Introduction

Vascular networks are formed by an array of morphogenetic processes, such as sprout outgrowth, anastomosis and pruning, which depend on the tight coordination of dynamic endothelial cell behaviors (Betz et al., 2016; Ellertsdóttir et al., 2010; Phng, 2018; Schuermann et al., 2014). In order to transport blood, the newly formed blood vessels need to form a continuous lumen. Studies of lumen formation in mammalian and zebrafish embryos have revealed that the early blood vessels of the trunk are formed by different morphogenetic processes. The larger axial vessels, the dorsal aorta (DA) and posterior cardinal vein (PCV), form their lumens between coalescing endothelial cells, a process referred to as cord hollowing (Ellertsdóttir et al., 2010; Jin et al., 2005; Strilić et al., 2009; Zeeb et al., 2010). In vivo time-lapse analyses of the formation of the intersomitic vessels (ISV) and the dorsal longitudinal anastomotic vessels (DLAV) in zebrafish have revealed the existence of two alternative mechanisms of lumenization. In one mechanism, called transcellular lumen formation, the luminal cell membrane invaginates into the cell. This cell hollowing process is driven by blood pressure and results in so-called seamless tubes, which contain a lumen surrounded by single cells (Betz et al., 2016; Gebala et al., 2016; Herwig et al., 2011 ; Lenard et al., 2013; Phng and Belting, 2021). Alternatively, local lumens, which are formed between individual cells within a sprout can be fused by cell rearrangements (Herwig et al., 2011; Sauteur et al., 2014). Such endothelial cell rearrangements generate multicellular tubes (Betz et al., 2016; Phng and Belting, 2021). Studies examining lumenization of ISV and DLAV have shown that transcellular lumen formation requires blood pressure and that, in the absence of blood flow, endothelial cell rearrangements appear to be the default mechanism of vascular tube formation (Herwig et al., 2011; Lenard et al., 2013).

Endothelial cells are very motile - they can move within the angiogenic sprout and also in patent vessels in the presence of blood flow (Blum et al., 2008; Franco et al., 2016; Jakobsson et al., 2010). A dynamic balance between endothelial cell-cell adhesion and plasticity allows angiogenic sprouting while maintaining the endothelial seal. Previous analysis on blood vessel formation and anastomosis in zebrafish has shown that junctional remodeling is central to many aspects of morphogenetic endothelial cell-cell interactions. In particular, the adhesion molecule VE-cadherin (Cdh5) is essential for coordinated cell shape changes during multicellular tube formation, as loss of VE-cadherin was shown to inhibit cell rearrangements (Sauteur et al., 2014). Furthermore, rather than supporting a passive role of VE-cadherin, which is consistent with the maintenance of vascular integrity, an active force-generating function for VE-cadherin in this process was suggested (Sauteur et al., 2014). By performing in vivo high-resolution time-lapse imaging and genetic analyses, we have shown that endothelial cells use oriented, oscillating junction-based lamellipodia (JBL) to move over each other, which thus provide the physical means for cell rearrangements to drive multicellular tube formation (Paatero et al., 2018).

To better understand the cellular and molecular mechanisms of JBL function we have undertaken a detailed analysis of junctional and actomyosin dynamics. Taken together, our data show that JBL provide forward movement by generating pushing and pulling forces at the poles of junctional rings thereby elongating the cell and promoting cell rearrangements.

Results

Junction-based lamellipodia generate luminal connectivity

Lumenization of vascular sprouts is central to blood vessel formation. Although different morphogenetic mechanisms of vascular tube formation have been described, in vivo analysis in zebrafish embryos has established cord hollowing as the default mechanism of blood vessel formation in the absence of blood flow (Herwig et al., 2011; Lenard et al., 2013). During the cord hollowing process neighboring cells form local lumens which become interconnected via cell rearrangements (Andrew and Ewald, 2010). To visualize the relationship between JBL dynamics and endothelial cell rearrangements, we used two different transgenic reporters: EGFP-ZO1 to label EC junctions and UCHD-mRuby to label F-actin in junctions as well as in JBL protrusions (Figure 1a, b). Time-lapse analysis revealed that JBL formed in front of elongating junctions as the junctional rings move towards each other. As a consequence of these movements, the junctional rings merge and the converging cells form a new contact leading to the fusion of the previously separated luminal pockets (Figure 1c, d) (Herwig et al., 2011). In order to visualize the fusion of such luminal pockets we followed the localization GFP-Podxl, which labels the apical cell membrane, together with F-actin reporter UCHD-mRuby2, to delineate endothelial cell-cell junctions and F-actin base protrusions (Figure 1c). Time-lapse recordings revealed that fusion of apical membrane compartments occured within a few hours upon initiation of DLAV formation and was associated with junctional rearrangements between the three associated cells (Figure 1c, d and video s2). In agreement with our previous work, higher temporal resolution imaging showed that cell junctions formed transient F-actin-based protrusion (JBL) at the pole with respect to junctional ring elongation (Figure 1 c, d and video s1, and video s3) (Paatero et al., 2018).

Figure 1.

JBL form distal junctions by de novo recruitment of VE-cadherin

We have previously shown that JBL are oscillating protrusions, which drive junction elongation and thereby cell rearrangements during tube formation. A salient feature of JBL is the formation of a distal junction in front of the JBL (Paatero et al., 2018).

To gain more insight into the dynamics of distal junction formation, we compared the distribution of F-actin (Ruby-UCHD) with that of VE-cadherin (VE-cad-Venus) and ZO1 (EGFP-ZO1), respectively (Figure 2). Time-lapse experiments revealed distinct and dynamic localization of these proteins. During early phases of JBL formation VE-cadherin was diffusely distributed throughout the protrusion and accumulated in aggregates leading to a characteristic “double junction” pattern at the distal tip of the junctional ring (Figure 2a and video s4). Compared to VE-cadherin and ZO1 the early demarcation of F-actin at proximal and distal junction was less sharp (Figure 2 b, c and videos s5, s6). However, early aggregates of VE-cadherin and ZO1 were always positive for Ruby-UCHD, showing that these three proteins were present from the onset of distal junction formation. At subsequent stages, as the distal becomes more mature, ZO1 and VE-cadherin at the proximal junction were maintained, while the Ruby-UCHD signal started to fade (Figure 2 c-e).

Figure 2.

Formation of the distal junction may occur in different ways. On one hand, it may form by a local junctional disassembly, followed by a reaggregation at the distal end of the JBL. This would essentially consist of a junctional rearrangement process. On the other hand, the distal junction may form de novo by recruitment of VE-cadherin from pools outside of the proximal junction. In order to distinguish between these scenarios, we performed color conversion experiments (Supplementary Figure 1 and video s7). Pools of photoconvertible VE-cadherin (Cdh5-mClav2) were differentially labeled by exposure to UV light within a narrow region of interest at the pole of the junctional ring (Supplementary Figure 1a), which resulted in efficient green-to-red conversion of fluorescence (Supplementary Figure 1 b). When we followed subsequent JBL dynamics over time, we noted that converted (red) VE-cad was absent from the protrusion as well as from the distal junction (Supplementary Figure 1c and video s7. representative of different 5 experiments). Instead, the distal junction contained nonconverted VE-cadherin. This indicates that the Ve-cad pool residing in the proximal junction does not substantially contribute to the JBL nor the distal junction but rather that the distal junction is formed de novo by recruitment of VE-cadherin from pools outside the proximal junction.

Arp2/3 complex dynamics spatiotemporally correlates with JBL formation

Membrane protrusions emerging from endothelial junctions are commonly seen in developing and newly formed blood vessels in the zebrafish embryo. Cell culture experiments, using human umbilical cord endothelial cells (HUVEC), have shown that such lamellipodial structures can control monolayer integrity and vascular permeability (Alves et al., 2018; Martinelli et al., 2013). In particular, so-called junction-associated intermittent lamellipodia (JAIL) are prominent in sub-confluent HUVEC, where they serve to repair junctional gaps (Taha et al., 2014). In contrast to JAIL, JBL are highly polarized relative to the vessel axis and they drive cell rearrangements rather than maintaining vascular integrity (Paatero et al., 2018). However, both lamellipodial structures are similar in their emergence from and association with endothelial cell junctions.

JAIL formation appears to be triggered by a local reduction in junctional tension and requires the activation of the ARP2/3 complex. To examine the role of the Arp2/3 complex in JBL formation, we used a reporter expressing Arpc1b-Venus fusion protein in the zebrafish vasculature. Arpc1b is an integral subunit of the active Arp2/3 complex and has previously been used as a reporter to localize Arp2/3 activity in endothelial cells (Malchow et al., 2024; Padrick et al., 2011). Time-lapse analyses showed a discrete spatiotemporal deployment of Arp2/3 at the leading edge of the forming JBL (Figure 3). Imaging JBL over several oscillation cycles revealed that Arp2/3 accumulation coincided with the initiation and the elongation of JBL, while it disappeared at the end of each JBL cycle (Figure 3a and video s8). Comparison with ZO1 showed that Arp2/3 localized distally with no overlap within the proximal junction (Figure 3a). When we imaged Arp2/3 together with mRuby2-UCHD at higher temporal resolution (30s/frame), we observed that Arp2/3 was maintained at the leading edge of the membrane protrusion (Figure 3b, c and video s9). Taken together, these data show that Arp2/3 specifically marks the distal edge of the JBL during formation and elongation (Figure 3d).

Figure 3.

As the above findings suggest that Arp2/3 is involved in JBL formation, we interfered with Arp2/3 function by pharmacological inhibition using CK666 (Hetrick et al., 2013). We used acute treatments to avoid secondary effects and performed live-imaging on rearranging endothelial cell junctions. Application of 0.2 mM CK666 at 32 hpf completely disrupted junctional Arp2/3 localization within 30 min. (Supplementary Figure 2). Analysis of F-actin organization revealed an increase of filopodia emanating from the junction in the area, which usually was occupied by JBL (Figure 4a-c and videos s10, s11). The abundance of these filopodia prevented the localization of JBL at the junctional poles - an observation consistent with previous cell culture studies (Henson et al., 2015), which suggests that filopodia are formed instead of lamellipodia in the absence of Arp2/3 activity. To verify that filopodia formation was caused by Arp2/3 inhibition, we performed wash-out experiments. Removal of the inhibitor from the medium effectively stopped filopodia formation at endothelial cell junctions, thus reverting the CK666 induced phenotype (supplementary Figure 3 and videos s12, s13). To test consequences of the loss of Arp2/3 activity on cell rearrangements, we measured the speed of junctional elongation in control and CK666 treated embryos. By imaging ZO1 labeled junctions over 1 hour we found that the speed of junction elongation was strongly reduced from an average of 8μm/h to 3μm/h (Figure 3 d,e and videos s14, s15). Taken together, the results show that JBL formation depends on Arp2/3 function and that the absence of Arp2/3 activity leads to a strong impairment of endothelial cell rearrangements during blood vessel formation.

Figure 4.

Proximal and distal junctions resolve by junctional merging and fusion

The appearance of “double junctions” (consisting of a proximal and a distal junction) is a distinctive junctional feature observed in JBL. Their spatiotemporal dynamics suggest that they represent part of the mechanism by which JBL promote junction elongation. Live-imaging revealed a cyclic turnover between the two junctional states: single junction and double junction pattern. Time-lapse imaging of VE-cad-Venus at high temporal resolution showed that this turnover is mediated by the cyclic formation of the distal junction followed by the merging and fusion of the proximal and distal junctions (Figure 5a and video s16).

Figure 5.

Myosin light-chain localization correlates with junctional merging

The cyclic junctional pattern in JBL suggests that it may represent a physical ratchet mechanism, which drives junction elongation. Therefore, we wanted to explore whether junctional mergence could be driven by actomyosin contractility. Increase in tension is accompanied by a recruitment of non-muscular myosin to the actomyosin cytoskeleton (Munjal et al., 2015). By coupling fluorescent reporters the myosin regulatory light chain dynamic changes of actomyosin tension can be visualized in vivo at subcellular resolution (Fernandez-Gonzalez et al., 2009). To monitor the spatiotemporal distribution of actomyosin contractility during JBL cycle, we utilized myosin regulatory light chain (MLC) (Myl9a-GFP and Myl9b-Cherry) as reporters for non-muscular myosin II (NMII) activity (Lancino et al., 2018). Myl9 accumulated along the junctions but was strongly enriched in JBL at both ends of the junctional ring (yellow arrowheads in Figure 5b).

During cell rearrangements JBL are formed at each pole of the junctional ring in a complementary fashion: each JBL is formed by one of the interacting cells corresponding to the forward orientation with respect to cell movement (Paatero et al., 2018)(see schematic in Figure 5c). To test whether Myl9 accumulation corresponds to the forward orientation of the JBL, we analyzed rare mosaic situations in which Myl9a-GFP was expressed in only one of the partnering cells. Indeed, we observed high levels of Myl9a-GFP in JBL formed by the high-expressing cell pointing into the forward direction of the elongating junction (Figure 5c-d).

Vascular tube formation by cell rearrangements has also been described in other vascular beds. We turned to the posterior cerebral vein (PCeV), which is generated by JBL-driven cell rearrangements converting the initial tube architecture from unicellular to multicellular (Lenard et al., 2013). When we compared Myl9a-EGFP distribution to mRuby2-UCHD, we found that the former localized within the JBL adjacent to the forming distal junction of the JBL (see yellow arrowheads in Figure 5e and video s17).

In order to gain a better understanding how actomyosin contractility may drive junctional mergence, we imaged the distribution of Myl9 at high spatiotemporal resolution. We first used VE-cadherin-EGFP to label proximal and distal junctions as spatial landmarks and thus demarcate the area of interest for the quantification of Myl9b-mCherry in space and time (Figure 6a, b, Supplementary figure 4, and videos s18 - s22). We defined three areas - proximal junction, interjunctional and distal junction and quantified Myl9b-mCherry at two time-points (0” and 45”) at the onset and during junctional mergence, respectively (Figure 6b). These quantifications reveal a redistribution Myl9 from the proximal and distal junctions towards the interjunctional space (Figure 6c). This increase indicates a strong recruitment of Myl9 and a concomitant increase of actomyosin contractility within the area between the proximal and distal junctions.

Figure 6.

Actomyosin contractility is required for junction elongation and junctional mergence

In order to assess the role of actomyosin contractility in junction elongation more directly, we used the Rock inhibitor Y-27632 (Uehata et al., 1997)and measured the length of junctional rings over 60 minutes. Rock inhibition strongly inhibited junction elongation compared to DMSO controls (Figure 7a, b and videos s23, s24). To test whether the lack of junction elongation under Rock inhibition may be caused by defects in JBL function, we performed acute short-term Y-27632 treatment and performed live-imaging over 30 minutes to image JBL dynamics during reduction of actomyosin contractility - with a focus on the merging of proximal and distal junctions. To this end, we imaged 20 (Y-27632) and 13 (DMSO) events of junctional merging, respectively, and tracked individual interjunctional distances over time (Figure 7c-e and videos s25, s26). Inhibition of actomyosin contractility led to a strong inhibition of junctional mergence, ranging from a reduction in speed to a complete block in junctional mergence. This blockage in junctional mergence was reflected by the overall increased occurrence of “double junctions” under Rock inhibition (20 (Y-27632) vs 13 events (DMSO-control) in 18 videos). Taken together, these results show that high levels of actomyosin contractility reside within interjunctional space and that these contractile forces are responsible for junctional mergence. Furthermore, these data suggest that JBL formation together with junctional mergence are the driving force of junctional elongation and endothelial cell rearrangements during angiogenic blood vessel formation.

Figure 7.

Discussion

Lumen formation is key in the formation of tubular organs such as blood vessels. Endothelial cells can use different morphogenetic pathways to generate a lumen, including membrane invagination, cord hollowing or lumen ensheathment (Eberlein et al., 2021; Phng and Belting, 2021). The dorsal longitudinal anastomotic vessel (DLAV) of the zebrafish embryo can be formed by either membrane invagination or cell rearrangements, which result in unicellular and multicellular tubes, respectively (Betz et al., 2016; Phng and Belting, 2021). Here, we have analyzed the cell rearrangements, which connect small luminal pockets within the initial endothelial rod, thereby generating a continuous lumen in a fully patent vascular tube.

The coalescence of local, extracellular lumens is commonly observed during the cord hollowing process of different biological tubes, including the notochord of Ciona intestinalis, the vertebrate dorsal aorta as well as the neural tube and intestine (Bagnat et al., 2007; Dong et al., 2009; Gladysheva et al., 2021; Jin et al., 2005; Strilić et al., 2009). In those tubular organs, in which lumen coalescence has been analyzed, it was found that it is mainly caused by lumen expansion, which in turn is driven by directed liquid transport and hydrostatic pressure (Bagnat et al., 2007; Lowery and Sive, 2005; Munson et al., 2008; Navis and Bagnat, 2015). In contrast, the lumen coalescence observed in the zebrafish ISV, DLAV and PCeV is accomplished by active cell intercalation. That is, endothelial cells actively elongate their cell-cell junctions until they form novel cell-cell contacts, which allow the connection of the two adjacent lumens.

We have previously shown that endothelial junction elongation is driven by junction-based lamellipodia (JBL) (Paatero et al., 2018). Careful analysis of junctional and cytoskeletal reporters revealed that JBL dynamics can be subdivided into 4 discrete steps: 1) JBL formation, 2) formation of a new, distal junction, 3) junction conversion and 4) junction stabilization (Paatero et al., 2018). Taken together, these studies showed that JBL provide an oscillating ratchet-like mechanism, in which endothelial cells employ F-actin-based protrusions and VE-cadherin-mediated cell-cell contacts to move over each other. JBL thus provide the physical means for cell rearrangements to drive multicellular tube formation.

In this study we have further explored the junctional and actomyosin dynamics to discern how JBL generate the motive force required for junction elongation and cell rearrangements. In essence, our data show that JBL require ARP2/3 dependent Factin polymerization for JBL formation as well as actomyosin contractility to merge the proximal and distal junctions at the end of the JBL cycle. Thus, by sequential deployment of pushing (F-actin polymerization) and pulling (junction merging) forces, JBL promote forward movement of endothelial cells (summarized in Figure 8).

Figure 8.

Oscillatory junctional protrusions of endothelial cells have also been described in cultured HUVECs (Cao et al., 2017; Taha et al., 2014). Here, so-called junction-associated intermittent lamellipodia (JAIL) form in similar intervals as JBL. However, differences in junctional dynamics suggest that JAIL and JBL represent related but distinct cellular activities. JAIL formation is thought to be triggerered by local dissolution of endothelial adherens junctions (i.e. by downregulation of VE-cadherin). Subsequently, these protrusions retract EC junctions are re-established (Cao et al., 2017; Taha et al., 2014). In this experimental paradigm, JAIL do not promote endothelial cell movement. JBL, in contrast, provide endothelial motility during blood vessel morphogenesis. Furthermore, formation of JBL is not preceded by a dissolution of EC junctions and a distal junction is formed at the leading edge of the protrusion leading to a “double junction” configuration as an intermediate state. Our studies show that these “double junctions” are an integral part of the cellular mechanism of junction elongation. We have previously shown that JBL are most prominent during multicellular tube formation and become less prevalent and smaller thereafter (Paatero et al., 2018). These salient and distinct features prompted us to adopt the term junction-based-lamellipodia (JBL), in order to differentiate them from JAIL. Although JAIL have also been implicated in endothelial cell migration (Cao and Schnittler, 2019; Cao et al., 2017; Seebach et al., 2021) neither junctional patterns nor junctional dynamics have been analyzed in this context. Thus, JAIL and JBL represent similar but different lamellipodia-like protrusions. JAIL are associated with the maintenance of endothelial integrity, as well as the control permeability and trans-endothelial cell migration, as has been suggested by several publications (Cao et al., 2017; Kipcke et al., 2025; Seebach et al., 2021; Taha and Schnittler, 2014). In contrast, JBL drive cell rearrangements, by step-wise elongation of cell junctions resulting in convergent cell movements.

Lamellipodial structures resembling JBL, displaying double junctions, have recently been described in the oak leaf-shaped endothelial cells of dermal lymphatics in the mouse (Schoofs et al., 2025). Here, the double junctions are thought to regulate endothelial shape and in particular to stabilize the interendothelial cell overlap, which is important to control the interstitial fluid transport.

Our photoconversion experiments (see Supplementary Figure 1) show that the distal junction is formed de novo from VE-cadherin pools originating from outside the existing (proximal) junction. During distal junction formation, we observe that VE-cadherin forms clusters resembling spot-junctions observed in the context of organ morphogenesis in other systems. Consistent with our observation such spot-junctions have been shown to be important in the formation of contractile F-actin fibers in these systems (Lecuit et al., 2011). While we are currently not able to monitor actin flow in our experimental in vivo system, our observation that the formation of the distal junction coincides with the formation of a distal F-actin belt is consistent with a clutch mechanism, in which transligation of VE-cadherin triggers reorganization of the cortical actin cytoskeleton similar to what has recently been described in cell culture studies (Noordstra et al., 2023). Furthermore, our previous findings from zebrafish cdh5 null mutants or embryos expressing a VE-cadherin protein (VE-cadΔCT) lacking the β-catenin binding site show that VE-cadherin mediated cell-cell interactions as well as VE-cad/F-actin interaction are important for JBL-driven junction elongation (Paatero et al., 2018; Sauteur et al., 2014).

The cell motility-promoting activity of VE-cadherin during tube formation may be quite unique, considering that in most studies cadherins have been described to stabilize epithelia, limit cell motility and to regulate contact inhibition of locomotion (Campàs et al., 2024; Maître and Heisenberg, 2013; Noordstra et al., 2023) Cadherin-based cell interactions have been shown to be essential for dynamic cell movements in several morphogenetic processes (Friedl and Mayor, 2017; Mayor and Etienne-Manneville, 2016). In the case of collective border cell migration during Drosophila oogenesis and migration of primordial germ cells in zebrafish, cadherins have been shown to provide traction for collective and single cell migration, respectively (Blaser et al., 2006; Cai et al., 2014).

A salient feature of JBL is their appearance at both ends of each rearranging cell. This bipolar activity causes the cells to elongate in both directions and rearrange rather than to migrate. Furthermore, and in contrast to lamellipodia-driven mechanisms of cell migration, the domain of actomyosin contractility is confined within the interjunctional area (between distal and proximal junctions, see Figures 6 and 8). This suggests that the restricted area of contractility is not only demarcated by the proximal and distal junctions, but that these junctions act as insulators to segregate compartments of different actomyosin tension along the cellular interface. We propose that this insulator function is essential to permit junction elongation.

In this study, we have investigated the cellular and molecular basis of force generation in JBL. Through detailed analysis of junctional, cytoskeletal, and regulatory proteins, we uncover a biphasic mechanism involving alternating protrusive and contractile forces. Notably, and in contrast to classical paradigms of cell migration, we show that actomyosin contractility is spatially restricted to a narrow zone at the junctional pole during endothelial cell movement. Together, our findings reveal a novel mechanism of cell rearrangement during vascular cord hollowing. JBL serve as localized modules that generate both protrusive and contractile forces during junctional remodeling and cell elongation. Moreover, JBL-localized junctions may act as insulators, spatially confining actomyosin tension to the junctional pole.

Material and Methods

Lead contact

Further information and request for resources and reagents should be directed and will be fulfilled by the Lead Contact; Heinz-Georg Belting (heinz-georg.belting@unibas.ch).

Materials availability

Materials and zebrafish lines generated and used in this study are available upon request to the lead contact.

Fish maintenance and stocks

Zebrafish (Danio rerio) were maintained according to FELASA guidelines (Aleström et al., 2020). All experiments were performed following institutional and ethical welfare guidelines and animal protocols in accordance with federal guidelines approved by the Kantonales Veterinäramt of Kanton Basel-Stadt (1027H, 1014HE2, 1014 G). Breeding and embryo collection were done according to standard protocols (Westerfield, 2007). Pain, distress, and discomfort were minimized as much as possible.

The following zebrafish lines were used in this study:

Sex was not considered in this study because zebrafish embryos used for the experiments were at developmental stages prior to sex differentiation. Therefore, the sex of the embryos was not determined or relevant to the outcomes.

Live imaging of zebrafish embryos

In DLAV experiments, zebrafish embryos between 30 to 32 hpf were dechorionated and anaesthetized with 0.16mg/ml (1×) tricaine methanesulfonate (Sigma). The embryos were selected for presence of fluorescence and mounted into microwell dishes within 0.55% low-melting-point agarose (ROTI) and covered with E3 buffer containing 1× tricaine. In PceV experiments, transgenic embryos selected for presence of fluorescence were anaesthetized in 1 × tricaine (0.08%) and mounted in a 35 mm glass-bottom Petri dish (0.17 mm, MatTek), using 0.7% low melting agarose (Sigma) containing 0.08% tricaine and 0.003% 1-phenyl-2-thiourea (PTU; Sigma-Aldrich). The embryos were mounted into microwell dishes within 0.55% low-melting-point agarose (ROTI) and covered with E3 buffer containing 1 × tricaine.

A Leica SP5 confocal microscope with a 40x water immersion objective was used for live-imaging experiments. High spatiotemporal resolution time-lapse images were acquired with an Olympus SpinSR spinning disc microscope using a 60X silicon oil (NA = 1.3) objective, with a z-stack step size of 0.6 μm at different frequencies depending on the experiment.

Photoconversion experiments

Photoconversion experiments were performed using a Leica SP5 confocal microscope with a 40x water immersion objective (NA = 1.1). Photoconversions were applied to manually selected regions of interest (ROIs) with a 405 nm laser for 10–20 seconds to induce conversion, monitored by simultaneous imaging at 488 nm to ensure complete conversion. All the imaging was performed at 28.5 °C.

Junction elongation assay

Junctional elongation was assessed by measuring elongation speed of isolated junctional rings during DLAV formation between 30 and 32 hpf. Inhibitor treatments CK666 (200 μM), Y-27632 (75 μM), or DMSO (1%) were applied 1h before mounting of embryos into 0.55% low-melting point agarose. The same concentrations of chemicals were applied to the low-melting-point agarose mounting medium and the E3 medium on top of it before imaging and imaging the junctions for 60–90 min on a Leica SP5 (Y-27632) Olympus SpinSR spinning disc microscope or Olympus SpinSR spinning disc (CK666) microscope.

Junctional merging tracking

Speed of junctional merge was evaluated by monitoring isolated junctional rings during DLAV formation. Inhibitor treatment Y-27632 (75 μM) or DMSO (1%) were applied 30 min before mounting. The same concentrations of chemicals were applied to the low-melting-point agarose mounting medium and the E3 medium on top of it before imaging and imaging the junctions for 10–15 min on Olympus SpinSR spinning disc microscope. Distances were measured using Fiji software. In each frame, the interjunctional distance was defined as the maximum distance between the proximal and distal junctions. A line was manually drawn between the proximal and distal junctions in Fiji, and its length was recorded. The same proximal and distal junction landmarks were used consistently across all time points.

Immunostaining

Zebrafish embryos of 30 to 32 hpf were fixed after dechorionation using Glyofixx (Thermo-Fisher Scientific) and incubated at 4°C overnight (ON) in 1.5 ml Eppendorf tubes (max. 20 embryos per tube). After fixation, embryos were washed four times for 5 minutes in 1x PBST at room temperature (RT) on a rocking platform. Embryos were then permeabilized using 0.5–1 % Triton-X100 in PBST for 30 min at RT. They were then blocked with 2% BSA and 5% goat serum in PBST overnight with continuous shaking at 4 °C. After blocking, the primary antibody solution was added. The primary antibodies were either diluted in blocking solution or Pierce Immunostain Enhancer Solution (Thermo Fisher Scientific). Primary antibody incubation was performed at 4 °C for 3 days. After incubation, the solution was removed, and the embryos were washed at least four times for 30 minutes each with PBST at 4 °C, or overnight. Subsequently, the embryos were incubated with secondary antibodies, diluted 1:2000 in blocking solution, and incubated at 4 °C.

Image preparation and analysis

Image analysis and measurements were performed using FIJI RRID:SCR_002285. Z-stacks were flattened by maximum slice intensity projections. Where needed, noise was reduced using Gaussian filtering (radius 0.7) and/or background subtracted (rolling ball radius 50) using ImageJ RRID:SCR_003070. Contrast and brightness of images were linearly adjusted. Particle Image Velocimetry (PIV) was employed to measure velocity fields in fluid flow as previously described (Yin et al., 2024). Image panels were created using the Open Microscopy Environment (OMERO). ARP2/3- Venus and mRuby2-UCHD localization were evaluated on a 4pixel-wide rectangle, and the average over length, corrected for background (evaluated as average of average intensity in 5 rectangles on the cell cytoplasm) and displayed in a graph using Graphpad Prism 10.4.1 RRID:SCR_002798. Proximal, distal, and interjunctional space domains were defined using VE-cad distribution, My9-mcherry distribution was evaluated as average intensity in the distinct domains The intensities are corrected for background (which has been evaluated as average of average intensities in 5 rectangles in the cell cytoplasm) and displayed in a graph using Graphpad Prism 10.4.1.

Statistical analysis

GraphPad Prism 10.4.1 was used to perform statistical analyses. For every experiment and analysis, sample size is specified in the figure legend. Sample sizes were not pre-determined using statistical methods, and all data meeting the quality control criteria were included in the analyses. Randomization was not applied to the experiments. Unless specified otherwise, default settings were used for all software tools in the analyses. Whenever a p-value is presented in the text or figures, the corresponding statistical test is indicated.

Supplementary Figures

Supplementary Figure 1.

Supplementary Figure 2.

Supplementary Figure 3.

Supplementary Figure 4.

Data availability

The data that support the findings of this study are available in the supporting information of this article. https://www.researchsquare.com/article/rs-7065344/v2

Acknowledgements

We thank Kumuthini Kulendra and Andre Rodriguez for fish care and the Imaging Core Facility of the Biozentrum (Universität Basel) for microscopy support. We thank Anne Schmid (Institute Pasteur, Paris, France) for providing Tg(kdrl: Myl9a-GFP)^p5Tg and Michel Bagnat (Duke University, USA) for providing TgKI(tjp1a-tdTomato)^pd1224. We thank Sven Andreas Belting for preparing illustrations and graphic animations. This work has been supported by the Kantons Basel-Stadt and Basel-Land and by grants from the Swiss National Science Foundation (310030_200701 and 310030B_176400) to M.A..

Additional information

Author contributions

M.A. and H.-G.B. conceptualized the project. L.M., J.Y., C.W., I.P. and H.-G.B. designed the experiments. L.M., J.Y., C.W. and I.P. performed experiments and analyzed the data. J.M. and C.H. generated Tg(flia:Arpc1b-mVenus-p2A-Arpc3- mTurq2)^mr25. L.M. and H.-G.B. wrote the manuscript. M.A., L.M. and H.-G.B edited the manuscript. All authors reviewed the manuscript.

Funding

Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung (SNF) (310030_200701)

• Markus Affolter

Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung (SNF) (310030B_176400)

• Markus Affolter

Additional files

Figure 1 - video s1. JBL drive endothelial cell movements. Time-lapse video of EGFP-ZO1 and mRuby2-UCHD at around 30hpf. Actin-rich protrusions (mRuby2-UCHD) form at the front of the elongating junction (EGFP-ZO1). Yellow asterisks point to JBL. Yellow double-headed arrows indicate the distance between the junction and the dorsal end of the DLAV. Scale bar: 10μm

Figure 1 - video s2. JBL drive local lumen fusion. Time-lapse video showing EGFP-Podxl1 (labeling apical cell membrane) and mRuby2-UCHD (labelling F-actin) during DLAV formation, starting at approximately 32 hpf. The leading edges of the converging junctions are indicated by white arrowhead. Yellow arrowheads point to JBL. Scale bar: 10 μm.

Figure 1 - video s3. JBL form adjacent to the apical cell compartment. Time-lapse video showing the oscillatory behavior of JBL (F-actin protrusions labeled with mRuby2-UCHD (green), yellow arrowhead) during DLAV formation (30hpf). The apical compartment is labeled by EGFP-Podxl1 (magenta). The white dashed line surrounds the UCHD-labeled JBL domain. Scale bar: 5μm.

Figure 2 - video s4. Formation of a new junction at the distal end of the JBL. Time-lapse of VE-cad-Venus imaged at rate of 1stack/12s during distal junction formation. New distal junctions emerge in clusters at the distal end of the JBL. Ve-cad is diffusely localized in early JBL, while it accumulates in big foci at later time points. Scale bar: 2 μm.

Figure 2 - video s5. Formation of a new junction at the distal end of the JBL. Time-lapse video showing F-actin (mRuby2-UCHD, magenta) and VE-cad-Venus (green) during JBL protrusion (0s) and distal junction formation (30s). The dashed line encircles the protrusion. White arrow heads point distal junction foci at the distal tip of the protrusion. Scale bar: 2 μm.

Figure 2 - video s6. Formation of a new junction at the distal end of the JBL. Time-lapse video showing dynamic EGFP-ZO1 (green) and mRuby2-UCHD (magenta) distribution during distal junction formation in the forming DLAV (approx. 32 hpf). After distal junction formation, F-actin gradually diminishes from proximal junction and interjunctional space, while maintaining strong localization at the distal junction. Scale bar: 2 μM. Blue and white arrowheads highlight proximal (PJ) and distal junction (DJ), respectively.

suppl. Figure 1 - video s7. The distal junction forms de novo. Time-lapse video s7 showing a DLAV junctional ring of an embryo expressing VE-cad-mClav2 before (0 min) and right after photoconversion (1, 2, 3 min). The red square delimitates the photoconverted area. scale bar: 5μm.

Figure 3 - video s8. Arpc1b-Venus oscillations occur distal to the junction. Time-lapse video of Arpc1b-Venus (green) and ZO1-tdTomato (magenta) at about 30hpf, showing deposition of Arpc1b-Venus at the distal side of the junctional ring over two JBL cycles. Scale bar: 2 μm.

Figure 3 - video s9. Arpc1b localizes at the distal end of JBL. Time-lapse of Arpc1b-Venus (green) and mRuby2-UCHD (magenta) during the protrusion phase of a JBL. scale bar: 2 μm.

Figure 4 - video s10. Arp2/3 activity is required for JBL formation. Time-lapse video of JBL labelled with EGFP-UCHD, at around 32hpf, in the presence of DMSO (1%) (control). scale bar: 5 μm.

Figure 4 - video s11. Arp2/3 activity is required for JBL formation. Time-lapse video of JBL labelled with EGFP-UCHD, at around 32hpf, after 1h incubation with CK666 (200μM). Note the ectopic filopodia and stress fibers. scale bar: 5 μm.

suppl. Figure 3 - video s12. CK666 mediated JBL inhibition is reversible upon washout. Time-lapse video showing F-actin dynamics at the junctional ring, labeled with EGFP-UCHD during incubation with CK666 (200μM). Recording starts 30 minutes into the incubation. Blue arrowheads point to ectopic filopodia. scale bar 2μm.

suppl. Figure 3 - video s13. CK666 mediated JBL inhibition is reversible upon washout. Time-lapse video showing F-actin dynamics in the same junctional ring as in video s12 labeled with EGFP-UCHD, 90 min after start of washout. Red arrowheads point to reemerging JBL. scale bar 2μm.

Figure 4 - video s14. Arp2/3 activity is required for junction elongation. Time-lapse of a ZO1-tdTomato labeled junctional ring in the presence of DMSO (1%) (control) recorded over 60 minutes. scale bar: 10μm.

Figure 4 - video s15. Arp2/3 activity is required for junction elongation. Time-lapse of a ZO1-tdTomato labeled junctional ring after 1 hour incubation and in the presence of CK666 (200μM). scale bar: 10μm.

Figure 5 - video s16. Proximal and distal junctions merge at the end of JBL cycle. Time-lapse video showing of VE-cad-Venus during junctional merging over 90 seconds. Blue and red arrowheads point to the proximal and distal junction, respectively. The two junctions are gradually moving closer until they merge. scale bar: 2μm.

Figure 5 - video s17. Myl9 is enriched within the JBL. Time-lapse video showing Myl9a-GFP and mRuby2-UCHD localization during JBL formation during anastomosis of the PCeV of the zebrafish hindbrain at around 60 hpf. White dashed lines delineate the JBL region. Yellow and white arrowheads point to the proximal and distal junction, respectively. Myl9-GFP is enriched inside the lamellipodia and localizes at the distal junction at later time points. scale bar: 2 μm.

Figure 6 - video s18. Dynamic redistribution of Myl9 during JBL formation and junctional merging. Time-lapse video s18 of VE-cad-EGFP (Cdh5ΔC-EGFP) and Myl9b-mCherry (shown as “fire”-LUT) during junctional merging at four timepoints (0, 45, 90 and 135 seconds). Blue dashed lines confine the applied mask. Yellow and white arrowheads point to the proximal and distal junction, respectively. scale bars 2 μm.

suppl. Figure 4 - video s19. Myosin light-chain recruitment at the inter-junctional space during merging. Time-lapse of VEcad-EGFP (Cdh5ΔC-EGFP) and Myl9b-mCherry (shown as “fire”- LUT) during “double junction” state (video s19). scale bar: 2μm.

suppl. Figure 4 - video s20. Time-lapse video s20 of VE-cad-Venus and Myl9b-mCherry (shown as “fire”-LUT) showing increase of MLC in the inter-junctional space prior to proximodistal junction mergence. scale bar: 2μm.

suppl. Figure 4 - video s21. Time-laps video s21 showing VE-cad-EGFP (Cdh5ΔC-EGFP) and Myl9b-mCherry (shown as “fire”-LUT) showing recruitment of MLC between proximal and distal junction during ‘‘double junction’’ state. scale bar: 2μm.

suppl. Figure 4 - video s22. Time-laps video s22 showing VE-cad-EGFP (Cdh5ΔC-EGFP) and Myl9b-mCherry (shown as “fire”-LUT) showing recruitment of MLC between proximal and distal junction during “double junction” state. scale bar: 2μm.

Figure 7 - video s23. Junction elongation during DLAV formation requires actomyosin contractility (control). Time-lapse video showing elongation of endothelial cell-cell junctions (EGFP-UCHD) during DLAV formation at around 32 hpf (1% DMSO - control). scale bar: 10μm.

Figure 7 - video s24. Junction elongation during DLAV formation requires actomyosin contractility (Y-27632). Time-lapse video showing elongation of endothelial cell-cell junctions (EGFP-UCHD) during DLAV formation at around 32 hpf after 60-minute incubation with Y-27632 (45μM). scale bar: 10μm.

Figure 7 - video s25. Merging of proximal and distal junctions requires actomyosin contractility (control). Time-lapse video showing VE-cad-Venus labelled proximal and distal junction, during junctional merging (1% DMSO control) or Y-27632 (75 μM). Blue and yellow arrowheads are pointing distal and proximal junctions respectively. scale bar: 2μm.

Figure 7 - video s26. Merging of proximal and distal junctions requires actomyosin contractility (Y-27632). Time-lapse video showing VE-cad-Venus labelled proximal and distal junction, during junctional merging under ROCK inhibition (75μM Y-27632). Blue and yellow arrowheads are pointing distal and proximal junctions respectively. scale bar: 2μm.

Figure 8 - video s27. Animation depicting the molecular mechanisms of junction elongation by junction-based lamellipodia (JBL). JBL formation is initiated by Arp2/3 activation. The JBL is pushed forward by F-actin polymerization. At the distal end a new cell-cell junction is formed. MyosinII is recruited to the interjunctional space. Actomyosin contraction merges the proximal and distal junctions resulting in an overall elongation of the junctional ring.

Note

This reviewed preprint has been updated to correct a corresponding author's email address, and include an author as corresponding author.