Spatiotemporal evolution and ecological dynamics of acute myeloid leukemia upon chemotherapy

University Of Pennsylvania, Philadelphia PA

Investigators

Abstract

ABSTRACT Acute myeloid leukemia (AML) is the most common and fatal form of adult leukemia, with less than a quarter of patients surviving past 5 years. Outcomes are especially poor in older patients, who are not candidates for traditional “7+3” chemotherapy. They are also often refractory to or relapse on newer therapies such as BCL-2 inhibitor Venetoclax and hypomethylating agent Azacitidine. Patient xenograft and mouse models have shown that the bone marrow (BM) microenvironment plays an important role in chemoresistance and relapse. The recent emergence of spatially-resolved proteomic and transcriptomic technologies opens possibilities to examine microenvironmental spatial relationships and cellular interactions at high resolution. Examination of the AML BM microenvironment with these technologies, particularly in the context of chemotherapy, has yet to be comprehensively explored. Previous work in our lab has included using CO-detection by InDEXing (CODEX), a multiplexed immunofluorescence spatial profiling technology, to map the biogeography of the normal BM environment, as well as a few NPM1 mutant chemoresistant AML samples. We will extend this effort to a cohort of 30 AML patients to examine spatial patterns of the BM microenvironment to reveal its role in AML chemoresistance and relapse. In addition, our lab has previously developed CytoCommunity, a method for detection of tissue cellular neighborhoods, or spatial regions of distinct and homogeneous cell type composition. These neighborhoods may indicate cell types organizing to form functional niches. Several other tools have also been developed to perform neighborhood detection. However, none yet can integrate neighborhood analyses across multiple samples to assess changes over time or in response to treatment. We will develop methods for tracing neighborhood changes over time and across conditions and apply them to our spatial analysis of AML. We will examine how AML cancer cells and other elements of the microenvironment spatially reassort in response to various treatments through analysis of neighborhood evolution and other spatial statistical methods. This will include subclone-level spatial analysis to identify how intrinsic (genetic) mechanisms of resistance correlate and interact with extrinsic ones. Finally, we will use CODEX-generated cell type labels to train a computer vision model to automatically annotate cell types on co-registered H&E images. This will facilitate automatic cell type labeling on the vast supply of H&E images without corresponding CODEX data, which will allow us to expand our spatial analysis to a much broader cohort. Overall, our project aims to generate biological insights into the microenvironment’s role in AML chemoresistance and relapse which will facilitate the improvement of treatment strategies and outcomes. In addition, we will develop neighborhood evolution and automatic cell type annotation methods that will enhance future research in spatial analysis of the microenvironment in leukemia and cancer more broadly.