Single Cell Transcriptomic Analysis Reveals Dysregulated Mesenchymal Stem Cells Linked to Disease Progression in Myelodysplastic/Myeloproliferative Neoplasms

Background

Myelodysplasia/ myeloproliferative neoplasms (MDS/MPN) are driven by heterogenous mutations involving splicing complexes and epigenetic regulator pathways, and carry an inherent risk of leukemic transformation. While leukemic stem cells harbor driver mutations, mesenchymal stromal cells (MSC) in the bone marrow (BM) microenvironment have been implicated in influencing the emergence and selection of leukemic clones and disease progression. However, specific transcriptomic signals of MSC at different stages of MDS/MPN disease progression remain unclear. We hypothesize that intrinsic differences in MSC stemness and differentiation at baseline influence the disease course.

Methods

Baseline clinical variables and BM samples from MDS/MPN patients treated at Singapore General Hospital were obtained. MSC were derived in MesenCult-ACF media (StemCell Technologies), and immunophenotypic characterization was done after 1-2 passages. RNA extracted from MSC was subjected to single-cell RNA sequencing (scRNA-Seq) (10X Genomics). Longitudinal BM MSC samples from a pair of MDS/MPN patients with rapid leukemic transformation in less than 2 years (“non-responder”) versus that with stable disease on 5-azacytidine (5-aza) treatment (“responder”) were analyzed. Gene expression was quantified using CellRanger against the hg38 reference transcriptome. Expression from different samples was integrated using reciprocal principal component analysis (RPCA), and analysis was performed in R with a cluster resolution of 1.2. Statistical significance of association between categorical variables was assessed using the chi-square test without Yates correction. Gene modules were inferred by hierarchical clustering on the correlation matrix of the top two thousand most highly variable genes. Biological process over-representation analysis was performed using PANTHER, with the highly variable genes as the background gene list. Statistical testing was performed with the Fisher Exact Test, with p-value correction performed using False Discovery Rate.

Results

High-resolution clustering of single-cell RNA-seq data identified 19 cell populations. Four clusters showed high expression of markers, including TWIST1 and PODXL, with stem cell-like MSC (scMSC) characteristics. Although the composition of scMSC at diagnosis was comparable between the responder and the non-responder (11.53% vs.10.32%), a marked depletion in sc-MSC was observed in the non-responder (10.32% to 1.95%) as compared to that of the responder (11.53% to 12.74%) after 5-Aza treatment.

Two gene modules comprising 545 genes associated with scMSC were identified among a total of 22 gene modules derived from the highly variable genes. Over-representation analysis demonstrated a significantly lower gene module score with DNA repair and cell cycle regulator genes in the scMSC obtained from the non-responder at baseline. Gene modules associated with Wnt signaling (WNT2, WNT2B), osteopontin (SPP1), SOX family of transcription factors (SOX5, SOX11, SOX18), interleukins (IL1, IL6, and IL10) and tumor necrosis factor superfamily members (TNFSF9, TNFSF15) were found in the non-responder.

Baseline MSC from the non-responder also showed significant overexpression of CXCL12 (chi-square=1630.5, p-value < 0.05). These CXCL12-expressing MSC had more defined transcriptomic profiles suggestive of lineage commitment, forming distinct clustering into osteogenic (identified by SPARC, COL1A1 and COL1A2 expression) or chondrogenic cells (identified by ACAN, CCN1 and CCN2 expression). These findings suggest the presence of dysregulated differentiation of scMSC at baseline in the patient who subsequently failed 5-Aza treatment.

Conclusion

MSC in the BM microenvironment are integral to the healthy HSC niche. Our high-resolution transcriptomic analysis highlights intrinsic dysregulation of MSC in MDS/MPN patients, with depletion of scMSC and impaired transcriptomic homogeneity as potential associations in developing drug resistance and disease progression. Further work to clarify the mechanism is underway.

Than:AbbVie: Consultancy, Honoraria, Other: Advisory fees; GSK: Consultancy, Honoraria, Other: Advisory fees; BMS: Consultancy, Honoraria, Other: Advisory fees; PharmaEssentia Corporation: Consultancy, Honoraria, Other: Advisory fees; Novartis: Consultancy, Honoraria, Other: Advisory fees.