Planar polarized force propagation integrates cell behavior with tissue shaping during convergent extension

Summary

Convergent extension (CE) is an evolutionarily conserved developmental process that elongates tissues and organs via collective cell movements known as cell intercalation. Here, we sought to understand the mechanisms connecting cell behaviors and tissue shaping. We focus on an often-overlooked aspect of cell intercalation, the resolution of 4-cell vertices. Our data reveal that imbalanced cellular forces are involved in a timely vertex resolution, which, in turn, enables the propagation of such cellular forces, facilitating the propagation of tissue-scale CE. Conversely, delayed vertex resolution leads to a subtle but significant change in tissue-wide cell packing and exerts a profound impact by blocking force propagation, resulting in CE propagation defects. Our findings propose a collaborative nature of local cell intercalations in propagating tissue-wide CE. It unveils a multiscale biomechanical synergy underpinning the cellular mechanisms that orchestrate tissue morphogenesis during CE.

The Blurb

Convergent extension is a conserved process that elongates tissues via collective cell intercalation. Weng et al. reveal that force-dependent vertex resolution affects cell packing and enables polarized force propagation to facilitate the propagation of convergent extension, highlighting a multiscale mechanical synergy in tissue shaping.

Introduction

Convergent extension (CE) is an evolutionarily conserved morphogenetic process crucial for the elongation of the body axis in most animals and for the development of multiple organ systems during embryogenesis (Figure 1A–C).^1–3 CE involves intricate cellular behaviors, including planar cell polarity establishment, polarized actomyosin activity, and directed cell intercalations (as shown in Figure 1C). These cell behaviors act collectively to narrow a tissue in one direction while elongating it in another direction.^2,4,5

Figure 1. 4-cell vertex resolves by t-junction extension and rotation.

(A-C) Schematic showing tissue-scale convergent extension in embryos (A) and explant culture (B), the latter of which enables observation of tissue-scale shape change (B) and cell intercalations (C).

(D) Schematic showing a 2-step process of cell intercalation: mediolateral cells shorten v-junctions (light gray) while forming a 4-cell vertex (middle panel), followed by nascent t-junction (dark gray) formation as cells continue to move (lower panel).

(E) Schematic showing a 2-step resolution of 4-cell vertices: nascent t-junctions extend anteroposteriorly first (middle panel) and rotate with continuous extension afterward (lower panel).

(F) Still images from representative movies showing the entire process of cell intercalation.

A major outstanding question relates to the mechanisms by which individual cell behaviors are integrated to achieve shape change at the level of whole tissues. An attractive candidate is the patterned spatial arrangement of cells, known as cell packing. At larger scales, packing serves as a remarkable predictor of collective cell motility^6–8 and tissue material properties.⁹ At the same time, packing is known to be influenced by cellular processes involving neighbor exchanges.^8,10

Neighbor exchange by cell intercalation, a so-called T1 transition (Figure 1D), is a key driver of CE, and it involves two pivotal steps: the formation of a 4-cell vertex, where so-called v-junctions connecting anteroposterior (AP) cell neighbors shorten to bring the mediolateral (ML) cell neighbors into contact (Figure 1D, middle panel), and the resolution of the vertex, in which a nascent t-junction forms and grows between the neighbors (Figure 1D, lower panel). While most studies of CE have focused on the essential first step of cell intercalation, it is the far less-explored second step, resolution, that contributes to the majority of total tissue elongation, as evidenced by studies on Drosophila germ band extension.^11,12 Nevertheless, our understanding of vertex resolution and its role in tissue shaping, especially in vertebrate models such as Xenopus, remains significantly limited.

Here, we combined embryology, cell biology, and non-invasive assessment of mechanical forces across different developmental scales to link the resolution of cell intercalation, cell packing configurations, and tissue scale CE in Xenopus notochord. Our findings illuminate the role of timely t-junction rotation, an overlooked process at the end of cell intercalation, in maintaining normal cell packing that facilitates polarized propagation of cellular forces and, thus, effective tissue-level CE. Our data provide multiscale mechanical insights into the propagation of tissue shaping through intricate and collaborative cell behaviors.

Results

Nascent t-junctions resolve 4-cell vertices by extension and rotation

We first characterized the resolution of 4-cell vertices during CE in the Xenopus notochord by live imaging of a membrane marker (Figure 1E, F). Curiously, in contrast to the single-step process depicted in Figure 1D, our data unveiled an integration of two steps in the resolution of a 4-cell vertex. Initially, nascent t-junctions extended perpendicularly to the connected v-junctions, as shown in Figure 1E, middle panel, and Figure 1F, 3^rd panel. Subsequently, they underwent a rotational movement while continuing to elongate (Figure 1E, F, lower panels).

The transition from extension to rotation occurred at different rates, with some instances showing a smooth transition, and others being acute, exhibiting a rotation angle exceeding 15° in 10 sec. To quantify this behavior, we defined the beginning of the t-junction rotation as the first frame when the rotation angle exceeds 15° away from its perpendicular position. This transition generally occurred 5 min after t-junction formation began. Up to this point, the overall extension rate was 1 μm/min, and the rotation rate was 5°/min. The rotation of a nascent t-junction seems not to have been described previously and prompted us to ask what cellular mechanisms drive this process.

T-junction rotation involves differential high tension on the connected v-junctions.

Previous work by our lab and others has shown that mediolaterally aligned v-junctions bear greater tension than the relatively anteroposterior aligned t-junctions.^13,14 This increased tension on v-junctions drives v-junction shortening in the first half of cell intercalation, so we reasoned that the rotation of nascent t-junctions might be driven by tension from their adjoining v-junctions. This idea was supported by laser ablation experiments. When a v-junction was ablated, the adjoining t-junction promptly rotated (Figure S1).

We then reasoned that normal t-junction rotation might occur as a result of a four-way tug-of-war. In this scenario, a new t-junction rotates counterclockwise (CCW) when tensions on the upper left (UL) and lower right (LR) adjoining v-junctions are higher than tensions on the upper right (UR) and lower left (LL) ones (Figure 2A; example in Figure 2E,F). The converse explains clockwise (CW) t-junction rotations (Figure 2B).

Figure 2. Differential high tension on v-junctions rotates the conjoining t-junction.

(A,B) Schematic depicting that 4-way tug-of-war between the two diagonal pairs of v-junctions controlling the rotation of the connected t-junction. A, if the tension on the upper left (UL) and lower right (LR) v-junctions are higher than the tension on the other diagonal pair, the connected t-junction will rotate in the counterclockwise direction. B, if the tension on the upper right (UR) and lower left (LL) v-junctions are higher, the t-junction will rotate clockwise.

(C,D) Schematic correlating the transverse fluctuation of a junction with the junctional tension on it. High tension correlates with low transverse fluctuation (C), and vice versa (D).

(E,F) Representative images of t-junction dynamics and transverse fluctuation analysis on the connected v-junctions. The transverse fluctuation on the adjacent 10-μm region of each connected v-junction during the first 5-min of the t-junction formation was measured (E). The diagonal junctions were paired for further analysis (F).

(G) The distribution of transverse fluctuation of the two diagonal pairs of the v-junctions in F, showing that the UR and LL v-junctions had significantly higher transverse movement. p-value was calculated using the Kolmogorov–Smirnov test. ***p < 0.0005

(H) The difference in tension between the two diagonal pairs was characterized by Δfluctuation, which was defined as the maximum difference between the two cumulative sums of transverse fluctuations.

(I) Plot showing t-junction rotation rate versus the tension difference on diagonal v-junctions represented by Δfluctuation. A positive value of Δfluctuation indicates higher tension on the upper right and lower left v-junctions, and vice versa. Cross markers represent data points, the solid line is the linear fitting, and the dashed lines are upper and lower boundaries with 95% confidence. N = 16.

(J,K) Plot showing the correlation between Δfluctuation and t-junction rotation depends on the force difference. The same data in (I) was separated into two subgroups based on their Δfluctuation and then replotted. N = 7 for greater Δfluctuation and N = 9 for less Δfluctuation.

See also Figure S1 and S2.

Laser cutting proved too destructive for longer-term observation of rotation, so we turned to our image-based, non-invasive method whereby transverse fluctuation of a cell-cell interface from time-lapse data serves as a proxy for junctional tension; junctions bearing high tension display smaller transverse fluctuations, while those with lower tension display more transverse movement (Figure 2C,D)(see Method for details).¹⁵

We measured transverse fluctuations in all four adjoined v-junctions within 10 μm of growing t-junctions (Figure 2E), paired the UL-LR and UR-LL diagonal v-junctions (Figure 2F,G), and assessed the difference in tension between the two pairs using the term Δfluctuation (Figure 2H). This term reflects the difference in fluctuation when the cumulative distribution of the two pairs reaches their maximum difference (Figure 2H); a positive value indicates higher tension on the UR and LL v-junctions, and a negative value, the reverse. We then plotted Δfluctuation against the overall t-junction rotation rate and found that 14 out of 16 data points were in the first and third quadrants (Figure 2I; Figure S2A,B), consistent with our hypothesis illustrated in Figure 2A,B.

Interestingly, we noticed that the two data points noted above that did not follow the expected pattern, both occurred when the diagonal difference in tensions was relatively small (Figure 2I). We, therefore, divided our data into two subgroups of higher and lower Δfluctuation. We found that the group with greater Δfluctuation had a much stronger correlation with t-junction rotation rate (Figure 2J), suggesting a more determined t-junction rotation being driven by high tensional difference. On the other hand, the subgroup with less Δfluctuation showed close to no correlation with t-junction rotation (Figure 2K), indicating a threshold that needs to be surpassed to rotate the conjoining t-junction. Combined with results from laser cutting data (Figure S1), these data suggest that the imbalance of tensions on the adjoining v-junctions determines the rotation direction of the conjoined t-junction in a threshold-dependent manner.

Lagging vertex resolution changes tissue-wide cell packing configuration.

Our data suggest that a threshold of cell cortex tension is required for normal t-junction rotation, and we tested this idea by examining cells lacking the catenin Arvcf, which is essential for normal CE and is required for the overall cell cortex tension in the Xenopus notochord.^15,16 Live imaging of Arvcf-deficient cells during 4-cell vertex resolution revealed that nascent t-junctions formed (Figure 3A) but displayed significantly altered dynamics (Figure 3B–E). Loss of Arvcf caused a significant delay in the transition from extension to rotation, leading to a 75% prolonged initial extension phase compared with wildtype (WT)(Figure 3B), and a 52% reduction of the overall rotation rate (Figure 3C). With the extension rate unaffected (Figure 3D), t-junctions were notably longer at the beginning of the extension-to-rotation transition in the Arvcf-depleted cells (Figure 3E).

Figure 3. Cell packing configuration is finely tuned by the t-junction rotation rate.

(A) Still images from representative movies illustrating t-junction formation in Arvcf deficient cells. Left panel, a t-junction starts to form when the ML cells meet. Middle panel, t-junction extends along the AP axis. Right panel, t-junction rotates at the end of the initial extension phase.

(B-E) Quantification of nascent t-junction dynamics in WT and Arvcf KD explants, including (B) the onset of junction rotation defined as the first frame when the junction is 15° away from its perpendicular position, (C) the overall rotation rate from the initial formation to the onset of junction rotation, (D) extension rate during the initial extension phase, and (E) junction length at the onset of junction rotation. N = 24 for t-junctions from WT explants and N = 23 from Arvcf KD explants. Explants were collected from at least three replicates. p values were calculated using Wilcoxon rank sum test (A.K.A. Mann-Whitney U test). ***p < 0.0005; NS, not significant.

(F) Plot showing the correlation between t-junction rotation rate and Δfluctuation was impaired with loss of Arvcf. Cross markers represent data points, the solid line is the linear fitting, and the dashed lines are upper and lower boundaries with 95% confidence. N = 16.

(G-J) Representative images of cell packing configurations in WT (G) and Arvcf KD (J) explants with membrane labeling. H and I are representative images showing the angle between t- and connected v-junctions (“tricellular angles”; solid lines). Arrowheads mark the so-called AP-aligned perpendicular t-junctions perpendicular to the surrounding v-junctions.

(K) Distribution of the tricellular angle, defined as the smaller angle between t- and connected v-junctions for each cell (examples in H and I). Data points were pooled from 11 explants for each condition from at least three replicates and approximately 50 cells from each explant. p-value was calculated using the Kolmogorov–Smirnov test. ***p < 0.0005.

See also Figure S2.

We then examined whether the reduced junctional tension is relevant to the lagging vertex resolution. We conducted the same analysis as shown in Figure 2E–I to assess the relationship between the tensional difference in the two diagonal pairs of the v-junctions and the rotation rate of the conjoining t-junction after Arvcf loss. As anticipated, the overall reduction of tension applied to the tricellular regions of v-junctions (Figure S2C), led to a decrease in tensional difference between the diagonal pairs. Interestingly, the absence of Arvcf completely eliminated the correlation between the tensional difference and t-junction rotation (Figure 3F), similar to the WT cells when the tensional difference is small (Figure 2K). These data suggest an Arvcf-dependent force threshold involved in rotating the t-junctions.

Interestingly, longer extension and slower rotation of t-junctions translated into noticeable differences in cell packing configurations between WT and Arvcf KD explants. At the population level, t-junctions in WT explants typically displayed a tilted orientation relative to their adjoining junctions (Figure 3G,H), with very few being perpendicular (i.e., strongly aligned in the AP axis at the cellular scale; Figure 3I, arrowheads). To quantify this configuration, we measured the smaller angle between each t-junction and its connected junctions (referred to as “tricellular angles”, Figure 3H,I, solid lines). The distribution of tricellular angles peaked at around 40° in WT explants (Figure 3K, blue), consistent with the tilted pattern described above.

Conversely, Arvcf depletion led to an increased prevalence of perpendicular t-junctions throughout the population (Figure 3J, arrowheads). Measurement of tricellular angles confirmed this observation: with a dominant peak around 90° (Figure 3K, orange). Coupled with the observations in Figure 3A–F, our data suggested that Arvcf deficiency prolonged the resolution of 4-cell vertices by delaying the extension-to-rotation transition in new t-junctions, and this, in turn, led ultimately to an overall shift in cell packing configuration characterized by an excess of perpendicular t-junctions within the tissue.

Cell packing configuration facilitates planar polarized force propagation in silico.

Embryos lacking Arvcf display severe defects in axis elongation despite relatively subtle defects in cell intercalation,^15–17 leading us to ask if the distinct cell packing configurations we observed might affect CE. We considered mechanical principles akin, for example, to how the pattern of beams and posts forming a truss bridge influences the distribution of loading. We first explored this idea by creating computational toy models to analyze relative force distributions in a tissue using a force balance-based inference technique (Figure 4A).¹⁸ In an initial experiment, we tested static models that represented extreme cell packing configurations, deliberately exaggerating features observed in WT and Arvcf KD explants (Figure 4B–E). The first model, termed Hexagon, employed cells with hexagonal shapes and tilted t-junctions (Figure 4B). The second model termed Brick-wall, featured rectangular cells with all t-junctions perpendicular to the neighboring junctions (Figure 4D).

Figure 4. Resolution of 4-cell vertices is required for polarized force propagation.

(A) Schematic explaining the cellular force interference technique, illustrating force balance at each tricellular vertex.

(B-G) Force propagation simulation in toy models with 3 representative cell packing configurations - Hexagon model (B,C), Brick-wall model (D,E), and Hybrid model (F,G). Left panels (B,D,F), cell packing configuration. Right panel (C,E,G), force distribution across the tissue model upon force stimulation in the center, indicating relative tension on each junction with color code. Dark red is the highest, and blue is the lowest. Insets show the overall force propagation pattern.

(H, I) Force distribution estimated using the cellular force inference in WT (H) and Arvcf KD (I) explants. Dark gray shading highlights regions with v-junctions having the maximum force (dark red), along with their adjacent v-junctions holding forces greater than the average (warm color). Arrowheads mark the t-junctions perpendicular to the apposed v-junctions.

(J, K) Quantification of force coupling in the ML (J) and AP (K) directions in WT and Arvcf KD explants. The number of adjacent cells having forces on v-junctions above average was quantified as the number of cells with force coupled. Data was pooled from 20 WT explants and 17 Arvcf KD explants from at least three replicates. p values were calculated using the Wilcoxon rank sum test (A.K.A. Mann-Whitney U test). ***p < 0.0005; NS, not significant.

We then computationally applied high tension to four v-junctions in the center of the tissue (marked by asterisks in Figure 4B, D) to mimic the ML converging forces of cell intercalation. The results revealed striking differences in force propagation between the two models, as indicated by force heatmaps in Figure 4C and E (index shown in Figure 4A). In the Hexagon tissue, forces applied in the center propagated in both ML and AP directions, creating high tensions across v-junctions throughout the tissue (Figure 4C, red/yellow). In contrast, the Brick-wall tissue allowed only ML force propagation, creating two ML-aligned chains of high tension (Figure 4E, red); neither t- nor v-junctions above or below the applied forces displayed any elevated tension (Figure 4E, cyan/blue).

As an additional test, we introduced a Hybrid tissue model, incorporating a single perpendicular t-junction in the Hexagon tissue (Figure 4F, arrowhead). Forces applied in this hybrid tissue (asterisks) propagated in the ML direction and also towards the anterior (up) (Figure 4G). However, force propagation towards the posterior (down) exhibited an immediate drop at the inserted perpendicular t-junction (Figure 4G). These preliminary model tests suggested that AP-aligned perpendicular t-junctions can efficiently block AP force propagation at the tissue level.

Aberrant cell packing configuration disrupts planar polarized force coupling during CE in vivo

We proceeded to validate our force propagation model in vivo by evaluating force distribution in tissue explants during CE using the same cellular force inference tool.¹⁸ We found that force distribution was highly heterogeneous across cells within WT explants, with regions of significantly higher and lower tension (Figure 4H). Interestingly, the regions of high tension (dark gray in Figure 4H) consistently spanned four to five cell diameters in both the ML (Figure 4J, blue) and AP directions (Figure 4K, blue), suggesting robust force coupling and propagation in both directions.

By contrast, the heterogenous force distribution in Arvcf KD explants displayed a different pattern. High-tension regions manifested as chains (Figure 4I), coupled in the ML direction (Figure 4J, orange) but not in the AP direction (Figure 4K, orange). This pattern resembled the force propagation pattern in the Brick-wall toy model (Figure 4E). Notably, a considerable number of perpendicular t-junctions surrounded the high-tension region in Arvcf KD explants (Figure 4I, arrowheads). These observations suggest that the excessive perpendicular t-junctions in the Arvcf deficient tissue effectively block the AP propagation of forces from local cell intercalations in vivo.

T-junction rotation promotes the planar polarized spread of cell intercalation in silico

Next, we aimed to examine the function of t-junction rotation in collective cell movement beyond individual T1 transitions, and we again turned to computational modeling for its ability to access experimentally inaccessible parameters.^19,20 We used a dynamic vertex model framework from our previous work²¹ in which junctional tensions are oscillatory with magnitude varying depending on the junction orientation to mimics our experimental observations.^13,21

We first examined an initial cell packing configuration such that all t-junctions aligned along the AP axis (Figure 5A), similar to the static brick-wall model shown in Figure 4D, and used synchronized junctional tension oscillations so that no diagonal tensional differences build up (Figure 5B). With this design, the in-silico tissue displayed no t-junction rotation, no cell intercalation, nor CE; rather, cells changed shape, but movement stalled soon after an initial narrowing driven by the higher tension along the ML axis (Figure 5C).

Figure 5. A single t-junction rotation triggers a cascade of cell movements in a vertex model.

(A) Initial configuration of the cells in the vertex model. All cells were hexagonal and packed so that all t-junctions were aligned along the AP axis. Cells numbered 1–6 were highlighted and their movement was tracked in the simulation.

(B-C) Simulation results showing no cell rearrangement when synchronized and orientation-dependent oscillatory tensions are applied to the model.

(C-E) Simulation results showing active cell intercalations across the tissue when one t-junction in the center was introduced to rotation. The tension between cells 1 and 3 and between cells 4 and 6 was temporarily reduced to induce the rotation of the t-junction between cells 3 and 4.

To explore whether and how t-junction rotation might rescue this stalled tissue, we induced one t-junction rotation in the center by temporarily reducing junction tension on a diagonal pair of v-junctions from the beginning of the simulation (Figure 5D,E, t=4) and lasted until the first neighbor exchange occurred. Soon afterward, we observed cell intercalations between cells near the site in which we induced asymmetric tensions (Figure 5E, t = 120). More excitingly, this local change in the otherwise stalled tissue was followed by multiple rounds of cell intercalations spreading across the AP axis, and these drove significant tissue-level CE (Figure 5E, t=285, t=555). These findings suggest that t-junction rotation was actively and functionally involved in the multicellular mechanical integration for the polarized propagation of cell intercalations in the vicinity.

Defects in t-junction rotation and cell packing disrupt the orderly propagation of convergent extension in the notochord.

Our data suggest that a delay in t-junction rotation and changes in cell packing at smaller scales result in differences in tissue shaping at larger tissue and embryo scales by hindering the polarized propagation of cell intercalation, as suggested in the dynamic vertex model (see Figure 5). To test this idea, we conducted live-imaging of notochord explants undergoing CE over a 2-hour period. At the beginning, WT explants displayed a trapezoidal shape (Figure 6A, left), consistent with anterior-to-posterior spreading of CE cell behaviors, including cell elongation and intercalation, a phenomenon reported decades ago²². At the end of the two-hour imaging, cells in both the anterior and posterior halves of the tissue were equally elongated along the ML axis (Figure S3A, blue), suggesting a uniform CE cell behavior across the tissue. Over the course of imaging, we also observed substantial AP extension and ML convergence across the entire AP axis (Figure 6B; Figure S3B, blue). Quantifying convergence at three positions along the AP axis (Figure 6E, left), revealed consistent tissue narrowing across the entire AP axis (Figure 6E, blue).

Figure 6. Propagation of tissue-scale CE is required for efficient tissue shaping.

(A-D) Convergent extension of WT (A,B) and Arvcf KD (B,D) explants during a two-hour incubation. Dashed lines mark tissue region at the equivalent stage of 12.5. Solid lines indicate the same region post-incubation. Two outlines were stacked together to show the shape change, highlighted with hatching shades (B, D). Hollow arrowheads mark the regions with local convergence, and solid arrowheads mark the regions with minimal to no convergence.

(E) Quantification of ML convergence at the anterior, middle, and posterior ends as illustrated in the left panel. N = 24 for WT explants (blue) and N = 23 for Arvcf deficient explants (orange) from at least three replicates. p values were calculated using the Wilcoxon rank sum test (A.K.A. Mann-Whitney U test). *p < 0.05; **p < 0.005; NS, not significant.

(F, G) Wildtype (F) and Arvcf KD (G) embryos at stage 14 stained by in situ hybridization for the notochord-specific probe Xnot. Left panels, representative images of embryos. Right panels, stacked grayscale images of over 70 embryos for each condition. Dashed lines outline the average shape of a notochord with 50% coverage.

(H) Quantification of normalized width difference between posterior and anterior notochords. N = 89 for WT embryos and N = 72 for Arvcf deficient embryos. p values were calculated using the Wilcoxon rank sum test (A.K.A. Mann-Whitney U test). ***p < 0.0005.

(I) Multiscale biomechanical model underpinning tissue-scale CE propagation. Cell intercalation causes local CE. In WT, timely t-junction rotation resolves 4-cell vertices, enabling forces from other cell intercalation events (asterisks) to propagate in the AP direction, promoting CE propagation. Arvcf loss impedes t-junction rotation, leading to excessive perpendicular t-junctions that block AP force propagation, hindering CE propagation.

See also Figure S3.

Interestingly, when we quantified the shape change in Arvcf deficient explants, there was a significant AP difference (Figure 6C–E). Our data revealed a relatively normal convergence at the anterior end of the tissue, but significantly reduced convergence toward the posterior (Figure 6E, orange), which together contributed to a reduced CE (Figure 6C,D, Figure S3B, orange). Notably, cells in anterior and posterior halves were still elongated to the same extent (Figure S3A, orange), ruling out the possibility that such an AP difference was a result of impaired anterior-to-posterior spreading of CE cell behaviors. Moreover, we consistently observed active cell intercalations at all positions (data not shown), leading to multiple instances of localized convergence along the AP axis, including the most posterior end (Figure 6D, hollow arrowheads). Intriguingly, these localized CE regions were surrounded by areas where convergence was minimal to nonexistent (Figure 6D, solid arrowheads). Such a discontinuous tissue-scale convergence suggested that localized convergence fails to propagate effectively along the AP axis.

Finally, we asked how this tissue-scale CE propagation defect impacted the notochord shape in the embryo. Using in situ hybridization for the notochord-specific marker Xnot,²³ we found that the notochord in WT embryos at early neurulation (NF stage 14) had reorganized into a slim rectangle (Figure 6F), with little difference in tissue width between its anterior and posterior (Figure 6F; 6H, blue). By contrast, the shorter and wider notochords of Arvcf KD embryos¹⁵ displayed a narrow anterior but significantly wider posterior (Figure 6G; 6H, orange), consistent with the observations in tissue explants (Figure 6E).

These findings suggested that the orderly propagation of cell intercalations along the AP axis allows the posterior end to catch up quickly in a WT embryo. While with Arvcf depletion, the impaired propagation of CE magnifies the normal difference in the onset time of cell behaviors, leading to an apparent AP difference. Our data suggest that properly resolving 4-cell vertices through timely t-junction rotation is crucial for normal cell packing, enabling polarized force propagation and efficient tissue-scale CE through synergistic cell intercalations (Figure 6I).

Discussion

Our overarching goal was to unravel the mechanisms connecting microscale and macroscale behaviors to achieve efficient tissue-scale morphogenesis. By employing a combination of live-imaging, image-based mechanical measurements, and computational modeling, we have pinpointed the nascent t-junction growth and rotation at the end of each cell intercalation event as the key to bridge CE behaviors across multiple scales.

To understand its implications, we delved into the realm of cell packing configuration. This concept has been linked to collective cell motility, as suggested by various vertex models designed for epithelia.^7,8 However, the parameters currently considered, such as cell shape index and cell alignment index, do not exhibit apparent differences during t-junction growth and rotation in Xenopus notochord or between the wild type and the Arvcf deficient tissues. This highlights the pressing need for new models that incorporate additional factors.

Our analysis of cell packing underscores its importance in the polarized propagation of cellular forces and cell movement, which collectively facilitate tissue shaping effectively (Figure 4–6). Specifically, we have demonstrated that cell packing configuration with finely tuned orientation of t-junctions affects the coupling and propagation of cellular forces along the AP axis (Figure 4). In addition to this relatively static perspective, our dynamic modeling in-silico and analysis of tissues in vivo suggest that t-junction rotation is critical in coordinating cell movements and facilitating the spreading of cell intercalations along the AP axis (Figure 5,6).

Interestingly, similar concepts of mechano-dependent morphogenetic propagation have been described in Drosophila endoderm invagination and cephalic furrow formation; in both, the propagation manifests a single-directional wavefront.^24,25 By contrast, our preliminary data presented here suggested a tide-like, shorter-lived mechanical wave in synergizing individual cell intercalation events in the neighboring regions. Nevertheless, it is worth exploring whether a tissue-scale wave of Rho1/MyoII activation, in addition to the known calcium wave,²⁶ exists in the Xenopus notochord, and how these waves correlate with local changes of cell packing configuration.

We also explored the cellular mechanisms driving t-junction rotation and demonstrated that high tension on the ML-aligned v-junctions, aside from its well-documented roles in v-junction shortening, is involved in t-junction rotation (Figure 2). We have observed a correlation between the difference in tension in the diagonal pairs of the v-junctions and the rotation of the connected t-junction (see Figure 2I). This correlation is stronger when the tensional difference is high and diminishes when the tensional difference is low (Figure 2J, K), indicating the presence of a threshold that needs to be surpassed. Considering the pulsatile and asynchronous nature of these high-tension on v-junctions,²¹ it is relatively easy to meet this mechanical requirement. By contrast, cells lacking Arvcf have an overall reduction of tension (Figure S2C), creating obstacles for junction rotation (Figure 3F).

Interesting questions arise regarding t-junction rotation. The first is whether the tension at the v-junctions is diagonally enriched to promote more efficient t-junction rotation. Our current data show that in half of the cases, diagonal pairs display the higher tension among the four connected v-junctions but this seemed not to relate to faster t-junction rotation (data not shown). This suggests that such diagonal enrichment is likely a result of stochastic pulsatile junctional tension. However, due to the limited number of datasets collected in this study, we could not fully rule out the possibility of such a diagonal enrichment.

The second question is whether boundary conditions play a role as suggested in other systems.^9,11 The Xenopus notochord is a multilayer non-epithelial structure, and individual cells are allowed to move in the anteroposterior, mediolateral, and dorsoventral axes. Although we did not impose a different AP boundary condition in the WT and Arvcf-deficient embryos or explants, the cells appeared to respond differently to the same boundary condition. One interesting observation is that the en face projection of the notochord tissue remained the same in the WT embryos (Figure S3C, blue). In contrast, Arvcf-deficient embryos displayed a progressive increase in the notochord en face projection (Figure S3C, orange), suggesting unordered cell movement along the mediolateral or dorsoventral axis. That said, our data do not yet allow us to rule out the possibility that boundary conditions also impact this process. Moreover, it remains possible that even if t-junction rotation were rescued in Arvcf morphants that force propagation would not be. Clearly, then, additional study is warranted.

To conclude, our data from in vivo observation, in vitro analysis, and in silico modeling connect force-dependent t-junction rotation with tissue packing configuration and force propagation to cell intercalation. Our data suggest a multifaceted biomechanical linkage between individual cell intercalation events and efficient tissue shaping and, moreover, demonstrate how a subtle mechanical defect at the cellular scale can escalate over time and length scales, leading to a significant change in tissue shape.

Resource availability

Lead Contact

Further information and reagent requests should be directed to Shinuo Weng (s.weng@jhu.edu).

Materials Availability

All reagents are available from lead contact with completed MTA upon request.

Data and Code Availability

• All data reported in this paper will be shared by the lead contact upon request.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

STAR★Methods

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Xenopus embryo manipulations

Ovulation was induced by injecting adult female Xenopus laevis with 600 units of human chorionic gonadotropin (HCG, MERCK Animal Health) and animals were kept at 16°C overnight. Eggs were acquired the following day by squeezing the ovulating females and eggs were fertilized in vitro. Eggs were dejellied in 2.5% cysteine (pH 7.9) 1.5 hours after fertilization and reared in 1/3x Marc’s modified Ringer’s (MMR) solution.^27,28 For micro-injection, embryos were placed in 2% ficoll in 1/3x MMR during injection and washed in 1/3x MMR 30 min after injection. Embryos were injected in the dorsal blastomeres at the 4-cell stage, targeting the C1 cell at the 32-cell stage and presumptive notochord. Keller explants were dissected at stage 10.25 in Steinberg’s solution using hair tools.²⁹ All animal work has been approved by the IACUC of UT, Austin, protocol no. AUP-2021–00167.

METHOD DETAILS

Plasmids and Morpholinos

The Arvcf morpholino has been previously described and characterized (5’-ACACTGGCAGACCTGAGCCTATGGC-3’,¹⁶ and was ordered from Gene Tools, LLC. Membrane-BFP plasmid was made in pCS2.²¹

mRNA and morpholino microinjections

Capped mRNAs were generated using the ThermoFisher SP6 mMessage mMachine kit (Catalog number: AM1340). Membrane-BFP mRNAs were injected at 75 pg per blastomere.

Imaging Xenopus explants

Explants were submerged in Steinberg’s solutions and cultured on glass coverslips coated with Fibronectin (Sigma-Aldrich, F1141) at 5 μg/cm². All images of membrane labeling (Membrane-BFP) were taken on a Nikon A1R or a Zeiss LSM700 confocal microscope, and at the focal plane 5 μm deep into the explant. For tracking tissue-scale CE, the first image was taken after a three-hour incubation at room temperature, that is, at the equivalent stage of 12. A second image was taken another two hours later, at the equivalent stage of 14. All other assays were started after a 5-hour incubation of the explant, at the equivalent stage of 14, if otherwise indicated. For tracking the dynamic of t-junction formation, we set the standard confocal time-lapse imaging at an interval of 5 sec. For the measurement of transverse fluctuation, we set the time interval at 1 sec.

In situ hybridization

Whole-mount in situ hybridization was performed at stage 12 and stage 14, as described previously using a DIG-labeled single-strand RNA probe against a partial sequence of Xnot.^29,30 This antisense probe has been well characterized, which shows Xnot expression in the prenotochordal region about the dorsal lip at stage 10.5 and along the dorsal midline exclusive to the notochord up to stage 16.²³ Bright field images were captured with a Zeiss Axio Zoom. V16 stereo microscope with Carl Zeiss Axiocam HRc color microscope camera or a Leica stereo microscope MDG41 with AmScope microscope digital camera WF200.

Transverse fluctuation analysis

Transverse fluctuation analysis is based on the idea that the transverse fluctuation of a junction is inversely proportional to tension on the junction (Figure 2C,D), and has been previously reported.¹⁵ Here, transverse fluctuation was measured in the 10 μm range of v-junctions conjoining to a forming t-junction. To track the junction movement, we labeled the cell membrane by injecting embryos at the 4-cell stage in both dorsal blastomeres with Mem-BFP mRNA. Keller explants were then dissected from early gastrula embryos, mounted on fibronectin coated coverslips, incubated in the Steinberg’s solution at room temperature for five hours, and then live-imaged on a Nikon A1R confocal microscope for 5 min. To capture the small transverse movement of a junction, we used a time interval of 1 sec and a pixel size of 0.20 μm. For image analysis, time-lapse movies were imported into Ilastik^®, and opposing cells were first segmented using Ilastik^® pixel classification and Ilastik^® Carving. 3D (xyt) meshes of the connected cells were then imported to a customized MATLAB script to detect the transverse fluctuation over time. Briefly, we first defined a base-line position of the junction at each time point to account for the overall movement of the tissue and individual cells. To do so, we performed a moving average over 2 μm along the junction length at each time point, and then another moving average over 20 sec. The transverse fluctuation was then measured at each time point as the distance from the original junction position to the baseline.

Vertex model simulating cell intercalations and convergent extension

A previously published framework of 2D vertex model was employed to simulate cell intercalation and convergent extension to study the function of t-junction rotation.^21,31 In this model, cell shapes and their packing configuration are defined by a network of polygons sharing vertices and edges. T1 transition occurs when two vertices connected by an edge come close, triggering a local network topology change known as the T1 transition rule. In essence, they change the connections with their neighboring cells. Cell shape changes are driven by the movements of individual vertices, which are governed by the equation of motion for the i-th vertex at time t:

$$\eta \frac{d{\mathit{r}}_i}{dt}=-{\nabla }_{i}U$$ (1)

where η is the friction coefficient, r_i denotes the position vector of the i-th vertex, and U is the energy function of the system. In our model, U is given by

$$U={\sum }_{\alpha }\frac{K_{\alpha }}{2}{\left(A_{\alpha }-{A_{\alpha }}^{(0)}\right)}^2+{\sum }_{(i,j)}{\Lambda }_{ij}{l}_{ij}+{\sum }_{\alpha }\frac{{\Gamma }_{\alpha }}{2}{L_{\alpha }}^2$$ (2)

In this equation, the first term expresses area elasticity, where A_α is the area of the α-th cell and K_α represents its elastic constant. The second term expresses the line tension along the edge 〈i, j〉, where l_ij is the length of the edge and Λ_ij is the line tension per unit length. The third term expresses the perimeter elasticity, where L_α is the perimeter of the α-th cell and Γ_α represents its elastic constant. These general formulations are based on previously established 2D vertex models.^6,32

Previous experimental work has found that contractile tension along the cell-cell interfaces is pulsatile and PCP-dependent^13,21. To keep our model simple, nevertheless, suffice to test our hypothesis, the line tension Λ_ij is expressed as

$${\Lambda }_{ij}={\Lambda }_b+p_{ij}{d}_{ij}\Lambda (\text{sin}\omega t+1)$$ (3)

where Λ_b is the baseline tension, Λ is the amplitude of oscillation, and ω is the angular frequency of the oscillation. p_ij is the orientation factor simulating the PCP dependency, calculated as $p_{ij}={\left(\frac{{\mathit{e}}_{\text{x}}\cdot {\mathit{l}}_{ij}}{l_{ij}}\right)}^2$, where e_x is a unit vector along the x-axis, and l_ij is a vector along the edge of the i-th vertex to the j-th vertex, and l_ij is the length of this edge. d_ij in equation (3) is the reduction factor with the default value of 1, indicating no reduction. To induce a t-junction rotation at the specified interface, d_ij associated with a diagonal pair of the conjoining v-junctions are reduced temporarily until a first T1 transition occurs.

All model constants are listed in the following table:

Symbol	Description	Value
K α	Area elastic constant	1.0
Γα	Perimeter elastic constant	1.0
Λb	Baseline tension	−6.0
Λ	Pulsatile tension	0.15
d ij	Reduction factor (1.0 when no reduction)	(0.2, 1.0)
A α (0)	Cell area at stress-free state	2.6
ω	Angular frequency	1.14 × 10−1
η	Friction coefficient	1.0
Δt	Time step size for numerical integration of Eq. (1)	1.0 × 10−4

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical analyses were carried out using Matlab software. Each experiment was conducted on multiple days and included biological replicates. Due to the non-normality and low n-numbers in some assays, we used non-parametric tests, such as the Wilcoxon rank sum test (A.K.A. Mann-Whitney U test) and the Kolmogorov–Smirnov test. In standard box plots, the median, 1^st and 3^rd quartiles were plotted together with individual data points. N values and statistics for each data set can be found in the figure legend.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Chemicals, peptides, and recombinant proteins
chorulon® Human chorionic gonadotropin	MERCK Animal Health	NADA NO. 140–927; RRID:SCR_018232
Fibronectin	Sigma-Aldrich	CAT# F1141
Ficoll	Fisher Scientific	CAT# BP525
Critical commercial assays
mMESSAGE mMACHINE™ SP6 Transcription Kit	ThermoFisher	CAT#AM1340
Experimental models: Organisms/strains
Wild Type Xenopus laevis (male and female)	Xenopus 1	RRID:XEP_Xla100
Oligonucleotides
Morpholino: MO-Arvcf 5’-ACACTGGCAGACCTGAGCCTATGGC-3’	Fang et al.16	Gene Tools
Recombinant DNA
Plasmid: Membrane-BFP	Shindo et al.21
Software and algorithms
ilastik	Berg et al.27	https://www.ilastik.org/
Matlab 2023b		https://matlab.mathworks.com/

Highlights.

• 4-cell vertex in T1 transition is resolved by t-junction extension and rotation.

• Slow t-junction rotation significantly affects how cells are packed in the tissue.

• T-junction rotation underpins polarized force propagation in silico and in vivo.

• Polarized force propagation is involved in the propagation of tissue shaping.