Imputing causality from single-cell transcriptomics in paroxysmal nocturnal hemoglobinuria

Abstract Background Paroxysmal nocturnal hemoglobinuria (PNH) is a rare life-threatening acquired disease, clinically manifest as intravascular hemolysis, bone marrow (BM) failure, and venous thrombosis. PNH originates in hematopoietic stem cells (HSCs) from somatic mutation of the phosphatidylinositol glycan class A (PIGA) gene, which interrupts synthesis of the glycosylphosphoinositol (GPI) moiety and failure to express GPI-anchored cell surface proteins (GPI-AP). In immune-mediated marrow failure, PIGA-mutated (mtPIGA) HSCs clonally expand and can represent the majority of peripheral blood mature cells in PNH patients, but most GPI-AP deficient (GPI(-)) PNH clones are small and stable over time. Several hypotheses have been proposed to explain the clonal expansion of mtPIGA clones, but definitive mechanisms remain unclear. We performed single-cell RNA sequencing (scRNA-seq) of BM cells from patients with large (≥ 50%) and small (10-50%) PNH clones. Method BM from eight patients (male, n = 2, female, n = 6; median age, 43 [range, 21-67]) were divided into two groups according to PNH clone size at sample collection: five patients had large PNH clones (≥ 50% of GPI(-) HSPCs) and three had small PNH clones (10-50% of GPI(-) HSPCs). BM samples were obtained from the eight patients and five age- and sex-matched healthy donors. Flow cytometric sorting of GPI(-) and GPI(+) cells from subjects was performed immediately following isolation of BM mononuclear cells (BMMNCs). Sorted cells were processed through Chromium Controller (10x genomics, Pleasanton, CA, USA). scRNA-Seq libraries were constructed using the Chromium Single Cell 3' Reagent Kits (10x genomics) and sequenced with the NovaSeq 6000 system (Illumina, San Diego, CA, USA). Results As expected, gene expression profiles of HSPCs from PNH patients and healthy controls were different by scRNA-seq analysis, but differences were also apparent between GPI(-) and GPI(+) HSPCs within each patient, and notably between HSPCs from patients with small and large PNH clones. We determined gene expression profiles of each HSPC subset in patients with large PNH clones: by gene set enrichment analysis (GSEA) of differentially expressed genes between GPI(+) vs GPI(-) HSPCs, upregulation of immune response pathways and downregulation of cell-cycling pathways were observed in GPI(+) HSPCs (consistent with our previous results from array data, Chen G et al, Leukemia. 2005;19(5) 862-8). When we compared patients with small and large clones, we were surprised to observe downregulation of genes involved in immune response pathways and upregulation of genes involved in cell-cycling pathways in unmutated GPI(+) HSPCs, as compared to GPI(-) HSPCs, a trend opposite to that seen for HSPCs from patients with large PNH clones. When gene module scores were calculated to directly compare transcriptional features across GPI(+) and GPI(-) HSCs of patients with small and large PNH clones, GPI(+) HSCs in patients with large PNH clones showed significantly higher IFN-γ response and activated HSC module scores than those of GPI(+) HSCs n patients with small PNH clones. However, these scores were equivalent between GPI(-) HSCs from patients with large PNH clones and those from patients with small PNH clones. In summary, transcriptional changes in GPI(+) HSCs, not in GPI(-) HSCs, explain differences in gene expression between GPI(+) and GPI(-) HSCs in PNH. We elucidated gene expression profiles in the PNH BM environment by examining transcriptomes of enriched BM immune cells by scRNA-seq. When gene expression of BM effector cells was compared between patients with small and large PNH clones, upregulation of IFN-γ response pathway was observed in most effector cells from patients with large PNH clones. In addition, cell-cell communications between effector cells and GPI(+) normal HSPCs were enhanced in patients with large PNH clones. In contrast, enhanced interactions between effector cells and GPI(+) HSPCs were not seen in patients with small PNH clones. These results indicate that strong negative selection by effector immune cells on GPI(+) HSCs is linked to clonal expansion of GPI(-) HSCs. Conclusion Our results provide direct evidence of functional evolution in clonal hematopoiesis, and immune escape as responsible for the co-occurrence of PNH and immune bone marrow failure.