Genome-wide analysis shows that Ldb1 controls essential hematopoietic genes/pathways in mouse early development and reveals novel players in hematopoiesis

Primitive erythroid and endothelial progenitors emerge almost simultaneously in mesodermal blood islands from a common hemangioblast progenitor¹  characterized by the expression of Flk1, the receptor for vascular endothelial growth factor (VEGF) (supplemental Figure 1). The other mesodermal-derived lineage expressing Flk1 is the cardiovascular progenitor cell, which is distinguishable from the hemangioblast by platelet-derived growth factor (PDGF) receptor expression. The hemangioblast has been studied in mouse, chick, and zebrafish embryos²  and its existence is supported by the observation that hematopoietic and endothelial progenitors share the expression of common genes.^3-7 

Deletion of Flk1 leads to embryonic lethality between 8.5 and 9.5 dpc, possibly because of the absence of yolk-sac blood islands.^8,9  The hemangioblast stage can be more specifically studied in vitro through the differentiation of embryonic stem (ES) cells into embryoid bodies (EBs), widely considered to be a relevant model of embryonic development. Within EBs, blast colony-forming cells (BL-CFCs) can generate blast colonies able to differentiate into hematopoietic and endothelial progenitors. They are therefore considered the in vitro equivalent of the hemangioblast.¹⁰ 

Recently, Li et al demonstrated that Ldb1 is also required during primitive hematopoiesis in the mouse for the development of megakaryocytes and the maintenance of long-term hematopoietic stem cells.^11,12 Ldb1 deletion is embryonic-lethal after 9.5 dpc, with severe developmental defects, including small size, anterior truncation, absence of heart/foregut structures, and impairment of yolk-sac hematopoiesis and vascular development.¹³  However, no functional studies or molecular mechanisms are known describing Ldb1 function in early hemangioblasts. Ldb1 molecular function has been well-studied in definitive erythropoiesis, where it plays an essential role by forming large transcriptional complexes with Gata1, Lmo2, Scl, E12/E47, HEB, Lyl1, E2-2, Cbfa2t3 (Eto2), Mtgr1, Cdk9, and Lmo4,¹⁴  binding E-box/GATA motifs genome-wide.¹⁵ 

In this study, we generated Ldb1^−/− ES cells and performed differentiation assays and genome-wide molecular and functional studies in BL-CFCs to unravel the function of Ldb1 in early primitive hematopoietic development. We showed that Ldb1 deletion actually results in 2 separate defects: a decreased number of hemangioblasts and their inability to differentiate further down the hematopoietic and endothelial lineages. To gain insight into the mechanism underlying these defects, we determined the direct and indirect targets of Ldb1 in these progenitor cells (in contrast to more advanced hematopoietic developmental stages). The genome-wide binding pattern of Ldb1 revealed that it is directly bound to a number of gene loci involved in mammalian hematopoietic development. Their expression is misregulated in the absence of Ldb1, placing it as one of the most upstream factors controlling hematopoietic development. Furthermore our study reveals novel genes and pathways that are potentially important for early hematopoietic/endothelial development and shows the complex interplay between transcription factors and signaling pathways in this very early primitive development stage, providing novel candidate targets that can be used to manipulate the differentiation of hematopoietic and/or endothelial cells.

This study was approved by Erasmus MC, Article 14.

Both Ldb1 alleles were inactivated using the same construct (supplemental Methods section, supplemental Figure 2A). After targeting the first allele, a single ES cell clone was transiently transfected with a Cre recombinase–expressing vector to delete exons 1/2 and the pMC-neo cassette, followed by targeting of the second allele. Southern blotting confirmed homologous recombination and deletion. Ldb1 protein levels were determined in nuclear extracts from Ldb1^+/+ and Ldb1^−/− ES cells with anti-Ldb1 N-18 (Cat. sc-11198, Santa Cruz Technologies, Santa Cruz, CA).

Ldb1^+/+ and Ldb1^−/− ES cells were grown in suspension at 10 000 cells/mL on nonadherent dishes in Iscove’s modified Dulbecco’s medium (without leukemia inhibitory factor) with 15% fetal calf serum (FCS), 1% P/S, 1% L-glutamine (Gibco, Cat.25030-08), 0.05 µg/mL transferin (Roche, Cat.652-202), 0.05 µg/mL ascorbic acid (Sigma, Cat.A-4544), and 3 µl/mL monothioglycerol (Sigma, Cat.M-6145) for day 4 EBs and 1.8 µl/mL for day 6/day 8 EBs. Five percent protein-free hybridoma medium II (Gibco, Cat.12040-077) was added for day 6/day 8 EBs.

Day 4 EBs were disrupted with trypsin–ethylenediamine tetraacetic acid and cells were transferred in methycellulose-based media with 10% FCS, 1% P/S, 1% L-glutamine, 0.25 µg/mL transferin, 0.25 µg/mL ascorbic acid, 2 µl/mL monothioglycerol, 0.01 µg/mL mIL6 (R&D Systems, Cat.406-ML), and 0.005 µg/mL hVEGF (R&D Systems, Cat.293-VE) for 3 days. Colonies were scored according to morphology and number under an inverted microscope.

Ldb1^+/+ and Ldb1^−/− day 6 EBs were disrupted with 2.5% collagenase. Cells were transferred in methycellulose-based media with 10% FCS, 1% P/S, 1% L-glutamine, 0.25 µg/mL transferin, 0.25 µg/mL ascorbic acid, 2 µl/mL monothioglycerol, 5% protein-free hybridoma medium II, 0.01 µg/mL mIL6, 0.001 µg/mL IL3 (R&D Systems, Cat.403-ML), 0.005 µg/mL hIL11 (R&D Systems, Cat.418-ML), 0.003 µg/mL GM-CSF (R&D Systems, Cat.415-ML), 4 U/mL EPO (R&D Systems, Cat.959-ME), 0.005 µg/mL TPO (R&D Systems, Cat.488-TO), and 0.1 µg/mL SCF (R&D Systems, Cat.455-MC). Red primitive erythroid colonies and white macrophage colonies (composed of round cells growing in clumps) were identified microscopically after 6 days according to morphology and color.

Day 4 EB single-cell suspensions were labeled with Flk1-PE clone Avas 12α1 (BD Pharmigen, Cat.555308) in 1% bovine serum albumin/phosphate-buffered saline. Dead cells were excluded by Hoechst 33258 (Molecular Probes). Flk1⁺ cells were sorted from day 4 EBs on FACSAria using the Diva 5.1 software (BD Pharmingen). Day 6/8 EB single-cell suspensions were labeled in 1% bovine serum albumin/phosphate-buffered saline with CD41-PE clone MWReg30 (Santa Cruz, Cat.sc-19963) or CD31-FITC clone MEC 13.3 (BD Pharmigen, Cat.553372). Dead cells were excluded with 7-AAD (Invitrogen). Flow cytometry (FACS) was performed on FACScan (Becton Dickinson) and analyzed with Cell Quest.

Day 4 EBs were disrupted with dissociation buffer (StemPro Accutase, Invitrogen). Dead cells were removed by Lymphoprep separation. Magnetic-assisted cell sorting was carried out with autoMACS Pro Separator (Miltenyi Biotec) using Anti-PE MicroBeads (Miltenyi Biotec) and the Flk1-PE clone Avas 12α1 (BD Pharmigen) antibody. Purity was assessed by FACS.

Chromatin immunoprecipitation (ChIP) sequencing of endogenous Ldb1 (using anti-Ldb1 antibody N-18, Santa Cruz) on Flk1⁺ BL-CFCs isolated from day 4 EBs was performed as previously described.^15,16 

RNA was isolated from Ldb1^+/+ and Ldb1^−/− Flk1⁺ cells with the QIAGEN RNeasy Mini Kit, and integrity was checked on the Agilent 2100 Bioanalyzer.

RNA was converted to biotin-labeled cRNA, hybridized on the Mouse Genome 430 2.0 Array, and analyzed with the Affymetrix GeneChip Scanner 3000 according to the manufacturer’s protocol.

RNA sequencing was performed on Illumina HiSeq 2000 platform according to the manufacturer instructions.

For ChIP/RNA sequencing and microarray bioinformatic analysis see the supplemental material online.

ChIP/RNA sequencing and microarray datasets described in this study have been deposited in the NCBI Gene Expression Omnibus (Accession Number: GSE43044).

Generation of the Ldb1^−/− mouse is presented in supplemental Figure 2A-D. The first striking observation in the embryos is the complete absence of any blood or vasculature from the extra-embryonic yolk-sac, which also fails to completely surround the whole embryo, showing that erythropoiesis and vascular development are severely impaired in the absence of Ldb1 (Figure 1A, in agreement with Li et al).¹¹ Ldb1^+/− mice were viable, able to breed, and did not exhibit any of the phenotypic characteristics of the Ldb1^−/− mice (not shown¹¹ ).

We additionally generated an Ldb1^−/− ES cell line to carry out differentiation and molecular and functional studies in hemangioblast-equivalent cells. The second Ldb1 locus allele in Ldb1^+/− ES cells was deleted (supplemental Figure 2E) and the absence of Ldb1 protein was confirmed by Western blotting and immunofluoresence (supplemental Figure 2F-G). The morphology and growth of Ldb1^−/− ES cells appeared similar to normal ES cells (not shown and Li et al).^11,17  We then tested their differentiation potential into EBs. Because Ldb1^+/− mice did not exhibit any phenotypic differences compared with Ldb1^+/+ mice, they were not used for further investigations. Ldb1^+/+ and Ldb1^−/− EBs did not show any phenotypic differences at days 3 and 4 and appeared similar in size and structure (Figure 1B). However, after 8 days we observed a striking absence of erythroid clusters within the Ldb1^−/− EBs (Figure 1B), correlating with the hematopoietic defect observed in the knock-out mouse. Of note, Ldb1^−/− EBs survived at least until day 10 without apparent apoptosis but were more adherent than the wild-type ones. After 10 days the majority was found attached to the bottom of the culture dish (not shown). CFC assays were performed to test the potential of Ldb1^−/− EBs to give rise to primitive erythroid and macrophage colonies. Both bright-red primitive erythroid colonies and white macrophage colonies, present in Ldb1^+/+ day 6 EB cultures, were absent in the Ldb1^−/− cultures. This defect was fully rescued by exogenous reexpression of Ldb1 (Figure 1C-D), indicating that those effects are specifically caused by the absence of Ldb1. In addition, the presence of primitive erythroid progenitors was examined by flow cytometry using CD41 (integrin alpha-IIb) as marker.¹⁸ Ldb1^+/+ EBs contained 2.6% and 12% of CD41⁺ cells at days 6 and 8, respectively, whereas Ldb1^−/− EBs did not contain any CD41⁺ erythroid progenitor cells in accordance with the absence of primitive erythroid colonies in Ldb1^−/− CFC cultures (Figure 1E). The presence of endothelial progenitors in day 6 and day 8 EBs was also examined using CD31 (PECAM-1) as a marker, because Ldb1^−/− embryos exhibit a severe defect in vascular development.¹⁹  Days 6 and 8 Ldb1^+/+ EBs contained 21% and 16% CD31⁺ cells, respectively. These cells were also found to be present in Ldb1^−/− EBs at the same time points but decreased to less than half when compared with the wild-type ones (Figure 1E).

We then investigated whether BL-CFCs, the in vitro hemangioblast equivalent, were present in the Ldb1^−/− EBs. Day 4 BL-CFCs were isolated using the VEGF receptor Flk1 as a marker (supplemental Figure 1). Flk1⁺ BL-CFCs were present in both Ldb1^+/+ and Ldb1^−/− EBs; however, a striking difference was observed because Ldb1^−/− EBs repeatedly contained half the number of Flk1⁺ cells (Figure 1F). In three independent experiments, BL-CFC assays also showed that blast colonies did form in the wild-type but not in the knockout cultures (Figure 1G). In conclusion, BL-CFCs are present in Ldb1^−/− EBs but fail to generate fully-grown blast colonies. Ldb1^−/− Flk1⁺ cells eventually die in culture (not shown).

To gain insight into the gene expression network and signaling pathways involved in the maintenance of the hemangioblast in vitro–equivalent cells, transcriptome analysis of Flk1⁺ BL-CFCs by RNA sequencing showed 12 468 expressed genes. Pathway enrichment analysis of the expressed genes revealed the signaling pathways potentially active in these progenitor cells (Table 1, Column A).

To identify the direct genomic targets of Ldb1 in Flk1⁺ BL-CFCs, ChIP sequencing in sorted Flk1⁺ cells showed 4252 significant Ldb1 binding peaks. Of all genes expressed in Flk1⁺ cells, more than one quarter, or 3264, genes contain Ldb1 peak(s) in their vicinity. The group of Ldb1-bound genes were found to participate in approximately half of the signaling pathways active in BL-CFCs (Table 1, Columns A vs B).

We compared the Ldb1 ChIP-sequencing dataset of primitive Flk1⁺ BL-CFCs to the Li et al dataset in more committed adult hematopoietic Lin^– bone marrow progenitors.¹¹  This revealed that the majority of the 4252 Ldb1-binding events in Flk1⁺ cells is specific to these early progenitor cells (79.6%, or 3384 binding events) (Figure 2A). Of note, among those specific binding events, a few (respectively, 35 and 168 events) have also been detected in previously published Ldb1 ChIP-sequencing datasets from fetal liver or definitive erythroleukemic cells.¹⁵  We therefore excluded those 203 binding events from the 3384 Flk1⁺-specific events for further investigation. The resulting 3181 Flk1⁺-specific binding events are referred to as “Set1” and will be compared with “Set2”, the common binding events between Flk1⁺ cells and Lin-bone marrow cells (20.4% or 868 Ldb1-binding events in Flk1⁺ cells).

We then performed a functional study to identify genes and associated signaling pathways regulated by Ldb1 in BL-CFCs. Day 4 Ldb1^−/− Flk1⁺ cells were used for microarray and RNA sequencing. The corresponding gene expression profiles were compared with the profile of wild-type Flk1⁺ cells. Of the 12 468 genes found to be expressed in Flk1⁺ cells, 2675 genes were identified as differentially expressed between the Ldb1^+/+ and the Ldb1^−/− Flk1⁺ cells (supplemental Table 1). Of those, 1424 genes were downregulated (53.23%) and 1251 were upregulated (46.77%). The differentially expressed genes were classified into 3 independent sets: Set1*, Set2*, and Set3, respectively, depending on whether they were included in the Ldb1-bound genes of Set1, the Ldb1-bound genes of Set2, or they were unbound in Flk1⁺ cells (Figure 2B). Interestingly, the majority of Ldb1 direct targets (Set1* and Set2*) were observed downregulated genes; 65% (483 genes) were downregulated and 35% (259 genes) were upregulated in Set1*, and 69.3% (149 genes) were downregulated and 30.7% (66 genes) were upregulated in Set2*. This trend is absent in indirect target genes (Set3), which present 46.1% (792 genes) and 53.9% (926 genes) of down- and upregulated genes, respectively. The fact that the majority of direct target genes is downregulated as a result of Ldb1 deletion shows that Ldb1 primarily acts as a positive transcriptional regulator in BL-CFCs. However, the one-third of upregulated genes in this group suggests an additional role for Ldb1 complexes as repressors of transcription. A search for enriched DNA-binding motifs carried out on sequences targeted by Ldb1 in Set1 revealed the known E-box(Tal1):Gata motif (motif 5) in only 3.6% of peaks (114 of 3181) (Figure 3A,C). The motifs found with the highest frequency were those associated with Klf transcription factors (motif 1: 513 peaks out of 3181 or 16%), SP1 (motif 2: 569 peaks or 18%), and PU.1 (motif 3: 333 peaks or 10%). In addition, 3% of Ldb1 peaks from Set1 are associated with a CTCF motif (motif 8: 98 of 3181 peaks) (Figure 3A,C). Conversely, in Set2 peaks (common with adult bone marrow progenitor peaks), a prominent E-box:Gata motif was found (motif 1: 419 peaks of 868 or 48%) as described,^11,15  and no CTCF motif was detected (Figure 3B,C). This suggests that Ldb1 forms different complexes in primitive Flk1⁺ cells, targeting different regulatory sites. Alternatively, Ldb1 complex activity may be regulated by different cofactors in these cells as suggested by the differences in recognition motifs, potentially explaining the different repressing and activating transcriptional functions of Ldb1 in BL-CFCs.

Among the misregulated genes in Ldb1^−/− Flk1⁺ BL-CFCs directly bound by Ldb1 (Set1*), we found genes previously shown to participate in the regulation of mouse embryonic development and hematopoiesis such as Gata2,^20,21 Gata1,²² Runx1,²³ Fli1,²⁴ Lyl1,²⁵ Id2, Id4,²⁶ Klf4,²⁷ Gfi1b,²⁸ Cbfa2t3/Eto2,²⁹ Myb,³⁰ Dkk1, Sfrp1,^31,32 Meis1,³³ Raldh2,³⁴  and Fgfr2.³⁵  The changes observed in some of those essential genes were confirmed by real-time polymerase chain reaction (supplemental Figure 3). Of interest is the downregulation of Stat5a and Sox7, Sox17, and Sox18 all having been reported as hematopoietic regulators.^36-38 Figure 4 shows 4 examples of hematopoietic-essential Set1* genes. Because BL-CFCs can differentiate into both the hematopoietic and endothelial lineages, we further analyzed the hematopoietic- and endothelial-specific genes.³⁹ Figure 5 shows that the majority of these is downregulated in the absence of Ldb1, illustrating the role of Ldb1 as an activator of transcription and to a lesser extent as a repressor in hematopoiesis and endothelial development (like the larger group of Set1* containing those genes, see Figure 2B).

To decipher the essential functions of Ldb1 in the maintenance/differentiation of BL-CFCs, we performed a pathway enrichment analysis on the total number of genes misregulated in Flk1⁺ cells in the absence of Ldb1 (Table 1, Column C). Approximately one-third (37 of 127) of the pathways enriched for genes expressed in Flk1⁺ cells are misregulated in the Ldb1^−/− Flk1⁺ cells. To distinguish between direct and indirect functions of Ldb1, we carried out in parallel pathway enrichment analyses on the Ldb1 direct-target genes (Set1*; Table 1, Column D) and the indirect Ldb1-regulated genes (Set3; Table 1, Column E). Approximately half of the misregulated pathways in the Ldb1^−/− Flk1⁺ cells were directly targeted by Ldb1 (Table1, Column C vs Column D). Among these, acute myeloid leukemia, basal cell carcinoma, interleukin signaling pathway, melanoma, and T-cell receptor signaling pathways were found. Four examples of perturbed signaling pathways are provided, illustrating how the genes are affected by the absence of Ldb1 (supplemental Figures 4-7). Signaling pathways such as WNT (Wingless), MAPK (mitogen-activated protein kinase), hedgehog, ErbB, focal adhesion, integrin, insulin, and angiogenesis were commonly enriched for both direct and indirect genes of Set1* and Set3. Not unexpectedly, we also found myeloid and lymphoid specific pathways in that context. In contrast, the VEGF-, epidermal growth factor (EGF)-, platelet-derived growth factor (PDGF), and endocytosis signaling pathways were enriched only in the indirect Set3 genes compared with Set1*.

In conclusion, the large number of pathways that are (potentially) affected in the Ldb1^−/− Flk1⁺ BL-CFCs easily explains the severe phenotype of the Ldb1^−/− embryos. Importantly, the fact that some Flk1⁺ cells still emerge during Ldb1^−/− ES cell differentiation suggests that most of these pathways may play some role in hemangioblast formation but become essential at subsequent developmental stages before the hematopoietic and endothelial lineages diverge.

Primitive hematopoiesis begins in the yolk-sac after the migration of brachyury expressing mesodermal cells through the primitive streak and their differentiation toward hematopoietic or endothelial precursors that collectively form blood islands. Hematopoietic precursors give rise to primitive erythroblasts and endothelial precursors to the vasculature. Both are absent in Ldb1^−/− embryo yolk-sacs, questioning the role of Ldb1 in this process.

Many of the misregulated factors identified in the gene-expression profiling of Flk1⁺ BL-CFCs from day 4 Ldb1^+/+ and Ldb1^−/− EBs act in the nuclear “response” of well-known signaling pathways. WNT signaling has a positive role in regulating primitive hematopoiesis.^32,40  Addition of the inhibitor Dkk1 (downregulated in the Ldb1^−/− Flk1⁺ cells) blocks the pathway, reducing the potential of BL-CFCs to generate colonies.³¹  Both WNT and Notch signaling pathways are involved in the differentiation of hemangioblast cells toward the primitive erythroid lineage. WNT signaling is active in Flk1⁺ BL-CFCs during the first hours of differentiation toward blast colonies, whereas Notch signaling remains inactive through Numb inhibition. Between 12 and 24 hours of differentiation, the inhibitory effect exerted on Notch signaling is reduced and its activation leads to the expression of WNT inhibitors blocking the WNT pathway.³¹  Both pathways were found to be active in Flk1⁺ cells; however, only WNT signaling was perturbed in Ldb1^−/− Flk1⁺ cells. We propose that WNT misregulation can be attributed to changes in expression of genes bound by Ldb1. In Drosophila, the importance of interplay among Notch, fibroblast growth factor (FGF), MAPK, and EGFR signaling pathways has been documented as a requirement for blood specification,⁴¹  with the last 2 pathways affected by the deletion of Ldb1 in Flk1⁺ cells.

Changes in gene expression of the WNT signaling–associated transcription factors Lef1, Tcf7, and Tcf7l1 in Ldb1^−/− Flk1⁺ cells confirm the importance of this pathway in early hematopoiesis. The downregulation of Etv2/Er71 in knockout cells leads to the same conclusion because it acts downstream of not only the WNT but also the bone morphogenic protein (BMP) and Notch pathways.^42,43  Also, expression of the transforming growth factor-beta (TGF-β) signaling receptors, endoglin and Alk-1 (Acvrlk1), was reduced in Ldb1^−/− Flk1⁺ cells. ES cell knockout and rescue experiments showed that these receptors are important for hemangioblast development and primitive hematopoiesis.⁴⁴  Recently, Ldb1 has been implicated in cell migration and focal adhesion through an interaction with the Ste20-like kinase (SLK), a microtubule-associated protein necessary for migration.⁴⁵  Our data show that expression of a number of focal adhesion proteins is affected, although SLK did not appear to change. Among the misregulated genes in Ldb1^−/− Flk1⁺ BL-CFCs, we found Stat5a, Sox7, Sox17, and Sox18. Interestingly, Stat5a overexpression in differentiating ES cells leads to an increase in hematopoietic progenitors.³⁶ Sox7, Sox18, and Sox17 participate in different stages of hematopoietic development in the mouse embryo, with Sox7 and Sox18 being involved at the earliest stages of yolk-sac hematopoiesis and Sox17 participating at later stages with the emergence of long-term repopulating hematopoietic stem cells.^37,38 

Ldb1 acts in a core complex with the essential hematopoietic regulators Scl, Lmo2, and Gata1/Gata2 and associates with other essential factors such as Runx1.^14,15  When Scl is absent, primitive hematopoiesis is impaired and hematopoietic colonies are absent in culture, similar to Ldb1^−/− embryo phenotype.^20,46,47  In vitro generation of Scl^−/− EBs further showed that this factor is dispensable for hemangioblast generation, but is essential for its commitment toward hematopoietic and endothelial lineages.⁴⁸ Gata1-deficient embryos fail to generate erythroid cells, but Gata1-null ES cells can form other hematopoietic lineages.^49,50 Gata2 deficiency shows reduced yolk-sac hematopoiesis and reduced number of ES cell–derived colonies. Finally, Lmo2 absence results in a phenotype similar to that of Scl and Gata2, although macrophage colonies were still present in colony assays.⁴⁷  Our data show that all of these factors are regulated by Ldb1, placing it as a key regulator of the transcription factor network at the origin of the hematopoietic system. Accordingly, Ldb1 deletion in the mouse results in a more severe hematopoietic phenotype than that observed with the factors mentioned before. The additional defect observed in endothelial development can also be explained by the EB results. There are still some CD31⁺ endothelial cells in the Ldb1^−/− EBs, but their number is severely reduced. We conclude that the defect in primitive hematopoiesis in the absence of Ldb1 occurs at a stage in development before the hematopoietic and endothelial lineages diverge, i.e., Ldb1 is involved in the proliferation and differentiation of Flk1⁺ BL-CFCs, resulting in half the number of BL-CFCs that do not develop into blast colonies when compared with normal.

It is not yet clear how the hemangioblast develops from the embryonic mesoderm and how differentiation into hematopoietic and endothelial cells is regulated. Our data show that this must be a process orchestrated by a number of signal transduction pathways and transcription factors, many of which are directly or indirectly affected by the loss of Ldb1. One of the best-studied transcription factors in hematopoietic development, Runx1, is affected by the absence of Ldb1. It is expressed at low levels after the first 24 hours of blast colony formation and increases as the first hematopoietic progenitors of the erythroid and macrophage lineages emerge.³¹  Although Runx1^−/− blast colonies and primitive hematopoietic colonies are still able to grow,²³  its absence results in the emergence of fewer colonies from EBs.³¹  Thus, downregulation of Runx1 in the Ldb1^−/− EBs would be yet another cause of the decrease seen in the number of BL-CFCs. A large number of key hematopoietic transcription factors is misregulated in the absence of Ldb1, showing that this factor is critically required to properly orchestrate the gene expression network of hemangioblasts and their progeny (Figure 6).

A number of factors are also upregulated, and it is interesting to note that expression of the mesodermal marker brachyury in Ldb1^−/− BL-CFCs and Ldb1^−/− EBs is upregulated (data not shown). Brachyury expression is lost within the first 24 hours of blast colony formation³¹  and our data show that it is suppressed by Ldb1. The brachyury upregulation also suggests that Ldb1^−/− BL-CFCs have not yet lost their mesodermal identity/properties and are therefore unable to differentiate into blast colonies.

Finally, our data also show that a number of genes involved in cardiovascular development are affected, suggesting that Ldb1 is also important in cardiovascular development, presumably in complex with the cardiac Gata factors.

In summary, Ldb1 was identified as an essential regulator of mouse hemangioblast proliferation and differentiation. The severity of the hematopoietic phenotype is the result of a decreased number of hemangioblast cells and their inability to differentiate further down the hematopoietic and endothelial lineages. The expression of essential transcription factors is directly affected by Ldb1, revealing its crucial role in the early hematopoietic/endothelial regulatory networks (Figure 6). In addition, the identification of Ldb1 target genes in Flk1⁺ cells reveals a number of novel pathways contributing to hematopoietic/endothelial development and provides better understanding of the early developmental steps of these lineages.

The authors thank the members of the Grosveld Laboratory for technical help and discussion. Cell sorting was done by Reinier van der Linden.

This work was supported by NWO (NL), the Dutch Academy of Sciences (KNAW), the EU (the integrated projects EUTRACC and SyBoss), and the European People Marie Curie actions Program (Marie Curie European Reintegration Grants ERG, Call: FP7-PEOPLE-2010-RG).

Contribution: A. Mylona, C.A.-S., and F.G. designed the research. A. Mylona, C.A.-S., A. Martella, E.S., and R.J. carried out the research. C.K. carried out Illumina sequencing. S.T. and J.H. performed bioinformatics analyses. W.v.I. and B.L. supervised Illumina data processing and bioinformatics analyses. A. Mylona, C.A.-S., E.S., S.T., and F.G. wrote the manuscript.

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

The current affiliation for A. Mylona is the Department of Hematology, Erasmus MC, Rotterdam, The Netherlands.

Correspondence: Frank Grosveld, Department of Cell Biology, Erasmus MC, Dr. Molenwaterplein 50, 3065 GE, Rotterdam, The Netherlands; e-mail: f.grosveld@erasmusmc.nl; correspondence for computational biology: Boris Lenhard, Imperial College London, London SW7 2AZ, United Kingdom; e-mail: b.lenhard@imperial.ac.uk.