Prickle2 regulates apical junction remodeling and tissue fluidity during vertebrate neurulation

Epithelial tissue fluidity is defined as the ability of epithelia to deform in response to stimuli while preserving structural integrity (Atia et al., 2021; Farhadifar et al., 2007; Guillot and Lecuit, 2013; Mongera et al., 2018; Park and Fredberg, 2016; Wang et al., 2020). In developing vertebrate embryos, an initially flat embryonic epithelium has a high potential to deform. In response to physical or chemical cues, individual epithelial cells change their shape and neighbors in a highly organized anisotropic manner, leading to the formation of tissues or organs. Directional apical junction (AJ) remodeling is a key regulator of epithelial tissue deformation. However, the underlying mechanisms are largely unknown (David et al., 2014; Founounou et al., 2021; Kim et al., 2021; Kuriyama et al., 2014).

The Xenopus ectoderm is a good model for studying tissue fluidity during morphogenesis. At the onset of neurulation, the neuroectoderm (NE) is “fluid-like” with frequent cell intercalation and apical constriction, which are required for neural tube closure (NTC) (Baldwin et al., 2022b; Christodoulou and Skourides, 2022; Matsuda et al., 2023; Williams et al., 2014). In contrast, the nonneural ectoderm (future epidermis) is “solid-like,” characterized by hexagonal packing and a low rate of cell intercalation. Notably, epithelia can reacquire “fluidity” during morphogenetic changes, allowing further deformation (Bocanegra-Moreno et al., 2023; Classen et al., 2005; Lawton et al., 2013; Mao and Wickström, 2024; Mitchel et al., 2020; Tetley and Mao, 2018). For example, the reduced fluidity of the embryonic ectoderm at the end of gastrulation is replaced by increased cell–cell rearrangements and apical domain heterogeneity during neurulation. Therefore, the spatiotemporal regulation of tissue and AJ fluidity is a hallmark of epithelial morphogenesis.

The planar cell polarity (PCP) pathway regulates the directionality of collective cell movements during morphogenesis. Originally defined in Drosophila genetic studies, core PCP components coordinate cell orientation in the tissue plane (Goodrich and Strutt, 2011; Humphries and Mlodzik, 2018; Peng and Axelrod, 2012). Vertebrate PCP is believed to be maintained by the feedback regulation of two complementary core protein complexes, Celsr–Vangl–Prickle (Pk) and Celsr–Frizzled (Fz)–Disheveled (Dvl), which are distributed at AJs on the opposite sides of the cell cortex (Aw and Devenport, 2017; Davey and Moens, 2017; Peng and Axelrod, 2012). In the vertebrate mesoderm, the core PCP pathway promotes mediolateral (ML) cell intercalation during convergent extension movements, leading to defects in embryonic axis elongation (Gray et al., 2011; Keller, 2002, 2012; Sokol, 1996; Wallingford et al., 2002). In addition to defects in axis elongation, mutations in the core PCP components result in neural tube abnormalities (Colas and Schoenwolf, 2001; Henderson et al., 2018; Montcouquiol et al., 2003; Williams et al., 2014). The PCP pathway was proposed to promote ML cell intercalations, extending the NE along the anteroposterior (AP) axis (Ybot-Gonzalez et al., 2007). PCP signaling may also involve defects in radial cell intercalations (Ossipova et al., 2015a) and apical constriction (Galea et al., 2021; Ossipova et al., 2015b), a reduction in the apical domain size that is essential for neural plate folding. Taken together, these studies suggest that PCP signaling may control tissue fluidity during epithelial morphogenesis.

The core PCP components are functionally linked to the actomyosin cytoskeleton. The Fz-Dvl complex binds RhoGEF and Daam1, a formin family protein that activates RhoA in the cell cortex (Habas et al., 2001; Nishimura et al., 2012). The Fz-Dvl complex therefore promotes the formation of supracellular actomyosin cables that help bend the neural plate (McGreevy et al., 2015; Nishimura et al., 2012; Röper, 2013). The Vangl-Pk complex is enriched at the anterior cell edge in various vertebrate models (Butler and Wallingford, 2018; Ciruna et al., 2006; Devenport and Fuchs, 2008; Mancini et al., 2021; Ossipova et al., 2015c; Yin et al., 2008). However, the cellular targets of Vangl and Pk that mediate tissue deformation largely remain to be identified.

In this study, we investigate the role of Prickle2 (Pk2) in AJ remodeling and tissue fluidity control during neurulation. Consistent with its presumed function as a core PCP protein, Pk2 is polarized at the anterior AJ of Xenopus neuroepithelial cells, and Pk2 morphants exhibit NTC defects (Butler and Wallingford, 2018). The same study revealed a correlation between Pk2 and actomyosin enrichment in shrinking AJs; however, the underlying mechanism and causation remain unclear. Here, we show that Pk2 knockdown (KD) in the NE, a fluid-like epithelium, attenuates AJ elongation and shrinkage. Conversely, Pk2 overexpression (OE) in the gastrula ectoderm, a solid-like epithelium, promotes AJ elongation and shrinkage, inhibiting hexagonal cell packing, a hallmark of the ectoderm at the end of gastrulation. This activity of Pk2 is unique and involves an evolutionarily conserved Ser/Thr-rich region (STR) of the protein. It also requires Rac1 activity and is accompanied by increased C-cadherin and actomyosin dynamics at AJs. Importantly, the effects of Pk2 on AJ remodeling are anisotropic in the NE and induce preferential elongation of cells along the AP axis. We propose that Pk2 increases epithelial tissue fluidity, leading to an enhanced response of the epithelia to anisotropic cues from the environment.

We assessed the role of Pk2 in AJ remodeling and tissue fluidity in Xenopus NE, a fluid-like epithelium characterized by intensive AJ remodeling, cell intercalation, and apical domain variability (Baldwin et al., 2022a; Christodoulou and Skourides, 2022; Matsuda et al., 2023; Williams et al., 2014). To unilaterally deplete Pk2 in the NE (Fig. 1 A), a previously characterized splicing–blocking morpholino (MO) (Butler and Wallingford, 2015, 2018; Devitt et al., 2024; Shindo et al., 2019) was coinjected with a lineage tracer RNA into one dorsal blastomere at the four- to eight-cell stage. Unilateral KD minimizes AP axis elongation defects, allowing a comparison of the behaviors of morphant and wild-type NE cells in the same embryo (Fig. 1 A). We assessed changes in AJs, cells, and cell‒cell arrangements in the posterior NE close to the brain‒spinal cord border, where ∼20% of cells per hour intercalate and replace mediolaterally oriented junctions with anteroposteriorly oriented junctions (Butler and Wallingford, 2018).

Pk2 KD disrupted NTC as reported previously (Fig. 1 B and Video 1) (Butler and Wallingford, 2018). The apical domain of Pk2-KD NE cells remained large (Fig. 1, B and C), suggesting defects in apical constriction. In Drosophila, medioapical actomyosin pulses have been implicated in apical constriction during epithelial folding (Martin et al., 2010). In the Xenopus NE, medioapical actomyosin increases in apically constricted cells, and the reduced medioapical F-actin correlates with apical constriction defects (Baldwin et al., 2022a). Consistent with these findings, the Pk2 KD NE presented a loss of apically constricted cells and a loss of cells with strong medioapical F-actin enrichment (Fig. S1, A–E). These results suggest reduced actomyosin contractility in Pk2 KD cells.

PCP signaling controls convergent extension in the NE (Ybot-Gonzalez et al., 2007). In Xenopus embryos, mediolaterally oriented cell intercalation extends the NE anteroposteriorly, followed by apical constriction, which folds the neural tube (Baldwin et al., 2022a; Christodoulou and Skourides, 2022). Therefore, we next tested the effects of Pk2 KD on cell intercalation. Time-lapse imaging revealed that Pk2 KD did not reduce the frequency of T1 transitions (white circles in Fig. 1, D–H and Video 2). It also did not affect the orientation of T1 transitions (a total of 277/277 ML-to-AP intercalation events in three Pk2 KD embryos). However, the elongation of newly formed AP junctions slowed after the resolution of T1 junctions (double arrows in Fig. 1, D–G). Tracking of randomly selected junctions revealed that Pk2 KD generally reduces the rates of elongation and shrinkage of AJs (Fig. 1, I and J). These results suggest that Pk2 is required for the efficient remodeling of AJs and that this may also contribute to the failure of apical constriction in the Pk2 KD NE.

Another change in the Pk2 KD NE is the loss of cell reorientation. Wild-type cells were more mediolaterally oriented in the early neurula and progressively acquired AP orientation during neural fold formation (myrRFP in Fig. 2, A and B; and Video 3). Pk2-depleted cells retained the ML orientation (myrGFP+Pk2MO in Fig. 2, A and B; and Video 3). We confirmed the defects in cell reorientation in the phalloidin-stained embryos (Fig. S1, G–G‴ and I). Although Pk1 depletion disrupted NTC (Fig. S1 H), Pk1-KD cells reoriented in the NE (Fig. 1, K; Fig. S1, H and J; and Video 4). Similarly, Vangl2 KD in the NE did not alter cell reorientation despite strong NTC defects (Matsuda et al., 2023). These results indicate that the observed cell orientation defect is specific to Pk2.

To further define defects in AJ remodeling, changes in AJ length and orientation were analyzed in more detail (Fig. 2, C–G). As shown previously (Baldwin et al., 2022a; Christodoulou and Skourides, 2022), AP junctions were longer than ML junctions in the control NE (Fig. S1 F). Individual AP junctions preferentially elongated, whereas ML junctions shrank (Fig. 2, F and H). Our analysis also revealed that individual AJs change orientation (Fig. 2 F), contributing to the ML-to-AP reorientation of cells. In the Pk2-depleted NE, ML junctions were longer than AP junctions (Fig. S1 F). The elongation and shrinkage of AP and ML junctions had similar frequencies (Fig. 2 H; and Fig. S1, K and L). These results suggest that Pk2 is required for the orientation-dependent control of AJ behavior.

In addition to AJ shrinkage and apical constriction, actomyosin contractile force has been reported to be important for AJ elongation in neighboring cells. During Drosophila AP axis elongation, paired pulsatile apical actomyosin at two opposite sides of T1 junctions was proposed to generate a pulling force, elongating newly formed AJs (Collinet et al., 2015; Finegan et al., 2019; Uechi and Kuranaga, 2019; Yu and Fernandez-Gonzalez, 2016). If a similar mechanism operates in Xenopus NE cells, the reduced actomyosin force in Pk2 KD cells may attenuate the extension of newly formed AP junctions. To test the effects of Pk2 KD on actomyosin, time-lapse imaging was conducted in embryos coexpressing sf9-mNeonGreen and Scarlet-UtrCH, which are live probes for nonmuscle heavy chain IIA and F-actin (Burkel et al., 2007; Hashimoto et al., 2015; Vielemeyer et al., 2010).

In the wild-type NE, sf9 preferentially accumulated in tricellular vertices, in which three or more cells meet (arrowheads in Fig. S2 A). During T1 transition (circles in Fig. S2 B), a pair of sf9-positive crescent-shaped actomyosin networks formed at the anterior and posterior sides of T1 junctions (arrowheads in Fig. S2, B and C; and Video 5), and newly formed AJs elongated between two sf9 crescents (double arrows in Fig. S2, B and C). These actomyosin networks (schematics in Fig. S2 C) are similar to those observed at T1 junctions in Drosophila embryos (Collinet et al., 2015; Finegan et al., 2019; Uechi and Kuranaga, 2019; Yu and Fernandez-Gonzalez, 2016). F-actin levels also increased at T1 junctions (arrows in Fig. S2, B and C); however, the current image resolution was not sufficient to separate F-actin networks in individual cells. In the Pk2 KD NE, medioapical actomyosin was pulsatile (Video 5). Pk2 KD did not affect the formation or spatial localization of sf9 crescents at T1 junctions (arrowheads in Fig. S2, D and E; and Video 5) (56/62 T1 junctions in wild type and 56/63 T1 junctions in Pk2 KD, three embryos each), although additional quantification may be necessary to fully evaluate the effects of Pk2 KD on actomyosin. Nevertheless, these observations suggest that reduced actomyosin contractility in tricellular vertices may not be the only reason for attenuated AJ elongation in Pk2 KD cells.

We next asked whether Pk2 activity in facilitating AJ remodeling is specific for the NE. The superficial ectoderm in Xenopus gastrula contains well-organized AJs and apical‒basal polarity (Cardellini et al., 2007; Fesenko et al., 2000). Importantly, the ectoderm is hexagonally packed at the end of gastrulation (control in Video 6), a hallmark of reduced tissue fluidity (Classen et al., 2005; Farhadifar et al., 2007). In early gastrula (00:00 h in Fig. 3 B), ectodermal cells were diverse in size and shape (control in Fig. 3, G and J) with fewer linear cell‒cell boundaries (control in Fig. 3 I). At the end of gastrulation (stage 12, 04:55 h in Fig. 3, C, G, I, and J), the cells became more uniform in size (control in Fig. 3 G) and more hexagonal in shape (control in Fig. 3 J) and had more linear junctions (control in Fig. 3 I).

Pk2 maternal and zygotic transcripts are broadly expressed in Xenopus gastrula (https://www.Xenbase.org). We used gain-of-function approaches to test whether Pk2 OE increases AJ and tissue fluidity in this solid-like embryonic ectoderm system (Fig. 3 A). Compared with the wild-type control ectoderm, Pk2 OE in early gastrula (00:00 h in Fig. 3 C) did not affect the variability of the apical domain size (Fig. 3 G), AJ linearity (Fig. 3 I), or the number of neighbors (Fig. 3 J). However, Pk2 OE cells continued to exhibit apical domain variability (04:55 h in Fig. 3, C, G, I, and J; and Video 6), accompanied by increased lateral membrane dynamics (Fig. 3, D–F and Video 7) and more frequent cell intercalations (Fig. 4, A–C). These results suggest that Pk2 increases tissue fluidity, inhibiting hexagonal packing. This finding is complementary to the Pk2 morphant phenotype in the NE.

Dynamic remodeling of AJs increases tissue fluidity (Pinheiro and Bellaïche, 2018). To assess the direct effects of Pk2 OE on junction remodeling, we tracked changes in AJ length in nonmitotic cells (Fig. 4, D and E) and in newly formed AJs between two daughter cells after cytokinesis (double arrows in Fig. S3, A and B). In both cases, Pk2 OE accelerated AJ elongation and shrinkage (Fig. 4, F–I and Fig. S3 C). Notably, the orientation of AJ remodeling was random in the gastrula ectoderm, and Pk2 OE did not affect directionality (Fig. S3 E). The mitotic rate was similar between the Pk2 OE and control ectoderms (Fig. S3 D), excluding the potential effect of cell division on increasing tissue fluidity (Devany et al., 2021; Firmino et al., 2016; Matoz-Fernandez et al., 2017; Ranft et al., 2010). These results indicate that Pk2 promotes AJ remodeling, increasing tissue fluidity in the gastrula ectoderm.

Cadherin turnover is important for the dynamic remodeling of cell junctions (Kowalczyk and Nanes, 2012). To test AJ dynamics more directly, time-lapse imaging of C-cadherin was conducted. In the Pk2 OE ectoderm, the distribution of YFP-C-cadherin was less uniform at bicellular junctions (BJs) (Fig. 5, A–C and Video 8). C-cadherin puncta in the cytoplasm indicate increased endocytosis of C-cadherin (arrowheads in Fig. 5 B). In the fluorescence recovery after photobleaching (FRAP) assay (Fig. 5, D–G), the increased fraction of C-cadherin fluorescence signals was recovered 80 s after photobleaching (Fig. 5 F). The average recovery of C-cadherin was faster in the Pk2 OE ectoderm (Fig. 5 G). Notably, C-cadherin recovery was highly variable in the Pk2 OE ectoderm (Fig. 5, E and F). We observed that the C-cadherin fluorescence intensity decreased or exceeded the level observed before photobleaching (Fig. 5 D). The constant drift in the baseline fluorescence of C-cadherin (Fig. 5 C) appears to contribute to the inconsistent recovery of C-cadherin. Taken together, these results suggest increased turnover of C-cadherin, which is consistent with increased AJ dynamics in the Pk2 OE ectoderm.

TCJs are hotspots of AJ remodeling and actomyosin tension and contain AJ and actomyosin components and TCJ-specific proteins (Bosveld et al., 2018; Finegan et al., 2019; Letizia et al., 2019; Uechi and Kuranaga, 2019; Vanderleest et al., 2018). TCJs are constantly remodeled; for example, E-cadherin levels at TCJs oscillate coordinately with changes in the apical domain size (Vanderleest et al., 2018). During AJ length changes, vertices at each end of BJs displace asymmetrically, and the displacement of vertices correlates with the increased fluctuation of E-cadherin (Huebner et al., 2021). In the gastrula ectoderm, Pk2 OE did not change the average abundance of C-cadherin (Fig. 5 H, bold lines in Fig. 5 I); however, it increased the variability of C-cadherin levels among TCJs (Fig. 5 I) and the fluctuation at individual TCJs (Fig. 5, J and K). These results suggest that Pk2 promotes C-cadherin dynamics at TCJs.

Next, we asked whether Pk2 OE specifically affects C-cadherin or TCJ dynamics in general. Tricellulin is a tricellular tight junction protein essential for the structural integrity and barrier functions of TCJs (Bosveld and Bellaïche, 2020; Ikenouchi et al., 2005). Compared with the wild-type tissue (Fig. 5, L–M′), the ectoderm overexpressing Pk2 presented more variable (Fig. 5 P) and overall reduced levels of tricellulin (Fig. 5 O) at the TCJs, which was mislocalized to the BJs more frequently (arrowheads in Fig. 5, L–N) and increased the variability of fluorescence signals at BJs (Fig. 5 Q). These results suggest a general change in TCJs in response to Pk2.

Anisotropic actomyosin network formation has been reported at tricellular vertices during junction remodeling in elongating tissues (Collinet et al., 2015; Finegan et al., 2019; Rauzi et al., 2008, 2010; Uechi and Kuranaga, 2019). We tested actomyosin dynamics in the Pk2 OE ectoderm and observed changes in the spatial distributions of F-actin and nonmuscle myosin II (NMII) in tricellular vertices (Fig. S4 and Video 9). In the wild-type ectoderm (Fig. S4 A), coherent NMII- and F-actin–positive puncta were observed near the center of the TCJs (arrows in Fig. S4, A-a and C). In the Pk2 OE ectoderm, NMII puncta split into two to three separable structures with variable sizes and intensities (arrowheads in Fig. S4, B-b and b′; and Fig. S4 C), reducing overall actomyosin abundance at TCJs (Fig. S4, D and E). The pattern and number of NMII puncta at TCJs varied over time in the Pk2 OE ectoderm (Video 9). These results provide further evidence that Pk2 promotes dynamic TCJ remodeling.

To determine the domains of Pk2 required for increased tissue fluidity, Pk2 deletion mutants (Fig. 6 A) were overexpressed in the gastrula ectoderm. Hexagonal cell packing was assessed by AJ linearity (Fig. 6 H), the variability in junction length (Fig. 6 I), and the number of neighbors (Fig. 6 J). Three domains were required to inhibit hexagonal packing. First, the C-terminal CAAX motif was required (Pk2ΔC30 in Fig. 6, G–J), a domain that mediates the membrane localization of Pk proteins (Fig. S5 N) (Cho et al., 2015). Second, it requires the Vangl-binding domain (VBD) (Butler and Wallingford, 2015; Jenny et al., 2003) (Pk2ΔVBD in Fig. 6, F and H–J), even though Pk2ΔVBD was localized at AJs (Fig. S5 M). Although Vangl2 OE alone did not inhibit hexagonal packing (Fig. S5, E and G–I), our finding that Pk2ΔVBD is inactive suggests that Pk2 cooperates with Vangl2 at AJs for full activity. The third required domain was the STR (Pk2ΔSTR in Fig. 6, D and H–J). The STR is an evolutionarily conserved domain among vertebrates Pk1 and Pk2, but it is absent in invertebrate Pk and vertebrate Pk3 (Fig. S5, A and B). We confirmed that Pk2ΔSTR localized at AJs in the gastrula ectoderm (Fig. S5 K). Other regions of Pk2, such as C4, were not required for AJ localization (Fig. S5 L) or the inhibition of hexagonal packing (Fig. 6, E and H–J).

Previous studies have implicated Pk1 in cell motility and F-actin reorganization (Carreira-Barbosa et al., 2003; Huang and Winklbauer, 2022; Tao et al., 2009; Veeman et al., 2003) through interactions with GEFs and GAPs for Rho family small GTPases (Daulat et al., 2019; Zhang et al., 2016; Zhang and Wrana, 2018). ARHGAP21/23, GTPase-activating proteins for RhoA, bind Pk1 via a region including the STR, inactivate RhoA, and are proposed to regulate the formation of actin-rich protrusions during mammalian cell migration in vitro (Zhang et al., 2016). In our assays, Pk1 KD in the NE disrupted NTC (Fig. S1 H); however, Pk1 KD cells reoriented from the ML to the AP direction (Fig. S1 J). Pk1 OE did not inhibit hexagonal cell packing in the gastrula ectoderm (Fig. 7, C and I–K). However, mosaically expressed Pk1 induced the formation of concave cells with protrusion-like structures adjacent to TCJs (arrowheads in Fig. S6 B). In contrast, mosaic Pk2 OE cells had a convex shape (Fig. S6 C). The mechanisms underlying these morphological changes are not known, but differences in cell adhesion or cortical tension between neighboring wild-type and Pk OE cells may contribute to these changes. These observations revealed distinct activities of Pk1 and Pk2 in AJ remodeling and cell shape regulation.

To determine the domain responsible for the functional difference between Pk1 and Pk2, chimeric Pk1 and Pk2 proteins (Fig. 7 A) were expressed in the gastrula ectoderm. First, the C-terminal half of Pk2 conferred the ability to promote tissue fluidity to Pk1 (Pk1N-Pk2C in Fig. 7, D and I–K). Pk2N-Pk1C had the opposite effect (Pk2N-Pk1C in Fig. 7, G and I–K). The STRs contain amino acids conserved among both Pk1 and Pk2, but they also contain amino acids conserved only among Pk1 or Pk2 (Fig. S5 B). Therefore, we tested whether these unique amino acids are responsible for the specific functional activities of Pk1 and Pk2.

The insertion of the Pk1 STR domain into the Pk2 backbone suppressed the tissue-fluidizing activity of Pk2 (Pk2(1STR) in Fig. 7, H and I–K), whereas the complementary construct had the opposite effect (Pk1(2STR) in Fig. 7, E and I–K). The rescue by the STR domain was less complete than the rescue by Pk2N-Pk1C or Pk1N-Pk2C. However, swapping the VBDs did not affect cell packing (Fig. S6, I–P). These results suggest that the STR domain is responsible for the tissue-fluidizing activity of Pk2, whereas it may require additional domains for full activity. For further assessment, the STR swap mutants were mosaically expressed in the ectoderm. The STR of Pk1 was sufficient to change the shape of mosaic cells to more concave (Pk2(1STR) in Fig. S6 E). The complementary construct changed the cell shape to more convex (Pk1(2STR) in Fig. S6 D). We quantified cell shape changes by circularity and solidity (Fig. S6, F–H), the parameters for compactness and convexity of shapes, via particle measurement tools in ImageJ. A maximum value of 1 represents a perfect circle (Ant et al., 2023; Cox, 1927; Msibi and Mabandla, 2019; Zanier et al., 2015). The OE of Pk3, which does not contain STR (Fig. S5 A), did not inhibit hexagonal packing with or without Vangl2 coexpression (Fig. S5, D and F–I). These results suggest that the STR domain is responsible for the functional differences between Pk1 and Pk2 in AJ remodeling or actomyosin.

Rho family small GTPases regulate AJs and actomyosin (Arnold et al., 2017; Citi et al., 2014; Heasman and Ridley, 2008). Our attempt to test the physical interaction of Pk2 with ARHGAP21/23 was not successful because of nonspecific binding. Alternatively, we coexpressed the dominant interfering forms of RhoA or Rac1 (DN-RhoA or DN-Rac1) (Hall, 1998) in the gastrula ectoderm to test the requirement of Rho or Rac1 for Pk2 ability to promote tissue fluidity.

DN-Rac1 suppressed, while DN-RhoA somewhat enhanced, the effects of Pk2 OE on cell packing (Fig. 8, D–G, J, and K). The complementary effects of DN-Rac1 and DN-RhoA on Pk2 activity are consistent with the mutually antagonistic roles of RhoA and Rac1 in actomyosin and AJs (Burridge and Wennerberg, 2004; Chauhan et al., 2011). Neither DN-RhoA nor DN-Rac1 had a substantial effect on their own (Fig. 8, B, C, and G–I). Taken together, these results suggest that Pk2 requires active Rac1 to promote AJ remodeling.

Pk2 KD attenuated AJ remodeling and reduced the orientation-based patterns of AJ behaviors in the NE (Fig. 2 H; and Fig. S1, K and L). Pk2 OE in the ectoderm promoted AJ remodeling, but the direction of AJ remodeling was random (Fig. S3 E), suggesting that factors other than Pk2 control the direction of AJ remodeling in the NE. To test this hypothesis, we expressed Pk2 in the NE at relatively high levels, leading to its uniform distribution at AJs (high Pk2 in Fig. 9 A). Compared with the control, uniform Pk2 OE in the NE increased the fraction of cells that extended along the AP axis (Fig. 9, B, C, and F; and Video 10). Although Pk1 OE did not have this activity (Fig. 9, D and G; and Video 10), Pk1(2STR) did (Fig. 9, E and H; and Video 10), further supporting the role of the STR domain of Pk2 in junction remodeling. These findings suggest that Pk2 has a permissive rather than instructive role in anisotropic AJ remodeling in the NE and requires another factor(s) for the directionality of the response.

Moderately expressed exogenous Pk2 preferentially localizes at anterior AJs in the NE (low Pk2 in Fig. 10 A) (Butler and Wallingford, 2018). When the increase in Pk2 was limited to anterior AJs, Pk2 OE did not elongate NE cells along the AP axis (compare Fig. 9 A and Fig. 10 A) or disrupt NTC (Fig. 10 B). However, tracking of individual AJs revealed that Pk2(low) further enhanced the anisotropy of AJ remodeling in the NE, as confirmed in three independent embryos (Fig. 10 C). The overall rate of AJ elongation increased (Fig. 10 D), whereas AJ shrinkage was not significantly affected (Fig. 10 E). Thus, while the Pk2 protein itself does not have an instructive role, its planar localization at anterior AJs enhances the anisotropy of AJ remodeling in the NE, cooperating with instructive cues in the environment.

This study reveals a previously uncharacterized role of Pk2 in tissue fluidity in the Xenopus ectoderm during neurulation. Loss-of-function and gain-of-function approaches have shown that Pk2 is required to maintain the fluidity of AJs and enhance the anisotropy of AJ remodeling in the NE, a tissue with high fluidity. In contrast, Pk2 OE helped us assess its inhibitory effects on cell packing in nonneural ectoderm, a tissue with low fluidity, and test the requirement of the STR domain in controlling tissue fluidity. In our model, Pk2 promoted AJ remodeling (Fig. 10 F), stimulating cell shape changes and cell‒cell rearrangements in the epithelia (Fig. 10 G). Vangl2 association is required for the full activity of Pk2. Given the anterior enrichment of Pk2 and Vangl2 in the NE (Fig. 10 A) (Butler and Wallingford, 2018; Ossipova et al., 2015c), Pk2 may locally increase junction remodeling at anterior BJs and TCJs, where the crescent-shaped actomyosin network promotes the elongation of AP junctions (Fig. S2). We propose that environmental chemical or mechanical cues cooperate with Pk2 to direct anisotropic AJ remodeling and tissue deformation. The anterior localization of Pk2 in the NE amplifies the effect of anteroposteriorly oriented environmental cues on AJ remodeling (Fig. 10 H). These cues may include tensile stresses originating from the elongation of the embryo body axis (Christodoulou and Skourides, 2022; Clausi and Brodland, 1993; Davidson and Keller, 1999; Handler et al., 2023; Hirano et al., 2022; Moon and Xiong, 2022; Smith and Schoenwolf, 1997; Sokol, 2016; Xiong et al., 2020; Zhou et al., 2015). Alternatively, diffusible morphogen gradients may be responsible for the directionality of Pk2 effects. In support of the latter possibility, several Wnt and FGF ligands are expressed predominantly posteriorly, in a graded manner during neurulation (Aulehla and Pourquié, 2010; Diez Del Corral and Morales, 2017; Edri et al., 2023; Niederreither et al., 1999; Nordström et al., 2002; Storey et al., 1998; Yamaguchi, 2001; Yoon et al., 2023).

Mechanistically, the Pk2 STR domain and Rac1 activity play key roles in Pk2-stimulated AJ remodeling. The presence of the STR domain in Pk1 and Pk2 parallels their ability to affect cell morphology (Carreira-Barbosa et al., 2003; Huang and Winklbauer, 2022; Tao et al., 2009; Veeman et al., 2003; Zhang et al., 2016). Rac1 is a well-known regulator of AJ remodeling (McCormack et al., 2013; Nelson, 2009; Watanabe et al., 2009) that can directly affect AJ components or via the reorganization of the actomyosin network (Chen et al., 2017; Daneshjou et al., 2015; Erasmus et al., 2015; Grikscheit et al., 2015; Izumi et al., 2004; Timmerman et al., 2015). Increased actomyosin dynamics in Pk2-overexpressing tissue supports the latter possibility. Pk2 KD may reduce actomyosin contractility at T1 junctions to attenuate AJ elongation as reported during tissue elongation in Drosophila embryos (Collinet et al., 2015; Finegan et al., 2019; Uechi and Kuranaga, 2019; Yu and Fernandez-Gonzalez, 2016). Alternatively, Pk2 may increase cadherin turnover through Rab11-mediated endocytic and recycling machinery (Akhtar and Hotchin, 2001; Amin et al., 2013; Bouchet et al., 2016; Erasmus et al., 2021; Frasa et al., 2010; Wayt et al., 2021; Woichansky et al., 2016). In our model, Pk2 binds RhoA-GAPs via its STR domain and locally activates Rac1, similar to ARHGAP21/23 binding to the STR domain of Pk1 (Zhang et al., 2016). Considering the functional differences between Pk1 and Pk2, the STR domain of Pk2 may interact with Rac1 regulators other than ARHGAP21/23. Identification of these GAPs or GEFs will be necessary to further characterize the function of the Pk2-Rac1 pathway in junction remodeling and actomyosin reorganization.

The role of Vangl2, a known binding partner of Pk2, in tissue fluidity remains unclear. Vangl2 OE did not affect cell packing in the gastrula ectoderm (Fig. S5, E and G–I) or cell reorientation in the NE (Matsuda et al., 2023), but its binding to Pk2 appears to be required for full Pk2 activity in AJ remodeling. Due to the proximity of the VBD to the STR, Vangl2 association may regulate Pk2 interaction with Rho-GAPs or Rac-GEFs. Vangl2 has also been reported to bind N-cadherin and promote its internalization at synapses (Nagaoka et al., 2014), indicating that Vangl2 may link Pk2 to the control of cadherin dynamics. Another candidate is Par3, which binds Pk3 and controls NTC in Xenopus embryos (Chuykin et al., 2018). Par3 is an essential regulator of apical‒basal polarity and has been implicated in PCP signaling, actomyosin, and epithelial folding (Bellaïche et al., 2001; Besson et al., 2015; David et al., 2013; Ishiuchi and Takeichi, 2011; Sawyer et al., 2011; Simoes et al., 2010; Wang et al., 2012). Further studies of Vangl2, other PCP components, and their cellular targets are needed to elucidate how Pk2 controls directional AJ remodeling in response to embryonic cues.

pCS107-GFP-xPk2 was kindly provided by John Wallingford (University of Texas, Austin, TX, USA) (Butler and Wallingford, 2015, 2018). Xenopus Pk1S cDNA was obtained from Open BioSystems (Horizon). pCS2-3xFlag-xPk3 was generated from pCS2-Flag-Pk3 (Ossipova et al., 2015a) via PCR with primers containing the corrected C-terminal sequence (Pk3-F; 5′-GAG​CTT​AAG​CGA​CCC​AGA​TG-3′; Pk3-R: 5′-CTC​GAG​TTA​TGA​AAG​AAG​GCA​ACT​TTT​G-3′). Subcloned Pk2 deletion mutants or Pk1S was inserted into either pCS107-Flag or pCS2-3xHA via restriction enzymes and T4 ligase. The domains in the C-terminal half of Pk2 are defined as follows: C1 (aa. 367–561), STR (aa. 562–665), C4 (aa. 666–733), and VBD (aa. 734–827). The NEBuilder HiFi DNA Assembly Mastermix (catalog # E2621; NEB) was used for the construction of chimeric Pk1 and Pk2 mutants using the following primers: FLAG-PK2-N-V-R: 5′-TTC​TCC​CAC​ACT​GCA​TGC-3′; FLAG-PK2-N-V-F: 5′-TCC​TAG​GCG​GCC​GCG​GC-3′; PK1S-C-I-F: 5′-GCA​TGC​AGT​GTG​GGA​GAA​GAT​GTG​CAT​GCA​TCG-3′; PK1S-C-I-R: 5′-GCC​GCG​GCC​GCC​TAG​GAA​ATA​ATG​CAG​TTT​TTG​CC-3′; HA-PK1S-V-F: 5′-GAT​GTG​CAT​GCA​TCG​GAC-3′; HA-PK1S-V-R: 5′-TGA​ATT​CGC​GTA​ATC​TGG-3′; PK2-N-I-F: 5′-CCA​GAT​TAC​GCG​AAT​TCA​ATG​TTT​AAC​CGG​AGC-3′; PK2-N-I-R: 5′-GTC​CGA​TGC​ATG​CAC​ATC​TTC​TCC​CAC​ACT​GCA​TG-3′; PK1S-C1-V-R: 5′-TTC​TTG​TAG​CTT​TCG​GC-3′; PK1S-C4-V-F: 5′-GAA​TTA​ACT​GCA​CAG​AAT​AAT​ATC​CGT​CGG​CCG​C-3′; PK2-C2-I-F: 5′-GCA​GCC​GAA​AGC​TAC​AAG​AAA​AGT​TTG​AGG​AAA​AGA​AGC-3′; PK2-C2-I-R: 5′-ATT​ATT​CTG​TGC​AGT​TAA​TTC-3′; PK2-C1-V-R: 5′-GGG​AAC​TCT​GAG​ACT​GG-3′; PK2-C4-V-F: 5′-GGA​GAC​TAC​AGC​AGC​ATA​GAA​ATA​AAA​ATG​ACC​CCG​ATG-3′; PK1S-C2-I-F: 5′-GAT​CCA​GTC​TCA​GAG​TTC​CCC​TGG​ATA​TGG​ATC​ACG​G-3′; PK1S-C2-I-R: 5′-TTC​TAT​GCT​GCT​GTA​GTC​TCC-3′. pCS2-GFP, myrGFP, and myrRFP were described previously (Matsuda et al., 2023). pCS2-3xGFP-mZO1 was subcloned from pCAGGS-GFP-mZO1 (Matsuda et al., 2004) by restriction enzyme and T4 ligase. pCS2-Lifeact-GFP was constructed from pCS2-GFP and annealed oligonucleotides encoding Lifeact probe peptides (Lifeact-F: 5′-CGA​TAC​CAT​GGG​CGT​GGC​CGA​CTT​GAT​CAA​GAA​GTT​CGA​GTC​CAT​CTC​CAA​GGA​GGA​GAT-3′; Lifeact-R: 5′-CGA​TCT​CCT​CCT​TGG​AGA​TGG​ACT​CGA​ACT​TCT​TGA​TCA​AGT​CGG​CCA​CGC​CCA​TGG​TAT-3′). Scarlet-UtrCH was inserted into the pCS2 vector from 3xmScarlet I-UtrCH, a gift from Dorus Gadella (University of Amsterdam, Amsterdam, Netherlands) (plasmid # 112962; Addgene) (Chertkova et al., 2020, Preprint) using restriction enzyme and T4 ligase. pT7T-C-cadherin-mYFP and pT7T-C-cadherin-GFP were generated from pT7T-CcadTSMod, a gift from Marc Tramier (University of Rennes, Rennes, France) (plasmid # 99874; Addgene) (Herbomel et al., 2017), by swapping TSMod to either mYFP or GFP. pCS2-mNeonGreen-Sf9 and pCS2-GFP-tricellulin were generous gifts from Ann Miller (University of Michigan, Ann Arbor, MI, USA) (Higashi et al., 2016) and Ed Munro (University of Chicago, Chicago, IL, USA). Sf9 is a single-chain variable fragment (scFv) of an antibody to NMII and is used as a biosensor of NMIIA (Hashimoto et al., 2015). DN-RhoA and DN-Rac1 in the pRK5 vector were gifts from Alan Hall (Sloan-Kettering Institute, New York, NY, USA) (Hall, 1998). Capped RNA was synthesized via mMESSAGE mMACHINE SP6 Transcription Kit (catalog # AM1340; Thermo Fisher Scientific) and purified with RNeasy Mini Kit (catalog # 74104; Qiagen). Pk2 MO (5′-GAA​CCC​AAA​CAA​ACA​CTT​ACC​TGT​T-3′) and Pk1 MO (5′-CTT​CTG​ATC​CAT​TTC​CAA​AGG​CAT​G-3′) (Gene Tools) were characterized in previous studies (Butler and Wallingford, 2018; Daulat et al., 2012; Huang and Winklbauer, 2022; Takeuchi et al., 2003).

Wild-type and albino Xenopus laevis were purchased from Nasco and Xenopus1, and maintained and handled following the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol for animal use was approved by the Institutional Animal Care and Use Committee (IACUC) of the Icahn School of Medicine at Mount Sinai. The sex of the animals was not considered in the study design or analysis because the study subjects were sexually indifferent embryos. In vitro fertilization and embryo culture were performed as previously described (Ossipova et al., 2014). Embryo staging was determined according to Nieuwkoop and Faber (Nieuwkoop and Faber, 1994). For microinjections, embryos were transferred to 3% Ficoll PM400 (catalog # 170300; Cytivia) in 0.5× Marc’s modified Ringer’s (MMR) solution (50 mM NaCl, 1 mM KCl, 1 mM CaCl₂, 0.5 mM MgCl₂, and 2.5 mM HEPES [pH 7.4]) (Peng, 1991).

RNA or MO in 5–10 nl of RNase-free water (catalog # AM9937; Thermo Fisher Scientific) was microinjected into one to two animal blastomeres of four- to eight-cell-stage embryos. For apical domain imaging of gastrula embryos, two ventral-animal blastomeres were injected with 150 pg of RNA encoding Flag- or HA-tagged Pk proteins and coinjected with tracer RNA, including 50 pg of myrGFP, 50 pg of Lifeact-GFP, 50 pg of Scarlet-UtrCH, 200 pg of mNeonGreen-Sf9, 20 pg of tricellulin-GFP, or 100 pg of 3xGFP-ZO1. For apical domain imaging of the neurula, one dorsal-animal blastomere was injected with either 10 ng of Pk2 MO, 150 pg of Flag-Pk2 or HA-Pk1 RNA for high expression, or 50 pg of Flag-Pk2 for low expression, together with 50 pg of myrGFP or 50 pg of Lifeact-GFP. The other dorsal-animal blastomere was injected with 50 pg of myrRFP or 50 pg of Scarlet-UtrCH. For imaging of actomyosin in the NE, 200 pg of mNeonGreen-Sf9 RNA and 50 pg of Scarlet-UtrCH RNA were coinjected with or without 10 ng of Pk2 MO into one dorsal-animal blastomere of a four- to eight-cell-stage embryo. For imaging to determine Pk2 distribution, RNA encoding GFP-Pk2 or Pk2 deletion mutants was coinjected with 50 pg RNA encoding myrRFP or Scarlet-UtrCH. RNA- or MO-injected embryos were cultured in 0.1x MMR until the early gastrula or neurula stage. Each injection included at least 20 embryos per condition. The experiments were repeated at least three times.

For phalloidin staining of the neurula, embryos were fixed in MEMFA (100 mM MOPS [pH 7.4], 2 mM EGTA, 1 mM MgSO₄, 3.7% formaldehyde) (Harland, 1991) for 1 h at room temperature. After permeabilization in 0.1% Triton X-100 in PBS for 10 min, the embryos were incubated with Alexa Fluor 555–conjugated phalloidin (1:400 dilution; catalog # A34055; Thermo Fisher Scientific) in PBS containing 1% BSA overnight at 4°C. The dissected neural plate was mounted on a glass slide with two coverglass spacers (0.13–0.17 mm) to minimize damage to the morphology of the NE.

Images of the fixed embryos were captured either by the BC43 spinning disk confocal microscope (Fusion Ver 2; Andor, Oxford Instruments) or by Zeiss LSM 980 with an Airyscan 2 confocal microscope (Zen [blue edition] Ver 3.7.97.04000, Zeiss). Nikon CFI Plan Apochromat Lambda D 20× (NA = 0.8 and WD = 0.8 mm), Nikon CFI Plan Apo Lambda S 40XC Sil (NA = 1.25 and WD = 0.3 mm), Zeiss Plan Apochromat 20×/0.8 (NA = 0.8 and WD = 0.55 mm), or Zeiss C-Apochromat 40×/1.2 W (NA = 1.2 and WD = 0.28 mm) were used for gastrula imaging. Nikon CFI Plan Apochromat Lambda D 20× (NA = 0.8 and WD = 0.8 mm) was used for imaging the neurula. The experiments were repeated at least three times. Each experiment included at least three embryos per experimental group. Z-stack images were projected into a single image via the maximum projection function in ImageJ2 (ver. 2.9.0/1.54 g), which was used for further analysis and quantification.

Embryos were injected with RNA encoding Pks and fluorescent protein-tagged marker proteins. 5 to 10 embryos per individual group were mounted in 1% low melting temperature agarose (catalog # 50101; Lonza) in 0.1× MMR on a glass slide attached to a silicone isolator (1.6 mm depth; catalog # 664204; Grace Biolabs) or on a glass-bottom dish (catalog # D35-28-1.5-N; Cellvis). Time-lapse imaging was carried out at room temperature either on an Andor BC43 spinning disk confocal microscope, a Zeiss LSM 980 confocal microscope, or a Zeiss Axio Zoom V16 fluorescence stereomicroscope equipped with an AxioCam 506 mono CCD camera. Images were taken every 1 s to 6 min for a period of 15 min to 5 h. The multiposition tool in Fusion or Zen (blue edition) imaging software was used for simultaneous time-lapse imaging. Z-stack images were projected into a single image via the maximum projection function in ImageJ.

FRAP was performed on stage 11–11.5 embryos via a Zeiss LSM 980 confocal microscope with Zeiss C-Apochromat 40×/1.2 W (NA = 1.2 and WD = 0.28 mm) at room temperature. A circular region 2 μm in diameter in the animal-side ectoderm was photobleached via a 488-nm laser (20% laser power; 1.02-μs/pixel dwell time, 10 iterations). The embryos were imaged every second after photobleaching for 80 s. GraphPad Prism (ver.10.3.1) was used to fit the C-cadherin FRAP data with a two-phase exponential curve and determine the halftime of recovery and the recovered fractions of cadherins. Statistical significance was determined by unpaired Student’s t test.

Image processing and quantification were performed as previously described (Matsuda et al., 2023). Briefly, grayscale images of the cell outline marker were segmented via the Python package CellPose v2.0.5 (Pachitariu and Stringer, 2022). Segmented cells were tracked across time points via the Python package Bayesian Tracker (btrack v0.4.5) (Ulicna et al., 2021) with manual correction. For NP segmentation, the NP area was defined by brighter signals of either Lifeact-GFP, Scarlet-UtrCH, or phalloidin staining than in the surrounding epidermis. Pigmented wild-type embryos were used for image segmentation.

The morphology and fluorescence intensity of individual junctions and apical domains were quantified as described previously (Matsuda et al., 2023). Briefly, after segmentation, each TCJ was first identified as a vertex or edge of the cell outline network. Each BJ was identified between two vertices. The mean fluorescence intensities were measured via dilated masks of the pixel or a set of pixels that defined BJs and TCJs (five pixels, 0.76155 μm in width or diameter). The medioapical domain was the area that did not correspond to TCJs or BJs.

Histograms and dot plots of the experimental data were generated via GraphPad Prism 10 (ver. 10.1.0) and Microsoft Excel. Individual experiments were repeated at least three times. The standard deviations (S.Ds.) were calculated via GraphPad Prism 10 and used to assess the variability among sample groups. Student’s t test (two-sided) or the Mann‒Whitney test was used to compare means or mean ranks between two groups, respectively. One-way ANOVA was used to compare the means of more than two groups, with the Bonferroni correction for pairwise comparisons. The two-sample Kolmogorov‒Smirnov test was used to compare distributions between two groups. The chi-square test was used for the categorical data.

Fig. S1 shows Pk2 KD inhibits AP cell orientation in the NE. Fig. S2 shows asymmetrical actomyosin network formation during T1 transition and AJ elongation. Fig. S3 shows Pk2 promotes the elongation of newly formed AJs after cytokinesis. Fig. S4 shows Pk2 changes the organization of the actomyosin network at TCJs. Fig. S5 shows neither Pk1 nor Pk3 increases tissue fluidity in the gastrula ectoderm. Fig. S6 shows the STR is responsible for the functional difference between Pk1 and Pk2 on AJs and actomyosin. Video 1 shows a representative time-lapse video showing the dynamics of NE cells during NTC in Pk2-KD embryos. Video 2 shows a representative time-lapse video of the control (left) and Pk2-depleted NEs (right). Video 3 shows a representative time-lapse video showing changes in cell orientation (green lines) in control NEs and Pk2-KD NEs. Video 4 shows a representative time-lapse video of the Pk1 morphant NE. Pk1 MO was coinjected with Lifeact-GFP in one dorsal blastomere. Video 5 shows a representative time-lapse video of the NE expressing Scarlet-UtrCH and sf9-mNeon, which are live probes for F-actin and NMII. Video 6 shows a representative time-lapse video showing hexagonal packing during gastrulation in wild-type (left) and Flag-Pk2 OE (right) embryos. Video 7 shows a representative time-lapse video showing hexagonal packing and lateral domain dynamics in the control wild-type (left) and Flag-Pk2 OE (right) ectoderms. Video 8 shows a representative time-lapse video showing YFP-C-cadherin dynamics in the control wild-type (left) and Flag-Pk2 OE (right) ectoderms. Video 9 shows a representative time-lapse video showing the dynamics of F-actin and NMII at TCJs. Video 10 shows a representative time-lapse video showing the dynamics of NE cells expressing high levels of Flag-Pk2 (left panel), HA-Pk1 (middle panel), or HA-Pk1(2STR) (right panel).

Data are available in the article itself and its supplementary materials. The imaging data that are not shown but used in the reported quantifications are available upon reasonable request from M. Matsuda or S.Y. Sokol.

We thank D. Gadella, A. Hall, A. Miller, E. Munro, M. Tramier, and J. Wallingford for plasmids. We also thank Jakub Harnos and Chih-Wen Chu for the construction of plasmids encoding Pk2 deletion mutants. We are grateful to Tanya Whitfield, Karen Kasza, and Sassan Ostwar for their comments on the manuscript and members of the Sokol laboratory for valuable discussions. We acknowledge the help from the ISMMS Microscopy Core facility.

This research was supported by the National Institutes of Health grant R35GM122492 to S.Y. Sokol.

Author contributions: M. Matsuda: conceptualization, data curation, formal analysis, investigation, methodology, validation, visualization, and writing—original draft, review, and editing. S.Y. Sokol: conceptualization, funding acquisition, methodology, project administration, resources, supervision, and writing—original draft, review, and editing.