2026 ASH®
Annual Meeting & Exposition December 12-15, 2026 ▪ New Orleans, Louisiana

Paper Details

Abstracts, authors, affiliations, and disclosures.

Saturday, December 12, 2026
04:00 p.m. - 05:30 p.m.
MCCNO - Great Hall BC
Technology and Procedures, Omics Technologies, Pathology, Bioinformatics, Computational Biology, Profiling, Treatment Considerations, Combination Therapy, Biological Therapies, Non-Biological Therapies, Chemotherapy, Artificial Intelligence (AI), Machine Learning, Biological Processes, Genomics, Multi-Systemic Interactions, Diseases, Myeloid Malignancies, Acute Myeloid Malignancies, Research, Fundamental Science, AML, Translational Research, Study Population, Human, Adult, Pediatric
PAPER

379 Microenvironmental drivers of AML treatment resistance revealed by spatial transcriptomics

Barbara W Xiong, BS, Shovik Bandyopadhyay, MD, PhD2, Kyung Jin Ahn, BS3, Rushabh Mehta, BS4, Yong Zhang, MD5, Sanjay Patel, MD, PhD6, Pamela J. Sung, MD PhD7, Martin Carroll, MD8, Kai Tan, PhD9
1 University of Pennsylvania, Philadelphia, Pennsylvania, United States
2 Brigham & Women's Hospital, Boston, Massachusetts, United States, University of Pennsylvania, Philadelphia, Pennsylvania, United States
3 Children's Hospital of Philadelphia, Philadelphia, Pennsylvania, United States
4 University of Pennsylvania, Philadelphia, Pennsylvania, United States
5 University of Pennsylvania, Philadelphia, Pennsylvania, United States
6 Weill Cornell Medicine/Presbyterian Hospital, New York, New York, United States
7 Roswell Park Comprehensive Cancer Center, Buffalo, New York, United States
8 University of Pennsylvania, Philadelphia, Pennsylvania, United States
9 Children's Hospital of Philadelphia, Philadelphia, Pennsylvania, United States, Children's Hospital of Philadelphia, Philadelphia, Pennsylvania, United States, University of Pennsylvania, Philadelphia, Pennsylvania, United States

Introduction

Acute myeloid leukemia (AML) is the most common adult leukemia, with the lowest overall survival. A third of patients are refractory to both 7&3 intensive therapy and venetoclax-hypomethylating agent (Ven/HMA) regimens, with most remaining patients relapsing within a few years. While previous studies have characterized the role of the bone marrow microenvironment (BMM)–particularly mesenchymal stromal cells (MSCs)–in AML progression and treatment resistance, they have not resolved how distinct leukemic cell (LC) phenotypes are spatially related to various BMM cell types. Addressing this gap requires methods that can distinguish malignant cells from phenotypically similar nonmalignant hematopoietic populations while preserving their spatial context. Here, we use the 10X Genomics Xenium spatial transcriptomics platform with mutation-specific probes as part of a custom probe panel to identify NPM1mut LCs in situ and study spatial relationships, comparing pre-therapy to post-therapy. We leverage this approach to characterize LC phenotypic shift and spatial interaction with the BMM in response to therapy.

Methods

We designed a custom Xenium panel with 436 probes for gene expression and 44 probes for mutations, including common NPM1fs cDNA variants. We generated longitudinal Xenium data from 136 samples from 51 NPM1mut AML patients on 7&3 or Ven/HMA regimens at available pre- and post-therapy timepoints, as well as 10 negative lymphoma staging marrow controls (NSM). Co-Detection by Indexing (CODEX) was also performed post-Xenium on the same slide, which quantified expression of select proteins including NPM1c. We identified probe-positive NPM1-mutant LCs which, along with comparator non-malignant cells from control samples, were then used to further impute LC status among remaining cells using a machine learning model. LCs and non-LCs were further manually annotated after clustering and DEG analysis. Ligand-receptor and cellular neighborhood analysis were performed using CellChat and N-Orbit, respectively.

Results

We first leveraged our longitudinal atlas to understand changes in LC phenotypic shifts following 7&3 or Ven/HMA therapy. We identified distinct 9 LC subpopulations based on comparison to previous single cell AML reports (PMID: 40294241) . Of these, erythroid-like LCs were enriched post-treatment regardless of therapy modality, but macrophages-like LC enrichment was specific to 7&3. Further investigation into these erythroid LCs revealed upregulation of OxPhos and fatty acid (FA) metabolism genes in both pre- and post-therapy compared to other LCs and NSM controls. Previous studies have suggested that such upregulation represents a resistance mechanism to Ven/HMA regimens, potentially mediated by MSCs. Therefore, we next analyzed the mesenchymal compartment and spatial relationships between LCs. Broadly, we found distinct LCs partitioned with distinct immune and mesenchymal cells. Macrophage-like LCs shared neighborhoods with adipogenic MSCs (adipo-MSCs), T-cells, and plasma cells after 7&3 in treatment-resistant patients. Erythroid-like LCs were found in neighborhoods with adipo-MSCs in post-treatment patient samples of either treatment regimen. Our ligand-receptor analysis showed this localization is associated with strongly predicted VCAM1-VLA4 and CD36-TSP1 mediated interactions that were significantly increased after treatment. Furthermore, erythroid-like LC proximity to Adipo-MSCs statistically correlated with increased OxPhos and FA metabolic activity, nominating this interaction as a potential target for abrogating Ven/HMA resistance in AML.

Conclusion

We present to our knowledge the largest longitudinal spatial transcriptomic analysis of the AML BMM comparing pre-therapy to post-therapy with either 7&3 or Ven/HMA-based therapy. Analysis shows that both LCs and their BMM constituents respond heterogeneously to the two therapies in this group of NPM1mut AML patients. Importantly, two LC subtypes were particularly enriched. Our study shows a transcriptomic shift towards erythroid lineage in treatment resistant AML cells along with reorganization into OxPhos-promoting neighborhoods. These metabolic features and spatial associations with Adipo-MSCs suggest a potential BMM target for overcoming resistance in NPM1mut AML. These results demonstrate for the first time the physiologic response of AML cells to cytotoxic therapy confirming some in vitro results but demonstrating the heterogeneity of response.

Barbara W Xiong, BS: No relevant conflicts of interest to declare.