Introduction Acute myeloid leukemia (AML) with mutations in TP53 (TP53-Mut AML) is a high-risk disease with a dismal prognosis. However, the molecular underpinnings of poor clinical outcome are incompletely understood. We have previously shown baseline enrichment in interferon-γ signaling and high expression of immune checkpoints and markers of immune senescence in diagnostic TP53-Mut AML samples, suggesting perturbations in immune signaling. The role of immune dysfunction in AML therapy resistance and overall survival remains unknown. We subsequently performed spatially resolved whole transcriptome analysis on TP53-Mut and -wild type (WT) AML bone marrow (BM) biopsies collected at diagnosis and over the course of treatment with hypomethylating agent (HMA) and BCL2 inhibitor, venetoclax (Ven).

Methods We obtained formalin fixed paraffin embedded BM biopsies from 24 TP53-Mut and 9 TP53-WT patients (5 IDH1/2-Mut, 4 FLT3-ITD-Mut) at the time of diagnosis (n=33) and serially post-HMA/Ven (n=87). These 3 AML subtypes were selected for their prognostic impact on response to HMA/Ven (ELN 2024). For spatial transcriptomics, regions of interest (ROIs) containing at least 100 nuclei (Cyto13 staining) were selected based on immunofluorescence staining for immature myeloid markers (CD117/CD123), p53 as a surrogate for TP53-Mut cells, and T-cells (CD3). ROIs were manually annotated for overall cellularity and percent blasts vs T-cells. We analyzed results from 536 ROIs, including 165 from TP53-WT and 371 from TP53-Mut AML. By treatment status, we obtained 176 diagnostic and 360 post-treatment ROIs. Spatial transcriptomics was performed using the Whole Transcriptome Atlas panel on the GeoMx Digital Spatial Profiling platform. Data was QC-ed using the standR package in R and comparison of gene expression was done using linear modeling in limma to adjust for random effects (i.e., sampling multiple ROIs from a single tissue section).

Results We analyzed differential gene expression and pathway enrichment of ROIs based on TP53 mutation status, overall survival (OS), response to HMA/Ven, TP53-WT co-mutations, and CD3 or myeloid cell high vs low (median split). Upon comparing diagnostic TP53-Mut vs -WT AML by gene set enrichment analysis (GSEA), the most significant difference (FDR < 0.05) in terms of Hallmark pathways was in reactive oxygen species. Comparison of diagnostic TP53-Mut AML ROIs by OS demonstrated significant interferon-α and -γ responses in patients with <6 months (m) vs >12m OS. Diagnostic TP53-Mut AML ROIs from primary refractory HMA/Ven patients had significant enrichment in heme metabolism compared to patients who initially responded to therapy who had significant enrichment for KRAS signaling. Notably, FLT3-ITD-Mut vs IDH1/2-Mut cases demonstrated multi-pathway inflammatory responses involving interferon, TNFα, IL2 and IL6 signaling as well as KRAS signaling.

To uncover transcriptional similarities among ROIs, we next performed clustering analysis using the Seurat suite in R. TP53-Mut ROIs separated into 6 clusters (Louvain algorithm), which correlated with response to therapy. Cluster 5 was present in diagnostic primary refractory samples and enriched post-therapy, suggesting this cluster may be associated with chemotherapy resistance. Hallmark-based GSEA of this cluster identified a significant enrichment in pathways involving inflammation and cytokine signaling (particularly IL2/6 via STAT3/5, respectively) as well as metabolism [oxidative phosphorylation (OXPHOS) and cholesterol homeostasis]. These pathways were also enriched in diagnostic ROIs from patients with an OS <6m, but not in patients with an OS >6m. This diversity across TP53-Mut samples led us to hypothesize that there is a subset of TP53-Mut AML with short OS (<6m) that may define the genes that best correlate with therapy resistance. In fact, expression analysis of ROIs from diagnostic TP53-Mut ROIs comparing <6m vs >12m, followed by LASSO feature selection, identified a 56 gene signature enriched in short survival TP53-Mut disease.

Conclusions Our spatial transcriptomic profiling of TP53-Mut AML reveals distinct immune and metabolic signatures associated with poor response to HMA/Ven and inferior survival. These findings highlight intratumoral heterogeneity within and across TP53-Mut AML patients and identify candidate pathways that may underly therapy resistance and inform future risk stratification and targeted interventions.

This content is only available as a PDF.
2025
Sign in via your Institution