Luminal pockets undergo oscillations with local constrictions along junctions

In zebrafish embryos, continuous lumen formation in the DLAV involves the creation and coalescence of luminal pockets between adjacent cells (Fig. 1A). During this process, initial contact between the tip cells at the top of sprouts is established by filopodial interactions (Fig. 1A(I)). Stable contacts are subsequently established through the deposition of adherens junction (AJ) proteins (Fig. 1A(II)). A de novo apical compartment is then formed at the adhesion site, creating a luminal pocket with junctions localized at the periphery (Fig. 1A(III)). The tip-tip luminal pocket expands through cell rearrangement and connects with other luminal pockets formed between tip and stalk cells, facilitating the establishment of vascular patency (Fig. 1A(IV)). By employing ZO-1 tdTomato and GFP-Podxl1 as markers for junctional and apical domains, respectively, we monitored the initiation and enlargement of tip-tip and tip-stalk luminal compartments (Fig. 1B; Video 1). Initially, a significant fraction of GFP-Podxl1 was observed outside the expected apical regions. However, GFP-Podxl1 predominantly relocated to the presumed apical compartments within the junctional rings as the luminal pockets expanded. To further investigate the dynamics of cell-cell contact establishment and luminal pocket expansion, we quantified changes in the size of the luminal compartments within junctional rings over a 4-hour period starting from 30 hpf (Fig. 1C). Both the tip-tip and tip-stalk junctional rings exhibited an overall increase in size. Nonetheless, intermittent decreases (blue regions) in size were noted, suggesting oscillatory behaviours characterized by alternating phases of expansion and constriction.

Fig. 1

Oscillatory constrictions along junctions maintain the shape of luminal pockets. (A) Schematic depicting the anastomosis of two tip cells in the formation of the DLAV. (I) Tip cells establish initial contact with filopodia. (II) Stable cell-cell contact is formed with the deposition of junctional proteins, including Cdh5. (III) At the cell-cell contact site, a luminal pocket is formed, with the apical membrane positioned centrally and encircled by a junctional ring. (IV) Cell rearrangement leads to the expansion of luminal pockets. (B and B’) Time-lapse imaging of ZO-1 tdTomato and Cdh5-Venus, alongside a corresponding schematic illustrating the formation and expansion of luminal pockets. (C) Changes in the size of luminal pockets displayed in (B), color-coded to match the regions in (B’). (D) Sequential imaging capturing the transition from initial filopodia contact to the establishment of well-defined luminal pockets, using mosaically labeled Myl9a-GFP and mRuby2-UCHD. (E) Time-lapse visualization of Myl9a-GFP and mRuby2-UCHD, revealing dynamic local enrichment of Myl9a-GFP along junctions across different spatial and temporal scales. White arrowheads indicate the localized enrichment of Myl9a. (F) Conversion of Myl9a-GFP signals observed in (E) into a 2D representation along junctions, with the x-axis denoting ring positions and the y-axis representing time. Dashed black lines indicate the progressive narrowing of Myl9a-GFP signals during each pulsatile event. (G) Time-lapse imaging showcasing localized constriction events as highlighted in (F), employing Myl9a-GFP and mRuby2-UCHD labeling. (H) Dynamic visualization of Myl9a-GFP and ZO-1 tdTomato illustrating how constrictions facilitate the straightening of junctions over time. White arrowheads indicate the localized enrichment of Myl9a

To investigate the oscillatory behaviors of junctional rings, we utilized mRuby2-UCHD and Myl9a-GFP as markers for the actin-myosin cytoskeleton (Fig. 1D). Mosaic labeling of UCHD and Myl9a in tip cells allowed for the precise distinction between the two cells during anastomosis. Initial contact between the two tip cells was established via filopodia (Fig. 1D(I)). Subsequently, Myl9a enrichment was observed at the stabilized contact site (Fig. 1D(II)). After the formation of the apical compartment, a significant portion of Myl9a localized along the junctional ring encircling the luminal pocket (Fig. 1D(III)). Detailed examination of Myl9a localization revealed its dynamic enrichment along various junctional regions at different time points (Fig. 1E; Video 2). Next, we unrolled the Myl9a distribution along the junctional ring and projected it onto a 2D map, with the x-axis representing relative positions and the y-axis representing time, to assess the dynamics of Myl9a along the junctions over time intervals ranging from 45 to 78 min (Fig. 1F). In line with prior observations, we detected concentrated Myl9a enrichment at specific junctional sites, exhibiting brief temporal pulses. Concurrently, we noted a reduction in the size of these enriched regions within each pulse, indicating localized constrictive events (dotted black lines). Consistent with this analysis, we observed Myl9a enrichment accompanied by localized constrictions at junctions during the pulse (Fig. 1G). Live imaging with ZO-1 indicated that these constrictive events straightened discontinuous or zigzag-shaped junctions following the expansion of luminal pockets (Fig. 1H; Video 3). Taken together, we observed oscillatory constrictions at the endothelial cell-cell junctions, as indicated by multiple junctional and apical reporters. The local accumulation of Myosin, demonstrated by Myl9a reporters, suggests the presence of actomyosin-mediated contractions at these junctions, hereafter referred to as junctional contractility.

heg1 and krit1 mutants display contorted junctions and fragmented apical domains

Rasip1, Heg1, and Krit1 are involved in the regulation of Rho GTPases, which are implicated in regulating endothelial contractility. Rasip1 and Krit1 bind to Heg1, a transmembrane protein, at different binding sites on the cytoplasmic domain of Heg1 (Fig. 2A) [19, 20]. We subsequently investigated the roles of Rasip1, Heg1, and Krit1 in the regulation of cell-cell junctions through analyses of rasip1ubs28, heg1m552, and krit1t26458 mutants [13, 32, 33]. Immunostaining of ZO-1 revealed that heg1 and krit1 mutants exhibited zigzag or ectopically contorted junctions, in contrast to the ring-shaped junctions observed in wild-type embryos (Fig. 2B and C). These observed phenotypes sharply contrasted with those of rasip1 mutants, which displayed ectopic junctional materials in the apical compartments. In contrast, the apical compartments in heg1 and krit1 mutants were largely devoid of junctional complexes (Fig. 2B and D). In wild-type embryos, Rasip1 was largely restricted to the presumed apical domains inside the junctional rings (Fig. 2C). Similarly, Rasip1 was found at the cell-cell interfaces in heg1 and krit1 mutants, indicating that its apical localization is largely independent of Heg1 and Krit1 (Fig. 2C). However, in heg1 and krit1 mutants, the apical domains exhibited fragmented distributions within the loops of contorted junctions, which were connected by bottlenecked regions (Fig. 2C). Thus, the contorted junctions observed in heg1 and krit1 mutants appeared to disrupt the integrity of the apical domains. We observed similar phenotypes using the Cdh5-Venus live reporter. In wild-type embryos, the junctional rings approached each other, with each ring straightened along the direction of the vessel (Fig. 2E). In contrast, the boundaries of the presumed apical compartments in rasip1 mutants appeared fuzzy and discontinuous, with widespread distribution of Cdh5-Venus (Fig. 2F). Although the junctions in heg1 and krit1 mutants appeared sharp, they were not straightened, and the ectopic twisting persisted, as seen in live imaging (Fig. 2G and H).

Fig. 2

Zigzag and contorted junctions occur in heg1 and krit1 mutants. (A) Diagram showing interactions between Rasip1, Heg1, and Krit1. (B and C) Immunofluorescence staining of ZO-1 alone (B) or in combination with Rasip1 (C) in wild-type embryos, rasip1, heg1, and krit1 mutants at 32 hpf. White arrowheads indicate the contorted junctions in heg1 mutants and krit1 mutants. (D) Quantification of the Cdh5 boundary-to-apical ratio in wild-type embryos (n = 53), rasip1 mutants (n = 55), krit1 mutants (n = 62), and heg1 mutants (n = 43), presented as mean ± SD. (E-H) Time-lapse imaging of Cdh5-Venus in wild-type embryos (E), rasip1 mutants (F), heg1 mutants (G) and krit1 mutants (H) at DLAV starting from 33 hpf. White arrows indicate the poorly defined junctions in rasip1 mutants. White arrowheads indicate the contorted junctions in heg1 mutants and krit1 mutants

Aberrant control of cell-cell interface shape in heg1 and krit1 mutants

To determine whether the contorted shape of junctions reflects extended/twisted cell-cell interfaces or an increased number of interconnections between cells, we performed live imaging on the initial adhesion sites and followed cell rearrangements (Fig. 3A-C). In wild-type embryos, most initial adhesion sites between two cells opened into circular junctional rings (Fig. 3A and D). In contrast, a significant proportion of single adhesion sites transitioned directly into “∞”-shaped junctions with two interconnected loops in heg1 (27/71) and krit1 (32/69) mutants, while the remaining adhesion sites transformed into more or less circular junctional rings (Fig. 3B-D). Thus, the ectopic “∞”-shaped junctions consisted of a continuous cell-cell interface between two adjacent cells, rather than multiple interconnected rings formed by three or more cells. To elucidate alterations in the shape of cell-cell interfaces in heg1 and krit1 mutants, we first quantified the shape of the one-loop junctional rings in the two mutants and compared them to those in wild-type embryos. The zigzag index (Z) was defined as the ratio of the perimeter of the junctional ring (P_junction) to the perimeter of the fitted ellipse (P_fitted_ellipse) (Fig. 3E). Our quantifications revealed an increased zigzag index in heg1 and krit1 mutants, even in the relatively normal portions of junctions (Fig. 3E’). To understand the geometry of these “∞”-shaped cell-cell interfaces and the shape of ECs, we performed live imaging with mosaic labeling of ECs in the DLAV (Fig. 3F; Video 4). Reconstructed orthogonal sections of the “∞”-shaped junctions showed that the labeled tip cell was positioned above the other cell in one loop, but below it in the other loop (Fig. 3F(III)). 3D reconstruction of “∞”-shaped junctions demonstrated that the two loops were in different z-planes, with an intersecting point where one junction lay on top of the other (Fig. 3G-G’’; Video 5). Thus, instead of being positioned opposite each other, the two tip cells intersected and lay on both the top and bottom sides of each other, leading to the formation of two interconnected but flipped interfaces.

Fig. 3

Ectopic shape control of endothelial cell-cell interfaces in heg1 and krit1 mutants. (A-C) Time-lapse imaging of Cdh5-Venus depicting the transition from junctional patches to circular rings in wild-type embryos (A) and “∞”-shaped junctions in heg1 mutants (B) and krit1 mutants (C). (D) Distribution of different junctional types in wild-type embryos (n = 57), rasip1 (n = 73), heg1 (n = 71), and krit1 (n = 69) mutants. (E and E’) Definition and quantification of the zigzag index in wild-type embryos (n = 22), heg1 (n = 22) and krit1 (n = 30) mutants. (F and F’) Time-lapse imaging and corresponding schematic illustrating “∞”-shaped junction formation in krit1 mutants with Cdh5-Venus and mosaically expressed GFP. Cross-sectional views along dashed lines in F(III) are shown in F’(III) and F’(IV), with blue asterisks indicating GFP-positive cells and white arrowheads highlighting crossing points between loops. (G) Visualization of “∞”-shaped junctions at various Z positions with GFP labeling in a single cell. GFP fluorescence at the upper loop was detected at 1.4 μm (blue), contrasting with the lower loop where GFP was visible at 0 μm (red), indicating differential Z-plane localization of the two loops. (G’) 3D diagrams of cell shape at contorted cell-cell interfaces, with color coding based on Z positions. (G’’) 3D reconstructions of (G) viewed from different angles. *** P < 0.001, ns P > 0.05 (t-test)

Heg1 and Krit1 are required for maintaining junctional contractility

The zigzag and twisted junctions observed in heg1 or krit1 mutants suggested that the actomyosin contractility along the junctions might be compromised. To test this hypothesis, we utilized Myl9a-GFP as a reporter of actomyosin activity in rasip1, heg1, and krit1 mutants, as we did in wild-type embryos (Fig. 4A and B). In wild-type embryos, Myl9a predominantly localized along the junctions. In rasip1 mutants, consistent with the role of Rasip1 as an inhibitor of actomyosin contractility at the apical compartment, Myl9a-GFP was ectopically enriched within the apical compartments. Our prior investigation revealed that heightened myosin activity at the apical compartments (hereafter referred to as apical contractility) pulled junctional complexes from the junctions into the apical compartments, thus destabilizing cell-cell junctions [31]. In contrast, there was no obvious apical or junctional enrichment of Myl9a-GFP in heg1 and krit1 mutants, suggesting a loss of junctional contractility as we hypothesized (Fig. 4A and B). Live imaging of heg1 and krit1 mutants revealed the rapid expansion of luminal pockets between the tip and stalk cells, characterized by zigzag boundaries, and the formation of “∞”-shaped junctions between tip cells (Fig. 4C-E; Video 6–8). Quantitative analysis showed a lack of oscillatory constrictions in heg1 and krit1 mutants, in sharp contrast to wild-type embryos (Fig. 4C’-E’). Thus, while Rasip1 inhibits actomyosin contractility within the apical compartments and along the junctions, Heg1 and Krit1 maintain actomyosin contractility along the junctions. These observations align with the previously identified role of Krit1 in recruiting Rock2 to VE-cadherin–β-catenin complexes [18].

Fig. 4

Heg1 and Krit1 control contractility along junctions. (A) Myl9a-GFP and Cdh5-Venus in wild-type embryos, rasip1, heg1, and krit1 mutants at 32 hpf. White arrowheads indicate junctional Myl9a, while white arrow point to apical Myl9a in rasip1 mutants. (A’) Quantification of Myl9a-GFP and Cdh5-Venus intensities along the dashed lines shown in (A). (B) Normalized Myl9a intensities at junctional regions and apical compartments in wild-type embryos (n = 25), rasip1 mutants (n = 18), heg1 mutants (n = 34), and krit1 mutants (n = 21), presented as mean ± SD. Myl9a-GFP intensity is standardized using the mean signal level from the entire cell. (C-E) Time-lapse imaging of Cdh5-Venus in wild-type embryos (C), heg1 mutants (D), and krit1 mutants (E), illustrating the expansion of luminal pockets between stalk and tip cells and the establishment of luminal pockets between tip cells. Stalk-tip luminal pockets appear zigzag (red arrows), while newly established tip-tip luminal pockets exhibit “∞”-shaped structures (white arrows) in heg1 (D) and krit1 mutants (E). (C’-E’) Changes in the size of stalk-tip luminal pockets in (C-E). (F) Illustration of recombined Heg1-GFP with GFP inserted in the middle of the extracellular domain. (G) Visualization of Heg1-GFP and mRuby2-UCHD in wild-type embryos at 32 hpf, showing the local recruitment of Heg1 to junctions during constrictions. White arrowheads highlight locally enriched Heg1-GFP. (H) Kymographs generated along the dashed lines in (G), demonstrating the recruitment of Heg1-GFP during local constrictions. White arrowheads indicate the recruitment of Heg1-GFP at the onset of constrictions

To investigate the involvement of Heg1 and Krit1 in oscillatory local constrictions at the protein level, we engineered a Heg1 reporter line with GFP fused to its extracellular domain [Tg(fli: gal4; UAS: Heg1-GFP61] (Fig. 4F). Notably, Heg1-GFP exhibited strong intermittent enrichment at local regions of the junctions, closely associated with instances of local constrictions (Fig. 4G; Video 9). Analysis using kymographs constructed from Heg1-GFP and mRuby2-UCHD signals indicated that Heg1-GFP was recruited to the constricting junctions at the onset of the constriction phase, persisting until the local region expanded once more (Fig. 4H). Similarly, Rasip1-Scarlet-I, akin to Heg1-GFP, exhibited dynamic recruitment to the sites of junctional constriction (Fig. S1A and S1B). Another potential interactive partner, Radil2a (Radil-B), showed similar dynamic shuttling between the apical compartments and junctions during local constrictive events [Tg(fli: gal4; UAS: GFP-Radil2a)ubs62] [19] (Fig. S1C and S1D). Combined with the functional studies, these observations indicate that Heg1 and its interactive partners are dynamically recruited to junctions and are required for regulating local actomyosin contractility.

Essential junctional contractility shapes cell-cell interfaces and apical integrity

To explore the link between junctional contractility and the regulation of cell shape and apical connectivity, we investigated the establishment of cell-cell interfaces and myosin activities during anastomosis in wild-type embryos and krit1 mutants. In wild-type embryos, the initial adhesion site underwent dramatic remodelling, with the transient establishment and subsequent collapse of immature junctional rings before the formation of stable luminal pockets (Fig. 5A). The immature junctional ring underwent constriction and collapse, accompanied by increased Myl9a enrichment, thereby further narrowing the interface between two tip cells (Fig. 5A(II and III)). Immature junctional rings were also observed in krit1 mutants within the partially opened cell-cell interfaces between two tip cells (Fig. 5A(I) and 5B(I)). However, in contrast to wild-type embryos, the immature junctional rings persisted in krit1 mutants (Fig. 5B(II)). In the absence of Myl9a enrichment and oscillatory constrictions, the linear cell-cell interface between two tip cells continuously extended (Fig. 5B(III), Fig. 3B and C). New junctional loops formed at the extended linear interface, likely on the reverse side of the old loop, resulting in “∞”-shaped junctions in krit1 mutants (Fig. 5B(IV)). We quantified the length of the linear interface between two tip cells before its transformation into junctional rings (white lines in Fig. 5A and B). Our analysis revealed that heg1 and krit1 mutants displayed significantly longer linear interfaces, possibly due to the absence of effective junctional contractility (Fig. 5C). This prolonged interface increased the likelihood of forming multiple junctional loops on both sides, leading to distorted junctions and fragmented apical domains.

Fig. 5

Essential junctional contractility refines cell-cell interactions. (A and B) Time-lapse imaging of Cdh5-Venus and Myl9a-GFP showing the transition from junctional patches to circular rings in wild-type embryos (A) and “∞”-shaped junctions in krit1 mutants (B). White arrows denote nascent or stabilized junctional rings, while white lines indicate the linear junctional interface before it opens into junctional rings. (C) Quantification of the length of linear junctional interfaces between tip cells prior to their conversion into junctional rings in wild-type embryos (n = 17), heg1 mutants (n = 13), and krit1 mutants (n = 15). (D and D’) Time-lapse imaging and corresponding diagrams depicting the untying of “∞” shape junctions in wild-type embryos with Cdh5-Venus and mosaically expressed Myl9a-GFP. White arrows highlight the resolution of loops. (E) Time-lapse imaging of Heg1-GFP and mRuby2-UCHD, revealing the local enrichment of Heg1-GFP during the retraction of junctional loops to disentangle “∞”-shaped junctions in wild-type embryos. (F) Schematic diagram of the single-component opto-RhoA system. Blue light activation induces dynamic membrane recruitment of cytosol-sequestered RhoA-BcLOV4-mCherry. (G) Activation of opto-RhoA straightens twisted junctions in krit1 mutants and corresponding diagrams. White arrowheads indicate the opto-RhoA clusters enriched along junctions after activation. *** P < 0.001, ** P < 0.01 (t-test)

Intriguingly, contorted junctions were also observed in a fraction of wild-type embryos (9/57) (Fig. 5D; Video 10). In wild-type embryos, most twisted rings (7/9) were disentangled by the retraction of one of the two loops. This process resulted in the formation of continuous luminal pockets encircled by circular junctions. Concurrently, we observed a local enrichment of Myl9a and Heg1 along cell-cell junctions close to the constricting site during the retraction (Fig. 5D and E). In contrast, the contorted junctions persisted (Fig. 2G and H), and even relatively normal circular rings twisted into contorted shapes in both heg1 and krit1 mutants (Fig. S2A-C). Hence, we hypothesized that junctional contractility is required to straighten junctions, thereby controlling the shape of cell-cell interfaces and ensuring proper apical integrity in ECs during anastomosis. To test this hypothesis, we applied opto-RhoA at the twisted junctions in krit1 mutants to induce local activation of RhoA signaling [35] (Fig. 5F). After continuous 450 nm illumination at the region of interest (ROI), the coiled junctions became straightened and untied with the recruitment of RhoA-BcLOV4-mCherry clusters to the junctions (Fig. 5G; Video 11). Hence, sufficient junctional contractility is required for the straightening of junctions and the removal of excessive cell-cell interfaces, thereby preserving the appropriate shape of cell-cell interfaces and apical integrity.

heg1 and krit1 mutants fail to establish stable cell-cell contact and interconnected luminal space

So far, our findings reveal that the excessive cell-cell interfaces and fragmented apical domains in heg1 and krit1 mutants are due to a lack of junctional contractility. These phenotypes were observed and studied in ECs during the anastomosis of the DLAV at around 32 hpf before the establishment of blood flow. We now ask how the proper control of cell-cell interfaces and apical integrity are relevant to the formation of perfused blood vessels. In wild-type embryos, the formation of interconnected luminal space within the vasculature involves two different cellular mechanisms referred to as type I and type II anastomosis [7, 30, 36]. In type I anastomosis, blood pressure pushes the luminal space from one cell to its connecting neighbour. The apical membrane on the expanding lumen fuses with the de novo inserted apical membrane within the luminal pocket, thereby generating a continuous intracellular lumen (Fig. 6A and A’). In contrast, type II anastomosis can occur in the absence of blood pressure. The connection of luminal space is driven by the expansion of luminal pockets via the rearrangement of cells. The de novo inserted luminal pockets eventually connect with each other, leading to lumen coalescence and the formation of a multicellular tube (Fig. 6B and B’).

Fig. 6

Actomyosin contractile forces and blood flow shape endothelial cell-cell interactions. (A and B) Time-lapse imaging of Cdh5-Venus and GFP-Podxl1 illustrating lumenization in type I and type II anastomoses in wild-type embryos. In type I anastomosis, the apical membrane invaginates due to blood pressure (white arrows). In type II anastomosis, luminal pockets expand through cell rearrangement, resulting in lumen coalescence (white lines). White arrowheads mark the distal end of junctional rings, while white arrows indicate the lumenized vessel. White lines label luminal spaces. (A’ and B’) Schematic diagrams depicting the establishment of continuous luminal space under type I and type II anastomoses. (C-E) Time-lapse imaging of Cdh5-Venus showing the collapse of contorted junctions in heg1 (C) and krit1 (D) mutants, with a corresponding diagram (E). White arrowheads denote junctions undergoing collapse. (F) Time-lapse imaging of ZO1-GFP in sih morphants at the same stage, with corresponding diagram (F’). (G and H) Lumenization at DLAV inflated “∞”-shaped junctions into circular junctions in perfused tubes as seen in wild-type embryos (2/45 with “∞”-shaped junctions before lumenization) (G) and krit1 mutants (2/33 with transient lumenization) (H). (I) Graphic diagrams illustrating blood pressure-driven lumenization through circular junctions or “∞”-shaped junctions

Zebrafish heg1 and krit1 mutants form massively dilated hearts and fail to establish blood flow [13, 33, 37]. Consequently, type I anastomosis is unlikely to occur in heg1 and krit1 mutants due to the absence of blood pressure. The contorted junctions in heg1 and krit1 mutants further twisted and eventually collapsed into dense junctional patches or linear structures (Fig. 6C-E; Video 12 and 13). The collapsed junctions failed to connect with other junctions, thereby hindering type II anastomosis and preventing the formation of interconnected luminal spaces. The fragmented apical domains collapsed with the surrounding junctions in krit1 mutants (Fig. 6D and E). These observations suggest that, without effective straightening and retraction, the excessive cell-cell interfaces and fragmented apical domains in both mutants are unstable and prone to collapse. To test whether blood flow is required for the straightening of junctions, we injected antisense morpholinos against cardiac troponin (silent heart, sih, tnnt2a) and found that junctional rings were straightened along the vessels in the absence of blood flow (Fig. 6F and F’). Thus, the formation of interconnected luminal space in heg1 and krit1 mutants was impeded by the absence of blood flow and inadequate cell rearrangement. The insufficient cell rearrangements are independent of blood flow but are subject to the excessive and unstable cell-cell contacts in the absence of junctional contractility in both mutants.

Actomyosin contractile forces and blood flow together shape endothelial cell-cell interactions

There is curiosity regarding the fate of the contorted junctions if blood flow were to resume, as a previous study has shown that blood flow suppresses vascular anomalies in zebrafish krit1 mutants [37]. In fact, a very small fraction (2/45) of wild-type embryos displayed “∞”-shaped junctions before lumenization (Fig. 6G; Video 14), and a very small portion (2/33) of krit1 mutants displayed transient lumenization at the DLAV (Fig. 6H; Video 15), providing opportunities to explore type I anastomosis with contorted junctions. Under blood pressure, the “∞”-shaped junctions were inflated, with the intersecting points of the two loops detaching and moving to opposite sides of the perfused compartment (Fig. 6I). The inflated “∞”-shaped junctions became circular in the tube, as the disentangling occurred in 3D, in contrast to the 2D disentangling with loop retraction (Fig. 6I). Thus, blood flow facilitates the formation of continuous luminal spaces through contorted junctions. Hemodynamic forces appear to render contorted junctions less hazardous, potentially by opening luminal spaces through contorted junctions and restoring tension.