MSC Reprogramming and Aberrant Inflammatory Signaling in Myelofibrosis: Insights from Single-Cell RNA-Sequencing of Trephine Bone Marrow Cells

Myelofibrosis (MF) results from mutation-induced activation of JAK/STAT signaling in hematopoietic stem, progenitor and mature cells, and features a progressive, fibrosing stromal reaction, splenomegaly, constitutional symptoms and reduced life expectancy. Identification of disrupted cellular pathways within bone marrow (BM) is pivotal for therapeutic breakthrough, but fibrosis severely limits access to stromal elements for single-cell analysis. To overcome this, we utilized collagenase digestion of fresh BM trephine biopsy material surplus to diagnostic needs and profiled isolated cells using scRNAseq.

9 samples were obtained from 8 patients with MF (4, JAK inhibitor naïve; 4, on ruxolitinib); 4 samples were from 4 patients with MF 4, 10, 12 & 12 months after allogeneic transplant; and 7 samples (termed “normal controls”) were from 6 patients who had received an allograft for myelodysplasia or acute leukemia, and 1 who had received CAR-T cells for B-ALL. Normal controls exhibited normal or near-normal blood counts and lacked BM fibrosis. Each sample was subjected to collagenase digestion and flow sorted to isolate non-erythroid hematopoietic (CD45+GlyA-) and microenvironmental (CD45+GlyA-CD71-) cells for single cell RNA sequencing and downstream comparative analysis.

105,231 transcriptomes met quality control criteria (median 4975/sample, range 1036-13277, no significant difference between sample categories) and 43 transcriptional clusters were identified. Compositional analysis identified the non-hematopoietic stromal cell fraction as harboring the biggest differences with, surprisingly, a proportional relative depletion of mesenchymal stromal cells (MSCs) and enrichment of endothelial cells (ECs) in MF vs. normal controls. While transcriptomes of ECs did not differ substantially, there were extensive transcriptional differences between MSCs recovered from MF vs. normal control samples. Subclustering revealed two subsets, MSC1 & MSC2 asymmetrically distributed between MF and normal controls. Intriguingly, MSCs recovered from MF patients after allograft occupied an intermediate position in principal component space indicating that reversion of disease-related MSC transcriptional reprogramming post-allograft occurs over a timescale of months.

The reprogrammed transcriptional state of MF MSCs is defined by reduced expression of hematopoietic support factor genes (eg, KITLG, CXCL12, ANGPTL2, NCAM2, CSF1) and increased expression of extracellular matrix (ECM) genes (eg, FN1, COL1A2, COL3A1, DCN, SPARC), features which suggest an explanation for the increased frequency of graft failure following allograft in MF versus other hematologic malignancies. Single-Cell rEgulatory Network Inference and Clustering (SCENIC) indicated transcription factors with either increased (eg, NR2F2, RUNX2, FOXP2) or decreased (eg, CEBPA, FOXC1, and PPARG) transcriptional activity as candidate master transcriptional regulators of the observed MSC reprogramming in MF. Transcriptional change in MSCs correlated strongly with histopathological assessment of BM reticulin.

Within the hematopoietic fraction, and agnostic of JAK inhibitor treatment status, comparison of MF vs. control cells unexpectedly revealed significant upregulation of interferon response genes, especially in CD14⁺ monocytes, but also in neutrophils and their progenitors, centred on expression of interferon-stimulated gene factor 3 (ISGF3) genes IRF9 and STAT1/2. Comparison of MF samples according to ruxolitinib treatment status indicated that ruxolitinib induced downregulation of an NFκB-centered inflammatory program, particularly in neutrophils.

Finally, ligand-receptor analysis using NicheNet identified TGFB1, IL1B, and TNF as top predicted ligands for induction of the MF signature in MSCs. While myeloid cells were top producers of these factors, their expression did not vary between MF vs. controls, suggesting other mechanisms, such as prolonged retention of ligand in the ECM, are important.

In summary, our comparative scRNA analysis of BM cells from MF patients and controls – the largest reported to date – reveals (i) transcriptional reprogramming of MSCs leading to reduced expression of hematopoietic support factor and increased expression of ECM genes; and (ii) increased expression of interferon and NFkB inflammatory pathways in myeloid cells.