NicheSphere reveals SPP1⁺ myeloid cells as central hubs of fibrotic remodeling in myeloproliferative neoplasms

Bone marrow fibrosis in myeloproliferative neoplasms arises through reciprocal interactions between mutant hematopoietic cells and fibrosis-associated stromal populations. Using dual lineage tracing, single-cell and multiplet RNA sequencing, spatial proteomics, and NicheSphere, a framework integrating condition-specific co-localization with ligand–receptor inference, we identified Spp1⁺ macrophages as central cellular communication hubs within a “Fibrosis Interacting core” comprising osteoCAR cells, fibroblasts, megakaryocytes, and additional hematopoietic populations. NicheSphere distinguished macrophage–vascular compartments enriched in inflammatory signaling from this Fibrosis Interacting core characterized by TGF-β, NF-κB, and extracellular-matrix programs. CODEX imaging confirmed increased proximity of SPP1⁺ immune cells, activated stromal cells, and megakaryocytes. Genetic deletion of SPP1 demonstrated complementary contributions of stromal- and hematopoietic-derived Spp1 to fibrosis, inflammation, and myeloproliferative features. Human myelofibrosis single-cell analysis identified SPP1⁺ cells as a prominent predicted communication hub, while elevated circulating SPP1 was associated with adverse clinical outcome. Together, these findings establish SPP1-centered multicellular communication as a conserved driver of myelofibrosis and a therapeutic vulnerability of the fibrotic niche.

This repository contains the following datasets:

• Physically Interacting Cells sequencing (PIC-seq) - /data/PICseq/
• CODEX - /data/CODEX/
• Human myelofibrosis (MF) scRNA-seq JAK2 subset from [https://doi.org/10.5281/zenodo.17210242] - /data/MF/

and the code used to produce the presented results (also available in GitHub: https://github.com/CostaLab/NicheSphere_paper_analysis ).

Final data files are in the "final_data" folders:

/data/PICseq/final_data/

• adata_singlets_final.h5ad

             Singlets data with final cell type names in the metadata "annotation" column (processing done in Python in PIC_data.ipynb) 

• adata_singlets_final_imp.h5ad

            (MAGIC) Imputated singlets data with final cell type names in the metadata "annotation" column (processing done in Python in PIC_data.ipynb)

• adata_doublets_final.h5ad

            Doublets data with final doublet cluster annotation in in the metadata "max_pair" column (processing done in Python in PIC_data.ipynb) 

• adata_doublets_final.rds

            Sceasy converted adata_doublets_final.h5ad (employed for ridge plots with doublet cluster names)

• adata_doublets_final_imp.h5ad

            (MAGIC) Imputated doublets data with final doublet cluster annotation in the metadata "max_pair" column (processing done in Python in PIC_data.ipynb) 

Phenotype (TPO or EV) is indicated in the metadata "stage" columns for these datasets.

/data/CODEX/final_data/

• CODEX_run45_anndata_final.h5ad

             Final CODEX dataset with final cell type names in the metadata "clusters_named" column, ohenotype (Thpo or EV) is in the "condition" column and sample and run are indicated in the columns with those names.

• zarr/

            Folder containing zarr files with cell shapes for each sample

/data/MF/

• MF_pre_treatment_Amplicon_object.h5ad

           JAK2 subset of human MF scRNA-seq dataset from [https://doi.org/10.5281/zenodo.17210242]

• MF_subcluster_subset_metadata.csv

            Metadata related to JAK2 scRNA-seq human data

• MF_JAK2_LR_data_final.Rds

            Resulting data from CrossTalkeR cell - cell communication analysis on JAK2 scRNA-seq human data

Intermediate data files which were used in the scripts can be found in the "intermediate_steps" folders ( eg: /data/PICseq/intermediate_steps/ ).

Analysis code is in the "code" folder and resulting data files and images are located in folders "res_data" and "figures" respectively.

Database related files are in the "csv" folder.