During development, hemogenesis occurs invariably at sites of vasculogenesis. Between E9.5 and E10.5 in mouse, endothelial cells in the caudal portion of the dorsal aortae generate hematopoietic stem cells 1, 2 and are referred to as hemogenic endothelium 3-8. The mechanisms by which hematopoiesis is restricted to this domain, and how the morphological transformation from endothelial to hematopoietic is controlled are unknown. We show here that HoxA3, a gene uniquely expressed in the embryonic but not yolk sac vasculature, restrains hematopoietic differentiation of the earliest endothelial progenitors, and reverts the earliest hematopoietic progenitors into CD41-negative endothelial cells. This reversible modulation of endothelial-hematopoietic state is accomplished by targeting key hematopoietic transcription factors for downregulation, including Runx1, Gata1, Gfi1B, Ikaros, and PU.1. Through loss-of-function, and gain-of-function epistasis experiments and the identification of antipodally regulated targets, we show that among these factors, Runx1 is uniquely able to erase the endothelial program set up by HoxA3. These results suggest both why a frank endothelium does not precede hematopoiesis in the yolk sac and why hematopoietic stem cell generation requires Runx1 expression only in endothelial cells.
Homeobox genes play an important role during vertebrate development, determining cell identity along the rostral-caudal axis. In early development, they are expressed along this axis in a manner colinear with their order within the Hox cluster 9. In some tissues, these rostral-caudal expression patterns are reset during organogenesis; for example, several Hox genes have been identified in hematopoietic stem cells, and the expression patterns change with differentiation 10-13. The expression of members of Hox paralog group 4, including HoxB4, promotes self-renewal of the HSC 14-16 and allows engraftment of early embryonic hematopoietic progenitors produced in vitro from ES cells 17. Hox paralog group 3, the next anterior, is best known for specifying the identity of tissues that originate from the pharyngeal arches, however it also plays an important role in endothelial development. Mice lacking HoxA3 display cardiovascular abnormalities 18 and vascular expression of paralog group 3 members is associated with angiogenesis and wound repair 19-21. Growing evidence suggests that the embryonic vasculature gives rise to hematopoietic progenitors via a specialized hemogenic endothelium 3, 22, 23. The posterior dorsal aorta is such a specialized site that plays a particularly important role in the origin of the adult HSC pool through a poorly understood process that is dependent on Runx1 5. The vascular role of paralog group 3 prompted us to evaluate the temporal expression of HoxA3 in the hemogenic domains of the dorsal aorta. In early development (E7.5), HoxA3 expression is restricted to the embryo proper, while Runx1 expression marks hematopoiesis in the yolk sac (Sup. Fig. 1A, B). HoxA3 and Runx1 display a remarkable pattern of mutually exclusive expression at hematopoietic/vascular sites in the early embryo. At E8.25 and E8.5, HoxA3 is highly expressed throughout the neurectoderm and mesenchyme, is present at intermediate levels in the endothelium of both the anterior and posterior dorsal aortae, and remains absent from the yolk sac (Fig. 1A, C and Sup. Fig. 1C, E). By contrast, Runx1 is found in the yolk sac but not in the aortae or other embryonic vessels (Fig. 1B, D and Sup. Fig. 1D, F) with the exception of the omphalomesenteric artery, as seen previously 24, 25. Notably, HoxA3 is not expressed in the omphalomesenteric artery (Fig. 1C). At E9.5, and E10.5 the time point at which the posterior dorsal aortae first begin to express Runx1 and become hemogenic, HoxA3 expression is clearly lost in the aortic endothelial cells (Fig. 1E, G Sup. Fig. 1G, I), while Runx1 is expressed (Fig. 1F, H, Sup. Fig 1H,J).
HoxA3 down-regulation thus marks the site of hemogenesis in the endothelium of the dorsal aorta. Does HoxA3 expression have any functional effect on the development of hemogenic endothelium? In order to answer this question, we generated a doxycycline (dox)-inducible HoxA3 murine ES cell line by cassette exchange recombination into a dox-inducible locus 16 and differentiated these cells as embryoid bodies (EBs). The kinetics of mesoderm differentiation in this system broadly mimics that of embryonic development 26 with bipotent hematopoietic-endothelial progenitors (hemangioblasts) identified in clonal assays as early as 2.75 days of differentiation 27, corresponding to embryonic bipotent progenitors of the posterior primitive streak, thought to contribute to yolk sac hematopoiesis 28. Vascular markers, eg. VE-cadherin, Tie2, and CD31, first appear two days later, coexpressed on many cells with the earliest hematopoietic marker, CD41 29, 30. The coexpressing population is capable of both endothelial and hematopoietic differentiation, thus defining it as hemogenic endothelium 31, 32. When we induced HoxA3 with dox just before this time (day 4-6), we noted a striking repression of the hematopoietic markers CD41+ and CD45+ (Fig. 2A, B). However, the total endothelial progenitor population identified as cells expressing both Flk1 and VE-cadherin 33, 34 (F/V population) was not reduced by HoxA3 expression (Fig. 2A, B). We assayed hematopoietic progenitor content in these EBs and found that HoxA3 dramatically suppressed hematopoietic colony-forming cell (CFC) content, (Fig. 2C) demonstrating that HoxA3 is not merely preventing expression of surface markers, but truly preventing hematopoietic differentiation. When hematopoietic progenitors (c-Kit CD41 double-positive cells) from uninduced EBs were sorted and plated in CFC assays, HoxA3 expression in the methylcellulose medium abolished hematopoietic colony-forming potential (Sup. Fig. 2A). To determine whether the hematopoietic repression of HoxA3 was due to cell death or a change in cell fate, hematopoietic (c-Kit+/CD41+; K/41), and endothelial (Flk1+/VE-cadherin+; F/V) fractions were purified from day 6 EBs and cultured on OP9 stromal cells, a system that support both hematopoiesis and endothelial development. In the absence of doxycycline both the F/V and K/41 fractions produced hematopoietic cells, consistent with the notion that the endothelial fraction is endowed with hemogenic capacity 32 (Fig. 2F, G, no dox). However when HoxA3 was upregulated, hematopoietic marker expression was significantly reduced (Figure 2F G, + dox). Remarkably, in the presence of doxycycline, not only were hematopoietic cells missing from the K/41 fraction, but colonies of cells with an epithelial morphology and expressing VE-cadherin were observed instead (Fig. 2E). The induction of endothelial markers and repression of hematopoietic markers was seen also in more committed progenitors already expressing the pan-hematopoietic marker CD45 (Sup. Fig. 2B). When HoxA3 expression was withdrawn, hematopoietic colonies developed again, in both K/41 and F/V-initiated cultures (Fig. 2D-G). This result shows that HoxA3 restrains hematopoietic development and maintains an endothelium, even in progenitors that have recently committed to hematopoiesis, indicated by expression of CD41 and CD45.
To test the effect of HoxA3 in bona fide hemogenic endothelium 6-8, we expressed HoxA3 with an ires-GFP reporter by retroviral transduction in disaggregated E10.5 AGM tissue cultured ex vivo. Both GFP+ hematopoietic and GFP+ adherent colonies could be detected in cultures transduced with control vector, however with the HoxA3 vector, only GFP-negative hematopoietic colonies were found while GFP+ colonies were adherent (Figure 2H). Flow cytometric analysis confirmed that cells expressing HoxA3 had elevated Flk-1 and VE-cadherin expression, and reduced CD41 and CD45 expression (Fig. 2I). Thus HoxA3 is able to repress hematopoietic development in both ES-derived hematopoieitic progenitors and in embryonic AGM-derived tissue.
To gain insight into the mechanism by which HoxA3 regulates the hemogenic transition we performed transcriptional profiling on both the endothelial and hematopoietic progenitor fractions from day 6 EBs treated with doxycycline for 6 hours (Fig. 3). In the hematopoietic (K/41) fraction HoxA3 regulated 585 genes of which 277 were repressed and 308 were upregulated (1.5 fold p-value 0.05 cut-off Fig. 3A). Because HoxA3 promotes the endothelial phenotype in this fraction and inhibits hematopoiesis, we evaluated HoxA3 upregulated genes based on their expression in control hematopoietic or endothelial fractions. Genes involved in endothelial development clustered together, were expressed at higher levels in the control F/V fraction compared to the control K/41 fraction, and were generally upregulated by HoxA3 in the dox-treated K/41 fraction (Fig. 3B). We performed a similar analysis for the F/V fraction in which 832 genes were regulated (421 down and 411 up; 1.5 fold cut off p-value 0.05, Fig 3A). Genes repressed by HoxA3 included many known hematopoietic regulators most of which are expressed at higher levels in control K/41 cells compared to control F/V cells (Fig. 3C). Within this cluster HoxA3 significantly repressed the expression of Runx1, Gata1, Gfi1B, Ikaros and Phemx and PU.1, all validated by qPCR (Fig. 3D). HoxA3 thus coordinately regulates a large number of genes involved in endothelial and hematopoietic development, preventing the hematopoietic program from arising in endothelial progenitors and reactivating an endothelial program in nascent hematopoietic progenitors. It is the first regulatory factor identified to date to have this effect.
One of the notable downregulated targets is Runx1. Loss of function analysis of Runx1 demonstrated it to be essential for definitive hematopoiesis 35, 36 and for the emergence of hematopoietic cells from the aorta 24. Conditional deletion of Runx1 has shown that its role is essential in VE-cadherin-positive cells of the aorta, but dispensable in the (Vav-1-expressing) hematopoietic stem cells themselves 5. To determine the functional relevance of Runx1 and other hematopoietic transcription factors repressed by HoxA3, we performed dominant epistasis experiments. Endothelial (F/V) progenitor cells from HoxA3-induced EBs were sorted and transduced with Runx1B, Ikaros, Pu.1, Gata1, Gfi1B, or control vectors bearing ires-GFP reporters, and cultured on OP9 stroma. Surface marker analysis of GFP-gated cells demonstrated that Runx1 strongly suppressed, GATA1 partially suppressed, and Gfi1B minimally suppressed the HoxA3 phenotype, while Pu.1 and Ikaros had no effect (Fig. 4A,B). Transcriptional analysis showed that Gfi1B, PU.1, Ikaros, and Phemx, were actually target genes of Runx1, but that GATA1 was not (Fig. 4C). To understand this epistasis, we sorted Runx1- or GATA1- transduced, HoxA3-expressing cells, and determined their global gene expression profiles in comparison to HoxA3-expressing cells (with control GFP vector), or cells in which HoxA3 expression was turned off by doxycycline withdrawal. Consistent with their ability to revert the HoxA3 phenotype, high level overexpression of both Runx1 and GATA1 reverted many of the gene expression changes induced by HoxA3, particularly those genes that were repressed by HoxA3, including Gfi1B and other hematopoietic transcription factors (Fig. 4D). However, Runx1 was more potent than GATA1 in repressing genes that were upregulated by HoxA3, including many key endothelial regulatory factors, as well as regulators of adhesion and polarity (Fig 4E).
We tested putative HoxA3 recognition sequences within the Runx1 locus by ChIP and found that several, including a conserved sequence upstream of the P2 enhancer, were sites of direct binding (Sup. Fig. 3A,B). We then tested whether the loss of HoxA3 would lead to derepression of Runx1 in vivo by performing in situ hybridization with Runx1 probe on HoxA3 mutant embryos 37. At E8.5, Runx1 expression was never detected in the dorsal aortae of wild-type (0/24) or heterozygous (0/41) embryos, however in a significant number of null embryos (14/29), we observed precocious expression of Runx1 in endothelial cells of the dorsal aorta, and occasionally in excess hematopoietic cells within the aortic lumen (Fig. 5A), demonstrating that HoxA3 represses Runx1 in vivo.
These data demonstrate that HoxA3 sits at the apex of a cascade of regulatory factors, such that the spatiotemporal regulation of HoxA3 expression determines where and when hemogenic potential arises within embryonic vessels. HoxA3 represses hemogenesis in part by blocking the expression of Runx1, which would otherwise activate numerous downstream transcription factors to promote hematopoietic development. This finding has implications for the two modes of hematopoiesis in early embryonic development: yolk sac mesoderm, which does not express HoxA3, undergoes synchronous hematopoietic and endothelial differentiation, while hematopoietic differentiation from lateral plate mesoderm, which expresses HoxA3, is delayed, and occurs by budding from a well-defined endothelium. It is notable that Runx1 was shown to be required only transiently, and in the endothelial cells, not the derived hematopoietic cells, of the AGM 5. The expression profile of cells expressing both HoxA3 and Runx1 indicates that an essential role of Runx1 is to extinguish an endothelial program. This is a transient requirement, and not necessary in the yolk sac where endothelial and hematopoietic cells develop concurrently, not sequentially, but required in the AGM where hematopoietic cells emerge from an endothelium. Our model is summarized in Fig. 5B: endothelial progenitors produced in lateral plate mesoderm are inhibited from further differentiation towards hematopoeisis by HoxA3 which represses Runx1 together with other key hematopoietic transcription factors. HoxA3 is then extinguished within a subset of endothelium, allowing hemogenesis to occur. Although the presence of a hemogenic endothelium has been appreciated for almost a century 38, the spatiotemporal patterning of the embryonic endothelium into hemogenic and non-hemogenic domains, and the regulation of the endothelial-hematopoietic transition have remained inscrutable. The data presented here demonstrates that HoxA3 plays a critical role in both processes by acting as a gatekeeper at the apex of the transcriptional hierarchy restricting entry into the hematopoietic program.