Rapid Fire & Poster Presentation 41st Lorne Genome Conference 2020

Identifying oncogene-specific cancer-immune cell interactions in acute myeloid leukaemia from in silico deconvolution of bulk patient bone marrow samples (#103)

Jasmin Straube 1 , Rebecca Austin 2 , Lihui Lin 1 , Matthew Witkowski 2 , Victoria Y. Ling 1 , Iannis Aifantis 2 , Megan Bywater 1 , Steven W. Lane 1 3 4
  1. QIMR Berghofer Medical Research Institute, Brisbane, QLD, Australia
  2. NYU School of Medicine, Department of Pathology, Laura & Isaac Perlmutter Cancer Center, New York City, USA
  3. The Royal Brisbane and Women's Hospital, Brisbane, QLD, Australia
  4. The University of Queensland, Brisbane, QLD, Australia

Acute Myeloid Leukaemia (AML) is a blood cancer with poor long-term survival after standard chemotherapy; hence, there is an urgent need for new therapies. Immunotherapies exploiting immune checkpoint blockade have shown great success in hypermutated cancers but further investigation is needed to determine whether checkpoint inhibition is effective in lowly mutated cancers, such as AML. AML is genetically heterogeneous with recurrent driver mutations and fusion oncogenes defining molecularly distinct subtypes. We have previously demonstrated in murine models that the immune system can control AML progression in an oncogene-dependent fashion. Here, we characterise transcriptional profiles of 1205 published primary patient AML bone marrow (BM) samples classified in 12 different oncogenic subgroups to determine their interaction with the immune system.

First, we deconvoluted transcriptional profiles from patient’s bulk BM to determine immune cell proportions in silico, using CIBERSORT. We found that differences in immune cell proportions are oncogene-dependent. Single-sample Geneset Enrichment Analysis (ssGSEA) on the same data also showed oncogene-specific functional differences in immune cell activation, differentiation and exhaustion. Furthermore, AMLs also exhibited oncogene-dependent expression of mediators of immune evasion, correlating with differences seen in the immune microenvironment.

We then used single-cell RNAseq to assess effects on 1460 immune cells (CD34-CD45+) by an immunogenic AML1-ETO-driven AML compared to 3859 healthy immune cells. We successfully classified single cells by their immune cell type and differentiation state. Compared to healthy, the BM showed depletion of NK-cells with Pseudotime analysis revealing altered T-cell differentiation. T-cells showed hallmarks of activation, through high mRNA expression of IFNG and exhaustion, evident through LAG3, HAVCR2 (TIM-3) and PDCD1 (PD1).

Altogether, these results are consistent with the active engagement of AML with the immune system in an oncogene-dependent manner, a contribution of T-cell exhaustion to immune evasion and disease progression and may help to inform future immunotherapy trials in AML.