As a tubular organ, the endothelial vasculature shares a lot of common features with epithelial tubes found in a number of other organs, such as lung, kidney, salivary glands and pancreas. A considerable amount of work has been carried out to elucidate epithelial tubulogenesis during the past several decades. Epithelial tubes are generally composed of a sheath of cuboidal or columnar epithelial cells, with defined apical membranes facing a central lumen, lateral edges interfacing with each other closely, and basal membranes making up the tube periphery (). Similarly, ECs of functional vessels consist of a luminal (apical) membrane facing the flowing blood and an abluminal (basal) membrane in contact with the basement membrane. The principal difference between epithelial and endothelial tubes is the smaller contact area comprising the junctional contacts (lateral membranes) between ECs ().
Junctional contacts in epithelial versus endothelial tubes
Nonetheless, a number of analogies between the two tissues can be made. Most epithelial tubulogenesis processes fall into one, or a combination of several, of the following categories: wrapping (i.e. vertebrate primary neurulation), budding (i.e. lung), mesenchymal-to-epithelial transition or MET (i.e. kidney tubules), cell hollowing (i.e. Drosophila trachea), cord hollowing (i.e. zebrafish gut), cavitation (i.e. salivary gland), and cell division and intercalation (i.e. zebrafish neural tube) [11
]. While MET reflects the overall process that angioblasts must accomplish to aggregate and form tubes, at least three additional mechanisms have been implicated in the later stages of endothelial lumen formation: budding, cell and cord hollowing [6
By definition, budding consists of formation and extension of a tube via growth from a pre-existing epithelium. During lung development, for instance, the pulmonary epithelium repeatedly buds and extends finger-like projections, until it becomes a contiguous, highly branched, ramifying, tubular tree. During blood vessel formation, sprouting angiogenesis is essentially synonymous with budding (). ECs along the wall of a blood vessel become locally activated, degrade the surrounding basement membrane, extend filopodia, change shape and migrate out, with tip cells at the leading front of the growing vessel, invading surrounding tissue. As angiogenic sprouting gives rise to patent vessels, it can thereby be considered one mechanism for lumen formation. However, what processes specifically drive stalk cells to organize, extend and maintain the parental lumen remain unclear.
2.2 Cell hollowing: vacuole fusion
In contrast to budding where new lumens extend directly from pre-existing lumens, cell hollowing
represents a different cellular mechanism whereby new lumens emerge intracellularly. In this case, lumens initiate as multiple small vesicles or larger vacuoles, that fuse to produce a central lumen, which in turn becomes connected with similar lumens in adjacent cells ( and
). This type of lumen formation has been observed in Drosophila terminal tracheal cells [15
], as well as in ECs. Indeed, until recently vascular lumen formation had primarily been studied in vitro
, and a large body of work using live imaging of ECs in 3D matrices has demonstrated that intraendothelial lumen formation occurs by cell hollowing via vacuole fusion [6
]. Vacuole fusion has also been observed in vivo
, during formation of intersegmental vessels in zebrafish [19
]. In these studies, live imaging of quantum dots injected intravascularly into circulating blood demonstrates intercellular fusion of vacuoles during ISV morphogenesis. Together, these observations suggested that cell hollowing is an important mechanism in both endothelial and non-endothelial tubulogenesis.
Vascular morphogenesis: Cell hollowing versus cord hollowing
Vascular tubulogenesis: mechanistic heterogeneity
2.3 Cord hollowing: vacuole fusion and cell rearrangement
A third type of mechanism by which some lumens form is termed cord hollowing. In this process, lumenal space is generated extracellularly between ECs, even as they remain joined peripherally (). Expansion of apical membranes to create an intervening extracellular space could be achieved either by addition of new lumenal membrane (), or by removal or clearance of junctions from the cord center. The result of either process is to build up the net surface area of the lumenal membrane.
During lumen formation, addition of new lumenal membrane has been to shown in some cases to occur via directed exocytosis of vesicles, which fuse with and expand the lumen at the cord center. This type of vectorial vesicle/vacuole fusion has been observed in cultured MDCK epithelial cells [12
], and has also been shown in ECs and suggested to be dependent on Rab7 [3
] (). Indeed, an interesting link was made between vectorial transport, cell adhesion and cell polarity in the ECs of arterioles in late gestation mouse embryos. In these vessels, loss of β1 integrin disrupted Par3 localization resulting in accumulation of excess Rab7+
cytosolic vesicles and failure of vascular lumen formation. This finding reflects the need for EC polarity during lumen formation, as ECs must be able to define and/or determine the ‘inside’ from the ‘outside’ of the cord to correctly target vectorial transport of vesicles/vacuoles.
Another possible mechanism for creating extracellular space between ECs during lumen formation is clearance of adhesion at the cell center. This can be accomplished either by de-adhesion at the apical/luminal membrane, and/or by redistribution of junctional molecules away from the cord center. In the former case, cells within a cord de-adhere from each other locally at the luminal membrane, but remain tethered at the lumen periphery. This differential control of adhesion thereby alters EC shape and rearranges ECs relative to each other. Control of de-adhesion, with precisely controlled regional removal of junctions at the cord center, is likely to be a highly regulated process. In the latter case, ECs actively redistribute existing junctions to the periphery, away from the lumen. Such junctional redistribution has been observed in the gut epithelium of zebrafish [22
] and more recently during vasculogenesis in mouse [23
To date no definitive experimental evidence has clearly distinguished endothelial lumenal membrane expansion (via centripetal vesicle/vacuole fusion) versus either de-adhesion or clearance of junctions during vascular tubulogenesis. It is likely that a number of cellular phenomena occur coordinately during this process. Indeed, in late mouse arterioles, both directed Rab7-directed vesicle transport to the lumenal membrane and junctional redistribution (away from the center) occur concurrently [3
]. It will be of great interest to assess whether tandem mechanisms apply more globally to forming vessels.
2.4. Vascular tubulogenesis: mechanistic heterogeneity
As the number of studies on vascular tubulogenesis has increased, our understanding of the variety of underlying mechanisms has evolved [7
]. Until recently, cell hollowing was considered a common, if not predominant, mode of de novo
endothelial lumen formation. However, additional mechanisms have since been identified. Here, we cover a few divergent examples.
Clear live imaging of ECs cultured in 3D matrices provided strong support for vacuole-based lumen formation [17
]. Cells were shown to generate intracellular vacuoles that would align at the cell center and fuse with each other to create lumens. Similarly, vacuole fusion has been observed in the growing vessels of vertebrate embryos [19
] (). Live imaging of growing zebrafish intersegmental vessels (ISVs) identified fusing vacuoles during lumen formation [19
]. This high resolution study examined the development of ISVs, which were thought to assemble stereotypically with three ECs in a head-to-tail cord along myotomal boundaries [28
]. Two photon live cell imaging identified vacuoles during angiogenic sprouting and suggested intra-cellular fusion of endothelial vacuoles at the center of ISV ECs suggesting cell hollowing.
More recent ISV studies have confirmed vacuole formation and fusion as a basis for ISV lumen formation [26
]. However, rather than intra-cellular vacuole fusion, they observed inter-cellular fusion in ISVs. Careful examination of membrane fusion events, using apical markers and injected dyes to outline functional lumens, revealed that vacuoles were fusing with and establishing the luminal membrane, ‘between’ rather than ‘inside’ cells, suggesting cord hollowing rather than cell hollowing. In addition, this study identified a requirement for the apical ERM domain protein moesin-1 and adherens junctions during this process. Cord hollowing was further supported by a recent study which examined tight junctions and adherens junctions in zebrafish ECs [27
]. Adjacent ISV ECs were observed to share large cell-cell junctional contacts, and lumens were shown to form extracellularly, between closely apposed cells. Tools and fish lines developed for these studies will undoubtedly continue to refine our understanding of angiogenic lumen formation in the future.
One strikingly unique example of a divergent mechanism of vascular lumen formation involves aortal and caudal vein lumen morphogenesis in zebrafish during arteriovenous segregation [29
] (). High-resolution live imaging of zebrafish vascular morphogenesis showed that the primary axial dorsal aorta (artery) arises first by vasculogenic aggregation of angioblasts. Secondarily, the caudal vein emerges via selective ventral sprouting and migration of angioblasts away from the dorsal aorta, a process they show to be regulated by EphrinB2-EphB4 signaling. Interestingly, the caudal lumen forms via angioblasts aggregating into a partially formed and dorsally ‘open’ vessel, which then rapidly fills with blood accumulated at the interface of the two vessels. Upon this ‘filling’ of the caudal vein, functional blood circulation is then rapidly established. The cellular mechanisms underlying this ‘open’ vessel formation remain to be examined, but support the idea that a variety of mechanisms underlie vascular tubulogenesis in different vessels.
Yet another unique mechanism for vessel lumen formation was reported during Drosophila heart morphogenesis [30
] (). In flies, the heart represents an open-ended and contractile endolymph ‘vessel’ of sorts, with tubular morphology and a distinct internal lumen. In these studies, formation of the cardiovascular lumen by two parallel rows of myoendothelial cells is shown to be modulated by membrane repulsion. Slit-Robo signaling downregulates E-cadherin at the lumenal cell-surface and prevents fusion between apposed cells. Instead, the two cells form junctions only at their dorsal- and then ventral-most regions, resulting in the formation of an internal central lumen, enclosed by two rows of “C-shaped” cells. Even more interestingly, and distinct from typical epithelial or endothelial tubulogenesis, the surface that lines the inner lumen in the fly heart lumen is the basal rather than apical cell surface. This difference suggests that fundamental differences can occur between different types of tube forming cells with respect to cell polarity features.
Similar apical/luminal surface repulsion has also been suggested during mouse dorsal aortae formation. Following cord hollowing, observations by Lammert and colleagues showed that accumulation of negatively charged sialomucins along apposed lumenal membranes results in the initial opening of slit-like spaces between ECs [32
] (). Specifically, they show that sialic acids of lumenal glycoproteins create repulsive electrostatic fields that result in membranes moving away from each other at the cord center. Neutralizing these charges with injection of cationic protamine sulfates inhibits normal lumen formation.
Taken together, these divergent examples of lumen formation mechanisms suggest that different blood vessels form via a range of different cellular mechanisms. It is not completely unexpected, as endothelium is known to display a high level of heterogeneity across different tissues [34
]. We propose that distinct or even yet to be discovered ‘novel’ mechanisms of lumen formation will likely be identified as vascular beds of different organs are more extensively examined and understood.
2.5. Cord hollowing example: Embryonic dorsal aortae
Cord hollowing has recently been demonstrated and thoroughly characterized during murine dorsal aortae lumen formation [23
] (). These cellular events are striking in that they constitute the first embryonic vasculogenic event, occurring prior to heart beating or hemodynamic flow [23
]. Aortic cords form as early as the 1 somite stage (1S), as 2 or 3 adjacent ECs contact each other and adhere along their entire interface. ECs at this stage are cuboidal and plump. Around 1-2S, junctions are cleared from the central region of the cord and EC luminal membranes de-adhere from each other. ECs then bend, forming a slit-like space between them. At 2-3S, the slit enlarges, eventually forming a lumen encircled by 3–4 more flattened ECs. From this time onward, the aortic lumens continue to enlarge and ECs continue to flatten and proliferate. By 8S, the aortic lumens are quite large and between 6–8 cells make up the aortic vessel circumference.
Cord hollowing appears to take place in a similar manner in the single midline aorta of zebrafish [35
]. However chick aortae display large vacuoles suggesting possible cell hollowing [36
]. Studies of aortae formation in additional species, and comparison with other vessel types, will determine the extent of the use of cord versus cell hollowing. It will also be interesting to establish whether this mechanism is distinct to early vasculogenic vessel formation, or whether it will prove more universal and be observed in additional vessel types.
2.6. Relationship of vascular tubes with surrounding microenvironment
One open question is the influence of surrounding cells on blood vessel tubulogenesis. An interesting relationship between ECs and surrounding mesenchyme has been recently noted in murine aortae, where ECs are tightly associated with the underlying endodermal epithelium and encased in the overlying paraxial mesoderm [23
]. In one model of vascular tubulogenesis failure, Rasip1 null embryos, aortic vessels never transition from cords to tubes, and interestingly, the mesoderm surrounding the failed aortic cords appears to autonomously ‘open’ a cavity at the location where aortic vessels would normally have formed. This suggests an active contribution by the mesoderm in providing support to ECs and perhaps ‘pulling open’ aortic lumens.
An alternative explanation for the open mesodermal ‘space’ is that ECs may digest surrounding mesenchyme-derived ECM to clear space for tube growth and expansion. This idea is supported by observations in an in vitro
3D system, where ECs cultured in a collagen matrix digest canals via membrane bound matrix metalloproteinase (MMP) dependent proteolysis [37
]. ECs subsequently flatten against the walls of these canals and organize into vessel-like tubes, assuming a characteristic cobblestone appearance along the wall of these spaces through EC-matrix contacts [39
]. The EC-generated physical spaces, or ‘guidance tunnels’ within the 3D ECM thus facilitate lumen and tube network formation [5
]. Additional studies however will be required to elucidate the reciprocal influence of ECs and surrounding cells or matrix in vivo,
during vasculogenesis and vascular tubulogenesis.