cases had previously been assessed for coding sequence mutations in BIRC3, EGR2, MYD88, NFKBIE, NOTCH1, POT1, SF3B1, TP53, and XPO1 [22]. Mutation screening for ATM was conducted using next-generation sequencing (NGS) in >99% of cases, primarily through the utilization of targeted gene panel sequencing, while only 22 cases were analyzed by Sanger sequencing or resequencing arrays (Supplementary Table S1). For cases investigated using NGS-based methods, sequence alignment, variant calling and annotation were performed at each participating center using a variant allele frequency (VAF) threshold of ≥5% to classify mutated cases. Additionally, polymorphic variants with a max population allelic frequency of ≥1% in the gnomAD database (v4.0) [24] were excluded, and the remaining variants were further filtered to only retain exonic non-synonymous variants and small insertions or deletions. All ATM variants are reported based on the reference transcript NM_000051.3/ENST00000278616.8 (GRCh37).

ATM data were available from paired tumor/germline samples for 427 cases (11.8%), while tumor-only data were obtained for the remaining 3204 cases. Since the majority of sequencing results were derived from tumor-only analyses, capturing both somatic and germline variants, a hierarchical ranking system was implemented to classify all variants into categories of putative somatic or putative rare germline/predicted neutral (Fig. 1A). The classification of these variants was based on their max population allelic frequency ( > 0.001) in the gnomAD database (v4.0) [24], annotation in ClinVar [25], OnkoKB [26, 27], variant type (nonsense/frameshift or missense), pathogenicity predictions by AlphaMissense [28] and CADD (v1.7) [29], as well as data from CLL datasets with available germline information [2, 30, 31], and is detailed in Supplementary Table S2. For patients with multiple reported ATM variants, the variant with the highest hierarchical rank was assigned to each patient for subsequent analyses. If multiple variants fell within the same tier of the hierarchy, the variant with the highest VAF was prioritized.

Fig. 1: Classification of reported ATM variants.
figure 1

A Hierarchical flowchart used to assign ATM variants to putative ‘germline/neutral’ or ’somatic/mutated’ categories. B Distribution of variant allele frequencies (VAFs) for ATM variants in each flowchart category, bar chart shows the proportion del(11q) positive cases for each category. C Distribution of VAFs for all variants assigned as ‘germline/neutral’ and ‘somatic/mutated’. D Graphical representation of amino acid changes and frequencies for all ATM variants in the cohort. Putative somatic variants are displayed above and putative germline/predicted neutral variants are shown below. *Confirmed germline by sequencing in published datasets [30, 31] and in unpublished data. **Four ATM variants, each observed in multiple patients and variably classified as either germline/neutral or somatic/mutated depending on the final step of the classification system, were reviewed and reassigned to a single classification category (Supplementary Table S2).

Full size image

The IGHV gene SHM status was available for 3402/3631 (93.7%) patients and was assessed using PCR amplification and sequence analysis of IGHV-IGHD-IGHJ gene rearrangements, utilizing a 98% identity threshold to germline sequences to define M-CLL (<98% identity) and U-CLL (≥98% identity) patients as previously described [32, 33]. Chromosomal aberrations were identified primarily through fluorescence in situ hybridization (FISH) and targeted probes for chromosomes 13q, 11q, 17p and 12 [11].

Statistical analysis

Correlations between ATM mutational status and clinicobiological variables were assessed using the chi-squared test. Two-sided Fisher’s exact tests were performed to assess co-occurrence of genomic alterations and p values were adjusted using the Benjamini–Hochberg method for multiple testing. The primary endpoint for survival analysis was TTFT, calculated from the date of diagnosis until date of initial treatment or date of administrative censoring or death if untreated, with a median follow-up of 5.1 years (95% CI 4.7–5.5) and was available for 3598/3631 (99.1%) patients. Additionally, OS was assessed, calculated from the date of diagnosis to either the date of death or last follow-up. The cohort had complete OS data for all patients, with a median follow-up time of 13 years (95% CI, 12.2–13.6 years). Kaplan-Meier survival curves were calculated to evaluate the effects of ATM mutations and/or del(11q) on TTFT and OS. Pairwise comparisons were performed to determine differences between subgroups by employing the Cox–Mantel log-rank test and p values were adjusted using the Benjamini–Hochberg method to account for multiple comparisons. Multivariable analyses using Cox proportional hazards models were employed to assess the prognostic strength of individual biomarkers. All statistical analyses were performed in R (v4.4.1) [34] and R studio (version 2024.04.0 + 735) [35]. Plots were created using ggplot2 (version 3.5.1), survminer (version 0.4.9), ComplexHeatmap (version 2.20.0) and Maftools (version 2.20.0) [36].

Results

Hierarchical ranking of somatic and germline ATM variants

A total of 445 ATM variants were identified across 362 patients in the cohort studied following mutational analysis and initial variant filtering. Among these, 309 variants (69%) were classified as putative somatic using our proposed hierarchical ranking system for ATM variant classification. To validate this approach, we reclassified all confirmed somatic mutations in samples with available germline information. Of the 27 confirmed somatic ATM mutations, 25 were consistently classified as somatic, while 2 were reclassified as predicted neutral based on multiple pathogenicity scores suggesting a neutral impact (Supplementary Table S2). Furthermore, among variants reported as rare germline based on prior studies, only 3/55 were predicted to be pathogenic or likely pathogenic according to the ClinVar database, all three cases harbored del(11q) on the alternate allele.

As the next step, we plotted the VAFs for all variants assigned to each category in the flowchart (Fig. 1B). Variants classified as putative germline/predicted neutral typically had VAFs centered around 50%, whereas putative somatic mutations displayed a broader distribution with the majority falling below 50% (Fig. 1B and Supplementary Table S2). These distinct VAF patterns became even more pronounced when pooling all samples with putative germline/predicted neutral or putative somatic mutations (Fig. 1C). Notably, variants with VAFs >50% were most often observed in tumors also harboring del(11q) (Fig. 1B, C). We identified 17 ATM variants classified as putative somatic with VAFs >60% in patients without del(11q) (Supplementary Table S2). For three of these cases heterozygosity status was available, revealing copy number-neutral loss of heterozygosity (cnLOH) in all. For all subsequent analyses, all predicted germline/neutral ATM variants were re-categorized as ATM wildtype.

Frequency of ATM mutations and association with clinicobiological features

Based on our classification, ATM mutations were identified in 246/3631 (6.8%) patients, with 57 (23.2%) of these individuals carrying more than one ATM mutation. The mutations were distributed across the entire coding sequence of the ATM gene, with the p.R3008C (n = 7, collectively p.R3008C/H/L/S n = 13) and the p.K468fs (n = 10) as the most frequently affected positions (Fig. 1D, Supplementary Table S3). Additionally, somatic missense mutations were primarily clustered toward the C-terminal region of the ATM protein, defined as the start of the FAT domain, while in contrast, somatic nonsense mutations were predominantly located in the N-terminal region (Fig. 1D and Supplementary Fig. S1). Putative germline/neutral variants were evenly distributed throughout the entire length of the ATM protein (Fig. 1D).

By combining our findings on ATM with sequencing data for nine other commonly mutated genes in CLL [22], ATM emerged as the fourth most frequently mutated gene, following NOTCH1, SF3B1, and TP53 (Fig. 2A). Mutations in ATM frequently co-occurred with other gene mutations, observed in 134 of 246 cases (54%). Among these, mutations in SF3B1 were the most common co-existing event, identified in 56 cases (23%, Fig. 2B). On the contrary, mutations in TP53 and MYD88 were rare occurrences in cases with ATM mutations, displaying mutual exclusivity (Fig. 2B, C). As expected, ATM mutations frequently co-occurred with del(11q) (112/246, 45.5%, Fig. 2B, C), while among cases with del(11q) (430/3631 cases, 11.8%), 112/430 (26.0%) carried ATM mutations. In addition, ATM mutations were significantly more frequent in U-CLL compared to M-CLL (p < 0.001), in patients with advanced stage at diagnosis (p < 0.001) and in those eventually in need of treatment (p < 0.001, Table 2). Notably, in U-CLL, ATM mutations were mutually exclusive with trisomy 12 and TP53 mutations (Fig. 3A, B), while SF3B1 and NFKBIE mutations were enriched in M-CLL (Fig. 3C, D).

Fig. 2: Overview of recurrent genetic aberrations in the entire cohort and ATM-mutated cases.
figure 2

A Overview of all 3631 CLL cases included in the study, sorted by frequency of mutations in 10 recurrently mutated genes. B Oncoplot showing detected mutations in recurrently mutated genes, IGHV somatic hypermutation status and chromosomal aberrations in 246 ATM-mutated cases. C Co-occurrence of recurrent gene mutations and chromosomal aberrations in the entire cohort. Odds ratios (OR) and BH-adjusted p values derived from two-sided Fisher exact tests. OR 0–1 indicates a trend towards mutual exclusivity, OR > 1 indicates a trend towards co-occurrence. U-CLL, CLL with unmutated IGHV genes, M-CLL, CLL with mutated IGHV genes.

Full size image
Table 2 Relation of ATM mutations to other clinicobiological markers.
Full size table
Fig. 3: Overview of detected recurrent genetic abnormalities in U-CLL and M-CLL.
figure 3

A Oncoplot and (B) co-occurrence plot displaying recurrently mutated genes and chromosomal aberrations in 1502 U-CLL cases. C Oncoplot and (D) co-occurrence plot with recurrently mutated genes and chromosomal aberrations in 1900 M-CLL cases. Odds ratios (OR) and BH-adjusted p values derived from two-sided Fisher exact tests. OR 0–1 indicates a trend towards mutual exclusivity, OR > 1 indicates a trend towards co-occurrence. U-CLL CLL with unmutated IGHV genes, M-CLL CLL with mutated IGHV genes.

Full size image

ATM aberrations and clinical outcome

We first evaluated the clinical impact of our ATM variant classification. Patients with putative germline/predicted neutral variants showed no significant differences in TTFT compared to ATM-wildtype cases while in contrast, cases with somatic ATM mutations displayed significantly shorter TTFT (Supplementary Fig. S2A–C). Results were also consistent when stratifying by del(11q) or limiting to Binet A cases (Supplementary Figs. S2D–F and S3), validating our re-categorization of germline/neutral variants with ATM-wildtype cases.

We assessed the clinical impact of ATM aberrations on TTFT by investigating del(11q), ATM mutations as well as bi-allelic ATM abnormalities (del(11q) and ATM mutations) independently in Binet A patients (2604/3631, 72% of the cohort). Here, any type of ATM abnormality resulted in significantly shorter TTFT when compared to ATM-wildtype cases (Fig. 4A). Among those with ATM mutations, no differences were detected when comparing missense versus nonsense/frameshift mutations (Fig. 4B and Supplementary Fig. S4) or when comparing mutations located in the C- versus N-terminal (Supplementary Fig. S5). We also investigated whether carrying multiple ATM mutations, as opposed to a single mutation, affected clinical outcome and found no evidence for a shorter TTFT in the overall cohort or in Binet A cases specifically (Fig. 4C and Supplementary Fig. S6).

Fig. 4: Clinical impact of ATM aberrations in Binet A CLL patients.
figure 4

Kaplan-Meier survival analysis using TTFT subdivided by (A) ATM aberration; somatic ATM mutation only, del(11q) only and combined ATM mutation and del(11q), (B) type of ATM mutation; frameshift/nonsense, missense/inframe indels with del(11q) displayed separately, and (C) comparing one versus multiple ATM mutations with del(11q) displayed separately.

Full size image

As we previously have shown that the prognostic impact of recurrent gene mutations in CLL is highly dependent on IGHV gene SHM status [22], we investigated if ATM mutations affected clinical outcome when patients were stratified into the U-CLL and M-CLL subgroups, again focusing primarily on Binet stage A CLL. Among patients with unmutated IGHV genes, only those carrying del(11q) but not ATM mutations had a significantly inferior TTFT when compared to ATM-wildtype cases (Fig. 5A). The same analysis in M-CLL cases revealed instead that any ATM abnormality, either mono-allelic, with or without del(11q) or bi-allelic, resulted in significantly shorter TTFT (Fig. 5B).

Fig. 5: Clinical impact of ATM aberrations in Binet A CLL patients stratified by IGHV somatic hypermutation status.
figure 5

Kaplan-Meier survival analysis using TTFT in Binet stage A in (A) U-CLL and (B) M-CLL patients. Pairwise comparisons were performed using the Cox–Mantel log-rank test. C Multivariable analysis using TTFT in Binet stage A CLL patients with U-CLL and M-CLL. U-CLL, CLL with unmutated IGHV genes, M-CLL, CLL with mutated IGHV genes. CI95, 95% confidence interval; * indicates a p value < 0.05, ** p < 0.01, and *** p < 0.001.

Full size image

Since mutations in ATM and SF3B1 often co-occurred, we also assessed if these aberrations in combination further affected TTFT. We noted no significant difference when comparing those carrying aberrations in ATM with or without an additional SF3B1 mutation (Supplementary Fig. S7). Furthermore, isolated ATM mutations (without other poor prognostic genetic markers) were rare, occurring in 1.2% of all Binet A cases and in 0.7% of Binet A M-CLL cases specifically. Of note, in Binet A patients, isolated ATM mutations did not have a significant impact on TTFT in either the U-CLL or M-CLL subgroups (Supplementary Fig. S8).

As the next step, we aimed to assess the specific contribution of ATM mutations on TTFT by employing a multivariable model. In addition to ATM mutations, the model included nine recurrently mutated genes [22], patient gender, age at diagnosis, del(11q) and trisomy 12 status, and IGHV gene SHM status. Patients carrying somatic mutations in TP53, or harboring del(17p), were combined as TP53 aberrant, while only cases carrying the hotspot mutation p.Y254fs/p.254* were considered mutated for NFKBIE [22]. Among Binet A cases, IGHV gene SHM status was the strongest significant biomarker of shorter TTFT, exhibiting the highest hazard ratio (Supplementary Fig. S9). In addition, mutations in SF3B1, EGR2, XPO1, TP53 aberrations, trisomy 12 and del(11q), but not ATM mutations, independently predicted shorter TTFT (Supplementary Fig. S9). In separate multivariable models for Binet A patients, stratified by IGHV status, mutations in SF3B1, TP53 aberrations, del(11q), trisomy 12 as well as EGR2 mutations were the only significant independent variables in U-CLL (Fig. 5C). For M-CLL patients, mutations in SF3B1, NOTCH1, XPO1, EGR2, trisomy 12 as well as del(11q) were significant biomarkers (Fig. 5C). ATM mutations had no independent significant effect on TTFT in either model.

Finally, we investigated the impact of different types of ATM mutations on OS across the entire cohort. Patients were stratified into four distinct groups based on their ATM status: ATM-wildtype, ATM-mutated, those carrying del(11q), and patients with biallelic alterations (del(11q) and ATM mutations). Any ATM aberration was associated with significantly reduced OS in the entire cohort and specifically among M-CLL patients. In contrast, within the U-CLL subgroup, only patients with del(11q) showed a significantly inferior outcome compared to those with wildtype ATM (Supplementary Fig. S10A–C). In multivariable analysis, del(11q), but not ATM mutations, emerged as a significant predictor of OS (Supplementary Fig. S10D–F).

Discussion

In the current study, we aimed to evaluate the prognostic significance of ATM mutations in CLL and their impact within the context of IGHV gene SHM status and other CLL-associated gene mutations, using data from a large, well-characterized, multi-center European cohort. Given that a large proportion of cases in the studied cohort included tumor-only data, we developed a hierarchical ranking system which integrates published data from databases on both cancer-associated variants and variants present in normal populations, enabling the classification of mutations as putative somatic or germline/neutral. Using our classification, we could accurately assign variants in samples within our dataset with available germline information. Putative somatic variants exhibited markedly different VAF distributions compared to those classified as putative germline or predicted neutral, with the majority of somatic variants showing frequencies <50% while those with VAFs >50% were most often detected in patients with del(11q). Notably, we identified 17 variants classified as putative somatic in patients without del(11q), all exhibiting VAFs >60%. In three of these cases, the elevated VAFs were attributable to cnLOH, consistent with what has been previously reported in CLL [37]. Finally, in concordance with a previous report, somatic missense mutations were predominantly found at the C-terminal domains while frameshift and nonsense mutations mainly affected the N-terminal [38].

We used TTFT as the primary clinical endpoint to evaluate how our classification of ATM variants affected clinical outcome and found cases harboring putative somatic variants to have a significantly inferior outcome in univariate analyses compared to those carrying putative germline/neutral variants, with the latter group exhibiting results comparable to those of ATM-wildtype cases. While our flowchart seems to effectively distinguish patients with somatic mutations from those classified as germline/putatively neutral in our analysis of TTFT, we cannot entirely exclude the possibility that some of the variants classified as somatic might be rare pathogenic germline mutations or that variants classified as neutral may in fact impact prognosis through disruption of normal ATM function. Additionally, selected germline variants excluded from our analysis may still influence tumor development in CLL, as suggested by recent publications [30, 31]. Ultimately, as a result of our classification, 246/3631 (6.8%) patients of our cohort had ATM mutations, while 57/246 cases (23.2%) carried multiple ATM mutations. The frequency observed is slightly lower than reported in previous studies and might be attributed to differences in classification strategies and the potential inclusion of germline or predicted neutral variants in other studies.

We noted that mutations in ATM often co-occurred with other recurrent abnormalities in CLL, predominantly del(11q) (45.5%) and SF3B1 mutations (23%). The former combination is well-known in CLL [7, 18, 39, 40], but the latter has also been reported previously in CLL, where both these mutations are suggested to be late-occurring CLL driver events [3, 19]. Interestingly, the combination of ATM deletions and SF3B1 mutations results in a CLL-like disease in elderly mice leading to genome instability, dysregulation of CLL-associated processes and providing mechanistic evidence for disease development [41].

We observed distinct patterns of co-occurrence or mutual exclusivity among ATM-mutated patients when stratified by IGHV gene SHM status. In samples harboring ATM mutations, trisomy 12 and TP53 mutations were found to be mutually exclusive in the U-CLL subgroup, while instead, SF3B1 mutations were enriched in the M-CLL subgroup. Focusing specifically on Binet A cases, which amounted to almost three-quarters of the cohort (72%, 2604 patients), isolated ATM mutations were a very rare finding detected only in 1.3% of cases and in 0.7% of M-CLL cases. This suggests that ATM mutation alone, excluding del(11q), may not be a key driver of CLL progression, in contrast to what is known about the relationship between TP53 mutation and del(17p) [42].

In order to explore the relevance of mono- and bi-allelic abnormalities of the ATM gene, we evaluated the clinical impact of ATM aberrations by separately analyzing del(11q), ATM mutations, and bi-allelic ATM abnormalities in Binet A patients. In this analysis, any type of ATM abnormality resulted in inferior outcome compared to wildtype cases. More interestingly, differences became more evident when we investigated U-CLL and M-CLL subgroups separately. In Binet stage A U-CLL patients, only isolated del(11q) resulted in inferior TTFT when compared to ATM-wildtype cases. Here, bi-allelic ATM abnormalities were not significant, possibly explained by the relatively low number of cases. On the contrary, the same analysis in M-CLL cases revealed instead that any ATM abnormality, either mono-or bi-allelic, resulted in significantly shorter TTFT. Of note, harboring multiple ATM mutations compared to a single mutation did not result in inferior TTFT in the entire cohort or in Binet A cases only. Additionally, TTFT analysis revealed no differences when comparing somatic missense to nonsense/frameshift mutations, when investigating C- versus N-terminal mutations or analyzing mutations in ATM with or without an additional SF3B1 mutation.

As most patients with ATM mutations also carried other genetic alterations, we assessed the specific contribution of ATM mutations, independent of del(11q), on TTFT by employing a multivariable model in Binet A cases stratified by IGHV status. We found that in U-CLL, del(11q) and not ATM mutations were an independent marker of TTFT, suggesting that ATM mutations alone do not independently contribute to disease progression. Notably, del(11q) was a highly significant marker for TTFT, adding similar risk as SF3B1 mutations or TP53 abnormalities. For M-CLL cases, again del(11q) but not ATM mutations was a highly significant marker for TTFT, similar to SF3B1 mutations. These findings highlight the prognostic impact of this aberration for both subgroups of patients. Similar findings were observed when analyzing OS, with ATM mutations emerging as a significant biomarker of poor prognosis in univariable models. However, in multivariable models, whether applied to the entire cohort or stratified by IGHV status, del(11q), but not ATM mutations, was identified as a significant predictor of OS. These latter results should however be interpreted with caution, given the substantial variability in treatment strategies among patients from the different participating centers.

The current retrospective study has several limitations which mainly arise from the multi-center composition of the cohort. Even though a significant proportion of patient samples were sequenced using targeted gene panels, a smaller subset underwent whole-genome (WGS) or whole-exome sequencing (WES) contributing to methodological variability. Additionally, for cases sequenced using NGS-based techniques, sequence alignment, variant calling, annotation and initial variant filtering were conducted independently at each participating center. Moreover, most samples were derived from unsorted peripheral blood mononuclear cells (PBMCs), and although unpurified CLL samples generally have high tumor content (>80%), mutations with VAFs near the 5% threshold could be missed, potentially underestimating the number of mutated cases. Finally, the proportion of cases requiring treatment seems slightly higher than in other published cohorts [43,44,45], likely explained by the long follow-up period (median 13 years) and the inclusion of samples from a clinical trial cohort.

In conclusion, our findings demonstrate that del(11q) significantly influences TTFT in CLL, impacting both the U-CLL and M-CLL subgroups. In contrast, somatic ATM mutations do not appear to have an independent prognostic impact in CLL. Isolated ATM mutations were rare, especially in M-CLL. Instead, ATM mutations often co-occurred with other poor-prognostic biomarkers, particularly del(11q) and SF3B1 mutations. Although del(11q) is strongly associated with a shorter TTFT, patients with this genetic aberration treated with chemoimmunotherapy can overcome the poor prognosis linked to del(11q), achieving prolonged OS [10]. Similar results are observed in ibrutinib-treated patients, where those with del(11q) were suggested to have a significantly longer PFS compared to patients without del(11q) [46,47,48]. Furthermore, relapsed/refractory patients with CLL carrying del(11q) were also shown to benefit from the combination venetoclax-rituximab compared to those receiving bendamustine and rituximab therapy [49]. These results highlight the need for large-scale, uniform cohort studies to better understand the overall prognostic and predictive roles of not only del(11q) and ATM mutations, but also other recurrent genetic abnormalities in CLL. Based on current evidence, we propose that del(11q), but not ATM mutations, should be considered a key factor in identifying patients at high risk for early disease progression and in determining therapeutic strategies. This is particularly important in the M-CLL subgroup, where del(11q) along with other recurrent mutations can help identify outlier cases with a poor prognosis despite otherwise indolent features.