Structural variants (SVs) are characteristic hallmarks of cancer genomes. The discovery of recurrent SVs altering protein-coding sequences has greatly improved our knowledge of oncogenesis and provided targets for diagnosis, prognosis, and therapies, especially in hematologic malignancies. Pathogenic non-coding SVs can disrupt higher-order genome architecture, leading to reorganization of chromatin domains to cause gene dysregulation. In particular, 'enhancer hijacking' is recognized as a common mechanism by which non-coding SVs reposition transcriptional enhancers from one chromatin domain to the vicinity of new cognate genes in a topologically distinct domain, causing rewired chromatin interactions to activate oncogene expression. However, due to the lack of tools to assess their effects on gene regulation and chromatin organization, the molecular determinants for the transcriptional output and the phenotypic consequence of enhancer hijacking remain poorly understood.

In this study, we developed a multimodal approach by combining whole genome sequencing, chromatin conformation, and sequence-based deep learning to quantitatively evaluate the transcriptional effects and structural reorganization imposed by SVs in leukemia genomes. We performed whole genome sequencing (WGS) and transcriptomic analyses of 50 primary human leukemia patients and cell lines, encompassing mixed phenotype acute leukemia (MPAL), AML, and T-ALL entities, to identify chromosomal breakpoints and correlate SVs with gene expression alterations, respectively. Next, we conducted sequence-based modeling of chromatin interactions for normal and mutated alleles for each sample using the Orca pipeline (Nat Genet 54, 725) trained with T-ALL and AML Hi-C and H3K27ac ChIP-seq datasets. Our approach revealed hundreds of non-coding SVs with potential functional consequence, and identified recurrent t(5;14) translocations associated with increased TLX3 expression in two cases of MPAL and in a T-ALL cell line. High-resolution chromatin interaction analysis revealed that the TLX3 promoter on chr5 form strong and frequent long-range interactions with enhancers downstream of BCL11B on chr14 in all t(5;14) cases. Moreover, CRISPRi-mediated perturbation of SV-hijacked BCL11B enhancers impaired TLX3 transcription, thus establishing direct evidence that enhancer hijacking regulates TLX3 activation in t(5;14)-associated leukemia.

To determine if repositioning BCL11B enhancers near TLX3 was sufficient for enhancer hijacking to activate TLX3-dependent oncogenic programs, patient-associated t(5;14) was engineered in Jurkat T-ALL cells using CRISPR/Cas9 genome editing. Formation of t(5;14) established de novo enhancer-promoter looping between TLX3 and BCL11B enhancers yet was not sufficient to activate TLX3. Importantly, the TLX3 locus contains a CpG island subject to DNA hypermethylation in Jurkat cells in contrast to the t(5;14)-containing leukemia cells. Moreover, allele-specific DNA methylation at the TLX3 promoter correlated with allele-specific enhancer-promoter interactions and transcription of TLX3 in patient-derived t(5;14) leukemia cells. Treatment with hypomethylating agents led to significant reduction of DNA methylation and over 500-fold TLX3 induction in SV-edited but not parental Jurkat cells, illustrating that loss of DNA methylation is required for enhancer hijacking-mediated TLX3 expression.

Taken together, we developed a multimodal computational strategy to identify non-coding SVs with potential functional consequence in leukemia genomes. We further leverage this new approach to interrogate a recurrent translocation that characterizes a subtype of pediatric acute leukemia. Our results not only provide the direct evidence of enhancer hijacking in t(5;14)-associated leukemia, but also uncover an under-appreciated mechanism whereby the transcriptional output of SV-mediated enhancer hijacking is dependent on the epigenetic state of target gene(s). These studies highlight the importance of the cooperation between genetic alteration and permissive chromatin as a critical determinant of pathogenic SV-mediated oncogenic transcription, with implications for understanding aberrant gene transcription following epigenetic therapies in human leukemia.

No relevant conflicts of interest to declare.

Author notes

*

Asterisk with author names denotes non-ASH members.

Sign in via your Institution