Sequencing of RNA in single cells reveals a distinct transcriptome signature of hematopoiesis in GATA2 deficiency

Abstract

Constitutional GATA2 deficiency caused by heterozygous germline GATA2 mutations has a broad spectrum of clinical phenotypes, including systemic infections, lymphedema, cytopenias, and myeloid neoplasms. Genotype-phenotype correlation is not well understood mechanistically in GATA2 deficiency. We performed whole transcriptome sequencing of single hematopoietic stem and progenitor cells from 8 patients, who had pathogenic GATA2 mutations and myelodysplasia. Mapping patients' cells onto normal hematopoiesis, we observed deficiency in lymphoid/myeloid progenitors, also evident from highly constrained gene correlations. HSPCs of patients exhibited distinct patterns of gene expression and coexpression compared with counterparts from healthy donors. Distinct lineages showed differently altered transcriptional profiles. Stem cells in patients had dysregulated gene expression related to apoptosis, cell cycle, and quiescence; increased expression of erythroid/megakaryocytic priming genes; and decreased lymphoid priming genes. The prominent deficiency in lympho-myeloid lineages in GATA2 deficiency appeared at least partly due to the expression of aberrant gene programs in stem cells prior to lineage commitment. We computationally imputed cells with chromosomal abnormalities and determined their gene expression; DNA repair genes were downregulated in trisomy 8 cells, potentially rendering these cells vulnerable to second-hit somatic mutations and additional chromosomal abnormalities. Cells with complex cytogenetic abnormalities showed defects in genes related to multilineage differentiation and cell cycle. Single-cell RNA sequencing is powerful in resolving transcriptomes of cell subpopulations despite a paucity of cells in marrow failure. Our study discloses previously uncharacterized transcriptome signatures of stem cells and progenitors in GATA2 deficiency, providing a broad perspective of potential mechanisms by which germline mutations modulate early hematopoiesis in a human disease. This trial was registered at www.clinicaltrials.gov as NCT01905826, NCT01861106, and NCT00001620.

Conflict of interest statement

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Figures

Graphical abstract

Figure 1.

Hematopoietic differentiation in healthy donors and GATA2 deficiency. (A) A tSNE plot of single-cell gene expression of healthy donors. (B) Reconstruction of the hematopoietic hierarchy pseudo-time ordering in Monocle. Cells are colored by types (HSC, MEP, GMP, ETP, and ProB). (C) Average expression of characteristic genes for lineages are highlighted in a differentiation tree. HSC genes: SPINK2, CRHBP, MEIS1, and MLLT3; MEP genes: GATA1, HBD, TFRC, UROD, NFIA, and KLF1; myeloid progenitor genes: MPO, CEBPA, ELANE, IRF8, and LGALS1; lymphoid progenitor genes: DNTT, CXCR4, CD79A, BLNK, IGLL1, and EBF1. The characteristic gene list is from our previous findings using scRNA-seq and validated by qualitative polymerase chain reaction and literature mining (supplemental Figure 2). (D) Expression of GATA2 and GATA1 highlighted in the differentiation tree. (E) Dynamic expression of GATA2 and GATA1 along pseudo-time of HSC differentiation into MEP. The x-axis shows pseudo-time ordering from HSC to MEP estimated by Monocle 2. The y-axis shows normalized gene expression by scaleData function in Seurat package. (F) Top 20 coexpressed (pink) and anticoexpressed (blue) genes with GATA2 and GATA1 (yellow). Red and blue lines connecting the genes indicate the top positive and negative correlations, respectively. Same plots showing positive and negative correlations of all values are in supplemental Figure 4.

Figure 2.

Gene mutations, blood cell counts, and hematopoietic differentiation in patients with GATA2 deficiency. (A) Schematic diagrams of GATA2 messenger RNA and protein. Mutations of the GATA2 gene or protein lesions in patients are indicated. (B) Patients’ hemoglobin levels (HGB), platelet counts (PLT), white blood cell counts (WBC), neutrophil counts, monocyte counts, B-cell counts, NK-cell counts, and T-cell counts. Background shading shows a normal reference range for each parameter. (C) Patients’ cells were mapped onto the differentiation tree (left) by the nearest neighbor method, based on transcriptome similarity with cells from healthy donors. A pie chart (right) in each panel shows percentages of subpopulations in HSPCs. (D) AUC scores were computed based on expression of lineage signature genes. Signature scores of MEP, GMP, ProB, and ETP in patients were compared with those in healthy donors using 2-sided, unpaired Mann-Whitney U test. A column plot shows means ± standard deviation. *P < .05. Sample patient 4-2 was obtained before HSC transplantation, 8 months after patient 4-1. During this interval, patient 4 had extensive progression of human papilloma virus infection, but there was no significant difference in blood counts and BM biopsy appearance. NS, not significant.

Figure 3.

Pathway analysis of differentially expressed genes in SPCs of GATA2 deficiency. (A-C) Results in total CD34+ HSPCs. (D-E) Results in HSCs. (A) GSEA of differentially expressed genes in GATA2 deficiency compared with those in healthy donors. (B) A network of upregulated genes in cell cycle and heme metabolism in HSPCs of GATA2 deficiency. (C) A network of downregulated genes in B and T cell signaling and immune responses in HSPCs of GATA2 deficiency. (D) Lists of upregulated (top) and downregulated (bottom) genes in HSCs in GATA2 deficiency. (E) Networks of upregulated (top) and downregulated (bottom) genes in different pathways in HSCs in GATA2 deficiency. FDRq, false discovery rate q value; NES, normalized enrichment score.

Figure 4.

Lineage priming programs in HSC and transcription factors switch. AUC scores of signature genes of early B progenitors (A) and MEP (B) in HSCs. The x-axis shows pseudo-time ordering of HSCs. The y-axis shows AUC scores of expression of lineage signature genes. Normalized expression levels of SPI1 in HSCs (C) and GMPs (D). Dynamic expression of GATA2 (blue), GATA1 (red), and SPI1 (also known as PU.1; brown) in differentiation from HSCs to MEPs determined by pseudo-time ordering with Palantir in healthy donors (E) and patients (F). The x-axis shows pseudo-time ordering from HSC to MEP estimated by Monocle 2. The y-axis shows normalized gene expression by scaleData function in Seurat package.

Figure 5.

Gene coexpression analysis in GATA2 deficiency. (A) Gene correlation distribution in healthy donors (pink; 0.390 ± 0.003) and patients (blue; 0.480 ± 0.110) The x-axis shows gene correlation score after imputations with MAGIC. The y-axis shows frequency. (B) A bar graph shows comparison of average correlation scores in patients with those in healthy donors. (C) A comparative correlation heatmap shows correlation between pairs of genes ranging from negative correlation (blue) to positive correlation (red) in healthy donors (upper left) and patients (lower right). The upper panel shows unsupervised modules of differentially coexpressed genes defined by weighted gene coexpression network analysis. Modules of genes are distinguished by colors. GO terms for each module of genes identified in the coexpression matrix are listed. Detailed information of individual gene modules is presented in supplemental Table 7. (D) Gene coexpression networks enriched in the KEGG_MTOR signaling pathway (left), the T-cell receptor signaling pathway (middle), and the B-cell receptor signaling pathway (right). Gene correlation networks are illustrated as the difference between patients and healthy donors.

Figure 6.

Aneuploidy in GATA2 deficiency. (A) Average gene expression for each chromosome in single cells from patients 1, 2, and 8. Average gene expression levels of individual chromosomes from 4 healthy donors were used for comparison. Chromosomal mapping read values were median centered. Top and bottom of the bars represent the 25% and 75% quartiles, respectively. (B) Heatmaps of chromosomal copy number variations (CNVs) signals of patient 8 obtained by the sliding window analysis. scRNA-seq data of patients were normalized against those of healthy donors. Copy number changes were examined in 22 chromosomes (columns) for patients’ individual cells (rows). Chromosome numbers are indicated on the top row. Three populations of cells are indicated on the right as +1,der(1;7); +1,der(1;7),+8; and diploid cells. (C) Percentages of aneuploid cells in subsets of HSPCs in patients 1, 2, and 8. The x-axis shows frequency of aneuploid cells. The y-axis shows HSPC subsets. (D) A heatmap of downregulated DNA repair genes in trisomy 8 cells in GATA2 deficiency. Networks of upregulated (E) and downregulated (F) genes in trisomy 8 cells in GATA2 deficiency.