Differential expression and homotypic enrichment of a classic Cadherin directs tissue-level contractile asymmetry during neural tube closure.

Summary

Embryos pattern force generation at tissue boundaries, but how they do so remains poorly understood. Here we show how tissue-specific expression of the type II cadherin, Cadherin2 (hereafter Cad2), patterns actomyosin contractility along tissue boundaries to control zippering and neural tube closure in the basal chordate, Ciona robusta. Cad2 is differentially expressed and homotypically enriched in neural cells along the Neural/Epidermal (Ne/Epi) boundary, where RhoA and Myosin are activated during zipper progression. Homotypically enriched Cad2 sequesters the Rho GTPase activating protein, Gap21/23, to homotypic junctions, and in turn, Gap21/23 redirects RhoA/Myosin activity to heterotypic Ne/Epi junctions. By activating Myosin II along Ne/Epi junctions ahead of zipper and inhibiting Myosin II along newly-formed Ne/Ne junctions behind zipper, Cad2 promotes tissue level contractile asymmetry to drive zipper progression. We propose that dynamic coupling of junction exchange to local changes in contractility may control fusion and separation of epithelia in many other contexts.

Introduction

Actomyosin contractility is a key force generator for cell movement and shape change in embryonic development (reviewed in Dahmann et al., 2011; Fagotto, 2015; Heisenberg and Bellaiche, 2013; Lecuit et al., 2011). Dynamic control of contractility underlies many different tissue-level morphogenetic behaviors, including invagination, elongation, and dynamic maintenance of tissue boundaries (reviewed in Fagotto, 2015; Heisenberg and Bellaiche, 2013), but the underlying mechanisms remain poorly understood.

One of the major ways in which embryos pattern contractility is by patterning local activation of myosin II. Recent work reveals two general modes of tissue-level control over myosin activity. The first involves local control of myosin II with respect to intrinsic apico-basal or planar polarity within a particular tissue or compartment. For example, apical localization and/or activation of factors such as Fog, T48 and Shroom family members can promote local activation of Myosin II to drive tissue bending/invagination (reviewed in (Gilmour et al., 2017; Heisenberg and Bellaiche, 2013; Lecuit et al., 2011; Martin and Goldstein, 2014)). Alternatively, polarized activation of Myosin II by core members of the Planar Cell Polarity (PCP) pathway can drive planar-polarized junction contraction, leading to cell-cell intercalation and tissue elongation (Nishimura et al., 2012, reviewed in (Harris, 2018; Heisenberg and Bellaiche, 2013; Shindo, 2018; Walck-Shannon and Hardin, 2014)).

A second mode of control, studied mainly in Drosophila, involves the differential expression of signaling or cell-adhesion molecules across tissue or compartment boundaries. For example, differential expression of the nectin-like homophilic binding protein Echinoid induces actomyosin cables at tissue boundaries during dorsal appendage formation in egg chambers and dorsal closure in embryos (Laplante et al., 2006; Laplante and Nilson, 2011). Similarly, local expression of the homophilic binding protein Crumbs within the salivary gland placode promotes the assembly of an actomyosin cable at the placode boundary to help drive invagination (Röper, 2012). Differential expression of Toll family receptors in stripes along the AP axis bias myosin activity to mediolaterally oriented junctions to promote cell intercalation and germband elongation (Paré et al., 2014). Finally, differential expression of signaling molecules such as Wingless, Hedgehog, Dpp and Notch across compartment boundaries in embryos and imaginal discs, activates Myosin II to prevent cell mixing across those boundaries (reviewed in (Dahmann et al., 2011)). Recent studies in chordates highlight a key role for differential expression of Eph/Ephrin signaling in localizing activation of myosin II at tissue boundaries ((Fagotto et al., 2013), reviewed in (Fagotto, 2015)). Potential roles for other molecules in controlling contractility at tissue boundaries in chordates remain poorly defined (reviewed in Fagotto, 2015). More generally, how differential expression of signaling molecules across tissue boundaries promotes polarized activation of myosin II in single cells remains poorly understood.

Neural tube closure in ascidians offers a powerful opportunity to study the dynamic control of Myosin II along tissue boundaries during epithelial morphogenesis. The ascidian neural tube comprises an axial nerve cord and a central vesicle, corresponding to the spinal cord and brain respectively, of higher vertebrates (reviewed in Lemaire et al., 2002; Schoenwolf and Smith, 1990; Wallingford et al., 2013; Yamaguchi and Miura, 2012); light blue and dark blue in Figure 1A). As in many vertebrates, the ascidian neural tube forms in three steps: (1) the neural plate invaginates, raising neural folds along its lateral boundary with the epidermis (the Ne/Epi boundary; Figure 1A); (2) The neural folds move towards and meet at dorsal midline (Navarrete and Michael Levine, 2016) and (3) the neural folds fuse, through local rearrangements of apical junctions (Ne/Epi -> Ne/Ne plus Epi/Epi), to separate a closed neural tube from a continuous overlying epidermis (Hashimoto et al., 2015; Nicol and Meinertzhagen, 1988a; 1988b; Ogura and Sasakura, 2016). The final fusion step proceeds directionally from posterior to anterior, and has thus been referred to as zippering (Hashimoto et al., 2015; Nicol and Meinertzhagen, 1988b; 1988a); Figure 1A.

Figure 1. Dynamic accumulation of myosin along the Ne/Epi boundary during zippering.

(A) Schematic overview of neural tube closure. Top row shows dorsal view and bottom row shows cross-sectional view of embryos at 5.75, 7, 8 and 9 hours after fertilization. Dashed line in top row indicates the axial positions of cross sections. Anterior is up in this and all subsequent figures. (B) Top: heterotypic Ne/Epi boundaries (orange) ahead of the zipper; homotypic Ne/Ne and Epi/Epi junctions (green) behind the zipper. Bottom: distribution of myosin activity (magenta) measured previously by antibody staining. (C-F) Spatiotemporal patterns of myosin accumulation in neural and epidermal cells. (C, D) Frames from Movies S1 showing distribution of iMyo::GFP when expressed under the control of epidermal-specific and neural-specific promoters. Note that expression is mosaic and schematics at left indicate the actual domains of expression (see also Figure S1C and S1D). Magenta arrowheads indicate zipper position. White arrowheads indicate increased myosin along the entire Ne/Epi boundary (filled arrows) and just ahead of the zipper (open arrows). Scale bars, 10 μm. (E, F) Measurements of average iMyo::GFP polarity in epidermal (E) and neural (F) cells as a function of cell position relative to zipper, indicated by color overlays in the first movie frames in C and D. Box plots show the distribution of average polarity values measured for individual junctions during the period in which an Ne/Epi junction next to the zipper (z + 1) starts and finishes its contraction. Horizontal line indicates median, + indicates mean, boxes indicate second and third quartiles, and whiskers indicate 95% confidence interval. (n = 23 junctions, from 8 embryos in E and n = 12 junctions, from 6 embryos in F; see STAR Methods for details).

Previously, we described dynamic patterns of myosin activity that underlie zippering and neural tube closure (Hashimoto et al., 2015). Myosin activity is high on Ne/Epi junctions ahead of the zipper, and low on Ne/Ne and Epi/Epi junctions behind the zipper (Figure 1A and 1B). In addition, strong activation of Myosin II on Ne/Epi junctions just ahead of the zipper drives a posterior to anterior sequence of rapid junction contractions (Figure 1B). Computer simulations suggested that tissue-level asymmetry in myosin activity (high ahead of the zipper and low behind the zipper) is essential to ensure that sequential contractions are converted efficiently into forward movement of the zipper (Hashimoto et al., 2015).

Here, we identify a molecular basis for this tissue-level asymmetry. We identify a classical Cadherin2 (hereafter Cad2, Figure S1A) that is expressed specifically in neural cells, and enriched at homotypic Ne/Ne junctions. Homotypically enriched Cad2 sequesters the putative Rho GAP, Gap21/23, to inhibit RhoA/Myosin II at Ne/Ne junctions, and redirect RhoA/Myosin II activity towards Ne/Epi junctions. Dynamic coupling of junction exchange to changes in RhoA/Myosin activity sustains the tissue-level contractile asymmetry that is required for zippering and neural tube closure.

Results

Myosin accumulates along the Ne/Epi boundary in both Neural and Epidermal cells

We previously showed that Myosin II accumulates along the entire Ne/Epi boundary during zippering (Hashimoto et al., 2015), but we did not determine in which cells (Ne or Epi or both) this accumulation occurs. To address this, we expressed a GFP-tagged intrabody (“iMyo-GFP”), that recognizes the non-muscle Myosin II heavy chain (Hashimoto et al., 2015; Nizak et al., 2003; Vielemeyer et al., 2010) under the control of either the epidermal-specific promoter pEpi1 ((Sasakura et al., 2009), Figure S1A and S1B) or an enhanced form of the neural-specific promoter pEtr1* ((Veeman et al., 2010) Figure S1A-S1C, STAR Methods). During zippering in both epidermal and neural cells, iMyo-GFP was enriched along the entire Ne/Epi boundary (filled white arrowheads in Figure 1C and 1D; Movie S1), and highly enriched on rapidly contracting junctions just ahead of the advancing zipper (white open arrowheads in Figure 1C and 1D; quantified in Figure 1E, 1F, S1D and S1E). Thus, myosin activity is patterned in both neural and epidermal cells during zipper progression.

Cad2 is differentially expressed and homotypically enriched in midline neural cells.

To identify mechanisms that polarize myosin activity along Ne/Epi boundary, we searched for genes encoding putative transmembrane proteins that are differentially expressed across the Ne/Epi boundary during zippering. We focused on the midline cells – two rows of neural and epidermal cells that flank the Ne/Epi boundary. The midline cells arise from a single pair of bilaterally symmetric founder cells called b6.5. The b6.5 cells cleave at the 64-cell stage to produce daughters b7.9 and b7.10 (Figure 2Ai), which divide unequally during gastrulation to produce pairs of neural (b8.17 and b8.19) and epidermal (b8.18 and b8.20) precursor cells (Figure 2Aii). The neural and epidermal precursor cells divide once and twice more, respectively, along the AP axis, to produce two rows of neural and epidermal midline cells (Figure 2Aiii).

Figure 2. Stage and lineage-specific expression of Cad2.

(A) Lineage tree for midline nerve cord and epidermal cells. Midline Ne/Epi precursors (purple) divide at ~4 hours after fertilization to produce founders of the midline neural (light blue) and epidermal (dark blue) lineages. Magenta fill indicates cells expressing Cad2 gene. Cell positions (top) and distributions of Cad2 mRNA (bottom) at the indicated stages. Orange bars show the orientation of a previous cleavage. Midline neural/epidermal precursors and their progeny are outlined in color, as in (A). Scale bars, 25 μm. (B) Zoomed view of neural cells ahead of the zipper expressing Cad2::GFP under the control of a neural-specific Etr1* promoter. Embryos were fixed and counter-stained with phalloidin. Orange arrowheads indicate the Ne/Epi boundary. Scale bars, 10 μm.

We searched the ascidian database ANISEED for genes that are expressed before the onset of zippering in the descendents of either b8.17 and b8.19, or b8.18 and b8.20 (Tassy et al., 2010). This search identified Cadherin2 (Cad2, also known as Cadherin.b, Figure S1A), one of two classical cadherins in the Ciona robusta genome (Sasakura et al., 2003). Cad2 encodes a Type II Cadherin most closely related to Human VE-Cadherin (CHD5) (Hulpiau and van Roy, 2009) and Cadherin-8 (CHD8) (Sasakura et al., 2003), which was previously shown to be expressed in the neural primordium (Noda and Satoh, 2008). To better characterize the spatiotemporal pattern of Cad2 expression, we performed simultaneous in situ hybridization and nuclear staining. Before gastrulation, Cad2 is expressed throughout the vegetal hemisphere in most neural precursor cells, but not in the neural midline precursors b7.9 and b7.10 ((Noda and Satoh, 2008); Figure 2Ai, ii). Cad2 is first expressed in neural midline cells b9.33, b9.34, b9.37 and b9.38 just before initiation of zippering (Figure 2Aiii), suggesting a specific role in neural tube closure. Because commercially available antibodies do not recognize endogenous Cad2, we used the neural-specific promoter pETR* to express a GFP-tagged form of Cad2 in the neural primordium, and examine its subcellular localization. Ahead of the zipper, Cad2::GFP was enriched at homotypic junctions between Cad2::GFP-expressing neural cells, and absent from heterotypic junctions between Cad2::GFP-expressing neural cells and non-expressing epidermal cells (Figure 2B). Cad2::GFP was also enriched at homotypic junctions between Cad2::GFP-expressing neural cells behind the zipper (Figure S2A), (Figure 2B; quantified in Figure S2A). We refer to this pattern of enrichment as homotypic enrichment. Using an antibody against β-catenin (Kawai et al., 2007), which binds stochiometrically to classical cadherins, we found that endogenous β-catenin was strongly reduced along Ne/Epi contacts, relative to Ne/Ne contacts, consistent with homotypic enrichment of endogenous Cad2 (Figure S2B).

Differential expression and homotypic enrichment of Cad2 directs myosin activity to expression boundaries.

We hypothesized that differential expression and homotypic enrichment of Cad2 directs myosin activity to heterotypic Ne/Epi junctions (Figure 3A). If so, then equalizing Cad2 expression across the Ne/Epi boundary should abolish homotypic enrichment and reduce or abolish myosin activity, while creating ectopic Cad2 expression boundaries in neural or epidermal territory should induce homotypic enrichment of Cad2 and activate myosin at those boundaries. We tested these predictions in two ways. First, we used a midline-specific promoter, pMsx (Roure et al., 2014), to express Cad2::GFP in all midline cells (Figure S1B), forcing similarly high levels of Cad2 expression across the Ne/Epi boundary and creating an ectopic expression boundary in epidermal territory (Figure 3B). Second, we used lineage-specific injection of morpholinos to inhibit Cad2 expression exclusively in midline neural cells, forcing similarly low levels of Cad2 expression across the Ne/Epi boundary, and creating an ectopic expression boundary between midline and non-midline neural cells (Figure 3C).

Figure 3. Myosin is activated at Cad2 expression boundaries.

(A) Schematic view of a wild type embryo near the onset of zippering indicating the Ne/Epi boundary (orange line) and endogenous Cad2 expression pattern (blue). Expanded view shows wild type distributions of Cad2 and myosin activity in neural cells. (B) Cad2::GFP (green) is over-expressed in all midline cells on one side of the embryo. (C) Co-injection of Morpholino Oligonucleotides (MO) against Cad2 with an expression marker (pEtr1*::Lifeact::mNeon Green (mNG)) into a single b4.2 blastomere at the 8-cell stage. Yellow indicates b4.2 descendants; green indicates neural b4.2 descendants that received the MO. (D-F) Effects of ectopic Cad2::GFP expression. (D) Top panel shows schematic dorsal view of a neurula over-expressing Cad2::GFP (green) over endogenous Cad2 (blue) in midline cells. Orange line: original Ne/Epi boundary. Blue line: boundary between Cad2 expressing and non-expressing epidermal cells. Bottom three panels show localization of Cad2::GFP, 1P-myosin, and their superposition, respectively, in the same embryo. Open and filled arrowheads: corresponding boundaries on Cad2::GFP expressing and non-expressing sides of the embryo. Dashed and solid lines indicate the boundary sections measured for roughness, as shown at the right. Scale bars, 10 μm. (E-F) Box plots showing ratios of (E) 1P myosin and (F) ratio of boundary roughness on Cad2::GFP or Lifeact::mNG expressing:non-expressing sides of the embryo for the indicated boundary types. (G-I) Effects of Cad2 knockdown in midline neural cells. (G) Top panel: schematic dorsal surface view of a neurula stage embryo, injected as shown in (C), showing neural cells inheriting co-injected morpholino and mNG marker (green) over endogenous Cad2 (blue). Orange lines: original Ne/Epi boundary. Blue lines: ectopic boundary between midline and non-midline neural cells. Second panel: localization of 1P-myosin in the same embryo. Open and filled arrowheads indicate corresponding boundaries on injected and non-injected sides of the embryo. Panels at the left show the left side of the embryo at the same magnification, but rotated clockwise by ~10° to display more clearly the boundary between MO-injected and control Ne cells. Third panel: superposition of mNG marker and 1P-myosin. Dashed and solid lines indicate the boundary sections measured for roughness, as shown below. Scale bars, 10 μm (H-I) Box plots showing ratios of (H) 1P myosin and (I) boundary roughness on MO-injected:non-injected sides of the embryo for the indicated junction or boundary types. (J-M) Measurements of zippering defects in Cad2::GFP over-expressing or MO-injected embryos. (J) Schematic showing measurement of “un-zippered” length L. (K-L) Representative phalloidin-stained embryos showing effects of Cad2::GFP over-expression (K) or Cad2 MO injection (L). Scale bars, 10 μm. (M) Measurements of “unzippered length” L. (n > 24 junctions, from > 10 embryos for each of the conditions in E and H, n > 9 embryos for each condition in F, I and H) *** p < 0.005, **** p < 0.0005, Student’s t test.

Overexpression of Cad2 in midline cells inhibits myosin activation on Ne/Epi junctions, zippering and neural tube closure.

When over-expressed in all midline cells, Cad2::GFP was enriched on junctions between midline neural and epidermal cells (filled orange arrowheads in Figure 3D), and absent from junctions between expressing and non-expressing epidermal cells (filled blue arrowheads in Figure 3D). Thus over-expressing Cad2 in all midline cells abolishes homotypic enrichment in midline neural cells and induces homotypic enrichment in midline epidermal cells.

To determine how the loss or gain of homotypic Cad2 enrichment affects myosin activity we examined the distribution of Ser19-phosphorylated Myosin II (1P Myosin) in fixed immunostained embryos over-expressing Cad2::GFP in the midline cells. Mosaic expression of the electroporated transgenes (Zeller et al., 2006) allowed us to compare cells overexpressing Cad2::GFP with paired controls on the opposite side of the same embryo (Figure 3B and D). Consistent with our predictions, 1P Myosin was sharply reduced on junctions between neural and epidermal cells over-expressing Cad2::GFP, and increased on junctions between Cad2::GFP-expressing and non-expressing epidermal cells, relative to paired controls (Figure 3D; quantified in 3E). Thus, abolishing homotypic enrichment of Cad2 in midline neural cells reduces myosin activity along Ne/Epi contacts, and inducing homotypic enrichment of Cad2::GFP in epidermal cells promotes activation of myosin II along heterotypic contacts.

Boundary roughness (defined as the total length of a boundary segment divided by the distance between its endpoints; see bottom panel in Figure 3D) correlates inversely with boundary tension in other contexts (Aliee et al., 2012; Landsberg et al., 2009). In embryos over-expressing Cad2::GFP in midline cells, we observed increased boundary roughness (indicating reduced tension) along the Ne/Epi boundary where 1P myosin is reduced, and decreased boundary roughness (indicating increased tension) along boundaries between Cad2 expressing and non-expressing epidermal cells, where 1P myosin is increased (bottom right panel in Figure 3D; quantified in Figure 3F). Significantly, the apparent reduction in boundary tension along the Ne/Epi boundary was accompanied by a complete loss of zipper progression and neural tube closure (Figure 3J, 3K and 3M). Thus equalizing high levels of Cad2 expression across the Ne/Epi boundary strongly reduces myosin activation and tension production along Ne/Epi boundaries, and blocks zipper progression and neural tube closure.

Inhibition of Cad2 in midline neural cells inhibits myosin activation on Ne/Epi junctions, zippering and neural tube closure.

In complementary experiments, we injected antisense morpholino oligonucleotides directed against Cad2 (Cad2 MOs) at the 8-cell stage into single b4.2 blastomeres, whose only neural progeny are the midline neural cells (Figure 3C). Consistent with our predictions, selective inhibition of Cad2 expression in midline neural cells caused a sharp decrease in myosin activity and a significant increase in boundary roughness along Ne/Epi boundaries, relative to un-injected sides of the same embryos (Figure 3G and Movie S2, quantified in Figure 3H and 3I). In contrast, myosin activity was significantly increased, and boundary roughness was significantly reduced, along Ne/Ne boundaries between MO-receiving and non-receiving neural cells (Figure 3G and Movie S2; quantified in Figure 3H and 3I), suggesting that creating an ectopic Cad2 expression boundary within neural territory is sufficient to induce myosin activation and increase tension along that boundary. Importantly, zipper progression and neural tube closure were strongly reduced in Cad2 MO-injected embryos (Figure 3J, 3L and 3M), and the residual zippering was associated with straightening and shortening of the Ne/Ne boundary between normal and Cad2 MO-injected neural cells, suggesting that it is due to active shortening of the ectopic boundary, rather than to a residual contribution of the normal Ne/Epi boundary (Movie S3). We observed similar effects with two different morpholinos (MO1 and MO2), suggesting that these effects are specific to Cad2 inhibition (Figure S2C-S2E). Thus inhibiting Cad2 expression in midline neural cells is also sufficient to block Myosin activation along the Ne/Epi boundary, zipper progression and neural tube closure.

Differential expression of Cad2 in anterior neural cells directs myosin activation and purse string closure of the anterior neural tube.

Following zippering and closure of the posterior nerve tube, the anterior neural tube closes to form the sensory vesicle. We found that anterior neural tube closes through a combination of “zippering” and “purse string” mechanisms, in which activated myosin accumulates along the entire anterior Ne/Epi boundary, and all Ne/Epi junctions shorten simultaneously (Figure S3A, S3B and Movie S4). Cad2 is expressed in the anterior neural plate before the onset of anterior closure (Figure S3C), suggesting that differential expression of Cad2 across the anterior Ne/Epi boundary controls myosin activation and purse-string closure of the sensory vesicle. To test this, we injected Cad2 MO into the a4.2 blastomere, whose neural progeny are the sensory vesicle precursor cells (Figure S3D). To mark the affected neural cells, we coinjected a marker for neural expression (pZicL::Lifeact::GFP; Figure S1A and S1B). Injection of Cad2 MO into a4.2 blastomeres sharply decreased myosin activity along the anterior Ne/Epi boundary, relative to the un-injected side of the same embryos (Figure S3E, quantified in Figure S3F), and completely abolished the formation and shortening of a contractile purse string on the injected side. In contrast, myosin activity was significantly increased along the boundary between MO-injected and un-injected anterior neural cells (Figure S3E, quantified in Figure S3F). Thus, differential expression of Cad2 across the Ne/Epi boundary directs myosin activation and boundary contraction during multiple distinct steps in neural tube closure.

Differential expression and homotypic enrichment of Cad2 directs RhoA activity to the Ne/Epi boundary

We previously reported that the RhoA/Rho kinase signaling is required for myosin activation along the Ne/Epi boundary during neural tube closure (Figure 4A) (Hashimoto et al., 2015). To test whether RhoA activity is spatially patterned in midline neural cells, we constructed the ascidian analogue of a previously validated biosensor for active RhoA by fusing GFP to the RhoA binding domain of Ciona Anillin (henceforth GFP::AHPH; (Munjal et al., 2015; Piekny and Glotzer, 2008; Priya et al., 2015; Tse et al., 2012)). When expressed in neural cells using the ETR1* promoter, GFP::AHPH was enriched along all Ne/Epi junctions ahead of the zipper, with highest accumulation just ahead of the zipper, as previously observed for myosin (Figure 4B and Movie S5, (Hashimoto et al., 2015)). Significantly, the GFP::AHPH signal decayed rapidly just behind the zipper, where heterotypic Ne/Epi contacts are replaced by homotypic Ne/Ne contacts, suggesting that differential expression and homotypic enrichment of Cad2 dynamically controls the local activation of RhoA (Figure 4B and Movie S5).

Figure 4. Differential expression of Cad2 directs polarized activation of RhoA in midline neural cells.

(A) RhoA/Rho kinase signaling pathway for myosin activation. (B) A sequence of images (from Movie S5) showing the dynamic distribution of a RhoA biosensor GFP::AHPH (magenta) with (top) and without (bottom) a membrane marker (FM4-64; blue). Yellow arrowheads indicate zipper position. Right: Measurements of average GFP::AHPH polarity in neural cells as a function of cell position relative to zipper, indicated by color overlays in the first movie frames in left schematic. (n = 12 junctions, from 4 embryos) (C, D) Effect of Cad2 MO on distribution of GFP::AHPH in neural cells. (E, F) Effects of ectopic expression of Cad2 in all midline cells on distribution of GFP::AHPH in midline neural cells. (G, H) Effects of ectopic expression of Cad2::mLumin in all midline cells on distribution of GFP::AHPH in midline epidermal cells. Green: cells expressing GFP::AHPH, yellow: cells receiving Cad2 MO, blue: distributions of Cad2 protein. In (D, F), orange lines and arrowheads indicate midline Ne/Epi junctions. Magenta lines and arrowheads indicate midline Ne/Ne junctions. In (H), orange lines and arrowheads indicate midline Ne/Epi junctions, blue lines and arrowheads indicate junctions along the ectopic Cad2::mLumin expression boundary between midline and non-midline epidermal cells, magenta lines and arrowheads indicate junctions between midline epidermal cells. In (D, F, H) graphs show GFP::AHPH polarity measured as ratio of intensities on junction types indicated below each graph. (n > 23 junctions, from > 9 embryos for each of the conditions in D, F and H) **** p < 0.0005, by Student’s t test. Scale bars, 10 μm.

To test this possibility, we co-injected Cad2 MO with DNA encoding neural expression of the Rho biosensor (pEtr1*::GFP::AHPH) into single b4.2 cells at the 8-cell stage (Figure 4C). Because we could not directly compare GFP::AHPH levels in injected and control sides of the same embryos, we quantified GFP::AHPH polarity, defined as the mean GFP::AHPH intensity along Ne/Epi junctions divided by half the mean intensity along Ne/Ne junctions (Ne/Epi::Ne/Ne intensity ratio; see STAR methods for details). GFP::AHPH polarity was completely abolished in midline neural cells receiving Cad2 MO, relative to embryos injected with a control morpholino (Figure 4D, Movie S6). The loss of GFP::AHPH signal polarity was not due to absence of GFP::AHPH expression, because GFP::AHPH was similarly enriched at cleavage furrows in both control and Cad2 MO injected embryos (Figure S4, Movie S6).

In complementary experiments, over-expressing Cad2::mLumin (variant mKate: (Chu et al., 2009)) in all midline cells (Figure 4E and G) abolished GFP::AHPH polarity in neural midline cells (Figure 4F), and induced enrichment of GFP::AHPH on junctions between Cad2 expressing and non-expressing epidermal cells (Figure 4H), relative to embryos expressing a control construct. Thus differential expression and homotypic enrichment of Cad2 is required to polarize RhoA activity in midline neural cells, and is sufficient to promote local activation of RhoA along ectopic expression boundaries.

Cad2 directs polarized enrichment of Gap21/23 in midline neural cells

How does homotypic enrichment of Cad2 direct activation of RhoA and Myosin II to Ne/Epi boundaries? We hypothesized that Cad2 sequesters an inhibitor of RhoA to Ne/Ne junctions, preventing it from accumulating on Ne/Epi junctions. Accordingly, we screened the Ciona genome for putative Rho GTPase activating proteins (RhoGAPs), whose orthologues localize to cell-cell junctions in other organisms, and which are homotypically enriched when expressed in neural cells using the Etr1* promoter. Of the five RhoGAPs we tested (Gap5, Gap11A, Gap21/23, Gap22/24/25, Gap29; see Figure S5A and S5B), only Gap21/23 was homotypically enriched when expressed in neural cells. Gap21/23 is the Ciona orthologue of Human RhoGAPs ARHGAP-21 and ARHGAP-23 (Peck et al., 2002) (previously ARHGAP-10 (Bassères et al., 2002)) (Figure S5A). Using in situ hybridization, we determined that Gap21/23 mRNA is expressed in all embryonic cells until the 32-cell stage; it decays before the onset of gastrulation, and then it is expressed again, specifically in posterior midline epidermal and neural cells, just before the onset of zippering (Figure 5A and S5C). When expressed using the Etr1* promoter, GFP::Gap21/23 was enriched on Ne/Ne junctions, including newly-formed Ne/Ne junctions behind the zipper, and virtually absent from Ne/Epi junctions during zippering (Figure 5B and S5D), suggesting that Cad2 may pattern RhoA/myosin activity by localizing Gap21/23.

Figure 5. Sequestration by homotypically enriched Cad2 polarizes Gap21/23.

(A) Gap21/23 expression (magenta) at late gastrula and early neurula stages. Light and dark blue indicate midline neural and epidermal cells respectively. Scale bars, 20 μm. (B) Localization of GFP::Gap21/23 in neural cells. Embryo was fixed and co-stained with phalloidin. Orange lines and arrowheads indicate Ne/Epi boundary. Scale bars, 10 μm. (C-K) Distributions of GFP::Gap21/23 for different patterns of Cad2 expression: (C-E) Cad2::mLumin expressed in all midline cells; GFP::Gap21/23 expressed in midline neural cells; (F-H) Cad2 inhibited and GFP::Gap21/23 expressed in midline neural cells. (I-K) Cad2::mLumin and GFP::Gap21/23 expressed in all midline cells. (C, F, I) Orange lines and arrows indicate Ne/Epi junctions. Magenta lines and arrows indicate Ne/Ne (C-H) or Epi/Epi (I,K) junctions in midline cells. Blue lines and arrowheads indicate junctions between midline and non-midline epidermal cells. (D, G, J) GFP::Gap21/23 polarity measured as ratio of intensities on junction types indicated below each y-axis. (E, H, K) relative enrichment of GFP::Gap21/23, measured as the ratio of junctional to cytoplasmic intensity for junction types indicated below each graph. Scale bars, 10 μm. (n > 23 junctions, from > 9 embryos for each of the conditions in D, E, G, H, J and K) *** p < 0.005, **** p < 0.0005, , Student’s t test. (L) Schematic summary of Gap21/23 distributions for different Cad2 expression and localization patterns.

To test whether Cad2 controls Gap21/23 localization, we manipulated Cad2 expression patterns and measured changes in GFP::Gap21/23 polarity and junctional enrichment in midline neural cells (see STAR methods for details). Over-expressing Cad2 in all midline cells induced a complete loss of GFP::Gap21/23 polarity in midline neural cells (Figure 5C and 5D), caused by increased enrichment of GFP::Gap21/23 on both Ne/Epi and Ne/Ne junctions (Figure 5E). Similarly, inhibiting Cad2 expression in midline neural cells induced a complete loss GFP::Gap21/23 polarity (Figure 5F and 5G), but in this case, the loss of polarity was associated with decreased enrichment on Ne/Ne junctions and increased enrichment on Ne/Epi junctions (Figure 5H). In contrast, morpholino-based inhibition of Gap21/23 had no detectable affect on localization of Cad2 in midline neural cells (Figure S5E and S5F).

In complementary experiments, we asked if differential expression and homotypic enrichment of Cad2 (Figure 3D) can induce polarized enrichment of Gap21/23 in midline epidermal cells, which do not normally express Cad2 (Figure 5I). When co-expressed with a control protein (Lifeact::GFP) in midline epidermal cells, GFP::Gap21/23 was uniformly enriched on all Epi/Epi junctions (Figure 5I and 5K). In contrast, co-expression with Cad2 induced polarized enrichment of GFP::Gap21/23 in midline epidermal cells (Figure 5I and 5J), and again, this gain of polarity was associated with increased enrichment of GFP::Gap21/23 on junctions between Cad2 expressing epidermal cells, and decreased enrichment on junctions between Cad2 expressing and non-expressing epidermal cells (Figure 5K). In summary, we find that when Cad2 is absent, Gap21/23 can accumulate at uniformly low levels on all junctions (Figure 5L, left). Uniformly enriched Cad2 induces uniformly high levels of Gap21/23 accumulation (Figure 5L, middle), while homotypically enriched Cad2 induces enrichment of Gap21/23 on homotypic junctions and depletion of Gap21/23 at heterotypic junctions (Figure 5L, right). These data support a simple model in which a limiting pool of Gap21/23 can associate weakly with all cell contacts independently of Cad2, and homotypically enriched Cad2 sequesters Gap21/23 to Ne/Ne junctions, preventing its association with Ne/Epi junctions.

Gap21/23 is required for polarized activation of RhoA and myosin II, zippering, and neural tube closure.

Next, we asked whether Gap21/23 is required in midline neural cells for polarized activation of RhoA and myosin II, zippering and neural tube closure. To bypass a possible requirement for Gap21/23 during early development, we injected Gap21/23 MO into b4.2 cells at the 8-cell stage (Figure 6A). These embryos developed normally until the onset of zippering but then failed to undergo zippering and neural tube closure (Figure 6A and 6B). Polarized enrichment of GFP:AHPH and 1P myosin was completely lost in midline neural cells that received Gap21/23 MO (confirmed with a second MO; Figure S6K), but not in cells that received a control MO (Figure 6C-F, Movie S7). Although we could not measure changes in RhoA activity relative to internal controls, the loss of polarized myosin activity was accompanied by increased 1P myosin on Ne/Ne junctions, decreased 1P myosin on Ne/Epi junctions, and increased roughness (indicating reduced tension) along the Ne/Epi boundary, relative to the un-injected sides of the same embryos (Figure 6G and 6H). These results suggest that homotypically enriched Gap21/23 displaces RhoA/Myosin activity away from Ne/Ne junctions and towards Ne/Epi junctions.

Figure 6. Gap21/23 is required for polarized activation of RhoA/myosin and zipper progression.

(A, B) Inhibition of zipper progression in Gap21/23 knockdown embryos, measured as in Figure 3. (A) Representative phalloidin-stained embryos embryos co-injected with control or Gap21/23 MO and neural-specific co-injection marker (pEtr1*::Lifeact::mNG). Scale bars, 10 μm. (B) Measurement of un-zippered length L in control and Gap21/3 MO injected embryos (n > 17 embryos for each condition). (C, D) Effects of Gap21/23 MO on distribution of GFP::AHPH in midline neural cells. Top schematic: yellow indicates cells receiving Gap21/23 MO; green indicates cells expressing pEtr1*::GFP::AHPH. Bottom panels: Orange lines and arrowheads indicate Ne/Epi junctions. Magenta lines and arrowheads indicate Ne/Ne junctions in midline neural cells. Scale bars, 10 μm. (D) GFP::AHPH polarity measured as the ratio of intensities on Ne/Epi::Ne/Ne junctions. (E, F, H) Effects of Gap21/23 knockdown on distribution of 1P-myosin in midline neural cells. (E) Top panel shows dorsal view of a neurula stage embryo, injected as shown in (A), then fixed and immunostained for 1P myosin. Color fill indicates neural cells inheriting co-injected morpholino and mNG marker (green) over endogenous Cad2 (blue). Orange lines and arrowheads indicates the original Ne/Epi boundary. Magenta lines and arrowheads indicate Ne/Ne junctions in midline neural cells. Second panel: superposition of mNG marker and 1P-myosin in the same embryo. Third panel: 1P-myosin alone. Left and right views show injected and control sides of the same embryo rotated clockwise or counter-clockwise by ~10°. Open and filled arrowheads indicate corresponding boundaries on injected and non-injected sides of the embryo. Scale bars, 10 μm. (G) Roughness ratio in Gap21/23 knockdown embryos. (n > 9 embryos for each condition) (H) 1P myosin ratio in Gap21/23 knockdown embryos (n > 14 junctions from > 9 embryos for each condition in D, F, and H). **** p < 0.0005, Student’s t test.

To assess further if Gap21/23 acts by locally inhibiting RhoA activity, we overexpressed Gap21/23 in midline neural cells. Embryos over-expressing GFP::Gap21/23 in neural cells developed normally to neurula stage, but then failed to undergo normal zippering (Figure S6A and S6B). GFP::Gap21/23 remained homotypically enriched, but with detectable accumulation on Ne/Epi junctions (Figure S6D). This ectopic accumulation was accompanied by complete loss of RhoA and myosin II polarity (Figure S6C-S6H), decreased enrichment of 1P myosin on both Ne/Epi and Ne/Ne junctions (Figure S6H), and increased roughness (indicating reduced tension) of the Ne/Epi boundary, relative to the unaffected sides of the same embryos (Figure S6I), These results suggest that Gap21/23 can act locally at individual junctions to inhibit RhoA and myosin activity.

To test if GAP activity is required for local inhibition of RhoA by Gap21/23, we mutated a conserved catalytic arginine residue (R1231) that is essential for the activity of other RhoGAPs (Figure S6J (Anderson et al., 2008; Dubois et al., 2005; Sousa et al., 2005)). When overexpressed in midline cells, Gap21/23^R1231A reduced RhoA and myosin polarity in midline neural cells (Figure S6D, S6E, S6H-I), and reduced zipper progression (Figure S6B), but these effects were notably weaker than those produced by over-expressing normal Cap21/23, suggesting that GAP activity contributes to GAP21/23’s role in polarizing RhoA and myosin during zippering, although other mechanisms may also be involved.

Cdc42 is thought to be the primary inhibitory target of Gap21/23 orthologues in other organisms (Dubois et al., 2005; Sousa et al., 2005), suggesting that Gap21/23 could act indirectly through Cdc42 to polarize RhoA activity. However, a biosensor for active Cdc42, obtained by fusing GFP to the Cdc42 binding domain of Ciona WASP, and validated by co-expression with a dominant negative inhibitor of Cdc42 (Figure S6L), showed the same pattern of homotypic enrichment as Gap21/23 in midline neural cells (Figure S6M). Thus Cdc42 does not appear to be an inhibitory target of Gap21/23 in these cells and is unlikely to mediate local inhibition of RhoA/Myosin II.

Discussion

Embryos use dynamic control of actomyosin contractility to maintain tissue boundaries and drive tissue morphogenesis (reviewed in (Dahmann et al., 2011; Fagotto, 2015; Heisenberg and Bellaiche, 2013; Lecuit et al., 2011)), but how they implement this control in single cells, and how it is coordinated with local tissue remodeling, remain poorly understood. Here we have described how tissue-specific expression of Cad2 in neural cells patterns RhoA/myosin II activity to create a dynamic tissue-level contractile asymmetry that is essential for zippering and neural tube closure. Our data support a model in which: (a) Differential expression of Cad2 drives its homotypic enrichment in midline neural cells, (b) homotypically enriched Cad2 sequesters Gap21/23 to neural cell contacts and away from the Ne/Epi boundary, and (c) polarized Gap21/23 redirects RhoA/Myosin II activity away from homotypic junctions and towards heterotypic junctions.

Differential expression directs homotypic enrichment of Cad2 in midline neural cells.

Previous cell-mixing experiments have documented a tendency for cadherins (including VE-Cadherin, the human orthologue of Cad2) to become enriched at boundaries between expressing cells, but not between expressing and non-expressing cells (Hirano et al., 1987; Klompstra et al., 2015; Navarro et al., 1998). Our data suggest that similar rules operate for Cad2 within an intact embryonic epithelium; differential expression of Cad2 dictates homotypic enrichment in neural or epidermal cells, independent of cell identity. The underlying mechanism(s) remain unclear, but likely involve multiple effects downstream of trans-homophillic engagement, including local stabilization of clustered Cadherins at cell-cell contacts (Cavey et al., 2008; Foote et al., 2013; Nose et al., 1988), and local inhibition of endocytosis (Izumi et al., 2004).

Gap21/23 is the key intermediary between homotypically enriched Cad2 and polarized activation of RhoA/Myosin II.

Three observations identify Gap21/23 as the key intermediary between homotypically enriched Cad2 and polarized activation of RhoA/Myosin II. First, Gap21/23 is expressed in neural cells from just before the onset of zippering. Second, homotypically enriched Cad2 dictates polarized enrichment of Gap21/23 in midline neural cells. Third, inhibiting Gap21/23 expression in midline neural cells prevents polarized activation of RhoA and Myosin II and blocks zippering and neural tube closure.

Our data support a model in which homotypically enriched Cad2 sequesters Gap21/23 to homotypic junctions. However, the molecular basis for sequestration of Gap21/23 by Cad2 remains unclear. E-cadherin acts through a-catenin and p120-catenin to recruit Gap21/23 orthologues to cell-cell contacts in early C. elegans embryos (Klompstra et al., 2015) and in vertebrate tissue culture (Sousa et al., 2005). The Cad2 orthologue VE-cadherin acts through p120-catenin to recruit other GAPs, such as p190 RhoGAP (Zebda et al., 2013). Thus an interesting possibility is that Cad2 uses a-catenin and p120-catenin to recruit Gap21/23 to Ne/Ne contacts in ascidian embryos.

How does Gap21/23 polarize RhoA and Myosin II activity in midline neural cells? In Gap21/23-depleted neural cells, RhoA/Myosin II activity is uniformly high on all junctions, suggesting that upstream activators are uniformly enriched in neural cells and that local inhibition of RhoA by Gap21/23 dictates its polarized activity. The finding that a GAP-deficient form of Gap21/23 is far less effective in blocking polarized activation of RhoA/Myosin II and zipper progression confirms that Gap21/23 acts at least in part by hydrolyzing GTP. Whether the residual effects of over-expressing a GAP-deficient form of Gap21/23 reflect additional GAP-independent roles for Gap21/23, or dominant-negative interference with endogenous Gap21/23’s interactions with other partners, such as Cad2, remains unclear.

ARHGAP21 can act as a GAP towards either RhoA or CDC42 in vitro, with preference for the latter (Dubois et al., 2005). Likewise, in cell culture or in vivo, ARHGAP21 and its orthologues inhibit either RhoA or CDC42, depending on cell type and subcellular localization (Barcellos et al., 2013; Klompstra et al., 2015; Lazarini et al., 2013; Marston et al., 2016; Zhang et al., 2016). Our data suggest that Cdc42 is not an inhibitory target of Gap21/23 in ascidian neural cells. Instead, we favor the simpler possibility that Gap21/23 inhibits RhoA directly, at least in part by stimulating GTP hydrolysis.

Importantly, Gap21/23 does not simply inhibit RhoA/Myosin II at Ne/Ne boundaries. Myosin II activity increases on Ne/Epi junctions when Gap21/23 is depleted. Thus polarized Gap21/23 appears to redirect Myosin II activity away from Ne/Ne junctions and towards the Ne/Epi boundary. The underlying mechanism remains unclear, but a simple hypothesis is that one or more factors required for Myosin II activation are limiting in midline neural cells such that inhibiting Myosin II activation on Ne/Ne contacts leads to increased activation at Ne/Epi contacts.

Although we have focused on the control of RhoA/myosin activity in midline neural cells, active RhoA and myosin are also enriched along the Ne/Epi boundary in midline epidermal cells (Figure 1C, E, and data not shown). The observation that ß-catenin is homotypically enriched in both midline neural and epidermal cells raises the intriguing possibility that another cadherin could operate in epidermal cells to polarize RhoA and Myosin II. An obvious candidate that we will explore in future studies is Ci-Cadherin – the Ciona orthologue of classical E-cadherin, which is maternally expressed in all cells of the early embryo (Noda and Satoh, 2008)

Comparison to other systems: Parallels and differences.

Tissue-specific Cadherin expression has long been thought to play a major role in patterning tissue morphogenesis (Takeichi, 1995), but the underlying mechanisms have remained poorly understood (Fagotto, 2015). Our study provides the first clear example in chordates of a mechanism by which tissue-specific expression of a classic cadherin localizes actomyosin contractility to a tissue boundary to control morphogenesis. Interestingly, variants of this mechanism have also been recently described in several invertebrate model systems. In early C. elegans embryos, an orthologue of E-cadherin recruits the orthologue of Gap21/23 (PAC-1) to all somatic cell contacts, where it inhibits CDC42, restricting CDC42 activity to non-contacting (apical) cell surfaces (Anderson et al., 2008; Klompstra et al., 2015). During gastrulation, endoderm-specific signal(s) couple apical CDC42 activity to the local recruitment of a CDC42 effector MRCK-1 to activate myosin II and drive endoderm cell ingression (Marston et al., 2016). In the developing Drosophila eye, differential expression and homotypic enrichment of N-cadherin directs myosin activity to heterotypic junctions between cone and primary pigment cells. In this system, N-cadherin may not act by sequestering an inhibitor of Myosin II to homotypic contacts. Instead, a small amount of unbound N-cadherin appears to promote contractility at heterotypic contacts, although the mechanism remains unknown (Chan et al., 2017).

Looking further afield to other homophilic binding proteins, it has been proposed that Drosophila Echinoid localizes myosin II to the epidermis/amnioserosa boundary during dorsal closure by sequestering PAR-3/Bazooka (Laplante and Nilson, 2011), although direct evidence for this mechanism is still lacking. During salivary gland invagination, the intracellular domain of Crumbs recruits atypical protein kinase C (aPKC) to homotypic contacts between Crumbs expressing cells, where it is thought to phosphorylate and inhibit Rho Kinase (Ishiuchi and Takeichi, 2012), restricting accumulation of Rho kinase and myosin activity to the boundary of the salivary gland rudiment (Röper, 2012). Thus dynamic sequestration of inhibitors by homotypically enriched transmembrane proteins may be a more general mechanism to localize contractility activity to tissue boundaries.

Implications for dynamic control of zipper progression in ascidians, and other chordates.

We previously showed that zipper progression involves two modes of control over myosin II (Hashimoto et al., 2015): Sequential activation of myosin II and rapid contraction of individual Ne/Epi junctions just ahead of the zipper provides the power stroke for zipper progression, while tissue level asymmetry in myosin activity and junction tension (high ahead of the zipper, and low behind the zipper) biases sequential contraction to favor net forward movement of the zipper (Hashimoto et al., 2015).

Our current findings explain how this tissue level force imbalance is dynamically maintained as the zipper progresses (Figure 7): Ahead of the zipper, sequestration of Gap21/23 and polarization of RhoA/Myosin II leads to local higher levels of myosin activity and tension along the Ne/Epi boundary. Behind the zipper, rapid engagement of Cad2 at newly formed Ne/Ne contacts leads to rapid recruitment of Gap21/23, local inhibition of RhoA/myosin II, and reduced tension along the Ne/Ne boundary. Thus dynamic coupling of RhoA/Myoasin activity to local remodeling of tissue contacts allows a tissue-level force imbalance to move with the advancing zipper.

Figure 7. Model for self-adjusting tissue level control of contractile asymmetry.

Ahead of the zipper, homotypically enriched Cad2 sequesters Gap21/23 to Ne/Ne junctions, directing RhoA/myosin II activity to Ne/Epi junctions. Behind the zipper, Cad2 recruits Gap21/23 to newly formed Ne/Ne junctions, inhibiting RhoA and myosin II at those junctions (see text for details).

Interestingly, we have found that in addition to abolishing tissue level contractile asymmetries, equalizing Cad2 expression across the Ne/Epi boundary (or inhibiting Gap21/23 in midline neural cells) also blocks strong activation of RhoA and myosin II just ahead of the zipper and completely blocks zipper progression. Thus local signaling by homotypically-enriched Cad2 is also required for the strong sequential activation of RhoA/myosin II just ahead of the zipper that provides the power stroke for zipper progression. The nature of this requirement, and how strong contractions are localized to a zone just ahead of the zipper, remains to be determined.

Finally, recent studies of zippering and neural tube closure in mouse embryos suggest that an actomyosin cable forms between epithelial and neuroectodermal cells, and contributes to forces that drive zippering (Galea et al., 2017). Thus it will be interesting to determine whether mechanisms similar to those we describe here also pattern myosin activity during neural tube closure in mouse embryos.

STAR Methods

Contact for Reagent and Resource Sharing

Lead contact information:

Edwin Munro, Department of Molecular Genetics and Cell Biology, University of Chicago, emunro@uchicago.edu

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact.

Experimental Model and Subject Details

Ciona robusta were used in this study. Ciona robusta adults were collected and shipped from Half Moon Bay, Oyster Point and San Diego (M-Rep, CA) and then maintained in oxygenated sea water at ~ 16°C. Fertilization, staging, dechorionation and electroporation were conducted as previously described (Bertrand et al., 2003; Corbo et al., 1997; Hotta et al., 2007). We cultured embryos in 5-cm plastic petri dishes coated with 1% agarose and filled with HEPES-buffered artificial seawater (ASWH) (Pasini et al., 2006).

Method Details

Tissue-specific expression of Fluorescent Protein (FP) fusions

To create Gateway destination vectors for tissue-specific expression of FP fusions, we first amplified GFP and mLumin (Chu et al., 2009), digested with Stu1, or EcoRV and BeglII together, and then inserted FP-encoding sequences into the 5’ or 3’ entry site of a standard Gateway pFOG::RfA cassette (Roure et al., 2007); FOG (Friend of GATA) is also known as Zfpm (Figure S1A), but we refer to it here as FOG, to create Gateway RfA cassettes: pFOG::RfA::GFP, pFOG::GFP::RfA pFOG::mLumin::RfA, and pFOG::RfA::mLumin. We amplified previously characterized promoter regions of Epi1, Etr1, Msx and ZicL from Ciona genomic DNA (Roure et al., 2014; Sasakura et al., 2009; Shi and Mike Levine, 2008; Veeman et al., 2010). We fused the 218 base pair basal promoter of pFOG (Rothbächer et al., 2007) to the C-terminus of the Epi1, Etr1, Msx and ZicL promoters by PCR. We digested the resulting fusions with Xho1 and Stu1, then used infusion (Takara) or HiFi DNA assembly (New England BioLab) kits to replace the full FOG promoter (Pasini et al., 2006) in pFOG::RfA::3xGFP (Hashimoto et al., 2015), pFOG::RfA::GFP, pFOG::GFP::RfA pFOG::mLumin::RfA, or pFOG::RfA::mLumin. Plasmid concentration for electroporation varied ~ 100 μg per plasmid.

We amplified full open reading frames of Cad2 and Gap21/23 from the unigene collection (Cogenics) clones VES99_P13 and VES100_J09 respectively. We amplified a fragment corresponding to the conserved AHPH domain of Ciona robusta Anillin from Cogenics clone VES88_P24. We amplified a fragment corresponding to the CDC42-interacting domain (amino acid residues 172–250, hereafter WASP-GBD (for GTPase-binding domain: (Kim et al., 2000))) of Ciona robusta WASP from Cogenics clone VES76_G22. We fused the sequence encoding Saccharomyces cerevisiae Lifeact: (Riedl et al., 2008) to the N terminus of GFP or mNeonGreen (mNG) or mLumin. We created Gateway entry clones containing Cad2, Gap21/23, Gap21/23^R1231A, Lifeact::GFP, Lifeact::mNG, Lifeact::mLumin, AHPH, and WASP-GBD, respectively, using pCR8/GW/TOPO TA Cloning Kits (Thermo Fisher) or HiFi DNA assembly kit (New England BioLab) or Q5® Site-Directed Mutagenesis Kit (New England BioLab). We then used the Gateway LR reaction (Thermo Fisher Scientific) to recombine these Gateway entry clones, or an entry clone for iMyo::GFP (Hashimoto et al., 2015), or entry clones containing a dominant negative form of Ciona intestinalis CDC42 ((Philips et al., 2003), hereafter CDC42 DN), into the tissue-specific destination vectors described above. Table S1 lists the primers that we used to amplify specific insert sequences from template clones or Ciona genomic DNA.

Whole-mount in situ hybridization

We performed detection of Cad2, Gap21/23 mRNAs by whole-mount in situ hybridization with digoxigenin (DIG)-labeled antisense RNA probes as previously described (Wada et al., 1995), using a DIG-RNA labeling kit (Sigma) to synthesis the probes.

Microinjection of plasmids, MOs and control experiments

We microinjected eggs as described previously (Bertrand et al., 2003), using 10-25 ng/μl for plasmid DNA, and 0.25–0.5 mM for morpholino oligonucleotides (MO). The sequences of translation-blocking MOs (Gene Tools, LLC) were as follows:

Cad2 MO1, 5’-ATCGTCTCCATGTTGTACTTTATGT-3’

Cad2 MO2, 5’-TTCGCAGTTTTAGTTTCACACTCTT-3’

Gap21/23 MO1, 5’-TGTGCCATTTCAAACACGTTGAAGC-3’

Gap21/23 MO2, 5’-ATTATCTCTTCACAAATCCATTCTT-3’

MO1s cover the starting methionine, while MO2s cover the 5’UTR and do not overlap MO1s. As control experiments with MO, we used a Morpholino Standard Control oligos offered from Gene Tools. As control experiments with overexpression of Cad2, GAP21/23 and Gap21/23^R1231A, Lifeact::mLumin was expressed.

Immunostaining and imaging of phosphomyosin.

We fixed and stained embryos with alexa 488 or 568 phalloidin (ThermoFisher) and a mouse polyclonal antibody raised against ser19-phosphorylated myosin regulatory light chain (1P myosin, Cell Signaling) as previously described (Hashimoto et al., 2015). A polyclonal antibody against β-catenin (Kawai et al., 2007) was a kind gift from Professor Hiroki Nishida of Osaka University, Japan. Whole mount immunostaining against β-catenin was carried out as follows. Embryo are fixed in 100 mM HEPES [pH 7.0], 100 mM EGTA, 10 mM MgSO4, 1% formaldehyde (Thermo Fisher Scientific), and 300 mM dextrose, for 2 hours on ice, washed with PBT (0.1% Triton) 3 times. Embryos were blocked in PBT 1% BSA overnight at 4°C, then placed in a 1/50 dilution of antibody in PBT/BSA overnight at 4°C, washed with PBT (3 times). Embryos were then incubated in a 1/1000 dilution of in secondary Cy3 Donkey Anti-Rabbit IgG (Jackson ImmunoResearch Laboratories) overnight at 4°C, washed with PBT (3 times). We collected Z-stacks of confocal images on a Zeiss LSM 880 confocal microscope with a 40×/1.3 oil-immersion objective at 0.5 μm intervals. We rendered 3D projections in ImageJ 3D Viewer (http://rsbweb.nih.gov/ij/plugins/3d-viewer/).

4D live time-lapse imaging

We performed 4D live time-lapse imaging as described previously (Hashimoto et al., 2015). Briefly, we labeled marker proteins by electroporating newly fertilized zygotes with ~50μg DNA. In some experiments, we labeled cell membranes using the fluorescent lipophilic dye FM4-64 (5μg/ml, Thermo Fisher Scientific, T13320). We settled embryos onto gelatin-formaldehyde coated glass bottom plates (TED PELLA) filled with seawater at ~18°C. We acquired images using a Nikon ECLIPSE-Ti inverted microscope equipped with 20x and 60x water-immersion lenses, solid-state 50 mW 481 and 561 laser excitation (Coherent), a Yokogawa CSU-X1 spinning disk scan head, and an Andor iXon3 897 EMCCD camera. We routinely used 20% of the laser power for imaging on both channels. We acquired Z stacks using an x-y motorized stage (for multiple-location imaging) and a fast piezoelectric Z-axis stepper motor with focus steps taken at 0.5 to 1 μm intervals.

Quantitative analysis of 1P Myosin and fluorescence protein intensities

We performed all image analysis and processing in ImageJ (https://imagej.nih.gov/ij/). To quantify junctional intensities of 1P myosin or FPs, we constructed maximum intensity projections of image stacks containing the apical surfaces of all relevant cells. We measured the average pixel intensity in three-pixel-wide lines drawn along a junction of interest, excluding regions very close to vertices. Then we subtracted the mean cytoplasmic intensity, measured in junction-adjacent regions, to estimate the average junction-specific intensity. For mosaically-expressed proteins, we scaled the average junctional intensity between two expressing cells by 0.5 to account for the double contribution. We quantified junctional enrichment as the ratio of average junctional intensity to the mean intensity of junction-adjacent cytoplasmic regions. For 1P myosin, we quantified changes in junctional intensities induced by experimental perturbations as the ratio I_exp/I_cont, where I_exp the average intensity along a junction on the perturbed side and I_cont is the average intensity of the corresponding junction on the contralateral (control) side of the same embryo. We quantified polarity of 1P myosin or FP-tagged proteins as I_AP/I_ML, where I_AP and I_ML are the average intensities for junctions oriented along the anterior-posterior and mediolateral axes, respectively. We measured junctional myosin intensity vs time during primary junction contractions and aligned these data across multiple embryos as described previously (Hashimoto et al, 2015).

Quantification and Statistical Analysis

In all box plots, the boxes indicate second and third quartiles, whiskers indicate 95% confidence, the horizontal line indicates median, and + indicates mean. Statistical methods were described in the Figure legends. Statistical analysis was performed in R (https://www.r-project.org). *p < 0.1, **p < 0.05, *** p < 0.005, **** p < 0.0005,

Supplementary Material

Movie S1

Movie S1 (related to Figure 1). Myosin II dynamics during zippering in live embryos expressing iMyo::GFP in epidermal cells on one side of the embryo (left) or in all midline neural cells (right). Each frame is the maximum intensity projection of 15 images collected at 0.75 μm intervals in Z near the apical surface; frames were collected at 60s intervals. The movie is displayed at 15 fps.

Movie S2

Movie S2 (related to Figure 3). 3D reconstruction of a neurula stage embryo in which Cad2 MO was injected into one b4.2 blastomere at the 8-cell stage. Orange arrowheads indicate Ne/Epi junctions on injected (filled orange arrowheads) and non-injected (open orange arrowheads) sides of the embryo. Blue arrowheads indicate junctions between injected and non-injected neural cells (filled blue arrowheads) and corresponding junctions on the non-injected control side (open blue arrowheads) sides of the embryo. See also Figure 3C.

Movie S3

Movie S3 (related to Figure 3). Time-lapse movies of embryos in which either a control (left) or Cad2 (right) morpholino was co-injected with pEtr1*::Lifeact::mNG into a single b4.2 blastomere at the 8-cell stage. Embryos were counter-stained with FM4-64 and imaged as described in STAR Methods. mNG expression marks neural cells receiving the morpholino. Each frame is a maximum intensity projection of 15 images collected at 0.75 μm intervals in Z near the apical surface; frames were collected at 60s intervals. The movie is displayed at 15 fps.

Movie S4

Movie S4 (related to Figure S4). Time lapse movie of junctional dynamics during zippering in an embryo expressing ZO1::GFP under the control of a promoter (pFOG) that drives expression in all epidermal cells and a subset of neural cells lying along the Ne/Epi boundary. Because of mosaic transgene expression, only half of the embryo expresses ZO1::GFP. Each frame is the maximum intensity projection of 15 images collected at 0.75 μm intervals in Z near the apical surface; frames were collected at 30s intervals. The movie is displayed at 25 fps.

Movie S5

Moive S5 (related to Figure 4). Visualization of active RhoA dynamics during zippering in live embryos expressing GFP::AHPH under the control of a neural-specific promoter (pEtr1*) in midline neural cells. Left panel shows GFP::AHPH (magenta) with a counter-stain FM4-64 (cyan). Right panel shows GFP::AHPH alone. Each frame is the maximum intensity projection of 15 images collected at 0.75 μm intervals in Z near the apical surface; frames were collected at 60s intervals. The movie is displayed at 15 fps.

Movie S6

Movie S6 (related to Figure 4). Time-lapse movies showing active RhoA dynamics in embryos in which either a control (left) or Cad2 (right) morpholino was co-injected with pEtr1*::GFP::AHPH into a single b4.2 blastomere at the 8-cell stage. Embryos were counter-stained with FM4-64 and imaged as described in STAR Methods. pEtr1*::GFP::AHPH is expressed only in midline neural cells, on the left side of the embryo, that received the morpholino. Each frame is the maximum intensity projection of 15 images collected at 0.75 μm intervals in Z near the apical surface; frames were collected at 60s intervals. The movie is displayed at 15 fps.

Movie S7

Movie S7 (related to Figure 6). 3D reconstruction of a neurula stage embryo in which Gap21/23 MO was injected into one b4.2 blastomere at the 8-cell stage. Orange arrowheads indicate Ne/Epi junctions on injected (filled orange arrowheads) and non-injected (open orange arrowheads) sides of the embryo. Magenta arrowheads indicate Ne/Ne junctions in the midline neural cells injected Gap21/23 MO (filled magenta arrowheads) and in the non-injected control side (open magenta arrowheads) sides of the embryo. See also Figure 6E.

Figure S2

Figure S2 (related to Figure 2). Localization of Cad2 and β-catenin in midline cells. (A) Neurula stage embryos expressing pEtr1*::Cad2::GFP, then fixed and counter-stained with phalloidin. Left panel: Manual tracing of midline neural cells on the left side of the embryo ahead of and behind the zipper. Color fill indicates apical surface (brown), Epidermal contact surfaces (grey) and newly-formed surfaces of contact with neural cells on the opposite (right) side just behind zipper (purple). Orange lines and arrowheads indicate the original Ne/Epi boundary. Magenta lines and arrowheads indicate Ne/Ne junctions ahead of zipper. Green lines and arrowheads indicate newly formed Ne/Ne junctions behind zipper. Yellow circle indicates zipper. Right three panels show Cad2::GFP signal, phallodin stain, and merge. Box plots show relative enrichment of Cad2::GFP, measured as the ratio of junctional to cytoplasmic intensity for junction types indicated below each graph. (B) Left: A neurula stage embryo fixed and immunostained with an antibody against β-catenin. Orange lines and arrows indicate Ne/Epi junctions, magenta and green arrows indicate Epi/Epi and Ne/Ne junctions, in midline neural cells. Top right: Schematic view indicating junction types. Bottom right: Graphs showing ratios of β-catenin intensities measured for the indicated junction types. (C) Co-injection of a second Morpholino Oligonucleotide (MO2) against Cad2 with an expression marker (pEtr1*::Lifeact::mNG) into a single b4.2 blastomere at the 8-cell stage. (D, E) Box plots showing ratios of (D) 1P myosin and (E) boundary roughness on injected:non-injected sides of the embryo for the indicated junction or boundary types. Control measurements were made on embryos injected with control morpholino. (n = 12 junctions, from 6 embryos in A; n > 12 junctions, from > 5 embryos for each condition in B; n > 23 junctions, from > 9 embryos for each condition in D; n >10 embryos for each condition in E.). *p < 0.1, **p < 0.05, *** p < 0.005, **** p < 0.0005, Student’s t test. Scale bars, 10 μm.

Figure S1

Figure S1 (related to Figure 1). Tools for driving tissue specific gene expression of iMyo::GFP and other factors. (A) Names of genes in this study (Stolfi et al., 2015). (B) GFP expression pattern driven by different promoters used in this study. Green cells indicate GFP-expressing cells. Orange line indicates Ne/Epi boundary. (C) We combined the ETR promoter characterized by (Veeman et al., 2010) with a basal pFOG promoter (Rothbächer et al., 2007) to produce an enhanced promoter pETR1* that drives Lifeact-mNeon Green (mNG) expression throughout the presumptive neural territory, including in midline neural cells b9. 33, b9.37 and b9.38, which are daughter cells of b8. 17 and b8. 19 in Figure 2A. Scale bars, 20 μm (D, E) Relationship between iMyo-GFP intensity and junction length during individual junction shortening events in embryos iMyo-GFP in epidermal cells (D) or neural cells (E). Data from individual junctions were aligned with respect to the onset of shortening (see STAR Methods for details). Red lines: normalized average junction length. Blue dashed lines: relative iMyo-GFP fluorescence intensity averaged along the junction, excluding the vertices. n = 16 junction shortening events from 10 embryos in (D), n = 10 junction shortening events from 8 embryos in (E). Error bars are SEM.

Figure S4

Figure S4 (related to Figure 4). RhoA activity accumulates at cleavage furrows in control and Cad2 knockdown embryos. Co-injection of Morpholino Oligonucleotides (MO) against Cad2 with an expression marker (pEtr1*::GFP::AHPH) into a single b4.2 blastomere at the 8-cell stage. Yellow indicates all descendants of the b4.2 blastomere; green indicates midline neural descendants that received the MO. Images extracted from Movie S6. Top row shows the biosensor (magenta) and a membrane marker (FM4-64; blue); bottom row shows the biosensor alone. White and orange arrowheads indicate cleavage furrows and Ne/Epi junctions. Scale bars, 10 μm.

Figure S6

Figure S6 (related to Figure 6). Gap21/23 over-expression in midline neural cells abolishes RhoA/myosin II polarity and blocks zippering. (A) GFP::Gap21/23 (green) is over-expressed in neural cells on one side of the embryo. Left: schematic showing measurement of “un-zippered length” L. Right: Embryos over-expressing pEtr1*::GFP::Gap21/23 or a control transgene (pEtr1*::Lifeact::GFP), then fixed and co-stained with phalloidin. Scale bars, 10 μm. (B) Quantitation of zipper progression in Gap21/23 or Gap21/23^R1231A over-expressing embryos (n > 14 embryos for each condition). (C-I) Effects of Gap21/23 or Gap21/23^R1231A over-expression on distribution of GFP::AHPH in midline neural cells. Top schematic: green indicates cells expressing pEtr1*::GFP::AHPH and pEtr1*::Gap21/23. Bottom panels: Orange lines and arrowheads indicate Ne/Epi junctions and magenta lines and arrowheads indicate Ne/Ne junctions in midline neural cells. Scale bars. 10 μm. (D) Relative enrichment of GFP::Gap21/23 or GFP::Gap21/23^R1231A, measured as the ratio of junctional to cytoplasmic intensity. (E) GFP::AHPH polarity measured in GFP::Gap21/23 or GFP::Gap21/23^R1231A over-expressing embryos as the ratio of intensities on Ne/Epi::Ne/Ne junctions. (F-I) Effects of Gap21/23 or Gap21/23^R1231A over-expression on distribution of 1P-myosin in midline neural cells. (F) Top panel shows dorsal surface view of a neurula stage embryo, over-expressing GFP::Gap21/23 in all neural cells except midline neural cells on one side of the embryo, then fixed and immunostained for 1P myosin. Green color fill indicates neural cells expressing GFP::Gap21/23. Orange line and arrowheads indicate the original Ne/Epi boundary, magenta lines and arrowheads indicate Ne/Ne junctions in midline neural cells. Second panel: superposition of GFP injection marker and 1P-myosin. Third panel shows localization of GFP::Gap21/23 (left) or 1P-myosin (right) in the same embryo. Open and filled arrowheads indicate corresponding boundaries on GFP::Gap21/23 over-expressing and control sides of the embryo. Scale bars, 10 μm. (G) 1P myosin polarity measured as the ratio of intensities on Ne/Epi::Ne/Ne junctions. (H) 1P myosin ratio in Gap21/23 or Gap21/23^R1231A -expressing embryos. (I) Roughness ratio in Gap21/23 or Gap21/23^R1231A -expressing embryos. (J) Sequence alignment of a portion of the RhoGAP domain from GAP21/23, GAP21, GAP23, Pac1 and GAP19D. Arginine 1231 of GAP21/23 is marked with an asterisk. (K) 1P myosin ratio in embryos injected with a second Morpholino Oligo (MO2) against Gap21/23. (L) Validation of Cdc42 biosensor; Reduction of GFP::WASP-GBD intensity in cells expressing dominant negative form of CDC42. Scale bars, 10 μm (M) Distribution of a Cdc42 biosensor GFP::WASP-GBD. Orange arrowheads indicate the original Ne/Epi boundary Scale bars, 10 μm. (n > 19 junctions from > 8 embryos for each condition in D, E, G, H and K, n > 9 embryos for each condition in I). **p < 0.05, *** p < 0.0005 **** p < 0.0005, Student’s t test.

Figure S5

Figure S5 (related to Figure 5). Phylogenetic analysis, expression pattern and localization of Gap21/23. (A) Sequence-based phylogeny for GAP domains from metazoan Rho GTPase activating proteins. Position of Ciona robusta Gap21/23 is marked with a magenta square. (B) List of Ciona robusta Rho GAPs tested in this study. (C) Gap21/23 gene expression pattern from eight-cell to gastrula stage. Scale bars, 25 μm. (D) Neurula stage embryos expressing pEtr1*::GFP::Gap21/23, fixed and counter-stained with phalloidin. Left panel: Manual tracing of midline neural cells on the left side of the embryo ahead of and behind the zipper. Color fill indicates apical surface (brown), Epidermal contact surfaces (grey) and newly-formed surfaces of contact with neural cells on the opposite (right) side just behind zipper (purple). Orange lines and arrowheads indicate the original Ne/Epi boundary. Magenta lines and arrowheads indicate Ne/Ne junctions ahead of zipper. Green lines and arrowheads indicate newly formed Ne/Ne junctions behind zipper. Yellow circle indicates zipper. Right three panels show GFP::Gap21/23 signal, phallodin stain, and merge. Box plots show relative enrichment of GFP::Gap21/23, measured as the ratio of junctional to cytoplasmic intensity for junction types indicated below each graph. (E) Co-injection of a Morpholino Oligonucleotide (MO1) against GAP21/23 with pEtr1*::Cad2::GFP into a single b4.2 blastomere at the 8-cell stage has no effect on Cad2 polarity. (F) Cad2::GFP polarity measured as ratio of intensities on junction types indicated below y-axis. (n = 12 junctions, from 6 embryos in D; n > 15 junctions, from > 6 embryos for each condition in F). **** p < 0.0005, Student’s t test.

Figure S3

Figure S3 (related to Figure 3). Tissue-specific expression of Cad2 in the anterior neural plate directs myosin activation to plate boundaries and anterior neural tube (brain) closure. (A) Sequence of frames from Movie S4 showing the progression of anterior neural tube (brain) closure in an embryo expressing ZO-1::GFP in anterior neural and epidermal cells, driven by the FOG promoter (Figure S1, (Pasini et al., 2006)). Magenta arrowheads indicate zipper position. Orange line indicates Ne/Epi boundary. Schematic on left indicates expression domain. Scale bars, 10 μm (B) Distribution of 1P-myosin at successive steps during brain closure. Orange arrowheads indicate Ne/Epi boundary; Magenta arrowheads indicate zipper position. Scale bars, 10 μm (C) Cad2 gene expression pattern at late gastrula, early neurula (during nerve cord zippering) and late neurula (during brain closure) stages. Nerve cord cells are outlined in light blue; anterior brain precursors are outlined in dark blue. Scale bars, 25 μm. (D) Co-injection of Cad2 MO with an expression marker (pZicL::Lifeact::GFP) into a single a4.2 blastomere at the 8-cell stage. Yellow indicates all a4.2 descendants; green indicates anterior neural descendants that received the MO. (E) Effects of Cad2 knockdown in anterior neural cells. Top panel shows schematic dorsal surface view of late neurula stage embryos that were injected with control MO (left) or Cad2 MO (right), then fixed and immunostained for 1P myosin. Yellow indicates neural cells inheriting co-injected morpholino and GFP marker (green) over endogenous Cad2 (blue). Orange line indicates the original Ne/Epi boundary. Blue line indicates boundary between injected and non-injected neural cells. Middle panel shows superposition of GFP marker and 1P-myosin in the same embryos. Bottom panel shows 1P-myosin alone. Open and filled arrowheads indicate corresponding boundaries on injected and non-injected sides of the embryo. Scale bars, 10 μm. (F) Box plots showing 1P myosin ratios for the indicated junction types. Ne/Epi ratios were measured on injected:non-injected junctions. Ne/Ne ratios were measured as the average intensity on junctions between injected and non-injected neural cells (filled blue arrowheads in E) divided by average intensity on junctions between non-injected neural cells (open blue arrowheads in E). Control measurements were made on embryos injected with control morpholino (n > 29 junctions, from > 7 embryos for each condition; *** p < 0.005, **** p < 0.0005, Student’s t test).