Molecular biomarkers as key determinants of outcome in mantle cell lymphoma: results from the FIL V-RBAC trial

Key Points

• •TP53 and CD36 mutations, CDKN2A loss, and ATM single-hit abnormalities independently predict outcomes in older patients with MCL.
• •The 4-factor molecular prognostic model underscores the key role of molecular biomarkers in older patients with MCL.

Visual Abstract

Abstract

This study comprehensively dissects the molecular landscape of older patients with mantle cell lymphoma (MCL) enrolled in the phase 2 FIL V-RBAC trial of the Fondazione Italiana Linfomi (FIL). Of the 140 patients enrolled, 132 had available genomic DNA extracted from lymph node biopsies or bone marrow aspirates and were included in the analysis. A cancer personalized profiling by deep sequencing assay targeting 146 genes relevant to MCL pathogenesis was used to identify gene mutations and copy number variations. ATM was the most frequently mutated gene, detected in 55 patients (41.7%), followed by TP53 and KMT2D in 31 patients (23.5%). ATM deletion was observed in 32 patients (24%), whereas CDKN2A loss was observed in 29 (22%). Beyond TP53 mutations, 3 other molecular lesions, including CDKN2A loss, CD36 mutations, and single-hit ATM abnormalities (either mutation or deletion), were independently associated with progression-free survival after adjustment for high-risk trial-defining features, namely Ki-67 ≥30% and blastoid variant. Notably, patients harboring single-hit ATM alterations without any additional risk factors achieved durable long-term remission, whereas CD36 mutations were associated with adverse survival. Both findings represent previously unrecognized aberrations in this cohort that were independently and inversely associated with survival. The 4 variables were integrated into a 4-factor molecular prognostic model internally validated using a bootstrapping approach, which identified 4 distinct patient subgroups with significantly different outcomes. These findings support the importance of (1) molecular profiling in MCL, (2) risk-adapted trials such as FIL V-RBAC, and (3) the integration of other biological markers with TP53 mutations for a more precise risk assessment in MCL. This trial was registered at www.clinicaltrials.gov as NCT03567876.

Introduction

The characterization of the molecular landscape of mantle cell lymphoma (MCL) has enabled the identification of biomarkers associated with poor outcomes, namely TP53 mutations, which are now incorporated into the routine diagnostic workup of newly diagnosed patients with MCL. Beyond TP53 abnormalities, the molecular landscape of MCL includes numerous additional genetic lesions, encompassing both gene mutations and chromosomal abnormalities, many of which have been linked to unfavorable prognosis.1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12 Importantly, most of these genetic alterations have been identified in the context of clinical trials enrolling young and fit patients eligible for autologous stem cell transplantation (ASCT), whereas data in older patients with MCL remain limited.

The therapeutic landscape of MCL is rapidly evolving, with the imminent introduction of pathway inhibitors, particularly Bruton tyrosine kinase inhibitors, in the frontline setting for both young and older patients.13, 14, 15 Despite these advances, clinical outcomes remain heterogeneous, particularly among older patients who are not candidates for ASCT. Although TP53 disruption has emerged as a major determinant of poor prognosis, additional genomic biomarkers capable of refining risk stratification and predicting progression-free survival (PFS) remain poorly defined in this population. The identification of such biomarkers could improve prognostic assessment and support risk-adapted treatment strategies in older patients with MCL.

The FIL V-RBAC trial represents, to our knowledge, the first prospective study to stratify patients with MCL to different treatments according to their molecular (TP53 abnormalities) and clinicopathological (Ki-67 levels and blastoid variant) risk profile, offering venetoclax consolidation and maintenance to high-risk (HR) patients.¹⁶ This platform provides a unique opportunity to comprehensively evaluate the molecular landscape of patients with MCL not candidate for ASCT, potentially supporting future treatment tailoring that maximizes efficacy and spares unnecessary toxicity.

Materials and methods

Study details

FIL V-RBAC is a multicenter phase 2 clinical trial of the Fondazione Italiana Linfomi (FIL) that enrolled previously untreated patients with MCL. Eligible patients were stratified into low-risk (LR) and HR groups based on Ki-67 expression, TP53 mutations/deletions, and morphology (presence of blastoid variant). All patients received 4 cycles of abbreviated R-BAC (rituximab, bendamustine, and cytarabine) induction therapy. LR patients received 2 additional cycles of R-BAC, whereas HR patients received fixed-duration oral venetoclax consolidation (4 months, 800 mg/d) and maintenance (20 months, 400 mg/d). The clinical results of the FIL V-RBAC have been recently reported.¹⁶

This study represents the molecular ancillary study of the FIL V-RBAC trial dedicated to dissecting the prognostic role of gene mutations and copy number variations (CNVs). Mutational analysis was performed on DNA extracted from formalin-fixed, paraffin-embedded tissue (n = 74), bone marrow aspirate (n = 47), or peripheral blood (n = 11) involved by MCL, as documented by flow cytometry.

Mutational analysis

The mutational landscape of the cohort was characterized using the LyV4.0 cancer personalized profiling by deep sequencing (CAPP-Seq) assay, targeting a panel of 146 genes (348 978 bp) relevant to MCL pathogenesis (supplemental Table 1). Libraries were sequenced on the Illumina MiSeq and NextSeq 550 platforms, achieving >2000× coverage in >80% of target regions, and analyzed using a previously validated bioinformatic pipeline.^2,17,18 A variant allele frequency of 1% was set for variant calling in peripheral blood/bone marrow samples and 5% on formalin-fixed, paraffin-embedded samples.

CNV analysis

Genome-wide somatic CNVs were analyzed using the CNVkit toolkit version 0.9.12, implemented in Python 3.10.^19,20 CNVkit is designed to infer copy number alterations from targeted capture sequencing data. Sequencing BAM files generated by the CAPP-Seq pipeline, including both on-target and off-target reads, were used as input.²⁰ Analyses were performed using default CNVkit parameters. A pooled reference file was constructed from 18 DNA samples extracted from granulocytes isolated with gradient density centrifugation and obtained from patients with lymphoma without peripheral blood involvement, as documented by flow cytometry. Read depths were median centered and bias corrected (GC content, sequence repeats, target density) to produce normalized log₂ read-depth values for each bin. To infer copy number changes, “observed” normalized log₂ read-depth values from patient samples were subtracted from the “expected” values in the reference file, and bins were segmented using circular binary segmentation with the “droplow-coverage” option.^21,22 Statistically significant recurrent somatic copy-number alterations (SCNAs) were identified using the GISTIC2.0 GenePattern module version 6.15.30.^23,24 The thresholds used in our analysis were 0.1 for amplifications and 0.3 for deletions.

Statistical analysis

The primary end point of the clinical study was PFS, and the secondary end point was overall survival (OS). Survival analysis for PFS and OS was performed using the Kaplan-Meier method and compared between groups using the log-rank test. Cox proportional hazards models were used to estimate adjusted hazard ratios for the association between exposure variables and time-to-event outcomes. Internal validation of the multivariate analysis was performed using the bootstrap approach with 1000 bootstrap samples. The Mann-Whitney U test for continuous variables and χ² test for categorical variables were used to compare patient characteristics and mutational profiles. The analysis was conducted using R version 4.4.2. (R Foundation for Statistical Computing, Vienna, Austria) and Statistical Package for the Social Sciences version 24.0 (IBM, Chicago, IL).

Results

Patient characteristics

Out of the 140 patients enrolled in the FIL V-RBAC trial, 132 (94.3%) were included in the present molecular analysis. The median age at diagnosis of the 132 analyzed patients was 72 years (interquartile range, 69-75), and most were male (n = 100 [75.8%]). Ki-67 ≥30% was present in 33 (25.0%) patients, and blastoid morphology in 13 patients (9.8%). The median follow-up of patients included in the molecular analysis was 30.5 months (interquartile range, 26.7-34.2). Importantly, patients analyzed and not analyzed for mutations and CNV showed superimposable PFS and OS (supplemental Figure 1) and superimposable baseline clinical characteristics (Table 1). Well-established clinical and pathological prognostic biomarkers of MCL, including mantle cell lymphoma international prognostic index combined (MIPI-c), Ki-67 ≥30%, and blastoid variant, were associated with PFS (supplemental Figure 2). The complete baseline clinical characteristics of the studied cohort are summarized in Table 1.

Table 1.

Patient characteristics

	Analyzed (n = 132)	Not analyzed (n = 8)	P value
Age, median (IQR), y	72.0 (69-75)	70.5 (67-72.5)	.35
Gender, n (%)			.68
Female	32 (24.2)	1 (12.5)
Male	100 (75.8)	7 (87.5)
ECOG performance status, n (%)			.377
0	92 (69.7)	4 (50.0)
1	31 (23.5)	4 (50.0)
2	6 (4.5)	0 (0)
MIPI risk group, n (%)			.01
Low (<5.7)	5 (3.8)	0 (0)
Intermediate (5.7 to <6.2)	53 (40.1)	7 (87.5)
High (≥6.2)	71 (53.8)	0 (0)
NA	3 (2.3)	1 (12.5)
Trial risk group, n (%)			.15
LR	79 (59.8)	7 (87.5)
HR	53 (40.2)	1 (12.5)
Ki-67 expression, n (%)			.67
<30%	99 (75.0)	7 (87.5)
≥30%	33 (25.0)	1 (12.5)
Blastoid variant, n (%)			.90
No	119 (90.2)	8 (100)
Yes	13 (9.8)	0 (0)
Bone marrow involvement, n (%)			.19
Negative	32 (24.3)	0 (0)
Positive	91 (68.9)	8 (100)
NA	9 (6.8)

ECOG, Eastern Cooperative Oncology Group; IQR, interquartile range; NA, not available.

Molecular landscape of the studied cohort

CAPP-Seq identified at least 1 nonsynonymous somatic mutation or 1 CNV in all patients (Figure 1A; supplemental Tables 2 and 3). The most frequently mutated gene was ATM, detected in 55 (41.7%) patients, followed by TP53 and KMT2D in 31 (23.5%) patients. Mutations in WHSC1 were observed in 19 (14.4%) patients, KMT2C in 15 (11.4%) patients, and both BIRC3 and UBR5 in 13 (9.8%) patients. NOTCH1 mutations occurred in 10 (7.6%) patients and CD36 mutations occurred in 7 (5.3%) patients (Figure 1B). In line with the biology of MCL, most mutations of ATM (n = 39 [54.2%]), TP53 (n = 29 [58%]), and WHSC1 (n = 19 [100%]) were nonsynonymous events. In contrast, mutations in KMT2D and UBR5 were predominantly truncating events (n = 30 [96.8%]; Figure 1C).

Figure 1.

Molecular landscape of FIL V-RBAC trial and its association with baseline clinical and pathological features. (A) Heat map representing the mutational profiles of 132 MCL cases, genotyped on tumor DNA. Each column represents 1 patient, and each row represents 1 characteristic. Ki-67 levels and blastoid variant are represented above the heat map. (B-C) Histograms representing the prevalence (B) and the different types (C) of mutations identified. (D) Histograms representing the prevalence of the CNVs identified in >10 patients; amplifications are shown in pink, and deletions are in purple. (E) Correlation map comparing the most recurrent gene mutations and CNVs with baseline clinical and pathological features. The red color scale of −log₁₀(P) points to a co-occurrence between 2 variables. LDH, lactate dehydrogenase; Ins/Del, insertion/deletion; SNV, single-nucleotide variant.

CNV analysis revealed at least 1 CNV in all patients. The most frequent alterations included 3q26.32 amplification in 43 (33%) patients, 13q34 deletion in 37 (28%) patients, 1p21.3 deletion in 35 (27%) patients, 11q22.3 deletion encompassing the ATM locus in 32 (24%) patients, 13q14.3 deletion in 32 (24%) patients, and 9p21.3 deletion encompassing the CDKN2A locus in 29 (22%) patients (Figure 1D).

The complete set of clinical information collected for all patients allows the potential role of different gene mutations in shaping the subsequent clinical presentation to be dissected. WHSC1 mutations were associated with adverse disease features, including Ki-67 ≥30%, elevated white blood cell count, higher MIPI score, and bone marrow involvement. TP53 mutations were associated with Ki-67 ≥30%, elevated white blood cell count, and elevated MIPI scores. NOTCH1 mutations were associated with the blastoid variant. CDKN2A loss was associated with the blastoid variant, Ki-67 ≥30%, and high lactate dehydrogenase. CD36 mutations were not associated with any HR baseline clinical and pathological feature (Figure 1D).

CD36 mutations, in addition to TP53 mutations, are associated with poor outcomes in older patients with MCL

TP53 mutations were associated with significantly inferior outcomes, with a 3-year PFS of 30.8% (95% confidence interval [CI], 16.6-56.9) and a 3-year OS of 45.7% (95% CI, 29.1-71.7), compared with 79.9% (95% CI, 70.7-90.2) and 78.3% (95% CI, 68.5-89.5) for patients with wild-type genotype, respectively (both P < .001; Figure 2A-B). To evaluate the role of the different types of TP53 mutations, we stratified them as loss of function (LOF) or gain of function (GOF), as previously described.²⁵ Patients with TP53 LOF mutations had a trend toward worse PFS, with a 3-year PFS of 24.8% (95% CI, 10.5-58.5) compared with 45.7% (95% CI, 20.3-100) for patients with TP53 GOF mutations. Similarly, patients with TP53 LOF mutations had worse OS, with a 3-year OS of 28.6% (95% CI, 12.8-64.0) compared with 88.9% (95% CI, 70.5-100) for patients with TP53 GOF mutations (Figure 2C-D).

Figure 2.

Prognostic impact of TP53 and CD36 mutations. Kaplan-Meier estimates of PFS (A) and OS (B) of patients with TP53 mutation (in red) vs WT genotype (in blue). Kaplan-Meier estimates of PFS (C) and OS (D) according to different TP53 mutation types. Patients with LOF mutations are represented in red, patients with GOF mutations in yellow, and patients with WT genotype in blue. Kaplan-Meier estimates of PFS (E) and OS (F) of patients with CD36 mutation (in red) vs WT genotype (in blue). P values are presented adjacent to the curves. (G) Forest plot multivariate analysis in terms of PFS for HR trial-defining features and CD36 mutations. mut, mutated; WT, wild-type.

In addition to TP53 mutations, CD36 mutations were sorted out as a potential novel prognostic biomarker in patients with MCL. Patients with CD36 mutations (n = 7 [5.3%]) presented a 3-year PFS of 14.3% (95% CI, 2.3-87.7) and a 3-year OS of 21.4% (95% CI, 4.2-100), compared with 71.1% (95% CI, 61.9-81.9) and 73.8% (95% CI, 64.6-84.2), respectively, among patients with wild-type CD36 (P < .0001 and P < .001, respectively; Figure 2E-F). A total of 7 CD36 mutations were identified, including 3 missense events, 2 splicing events, 1 frameshift, and 1 nonsense event (supplemental Figure 3A-B). All the mutations identified were predicted to alter the function of the CD36 protein. Interestingly, consistent with the lack of association between CD36 mutations and any HR baseline clinical or pathological features, CD36 mutations also did not significantly co-occur with any other HR mutations and/or CNVs (supplemental Figure 3C).

To further clarify the independent prognostic impact of CD36 mutations, they were tested in multivariate analysis with the presence of any HR trial-defining features (ie, TP53 disruption and/or Ki-67 >30% and/or blastoid variant) and retained independent prognostic value (HR, 4.78; 95% CI, 1.95-11.7; P < .001; Figure 2G).

CDKN2A loss and ATM single-hit abnormalities represent independent prognostic biomarkers in older patients with MCL

In addition to gene mutations, the prognostic role of CNVs was also evaluated. Patients harboring CDKN2A loss presented a 3-year PFS and OS of 43.1% (95% CI, 25.7-72.2) and 54.8% (95% CI, 41.9-81.6) compared with 75.0% (95% CI, 65.3-86.1) and 73.7% (95% CI, 63.1-86.0) for patients with wild-type genotype (P = .003 and P = .036, respectively; Figure 3A-B). Importantly, by performing a multivariate analysis that included the presence of any HR trial-defining features (ie, TP53 disruption and/or Ki-67 >30% and/or blastoid variant) and CD36 mutations, CDKN2A loss (HR 2.35; 95% CI, 1.17-4.76; P = .02) retained independent prognostic value for shorter PFS (Figure 3C).

Figure 3.

Prognostic impact of CDKN2A deletion and ATM abnormalities. Kaplan-Meier estimates of PFS (A) and OS (B) of patients with CDKN2A loss (in red) vs WT genotype (in blue). (C) Forest plot multivariate analysis in terms of PFS for HR trial-defining features, CD36 mutations, and CDKN2A loss. Kaplan-Meier estimates of PFS (D) and OS (E) according to ATM SH abnormalities. Patients with either ATM mutations or deletion are represented in red, and all the other patients (ie, patients with both ATM mutation and deletion and patients with WT genotype) in blue. P values are presented adjacent to the curves. (F) Forest plot multivariate analysis in terms of PFS for HR trial-defining features, CD36 mutations, CDKN2A loss, and SH ATM abnormalities. (G) Forest plot multivariate analysis in terms of PFS for TP53 mutations, CD36 mutations, CDKN2A loss, SH ATM abnormalities, and MIPI-c. mut, mutated; SH, single-hit.

Subsequently, the potential different prognostic role of ATM mutation and 11q22.3 deletion encompassing the ATM locus was evaluated. Patients harboring single-hit ATM abnormalities (ie, either mutation or deletion) were characterized by longer PFS compared to patients with double-hit ATM abnormalities (ie, both mutations and deletions) or wild-type patients. More precisely, patients with ATM mutations only had a 3-year PFS of 89.0% (95% CI, 77.5-100), patients with ATM deletions only had a 3-year PFS of 76.9% (95% CI, 52.0-100), patients with double-hit ATM had a 3-year PFS of 71.1% (95% CI, 52.3-96.6), and patients with wild-type ATM had a 3-year PFS of 50.5% (95% CI, 36.0-70.8; supplemental Figure 4). Therefore, patients with single-hit ATM abnormalities were grouped together and presented a 3-year PFS of 86.0% (95% CI, 74.7-98.8) compared with 56.1% (95% CI, 43.8-71.8) in other patients (P = .002) and a 3-year OS of 89.8% (95% CI, 80.3-100) compared with 59.1% (95% CI, 46.8-74.6) in other patients (P = .002; Figure 3D-E). Importantly, single-hit ATM abnormalities retained an independent association with better outcomes (HR, 0.26; 95% CI, 0.10-0.67; P = .005) when adjusted for HR trial-defining features (ie, TP53 disruption, Ki-67 >30%, and/or blastoid variant), CD36 mutations, and CDKN2A loss (Figure 3F).

The 4-factor molecular prognostic model and the impact of venetoclax

Since the HR trial-defining features lost the independent prognostic factor in multivariate analysis when adjusted for CD36 mutations, CDKN2A loss, and single-hit ATM abnormalities (Figure 3F), we tried to extrapolate the strongest molecular biomarker enclosed in the HR trial-defining features, namely, TP53 mutation, and include it in a multivariate analysis. Importantly, CD36 mutations, TP53 mutation, CDKN2A loss, and single-hit ATM abnormalities also retained independent prognostic value after adjustment for the validated MIPI-c score (Figure 3G).

Therefore, a multivariate analysis was performed including only the 4 molecular biomarkers that retained independent prognostic value (Figure 4A). Based on the β coefficients rounded to the nearest integer, 1 point was assigned to TP53 mutations, CD36 mutations, and CDKN2A loss. Because single-hit ATM abnormalities were associated with a better outcome, this variable was recoded in the model as “no ATM single hit” to ensure it carried a positive rather than negative value. Accordingly, 1 point was also assigned to “no ATM single hit.” This scoring system identified 4 distinct risk groups with significantly different PFS and OS. LR MCL (0 points; n = 31), composed entirely of patients carrying single-hit ATM abnormalities without any additional HR molecular lesion, showed 3-year PFS and OS rates of 100%. Intermediate-1–risk MCL (1 point; n = 63) had 3-year PFS and OS rates of 75.7% (95% CI, 62.5-91.6) and 74.9% (95% CI, 61.3-91.6), respectively. Intermediate-2–risk MCL (2 points; n = 28) demonstrated 3-year PFS and OS rates of 46.0% (95% CI, 29.9-70.5) and 47.0% (95% CI, 30.2-73.1), respectively. HR MCL (3-4 points; n = 10) had poor outcomes, with 3-year PFS and OS rates of 0% (95% CI, 80.3-100) and 33.3% (95% CI, 13.2-83.9), respectively (Figure 4B-C).

Figure 4.

The 4-factor molecular prognostic model. (A) Multivariate analysis with bootstrapping validation of PFS, including TP53 mutations, CD36 mutations, CDKN2A loss, and single-hit ATM abnormalities. (B) Table showing how patients were grouped into risk classes based on their total score. Kaplan-Meier estimates of (C) PFS and (D) OS according to the 4-factor score. LR patients (0 points) are represented by the blue curves, intermediate-1–risk patients (1 point) are represented by the gray curves, intermediate-2–risk patients (2 points) are represented by the yellow curves, and HR patients (3-4 points) are represented by the red curves. Pairwise P values are presented in the tables adjacent to the curves. Int-2, intermediate-1-risk; Int-2, intermediate-2-risk; LCI, lower confidence interval; mut, mutated; UCI, upper confidence intervel.

Minimal residual disease (MRD) negativity rates at the end of R-BAC induction and after venetoclax consolidation were compared across risk groups. This analysis was limited by the small sample size and by the fact that MRD assessment in the FIL V-RBAC trial was performed per protocol only in patients classified as HR according to the HR trial-defining features. Although statistical correlation could not be performed because of the limited number of evaluable patients, MRD negativity rates appeared to decrease progressively across increasing risk categories of the 4-factor model (supplemental Figure 5).

Notably, the variables of the 4-factor prognostic model were all frequently selected by internal 1000× bootstrap validation (Figure 4A). Also, the C-index of the model for PFS and OS was 0.783 and 0.759, respectively, and 0.782 and 0.756 after optimism correction using 1000 bootstrap resamples.

The model was subsequently evaluated on an explorative basis in patients who did and did not receive venetoclax according to trial indication, aware of the nonrandomized nature of venetoclax treatment and the limited number of cases in this subgroup analysis. Importantly, some patients (n = 11) who were intended to receive venetoclax did not ultimately receive the drug because of disease progression before its planned initiation. Therefore, in the group that received venetoclax, the starting point of the survival analysis was defined by the date of venetoclax initiation and not by the time of trial enrollment. Overall, the model consistently stratified patients regardless of venetoclax exposure (supplemental Figure 6A-B). However, venetoclax had a potential different impact according to the different risk groups of the 4-factor molecular prognostic model. Patients belonging to the HR or intermediate-2–risk groups that received venetoclax had a trend toward improved PFS, with a 3-year PFS of 38.4% (95% CI, 22.1-66.9) compared with 26.8% (95% CI, 10.8-66.0) among patients who did not receive venetoclax (P = .100; supplemental Figure 6C). This PFS trend translated into longer OS, with a 3-year OS of 53.2% (95% CI, 34.1-82.9) among patients who received venetoclax, compared with 25.7% (95% CI, 10.1-65.6) among patients who did not receive venetoclax (P = .011; supplemental Figure 6D). Conversely, venetoclax does not affect PFS or OS in patients classified as intermediate-1 risk or LR, accordingly (supplemental Figure 6E-F).

An additional exploratory analysis was conducted to describe PFS according to each molecular and clinical biomarker among patients who received venetoclax and those who did not (supplemental Figures 7 and 8). Overall, in both scenarios, MCL prognostic biomarkers retained prognostic value with some slightly different signals (eg, Ki-67 levels). The results should be interpreted with caution, given the nonrandomized administration of venetoclax, the limited sample size of this subgroup analysis, and importantly, the fact that a proportion of patients who were intended to receive venetoclax according to HR trial-defining features (TP53 mutations, Ki-67 ≥30%, and blastoid variant) did not ultimately receive the drug because of early disease progression and subsequent discontinuation from the trial.

Discussion

This study provides, to our knowledge, the first comprehensive molecular characterization of the phase 2 FIL V-RBAC clinical trial, which was specifically designed for older patients with MCL. By integrating targeted sequencing and CNV, the data show that (1) CD36 mutations were sorted out as a potential novel biomarker of poor prognosis in MCL; (2) ATM single-hit abnormalities represent the first biomarker associated with long-lasting remission in patients with MCL; (3) a 4-factor prognostic model integrating TP53 mutations, CDKN2A loss, CD36 mutations, and ATM single-hit abnormalities outperforms clinical and pathological variables in outcome prediction of older patients with MCL; and (4) molecularly HR patients benefit the most by the addition of venetoclax.

The prevalence of gene mutations and CNV and their association with both baseline clinical features and outcomes are in line with previous reports for patients with MCL.1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 Despite retaining independent prognostic value, TP53 mutations and CDKN2A loss were enriched in patients with high Ki-67, elevated MIPI scores, and blastoid morphology, further highlighting their role in shaping the subsequent clinical aggressive presentation. In addition, emerging evidence suggests that not all pathogenic TP53 mutations are associated with the same outcome.^26,27 Despite the limited number of cases, the distinction between LOF and GOF TP53 mutations further refined prognostication, with LOF variants conferring particularly poor outcomes. These findings require further validation in larger cohorts of patients, including young patients treated with ASCT and/or patients treated with Bruton tyrosine kinase inhibitor as first-line therapy.

This molecular analysis identifies CD36 mutations as a potential novel and independent adverse prognostic factor in older patients with MCL. Despite their low frequency, patients with CD36 mutations experienced inferior PFS and OS, and this effect persisted in multivariate analyses adjusted for established HR clinical and molecular features. In contrast to TP53 and CDKN2A loss, which are associated with HR clinical and molecular features, CD36 mutations are not linked to HR characteristics, suggesting an independent mechanism of aggressiveness. It is notable that CD36 codes for a transmembrane protein that regulates lipid metabolism, tumor-microenvironment interactions, and NF-κB signaling, which are key pathways in MCL biology.^28,29 In addition, CD36 surface expression has already been associated with unfavorable outcomes in patients affected by MCL or acute myeloid leukemia.30, 31, 32, 33 However, the precise biological mechanisms by which CD36 mutations may promote MCL proliferation and progression require dedicated functional studies and are beyond the scope of the present translational analysis.

Recent evidence from different B-cell malignancies suggests that not all ATM abnormalities are associated with the same outcomes.^7,34,35 This analysis shows that single-hit ATM abnormalities, defined as either mutation or deletion alone, were associated with superior outcomes compared with both double-hit ATM alterations and ATM wild-type status. This finding suggests a complex, nonlinear relationship between ATM dysfunction and treatment response, potentially reflecting partial impairment of DNA damage response pathways that enhance chemosensitivity without fully abrogating genomic stability.³⁶

The molecular analysis of the FIL V-RBAC phase 2 clinical trial documents that the well-established prognostic biomarkers of MCL in younger patients (TP53 mutations and CDKN2A loss) retain independent prognostic value also in older patients with MCL. This finding renders the cohort suitable for the identification of other novel potential biomarkers for this clinical setting. By integrating TP53 mutations and CDKN2A loss with CD36 mutations and single-hit ATM abnormalities, a 4-factor molecular prognostic model was devised that stratified patients into 4 distinct risk groups with unique outcomes. This model outperformed traditional HR trial-defining features, and all the 4 variables were selected by bootstrapping validation. Furthermore, to mitigate the lack of an external validation cohort, we applied a bootstrapping approach that strongly supported the independent prognostic significance of CD36 mutations and single-hit ATM abnormalities. Because the FIL V-RBAC trial did not include rituximab maintenance, the potential of this strategy to mitigate the adverse impact of HR molecular features warrants investigation in dedicated cohorts treated with rituximab maintenance.^14,15,37 Lastly, despite the nonrandomized design of the trial, the model provided insights for a differential benefit of the addition of venetoclax across risk groups. HR patients appeared to benefit the most from venetoclax, consequently corroborating the role of molecularly guided venetoclax use in HR patients with MCL.38, 39, 40

In summary, this study represents, to our knowledge, the largest comprehensive molecular analysis of older patients with MCL enrolled in a clinical trial. The results further reinforce the notion that molecular analysis is essential to guide treatment strategies in MCL. Beyond TP53 mutations and CDKN2A loss, additional molecular features, including CD36 mutations and ATM single-hit abnormalities, emerge as novel biomarkers that, if independently validated, may be included in dedicated clinical trials to further improve precision medicine for patients with MCL.

Conflict-of-interest disclosure: R.M. is on advisory boards and speaker’s bureaus for Johnson & Johnson, AbbVie, AstraZeneca, Eli Lilly, and BeOne Medicines. G.G. is on advisory boards and speaker’s bureaus for AstraZeneca, AbbVie, BeOne Medicines, Incyte, Johnson & Johnson, and Eli Lilly; and has received research support from Gilead. P.Z. is on advisory boards and speaker’s bureaus for Secura Bio, Recordati, Kite, Gilead, Janssen-Cilag, Bristol Myers Squibb, Sobi, Sandoz, MSD, Regeneron, Takeda, Roche, AstraZeneca, Kyowa Kirin, Novartis, ADC Therapeutics, Incyte, BeOne Medicines, and Eli Lilly. V.R.Z. is on advisory boards for AbbVie, AstraZeneca, Bristol Myers Squibb, Kite/Gilead, Novartis, Roche, Sobi, and Takeda; has consultancy roles with Roche; has received research funding from Kite/Gilead; participates in speaker’s bureaus for AbbVie, AstraZeneca, Janssen, Eli Lilly, Sobi, and Takeda; and has received travel support from AbbVie, Janssen, Eli Lilly, Roche, and Takeda. P.M.S. is on advisory boards for Bristol Myers Squibb, Kite/Gilead, Roche, Sobi, and Takeda; participates in speaker’s bureaus for Sobi, Takeda, and Incyte; and has received travel grants from Janssen, Roche, Takeda, Incyte, and Sobi. F.P. is a consultant and is on advisory boards and speaker’s bureaus for AbbVie, AstraZeneca, BeOne Medicines, Bristol Myers Squibb, Janssen, Kite/Gilead, Roche, and Sobi. G.L. is on advisory boards and speaker’s bureaus for Janssen-Cilag, Gilead, Roche, Italfarmaco, Takeda, AbbVie, Incyte, AstraZeneca, BeiGene, GSK, and Eli Lilly. G.M. is on advisory boards and speaker’s bureaus for Janssen, Incyte, Roche, AbbVie, Takeda, Genmab, AstraZeneca, Eli Lilly, Sobi, and BeOne Medicines. A.D.R. is on advisory boards for AbbVie, AstraZeneca, Johnson & Johnson, Bristol Myers Squibb, Kite/Gilead, Roche, Sobi, and Takeda; is a consultant for Incyte, AbbVie, Roche, Novartis, Johnson & Johnson, and Kite/Gilead; and participates in speaker’s bureaus for AbbVie, Johnson & Johnson, Eli Lilly, Sobi, Roche, Kite/Gilead, Bristol Myers Squibb, Incyte, and Novartis. J.O. is on advisory boards for AbbVie, AstraZeneca, and Kite; and participates in speaker’s bureaus for AbbVie, AstraZeneca, Janssen, and Kite. C.P. is on advisory boards for AbbVie, AstraZeneca, Johnson & Johnson, Roche, Sobi, Takeda, Incyte, and Eli Lilly. S. Ferrero. is a consultant for Janssen, EUSA Pharma, AbbVie, and Sandoz; is on advisory boards for Janssen, EUSA Pharma, Recordati, Incyte, Roche, AstraZeneca, Italfarmaco, and CSL Behring; has received speaker’s honoraria from Janssen, EUSA Pharma, Recordati, Eli Lilly, BeiGene, Gilead, and Gentili; and has received research funding from Gilead and MorphoSys. C.V. is on advisory boards and speaker’s bureaus for, and has received travel grants and research support from, AbbVie, Bristol Myers Squibb, AstraZeneca, Servier, Incyte, Roche, Pfizer, Novartis, Gentili, Johnson & Johnson, Kite/Gilead, BeiGene, Eli Lilly, and Kyowa Kirin. The remaining authors declare no competing financial interests.