The transcriptional landscape of hematopoietic stem cell ontogeny

Abstract

Transcriptome analysis of adult hematopoietic stem cells (HSCs) and their progeny has revealed mechanisms of blood differentiation and leukemogenesis, but a similar analysis of HSC development is lacking. Here, we acquired the transcriptomes of developing HSCs purified from >2,500 murine embryos and adult mice. We found that embryonic hematopoietic elements clustered into three distinct transcriptional states characteristic of the definitive yolk sac, HSCs undergoing specification, and definitive HSCs. We applied a network-biology-based analysis to reconstruct the gene regulatory networks of sequential stages of HSC development and functionally validated candidate transcriptional regulators of HSC ontogeny by morpholino-mediated knockdown in zebrafish embryos. Moreover, we found that HSCs from in vitro differentiated embryonic stem cells closely resemble definitive HSCs, yet lack a Notch-signaling signature, likely accounting for their defective lymphopoiesis. Our analysis and web resource will enhance efforts to identify regulators of HSC ontogeny and facilitate the engineering of hematopoietic specification.

Figure 1. Acquisition of global gene expression profiles of HSC compartments throughout murine ontogeny

(A) Outline of the developmental time-points, embryonic tissues, and ESC-derived populations examined. See also Figure S1. (B) RNA was collected from double sorted cells, amplified, and then hybridized to 430 2.0 Affymetrix gene chips. The resulting data was analyzed by unsupervised hierarchical clustering, PCA, Naïve Bayesian classifier, WGCNA, GRN reconstruction by CLR, differential expression analysis, and GSEA prior to functional studies in zebrafish.

Figure 2. Identification of distinct transcriptional and regulatory stages of HSC ontogeny

Results of unsupervised hierarchical clustering (A) and PCA (B). (C) Schematics of the HSPC gene regulatory network (GRN) in YS, AGM, FL12 F, FL13, and FL14 relative to WBM HSC. This GRN is composed of HSPC expressed genes (center rectangle), transcription factors (TFs) predicted to regulate these genes (circles), and cooperating gene sets that must be highly expressed for the TFs to exert a regulatory influence (contexts, boxed areas). Stimulatory TFs are shown as arrows and inhibitory TFs are shown as blunt lines. Differences in TFs expression are shown as red circles (up-regulated) or blue circles (down-regulated). Only TFs where the absolute value of the log2 ratio of the given sample versus WBM HSCs exceeds 1 are shown. See also Figure S2.

Figure 3. Comparison of HSCs in ontogeny to adult tissues and cell types

(A) A Naïve Bayesian classifier was used to assess transcriptional overlap with 44 tissues and cell types. The results of 20 of these comparisons are displayed (all other reference tissues and cell types were negative). NPC, neural progenitor cells; RPE, retinal pigment epithelium; SM, skeletal muscle; CM, cardiac muscle; Mac, macrophage; MEP, megakaryocyte-erythrocyte progenitor; CMP, common myeloid progenitor; GMP, granulocyte-monocyte progenitor. Each row is a biological group (i.e. WBM HSCs), and each column is a known tissue or cell type. The classifier determines the posterior probability that a sample is indistinguishable from each of the tissues or cell types in the reference data set. Higher probabilities are bright yellow and low probabilities are dark green and black. (B) Schematics of the macrophage GRN in AGM, FL12 A, FL12 F, FL13, and FL14 relative to primary macrophages (see Figure Legend 2C for details). Only TFs where the absolute value of the log2 ratio of the given sample versus primary macrophages exceeds 2 are shown. See also Figure S3.

Figure 4. Pathway enrichment analysis of pair-wise comparisons between major developmental hematopoietic groups

(A) GSEA of pair-wise comparisons to find GO Biological Processes enriched (red) or depleted (blue) between developmental populations. (B) GSEA of pair-wise comparisons to identify NetPath-annotated signaling pathways transcriptionally activated or suppressed. “In vivo Definitive HSC” includes WBM, FL12 F, E13.5 FL and E14.5 FL; “Specifying HSC” includes AGM, FL12 A, and placenta; “YS-like” includes EB and YS; and “All definitive HSC” includes ESC-HSC, WBM, FL12 F, E13.5 FL and E14.5 FL. Only significant gene sets are shown (Family wise error rate<0.05). See also Figure S4.

Figure 5. Identification of stage-specific gene sets

Co-regulated gene sets were identified via WGCNA. 66 distinct modules were discovered. A module was considered “Definitive HSC” if its expression was significantly higher in definitive HSC (FL12 F, FL13, FL14, ESC-HSC, and WBM) relative to other samples (Holm-corrected p-value<0.01). Similarly, modules were annotated as “stage-enriched”, “specifying”, or “in vitro”. Each row represents the module profile: a summary of the expression pattern of all genes within a module. See also Figure S5 and Table S1.

Figure 6. Identification and validation of novel transcriptional regulators of discrete stages of HSC ontogeny

(A) CLR was applied to identify putative TRs for each module. A network schematic of the CLR-derived predictions at the 0.05 FDR for all “Definitive HSC” modules is shown. Pink squares represent modules and blue circles represent predicted TRs. “Hub” genes are labeled black. Genes assessed functionally in zebrafish are highlighted in red. A list of all genes predicted to regulate each module can be found at http://hsc.hms.harvard.edu/. (B) Whole-mount in situ hybridization for c-myb and runx1 was performed at 36 hpf on uninjected embryos or embryos injected with morpholinos (MO) targeting mllt3, gfi1, atf3, tulp4, or prdm16. Bars in gfi1b panels designate the posterior ICM. (C) The CHT of prdm16 morphants was examined 4 dpf for c-myb/runx1 expression via whole-mount in situ hybridization. (D) Prdm16 morphants were examined via whole-mount in situ hybridization for mpo or l-plastin expression at 38 hpf. (E) Prdm16 morphants were examined via whole-mount in situ hybridization for rag-1 at 4 dpf. See also Figures S6, S7.