A Transcriptional Switch Point During Hematopoietic Stem and Progenitor Cell Ontogeny

Abstract

During mammalian embryogenesis, hematopoietic stem and progenitor cells (HSPCs) originate from mesoderm-derived endothelial cells in the aorta-gonad-mesonephros (AGM) region and placenta (PL). Later, HSPCs expand in fetal liver (FL) and migrate to bone marrow (BM) shortly before birth. Understanding global transcriptional regulation governing HSPC emergence from embryonic stem/induced pluripotent stem cells is necessary to devise clinical applications, such as novel transplantation approaches. In this study, to assess transcriptional dynamics during development, we performed cap analysis of gene expression on 10 developmental murine HSPC populations isolated from the AGM region, PL, FL, and BM and identified 15,681 transcripts across HSPC ontogeny. We performed microarray analysis of AGM-derived HSPCs at 9.5 and 10.5 days postcoitum (dpc) and identified 40 differentially expressed genes, 23 confirmed as significantly changed by real-time polymerase chain reaction. We conclude that a transcriptional switch point occurs in HSPC ontogeny between 9.5 and 10.5 dpc in the AGM region.

Introduction

Hematopoietic stem cells (HSCs) either self-renew or differentiate into all blood cell types. How HSCs maintain a balance between self-renewal and differentiation capacity throughout their lifespan is a key area of investigation [1]. During mammalian embryogenesis, HSCs emerge through a complex process involving ontogenesis at distinct anatomical sites, including the yolk sac, the aorta-gonad-mesonephros (AGM) region, placenta (PL), and fetal liver (FL), culminating at birth in colonization of bone marrow (BM) [2]. Understanding molecular mechanisms regulating these steps could have direct clinical applications for HSC transplantation therapy. Decades of research investigating HSC developmental stages and biology demonstrate that intrinsic signals and niche factors regulate these transcriptional programs [3 –5], but the genome-wide picture of the transcriptional network governing HSC development and maturation remained far from complete.

In combination with sensitive well-defined assays based on microarray technology, RNA sequencing, and serial analysis of gene expression, researchers previously defined transcriptional control mechanisms regulating transient hematopoietic stem and progenitor cell (HSPC) populations [6,7]. However, difficulty in obtaining sufficient amounts of nucleic acid material for subsequent analysis has limited research progress. In addition, researchers found it challenging to compare lists of differentially regulated genes due to use of different cell populations or HSC classification criteria.

Cap analysis of gene expression (CAGE) sequencing is a method used to identify the 5′ ends of capped RNAs based on cap trapping and hence provides a means to detect likely promoter regions [8]. In this study, as part of the FANTOM5 project [9 –11], we utilized CAGE sequencing to examine primary murine HSPCs derived from 10 spatially or temporally critical locations during HSPC development, including the para-aortic-splanchnopleural (p-Sp) region at 8.5 days postcoitum (dpc); AGM region at 9.5, 10.5, and 11.5 dpc; PL at 11.5 dpc; FL at 12.5, 14.5, and 19.5 dpc; and BM at 2- to 3-month old and at 2-year old.

Genome-wide expression profiles of HSPC samples generated from single-molecule CAGE sequencing [12] revealed 15,681 transcription start sites (TSSs). The 10 groups were clustered as pre-HSPCs, definitive HSPCs, fetal HSPCs, and adult HSPCs, allowing further generation of signature gene lists for each stage. The 15,681 TSSs mapped to 10,385 genes, highlighting an abundance of alternate transcripts and indicating that major changes in the transcriptome occur in the AGM region from 9.5 to 10.5 dpc. Due to the requirement for a large number of embryos for CAGE sequencing, we performed microarray and real-time polymerase chain reaction (PCR) analysis to confirm the CAGE sequencing data. Consistent with the CAGE sequencing, microarray and real-time PCR confirmed that a transcriptional switch point exists in HSPC ontogeny from 9.5 to 10.5 dpc at the AGM region. This work is part of the FANTOM5 project [9 –11]. Data download, genomic tools, and copublished articles are summarized at http://fantom.gsc.riken.jp/5/

Materials and Methods

Animals

ICR and C57BL/6J mice were purchased from Nihon SLC (Hamamatsu, Japan) and Kyudo (Tosu, Japan), respectively. Noon of the day of the plug was defined as 0.5 dpc. Embryos at various developmental stages were dissected in phosphate-buffered saline (PBS) under a stereomicroscope and the number of somite pairs (SP) counted [13,14]. Animals were handled according to Guidelines for Laboratory Animals of Kyushu University. This study was approved by the Animal Care and Use Committee, Kyushu University (Approval ID: A21-068-0).

Cell preparation

The caudal portion of embryos containing the p-Sp/AGM region was used to obtain a single cell suspension. ICR embryos were used at 8.5 dpc, whereas C57BL/6J embryos were used at 9.5, 10.5, and 11.5 dpc. Single cells were prepared from p-Sp/AGM at 8.5 dpc and the AGM region at 9.5, 10.5, and 11.5 dpc by collagenase treatment (Supplementary Data; Supplementary Data are available online at www.liebertpub.com/scd). PLs at 11.5 dpc without deciduas and umbilical vessels were passed through 21-gauge needles and incubated with collagenase. To isolate mononuclear cells, density gradient centrifugation using Lympholyte^®-M Cell Separation Media (Cedarlane Laboratories, Ontario, Canada) was performed according to the manufacturer's instructions.

To obtain FL HSPCs, FL cells from 12.5, 14.5, and 19.5 dpc C57BL/6J embryos were filtered through 40-μm nylon mesh (BD Biosciences) and washed once with PBS. Mononuclear cells were isolated as stated above. Mature blood cells were removed by cell sorting after staining with biotin-conjugated antilineage markers (Supplementary Data).

To obtain adult BM HSPCs, femurs, tibias, and humeri of 2- to 3-month-old and 2-year-old C57BL/6J mice were dissected out. BM cells were harvested by flushing with PBS and passed through 40-μm nylon cell strainers (BD Biosciences). Mononuclear cells were isolated and mature blood cells removed by magnetic-activated cell sorting. Cells were incubated with biotin-conjugated antibody as described above. Cells were then incubated with anti-Biotin MicroBeads (Miltenyi Biotec, Bergisch Gladbach, Germany) and passed through MACS^® Separation Columns (Miltenyi Biotec).

Flow cytometry and cell sorting

Antibodies used for cell sorting are shown in Supplementary Data. After gating for propidium iodide–negative (living) cells, mesodermal cells, pre-HSPCs, and HSPCs were isolated from hematopoietic organs using the following protocol. For the 8.5 dpc p-Sp sample, among E-cadherin-negative nonendodermal cells, mesodermal cells expressing Flk-1 (VEGF receptor 2) and c-Kit (stem cell factor receptor) were sorted out. For the 9.5 dpc AGM region sample, cells double positive for CD31 (PECAM-1) and CD34 (mucin-like glycoprotein), which include both HSPCs and vascular endothelial cells, were selected. Among CD31⁺/CD34⁺ cells, hematopoietic cells expressing c-Kit were sorted out as pre-HSPCs. For the 10.5 dpc AGM region, cells double positive for CD31 and CD34, which include both HSPCs and vascular endothelial cells, were selected. Among CD31⁺/CD34⁺ cells, c-Kit⁺ hematopoietic cells were sorted.

To remove macrophages among hematopoietic cells, we used glycoprotein F4/80. In addition to the AGM region, PL reportedly generates adult repopulating HSCs. Thus, we collected a sample from PL expressing c-Kit, CD31, and CD34 at 11.5 dpc [15]. For 12.5 dpc FL [16], we sorted HSPCs expressing Sca1 (stem cell antigen 1), c-Kit, and CD45. The common leukocyte marker, CD45, was used as an HSC maturation marker. To remove differentiated cells in 14.5 dpc FL [16], we used Ter119 (erythroid cells), CD45 (leukocytes), CD19 (B-lymphocytes), CD4, CD8 (T-lymphocytes), Gr-1 (granulocytes), and F4/80 (macrophages) markers for cell sorting, and all negative cells were classified as lineage negative (Lin⁻). Among Lin⁻/Sca1⁺ cells, c-Kit⁺/CD45⁺ cells were sorted as HSPCs. To examine the effect of aging, BM HSPCs from 2- to 3-month-old or 2-year-old mice were collected by selecting CD34⁻/Sca1⁺/c-Kit⁺ cell populations. Among Lin⁻/Sca1⁺ cells in 2- to 3-month-old BM, c-Kit⁺ cells were sorted out, regardless of CD34 expression.

RNA extraction and CAGE analysis

Total RNA was isolated and treated with DNase I using an RNeasy^® Plus Micro Kit (Qiagen, Hilden, Germany) according to the manufacturer's instruction. CAGE analysis was carried out as part of the Functional Annotation of the Mammalian Genome 5 (FANTOM5) project at the RIKEN Omics Science Center in Yokohama, Japan.

For real-time PCR, total RNA was isolated and treated with DNase I using an RNAqueous^®-4PCR Kit (Ambion, Inc., Austin, TX), according to manufacturer's instruction. DNase I-treated RNA was quantitated by NanoDrop 2000/2000c (Thermo Scientific, DE). cDNA was prepared using a High Capacity RNA-to-cDNA Kit (Life Technologies, Carlsbad, CA) according to the manufacturer's instruction. Briefly, the cDNA synthesis reaction included random octamers, dNTPs, an RNase inhibitor, MuLV reverse transcriptase, and DNase I-treated RNA. cDNA was synthesized at 37°C for 60 min followed by denaturation at 95°C for 5 min and holding at 4°C until use.

CAGE bioinformatics analysis

TSSs were assigned to known genes by the FANTOM5 consortium. If the CAGE peak was within 500 bases of the 5′ end of a known transcript, it was annotated with the gene name from which that transcript was derived. Enrichment of differentially expressed genes with respect to transcription factor ChIP-seq datasets was calculated using the GSCA tool [17], while functional and pathway enrichment was calculated using Database for Annotation, Visualization and Integrated Discovery (DAVID) [18,19]. ChIP-sequencing data for multiple transcription factors in HSCs and HPLCs were collected from gene expression omnibus [10,20 –22]. Enrichment for known sequence motifs was performed using HOMER [23].

Genome-wide chromatin modifications in murine HSC samples were downloaded from Mouse ENCODE [19], and methylation data were downloaded from Hogart et al. [24]. The seqMINER tool was used to cluster epigenetic marks [18]. P values were calculated using a hypergeometric test. Data analysis was done using a combination of R, Perl, and shell scripts. To provide a snapshot of global transcription (the transcriptome) in HSCs across different times and locations, GEDI (Gene expression dynamics inspector) plots (maps) were generated using GEDI23 software (www.childrenshospital.org/research/ingber/GEDI/gedihome.htm).

Microarray analysis

Microarray analysis of three independent samples each of AGM-derived HSPCs at 9.5 and 10.5 dpc was performed and compared. Total RNA was isolated from sorted hematopoietic cells using an RNAqueous^® Total RNA Isolation Kit (Thermo Fisher Scientific, Inc., MA). Total RNA was linearly amplified in two rounds of T7 in vitro transcription to generate antisense amplified RNA (aRNA) using a MessageAmp™ II aRNA Amplification Kit (Thermo Fisher Scientific, Inc., MA) and an Illumina^® TotalPrep RNA Amplification Kit (Thermo Fisher Scientific, Inc., MA) according to the manufacturer's instructions. During the second round of amplification, aRNA was labelled with Biotin-16-UTP. aRNA was purified and verified by spectrophotometry. Subsequently, the Illumina Gene Expression system (Illumina, Inc., CA) was used for direct hybridization of labelled aRNA to gene-specific 50-mer oligonucleotide probes attached to microbeads according to the manufacturer's instruction. After hybridization and washing, BeadChips were immobilized with Cy3-streptavidin (GE Healthcare, Buckinghamshire, UK) and scanned using an Illumina BeadArray Reader.

To analyze microarray data and filter criteria, raw signal intensities of six samples were normalized using the quantile algorithm with “lumi” [25] and the “preprocessCore” library package [26] on Bioconductor software [27]. Probes called by the “Detection P value <0.05” flag in at least one sample were selected. Then, Linear Models for Microarray Analysis (limma) package [28] of Bioconductor software were applied. Differentially expressed genes were shown on a heat map generated by MeV software [29]. Hierarchical clustering (HCL) analysis was used to sort genes. Color coding indicated distance from the median of each row. DAVID was used to investigate gene ontology categories enriched for function of differentially expressed genes. Gene encoding factors functioning in transcription, either upregulated or downregulated, were selected and validated by real-time PCR. Primer sets used are shown in Supplementary Table S1.

Real-time PCR analysis

Gene expression levels were measured by real-time PCR using Fast SYBR^® Green Master Mix (Life Technologies, Carlsbad, CA) and StepOnePlus™ real-time PCR (Life Technologies, Carlsbad, CA). Forward and reverse primers were designed using Primer Express^® version 3 (Applied Biosystems) and are listed in Supplementary Table S1. Primer specificity was assessed in silico using BLAST (Supplementary Table S1). Primer efficiency was calculated from the slope of the calibration curve using fivefold serial dilution of cDNA prepared from whole embryos or whole fetal organ-derived cDNA in real-time PCR. Amplification conditions were an initial denaturation at 95°C for 20 s, followed by 40 cycles of denaturation at 95°C for 3 s, and annealing and extension at 60°C for 30 s.

To ensure specific amplification, melting curve analysis was evaluated in all analyses and in a negative control lacking cDNA template. Melting curve analysis consisted of denaturation at 95°C for 15 s and annealing at 60°C for 1 min/cycle, and annealing temperature was increased 0.3°C/cycle until 95°C. All analyses were performed in triplicate wells; mRNA levels were normalized to Actb mRNA, and the relative quantity of expression was calculated by delta delta Ct method and compared with a reference sample. Differences were statistically evaluated using Student's t-test. P values less than 0.05 indicated a statistically significant difference.

Results

Collection of HSPC samples from murine tissues at different developmental stages

To characterize changes in the HSPC transcriptome during hematopoietic development, we collected 10 samples from diverse murine organs at selected developmental time points (Fig. 1A). Relevant to marker analysis, E-cadherin⁻ Flk-1⁺ cells represent mesodermal cells [30], and c-Kit marks intra-aortic clusters of the AGM region [13,31]. Hematopoietic multipotent progenitors in the p-Sp region at 8.0 dpc [32] and Flk-1⁺ c-Kit⁺ cells have been observed at p-Sp/AGM region at 9.5 dpc [33]. Therefore, we collected E-cadherin⁻/Flk-1⁺/c-Kit⁺ mesodermal cells, the ancestors of hematopoietic cells, from the caudal region (p-Sp) of 8.5 dpc embryos. Cells capable of reconstituting neonatal recipients, known as “pre-HSPCs,” have been detected in the p-Sp/AGM region at 9.5 dpc [34,35], while cells present at 10.5–11.5 dpc acquire the capacity to reconstitute adult recipients and are known as “long term-repopulating HSCs” [36 –38]. We obtained both pre-HSPC and HSPC samples from the AGM region by selecting cells expressing the HSPC marker c-Kit and the endothelial markers CD31 and CD34 [39,40].

FIG. 1.

Isolation of HSPCs from mouse embryos and adults. (A) A total of 10 tissues (indicated in red) derived from 9 indicated developmental time points served as HSPC sources for CAGE analysis. They include (1) caudal regions (p-Sp, dotted line) of 8.5 dpc embryos (mesoderm); (2) 9.5 dpc p-Sp/AGM tissue for pre-HSPCs; (3) AGM tissue at 10.5 dpc; (4) 11.5 dpc; (5) 11.5 dpc PL; (6) 12.5 dpc FL; (7) 14.5 dpc FL; (8) 19.5 dpc FL; (9) BM from 2- to 3-month-old mice; and (10) BM from 2-year-old mice. (B) Flow cytometric analysis of hematopoietic tissues based on surface expression of hematopoietic cell markers. Single cell suspensions of indicated embryonic and adult tissues were prepared and analyzed by flow cytometry. Isotype control is not shown. To obtain mesodermal cells at 8.5 dpc, E-cadherin-negative cells were gated first. Then among them, Flk-1⁺ and c-Kit⁺ cells were analyzed. To obtain pre-HSPCs from p-Sp/AGM tissue at 9.5 dpc and HSPCs from AGM tissue at 10.5 dpc and from PL at 11.5 dpc, CD31⁺ and CD34⁺ cells were gated first. Then c-Kit⁺/F4-80⁻ cells were analyzed on CD31⁺/CD34⁺ cells. To obtain HSPCs from PL at 11.5 dpc, CD31⁺ and CD34⁺ cells were gated first. Then among them, c-Kit⁺ cells were analyzed. To obtain HSPCs from FL at 12.5 dpc, Sca-1⁺/c-Kit⁺ cells were gated first and then CD45⁺ cells were analyzed among them. To obtain HSPCs from FL at 14.5 and 16.5 dpc, Lin⁻ and Sca-1⁺ cells were gated first. Then c-Kit⁺/CD45⁺ cells were analyzed among them. To obtain HSPCs from BM at 2- to 3-month-old and 2-year-old mice, Lin⁻ and Sca-1⁺ cells were gated first. Then among them, c-Kit⁺/CD34⁻ cells were analyzed. AGM, aorta-gonad-mesonephros; BM, bone marrow; CAGE, cap analysis of gene expression; dpc, days postcoitum; HSPCs, hematopoietic stem and progenitor cells; FL, fetal liver; mo, month; PL, placenta; yr, year.

Among the 10 HSPC populations identified, we observed that HSPCs form a very small proportion of cells from a given niche (Fig. 1B). For the 8.5 dpc p-Sp sample, E-cadherin⁻/Flk-1⁺/c-Kit⁺ cells represented 2.5% ± 0.18% of the population; for the 9.5 dpc AGM region, CD31⁺/CD34⁺/c-Kit⁺ cells represented 0.68% ± 0.32%; and for 10.5 and 11.5 dpc AGM samples, CD31⁺/CD34⁺/c-Kit⁺/F4/80⁻ cells represented 0.12% ± 0.08% and 0.09% ± 0.14%, respectively. For 11.5 dpc PL, CD31⁺/CD34⁺/c-Kit⁺/F4/80⁻ cells represented 0.79% ± 0.67%; for 12.5 dpc FL, Sca-1⁺/c-Kit⁺/CD45⁺ cells represented 0.75% ± 0.12%; and for 14.5 and 19.5 dpc FL, Lin⁻/Sca-1⁺/c-Kit⁺/CD45⁺ cells represented 6.57% ± 0.95% and 3.95% ± 1.0%, respectively. Finally, for 2- to 3-month-old and 2-year-old BM, Lin⁻/Sca-1⁺/c-Kit⁺ cells represented 0.044% ± 0.009% and 0.14% ± 0.09%, respectively. Surface markers used to sort each sample are shown in Fig. 2A.

FIG. 2.

CAGE analysis over HSPC development. (A) Table showing indicated groups depicted in Fig. 1A, with respect to surface phenotypes, number of cells isolated, number of mapped reads, TSSs identified, and total number of genes mapped to each CAGE sample, as well as GEDI plots of the global transcriptional landscape. Each pixel within a GEDI plot represents a mini-cluster of genes with similar expression pattern across all analyzed samples. Pixel color indicates gene expression level for each cluster (blue to red indicates low to high levels, respectively). CAGE libraries were generated and sequenced to an average depth of ∼250,000 reads for each sample. Analysis detected use of 15,681 distinct TSSs in 10 samples with an average of 10,000 TSSs in each. (B) Number of differentially expressed TSSs between two consecutive stages of HSPC development. Given at least a twofold difference transcript expression, >3,000 TSSs were differentially expressed among p-Sp-8.5 dpc-derived mesoderm, AGM-9.5 dpc-derived pre-HSPCs, and AGM-10.5 dpc-derived HSPCs, whereas ∼1,000 TSSs were differentially expressed when comparing AGM-11.5 dpc-, PL-11.5 dpc-, FL-12.5 dpc-, FL-14.5 dpc-, FL-19.5 dpc-, BM-2- to 3-month, and BM-2 year-derived HSPCs. (C) 15,681 TSSs from all samples were validated using previously published ChIP-sequence data of chromatin modification relevant to HSPCs derived from bone marrow [53]. High levels of H3K4me3 and H3K27ac and lack of H3K4me1 confirmed TSSs as promoter regions. The figure was generated using seqMINER software. GEDI, gene expression dynamics inspector; TSSs, transcription start sites.

CAGE profiling of murine HSPCs identifies stage-specific transcripts

To study genome-wide transcriptional dynamics during HSPC development, we performed single molecule CAGE [12] sequencing of the 10 samples identified. The rarity of HSPCs during early development represents a challenge requiring collection of large numbers of mouse embryos. Thus, we opted not to generate replicates for each population. By generating ∼250,000 reads per sample, we identified a total of 15,681 distinct TSSs, which were detected (≥10 tags per million) in at least one of 10 samples, with an average of 8,037 TSSs per sample. We then used GEDI plots [41] to provide a global gene expression overview of each sample. GEDI plot analysis revealed a distinct transcriptome signature in each of the 10 samples (Fig. 2A). More than 3,000 TSSs showed at least a twofold difference in expression among p-Sp-8.5 dpc-derived mesoderm, AGM-9.5 dpc-derived pre-HSPCs, and AGM-10.5 dpc-derived HSPC samples, whereas ∼1,000 TSSs were differentially expressed in HSPCs derived from AGM-11.5 dpc, PL-11.5 dpc, FL-12.5 dpc, FL-14.5 dpc, FL-19.5 dpc, 2- to 3-month-old BM, and 2-year-old BM (Fig. 2B). These observations suggest that major transcriptional changes likely occur in the AGM region between 9.5 and 10.5 dpc.

Of TSSs, 15,681 were within 500 bases of a known transcript and were thus annotated with the corresponding gene symbol (representing 10,385 separate genes). The remaining 1,681 TSSs were unannotated and may represent novel HSPC-specific transcripts. The ratio of TSSs to genes was greater for transcription factors (TFs), with 1,518 TSSs mapping to 880 TFs (ratio 1.7 compared to 1.5 for all genes), suggesting that alternate transcription of TFs contributes to the regulatory complexity of the mammalian genome [42].

To validate this data using an independent source, we collected ChIP sequencing data for five chromatin modifications (H3K27me3, H3K4me3, H3K4me1, H3K79me2, and H3K27Ac) and for CTCF (CCCTC-binding factor) binding in murine HSCs derived from BM [43]. Over the 15,681 TSSs flanking known transcripts, H3K4me3 and H3K79me2 (predictors of transcription initiation) were enriched near CAGE peaks, while H3K4me1, an enhancer signature, and H3K27me3, a signature of inactive promoters, were depleted, supporting the idea that our analysis detects active transcription initiation events (Fig. 2C). All TSSs also overlapped with binding of CTCF binding, which reportedly preferentially binds near promoters [44].

Finally, to confirm the identity of each sample, we checked expression profiles of genes encoding HSPC surface markers (Flk-1, c-Kit, CD31, CD34, and Sca-1) and lineage markers (E-cadherin, F4/80, Gr-1, CD4, CD8, Ter119, and CD19) used for cell sorting (Supplementary Fig. S1A). As expected, we detected high Cd34 levels in all samples. Kdr (also known as Flk-1) was detected only in AGM samples and was downregulated in PL, FL, and BM-derived populations. Similarly, Pecam1 (also known as Cd31) was expressed at low levels in non-AGM samples. Conversely, c-Kit, Ly6a (also known as Sca-1), and Ptprc (also known as CD45) were more highly expressed in PL, FL, and BM-derived populations relative to AGM. We also detected low levels of Prom1 (also known as CD133), a marker of some early HSPC populations [45], in mesodermal and pre-HSPC samples.

A transcriptional switch point in HSPC ontogeny occurs in AGM between 9.5 and 10.5 dpc

HCL of the 10 HSPC populations assigned them to four clusters: (1) pre-HSPCs (p-Sp/8.5 dpc and AGM/9.5 dpc), (2) early HSPCs (AGM/10.5 dpc, AGM/11.5 dpc, PL/11.5 dpc, and FL/12.5 dpc), (3) fetal definitive HSPCs (FL/14.5 dpc and FL/19.5 dpc), and (4) adult definitive HSPCs (BM in 2- to 3-month-old mice and BM in 2-year-old mice) (Fig. 3A). Principal component analysis of the 10 samples was in agreement with partitioning of populations into four corresponding groups (Fig. 3B). In short, this analysis suggests that HSPCs are committed in group 1; become mature and prepared to move to FL in group 2; proliferate, differentiate into mature hematopoietic cells, and prepare to move to BM in group 3, and settle in BM and become quiescent in group 4.

FIG. 3.

Identification of gene signatures indicative of HSPC development. (A) Hierarchical clustering of 10 CAGE samples. (B) Principal Component Analysis of 10 CAGE samples showing indicated groups. (C) Summary of genes differentially expressed among the four stages with respective GO, as well as pathway enrichment analysis. GO, gene ontology.

To understand differences between groups, we identified sets of differentially expressed genes and annotated them using functional and pathway enrichment analysis (Fig. 3C). Although 9.5 and 10.5 dpc HSPCs exhibit similar surface markers, they were clustered into pre-HSPCs and early HSPCs, respectively, based on global gene expression patterns. Consistent with this clustering, the transition from endothelial to HSPC phenotype occurs after 9.5 dpc [13]. Therefore, genes related to vascular development are more highly expressed in group 1 (pre-HSPCs) than in group 2 (early HSPCs) (Fig. 3C). In addition, intra-aortic clusters containing embryonic HSPCs in the AGM region at 9.0–10.5 dpc are likely released into circulation to home to FL at 10.5–11.5 dpc based on β1-integrin expression [46,47].

We observed upregulation of gene functioning in transendothelial migration in early rather than pre-HSPCs (Fig. 3C; see“2 versus 1_up”), implying that early HSPCs are prepared to home. Hematopoietic genes were upregulated in HSPCs in groups 2–4, suggesting that group 1 pre-HSPCs are not yet committed to an adult HSPC program. Accordingly, Pecam1 and Cdh5 (also known as VE-cadherin) were downregulated in groups 2, 3, and 4 relative to group 1, whereas Itga2b (also known as CD41) was upregulated in those groups, as expected. Two key regulators, Ccnd1 (cyclin D1) and Twist1, were downregulated in groups 2–4 relative to group 1. Twist1 downregulation suggests that it may act as a master regulator of HSPC generation, while Ccnd1 downregulation suggests that proliferative status of group 1 pre-HSPCs changes as development proceeds (Supplementary Fig. S1B).

Identification of TFs regulating HSPC ontogeny

To identify stage-specific TFs governing HSPC development and maturation, we randomly selected 43 TFs differentially expressed (based on at least a twofold expression change) among the 10 samples (Fig. 4A). Of these, 9 (Sox18, Hmga2, Sox17, Sox7, Peg3, Hey1, Sox11, Snai1, and Fhl2) were downregulated during HSPC maturation (Fig. 4A, blue box). Expression of Sox17, Sox18, and Sox7 in AGM/9.5 dpc-derived pre-HSPCs suggests that these cells represent either endothelial/hematopoietic cell progenitors or cells in a transition state [13,48]. About a third (11 of 43) of the TFs (Hes6, Nr2c2, Tob1, Arhgap17, Irf1, Runx1, Cebpa, Nrip1, Maz, Mta1, and Aes) (Fig. 4A, red boxes) was differentially expressed between AGM-9.5 dpc and AGM-10.5 dpc.

FIG. 4.

Changes in global transcription over the course of HSPC ontogeny. (A) Heat map showing dynamically expressed TFs with their TSS location over HSPC development indicating differential TF expression in the AGM between 8.5 and 9.5 dpc. Blue box indicates genes downregulated during HSPC maturation; red boxes indicate upregulated genes, including Hes6, Nr2c2, Tob1, Arhgap17, Irf1, Runx1, Cebpa, Nrip1, Maz, Mta1, and Aes. (B) Overrepresented sequence motifs (black boxes) in each sample. Specifically, ISRE, CEBP, and Runx1 show dynamic behavior that reflects a particular pattern during HSPC ontogeny. TFs, transcription factors.

We reasoned that dynamically expressed gene loci should be enriched for cis-regulatory motifs recognized by these TFs. Most enriched cis-regulatory motifs for known factors obtained using HOMER software [23] (Fig. 4B) were overrepresented relative to random background sequences with the same GC content across all samples. They included motifs recognized by key HSPC TFs, including E twenty-six (ETS), bHLH proteins, JUN, MYB, PU.1, and STAT proteins. The Gfi1b motif was enriched in 11.5 dpc PL, but with P value <1e-3, which was lower than the strict cutoff (P = 1e-5). Three motifs, the ISRE (IFN-stimulated response element) [49] and sequences recognized by Cebp and Runx1, showed progressively greater enrichment during HSPC ontogeny (Fig. 4B). The ISRE motif is found in promoters of genes induced by interferon, which activates dormant HSCs [50]. Moreover, Ifnar2, a target of INF alpha, was upregulated in both FL and BM samples (Supplementary Fig. S1B).

Microarray analysis and gene selection and validation

To identify TFs that are differentially expressed in AGM-derived HSPCs at 9.5 and 10.5 dpc tissues, we selected a total of 370 differentially expressed genes after statistical analysis with limma (Supplementary Table S2). A heat map of these genes is shown (Fig. 5A). We then conducted enrichment analysis of gene function (Enrichment score >1.3) of 370 genes. Among them (Supplementary Table S2), 257 genes matched with gene identifier of the DAVID, and seven annotation clusters were enriched (Fig. 5B). We selected 40 genes for real-time PCR analysis and those genes are shown in Table 1. These 40 genes were normalized intensities based on three independent samples and consisted of 20 up- and 20 downregulated genes whose sequence information was obtained through the NCBI website and for which primer sets could be designed for real-time PCR. Then real-time PCR analysis was conducted to analyze gene expression in AGM-derived HSPC at 9.5 and 10.5 dpc. Among 20 upregulated genes, 11 were significantly upregulated (P < 0.05) at 10.5 dpc relative to 9.5 dpc (Fig. 5C). Among 20 downregulated genes, 12 were significantly downregulated (P < 0.05) at 10.5 dpc relative to 9.5 dpc (Fig. 5D).

FIG. 5.

Microarray analysis of AGM-derived HSPCs at 9.5 and 10.5 dpc. Three independent samples of the AGM-derived HSPCs at 9.5 and 10.5 dpc were compared. (A) Heat map of 370 differentially expressed genes. (B) Function enriched (Enrichment Score >1.3), as indicated by DAVID analysis. Functions of the 257 of 370 genes were annotated into seven clusters: cytoskeleton, intracellular organelle lumen, diacylglycerol binding, lysosome, GTPase regulator activity, mitochondrial lumen, and phagocytosis-related genes. Based on three independent analyses, gene encoding factors related to transcription (either upregulated or downregulated) were selected and validated by real-time PCR. A total of 20 each of upregulated and downregulated genes from the microarray were selected and validated by real-time PCR. (C) Among upregulated genes, 11 of 20 showed significant changes (P < 0.05) at 10.5 dpc relative to 9.5 dpc in independent samples. (D) Among downregulated genes, 12 of 20 showed significant changes (P < 0.05) in the same period in independent samples. Unfilled and gray filled bars represent gene expression levels in AGM-derived HSPCs at 9.5 and 10.5 dpc, respectively. Actb served as internal control. DAVID, Database for Annotation, Visualization and Integrated Discovery; PCR, polymerase chain reaction.

Table 1.

Differentially Expressed Candidate Genes

		AGM-derived HSCs at 9.5 dpc			AGM-derived HSCs at 10.5 dpc
Probe ID	Symbol	Sample no. 1	Sample no. 2	Sample no. 3	Sample no. 1	Sample no. 2	Sample no. 3	Ratio	P	Adjusted P value
2970324	Hdc	830.25	935.83	231.57	9519.32	9364.16	7794.57	15.68899	0.00014	0.07378
4880386	Muc13	291.20	293.79	650.28	3445.93	2988.55	2794.07	8.02708	0.00005	0.06115
5720609	Lyz	355.62	371.78	610.36	1701.37	2622.91	2588.35	5.23103	0.00007	0.06715
150458	Trp53rk	458.18	409.77	277.21	1917.66	2082.78	1026.80	4.28718	0.00041	0.08299
6200719	Acss1	410.76	446.32	252.09	1630.10	1529.18	1126.99	3.93188	0.00018	0.07378
4540564	Dok2	261.14	265.47	342.33	1220.66	1229.69	881.23	3.81983	0.00003	0.06115
130634	Ddit4	568.31	588.80	1092.67	3363.15	2332.50	2439.05	3.74040	0.00042	0.08299
6580021	Mpo	3535.14	3506.52	3883.81	9975.33	12586.18	17879.26	3.59923	0.00010	0.07378
2940681	Gata2	3451.52	3246.33	4606.56	16151.05	8857.75	13501.76	3.34486	0.00035	0.08299
4060593	Vangl2	450.57	481.80	579.40	1761.05	1908.88	1337.61	3.29427	0.00005	0.06115
1450333	Clec7a	246.66	231.57	251.49	753.09	821.82	648.83	3.03497	0.00002	0.06115
940541	Fcrls	423.85	433.03	326.64	1168.86	1097.15	1270.86	3.00681	0.00004	0.06115
6590167	Fcgrt	405.36	404.52	274.40	1159.82	1011.28	975.14	2.94030	0.00013	0.07378
940309	Mpo	11146.78	10513.86	11450.52	31002.28	29213.21	31831.42	2.77991	0.00001	0.06115
1010196	Zfp341	303.61	315.71	246.30	806.63	856.22	719.41	2.76093	0.00005	0.06115
5860154	Lst1	643.90	674.92	767.84	1547.38	2280.48	1888.36	2.71304	0.00013	0.07378
2140221	Slc40a1	306.05	313.44	303.61	756.40	692.31	1107.46	2.71045	0.00023	0.07845
6280168	Mpo	526.28	528.24	452.79	1190.50	1189.41	1168.86	2.36025	0.00005	0.06115
110717	AI428936	333.19	363.03	302.05	815.43	884.37	656.69	2.34904	0.00020	0.07472
1070224	Emp3	255.03	263.50	224.43	495.00	657.36	572.78	2.31197	0.00018	0.07378
130661	5830436K05RIK	241.20	253.50	239.45	461.90	569.42	529.51	2.12	0.00016	0.07378
1090180	Arhgdib	1187.16	1255.51	1424.35	2509.13	2708.39	2936.27	2.11038	0.00019	0.07378
6660707	Klhl5	529.25	557.07	523.35	332.03	242.45	239.44	0.49989	0.00079	0.08614
5270452	Tiam1	3596.56	3369.53	2392.86	1724.77	1283.19	1530.37	0.48882	0.00194	0.10631
1500768	Dcbld1	881.68	921.93	710.39	342.22	462.24	422.83	0.48747	0.00086	0.08804
1820224	Marcks	1643.34	1516.90	1400.90	690.10	769.78	743.61	0.48363	0.00016	0.07378
4880554	Cnih4	2341.05	2386.38	1591.44	872.82	975.52	1036.81	0.46306	0.00098	0.09054
6940068	X99384	1487.66	1339.79	1286.74	755.13	501.51	657.07	0.45951	0.00066	0.08413
450372	Bok	2243.17	1962.47	2329.19	842.41	945.04	1235.39	0.45776	0.00060	0.08413
6330445	Mogat2	698.27	716.26	772.59	307.48	344.58	344.01	0.45521	0.00008	0.07262
3190408	Mest	2336.51	2268.08	2692.25	814.69	1242.20	1239.92	0.44471	0.00091	0.09005
3440343	Cap1	952.25	998.10	550.70	310.26	330.02	393.64	0.42545	0.00225	0.11264
6350044	Slc2a1	7385.06	7320.20	4586.46	2559.53	2795.69	2645.31	0.42422	0.00083	0.08659
3290747	Bex2	3216.97	3492.91	2793.40	1537.72	1210.87	1237.43	0.41871	0.00016	0.07378
3990484	3110004L20Rik	1740.60	1768.32	1503.83	638.90	752.33	686.64	0.41	0.00007	0.06715
6450520	7530408C15Rik	4395.28	4543.87	5979.14	2324.44	2188.63	1465.30	0.40	0.00077	0.08610
6770356	Mest	7257.15	6954.87	7505.60	2006.52	2951.52	3837.83	0.39148	0.00100	0.09078
6290133	Endod1	1245.66	1289.43	950.12	398.92	462.24	488.60	0.38938	0.00015	0.07378
1690019	Mest	6712.69	6420.07	6332.16	1794.83	2574.71	3281.06	0.38159	0.00063	0.08413
2480059	Anxa5	1860.18	1826.27	1097.72	649.96	544.17	558.57	0.37557	0.00065	0.08413
2690435	Peg3	9670.34	9364.16	5634.74	2278.11	2916.39	3233.71	0.34789	0.00069	0.08413
160519	Tmem65	2148.18	2243.17	1723.06	640.14	658.88	770.88	0.33958	0.00004	0.06115
610092	Grm6	1041.87	1021.59	538.14	234.23	245.66	386.25	0.33855	0.00213	0.10902
4060521	Xist	2281.13	2399.39	1088.77	471.60	543.83	684.85	0.30890	0.00157	0.10062

Forty genes were selected for microarray validation. Shown are normalized intensities in both 9.5 and 10.5 dpc AGM-derived HSCs (n = 3), ratios (non-log fold-change), P values, and adjusted P values. Up- and downregulated genes (20 each) sorted by ratio are shown. Mpo and Mest genes, each ranked by three probes, were counted as one gene.

AGM, aorta-gonad-mesonephros; dpc, days postcoitum; HSC, hematopoietic stem cell.

Discussion

Genome-wide datasets have been generated to address how transcriptional networks govern numerous biological processes. Although microarray-based expression profiling is widely used for this purpose, the standard array does not provide information relevant to transcript levels. We therefore used CAGE to construct a global picture of the transcriptional landscape regulating HSPC development, including transcript levels. The CAGE shows that a transcriptional switch point occurs between 9.5 and 10.5 dpc in the AGM region.

Due to the rarity of HSPC samples, we could not generate replicates in CAGE. Based on this outcome, in analyzing rare samples, CAGE sequencing could be used for screening purposes to predict stage-specific transcripts, while microarray analysis could be useful for statistical evaluation regardless of TSSs.

Others have reported bias in the nonspecific guanine at the 5′ end of the CAGE tag [51]. However, such bias is unlikely to underlie the large number of novel unannotated TSSs in HSPCs. First, FANTOM5 CAGE tags are longer, which can multimap and form more stringent mapping procedure. In addition, the HeliScope CAGE protocol does not use PCR, so tags are not amplified. In FANTOM5, the HeliScope CAGE protocol generates a much longer CAGE tag (∼32 bases or longer) rather than 18–21 bases in FANTOM3, allowing unequivocal unique mapping of most tags. In addition, in FANTOM5, a more advanced probabilistic aligner known as Delve is used and only reports uniquely mapping tags.

We collected 10 developmental murine HSPC populations isolated from the AGM region, PL, FL, and BM. Principal component analysis of TSSs defined 4 HSPC groups among 10 samples (Fig. 3). Previously, Daley's group reported microarray analysis of gene expression during HSC ontogeny [52]. They collected embryonic and adult HSC samples from different stages and sites, in addition to embryonic stem (ES) cell-derived HSCs. Both of our studies demonstrate that HSPC specification occurs from 10.5 to 12.5 dpc, regardless of cellular location. Their study used CD150 as an HSC marker in FL and BM; thus clustering differences between our groups are likely due to cell surface phenotypes used for cell collection. In addition, they evaluated a 9.5 dpc yolk sac sample, whereas we assessed both 8.5 dpc mesoderm and 9.5 dpc AGM samples. Both of our studies suggest that dynamic gene expression changes occur in HSPCs from 9.5 to 10.5 or 11.5 dpc, as the transition from endothelial to hematopoietic HSPC phenotypes occurs [13,37], implying that HSPC commitment is programmed by this time point. Based on this data, it is likely that a major transcriptional switch occurs from 9.5 to 10.5 dpc.

To confirm this transcriptional switch, we performed microarray analysis using triplicate samples. Clustering analysis identifying changes in cytoskeletal genes suggests a transition from endothelial to hematopoietic HSPCs, an outcome consistent with prediction of CAGE sequencing and phenotypic changes observed in HSPCs [13]. Based on ratios, P values, and adjusted P values, we chose 40 differentially expressed genes to validate by real-time PCR. Among them, 23 were significantly altered (either upregulated or downregulated), demonstrating the utility of the dataset. The function of some of these genes remains unclear in hematopoiesis.

Overall, the dataset presented in this study should foster identification of novel genes involved in HSPC development and further our understanding of HSPC biology. Our work could also suggest novel approaches to culture and manipulate HSPCs in vitro or ex vivo in future studies.

Footnotes

Acknowledgments

The authors thank Miss. Yuka Tanaka and Yuka Horio for technical support, Dr. Elise Lamar for critical reading of the article, and the Ministry of Education, Culture, Sports, Science and Technology, the Ministry of Health, Labor and Welfare, and the Japan Society for the Promotion of Science for funding. K.S.T. is a recipient of a scholarship from the Tokyo Biochemical Research Foundation, Japan, and a MyPhD scholarship from the Ministry of Higher Education (MOHE), Malaysia. A.J. is a Chancellors Fellow at the University of Edinburgh. Dr. A.J.'s laboratory is supported by strategic funding from the Biotechnology and Biological Sciences Research Council. FANTOM5 analysis was made possible by a Research Grant to the RIKEN Omics Science Center from MEXT to Y.H. Y.H. is also a recipient of a grant for Innovative Cell Biology by Innovative Technology (Cell Innovation Program) from MEXT, Japan. The authors thank all members of the FANTOM consortium for contributing to sample generation and dataset analysis. The authors also thank GeNAS for data production.

Author Disclosure Statement

No competing financial interests exist.

References

Seita

and Weissman

. (2010). Hematopoietic stem cell: self-renewal versus differentiation. Wiley Interdiscip Rev Syst Biol Med, 2:640–653.

Mikkola

and Orkin

. (2006). The journey of developing hematopoietic stem cells. Development, 133:3733–3744.

Sugiyama

, Inoue-Yokoo

, Fraser

, Kulkeaw

, Mizuochi

and Horio

. (2011). Embryonic regulation of the mouse hematopoietic niche. ScientificWorldJournal, 11:1770–1780.

Sasaki

, Mizuochi

, Horio

, Nakao

, Akashi

and Sugiyama

. (2010). Regulation of hematopoietic cell clusters in the placental niche through SCF/Kit signaling in embryonic mouse. Development, 137:3941–3952.

Swain

, Inoue

, Tan

, Nakanishi

and Sugiyama

. (2014). Intrinsic and extrinsic regulation of mammalian hematopoiesis in the fetal liver. Histol Histopathol, 29:1077–1082.

Kim

, Wu

, Chen

, Xuan

, Jung

, Zhang

, Rowley

and Wang

. (2009). The transcriptome of human CD34+ hematopoietic stem-progenitor cells. Proc Natl Acad Sci U S A, 106:8278–8283.

Wilhelm

, Briau

, Austin

, Faubert

, Boucher

, Chagnon

, Hope

, Girard

, Mayotte

, et al. (2011). RNA-seq analysis of 2 closely related leukemia clones that differ in their self-renewal capacity. Blood, 117:e27–e38.

Lenhard

, Sandelin

and Carninci

. (2012). Metazoan promoters: emerging characteristics and insights into transcriptional regulation. Nat Rev Genet, 13:233–245.

Andersson

, Gebhard

, Miguel-Escalada

, Hoof

, Bornholdt

, Boyd

, Chen

, Zhao

, Schmidl

, et al. (2014). An atlas of active enhancers across human cell types and tissues. Nature, 507:455–461.

10.

Carninci

, Kasukawa

, Katayama

, Gough

, Frith

, Maeda

, Oyama

, Ravasi

, Lenhard

, et al. (2005). The transcriptional landscape of the mammalian genome. Science, 309:1559–1563.

11.

Forrest

, Kawaji

, Rehli

, Baillie

, de Hoon

and Haberle

and Lassman

and Kulakovskiy

and Lizio

, et al. (2014). A promoter-level mammalian expression atlas. Nature, 507:462–470.

12.

Kanamori-Katayama

, Itoh

, Kawaji

, Lassmann

, Katayama

, Kojima

, Bertin

, Kaiho

, Ninomiya

, et al. (2011). Unamplified cap analysis of gene expression on a single-molecule sequencer. Genome Res, 21:1150–1159.

13.

Mizuochi

, Fraser

, Biasch

, Horio

, Kikushige

, Tani

, Akashi

, Tavian

and Sugiyama

. (2012). Intra-aortic clusters undergo endothelial to hematopoietic phenotypic transition during early embryogenesis. PLoS One, 7:e35763.

14.

Kaufman

. The Atlas of Mouse Development. (1992). Elsevier Academic Press, London.

15.

Ottersbach

and Dzierzak

. (2005). The murine placenta contains hematopoietic stem cells within the vascular labyrinth region. Dev Cell, 8:377–387.

16.

Ema

and Nakauchi

. (2000). Expansion of hematopoietic stem cells in the developing liver of a mouse embryo. Blood, 95:2284–2288.

17.

Joshi

, Hannah

, Diamanti

and Gottgens

. (2013). Gene set control analysis predicts hematopoietic control mechanisms from genome-wide transcription factor binding data. Exp Hematol, 41:354–366.e14.

18.

, Krebs

, Choukrallah

, Keime

, Plewniak

, Davidson

and Tora

. (2011). seqMINER: an integrated ChIP-seq data interpretation platform. Nucleic Acids Res, 39:e35.

19.

Huang da

, Sherman

and Lempicki

. (2009). Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat Protoc, 4:44–57.

20.

Wilson

, Foster

, Wang

, Knezevic

, Schutte

, Kaimakis

, Chilarska

, Kinston

, Ouwehand

, et al. (2010). Combinatorial transcriptional control in blood stem/progenitor cells: genome-wide analysis of ten major transcriptional regulators. Cell Stem Cell, 7:532–544.

21.

Fan

, Bonde

, Gao

, Sotomayor

, Chen

, Mouw

, Zavazava

and Tan

. (2012). Dynamic HoxB4-regulatory network during embryonic stem cell differentiation to hematopoietic cells. Blood, 119:e139–e147.

22.

Tanaka

, Joshi

, Wilson

, Kinston

, Nishikawa

and Gottgens

. (2012). The transcriptional programme controlled by Runx1 during early embryonic blood development. Dev Biol, 366:404–419.

23.

Heinz

, Benner

, Spann

, Bertolino

, Lin

, Laslo

, Cheng

, Murre

, Singh

and Glass

. (2010). Simple combinations of lineage-determining transcription factors prime cis-regulatory elements required for macrophage and B cell identities. Mol Cell, 38:576–589.

24.

Hogart

, Lichtenberg

, Ajay

, Anderson

, Margulies

and Bodine

. (2012). Genome-wide DNA methylation profiles in hematopoietic stem and progenitor cells reveal overrepresentation of ETS transcription factor binding sites. Genome Res, 22:1407–1418.

25.

, Kibbe

and Lin

. (2008). lumi: a pipeline for processing Illumina microarray. Bioinformatics, 24:1547–1548.

26.

Bolstad

, Irizarry

, Astrand

and Speed

. (2003). A comparison of normalization methods for high density oligonucleotide array data based on variance and bias. Bioinformatics, 19:185–193.

27.

Gentleman

, Carey

, Bates

, Bolstad

, Dettling

, Dudoit

, Ellis

, Gautier

, Ge

, et al. (2004). Bioconductor: open software development for computational biology and bioinformatics. Genome Biol, 5:R80.

28.

Gentleman

, Carey

, Dudoit

, Irizarry

, and Huber

(2005). Limma: linear models for microarray data. In: Bioinformatics and Computational Biology Solutions Using R and Bioconductor. Smyth

, ed. Springer, New York.

29.

Saeed

, Sharov

, White

, Li

, Liang

, Bhagabati

, Braisted

, Klapa

, Currier

, et al. (2003). TM4: a free, open-source system for microarray data management and analysis. Biotechniques, 34:374–378.

30.

Nishikawa

, Nishikawa

, Hirashima

, Matsuyoshi

and Kodama

. (1998). Progressive lineage analysis by cell sorting and culture identifies FLK1+VE-cadherin+ cells at a diverging point of endothelial and hemopoietic lineages. Development, 125:1747–1757.

31.

Yokomizo

and Dzierzak

. (2010). Three-dimensional cartography of hematopoietic clusters in the vasculature of whole mouse embryos. Development, 137:3651–3661.

32.

Cumano

, Ferraz

, Klaine

, Di Santo

and Godin

. (2001). Intraembryonic, but not yolk sac hematopoietic precursors, isolated before circulation, provide long-term multilineage reconstitution. Immunity, 15:477–485.

33.

Yoshida

, Takakura

, Hirashima

, Kataoka

, Tsuchida

, Nishikawa

and Nishikawa

. (1998). Hematopoietic tissues, as a playground of receptor tyrosine kinases of the PDGF-receptor family. Dev Comp Immunol, 22:321–332.

34.

Kumano

, Chiba

, Kunisato

, Sata

, Saito

, Nakagami-Yamaguchi

, Yamaguchi

, Masuda

, Shimizu

, et al. (2003). Notch1 but not Notch2 is essential for generating hematopoietic stem cells from endothelial cells. Immunity, 18:699–711.

35.

Yoder

, Hiatt

, Dutt

, Mukherjee

, Bodine

and Orlic

. (1997). Characterization of definitive lymphohematopoietic stem cells in the day 9 murine yolk sac. Immunity, 7:335–344.

36.

Medvinsky

and Dzierzak

. (1996). Definitive hematopoiesis is autonomously initiated by the AGM region. Cell, 86:897–906.

37.

Rybtsov

, Sobiesiak

, Taoudi

, Souilhol

, Senserrich

, Liakhovitskaia

, Ivanovs

, Frampton

, Zhao

and Medvinsky

. (2011). Hierarchical organization and early hematopoietic specification of the developing HSC lineage in the AGM region. J Exp Med, 208:1305–1315.

38.

Taoudi

, Gonneau

, Moore

, Sheridan

, Blackburn

, Taylor

and Medvinsky

. (2008). Extensive hematopoietic stem cell generation in the AGM region via maturation of VE-cadherin+CD45+ pre-definitive HSCs. Cell Stem Cell, 3:99–108.

39.

Garcia-Porrero

, Manaia

, Jimeno

, Lasky

, Dieterlen-Lievre

and Godin

. (1998). Antigenic profiles of endothelial and hemopoietic lineages in murine intraembryonic hemogenic sites. Dev Comp Immunol, 22:303–319.

40.

Fraser

, Ogawa

, Yokomizo

, Ito

, Nishikawa

and Nishikawa

. (2003). Putative intermediate precursor between hematogenic endothelial cells and blood cells in the developing embryo. Dev Growth Differ, 45:63–75.

41.

Eichler

, Huang

and Ingber

. (2003). Gene Expression Dynamics Inspector (GEDI): for integrative analysis of expression profiles. Bioinformatics, 19:2321–2322.

42.

Davuluri

, Suzuki

, Sugano

, Plass

and Huang

. (2008). The functional consequences of alternative promoter use in mammalian genomes. Trends Genet, 24:167–177.

43.

Stamatoyannopoulos

, Snyder

, Hardison

, Ren

, Gingeras

, Gilbert

, Groudine

, Bender

, Kaul

, et al. (2012). An encyclopedia of mouse DNA elements (Mouse ENCODE). Genome Biol, 13:418.

44.

Phillips

and Corces

. (2009). CTCF: master weaver of the genome. Cell, 137:1194–1211.

45.

Horn

, Tesch

, Staib

, Kube

, Diehl

and Voliotis

. (1999). Expression of AC133, a novel hematopoietic precursor antigen, on acute myeloid leukemia cells. Blood, 93:1435–1437.

46.

Hirsch

, Iglesias

, Potocnik

, Hartmann

and Fassler

. (1996). Impaired migration but not differentiation of haematopoietic stem cells in the absence of beta1 integrins. Nature, 380:171–175.

47.

Sugiyama

, Kulkeaw

and Mizuochi

. (2013). TGF-beta-1 up-regulates extra-cellular matrix production in mouse hepatoblasts. Mech Dev, 130:195–206.

48.

Levy

, Kessler

, Pine

, Reich and

, Darnell

Jr. , (1988). Interferon-induced nuclear factors that bind a shared promoter element correlate with positive and negative transcriptional control. Genes Dev, 2:383–393.

49.

Essers

, Offner

, Blanco-Bose

, Waibler

, Kalinke

, Duchosal

and Trumpp

. (2009). IFNalpha activates dormant haematopoietic stem cells in vivo. Nature, 458:904–908.

50.

Joshi

, Van de Peer

and Michoel

. (2008). Analysis of a Gibbs sampler method for model-based clustering of gene expression data. Bioinformatics, 24:176–183.

51.

Zhao

, Valen

, Parker

and Sandelin

. (2011). Systematic clustering of transcription start site landscapes. PLoS One, 6:e23409.

52.

McKinney-Freeman

, Cahan

, Li

, Lacadie

, Huang

, Curran

, Loewer

, Naveiras

, Kathrein

, et al. (2012). The transcriptional landscape of hematopoietic stem cell ontogeny. Cell Stem Cell, 11:701–714.

53.

Bernt

, Zhu

, Sinha

, Vempati

, Faber

, Krivtsov

, Feng

, Punt

, Daigle

, et al. (2011). MLL-rearranged leukemia is dependent on aberrant H3K79 methylation by DOT1L. Cancer Cell, 20:66–78.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.24 MB

0.00 MB