Intra-subtype heterogeneity shapes treatment response in KMT2A-rearranged ALL across all age groups

Abstract

KMT2A-rearranged B-cell acute lymphoblastic leukemia (KMT2Ar B-ALL) exhibits significant heterogeneity in age of onset, developmental origins, and clinical outcomes. The interplay of individual factors influencing early treatment response within this high-risk molecular subtype remains poorly elucidated. To identify determinants of early treatment response to induction chemotherapy, we analyzed 465 KMT2Ar B-ALL cases spanning a wide age range (1 month to 89 years) by integrating transcriptomic and genomic profiling with functional drug response and measurable residual disease (MRD) kinetics. We observed a strong inverse correlation between MRD clearance with advancing age (P = 2.1E-04), proximity to early B-cell-precursor developmental state (low maturity score, P = 1.3E-03), and AFF1 as fusion partner (P = 7.0E-04). A multivariable analysis confirmed the strong impact of maturity (P = 0.02) and KMT2A fusion partner (P = 0.03) on MRD clearance, supporting the concept that the cell's developmental state defines therapy response. Gene expression analysis identified cellular traits that relate to MRD clearance (e.g., chromatin organization, immune modulation, and proliferation). This gene expression classifier grouped cases not only by MRD clearance but also by ex vivo sensitivity to induction therapy drugs. Notably, good responders to ex vivo induction drugs were characterized by a higher maturity score (P = 1.8E-03), whereas for less mature KMT2Ar B-ALL cases, response profiles suggested higher Venetoclax sensitivity. Our study provides an integrative framework linking developmental phenotype, fusion partner, and MRD kinetics across the full age spectrum of KMT2Ar B-ALL. These insights may support future risk-adapted strategies and therapeutic targeting, particularly in immature KMT2Ar B-ALL.

Figure 1

Age‐overriding KMT2A‐rearranged B‐cell acute lymphoblastic leukemia (KMT2Ar B‐ALL) patient cohort. (A) Included patients (n = 465) of different age groups (n = 445/465 with available age information: ≥55 n = 79, 18–54 n = 207, 1–17 n = 96, and <1 n = 63) from three cohorts (UKSH n = 288, St. Jude n = 136, and MLL n = 41). (B) Distribution of common KMT2A fusions across age groups (n = 422, 96.6%). (C) Distribution of rare KMT2A fusions (n = 15, 3.4%). (D) Overview of available data layers per patient (see also Supporting Information S1: Table 1) by cohort (gray: internal discovery cohort, brown: St. Jude cohort, and beige: Munich Leukemia Laboratory cohort ¹⁸ ). Available layers are marked by black columns: fusion partner known, age known, RNAseq counts, measurable residual disease (MRD) at two time points available, white blood cell count at diagnosis (WBC) available, virtual karyotypes by single nucleotide polymorphism (SNP) array, DNA capture, drug response profiling (DRP).

Figure 2

Transcriptional maturity score in KMT2A‐rearranged acute lymphoblastic leukemia (KMT2Ar ALL) correlates with age and fusion partner. (A) Gene set enrichment analysis of n = 325 KMT2Ar ALL cases. Enrichment score represents proximity to five healthy B‐cell‐precursor developmental stages (pro‐B, pre‐B‐I, pre‐B‐II‐large, pre‐B‐II‐small, and immature‐B), calculated based on the RNAseq data using ALLCatchR tool. (B) Developmental trajectories of KMT2Ar ALL were condensed into a single maturity score ranging from −1 to 1 by fitting linear regression to the enrichment scores in (A) and calculating the slope. Low maturity scores represent immature transcriptional developmental state and proximity to pro‐B cells, high maturity scores represent more mature (pre‐B‐II, immature‐B) transcriptional developmental states. (C) KMT2Ar samples were mapped to the expression matrix of the human single‐cell B‐cell developmental atlas and colored by maturity score (left). Reference cell states are annotated in the right panel. Samples with low maturity score map to early cell stages (pro‐B‐V(D)J/pre‐pro‐B cycling), samples with higher maturity score map to mature B‐cell stages (pre‐B/immature‐B). (D) Maturity score distribution between age groups within driver fusion groups (AFF1r n = 221, MLLT1r n = 43, MLLT10r n = 14, and MLLT3r n = 28). Multi‐comparison analysis of variance (ANOVA) between age groups within each fusion group reveals significantly different maturity scores between age groups in AFF1r (P = 2.6E−05) and MLLT1r (P = 0.0035) cases.

Figure 3

Measurable residual disease (MRD) clearance is modulated by age, gene fusion, and maturity score. (A) Distribution of age within MRD clearance categories (MRD fast clearance: n = 39, median age: 9.4 years; MRD intermediate clearance: n = 53, median age: 32 years; and MRD slow clearance: n = 122, median age: 37 years). Age is significantly higher in MRD slow clearance patients (multi‐comparison analysis of variance [ANOVA], P = 0.00021). (B) Distribution of maturity scores within MRD‐clearance categories (MRD fast clearance: n = 30; MRD intermediate clearance: n = 29; and MRD slow clearance: n = 38). Maturity scores are significantly higher in MRD fast clearance patients (multi‐comparison ANOVA, P = 0.0013). (C) Multivariable logistic regression results. The model tested the effect of white blood cell count at diagnosis (WBC), maturity score, fusion partner (MLLT3 and MLLT1 compared to AFF1 as reference, respectively), sex, and age group (infant and adult/elderly compared to pediatric as reference). x‐Axis shows log odds ratio and 95% confidence intervals, P‐values are annotated for each factor (WBC: P = 0.00916, maturity score: P = 0.0455, fusion MLLT3: P = 0.315, fusion MLLT1: P = 0.033, sex: P = 0.21, age group infant: P = 0.0892, and age group adult/elderly: P = 0.803). (D) Machine‐learning trained decision tree, trained to predict MRD‐clearance categories based on age, fusion, and maturity score. The highest level split predicts patients with a maturity score >−0.04 to have fast clearance, the second split predicts patients with a maturity score ≤−0.04 and MLLT1r to have fast MRD clearance, the third split predicts patients with maturity scores between >−0.28 and −0.04 into intermediate MRD clearance, and patients with maturity scores ≤−0.28 and no MLLT1r to have slow MRD clearance. (E) Gene expression signature detected from a proportional odds linear regression model grouping patients by MRD clearance. Cluster 1 is enriched for patients with fast MRD clearance (Cluster 1: n = 12/18 [67%] fast, n = 5/18 [28%] intermediate, and n = 1/18 [6%] slow; Cluster 3: n = 1/28 [4%] fast, n = 7/28 [25%] intermediate, and n = 20/28 [71%] slow; chi‐squared P < 0.0001). (F) Resistance score (total score from single sample gene set enrichment analysis using genes upregulated in Cluster 3 as up‐set and genes upregulated in Cluster 1 as down‐set).

Figure 4

Ex vivo drug response to induction therapy drug mirrors slow measurable residual disease (MRD) clearance. Drug response of ex vivo KMT2Ar samples (n = 61). (A) normalized log2 area under the curve (norm.logAUC) for 27 compounds with at least n = 40 cases measured., Top six rows represent induction therapy drugs. Patients were grouped according to MRD clearance (n = 49) and annotated with resistance score from Figure 3F. (B) Circos plot of patients (n = 59) ordered and separated into quartiles by their mean response (norm.logAUC) to induction phase drugs, and annotated with maturity score, fusion partner, age group, and MRD clearance. Quartiles 1 and 2 (Q1/Q2, n = 15, respectively; most sensitive) were enriched for fast MRD clearance, high maturity score, non‐AFF1r, and pediatric age. Quartiles 3 and 4 (Q3/Q4, n = 15, respectively; most resistant) were enriched for slow MRD clearance, low maturity score, AFF1r, and older age (chi‐squared tests: MRD clearance P = 0.003, maturity score group P = 0.064, fusion P = 0.153, and age group P = 0.025, Supporting Information S3: Figure 7). (C) Response (norm.logAUC) to induction therapy drugs was poorer in patients with slow MRD clearance (n = 17; red box) compared to patients with fast MRD clearance (n = 16; green box; mean norm.logAUC across drugs: 0.61 [slow MRD] vs. 0.52 [fast MRD], Wilcox‐test per drug: Asparaginase P = 0.02, Dexamethasone P = 0.0027, Doxorubicin P = 0.041, Vincristine P = 0.012, and Cytarabine and Daunorubicin not significant). (D) Response to induction phase drugs (left) and Venetoclax (right). Patients grouped by maturity score group (cutoffs −0.04 and −0.28 based on cutoffs from trained decision tree (Figure 3D). Patients in the lowest maturity score group (≤−0.28) are significantly more resistant to patients in the highest maturity score group (>−0.04) (Wilcox‐test, P = 0.0018). The effect is inverse for Venetoclax (Wilcox‐test P = 0.18). DRP, drug response profiling.