|Home | About | Journals | Submit | Contact Us | Français|
Arteriovenous (AV) differentiation is a critical step during blood vessel formation and stabilization. Defects in arterial or venous fate lead to inappropriate fusion of vessels, resulting in damaging arteriovenous shunts. While many studies have unraveled the molecular underpinnings that drive AV fate, surprisingly, the spatiotemporal emergence of arteries and veins in mammalian embryos remains unknown. Here, we examine artery and vein specification and differentiation during vasculogenesis. We show that the first intraembryonic vessels formed are arteries, which differentiate in a stepwise manner. By contrast, veins emerge later, progressively forming after embryonic turning. In addition, we demonstrate that hemodynamic flow is not required for arterial specification, but is required for maintenance of select arterial markers. Together, our results provide a first spatiotemporal analysis of mammalian AV cell fate establishment and anatomy, as well as a delineation of a molecular toolkit for analysis of arteries and veins during early vessel development.
The cardiovascular system, the first and most critical organ network to develop in vertebrates, is composed of two distinct but interconnected networks of arterial and venous blood vessels. Although these fundamentally different types of blood vessels have been recognized for thousands of years (since Erasistratus, 304–250 B.C.), we have only begun to understand their ontogeny at the cellular and molecular levels. Many questions remain unanswered, especially in mammalian systems. When and how do blood vessel endothelial cells (ECs) undergo arteriovenous (AV) ‘specification’ (initial acquisition of AV fate) or ‘differentiation’ (functional AV identity)? How and when are the two separate networks assembled and subsequently interconnected? Is AV fate dependent on endothelial morphogenesis or hemodynamic circulation? Is AV fate an on-off switch, or is fate acquired incrementally? In part, the dearth of information regarding mammalian AV development is likely due to the small size of embryos and to the speed at which the initial vascular plexus is formed (via de novo blood vessel formation, or vasculogenesis) and expanded (via sprouting and remodeling, or angiogenesis) (Risau and Flamme, 1995). Indeed, rapid morphogenesis of the early embryonic vasculature, occurring in a matter of hours, has resulted in our failure to understand much of the intervening dynamic anatomy and molecular signature of vessels, including the first arteries and veins.
Studies in avian and fish systems have addressed the underlying mechanisms that modulate AV cell fate. Two key ideas have emerged: 1) angioblasts (endothelial progenitors) are specified/predetermined to contribute to either arteries or veins as they emerge in the mesoderm, prior to blood vessel formation, and 2) AV identity within a vessel is affected by hemodynamic flow, and is thus plastic. One study in the chick chorioallantoic membrane illustrated the first idea, as yolk sac blood island ECs were shown to express AV markers before blood flow was established (Herzog et al., 2005). In addition, zebrafish studies underlined the genetic basis for AV fate when they identified an important role for Notch signaling during arterial specification (Lawson et al., 2002; Zhong et al., 2001). Other studies have favored the second idea. In chick-quail chimeras, early arterial or venous endothelial grafts demonstrated the capacity of these cells to contribute to either arterial or venous host vessels and to adapt their fate to the local vessel environment and its flow parameters (le Noble et al., 2004). Together, these studies, whose conclusions need not be mutually exclusive, suggested that AV fate is determined early within angioblasts, but that it is plastic during a window of developmental time, and depends in part on hemodynamics. The question remains open as to whether these findings hold true in mammalian embryos.
To study AV fate, it is critical to delineate basic embryonic vascular anatomy and determine exactly when, where and how the first embryonic vessels form. Such detailed analyses have been carried out in fish and avian model systems. Examination of the developing zebrafish vasculature has catalogued and identified most early vessels in great detail, including arterial and venous networks, in part due to useful vascular reporters and embryonic transparency (Blum et al., 2008; Herbert et al., 2009; McKinney and Weinstein, 2008). In chick, the development of the first arteries, the dorsal aortae, has received recent attention, with breakthrough studies addressing their formation and patterning (Reese et al., 2004). While a number of studies in mammals (mouse) have provided spatiotemporal analysis of embryonic blood vessels (Drake and Fleming, 2000; Walls et al., 2008) studies of mammalian AV fate, anatomy and molecular signature are still needed.
Here, we demonstrate that mouse vessels form and differentiate into arteries and veins in a stepwise manner. We show that embryonic vessels emerge as blind-ended vessels, progressively extending, forming tubes and acquiring their AV fate. The first intraembryonic vessels, the dorsal aortae, progressively display expression of a growing subset of arterial markers prior to blood circulation. At this stage, other vessels exhibit ambiguous AV fate. By contrast, the first veins form and acquire their fate later than the first arteries, during and after embryonic turning (almost a full day later). In addition, using both explanted embryonic tissues and flow defective mutant embryos, we demonstrate that hemodynamic flow is not required for arterial specification but rather for maintenance of select arterial gene expression. Together, our data questions the criteria by which we assign AV fate and reveals unexpected aspects of initial blood vessel development. This comprehensive analysis shows that acquisition of AV fate in the mammalian embryonic vasculature is incremental, and provides a timeline of AV gene expression onset to guide future studies for mouse models of cardiovascular malformations.
To characterize the developing anatomy of the embryonic mammalian vasculature, including the newly forming arteries and veins, we utilized the Flk1-LacZ mouse line (Shalaby et al., 1995). Flk1 (VEGFR2) is a tyrosine kinase receptor for VEGFA and is expressed in all ECs, including angioblasts (Yamaguchi et al., 1993). This line allowed us to visualize all blood vessels, from their initial aggregation of angioblasts -first into cords, and then patent vessels- to their subsequent remodeling/arborization and extension. Because yolk sac vessel remodeling has been characterized by many groups (Jones et al., 2008; Lucitti et al., 2007), we focused on intraembryonic vessels which were identified based on their relative anatomical morphology (Kaufman, 1992; Nelsen, 1953).
At the 0 somite stage (0s, at embryonic day 7.5 or E7.75), angioblasts emerged first in the extraembryonic yolk sac plexus, and were not detected within the embryo proper (Fig. 1A). Just hours later, at 1s, two key intraembryonic vascular features became distinguishable in the embryo: two parallel tracts of angioblasts, presaging the paired dorsal aortae, and a thin crescent of angioblasts along the developing anterior intestinal portal (AIP) termed the ‘cardiac crescent’ (Fig. 1B,C). This cardiac crescent contains precursors to the heart and its endothelial lining, the endocardium. By 2s, pre-aortic angioblasts formed cohesive, but lumenless cords (inset, Fig. 1D). By 3s, aortic cords opened central lumens and formed the first patent embryonic vessels (Fig. S1). By 5s, dorsal aortae displayed expanded and continuous lumens (inset, Fig.1E).
We identified vein primordia based on anatomical location (Kaufman, 1992) and noted that they appeared shortly after aortae formation. Just posterior and lateral to the cardiac crescent, two short tracts of angioblasts along the yolk sac-embryo interface also formed cords, at the future location of the sinus venosus/heart inflow tracts (Fig. 1E, F). Similar to the primitive aortae, these vessels comprised blind-ended vessels, patent only anteriorly in proximity to the sinus venosus, (Fig. S1). Notably, these ‘partial’ vessels were the first and only anatomically distinguishable veins at this early time point, as we could detect neither cardinal or ophalomesenteric veins in any embryonic location. Thus, during vasculogenesis, the first intra-embryonic vessels present in the embryo were the dorsal aortae, followed by the sinus venosus/cardiac inflow tracts.
By 8s, cardiac crescent angioblasts formed the primitive heart tube (pre-endocardium) and the associated sinus venosus (Fig. 1G). By this stage, a heartbeat could be visually detected. The dorsal aortae remained two paired vascular tubes, which by this stage extended anteriorly into the cranial mesoderm and posteriorly to the tail and allantois. At 9s, the embryo initiated turning (Fig. 1H,I). During this embryonic morphogenetic transformation, the embryo turned towards its right, as the AIP and caudal intestinal portal (CIP) began to constrict towards the future umbilical cord. This process both forms the gastrointestinal tract and transforms the cup-shaped embryo into a more familiar fetal body shape.
During turning, the primary vascular plexus underwent dramatic changes. At the onset of turning, we observed progressive extension of the vitelline vein primordial (vitelline veins are vessels that will eventually drain blood from the yolk sac). While our static observations could not resolve whether this extension was the result of vasculogenesis or angiogenesis, scattered angioblasts appeared to be recruited into the forming vessels, implying EC aggregation. Sinus venosus and pre-vitelline angioblasts aligned in two tracts, posterior to the heart tube, from 2s to 9s (Fig. 1D, E, G, H, blue arrowheads), and by 11s (Fig. 1J) formed two lumenless cords along the interface of the embryonic body wall and the yolk sac. Halfway through turning, the paired dorsal aortae were prominent and established vessels, of relatively wide diameter. By contrast, the forming vitelline veins at this time remained thin, cordlike vessels. However, we continued to observe no evidence of cardinal vein formation (Fig. 1J).
Following turning, the largest vessels in the embryo were the dorsolateral paired dorsal aortae and endocardium. However, by this time, we began to distinguish the cardinal veins along the flank of the embryo. At 13s, the common cardinal veins extended laterally from the sinus venosus, towards the ventral aspect of anterior somites. The anterior cardinal veins also became apparent as extensions from the common cardinal vein, towards the head (Fig. 1K, L). Interestingly, unlike the aortae, the cardinal vein never transitioned through an obvious cord phase, but instead initially presented as a loose plexus. In addition, aggregating angioblasts could also be seen posterior to the sinus venosus, at the expected location of the future posterior cardinal veins. Together, this spatiotemporal analysis of initial blood vessel formation delineated the stepwise anatomical emergence of arteries and veins, where the dorsal aortae form first, then the vitelline veins, and finally the cardinal veins.
To correlate molecular AV differentiation with our anatomical observations, we examined expression of known AV markers (Eichmann et al., 2005; Rocha and Adams, 2009) during vasculogenesis. Analysis of AV marker expression in early vessels was carried out to establish whether angioblasts are specified before they aggregate into vessels, or if vessels form before their fate is determined.
We first examined the expression of all standard arterial markers, such as ephrinB2 (efnB2), Cx40, Cx37, Hey1, Hey2, Nrp1, Notch1, Notch4, Dll4 and Jag1 during early vessel formation at E8.0 (4–5s stage). Surprisingly, ECs of the paired dorsal aortae at this stage lacked any hint of most arterial markers assayed, including ephrinB2-LacZ, Cx40, Hey2, Nrp1, Notch1, Notch4, or Jag1 (Fig. 2A, B, E, F, G, H, J). Weak Cx37 and Hey1 aortic expression was detected however (Fig. 2C, D), and Dll4 was the only robust arterial marker in the forming dorsal aortae at this stage (Fig. 2I). It was interesting that although the dorsal aortae were patent vessels by 4–5s (Fig. 1E), most arterial markers tested were not detected in aortic ECs at this stage. Also surprising is that many arterial markers (Cx40, Cx37, Hey1/2, Nrp1, Notch4, and Dll4) were expressed in the cardiac crescent and the future inflow tracts within the sinus venosus.
Slightly later, select arterial markers became detectable in the dorsal aortae, while others remained weak. EphrinB2, Cx40, Cx37, Hey1, Nrp1, Notch1, Notch4, and Dll4 initiated aortic expression by E8.25 (Fig. 2A′–D′, F′–I′). Sections of E8.25 aortae revealed expression in ECs, and in some cases, surrounding mesenchymal cells (Fig. 2A″–D″, F″–I″ and Fig. S2). In contrast, Hey2 and Jag1 lacked EC expression at this stage (Fig. 2E′, E″, J′, J″). Again, surprisingly, many reportedly arterial markers were also expressed in the sinus venosus, a region that by nature would be predicted to be ‘venous’. This suggested that sinus venosus ECs may not yet have acquired venous identity at this stage, as they exhibit a partial arterial molecular signature. Indeed, most standard arterial markers examined were not restricted to arterial endothelium (dorsal aortae), if expressed there at all, but were expressed in other ECs including those of the sinus venosus, yolk sac and endocardium.
Shortly before embryonic turning, some arterial markers finally became arterial-enriched, while others still had not yet initiated endothelial expression. By this stage, ephrinB2, Cx40 and Cx37 were robustly expressed in the aortae (Fig. S3A–C). By contrast, Hey1, Nrp1, Notch1, Notch4 and Dll4 expression was downregulated in the aortae (Fig. S3D, F–I), while Hey2 and Jag1 were never expressed in the aortae, even at this late time-point (Fig. S3E, J). Of note, many of these markers were also downregulated in the sinus venosus over time. This may suggest that ECs of the sinus venosus are initially ambiguous in their AV fate, displaying markers of both fates, and only secure their venous identity later. Together, our observations demonstrate that stepwise arterial differentiation during vasculogenesis can be detected with a subset of established arterial markers, but that there is some ambiguity in fate at certain timepoints in certain vessels.
To determine onset of venous specification, we assessed venous markers during vasculogenesis. Overall, none of the venous markers examined displayed robust vascular expression at this early timepoint (Fig. 3A–E). EphB4, Nrp2 and Flt4 exhibited faint expression in the short inflow tracts of the sinus venosus (Fig. 3A, C, E). However, most genes examined had prominent expression in non-endothelial tissues (Fig. S2). CoupTFII, Nrp2, and APJ, for instance, were expressed in somites and lateral plate mesoderm (Fig. 3B, C, D). Surprisingly, although Flt4 was faintly expressed in the vitelline vein primordia, it was also transiently enriched in the dorsal aortae (Fig. 3E), suggesting possible early ambiguity in fate in aortic ECs. The delay in venous expression initiation, compared to markers of arterial fate, suggest that venous specification initiates after arterial specification.
By E8.25, expression of venous markers changed only marginally. Nrp2 and Flt4 became more prominently expressed in angioblasts that will later contribute to the vitelline veins, located at the border between the embryo and the yolk sac (blue arrows, Fig. 3C′, E′). EphB4 expression also increased in scattered cells of the sinus venosus (Fig. 3A′). However, overall at this stage, most venous markers examined were ambiguous, faint and widely expressed in non-endothelial tissues. The only easily identifiable veins at this stage remained the inflow tracts of the sinus venosus, which were identifiable by anatomical location and low levels of venous marker expression.
By E8.5, many of the venous markers tested became expressed in pre-vitelline vein angioblasts, near the posterior aspect of the sinus venosus and along the interface of the embryonic and yolk sac tissues (Fig. S3K–O). CoupTFII and Nrp2 were also strongly expressed in the somites at this stage. Of note, by this stage Flt4 expression became downregulated in the dorsal aortae and EphB4 became more robust in the sinus venosus.
Together, our results indicate that few tools are available to identify early endothelial cells as unambiguously venous. We propose that the absence of clear venous identity, as assessed by lack of venous marker expression, is due to the anatomical absence of major bona fide veins at this early stage. While anatomic texts (Kaufman, 1992; Nelsen, 1953) mention the presence of paired ‘head veins’ at these early stages (axial vein present along the cranial neural tube), we observed no expression of standard venous markers in the head region. The large arteries of the primary vascular plexus are thus the first differentiated vessels to form during vasculogenesis, while the first major veins (vitelline veins) form later.
To analyze AV vessel identity with greater resolution, we used immunofluorescence against the earliest AV markers (identified above) for which antibodies were available. We examined Cx40, Nrp1 and Nrp2 expression in embryos bearing a Flk1-EGFP allele. These markers readily distinguished arteries from veins in established vascular beds (later during development), with Cx40 and Nrp1 in arteries and Nrp2 in veins (Fig. S4 and data not shown), as previously reported (Ema et al., 2006).
We then examined AV fate establishment, both during and after vasculogenesis. At E8.25, Cx40 was primarily restricted to the arterial endothelium of the dorsal aortae (Fig. 4A–A″), as expected. Similarly, Nrp1 was expressed by aortic ECs (Fig. 4B–B″). Interestingly, low levels of Cx40 (Fig. 4A′) and Nrp1 (Fig. 4B′) were observed in smaller blood vessels, including the paired ‘head veins’ (non-arterial expression indicated with white arrowheads). In addition, ECs of the sinus venosus also expressed low levels of both arterial proteins (as seen in Fig. 2). We propose that early veins exhibit ambiguous AV identity and have not yet stabilized venous identity at this point. By contrast, the venous marker Nrp2 was expressed in sinus venosus endothelium (Fig. 4C–C″), as expected, and was not expressed within aortic ECs. Together, these observations suggested that the AV identity of the first veins is at first promiscuous, as they express arterial markers inappropriately.
One day later, similar analyses revealed that the AV identity of the largest vessels became solidified, while that of the secondary vessels still lagged behind. By E9.5, Cx40 was robustly and exclusively expressed in the dorsal aortae, with no trace detected in veins (Fig. 4D–D″). Similarly, Nrp2 endothelial expression was restricted to the sinus venosus ECs (Fig. 4F–F″). Nrp1, by contrast, was strongly expressed in the dorsal aortae but still displayed low levels in the sinus venosus and head vein of the PNVP (perineural vascular plexus) (Fig. 4E–E″), further suggesting that veins at this stage still possessed some level of ambiguity in their AV fate. Overall, these data demonstrate the stepwise and progressive restriction of fate involved in the establishment of initial vessel identity, which occurs earlier in arteries than in veins.
Because we failed to identify the expected array of veins in pre-turning embryos, we examined venous genes both during and after turning. Between 6 to 10s, we found scattered angioblasts, posterior to the sinus venosus and at the location of the future vitelline veins (Fig. 5A–C). This observation suggested that vitelline veins might arise from aggregation of angioblasts at the border of the embryonic and yolk sac tissues (Fig. S5).
Prior to turning, a wide flat endothelial tube could be observed lateral to the sinus venosus, at the expected position of the common cardinal vein (or Duct of Cuvier), as early as 8s (asterisk, Fig. 5B, C, and data not shown). Following turning, angioblast cords could be distinguished at the expected position of the cardinal veins, laterally along each flank of the embryo. At 14s, a loose but narrow plexus extended anteriorly from this vessel into the head (white arrow, Fig. 5D), constituting the first evidence of cardinal vein formation. Shortly thereafter, at 16s, the anterior cardinal vein remodeled into a more recognizable vessel and acquired a central lumen. During this time, the posterior cardinal vein also appeared and presented as a loose and partial cord of angioblasts, extending from the common cardinal vein towards the tail (Fig. 5E). By 20s, both the anterior and posterior cardinal veins were distinguishable vessels, running the length of the anteroposterior axis, albeit extensively interconnected to the flank vascular plexus (Fig. 5F).
The sequence of artery and vein emergence thus suggests a stepwise process (Fig. S6). Arterial specification occurs earlier than veins, in the angioblasts of the paired dorsal aortae. Veins, however, appear shortly thereafter, as pre-vitelline vein extensions from the ventroposterior aspect of the cardiac crescent (later sinus venosus). Following turning (9–11s), the cardinal veins appear as a lateral vessel plexus, which remodel into more cohesive vessels over time. Arteries and veins thus form differently and differentiate progressively, over the period of approximately a day.
Once we identified forming veins during and after turning, we examined venous markers at these later embryonic stages. We confirmed that Nrp2, APJ and Flt4 were expressed in early vitelline angioblasts (Fig. 6A–C), and further found that APJ and Flt4 marked the forming cardinal and sprouting intersomitic veins (Fig. 6D–F). Expression was detected near the common cardinal vein, at the level of the heart along the axis, and extended anteriorly/posteriorly following turning. EphB4-LacZ expression was observed much later in the common cardinal following turning (Fig. 6G), and expanded along the forming cardinal veins in time (Fig. 6H). By 20–22s, EphB4-LacZ could be observed along most of the anterior cardinal vein, and the nascent posterior cardinal vein (Fig. 6I). Thus, although the venous vessels are first present during early stages (2s, see Fig. S1), full differentiation as assessed by EphB4 expression, was not established until after turning.
As AV differentiation appeared to initiate prior to, or concurrent with, the known onset of blood flow, we asked whether AV marker expression depended on hemodynamic flow. We carried out explant studies to assess arterial fate in the absence of blood flow. To generate “flowless” embryonic explants, we dissected and isolated the anterior and posterior halves of 5s embryos (Fig. S7A,B). These embryonic halves were then cultured for 10 or 18 hours. At 5s, arterial fate had been specified in the dorsal aortae, as assessed using Cx40 and Dll4 (Fig. S7C), which both display early, robust and largely specific arterial expression at this early stage (Cx40, Fig. S8).
Using 5s Flk1-LacZ embryos as controls, at time 0, we established that the dorsal aortae were lumenized vessels by this stage, both in the anterior and posterior portions of the embryonic axis (Fig. 7A, D and data not shown). Following 10 hours in culture, both the dorsal aortae and cardiac crescent largely maintained their morphology in culture (Fig. 7B,E). After 18 hours of culture however, the aortae broke up into endothelial pockets (Fig. 7C,F). This was further accentuated in the posterior explant (Fig. 7F). Of note, by this stage in vivo, the embryo would normally be completely turned and the aortae would soon fuse, however in the explants the two aortae remained widely separated.
We assessed arterial fate in these explants and found that Cx40 was dramatically downregulated after 10hrs, and extinguished by 18 hours (Fig. 7A′–C′). In the posterior region, Cx40 expression was largely gone in aortae after 10hrs, although it remained expressed at low levels in yolk sac vessels (Fig. 7E′). However, after 18 hours, Cx40 was completely undetectable (Fig. 7F′). By contrast, Dll4 remained robustly expressed after 10 hours, in both anterior and posterior aortae, and only slightly diminished after 18 hours (Fig. 7A″–C″,D″–F″). Using the same experimental conditions, we also tested venous fate in the “flowless” vessels of cultured embryonic explants. In our explant cultures, we expected some level of expression of venous markers in the sinus venosus, however EphB4-LacZ explants exhibited no detectable positive cells either at the onset of culture or later (Fig. S7D).
These results support the contention that flow is not required for arterial specification, but rather for the expression of subsets of AV markers. While flow is clearly required for maintenance of certain arterial markers (i.e. Cx40), it is not required for others, such as Dll4. From these experiments however it is unclear what role blood flow plays in venous differentiation, as measurable venous marker expression initiates later than the developmental timeframe covered in these experiments.
To test whether blood flow was required for arterial specification in vivo, we assessed our panel of arterial markers in Rasip1 null embryos that we have previously reported lack all vascular lumens, and consequently lack all blood circulation (Xu et al., 2011). As previously described, these embryos display normal angioblast specification, patterning and differentiation. Similarly, we observed that Flk1-LacZ expressing angioblasts were present and normally distributed in Rasip1 null embryos (data not shown), however they failed to undergo the normal cord to patent vessel transition and appeared as cords rather than tubes.
We performed in situ hybridization on Rasip1 mutant embryos, using Cx40 and Dll4 to test whether arterial specification occurred normally in the absence of blood flow. Assaying Cx40 expression, we detected faint and discontinuous expression of Cx40 in the dorsal aortae at early stages (E8.25, Fig. 8A,B). As development proceeded, however, expression of Cx40 rapidly declined. By E9.0, no trace of Cx40 could be detected in the lumenless aortic cords or other vessels, although low levels of expression could be seen in the vitelline artery (Fig. 8C,C′,D,D′). This corroborated our ‘flowless’ in vitro explant experiments. Interestingly, again similar to the explants, we observed relatively normal or increased levels of Dll4 expression in Rasip−/− aortic cords (Fig. 8E–H′). This is in agreement with the increase in Dll4 expression previously observed in flowless mouse models (Jones et al., 2008).
Together, these data show that at least one arterial gene (Cx40) appeared to depend on blood flow, while another one (Dll4) was expressed regardless of blood flow status. These results support the contention that flow is not required for initial arterial specification, but is required for expression of the full range of arterial genes, suggesting differentiation and/or maintenance of arterial fate in blood vessels depends on hemodynamic flow.
Our spatio-temporal analysis of initial murine blood vessel formation shows that anatomical emergence and molecular differentiation of embryonic arteries and veins both occur in an incremental, stepwise manner. To date, over 80 mouse mutants have been identified as having “vascular remodeling defects” (M.Dickinson, personal communication) and usually incurring early lethality, yet there has been little effort to delineate their exact vascular defects, or possible AV defects. This is likely due to a dearth of useful characterized markers at these early stages and to the complex, rapidly changing initial embryonic vasculature. With this comprehensive analysis of mouse vascular development, we provide a useful guide for such studies. In this report, we show that the primary vasculature consists of only a few major vessels, including the main embryonic arteries (dorsal aortae) and incomplete vein primordia (sinus venosus/pre-vitelline veins). In addition, we find that most widely used AV markers are not useful for recognizing early arteries and veins during vasculogenesis. Acquisition of AV marker expression occurs in a stepwise manner, with only a subset requiring blood flow for maintenance. While studies in mouse have characterized aspects of the early vascular anatomy (Nelsen, 1953; Walls et al., 2008), none has correlated AV fate acquisition with vascular development or defined its molecular signature. This work is thus a first examination of AV ‘onset’ in mammalian vessels, which correlates AV gene expression with vessel emergence.
To assess AV fate onset, we first delineated the anatomy of the embryonic vasculature. The first mouse vessels form via the aggregation of angioblasts around E8.0, as previously described (Drake and Fleming, 2000), however we made the striking finding that these vessels emerge in a progressive and stereotyped manner. In addition, we noted that vessels ‘open up’ lumens in an anteroposterior manner. Indeed, the dorsal aortae and vitelline veins initially appear as solid cords, but then quickly transform (aortae first), forming lumens proximally near their connections to the heart, but not distally. Surprisingly, this type of ‘blind ended’ vessel emergence has not been previously detailed. For almost a day, these vessels grow in this conformation, in the absence of any type of closed circulation. Our static observations of fixed tissues do not distinguish whether these vessels form via bona fide vasculogenesis (angioblast aggregation), or by angiogenic sprouting. Live imaging of mouse embryos carrying endothelial reporters will be useful to answer these questions.
While a large body of work exists on the molecular basis for arteriovenous fate (Herbert et al., 2009; Kim et al., 2008; Lawson et al., 2001; Wang et al., 1998; You et al., 2005), we have focused on the ontogeny of arteries and veins at the earliest stages of mammalian vasculogenesis. A principal finding of our studies is that AV fate is acquired in a stepwise, rather than ‘all-or-none’, manner. Few standard arteriovenous markers are initially expressed in early vessels, but this subset incrementally grows over time. In addition, co-expression of arterial and venous markers in early vessels suggests an early ambiguity or ‘toggling’ in endothelial fate, which resolves as vessels become established. We propose that arteries solidify their fates slightly earlier than veins do, possibly as a direct consequence of their earlier formation and continuity with the heart and hemodynamic pressure.
These findings raise questions as to the true nature of ‘artery’ versus ‘vein’ fate. What comprises the molecular signature that definitively characterizes arterial or venous endothelium? Is expression of just one marker, say efnb2 or Nrp1, enough to define a vessel as having arterial identity, or is the full range of markers required? We would argue that the answer is simply semantic, but that it is important to acknowledge that there is a range of arteriovenous ‘states’.
We have found that assignment of embryonic vessel identity is a difficult task. While there is growing awareness that endothelium is highly heterogeneous in diverse adult tissues (Aird, 2007), few studies have examined the emergence of different types of endothelial cells in the embryo, including those of arteries and veins. This study identifies a toolkit of markers that are useful for identifying the first arteries and veins, as well as resolving the timing of their specification (onset of AV identity) and differentiation (functional AV identity). These tools will be critical for analysis, both future and retrospective, of vascular mutant mouse embryos.
The role of flow on many aspects of vascular development has long been a confounding parameter when studying the development of blood vessels. Many cardiovascular mutants display a range of defects that are secondary to anatomical abnormalities, including alterations in blood flow. Shear stress and stretch, caused by local blood flow conditions, are sensed by endothelial cells that line vessels and these physical forces drive morphological, as well as transcriptional, changes within these cells (Andersson et al., 2005; le Noble et al., 2004).
It is therefore important to establish the timing of intraembryonic hemodynamic circulation initiation, the paths taken by blood during initial circulation and the exact contributions of blood flow or pressure in each new emerging embryonic vessel as it forms (Jones et al., 2004; Jones et al., 2002; Lucitti et al., 2007). A number of reports identify plasma circulation in murine embryos beginning as early as the 3 somite stage (Lucitti et al., 2007). At this stage, the heart myocardium has begun to beat, however few erythrocytes are identifiable and primarily plasma is thought to flow. Another study speculates that although blood cells and flow are present in the embryo after heart formation, a full functional circulatory loop is not complete until E10.0 (McGrath et al., 2003). Interestingly, it is still unclear when a true physical intraembryonic circulatory loop forms, as no studies have characterized AV vascular circuits or the anatomy of the first intraembryonic blood flow in mammalian embryos during vasculogenesis.
We propose that intraembryonic circulatory loops do not form until well after initial vasculogenesis. Our observations show that during this time, the main arteries have formed and yet there are no veins to complete the circuit. Given that the heart begins to beat prior to circulatory loop formation, the question arises as to the path taken by blood/plasma, which is pumped by the nascent heart. We postulate that vessels at this time experience pulsative pressure, rather than true hemodynamic flow. It is even possible that this pressure contributes to initial vascular lumen formation and/or maintenance. Alternatively, a circulatory loop may in fact exist, however it may return blood to the heart not via embryonic veins (as these are not yet formed), but instead via the yolk sac plexus that is contiguous with the developing embryonic vasculature.
The possibility that our analyses of AV gene expression were conducted prior to onset of true embryonic circulation brings into question the role of blood flow during artery or vein formation and differentiation. To query whether blood flow played a role in early AV fate, we assessed our toolkit of AV markers in the presence or absence of hemodynamic flow. Because the timeframe of our studies was early and relatively short (vasculogenesis), we were limited to assessing arterial differentiation, as veins develop later.
We find that when flow is abrogated (in explants) following a short period of flow, Cx40 expression is initiated, but soon extinguished. Similarly, when flow never initiates, in Rasip1−/− embryos (as lumens never open), Cx40 expression initiates, but then rapidly diminishes. Dll4 expression, on the other hand, shows no downregulation under either set of conditions. These results suggest that some ‘arterial’ genes depend on flow while others do not, again bringing into question what constitutes ‘true’ arterial cell fate. If we consider Dll4 expression to represent at least partial arterial identity (a first step on the road to being ‘arterial’), then we can infer that arterial specification does not require flow. Expression of more than one marker could then be considered as being farther ‘along the road’ to full differentiation, and we can thus reason that differentiation does require flow.
Regarding the downregulation of Cx40 in no-flow conditions, we note that while some previous reports observed dependence of Cx40 expression and function on flow (Ebong et al., 2006), others did not (Jones et al., 2008). In the latter study, no-flow mouse embryos were created by clipping the inflow tracts to the heart, and in these embryos Cx40 expression declined slightly, but not significantly (Jones et al., 2008). We propose that perhaps in this model, pulsative pressure from the beating heart to the aortae remained and this residual ‘flow’ stimulated Cx40 expression. Further studies will be needed to resolve the dependence of initial versus full arterial identity on hemodynamic flow.
Finally, we surmise that hemodynamic flow does not play a role during initial vein formation, as it is unclear how blood could enter veins if a complete circuit is not present. In addition, most venous markers initiate in pre-venous angioblasts, prior to vein tube formation or blood flow. Interestingly, the putative lack of flow in early veins might explain their more plexus-like organization during vasculogenesis (for example the anterior cardinal vein, Fig. 5D), as flow tends to drive plexus remodeling and formation of larger vessels. Live imaging will be useful in future studies to directly examine how flow changes in coordination with the dramatic network changes that occur in the developing vasculature. The timeline and molecular signature of arteries and veins provided by our study will provide tools to examine the role of blood flow specifically in mammalian arteries or veins.
This study contributes three new findings to our understanding of the initial, developing vascular anatomy in mammalian embryos: 1. AV fate establishment (differentiation) occurs in a stepwise manner; 2. arteries form prior to veins, and both emerge via the progressive formation of blind-ended vessels; and 3. hemodynamic flow is required for some, but not all, arterial gene expression suggesting that arterial specification does not depend on flow, but full differentiation does.
Together, our data indicates that blood vessels emerge and become specified in a stepwise manner. The first murine blood vessels are arteries and their specification occurs early, prior to the establishment of blood flow, but likely concurrent with the onset of pulsative pressure from the heart beating. Indeed, the expression of a subset of arterial genes in angioblasts occurs during aortic cord formation, at the onset of vascular lumen formation (Strilic et al., 2009; Xu et al., 2011), but prior to embryonic circulation. These findings suggest that AV specification is driven by a cell intrinsic program or microenvironmental cues, rather than by hemodynamic cues. However, in the absence of hemodynamic flow some, but not all, arterial gene expression is extinguished, suggesting that some aspects of AV differentiation/maintenance depend on cell extrinsic factors such as the mechanical force of blood flow. While classical work has demonstrated this to be true in the chick yolk sac (le Noble et al., 2004), ours is a first such analysis in mammalian embryos and in intraembryonic blood vessels. Furthermore, this study defines both the spatiotemporal molecular signature of developing arteries and veins, as well as the selective dependence of certain arterial markers on blood flow. This work will instruct future studies of the cardiovascular system and provide greater resolution to studies of mammalian vasculogenesis.
CD1 embryos were collected from pregnant females (E7.5 through E10.5) by dissection in ice-cold 1xPBS buffer phosphate buffered saline and fixed in 4% paraformaldehyde (PFA) in PBS solution overnight at 4°C with gentle rocking. Embryos were then washed twice in PBS for 10 min at 4°C, and dehydrated using an ethanol series, and stored in 75% ethanol at −20°C. To generate Rasip1 null embryos, Rasip1+/− males were mated to Rasip1+/− females (Xu et al., 2011). Embryos were collected at E8.25, fixed, rinsed, and stored similar to CD1 embryos. All animal handling was performed in accordance with IACUC regulations.
Whole mount in situ hybridization was carried out using a protocol adapted from D. Wilkinson’s method (Wilkinson, 1999). Briefly, embryos stored in 75% EtOH at −20°C, were rehydrated in stepwise fashion to PBST. Then, the embryos were treated with proteinase K, fixed in a 0.25% gluteraldehyde/4% PFA solution, and pre-hybridized at 60°C for 1 hr. The samples were transferred into hybridization mix, containing a Digoxigenin-labeled probe. The in situ hybridization steps were carried out using a Biolane HTI automated incubation liquid handler (Holle & Huttner). Development of color reaction was done using BM Purple (Roche). Images were taken using a Lumar dissecting microscope (Zeiss) and a DP-70 camera (Olympus).
Dig-RNA probes were generated from clones (Open Biosystems): Cx40 (BC053054), Cx37 (BC056613), Nrp1 (BC060129), Nrp2 (BC098200), Notch1 (BC010325), Notch4 (BU703407), Dll4 (BC042497), Jagged1 (BC058675), CoupTFII (BC094360), APJ (BC039224) and Flt4 (BI557457). Hey1 and Hey2 clones were generously provided by Dr. Eric Olson, UTSW).
Flk1-lacZ, ephrinB2-LacZ, and EphB4-LacZ embryos were generated by mating heterozygous males to CD1 females. Embryos were dissected manually and fixed in a 0.2% gluteraldehyde/2mM MgCl2/5mM EGTA/PBS solution for 20 min on ice. The embryos were then rinsed 3 times in PBS for 5 min each. After rinsing embryos, a LacZ staining solution was made using the following: 20 mM K4Fe(CN)6, 20 mM K3Fe(CN)6, 2 mM MgCl2, 0.02% NP-40, and 1X PBS to a final volume of 500 ul. Addition of 4 ul of 100 mg/ml X-Gal stock (in dimethyl formamide) was added after the staining solution was warmed to 37°C to avoid X-Gal precipitation. Embryos remained in staining solution overnight, shielded from light, to allow the color reaction to develop. Then, the embryos were washed with PBS 3 times for 5 min. each, and fixed in 4% paraformaldehyde/PBS for one hour. The embryos were again rinsed in PBS 3 times for 5 min. each and transferred to 80% glycerol for viewing using a NeoLumar stereomicroscope (Zeiss) and photographed using a DP70 camera (Olympus).
For paraplast sectioning of embryos following in situ hybridization, the embryos were post-fixed in 4%PFA and dehydrated to 75% ethanol. Embryos were then rinsed twice in 80, 95, and 100% ethanol for 5 min. each, twice in xylene at room temperature for 5 min, then a series of rinses in 100% Paraplast Plus tissue embedding medium (McCormick) at 60°C. The embryos were then embedded and sectioned with a 2030 Reichert-Jung microtome. For examination, the sections were mounted on SuperfrostPlus glass slides (Fisher), deparaffinized in Xylene twice for 5 min. each, and mounted with glass coverslips using Permount (Fisher). Images were taken on a Zeiss Axiovert microscope using a DP70 camera (Olympus).
To generate Flk1-EGFP embryos, heterozygote Flk1-EGFP males were mated to CD1 females. E8.5-E9.5 embryos and E15.5 pancreata were dissected and fixed for one hour in 4% paraformaldehyde/PBS. The samples were then rinsed in PBS, dehydrated through an ethanol series, rinsed in xylene twice for ten min. each followed by a series of washes in Paraplast Plus tissue embedding medium (McCormick) and then embedded and sectioned on a 2030 Reichert-Jung microtome. After the tissues were sectioned and mounted on SuperfrostPlus glass slides (Fisher), they were deparaffinized in xylene twice for 10 min. each, rehydrated through an ethanol series, 100% twice for 2 min. each, 90%, 80%, 70%, 40% for one minute each and then placed under running water for 3 minutes. The slides were then equilibrated in a 1x Buffer A solution (Electron Microscopy Sciences) and antigen retrieval was performed in a 2100 Retriever (Pickcell Laboratories) overnight. The next day, the slides were rinsed in 1X PBS and blocked for 2 hours with Cas-Block (Invitrogen) before the primary antibody was added for overnight incubation at 4°C. The following primary antibodies were used at indicated concentrations: Cx40 1:200 (Santa Cruz), Nrp1 1:200 (R&D), Nrp2 1:200 (Cell Signaling), GFP 1:500 (Aves Labs). Signal was detected the following day using AlexaFluor 555 and AlexaFluor488 (Invitrogen). The sections were mounted with ProLong Gold Antifade (Invitrogen) and images were acquired on a LSM510META (Zeiss) confocal microscope.
CD1 embryos at the 5s stage were isolated in ice cold 1X PBS at E8.0. Embryos were then bisected using Dumont #55 fine forceps (Fine Science Tools). The anterior and posterior halves were placed on separate 13mm filters (Whatman) over 500ul of media in 4-well plates (Nunc). This created an air-surface interface and allowed the explants to grow flat to facilitate analysis of vessel morphology. Media used was DMEM (ATCC), 10% FBS (ATCC), 1% penicillin/streptomycin (Invitrogen), and 2mM glutamine (Invitrogen). After 10 hours or 18 hours, explants were rinsed in 1X PBS twice, fixed in 4% PFA/PBS overnight, then rinsed in 1X PBS and dehydrated to 70% ethanol. Whole-mount in situ hybridization was then performed as previously described (Villasenor et al., 2008). Images were taken using a NeoLumar stereomicroscope (Zeiss) and photographed using a DP70 camera (Olympus).
Transverse sections through a β-galactosidase staining of a Flk1-LacZ embryo reveals established and lumenized dorsal aortae, before inflow tract/vitelline vein formation. (A) Anterior view of Flk1-LacZ 3s embryo. Red arrow points to dorsal aorta and blue arrow points to inflow tract/vitelline vein. B) Schematic of boxed area in (A) showing where sections of (C,D) are located relative to the embryo. (C) Transverse section right below the cardiac crescent shows dorsal aorta and inflow tract contain lumens. (D) A more posterior transverse section illustrates that the dorsal aorta remains lumenized, but the emerging vitelline vein is a cord, with no lumen. da, dorsal aorta; it, inflow tracts; vv, vitelline vein.
Transverse sections through E8.25 embryos. All images display the paired dorsal aortae and the neural folds. (A–J) Transverse sections through embryos stained for arterial markers reveal that some markers are specific to endothelial cells (Cx40, Cx37, Hey1, Notch4, Dll4) while others are also expressed in other tissues (ephrinB2, Nrp1, Notch1). However, Jag1 was never expressed in endothelial cells at this stage. (K–O) In contrast to arterial markers, venous markers were expressed in multiple tissues (CoupTFII, Nrp2, APJ) with the exceptions of Flt4 which was EC-specific and EphB4 which had undetectable levels.
In situ hybridization or β-galactosidase staining of established AV markers at E8.5. (A–J) Dorsal aortae expression is robust using ephrin-B2, Cx40, and Cx37. Hey1, Notch4 and Dll4 expression become downregulated. (K–O) Vitelline vein expression can be identified using the venous markers Nrp2 and Flt4 (blue arrows). All somite stages are indicated.
Co-immunofluorescence for Flk1 and arteriovenous markers on E15.5 pancreas cross-sections. Single channels for Cx40, Nrp1, and Nrp2 (A′–C′) and Flk1 (A″–C″). Expression of Cx40 and Nrp1 is restricted to arterial endothelium and surrounding smooth muscle cells (A–B) while Nrp2 is expressed only on venous endothelial cells (C) at this late stage. a, artery; v, vein.
The first veins, vitelline veins, emerge along the embryonic-extraembryonic boundary, along the lateral edge of the extraembryonic vascular plexus. (A) By 6s, primordial veins appear blunt-ended, with scattered angioblasts more posteriorly. (B) By 8s, vitelline veins extend more posteriorly, but are thin cords. (C) By 10s, vitelline veins extend farther posteriorly.
Progression of arterial and venous development over time. Arteries – The cardiac crescent (cc), a bilateral patch of splanchnic mesoderm comprising both cardiac and endothelial mesoderm, is formed at about 1–2s, before dorsal aortae (da) formation. The aortae arise from angioblast aggregations, form from anterior to posterior, and become lumenized at 3s. In the cephalic region, the aortae also form, ventral to the neural folds, and connect to the heart (h) via the first and most anterior aortic arch (aa). By 7–8s, the linear heart tube acquires an s-shaped and additional endothelial arches begin to form by 9–11s. Veins - The first veins, the vitelline veins (vv), arise as extensions of the cardiac crescent and sinus venosus (sv) at about 4–5s. Shortly thereafter, between 6–8s, ECs of the sv form the common cardinal veins (ccv) and the anterior cardinal veins (acv) begin to emerge anteriorly along ventrolateral to the somites. By 9–11s, the posterior cardinal veins (pcv) then arise and extend posteriorly. They continue to emerge posteriorly until they connect to the posteriormost aspect of the paired dorsal aortae, forming a circulatory loop.
Schematic of explant procedure. A) 5s embryos were bisected and each half was grown on a filter, at the air-surface interface. B) Brightfield images of the isolated anterior (a) and posterior (b) halves, as well as a whole embryo (c) at 5s. C) Cx40 and Dll4 are expressed in the dorsal aortae at 5s in unmanipulated embryos. D) EphB4-LacZ embryonic halves, cultured for 10 hours, lacked EphB4 expression suggesting a role for circulation in venous gene expression maintenance.
Wholemount and sections of in situ hybridization on embryos. Cx40 expression is absent at 0s (A) but is robustly expressed in the cardinal crescent at 3s (B). By 8s (C), Cx40 is expressed in the dorsal aortae and the arteries in the allantois (inset). After turning, Cx40 is robustly expressed in the aorta (D–H, J) and the developing heart and head arteries. Notably, Cx40 is expressed strongly in the yolk sac arteries (I) and is only in three chambers of the E11.5 heart (K). Sections reveal the enriched expression of Cx40 to the endothelium (J), including the endocardium (L). Red arrows point to dorsal aortae.
We thank Janet Rossant and Eli Keshet for the Flk1-LacZ mice; Janet Rossant for the Flk1-EGFP mice; and Mark Henkemeyer for the ephrinB2-LacZ and EphB4-LacZ lines. We also thank Aly Villasenor, Victoria Bautch, Jenny Hsieh, Hyun Lee, Jessica Harrell and Stryder Meadows for critical reading and discussions. This work was supported by NIH DK079862 and AHA Grant-in-Aid 0755054Y to OC.