Development and origins of Zebrafish ocular vasculature

The developing eye receives blood supply from two vascular systems, the intraocular hyaloid system and the superficial choroidal vessels. In zebrafish, a highly stereotypic and simple set of vessels develops on the surface of the eye prior to development of choroidal vessels. The origins and formation of this so-called superficial system have not been described. We have analyzed the development of superficial vessels by time-lapse imaging and identified their origins by photoconversion experiments in kdrl:Kaede transgenic embryos. We show that the entire superficial system is derived from a venous origin, and surprisingly, we find that the hyaloid system has, in addition to its previously described arterial origin, a venous origin for specific vessels. Despite arising solely from a vein, one of the vessels in the superficial system, the nasal radial vessel (NRV), appears to acquire an arterial identity while growing over the nasal aspect of the eye and this happens in a blood flow-independent manner. Our results provide a thorough analysis of the early development and origins of zebrafish ocular vessels and establish the superficial vasculature as a model for studying vascular patterning in the context of the developing eye.


Background
The formation of correctly patterned vasculature is necessary for normal organ development and function. Two main mechanisms underlie formation of vascular networks: vasculogenesis and angiogenesis. In vasculogenesis, angioblasts, the precursors of endothelial cells, coalesce at a specific region where they differentiate into endothelial cells and rearrange to form new blood vessels. In angiogenesis, new blood vessels arise from an existing vessel, either by splitting of the vessel (intussusceptive angiogenesis) or by sprouting of new vessels (sprouting angiogenesis) [1,2].
The developing eye receives blood supply from the hyaloid and choroidal vascular systems. The hyaloid system is intraocular, and forms a dense network of vessels that reaches and partially covers the posterior aspect of the lens [3][4][5]. Hyaloid vessels are transient and are replaced by retinal vasculature [3]. The more superficial part of the developing eye receives blood supply through the choroidal system, which forms a plexus of capillaries that are located outside the retinal pigmented epithelium (RPE) layer and encircles the optic cup [3]. Abnormal formation of ocular vasculature in the embryo as well as abnormal vascularization in the mature eye are implicated in several blinding diseases [6,7], underscoring the importance of understanding the development of these vessels.
Zebrafish embryos have a hyaloid system as well as a simple array of vessels on the surface of the eye, which develops by two days post-fertilization (dpf ) and is referred to hereafter as the superficial system [4,8,9]. Development of the superficial system precedes the appearance of an elaborate network of choroidal vessels on the surface of the eye at 9 dpf [9]. It is currently unclear how superficial vessels are related to choroidal vessels. In the hyaloid system the first vessel to form is the hyaloid artery (HA), which, as in other vertebrates, enters the eye via the optic fissure before the latter closes [4,9].
The hyaloid artery branches into capillaries inside the eye and blood flowing in these vessels drains through the hyaloid vein, which exits the eye through the optic fissure and connects to the superficial system [5,9,10]. The branching pattern of hyaloid capillaries varies such that the hyaloid system appears slightly different in individual embryos. The superficial system comprises the nasal radial vessel (NRV) through which blood enters, and the dorsal radial vessel (DRV) and ventral radial vessel (VRV) through which blood drains from the system (naming according to [9]). These vessels are interconnected by a ringshaped vessel, the superficial annular vessel (SAV) [9] (also see Figure 1F). Unlike the hyaloid system, superficial vasculature appears highly similar between embryos suggesting it develops in a highly stereotypic manner.
In contrast to the more extensive knowledge on hyaloid system development, little is known about the development of superficial ocular vessels and the earliest stages of their formation have not been described; hence the origins of these vessels, the dynamics of their formation and the signals that influence their development are unknown. Additionally, it is unknown whether there are molecular differences between vessels that are anatomically referred to as arteries or veins. Here we provide a thorough analysis of superficial system development from its earliest stages. We show that the entire system forms by angiogenesis from a venous origin and surprisingly, we also find that the hyaloid system has, in addition to its previously described arterial origin, a different source for its venous part. We show that endothelial cells of the NRV, which serves as the artery for the superficial system, turn on Notch pathway activity as they grow, suggesting they acquire arterial identity. Together, our data add new insights into the development of ocular vasculature and show that the zebrafish superficial vascular system can serve as a model for identifying mechanisms of vascular patterning.

Development of superficial ocular vasculature in zebrafish
To obtain a detailed understanding of how superficial vasculature forms from its earliest stages, we studied development of these vessels over time by imaging transgenic kdrl:EGFP embryos, whose endothelial cells express EGFP [11]. At 20 hours post-fertilization (hpf ) (22 somite-stage), the primordial midbrain channel (PMBC) [8] can be seen growing dorsoanteriorly along the posterior margin of the eye, whereas the cranial division of internal carotid artery (CrDI) [8] is growing in a dorsoposterior direction along the anterior margin of the eye. The PMBC meets and connects with the CrDI at approximately 21-22 hpf (see movie in Additional file 1). From 22-23 hpf, sprouting from the PMBC over the dorsal retina can be seen where the DRV will be located ( Figure 1A showing a sprout at 24 hpf and movie in Additional file 1). Typically, 1-2 sprouts grow in a ventral-anterior direction. If two sprouts grow, they bridge and continue to grow anteriorly as a single vessel dorsal to the lens until reaching and connecting to the CrDI at approximately 31-33 hpf, thereby forming the NRV ( Figure 1B,C; see movie in the Additional file 2) [12]. Often one of the two sprouts will be pruned, but occasionally two DRVs remain. At approximately the same time of NRV formation, ventral sprouting begins from the junction of the NRV and SAV, concomitant with dorsal sprouting from the VRV. These sprouts send long filopodial extensions and grow towards each other to form the anterior part of the SAV, which is completed at around 40 hpf ( Figure 1D; see movie in Additional file 3). At approximately 35 hpf, ventral sprouting from the junction of the DRV and SAV, together with less pronounced dorsal sprouting from the VRV, begin to form the posterior part of the SAV ( Figure 1E showing growth of posterior SAV at 49 hpf; see movie in Additional file 4), thus completing formation of the early superficial system at approximately 53-54 hpf (see Figure 1F for a completed system at 59 hpf ).
Formation of the superficial system is independent of blood flow. We injected antisense morpholino oligonucleotides (MO) that block translation of troponin T2a gene (tnnt2a; silent heart), which is specifically expressed in the heart and is essential for heart contraction [13]. Despite lack of blood flow in the morphants, superficial vessels formed in a correct pattern (see below).

A single venous origin for superficial vessels and a mixed origin for hyaloid vasculature
The documentation of superficial vessel development suggests that most of these vessels originate from the dorsal region of the PMBC, beginning with sprouting of the future DRV. However, the VRV is present before sprouts from the dorsal part of the system grow ventrally, suggesting it could have a different origin. Moreover, the superficial and hyaloid systems are connected via the hyaloid vein (HV), which raises the question of whether these systems have any shared origins. To clarify these issues we generated transgenic fish expressing the photoconvertible protein Kaede in endothelial cells under the regulation of kdrl promoter, and used this transgenic line to map the origins of all early ocular vessels.
All head vasculature develops from two clusters of angioblasts, which can be visualized from early segmentation stages: the rostral organizing center (ROC) and the midbrain organizing center (MOC) [14]. According to previous descriptions, posteriorly migrating ROC cells give rise to the CrDI, whereas anteriorly migrating cells give rise to the optic (ophthalmic) artery (OA) that grows into the eye to generate the hyaloid artery [8,14]. Consistent with the above mentioned previous studies [8,14], photoconversion of the entire ROC at 17 hpf (16-somite stage) (Figure 2A,B) and imaging at 40 hpf ( Figure 2C,D) showed that the CrDI and central vessels of the hyaloid system were clearly labeled by photoconverted Kaede. Importantly, peripheral vessels of the hyaloid system, which are connected to the HV, as well as superficial vessels that developed by 40 hpf did not express photoconverted Kaede ( Figure 2C,D) (n = 10/10). Photoconversion of endothelial progenitors migrating posteriorly or anteriorly from the ROC showed that these cells contributed to the CrDI or the central hyaloid vessels, respectively (data not shown). These results indicate that at least the central vessels of the hyaloid system originated from the ROC via the OA.
The PMBC is derived from the MOC [14] and, as we show above, gives rise to dorsal vessels of the superficial system. Indeed, when we photoconverted the entire PMBC at 20 hpf (22 somite-stage) and imaged embryos at 50 hpf, dorsal superficial vessels expressed photoconverted Kaede. Moreover, the VRV also contained photoconverted cells ( Figure 2E-H) (n = 5/5), suggesting that the PMBC is the origin of all superficial vessels. In this experiment, none of the hyaloid vessels appeared to contain photoconverted cells. As the data suggested that the PMBC is also the source of the VRV, we re-examined growth of vessels from the PMBC. We noticed that during a short time window between 20-23 hpf, long extensions are sent from the ventral region of the PMBC towards the ventral optic fissure and growing hyaloid artery ( Figure 3A,B). High magnification imaging suggests that these extensions are formed by a trail of individual endothelial cells, which migrate from the PMBC ( Figure 3C). Indeed, when we repeated photoconversion specifically of the ventral PMBC at 22-23 hpf and imaged embryos at 35 hpf, we detected photoconverted cells in the VRV (Figure 2I-L). Surprisingly, peripheral hyaloid vessels were also photoconverted ( Figure 2K,L) (n = 8/8), indicating that these vessels also originated from the PMBC. This conclusion is supported by timelapse imaging showing growth of peripheral hyaloid vessels from the hyaloid vein and their connection to central hyaloid vessels that extend from the hyaloid artery, thereby demonstrating that these two groups of vessels indeed have different origins (see movie in Additional file 5). Together, the data show that the entire superficial system is of venous origin, the PMBC, whereas the hyaloid system originates both from arterial (OA and HA) and venous (PMBC) origins, which give rise to central and peripheral components of the system, respectively. The different origins of hyaloid vessels,  together with direction of blood flow into the eye via the HA, through central vessels and then draining to the HV via peripheral vessels, raise the idea that central and peripheral hyaloid vessels represent arterial and venous components of the hyaloid system, respectively.
The NRV appears to acquire arterial identity while growing Given that the entire superficial system arises from a venous origin, the question arises whether the NRV, which according to direction of blood flow serves as the artery of this system [9], acquires at some point arterial identity. To address this question we first attempted to label embryos by in situ hybridization with various artery-or vein-specific markers. However, we could not detect any signals in superficial vessels, possibly due to their very small size and/or low levels of transcripts. We therefore asked if Notch signaling, which is known to be required for artery specification [15], was activated in the NRV. To observe Notch pathway activation we used the transgenic line Tg(EPV.Tp1-Mmu.Hbb:EGFP)um14 (hereafter abbreviated as Tp1bglob:eGFP), which reports on Notch pathway activation by EGFP expression [16]. We used double transgenic embryos carrying Tp1bglob: eGFP and kdrl:Hsa.HRAS-mCherry transgenes to identify blood vessels showing Notch pathway activation. At 23 hpf the CrDI has connected to the PMBC, but superficial vasculature has not yet developed. Consistent with their artery and vein identities, the CrDI (an artery) expresses high levels of EGFP whereas no EGFP expression is detected in the PMBC (a vein) ( Figure 4A-C). Moreover, the PMBC, but not the CrDI, expresses the veinspecific marker dab2 [17,18], confirming that at this stage it already has a venous identity (Additional file 6). By approximately 31-33 hpf, the NRV connects to the CrDI. Imaging embryos shortly before this connection occurs revealed that EGFP expression is present in some of the leading cells of the growing NRV ( Figure 4D-F) (n = 4/4). Immediately after it connects to the CrDI, EGFP expression can be clearly seen throughout the NRV ( Figure 4G-I). These results show that the NRV, which arises from a vein, upregulates Notch pathway activity while growing, shortly before it connects to the CrDI. We hypothesize this activation of Notch reflects acquisition of arterial identity by the NRV. To further test this hypothesis we used a second transgenic line, Tg(flt1_9a_cFos:GFP)wz2, which utilizes a specific flt1 enhancer previously described to drive strong expression of GFP in arteries and weak expression in veins [19]. Consistent with our hypothesis, at 52 hpf the NRV as well as few endothelial cells in the dorsal SAV expressed high GFP levels, whereas other superficial vessels expressed very low levels of GFP (n = 7/7) (Additional file 7). Moreover, central hyaloid vessels, which originate from an artery, also expressed high GFP levels whereas in peripheral hyaloid vessels, which originate from a vein, GFP expression was undetectable (Additional file 7). These results support the idea that the NRV acquires an arterial identity and that the hyaloid system has a central, arterial component and a peripheral, venous component.
It has been shown that in specific contexts, blood flow determines the identity of vessels [20]. Because Notch signaling is activated in the NRV before it connects to the CrDI, blood flow from the CrDI is not required for it to upregulate Notch pathway activity; however, it could be required for its maintenance. We therefore blocked blood flow by injecting Tg(Tp1bglob:eGFP; kdrl: Hsa.HRAS-mCherry) embryos with antisense morpholino oligonucleotides targeting tnnt2 and examined them at 2 dpf. Despite lack of blood flow in injected embryos, the NRV expressed EGFP, indicating that maintenance of Notch activity was independent of blood flow ( Figure 4M-O). Hence, the data suggest that local cues, rather than physiological changes, determine the apparent arterial identity of the NRV.

Discussion and conclusions
In this work we provide a detailed description of the development of zebrafish superficial ocular vasculature and identify the origins of hyaloid and superficial vessels (summarized schematically in Figure 5). Our results provide a basis for better understanding the molecular mechanisms that influence ocular vessel development.
The origins and development of hyaloid vessels have been previously described [4,5,8,9]. Consistent with these studies, our analyses show that central vessels of the hyaloid system originate from the OA. However, we also found, unexpectedly, that peripheral hyaloid vessels and the HV originate from the PMBC. Given that peripheral vessels empty into the HV and that the hyaloid system has two different origins for its central and peripheral vessels, we propose that central and peripheral parts of the system represent arterial and venous components, respectively. It will be interesting to find whether two distinct origins for hyaloid vessels are found in other vertebrates as well.
In contrast to the hyaloid system, superficial vessels have a single venous origin but the NRV acquires arterial identity, as judged by upregulation of Notch pathway activity and differential expression of GFP in flt1_9a_cFos: GFP transgenic embryos. This finding is consistent with the direction of blood flow in the NRV, suggesting it serves as the artery of the superficial system. Hence, unlike the hyaloid system where presumptive arteries and veins have arterial and venous origins, respectively, the superficial system presents a scenario where one of the vessels comprising it acquires an identity, which is different from its vessel of origin. This finding suggests there is plasticity in artery-vein specification. Indeed, such plasticity has been shown in chick embryos during a limited early time window in development [21,22]. Interestingly, change of identity from veins to arteries was also shown during formation of coronary arteries, that form from differentiated venous endothelial cells in the sinus venosus [23]. Together with the fact that the NRV upregulates Notch pathway activity as it grows and before connecting to the CrDI, these findings suggest that local cues in the environment promote this change of endothelial cell type. The most likely tissue to mediate the environmental cues is the developing eye on which superficial vessels grow and it will be challenging to identify the signals that direct both the patterning of superficial vessels and the molecular changes in the NRV.
Previous studies have identified genes and drugs that influence hyaloid system development without affecting development of other vascular systems, e.g. intersegmental vessels (ISVs) [4,9]. These results demonstrate heterogeneity in the molecular mechanisms that underlie hyaloid and ISV formation. It will be important to find whether similar heterogeneity exists between molecular mechanisms governing superficial vessel development and other vascular systems and whether superficial vessels and hyaloid vessels are similarly affected by specific molecular cues.
The simple, yet highly stereotypic pattern of the superficial vasculature, makes it an excellent model for studying vascular patterning. We hypothesize that the molecular landscape of the eye provides cues for directing the growing vessels. Eye tissues that are in close proximity to developing vessels are likely to influence their growth and include neural retina, retinal pigmented epithelium, lens and surface epithelium that will develop into cornea. Relevant molecular cues can be attractants or repellents, influencing vessels directly or acting as modifiers of an angiogenesis-promoting signal such as vascular endothelial growth factor (VEGF). Candidate molecules would be, for example, cell adhesion molecules and cell-cell signaling molecules. By manipulating expression of such molecules within eye tissues, it will be possible to gain new insights into the molecular mechanisms that shape vascular systems.

In situ hybridization and morpholino injection
Whole-mount in situ hybridization (ISH) using riboprobes was performed according to standard protocols [28]. BMPurple (Roche) was used as blue substrate. dab2 probe has been described [17]. MO1-tnnt2a [13] has been described and was injected at the 1-2 cell stage.