Differential Gene Expression in KMT2A::AFF1 Leukemia Is Driven By Enhancer Heterogeneity

Although genetic alterations drive carcinogenesis (PMID: 29439951), they alone cannot account for the diverse phenotypes of cancer cells. Even cancers with the same driver mutation show significant transcriptional heterogeneity and varied responses to therapy (PMID: 32807900). However, the mechanisms underpinning this heterogeneity remain under-explored. Aberrant enhancer activity is a hallmark of many cancers, including KMT2A::AFF1 acute lymphoblastic leukemia (ALL; PMID: 37626123), an aggressive leukemia subtype with a poor prognosis (PMID: 32376390) and a nearly mutationally silent genetic landscape (PMID: 25730765). Despite the low mutational burden, KMT2A-rearranged leukemias exhibit substantial heterogeneity between individuals, providing an ideal model to study transcriptional heterogeneity without the confounding effect of numerous cooperating mutations.

To validate the extent of transcriptional heterogeneity we first compared two patient-derived KMT2A::AFF1 cell lines (RS4;11 and SEM), identifying 2,361 differentially expressed genes via RNA sequencing. Next, we compared enhancer activity using Assay for Transposase-Accessible Chromatin using sequencing (ATAC-seq), identifying 3,342 enhancers with significantly altered activity in these KMT2A::AFF1 ALL cell lines. CRISPR-Cas9-based deletion of representative enhancers confirmed cell line-specific activity on their predicted target genes. Using supervised machine learning, we developed a predictive model that identified the KMT2A-complex as a key factor in enhancer activity differences between cell lines. Our previous work showed that KMT2A::AFF1 binds enhancers and regulates them by recruiting a complex that facilitates transcriptional elongation (PMID: 37626123). In line with this, loss of KMT2A::AFF1 or components of the transcription elongation complex significantly diminishes enhancer-promoter interactions and features of active enhancers, implying that it is essential for maintaining enhancer activity (PMID: 37626123).

To confirm differential enhancer usage in patients, we used single-cell multiomics (RNA+ATAC) on KMT2A::AFF1 patient blasts, identifying substantial transcriptional heterogeneity and 3,162 differentially active putative enhancers. As ATAC-seq identifies genomic boundary elements in addition to enhancers and promoters, we performed detailed epigenetic profiling of nine KMT2A::AFF1 patient samples (four infants and five children) with a low-cell number optimized chromatin-immunoprecipitation protocol. Employing a deep-learning-based strategy, we reduced intrinsic sample-to-sample noise and dramatically improved the stringency of differential enhancer identification. This approach revealed 2,647 differentially active enhancers, with 61 displaying activity unique to a single patient.

We examined whether any unique enhancers were located near prognostically relevant genes and identified putative patient-specific enhancers near MEIS1 and RUNX2. Single-cell expression data revealed elevated MEIS1 and RUNX2 expression in the patient with these unique enhancers. Using Micro-Capture-C (MCC), we confirmed that these enhancers contact the MEIS1 and RUNX2 promoters, implicating these patient specific enhancers in the overexpression of these genes.

Overall, our data suggests that enhancer heterogeneity is highly prevalent in KMT2A::AFF1 ALL and appears to be driven by differential KMT2A-complex binding. We identify differential enhancer activity at key oncogenes such as MEIS1 and RUNX2. Analysis of published microarray data shows that overexpression of either gene is associated with significantly reduced overall survival (PMIDs: 20699438, 19880498, 17312329). Taken together, this indicates that enhancer heterogeneity may significantly contribute to the phenotypic diversity observed between patients in ALL.