Bone Marrow Spatial Multi-Omics Identifies Reduced Interactions between Hematopoietic Stem Cells and Myeloid Cells in Myelodysplastic Syndrome

Introduction: Myelodysplastic syndrome (MDS) is driven by aberrant hematopoietic stem and progenitor cells and, therefore, affects the cellular constitution of bone marrow. While it is known that cell proportions change in MDS, how cell signaling and cell-cell interactions change is poorly understood. Alterations are thought to involve interactions of hematopoietic stem and progenitor cells (HSPCs) with myeloid and lymphoid cells. Understanding these changes is crucial for uncovering the mechanisms underlying MDS pathogenesis and identifying therapeutic targets.

Methods: We collected bone marrow biopsies from 12 MDS patients at a single site. Utilizing Akoya’s spatial multiplex immunohistochemistry method, co-detection by indexing (CODEX), we characterized the bone marrow at single-cell resolution using a panel of 24 markers defining cell types or cell states. CODEX multiplex imaging was performed using the Islet and Pancreas Analysis (IPA) Core supported by the Vanderbilt Diabetes Research Center. Cell types and neighborhoods were annotated based on label-projections from publicly available data, using the Python package MaxFuse and the R package Seurat. We compared cell populations and neighborhoods between samples from 12 MDS patients and from 12 publicly available controls (Bandyopadhyay et al. Cell, 2024). Additionally, we performed single-cell RNA sequencing (scRNAseq) on bone marrow biopsies from these 12 MDS patients and an additional 5 MDS patients. Analysis of scRNAseq data was performed with the R packages SoupX, Seurat, DoubletFinder, and CellChat.

Results: We identified several alterations in cell proportions and cell interactions in MDS that were consistent across spatial and scRNAseq data. First, in the spatial data, we observed an expansion of HSPCs with MDS (p-value = 4.11x10^-6). We also observed a reduction in the myeloid:erythroid ratio in MDS (5:1) compared to normal (8.1:1). Furthermore, analysis of cell neighborhoods in the spatial data highlighted reduction of “HSPC/intermediate myeloid” neighborhoods (p-value = 0.003). Myeloid cells were on average further from HSPCs in MDS (p-value = 0.04). Lymphoid cells were also on average further from HSPCs in MDS (p-value = 0.02). In the scRNAseq data, we similarly identified a reduction in the myeloid:erythroid ratio in MDS (2.7:1) compared to normal (3.1:1). Additionally, in MDS samples, likelihood of signaling between hematopoietic stem cells and myeloid progenitors was reduced, as evaluated by expression of ligand and receptor pairs.

Conclusion: Our findings reveal significantly reduced interactions between HSPCs and myeloid cells in MDS. We found that HSPCs and myeloid cells are farther apart and have reduced likelihood of interacting in MDS. These results provide valuable insights for in situ evaluation of MDS.