Impact of KMT2A Rearrangement on Peripheral T‑Cell Lymphoma, Not Otherwise Specified, and Angioimmunoblastic T‑Cell Lymphoma

1. Introduction

Peripheral T-cell lymphomas (PTCL) account for 5–20% of non-Hodgkin lymphomas and are relatively more frequent in Asia. In Asia, PTCL, not otherwise specified (PTCL-NOS) and angioimmunoblastic T-cell lymphoma (AITL) represent the second and third most common PTCL subtypes [1]. Histologically, they partly overlap, and diagnostic reproducibility is imperfect even among experts [2]. The 2016 WHO revision and subsequent 5th edition emphasize the concept of a T-follicular helper (TFH) lineage for AITL and a subset of PTCL-NOS, defined by variable expression of PD-1, ICOS, CXCR5, CD10, and/or BCL6. These lymphomas are usually accompanied by a distinctive microenvironment with arborizing high-endothelial venules, follicular dendritic cell meshwork, and EBV-positive B-immunoblasts [3,4]. The TFH framework is supported genomically by frequent mutations in TET2 and DNMT3A, the IDH2 R172 hotspot, and the RHOA G17V switch, often co-occurring and tightly linked to TFH differentiation [5,6,7,8]. Patients with TFH lymphomas have shown sustained response to hypomethylating agents and proximal TCR pathway inhibition, whereas non-TFH PTCL-NOS exhibit distinct dependencies [9,10]. These genetic differences in tumors may impact chemotherapy sensitivity.

This study aimed to explore the contribution of KMT2A rearrangements (KMT2A-r) to nodal PTCL biology, which remains poorly defined. Although KMT2A-r is a founding lesion in acute leukemias, mechanistic and clinical data in mature T-cell neoplasms are scarce [11]. Using next-generation sequencing (NGS), we conducted a bi-center analysis to identify clinicopathologic and genomic features across AITL and PTCL-NOS and assess the prognostic impact of KMT2A-r and treatment options in first-line chemotherapy.

2. Materials and Methods

2.1. Patient Selection

We conducted a retrospective cohort study of patients diagnosed with de novo AITL or PTCL-NOS between January 2021 and September 2024 at Seoul St. Mary’s Hospital and Yeouido St. Mary’s Hospital. All biopsies were reviewed by hematopathologists with expertise in T-cell lymphomas. This study was approved by the Institutional Review Board and Ethics Committee of the Catholic Medical Center, South Korea (XC25RIDI0050). The requirement for informed consent was waived because only anonymized archival material and routine clinical data were used.

2.2. NGS

Targeted NGS was conducted using the QIAseq Pan-cancer Multimodal Panel (Qiagen, Hilden, Germany), which assesses DNA aberrations spanning 523 cancer-associated genes. At diagnosis, RNA and DNA were extracted from Formalin-fixed, paraffin-embedded (FFPE) tissue blocks using the QIAamp DNA FFPE Tissue Kit (Qiagen, Hilden, Germany) according to the manufacturer’s protocol. Input DNA quantity and integrity were evaluated using the QIAseq DNA QuantiMIZE Array, ensuring that the material met the requirements for downstream library construction.

Library construction was performed according to the QIAseq Multimodal Panel HT Handbook (Qiagen). The workflow comprised enzymatic DNA fragmentation, end-repair, A-tailing (adenine overhang addition), adapter ligation, and polymerase chain reaction (PCR) amplification. Unique molecular identifiers were incorporated during library build to minimize PCR duplicates and sequencing artifacts, thereby improving the accuracy of variant detection. Libraries were sample-indexed, quantified using quantitative PCR, and quality-checked on an Agilent Bioanalyzer to confirm fragment-size distribution and effective adapter trimming.

Sequencing was performed on Illumina instruments (Illumina, San Diego, CA, USA) using Qiagen-supplied custom sequencing primers. Primary data processing and variant calling were executed in the QIAGEN CLC Genomics Workbench, and downstream variant annotation and clinical interpretation were performed using QIAGEN Clinical Insight Interpret. A comprehensive list of all molecular alterations detected in the cohort is provided in Supplementary Table S1.

2.3. Therapy and Response Assessment

First-line therapy was selected at the physician’s discretion and included CHOP/CHOEP, ProMACE-CytaBOM, or brentuximab vedotin (BV) plus cyclophosphamide, doxorubicin, and prednisone (CHP) [12,13,14]. Regimen selection was done based on age, comorbidity, CD30 expression, and performance status. Responses were evaluated according to the Lugano 2014 criteria. Consolidation with autologous stem-cell transplantation (ASCT) was considered for fit patients attaining complete remission (CR). Staging was consistently evaluated using the Lugano classification and Deauville score. CR was defined as the disappearance of target lesions on computed tomography with normalization of 18F-fluorodeoxyglucose positron emission tomography-computed tomography uptake at all sites (Deauville score, 1–3) [15]. BuMelTT protocols (busulfan, melphalan, and thiotepa) were initiated as previously described [16].

Survival endpoints were overall survival (OS) and progression-free survival (PFS). Curves were generated using Kaplan–Meier estimates with log-rank tests. Univariate Cox models were used to screen clinical and genomic variables. Owing to sample size, multivariable modeling was limited to prespecified factors (age, Eastern Cooperative Oncology Group [ECOG], lactate dehydrogenase [LDH], marrow involvement, histology, KMT2A-r). Pearson correlation matrices were constructed to evaluate variable co-movement and visualized using heatmaps. Network analysis was performed using the igraph package, with absolute correlation as edge weights; node degree and betweenness were used to identify hubs [17]. A decision tree (package named rpart) based on binary features (1=present, 0=absent) was used to distinguish AITL from PTCL-NOS. To reveal multi-gene rules in this small cohort, we report the maximal, unpruned tree (very low cp; reduced minsplit/minbucket; increased maxdepth), showing node counts and class probabilities [18]. All analyses were performed using R version 4.2.3 software (R Core Team, Vienna, Austria), and two-sided P-values <0.05 were considered statistically significant.

3. Results

3.1. Patient Cohort, Baseline Features, and Survival Outcomes

A total of 37 patients were included in the study (AITL n=14; PTCL-NOS n=23). The median age was 61 years (interquartile range: 54–69), and 18 (48.6%) patients were women. Most patients had advanced-stage disease (stage III–IV, 91.9%), and 40.5% had bone-marrow involvement. LDH levels were elevated above the upper limit of normal in 62.2% of cases. TFH markers, including CD10, BCL6, and PD-1, were significantly more frequent in AITL than in PTCL-NOS. Additionally, RHOA mutations were enriched in AITL (95.7% vs. 57.1%; P=0.014). KMT2A-r was present in nine patients (9/37, 24.3%), without strong enrichment based on histology. Additional baseline characteristics are presented in Table 1.

At a median follow-up of 24 months among survivors, 2-year OS was higher in AITL versus PTCL-NOS (70.7% vs. 38.8%; P=0.040), whereas PFS was similar. BV-CHP did not significantly improve OS or PFS in the overall population, and differences in CD30 did not affect treatment outcomes. This is consistent with uncertainty regarding the benefit of BV-CHP in PTCL subtypes except for anaplastic large cell lymphoma [19]. Patients undergoing ASCT consolidation had longer OS and PFS than those managed without ASCT (Figure 1). Although improved survival outcomes with ASCT have been consistently reported, careful interpretation is required owing to the risk of selection bias favoring chemotherapy responders [20,21].

3.2. Impact of Genetic Alteration on Survival Outcomes

Across the cohort, ATR mutations were the most frequent (43.2%), followed by KMT2A-r, RHOA, DNMT3A, and IDH2 (Figure 2A). Univariate Cox analyses revealed that LDH elevation (hazard ratio [HR]: 3.10, 95% confidence interval [CI]: 1.27–7.58), bone marrow involvement (HR: 2.98, 95% CI: 1.25–7.08), and KMT2A-r (HR: 2.56, 95% CI: 1.02–6.45) predicted shorter PFS. DNMT3A mutation, ECOG≥2, LDH elevation trended toward inferior OS (Table 2 and Table S2). DNMT3A mutations often correlated with higher ECOG and older age, both of which contribute to poorer OS. This finding aligns with prior reports that clinical indices (e.g., ECOG, LDH) affect survival outcomes across PTCL subtypes [4]. Median PFS was 5.9 months in patients with KMT2A-r, compared to 12.5 months in those without KMT2A-r (log-rank P=0.039). Figure 2B and 2C illustrate survival according to KMT2A-r and treatment regimen. Among the six patients with KMT2A-r who received BV-CHP, median PFS was not reached at 12 months, compared with 4.8 months (95% CI: 2.1–7.5) in three patients with KMT2A-r treated with other regimens (P=0.011); however, numbers are small, with wide confidence intervals. Baseline CD30 levels did not differ substantially between the KMT2A-r and non-KMT2A-r groups.

3.3. Interrelationship Among the Molecular Landscape

Heatmap analysis revealed a cluster linking KMT2A-r with elevated LDH (correlation coefficient=0.41) and bone marrow involvement (correlation coefficient=0.36), whereas the canonical TFH cluster (RHOA, IDH2, TET2/DNMT3A) tended to anti-correlate with KMT2A-r. In the network view, KMT2A and ATR emerged as high-betweenness nodes bridging a DNA-damage repair motif (ATR–MLH1) and receptor tyrosine-kinase signaling (KIT), suggestive of crossed stress and growth cues. The decision tree split first on RHOA (1→AITL), then used EBV-encoded RNA in situ hybridization and EZH2 for refinement. Within the RHOA=0/EZH2=0 branch, KMT2A=1 predicted PTCL classification, indicating that KMT2A is informative, particularly when TFH signals are absent (Figure 3).

4. Discussion

This two-center study integrated clinicopathologic and genomic data in patients with PTCL-NOS or AITL to examine the impact of KMT2A-r in nodal PTCL. Our findings align with established TFH biology in AITL and its overlap with PTCL-NOS. Sequencing studies have shown that RHOA mutations frequently co-occur with mutations in TET2 (and often DNMT3A), and that IDH2 mutations provide a convergent methylation program; together these lesions drive TFH differentiation and the characteristic microenvironment [22]. The Human Pathology multicenter study further demonstrated that increasing numbers of positive TFH markers track with AITL-like histology and that PTCL-TFH overlaps with, but is not identical to, AITL [23]. Within that framework, our data place KMT2A-r outside the TFH cluster and closer to non-TFH PTCL-NOS behavior—consistent with shorter PFS and higher tumor-burden markers.

KMT2A rearrangements have been sparsely investigated in PTCLs. In experimental models, enforced expression of the KMT2A–AF9 fusion is sufficient to induce leukemia, highlighting the strong oncogenic potential of KMT2A fusion proteins [24]. In our cohort, KMT2A-rearranged cases correlated with elevated LDH and bone-marrow involvement, features that typically accompany advanced disease. Moreover, correlation-network analysis revealed that KMT2A alterations frequently co-occurred with other genetic events (for example, ATR), consistent with a heavier overall genomic alteration burden.

Although ASCT in first remission was associated with superior survival in our cohort, a major practical challenge is that durable CR is not reliably achieved with current induction regimens. Therefore, prolonging the first PFS—by selecting front-line chemotherapy that delivers rapid, deep, and sustained disease control—becomes the key determinant of long-term outcome.

BV is an antibody–drug conjugate that targets CD30 and delivers the microtubule poison Monomethyl auristatin E after receptor-mediated internalization, culminating in mitotic arrest and apoptosis. In ECHELON-2, adding BV to CHP significantly improved OS in CD30-positive PTCL; however, the trial was dominated by Anaplastic large cell lymphoma (ALCL), leaving the effect size in non-ALCL subtypes (AITL and PTCL-NOS) less certain. Moreover, across non-ALCL cohorts the predictive value of CD30 intensity has been inconsistent, and CD30 negativity does not invariably preclude benefit [13,25]. In our cohort, survival outcomes did not differ by CD30 status, reinforcing the view that CD30 expression—while biologically relevant to the mechanism of BV—may be an imperfect surrogate for clinical benefit in AITL or PTCL-NOS. The positive signal for BV-CHP in KMT2A-r cases—although exploratory and sample-limited—supports the hypothesis that microtubule-disrupting payloads may be particularly effective when the DNA damage response (DDR) stress is high [26].

Regarding future treatment strategies, menin inhibitors such as revumenib, which is currently used for the treatment of KMT2A-rearranged leukemias, may be a potential option [27]. From a biology-driven standpoint, 5-azacitidine—alone or combined with romidepsin—has shown activity in TFH-derived nodal PTCL (AITL and PTCL-NOS with a TFH phenotype), in which epigenetic lesions involving DNMT3A and TET2 are frequent [9,28,29]. Given its comparatively low-intensity profile, this epigenetic approach is a rational option for older, poor-performance patients with DNMT3A-mutated PTCL-NOS. In contrast, in our series, KMT2A-rearranged cases clustered away from the canonical TFH program, suggesting limited sensitivity to hypomethylating or histone deacetylase inhibitor combinations and a need for alternative strategies.

This study has certain limitations, including the small sample size, retrospective design, heterogeneous frontline regimens, and the lack of centralized pathologic review for all cases. Additionally, NGS was uniform but targeted; thus, structural variants outside the panel’s scope may have been under-captured. Moreover, our exploratory network and decision-tree analyses illustrate associations rather than causation. Nonetheless, the integration of molecular profiling, immunohistochemistry, and outcomes provides a coherent hypothesis: KMT2A-r identifies a clinically aggressive, TFH-sparse subset of nodal PTCL in which BV-CHP and ASCT may attenuate risk.

Prospective, multicenter validation is needed to refine the prognostic impact of KMT2A-r, delineate partner genes and breakpoint biology, and evaluate tailored regimens, while recognizing that lineage-specific context will determine translatability.

5. Conclusions

KMT2A-r is associated with adverse PFS and clinicogenomic features of aggressive disease in nodal PTCL. It lies largely outside the TFH program and shows connectivity to DDR and receptor tyrosine kinase modules in systems analyses. BV-CHP and ASCT may mitigate risk in selected patients. Larger studies are warranted to validate KMT2A-r as a biomarker and clarify its therapeutic implications.