The six sections below present a detailed discussion of the formation of different vascular systems in the zebrafish, and contrast them with the vascular anatomy of other developing piscene and non-piscene vertebrates. One of the primary goals of this work is to provide a firm basis for comparing blood vessels and vascular defects in the zebrafish to those in other vertebrate species, particularly mammals. Implicit in this is an in-depth understanding of the ways in which vessel formation in the zebrafish is homologous with that in other vertebrates, and the ways in which it is not. The comparative analysis presented below is an attempt to synthesize some of the data in this website and facilitate this understanding.

In mammals and many other vertebrates transverse vessels branching from the dorsal aorta have an important role in the primary vitelline circulation. Earlier authors reported that transverse vessels do not form in teleosts ((Ballard, 1964; Ura, 1949a)), but Isogai and Horiguchi demonstrated that vestigial transverse vessels do exist during very early stages of the rainbow trout circulation (Isogai, 1997). They suggested that formation of these vessels is initiated in most teleosts as in other vertebrates, but that they do not become fully functional and other vessels are actually used for yolk absorption. In salmonids, the subintestinal vein plays a major role in yolk absorption. Interestingly, this vessel is fed by venous blood from the caudal vein, while the transverse vessels of other vertebrates are fed by arterial blood directly from the dorsal aorta (Ballard, 1964; Ura, 1949a). Detailed reexamination of the development of the subintestinal circulation in the rainbow trout (Isogai, 1997) showed that it initially receives its blood supply only from the caudal vein (with the exception of the transient vestigial supply from the transverse vessels at a very early stage, as noted above). With the formation of the supraintestinal artery, the subintestinal vein connects to and receives blood from this vessel. Eventually, the subintestinal vein becomes fed exclusively by the supraintestinal artery, and the connection with the caudal vein is lost entirely. This same supraintestinal circulatory pathway is found in other vertebrates, including mammals (Aoyama, 1956; Isida, 1956; Kawanishi, 1956; Saito, 1984; Tada, 1956; Ura, 1956a; Ura, 1956b). In addition to the subintestinal veins, enlarged common cardinal veins (also termed "ducts of Cuvier") also play a major role in yolk absorption in many teleosts. The common cardinals and subintestinals are both used in the medaka Oryzias latipes (S. Isogai, unpublished), in the pike Esox lucius (Zieba, 1956) and in many other teleosts.

In zebrafish, we have not been able to detect transverse vessels using angiographic methods, even transiently, although it is possible that, as in rainbow trout, unused vestigial transverse vessels might exist that could be detected by molecular methods (see "limitations and caveats"). The paired common cardinal veins are the sole vessels providing yolk absorption until approximately 2.5 dpf, at which point subintestinal veins finally develop over the yolk ball. Unlike many other teleosts, the subintestinal veins of zebrafish receive their blood directly from the supraintestinal artery from the outset. This route corresponds to the secondary pathway described in rainbow trout; the primary route from the caudal vein found in many teleosts has apparently been abandoned in zebrafish.

Miki (Miki, 1968) previously described the developmental origins of the hepatic portal system and compared its formation in different vertebrate species (Aoyama, 1956; Isida, 1956; Kawanishi, 1956; Saito, 1984; Tada, 1956; Ura, 1949a; Ura, 1956a; Ura, 1956b; Yamada, 1951). The single (or merged) subintestinal vein initially runs cranial ward along mid-ventral wall of the gut, draining into the hepatic sinusoid as the primary hepatic portal vein. At slightly later stages this subintestinal vein makes a new connection, detouring around the fore- and mid-gut boundary from left to right side via the dorsal gut wall and then draining into the hepatic sinusoid again alongside the bile duct, as the secondary portal hepatic vein. After this secondary hepatic portal connection becomes established the primary hepatic portal route is severed, although the most proximal portion of the subintestinal vein (now designated "Rusconi’s vein," (Aoyama, 1956)) remains. At still later stages, a supraintestinal vein forms collateral to the supraintestinal artery, also draining into the secondary hepatic portal vein. After this new vein develops, all but the most cranial portions of the subintestinal vein degenerate, and the supraintestinal vein becomes the sole route for venous return from the supraintestinal artery. This basic program is followed in the rainbow trout (Isogai, 1997).

In the 2.5 dpf zebrafish, a pair of right and left subintestinal veins each drains into the hepatic sinusoid via independent hepatic portal veins on either ventral side of the cranial midgut. By 4-4.5 dpf, caudal (distal) portions of the left subintestinal vein become gradually divided into separate parts, which then merge into the right subintestinal vein. With the reduction of yolk and postural changes in the cranial midgut, the left cranial SIV also degenerates, except for the most proximal portion (Rusconi’s vein). By 7 dpf the right subintestinal vein has become the dominant route for venous drainage into the liver, although additional hepatic portal veins appear on the ventral wall of cranial midgut. All of the SIVs and supplementary intestinal veins drain into the liver along the ventral midgut wall. Comparing this to the "general vertebrate plan" described above, the 7 dpf zebrafish has only completed the initial phases of subintestinal vein morphogenesis. That is, merging the caudal left and right SIV and disconnecting the most proximal portion of the left SIV to become a distinct, separate entry point to the liver. The subsequent phases described above have not yet occurred in zebrafish by 7 dpf. However, these later stages have been demonstrated in a wide variety of vertebrates, including rainbow trout (Isogai, 1997) and Medaka (S. Isogai, unpublished), so it is likely that they also occur in zebrafish. Further examination of older larvae, juvenile fish, and adults will reveal the complete sequence of development of the portal hepatic vascular system in zebrafish and its correspondence to that of other vertebrates.

In most vertebrates, paired posterior cardinal veins extend caudally from the common or anterior cardinal veins, running adjacent to the pronephric ducts. At their caudal end they connect to the caudal vein. When this connection occurs, blood from the caudal vein begins to flow cranial ward, and the posterior cardinal vein functions as a pronephric portal vein. Swaen and Brachet (Swaen and Brachet, 1899) reported that in Salmonids there is only a single posterior cardinal vein ("median vein"), derived directly from the ventral half of Oellacher’s intermediate cell mass (Swaen and Brachet, 1899; Vernier, 1969). More recently, however, Isogai and Horiguchi showed that at the very early stages of circulation in the rainbow trout, there are in fact a pair of posterior cardinal veins, as in other vertebrates (Isogai, 1997). These subsequently merge into single posterior cardinal vein (also called the "axial," "central," "stem," or "trunk vein").

In zebrafish, the posterior cardinal vein is present and functioning as soon as circulation initiates in the embryo and only the most cranial portion of the posterior cardinal vein is paired. As noted above, the subintestinal vein does not connect to the caudal vein in the zebrafish, and so the posterior cardinal vein serves as the sole route for venous return from the caudal vein. This might help to explain why it is necessary for this vessel to form at such an early stage as a distinct, functional midline vessel compared to most other vertebrates, and even many teleosts. Beginning at 2 dpf, the left PCV starts to regress, and the right PCV becomes the dominant vessel carrying most of the venous return from trunk. The same phenomenon occurs in all vertebrates, including humans. Also like other vertebrates, a mesonephros takes the place of the embryonic pronephros. We have no data on the formation of the mesonephric vascular system yet. At approximately 3.5 dpf a profusion of short vessels appears in the trunk between the DA and PCV, just where the mesonephros develops in later stages. Most of the vessels are transient, and largely disappear by 4.5-5 dpf. The nature of these vessels is not at present understood, and awaits further analysis.

A paired intersegmental artery and intersegmental vein appear at each vertical transverse myoseptum in reptilian, avian, and mammalian embryos (Evans, 1909a; Evans, 1912). In lower vertebrate embryos such as teleosts, however, only a single artery or vein develops at each transverse septum. It has been generally believed that single intersegmental arteries and single intersegmental veins alternate at each successive transverse septum with regularity (Furuyama, 1960; Grodzinski and Hoyer, 1938; Hashimoto, 1960). The regularity of artery-vein alternation from one segment to the next has never been experimentally substantiated, however, and it is in fact not the case in the zebrafish.

In the zebrafish, intersegmental arteries and veins in the caudal trunk and cranial tail have fully extended by 1.5 dpf to meet one another dorsal to the trunk as the dorsal longitudinal anastomotic vessels. At this time, and thereafter, the arrangement of intersegmental arteries and intersegmental veins at each intersegment appears to be random (with the exception of the first four segments; see below). Multiple veins or multiple arteries frequently follow one another at successive myotomal boundaries. Furthermore, the patterns of intersegmental arteries and veins vary from one embryo to the next without apparent regularity. However, numerous reticular anastomoses develop between right and left dorsal longitudinal anastomotic vessels, permitting blood flow between intersegmental vessels on both sides of the trunk. This communication between the two sides makes it possible to have efficient blood flow to the entire trunk without the need for a regular alternating pattern of intersegmental veins and arteries on one side of the embryo, a feature that may have been overlooked by previous authors. In the most rostral four trunk segments the dorsal longitudinal vessels remain paired and are not connected by anastomoses, and in these segments there is in fact regularity in the pattern of the intersegmental vein and arteries. The first set of intersegmental vessels on either side are always veins, draining blood coming from both the (head) basilar artery and the (trunk) dorsal longitudinal anastomotic vessels into the posterior cardinal vein. The second intersegmental vessels are always arteries. The pectoral (subclavian) arteries originate from the dorsal aorta adjacent to the root of these second intersegmental arteries. The third and the fourth intersegmental vessels on both sides appear to always be veins.

The intersegmental vessels are the only vessels functioning in the trunk until approximately 3.5 dpf. Sprouts for the vertebral artery (associated with the spinal cord) and the parachordal vessel (adjacent to the notochord, at the level of the horizontal myoseptum) begin to appear at 3.5 and at 4 dpf respectively. As these two longitudinal vessels become established across myotomal segments, they frequently cause changes in the direction of blood flow in distal (dorsal) portions of the intersegmental vessels, causing flow patterns to become more complex. At 4-4.5 dpf, additional lateral branches sprout from the intersegmental vessels at the level of the horizontal myoseptum, often directly from the parachordal vessel sprouts themselves. These lateral branches further subdivide into dorsal and ventral (intercostal) branches, which extend obliquely along the transverse septa in a "herringbone" pattern only partially visible by 7 dpf (Fig. 11 B,C). We have not detailed the course of the intercostal vessels and their further development post-7 dpf as a part of this work. However, preliminary microscopic observations at later stages (S. Isogai et al, unpublished results) indicate that the dorsal branches have their arterial root at the dorsal tip of intersegmental arteries while the ventral intercostals have their root at the dorsal aorta. Both sets of vessels drain into the intersegmental veins at the level of the horizontal myoseptum, as noted above. Many of the details of how these later trunk vessels form are still unknown in higher vertebrates.

The early development of the vasculature of the human brain has been described in detail by Padget (Padget, 1948; Padget, 1957). In humans, the cerebral artery system establishes its basic pattern in a number of stages. Initially, two bilateral rostro-caudally aligned sets of major vessels, the primitive internal carotid artery and the primordial hindbrain channel (in this study, we have defined the zebrafish vessel between the anterior cerebral vein and the mid-cerebral vein as the primordial midbrain channel), form first in the head. The primordial hindbrain channels extend caudally along the lateral walls of the hindbrain, draining into the anterior cardinal veins just caudal to the otic capsules via a short segment. Next, a capillary network develops from (and between) the primordial hindbrain channels, and a pair of bilateral longitudinal neural arteries arise from the medial edges of these networks, eventually connecting with the vertebral arteries in the trunk. Slightly later, the medial basilar artery forms at the ventral keel of the hindbrain by merger of the two bilateral longitudinal neural arteries. The primordial hindbrain channels intermingle with the cerebral capillary network, then connect to the vertebral arteries.
A similar process occurs in the zebrafish, with a few notable differences. No trace of the bilateral longitudinal neural arteries is apparent, at least by angiographic methods. The basilar artery forms as a single medial vessel from the beginning, apparently bypassing bilateral longitudinal neural artery formation. However, a recent study of the formation of the basilar artery in zebrafish indicates that it arises by a mechanism similar to that described for the bilateral longitudinal neural arteries of humans. That is, by angiogenic vessel growth from the primordial hindbrain channels (B. Roman and B. Weinstein, unpublished results). The connection between the cranial arteries and the vertebral arteries appears to be occurring at a later stage compared to formation of the basilar artery. As noted briefly above in the Results section, the circle of vessels comprised of the basal communicating artery (BCA) and posterior communicating segments (PCS) superficially resemble but are not homologous to the human circle of Willis. With the enormous hypertrophy of the human cerebrum, the circle of Willis branches out a few robust cerebral arteries to supply the forebrain. The names of these vessels (or parts of the circle) derive from anatomical features related to the human cerebrum. The zebrafish circle supplies the mid- and hindbrain, and has no direct connection with the forebrain until 7 dpf at least. Therefore, we have given new names to the elements of the zebrafish circle (i.e., BCA and PCS). To further compare the zebrafish BCA-PCS vascular circle to other vertebrates, we will need additional anatomical data from later larva, juveniles, and adult zebrafish.

Following establishment of the initial plan of major vessels surrounding the brain, in the next developmental stage of the cerebral vascular system the vasculature penetrates into the brain substance. Many of these new internal vessels take over major circulatory roles from the earlier-formed vessels on the cerebral surface (see below, "Vessel formation is a highly dynamic process, but it follows a stereotypic program."). The data we have provided here, detailing how this process occurs in zebrafish, are difficult to relate to other vertebrates, since no comparable data exist for other vertebrate models. The profound optical clarity of the zebrafish embryo and our confocal microangiography technique (Weinstein et al., 1995) have enabled us, for the first time in any vertebrate species, to make a detailed examination of all cranial vessels, from the outside surface to the very center of the brain. As we expostulate more fully below, our findings highlight the important role that intrinsic, programmed changes in vascular connections play in the transition between early and later stages of cranial and other vessel patterns, and in the establishment of the final plan of major vessels in the adult.

In the zebrafish, the cranial division of the primitive internal carotid artery (CrDI) branches out a number of arteriolae which distribute on the fore- and mid- brain surface in the 1.5 dpf embryo. Central arteries begin to penetrate into the fore-, mid-, and hindbrain directly from the basal communicating artery and posterior communicating segments and from the basilar artery. Some of these central arteries take over portions of vessels formerly fed by the cranial division of primitive internal carotid artery. These central arteries make connections to and begin to feed existing venous vessels on the brain surface. The primordial hindbrain channel and the basilar artery both reach the caudal end of the medulla oblongata by 2-2.5 dpf and make a new connection with the dorsal longitudinal anastomotic vessel in the trunk. This corresponds to one of the primitive patterns of the human cerebral circulation described by Padget (Padget, 1948). However, at up to 7 dpf we have not observed the direct connections between the basilar artery and vertebral arteries that eventually form in humans, although the vertebral arteries do connect to the basilar artery by a temporary bypass. With the formation of the primary head sinus ("lateral head vein") this vessel becomes the main route for venous drainage of the anterior- and mid-cerebral veins and rostral portions of the PHBC (which reverse their direction of flow). Circulation through the primordial hindbrain channel ("medial head vein") becomes weakened, and in some portions flow is reversed. As development continues, the primordial hindbrain channel remodels, elaborates, and intermingles with the capillary network of the medulla oblongata, providing most of the venous drainage for this network. The rostral, middle, and caudal remnants of the PHBC become separated from one another. The connection between the primordial hindbrain channel and the anterior cardinal vein becomes incorporated as a segment of the posterior cerebral vein, which drains both the dorsal longitudinal vessel and caudal portions of the venous capillary network of the medulla oblongata (largely derived from the remodeled primordial hindbrain channel).

As in other vertebrates, in teleosts the mandibular (first) and the hyoid (second) aortic arches do not contribute to the gill circulation, and become greatly modified during their development (Goodrich, 1958). In the 3-7 dpf zebrafish embryo, the mandibular aortic arch elongates rostral-ventrally to match the rostral boundary of the ventral aorta. One branch, the hypobranchial artery, takes off rostrally from the mandibular arch at 3 dpf. This vessel will supply the ventral branchial region and the heart. At 3.5 dpf, another vessel, the pseudobranchial artery, arises as a vascular loop attached to the efferent part of the mandibular arch. The hyoid (second) aortic arch appears only transiently as a vestigial connection between the mandibular arch and the lateral dorsal aorta at 1.2 dpf, disappearing a short time thereafter. A new vessel, the opercular artery, arises from the hyoid stump on the mandibular aortic arch, then follows the lateral margin of the operculum (gill covering) ventrally to reconnect into the proximal part of the mandibular aortic arch. In most teleosts, the so-called orbital artery comes off the lateral dorsal aorta close to the hyoid aortic arch, or directly from this arch, to supply the operculum region. Because of its location some authors have called this vessel the "external carotid artery", but this vessel is not actually homologous to the external carotid artery of tetrapods (Goodrich, 1958). In fish, the name "internal carotid" has been generally applied to that region of the vessel beyond the origin of the orbital artery. We have not yet observed the orbital artery or the ophthalmic artery in the zebrafish (up to 7 dpf). However, since these vessels are known to form and have their roots in this locale in other teleosts (Goodrich, 1958), we have applied the name "primitive internal carotid" to that region of the vessel beyond the mandibular aortic arch and the psuedobranchial artery.

In teleosts the gross remodeling of the caudal four arches that occurs in mammals does not take place, and these arches are retained largely intact as the branchial arteries providing circulation through the gills. The third and the fourth aortic arches arise at approximately 2 dpf as direct branches from the ventral aorta to the lateral dorsal aorta. In contrast, the fifth and sixth arches, which come on-line at approximately 2.5 dpf, have a common trunk from the ventral aorta and drain to the midline dorsal aorta via their own independent route. The third and fourth aortic arches will serve primarily to supply the cranial vasculature, while the fifth and sixth aortic arches supply the trunk and tail. The third through sixth aortic arches divide into afferent and efferent vessels, with the two eventually communicating only by delicate loops in the gill lamellae where blood is oxygenated. These loops begin to form at 3.5 dpf in the third and forth arches, and at 4 dpf in the fifth and sixth arches.

1. Aoyama, M. (1956). Disvolvigo de sangvazoj de la mez- kaj finintesto ce hynobius naevius (in Japanese, Esperanto summary). Kaibogaku Zasshi 31, 573-598.
2. Ballard, W. W. (1964). "Comparative Anatomy and embryology." Ronald Press, USA.
3. Evans, H. M. (1909). On the development of the aortae, cardinal and umbilical vein, and the other blood vessels of vertebrate embryos from capillaries. Anat. Record 3, 498-519.
4. Evans, H. M. (1912). The development of the vascular system. In "Manual of human embryology" (F. Keibel and F. P. Mall, Eds.), Vol. 2.
5. Furuyama, Z. (1960). Disvolvigo de la angiosistemo de la korpomuro ce caretta olivacea (in Japanese, Esperanto summary). Kaibogaku Zasshi 35, 302-330.
6. Goodrich, E. S. (1958). "Studies on the structure and development of vertebrates." Dover Publications, New York.
7. Grodzinski, Z., and Hoyer, H. (1938). Das Blutgefäßsystem. In "Klassen und Ordnungen des Tierreichs, Echte Fische" (H. G. Bronns, Ed.), Vol. Bnd 6 Teil 2, pp. 1-77.
8. Hashimoto, N. (1960). Disvolvigo de la angiosistemo de la spinalmedolo ce caretta olivacea (in Japanese, Esperanto summary). Kaibogaku Zasshi 35, 331-369.
9. Isida, Z. (1956). Disvolvigo de kalateralaj vejnoj sur la ovoflavsako de kokojo. (in Japanese, Esperanto summary). Kaibogaku Zasshi 31, 334-348.
10. Isogai, S., Horiguchi, M. (1997). The earliest stages in the development of the circulatory system of the rainbow trout Oncorhynchus mykiss. J. Morph. 233, 215-236.
11. Kawanishi, H. (1956). Disvolvigo de sangvazoj de ovoflavsako ce testudoj, Caretta olivacea (in Japanese, Esperanto summary). Kaibogaku Zasshi 31, 517-550.
12. Miki, S. (1968). Über die genetische Beziehung der Milz zu den Gefässen des Magens besonders zu seinen sekundären Venen beim Riesensalamander (Megalobatrachus japonicus) (in Japanese, German summary). Kaibogaku Zasshi 38, 140-155.
13. Padget, D. H. (1948). The development of the cranial arteries in the human embryo. In "Contributions to Embryology", Vol. 212, pp. 207-261.
14. Padget, D. H. (1957). The development of the cranial venous system in man, From the viewpoint of comparative anatomy. In "Contributions to Embryology", Vol. 247, pp. 81-140.
15. Saito, H. (1984). The development of the spleen in the Australian Lungfish Neoceratodus forsteri Krefft, with special reference to its relationship to the "gastro"-enteric vasculature. Am. J. Anat. 169, 337-360.
16. Swaen, A., and Brachet, A. (1899). Etude sur les premières phases du développement des organes dérivés du mésoblaste chez les poissons téléostéens. Arch. Biol. 16, 173-311.
17. Tada, Y. (1956). Disvolvigo de sangvazoj sur la ovoflavsako kaj ankau de la mesintesto ce hamsteroj (Cricetus anuratus) (in Japanese, Esperanto summary). Kaibogaku Zasshi 31, 388-417.
18. Ura, R. (1949). The development of the blood vascular systems. (in Japanese). Nisshin Igaku 33, 130-143.
19. Ura, R. (1956a). Kontribuo al ontogenio de vazsistemo de antauj mmbroj, precipe ce hynobius naevius- Unu provo solvi membroproblemon (in Esperanto). Okajimas Folia. Anat. Jpn. 28, 257-285.
20. Ura, R. (1956b). Pri la unuaj vojoj de sangvazoj kaj ilia sorto ce rajoj, trygon stephen (in Esperanto). Okajimas Folia. Anat. Jpn 28, 413-419.
21. Vernier, J.-M. (1969). Table chronologique du développement embryonnaire de la truite arc-en-ciel, Salmo gairdneri Rich. 1863. Ann. Embryol. Morph. 2, 495-520.
22. Weinstein, B. M., Stemple, D. L., Driever, W. D., and Fishman, M. C. (1995). gridlock, a localized heritable vascular patterning defect in the zebrafish. Nat. Med. 11, 1143-1147.
23. Yamada, H. (1951). The blood vascular system of the post-branchial alimentary canal in the lamprey (in Japanese). Okayama Igakukai Zasshi 63, 1-51.
24. Zieba, J. (1956). Die blutgefäße des Dottersackes des Hechtes (Esox lucius L.). (Poln., Dt. Zusammenfass.). Builet. Zakladu Biol. Stawow Polsk. Akad. Nauk. 3, 5-18.