Deep Multi-Omics Profiling in Cytogenetically Poor-Risk AML

Introduction: Acute myeloid leukemia (AML) patients with adverse cytogenetics have particularly poor outcomes with median survival of only 7.6 months. The treatment of these patients has lagged behind other AML subtypes, which have seen substantial improvement with novel therapeutic approaches. We reasoned that deeper phenotyping of poor-risk primary AML cases samples may offer new insights into the biology of these leukemias and allow the development of novel treatments. Here we present the most comprehensive multi-omic approach reported to date in AML comprising genomics, transcriptomics, (phosphor)proteomics, drug screening (>500 compounds) and CyTOF analyses from 57 cytogenetically poor-risk AML primary samples.

Aim: Application of multi-omics and integrative approaches to decipher the complexities of cytogenetically poor-risk AML

Methods: We obtained multi-omic profiles for a retrospective series of 57 untreated primary AML samples from patients aged between 17 and 84 years (average 52.5) treated at Barts Health NHS Trust designated as poor risk disease based on their pre-treatment cytogenetics. Altogether the series included cases with complex karyotype (20: 35%), -7/del(7) (15: 26.3%), KMT2A rearrangements (not including t(9;11)) (12: 21%), t(6;9) (4: 7%) and 6 with other poor-risk karyotypes (inv(3), -5/del(5), t(3;12)/+8 or -17/del(17)) (10.5%). Multi-omic experiments included whole genome sequencing (WGS, in 33 cases where matching germline material was available, 60X for tumour and 30X for germ-line controls), targeted deep sequencing of 54 myeloid loci, total RNA-seq (100 million reads per bulk sample), mass spectrometry proteomics and phosphoproteomics (with >6,000 proteins and > 25,000 phosphorylation sites detected and quantified), mass cytometry (CyTOF, 39 markers) and in vitro drug screening (ranging from 200-500 approved or investigational compounds).

Results: We first sought to assess the quality of the omic data and its potential as a discovery tool. Using FusionCatcher, 94 unique in-frame fusion genes were identified, including the 12 KMT2A rearrangements and 4 t(6;9)/DEK-NUP214 characterized by karyotype. WGS analysis was used to resolve the corresponding structural variant, such as a novel intra chromosomal deletion of 8.7kb on chromosome 19 leading to a COX6B1-UPK1A fusion transcript. Exploring how chromosomal aberrations may impact gene expression, we detected a set of genes located in chromosome 7 that were strikingly overexpressed in monosomy 7 samples, including ABCB1, a member of the superfamily of ATP-binding cassette transporters. WGS data was also used to explore non-coding mutations in our series of patients, resolving novel variants in the regulatory regions of GNAS, ETV6 or RASA3 that impact downstream expression as evidenced by luciferase assays. The integration of WGS and RNA-seq allowed detailed exploration of allele-specific expression (ASE) by quantifying in individual transcriptomes, the relative reads from different alleles as defined by differential SNP genotypes at the genomic level. 312 protein-coding genes showed ASE in >20% of samples, including GATA2, PBX2, PBX3, HOXB genes and other RNA binding proteins involved in RNA-splicing and protein synthesis. Finally, we used our multi-omic data to uncover patient-specific drug targets and introduce a biomarker toolkit (genomic, (phospho)proteomics and/or CyTOF signatures) that indicate the most appropriate therapy for an individual or group of patients. For example, we validated TP53 WT status as a determinant of response to MDM2 inhibitors (AMG-232, idasanutlin, SAR405838 and NVP-CGM097) and we found that within the TP53-WT group of patients, good-responders have a significantly lower expression of TP53 pathway genes at diagnosis compared to non-responders. A resultant 14-gene expression signature identified TP53-WT patients that are sensitive to MDM2 inhibitors in larger series of AML cases from all cytogenetic groups.

Conclusion: Altogether, these findings demonstrate the feasibility of simultaneously generating multi-omics data from several different platforms in AML primary samples and highlights that integrative analysis will increase our understanding of the biology of the disease and its therapeutic vulnerabilities.