The Baseline Molecular Profile of Late Relapse: Identifying Drivers of Delayed Recurrence in DLBCL

Conflicts of Interest

Michal Masar: none. Joseph Novak: none. Janek Walker: none. Prokop Vodicka: none. Abigail Dropik: none. Zilin Xianyu: none. Matthew Maurer: Research funding: Roche, BMS, Arima. Thomas Habermann: Data Monitoring Committee: Seagen, Tess Therapeutics, Eli Lilly & Co.; Scientific Advisory Board (no personal compensation): Morpohsys, Incyte, Beigene, Loxo Oncology; Research funding (to institution): Genentech, Sorrento, BMS. Eric Mou: none. Rebecca L. King: none. Kerstin Wenzl: employment by BMS. Thomas Witzig: Research funding (to institution): Karyopharm, Kura Oncology; Advisory board (compensation to institution): ADC Therapeutics; Honoraria: Curio Science (compensation to institution and personal). Stephen Ansell: Research funding (to institution): Takeda, BMS, Regeneron, Affimed, Pfizer, AstraZeneca, StepPharma, ADC Therapeutics. Pavel Klener: none. James Cerhan: Research funding (to institution): BMS, Genentech, Genmab; SMB (personal compensation): Protagonist Therapeutics. Marek Trneny: none. Yucai Wang: Research funding (to institution): Incyte, InnoCare, Loxo Oncology, Eli Lilly, MorphoSys, Novartis, Genentech, Genmab, AbbVie, BeiGene, BeOne, Merck, AstraZeneca, Bristol Myers Squibb. Advisory board (compensation to institution): Eli Lilly, Loxo Oncology, TG Therapeutics, Incyte, InnoCare, Kite, Janssen, BeiGene, BeOne, AstraZeneca, Genmab, AbbVie. Consultancy (compensation to institution): InnoCare, AbbVie. Honorarium (to institution): Kite. Magdalena Klanova: none. Anne Novak: Research funding (to institution): BMS/Celgene.

To the Editor,

The clinical management of diffuse large B‐cell lymphoma (DLBCL) is increasingly directed toward identification of early events. Primary treatment resistant patients (PTR) with a lymphoma‐related event within 6 months from diagnosis [1], or those who do not achieve event‐free survival within 24 months from diagnosis have very poor outcomes [2]. Molecular models for prediction of early clinical failure have been recently developed [3]; however, 8%–10% of EFS24 achievers subsequently relapse; these cases are classified as late relapse (LR), defined as recurrence occurring more than 24 months after diagnosis (∼5% of DLBCL) [4]. LR cases often exhibit a distinct genetic evolution which accounts for more favorable outcomes (median OS ranging 72–83 months) and sustained chemosensitivity compared to early events [5, 6]. Despite better prognosis, LR reflects a failure of frontline treatment durability, with a clinical and biological trajectory distinct from non‐relapsed (NR) patients. Furthermore, in clinical practice, the frequency of routine follow‐up visits is often reduced in patients achieving EFS24. To date, a comprehensive characterization of the baseline molecular profiles differentiating those that will progress with LR from those that are NR and PTR is lacking.

We aimed to define the genomic and transcriptomic profile of LR patients to identify distinct biological hallmarks at diagnosis that differentiate these cases primarily from long‐term remission achievers (i.e., NR; primary endpoint), to compare LR cases with the PTR group (secondary endpoint).

A total of 289 patients with newly diagnosed DLBCL who received frontline immunochemotherapy from the Mayo Clinic/University of Iowa Specialized Program of Research Excellence (SPORE) Molecular Epidemiology Resource (MER), enrolled between 2002 and 2015 [3, 7] were included in the study and stratified into three groups based on clinical outcome: LR (n = 19), PTR (n = 56), and NR (n = 214; event‐free with ≥ 60 months follow‐up); individuals with an event between 6 and 24 months, or non‐lymphoma‐related event < 6 months were excluded. Median follow‐up of included patients was 8.9 years (LR 10.1, NR 8.9 years).

Whole Exome Sequencing (WES) data were available for 264 cases (LR n = 17, PTR n = 51, NR n = 196), RNA‐seq data were available for 202 (LR n = 14, PTR n = 44, NR n = 144) and copy number alterations were available on 240 cases (LR n = 12, PTR n = 49, NR n = 179). Flowchart of the cohorts provided as Supporting Information S1: Figure S1. Molecular classification was performed as previously described, with detailed methodologies provided in the Supporting Information S1: Material [3, 5].

Clinical characteristics at diagnosis were significantly different between the three groups (Supporting Information S1: Table S1), although they did not distinguish the LR and NR groups. IPI 3‐5 was observed in fewer LR cases than PTR (27% vs. 63%; p < 0.01), and similar to NR cases (28%, p = 0.87). Patients in the LR group also had longer diagnosis‐to‐treatment interval compared to PTR group (median 20 vs. 11 days, IQR 12.5–25 vs. 6.5–21 days, p = 0.02).

The genomic profile of LR cases identified variants previously described in DLBCL (Figure 1A–B). Despite clinical similarities to NR, the LR cohort exhibited a distinct genomic profile enriched in epigenetic and chromatin‐modifying genes, particularly SETD2 and KMT2D mutations (Figure 1C). Similarly, 3p21.31 loss (SETD2) was more frequent in LR (Figure 1D). Other significantly enriched alterations in LR included gains of 8q24.21 (MYC), 18q21.33 (BCL2), and 9p21.3 loss (CDKN2A/B). These six alterations also showed a trend toward higher prevalence in LR when compared to PTR (p < 0.1; Figure 1C–D). Furthermore, when evaluating the genetic classifiers, we noted that the LR cases displayed a lack of DLBClass C1 and LymphGen BN2 cases (Supporting Information S1: Figure S2).

FIGURE 1.

Genomic Alterations in Late Relapse DLBCL. (A) Oncoplot of the top 25 mutated lymphoma driver genes in LR cases. (B) Oncoplot of the top 25 copy number alterations in LR cases. (C) Statistical comparison of alterations in NR versus LR cases (C) and PTR versus LR cases (D) using Fisher's exact test. The 20 most different alterations (based on frequency differences) are displayed. Alterations with p < 0.1 in both comparisons marked as red; p < 0.1 *, p < 0.05 **, p < 0.01 ***.

Gene Set Enrichment Analysis (GSEA) of differentially expressed genes between the LR and NR cohorts revealed enrichment of upregulated pathways in LR related to translation and the B‐cell state S04 and S05 Ecotyper signature (BEcoS4, BEcoS5; Figure 2A–C). When compared to PTR cohort, LR cases demonstrated enrichment of genes within epigenetic regulatory pathways, specifically DNA methylation and chromatin organization (Figure 2D–F). Supporting this finding, RNA‐based molecular classifiers revealed that only the B‐cell state distinguished the LR cohort from NR at baseline (p = 0.01). Notably, the S04 and S05 pre‐plasmablast states encompassed 50% of the LR cases compared to 22% of the NR, supporting the GSEA results (Supporting Information S1: Figure S2).

FIGURE 2.

Gene Set Enrichment Analysis (GSEA) of LR DLBCL. GSEA results for LR versus NR (A–C) and LR versus PTR (D–F). The top row displays the most up‐ and downregulated pathways for each comparison and the middle and lower panels highlight individual pathways upregulated in LR for each comparison.

Effective risk stratification in DLBCL requires a comprehensive understanding of the molecular profiles underlying distinct clinical trajectories, including those of patients who develop LR. While the integration of targeted therapies to specific molecular classifiers represents a potential strategy to improve outcomes in high‐risk subgroups, LR cases frequently retain significant chemosensitivity at the time of relapse and often achieve favorable responses to the current standard of care, yet new therapeutic targets are needed [6]. Multiomic profiling reveals that LR cases have a biological signature characterized by enrichment in unique epigenetic and chromatin‐remodeling alterations which may explain the long‐term persistence of the malignant clone leading to delayed recurrence. SETD2 depletion/mutation leads to decreased recruitment of DNA replication machinery and replication fork instability which may impact response to therapy [8]. Consistent with the broader epigenetic landscape of DLBCL, in which KMT2D mutations occur in approximately 30% of cases [9], KMT2D mutations were markedly enriched in our LR cohort (53%) compared to NR (24%) and PTR (27%). LR may be further be driven by a pre‐plasmablast B‐cell state gene expression signature (S04 and S05), which are ABC‐enriched and associated with the most adverse prognosis [10]. This is supported by our finding that LR cases were enriched for ABC (36%) compared to NR (27%), although not statistically significant. Notably, lack of BN2 subtypes associated with prolonged survival were identified among LR patients, consistent with the low prevalence of this cluster reported in prior studies [6]. In conclusion, our findings identify molecular features associated with LR, indicating that, despite comparable baseline clinical characteristics, LR cases are characterized by an enrichment of unfavorable genetic and molecular alterations at diagnosis compared to NR. These results highlight the importance of molecular profiling for risk stratification beyond conventional clinical parameters. Limitation of this study is the modest number of LR cases; larger cohorts will be necessary to perform multivariate analyses and confirm the independent prognostic value of these factors in predicting LR.

Author Contributions

M.M. designed and conducted analyses, generated data and figures, interpreted results, and wrote the manuscript. J.P.N., J.S.W., A.R.D and Z.X. conducted experiments and interpreted results, together with P.V. who generated figures. M.J.M. and K.W. collected and assembled data. S.M.A., T.M.H., T.E.W., and R.L.K. provided study materials or patients. J.R.C. and A.J.N. planned the project, acquired funding, together with M.Z., P.K., and M.T. who supervised the study, interpreted results, and reviewed the manuscript. All authors read and approved the final version of the manuscript.

Funding

This work was supported in part by National Institutes of Health (NCI/NIH) grant SPORE‐P50 CA97274 (J.R. Cerhan and A.J. Novak; Y. Wang), R01 CA212162‐01A1 (A.J. Novak and J.R. Cerhan), U01 CA195568 (J.R. Cerhan), NU23‐03‐00127 (Czech Ministry of Health; MM), the Predolin Foundation Biobank, and the generosity of the Tanoto Foundation.

Supporting information

Supporting Information S1

Data Availability Statement

The data that support the findings of this study are openly available in dbGaP at https://www.ncbi.nlm.nih.gov/projects/gap/cgi‐bin/study.cgi?study_id=phs003634.v1.p1, reference number phs003634.v1.p1.