Wnt/PCP signaling controls intracellular position of MTOCs during gastrulation convergence and extension movements

Cell movements are essential in the embryo to shape the axial body plan and in the adult for homeostasis and wound healing. During vertebrate gastrulation, convergence and extension (C&E) movements narrow and elongate the germ layers to form the anteroposteriorly (AP) elongated body axis that is common to vertebrate embryos (Keller, 2002; Solnica-Krezel, 2005). This morphogenetic process involves multiple cell behaviors that change according to stage and location (Roszko et al., 2009).

Non-canonical Wnt/planar cell polarity (Wnt/PCP) signaling regulates C&E in chordates from ascidians through mammals by polarizing the morphology and behavior of cells (Jiang et al., 2005; Solnica-Krezel, 2005). Wnt/PCP signaling organizes sheets of cells to exhibit a consistent polarity perpendicular to the axis of the apical-basal polarity (Axelrod and McNeill, 2002; Eaton, 2003; Fanto and McNeill, 2004). In Drosophila, where planar cell polarity was discovered, PCP signaling is required to coordinate the orientation of epithelial structures in the cuticle, wing, and eye (Gubb and Garcia-Bellido, 1982; Adler, 2002). Downstream of an asymmetry cue (Strutt and Strutt, 2009), transmembrane proteins Frizzled (Fz) and Vangogh (Vang) are required in adjacent cells to initiate planar polarization. Other components of the Wnt/PCP pathway, Dishevelled (Dvl) and Prickle (Pk), amplify the difference leading to visible asymmetry in protein distribution; Fz and Dvl are found at the distal side of the cell where a hair forms, while antagonistic factors Vang and Pk are enriched at the opposite cell membrane (Usui et al., 1999; Axelrod, 2001; Feiguin et al., 2001; Shimada et al., 2001; Strutt, 2001; Tree et al., 2002; Bastock et al., 2003; Narayan et al., 2004; Simons and Mlodzik, 2008).

Wnt/PCP signaling most commonly controls polarity in epithelia, where cells tend to maintain close cell-to-cell contact. However, during vertebrate gastrulation, Wnt/PCP also influences polarity in mesenchymal cells that form transient cell contacts (Jessen et al., 2002; Marlow et al., 2002). Wnt/PCP signaling promotes polarized cell migration and rearrangements and coordinates this polarity with embryonic axes. These polarized movement behaviors include directed migration, mediolateral intercalation (Wallingford et al., 2000; Jessen et al., 2002) and polarized radial cell intercalation (Yin et al., 2008). Furthermore, oriented cell divisions also require Wnt/PCP signaling (Gong et al., 2004; Ciruna et al., 2006). Under the influence of Wnt/PCP signaling, cells elongate and align mediolaterally, and thus perpendicular to the AP gastrula axis. In addition, the direction and stability of protrusions, and movement direction of cells, are mediolaterally biased (Wallingford et al., 2000; Topczewski et al., 2001; Jessen et al., 2002; Montero et al., 2005). Mechanistically, Wnt/PCP signaling influences dynamic cohesion of anterior mesoderm cells (Montero et al., 2005), endocytosis of E-cadherin molecules (Ulrich et al., 2005), and the distribution of Fibronectin (Goto et al., 2005; Yin and Solnica-Krezel, 2007).

Such striking polarization of cell behaviors suggests Wnt/PCP influences the microtubule (MT) cytoskeleton. Evidence implicates Wnt/PCP signaling in orienting the basal body in cochlea, floor plate of the neural tube, and epidermis, suggesting an interaction with the centrosome or microtubule organizing center (MTOC) (Jones and Chen, 2008; Park et al., 2008; Borovina et al., 2010). As the MTOC organizes the MT cytoskeleton, Golgi complex, cilium and the proteasome, its position polarizes many cellular functions (Badano et al., 2005). Here, we begin to examine the influences of Wnt/PCP on the MT cytoskeleton during gastrulation by testing whether Wnt/PCP signaling can control the position of the MTOC.

The asymmetric accumulation of PCP components, so obvious in Drosophila epithelia (Simons and Mlodzik, 2008), has been difficult to document in migrating cells of vertebrate gastrulae. However, recent studies showed anterior membrane localization of exogenous Drosophila Prickle-GFP in the neural plate and mesoderm (Ciruna et al., 2006; Yin et al., 2008), and posterior accumulation of Xenopus Dishevelled-GFP in mesoderm during zebrafish gastrulation (Yin et al., 2008). The molecular mechanisms that establish the asymmetric distribution of the Wnt/PCP pathway components remain incompletely understood (Strutt and Strutt, 2007). However, in Drosophila epithelia, intact MTs are needed for proper Fz localization. Furthermore, Fz-GFP protein travels along MTs toward the distal cell edge before its asymmetric accumulation is apparent (Shimada et al., 2006).

Here we have investigated interactions between Wnt/PCP signaling and the MT cytoskeleton in zebrafish gastrulation cell movements that differ in their dependence on Wnt/PCP signaling (Fig. 1A,B). First, we find that MTOCs have a biased location along the superficial-deep axis of the embryonic tissues, suggesting an emerging apical-basal polarity. Second, we find that MTOCs occupy a polarized position within the plane of the ectoderm and mesoderm, becoming biased to the posterior and dorsal/medial side of the cell between mid and late gastrulation. We find that interfering with the Wnt/PCP signaling components Knypek/Glypican4/6 (Kny/Gpc4) and Dvl randomizes MTOC position in the planes of the ectoderm and mesoderm.

Third, we asked whether the establishment and maintenance of mediolateral and AP cell polarity mediated by Wnt/PCP signaling requires an intact MT cytoskeleton. We find that disruption of the MT cytoskeleton at midgastrulation, before the onset of Wnt/PCP signaling, blocks clustering of Prickle-GFP at the anterior cell membranes. By contrast, disruption of the MT cytoskeleton when cells are mediolaterally elongated did not eliminate Pk-GFP puncta from the membrane, or block their anterior location. Hence, MTs are needed for the establishment, but not for the maintenance, of the anterior Pk localization.

Together, our studies of MTOC intracellular position show a nascent polarity aligned with the future apicobasal polarity and planar cell polarity in cells undergoing C&E during gastrulation. Moreover, we reveal complex mutual interactions between components of Wnt/PCP signaling and the MT cytoskeleton; some positively acting components (Dvl and Kny/Gpc4) are required for polarized MTOC position. In turn, an intact MT cytoskeleton is not necessary for the maintenance of Prickle, a Wnt/PCP protein, at the anterior cell membrane during C&E, but is needed for the establishment of this polarity.

AB^* wild-type zebrafish and fish carrying mutations in knypek^fr6 (Topczewski et al., 2001) were maintained as described (Solnica-Krezel et al., 1994). Naturally spawned embryos were raised at 26-32°C, and staged according to their morphology (Kimmel et al., 1995). Transgenic Tg(XlEef1a1:dclk2DeltaK-GFP)io007 fish (a gift from Marina Mione, Milan, Italy) were used to test nocodazole effectiveness. We use descriptive terms to denote gastrulation stages: `start of gastrulation' refers to shield stage (6 hpf), `midgastrulation' refers to 75 to 85% epiboly (8-8.7 hpf) and `late gastrulation' refers to yolk plug closure to the one-somite stage (9.5-10.2 hpf).

RNA was synthesized with (mMESSAGE mMACHINE; Ambion) and injected into one- or two-cell embryos at the following doses: Xdd1 (400 pg), membrane-localized (i.e. CAAX/Ras membrane-localization domain) RFP or Cherry (200-400 pg) (a gift from Fang Lin, University of Iowa, IA, USA), Xenopus EGFP centrin (5-40 pg) (a gift from Adrian Salic and Kristin Kwan, Harvard Medical School, MA, USA). Embryos were fixed overnight in 4% paraformaldehyde in PBS, permeabilized in 0.5% Triton in PBS 30-60 minutes at room temperature and labeled with anti-acetylated tubulin MAb (1/800 Sigma T6557) or anti γ-tubulin (1/250 Sigma T6793).

Embryos were injected with RNAs coding for RFP-CAAX (Jason Jessen, Vanderbilt University Medical Center, TN, USA) at the one-cell stage and for Drosophila Pk-GFP (Jenny et al., 2003) with or without Histone2A-RFP at the 8- to 32-cell stage and allowed to develop until mid or late gastrulation. Dechorionated embryos were grown for 30-120 minutes in 28 or 32°C Danieau's buffer with 0, 5 or 10 μg/ml nocodazole.

Live embryos were mounted in 2% methylcellulose or 0.5% low-melting-point agarose (Seaplaque catalog number 50100) in 30% Danieau's buffer (Saude et al., 2000) on glass-bottomed dishes (Matek PG35G-0-10-C or Willco Wells BV-0310-35). Confocal and DIC z-stacks (0.5-4 μm spacing) were collected from dorsal regions with a Zeiss LSM 510 microscope using 20× or 40× oil-immersion (NA 1.3) objectives, or spinning disk confocal using a 40× oil-immersion (NA 1.3) objective mounted on a motorized Zeiss Axiovert 200 and a PerkinElmer ERS spinning-disk confocal system. The positions of the cell nucleus (visible in DIC or by diffuse Centrin labeling) and MTOC (Centrin-labeled) were marked in Object-Image (NIH image) and exported to Excel (Microsoft), where angle was calculated. Cells containing two well-separated MTOCs (greater than 7-10 μm) were not counted. Cell divisions (cells with two opposite MTOCs and furrowing of membrane) were described as `aligned with the AP axis' if they were within 45° of the axis. Further analysis was performed using LSM-Browser (Zeiss), Image J (NIH), Object Image (Norbert Vischer), Volocity (Improvision) and Excel (Microsoft). Standard error of the mean (s.e.m.) was calculated as (standard deviation/√sample size). VectorRose (Watson's two-sample U² test, PAZsoftware, Pierre A. Zippi) was used for statistical analysis.

To gauge polarity of the MT cytoskeleton in gastrula cells, we examined the location of the centrosome, labeled by injection of Xenopus centrin-gfp RNA. Centrin is a component of the distal side of the centriole, a structure that serves several functions in the cell, including organizing MTs during cell division and interphase, and acting as the basal body of a cilium (Dawe et al., 2007).

Antibody against γ-tubulin and fluorescently labeled Xenopus Centrin (eGFP-Xcentrin) fusion protein co-labeled (data not shown, DNS) small, tightly focused and usually paired spots that could be found in nearly all cells of the zebrafish gastrula; we interpret these to be centrioles (Fig. 1E). We injected synthetic RNA encoding eGFP-Xcentrin into 1- to 16-cell stage embryos. Centrosomes were often found close to the cell membrane, and were sometimes close to centrosomes in an adjacent cell (Fig. 1F,G). Furthermore, some centrioles localized to the tips of intracellular projections of the plasma membrane, which followed the centrioles as they moved (Fig. 1G, blue arrow). Time-lapse recordings showed that centrioles moved rapidly in cells, as has been reported for cultured cells [Fig. 1G, average instantaneous speed 2.14 μm/minute, s.e.m. 0.53, average net speed 0.61 μm/minute s.e.m. 0.26, n=1 YPC embryo, 122 cells (Piel et al., 2000)].

To learn whether centrosomes serve as basal bodies during late gastrulation, we labeled centrosomes with eGFP-Xcentrin and cilia with an antibody against acetylated tubulin. Short presumptive cilia are seen on the outermost enveloping layer cells from as early as midgastrulation (80% epiboly, 8 hours post fertilization, hpf, DNS). By segmentation, some centrosomes were found abutting presumptive cilia (Fig. 1H, blue arrowhead) in neural plate and mesoderm, and in the interior of the notochord (DNS). The plasma membrane attached to centrosomes may enclose the intracellular parts of cilia, as these structures are approximately the same length (compare Fig. 1F, blue arrowhead with Fig. 1H).

Cells in an epithelial tissue can be polarized over multiple axes, for example, across the plane of an epithelial sheet (i.e. planar polarity), as well as across the thickness of the tissue (i.e. apical-basal polarity). In epithelia, centrosomes are typically located near the apical surface (Hay, 2005; Baum et al., 2008; Hong et al., 2010), whereas in migrating mesenchymal cells, centrosomes frequently assume a position between the leading edge and the nucleus. During zebrafish gastrulation, lateral ectoderm and mesoderm cells migrate towards the dorsal midline as a cohesive sheet or as individual cells, respectively (Solnica-Krezel, 2005). Near the midline, mesenchymal mesoderm cells cohere to form a cohesive body axis. We wondered whether MTOC position was polarized relative to the superficial-deep axis of the germ layers during these migrations. We labeled cell membranes and centrosomes by injection of synthetic RNA encoding membrane-localized red fluorescent protein (membrane-RFP) and green fluorescent protein Xenopus centrin fusion (GFP-centrin) protein, respectively. We collected z-stacks from the resulting mid and late gastrulae and made orthogonal reconstruction stacks of the lateral region (above the blastoderm margin, about 13 cells wide) at midgastrulation and the region immediately adjacent to the notochord (seven cells wide, at the equator) at late gastrulation (Fig. 1A,B). These images showed the position of MTOCs relative to the germ layer boundaries in both ectoderm and mesoderm (Fig. 2A-F). We observed that MTOCs were present preferentially near the superficial side of the ectodermal cells. Intriguingly, MTOCs were largely absent from the ectoderm/mesoderm boundary. Within the mesoderm, MTOCs were found toward the interior of the germ layer. Quantification of the average MTOC position relative to the germ layer thickness position in three mid- and two late wild-type gastrulae and in two kny^–/– gastrulae was consistent with this visual impression from previous embryos (Fig. 2G). Based on these observations, we conclude that gastrula cells exhibit polarity along the superficial-deep axis of the germ layers, possibly representing a nascent apical-basal polarity by midgastrulation that is independent of Wnt/PCP signaling.

Cells engaged in C&E also become polarized within the tissue plane, i.e. planar cell polarity (Wallingford et al., 2000; Topczewski et al., 2001; Jessen et al., 2002; Marlow et al., 2002; Myers et al., 2002). Our current and previous studies have examined cells that begin gastrulation at the lateral margin. These mesodermal cells demonstrate different migration behaviors and dependence on Wnt/PCP signaling at mid and late gastrulation. At midgastrulation, these moderately elongated and loosely packed cells (Jessen et al., 2002; Sepich et al., 2005) converge dorsally and become distributed in an anteroposteriorly elongated array of highly packed and mediolaterally elongated cells. Our observations were centered on cells at the embryonic equator; by the one-somite stage, these equatorial cells contribute to the first two somites and the anterior mesoderm. Consequently, we examined mesoderm at midgastrulation when cells are about 75° from the midline (Fig. 1A), and again at late gastrulation when cells quickly traverse a region 55-20° from the midline (Fig. 1B). At midgastrulation, the germ layers are one or two cells thick with irregularly shaped cells, consistent with lower cell density. Thickness of the germ layers remains similar at late gastrulation, as cells become densely packed; when somites form, well-organized two-cell-thick tissues appear (Fig. 2A,B and DNS).

To determine whether the MTOCs have a polarized position within the plane of the germ layer in this one cell population at mid and late gastrulation, we collected simultaneous confocal and differential interference contrast (DIC) z-stack images to reveal the orientation of the nucleus-MTOC axis relative to the body axes of the embryo (Fig. 1A-E). As illustrated, we calculated the position of the MTOC relative to the nucleus and the body axes, and described this relationship as an angle relative to the anterior side of the embryo (see Fig. 1E). At midgastrulation, when lateral cells are only moderately elongated (length-to-width ratio, LWR=1.4±0.28) (Jessen et al., 2002), we found no evidence of polarization of MTOC position. MTOCs were randomly oriented with respect to the nucleus and embryonic axes (Fig. 3; Tables 1 and 2). However, we found that by late gastrulation, MTOC position became biased in both ectodermal and mesodermal cells (Fig. 3). At yolk plug closure (9.5 hpf) and tailbud (10 hpf) stages, the most common position for MTOCs was posterior of the nucleus or posterior and slightly dorsal/medial of the nucleus (Fig. 3, Table 1). This bias was not correlated with the direction of migration, which is dorsalward at this time (n=3 embryos, n=58 ectodermal cells, average direction 266° from anterior, n=57 mesodermal cells, average direction, 263°). By the one-somite stage (10.2 hpf), the most common position for the MTOC was posterior-dorsal or posterior-ventral of the nucleus relative to the body axes. The distributions of MTOC orientations were statistically different from a random distribution, and different from the orientation of MTOC found in cells at earlier stages of gastrulation (Tables 1, 2). Therefore, we conclude that MTOCs acquire a biased position with respect to the AP axis within cells engaged in C&E movements at late gastrulation.

The labeling method did not disturb the normal orientation of cell division, which appeared to be normally aligned with the AP axis. Dividing cells were counted separately and were defined as round cells or cells with an obvious cleavage furrow and two centrioles, which defined spindle orientation. In addition, cells with two centrosomes separated by more than a small distance (∼5 μm) were excluded from orientation analysis, as nuclear/centrosome orientation could not be uniquely defined. Previous work has shown that more than 90% of dividing cells in ectoderm during gastrulation were aligned within 45° of the AP axis (Gong et al., 2004). Accordingly, we observed that more than 82% of dividing ectodermal and 69% of mesodermal cells aligned their division plane with the AP axis at late gastrulation (9.5-10.2 hpf, 17 embryos, n=101/123 and 32/46 divisions, respectively, composite 78%, n=133/169) and at midgastrulation (75-85% epiboly, 11 embryos, 8-8.7 hpf, 72 and 75% aligned in ectoderm and mesoderm, respectively, n=13/18 and 6/8 divisions; Table 3).

Wnt signaling is required for reorientation of the MTOC in fibroblasts in the cell culture scratch assay (Schlessinger et al., 2007). The implicated Wnt pathway involves elements of canonical signaling (the DIX domains of Dvl and Axin, and inhibition of GSK3β) and Wnt/PCP signaling (the ligand Wnt5, which can stimulate canonical or Wnt/PCP signaling) (Mikels and Nusse, 2006; Cha et al., 2008; Mikels et al., 2009).

To test whether Wnt/PCP signaling regulates MTOC polarity in dorsally migrating cells during zebrafish gastrulation, we manipulated Wnt/PCP signaling by two means and analyzed the consequences on the intracellular position of MTOCs during late gastrulation. We found that when Wnt/PCP signaling was impaired, the distribution of MTOC positions differed from that found in wild-type gastrulae and often was indistinguishable from the random distribution. First, we altered the function of Dvl, a pathway component common to both Wnt/β-catenin and Wnt/PCP signaling, by overexpressing synthetic RNA coding for the dominant-negative Xdd1 mutant protein. According to earlier studies, overexpression of Xdd1 in the dose range used here inhibits the Wnt/PCP pathway and C&E without affecting patterning and thus canonical Wnt signaling. Accordingly, embryos overexpressing Xdd1 exhibited shortened bodies, synopthalmia or cyclopia, and reduced elongation and alignment of mesodermal cells (Tada and Smith, 2000; Wallingford et al., 2000; Jessen et al., 2002) (see Fig. S1 in the supplementary material; 400 pg Xdd1). Notably, MTOCs were positioned randomly in ectoderm and randomly or anteriorly in mesoderm in gastrula cells at late gastrulation (YPC and TB stages, Fig. 4, Tables 1 and 2). Moreover, dividing cells were less well aligned with the AP axis compared with wild-type embryos (59% of 46 dividing cells in five Xdd1-injected versus 78% aligned in 17 wild-type embryos, Table 3), but not as disturbed in this experimental setting as previously reported, suggesting Wnt/PCP signaling was reduced but not absent (Gong et al., 2004).

Second, we examined MTOC orientation in embryos homozygous for the null fr6 allele of knypek (kny^fr6/gpc4), a positively acting component of the Wnt/PCP pathway (Topczewski et al., 2001). In these zygotic mutants, we found that MTOCs were positioned randomly in ectoderm and randomly or anteriorly in mesoderm at late gastrulation (Fig. 4, Tables 1, 2). We observed disturbed cell division orientation in kny embryos compared with wild-type siblings (64% aligned, 121 dividing cells in eight embryos, Table 3), consistent with impaired Wnt/PCP activity.

Are MTOC distributions in the Wnt/PCP gastrulae different from wild-type distributions and different from random ones? Relatively few observations came from individual embryos; distributions only emerged when data was combined from several embryos. Each Wnt/PCP-depleted scenario was compared with a random distribution and to the sample collected from stage matched control wild-type embryos, using the Watson's U² test of circular data and a significance limit of 1% (Table 1). We found that the distributions of MTOCs in Wnt/PCP-depleted embryos differed at every stage from those in control gastrulae. Moreover, distributions of MTOCs in the ectoderm of Wnt/PCP-depleted embryos were not different by statistical analysis from data expected for a random distribution. The results for mesoderm were mixed. Data from kny at 1 somite stage and from Xdd1 at YPC stage were indistinguishable from a random distribution. Statistical analysis indicated data from tailbud stages barely exceeded the critical value of U², indicating difference from a random population. These distributions appeared to have an anterior bias. Consequently, we observe that distribution of MTOCs in Wnt/PCP-depleted embryos is altered from that observed in control gastrulae, and this probably represents randomization or an anterior bias of MTOC position.

In summary, reducing the function of two positively acting Wnt/PCP signaling components, Dvl and Kny/Gpc4, resulted in randomization of a polarized trait, the intracellular position of the MTOC during late gastrulation. Based on these observations, we propose that Wnt/PCP signaling can bias the intracellular localization of MTOCs.

PCP signaling in Drosophila wing primordia results in the asymmetric localization of PCP components across each cell. Fz and Dvl proteins accumulate distally, whereas Vang and Pk localize proximally (Strutt and Strutt, 2007; Axelrod, 2009; Amonlirdviman et al., 2005; Wu and Mlodzik, 2008). Frizzled protein is proposed to initially arrive at the distal side through an asymmetric transport of vesicles along MTs aligned with the proximo-distal wing axis (Shimada et al., 2006). Accumulation of the Wnt/PCP component, Pk, at the anterior cell edges has also been observed during C&E in ascidians and zebrafish gastrulae (Jiang et al., 2005; Ciruna et al., 2006; Yin et al., 2008). This anterior localization is lost or reduced in embryos deficient in Wnt/PCP components Dvl, Kny/Gpc4 and Tri/Vangl2 (Ciruna et al., 2006; Yin et al., 2008). Moreover, in tri/kny compound mutants, Pk-GFP remains mostly unclustered in the cytosol (Yin et al., 2008). To begin to address the cellular mechanism by which Wnt/PCP components become asymmetrically localized during C&E, we asked whether an intact MT cytoskeleton was needed for the establishment and/or maintenance of Pk localization at the anterior cell membrane in dorsal mesoderm and ectoderm. To find a treatment capable of depolymerizing MTs, we exposed dechorionated embryos carrying a transgene for a GFP-tagged microtubule-binding protein (TgXlEef1a1:dclk2DeltaK-GFP) to nocodazole (5 or 10 μg/ml) starting at midgastrulation or at YPC, continuing 1 or 2 hours. We found that 5 μg/ml nocodazole efficiently depolymerized MTs in mid or late gastrulae, as previously reported (Strahle and Jesuthasan, 1993; Solnica-Krezel and Driever, 1994). Fine MTs were completely absent. Infrequent large bundles of MTs were seen in the enveloping layer cells at all stages and in deep cells at late gastrulation. With longer exposure, even these few MT bundles disappeared, and diffuse fluorescence shifted into the nuclei (see Fig. S2 in the supplementary material).

To determine whether MTs were needed for the initial clustering of Pk at the anterior membrane, Pk-expressing dechorionated embryos were exposed to nocodazole from midgastrulation (8 hpf) until their siblings completed epiboly (about 2 hours, 10.2 hpf). Images were collected by confocal microscopy at the dorsal midline of the embryo to visualize Pk protein accumulation in the notochord, presomitic mesoderm and neural ectoderm. Embryos treated at 80% epiboly failed to close the blastopore; C&E continued as evidenced by the formation of a thicker axis and a small head bulge. Essentially all deep cells became rounded (five embryos, 201 cells, LWR 1.22±0.22) and appeared to lose cell-cell contacts (Fig. 4C and DNS). GFP-Pk clusters were absent. Based on these observations, we conclude that cell-cell contact and subsequent clustering of Pk require intact MTs.

To determine whether MTs were needed to maintain anterior Pk localization, we treated embryos with nocodazole from YPC to the one-somite stage (9.5-10.3 hpf). C&E was partly delayed (DNS) and the embryos were sharply elongated, similar to weakly dorsalized mutants (Mullins et al., 1996). This morphological defect demonstrates a requirement for MTs even after blastopore closure. However, cells maintained mostly elongated morphology (LWR in wild-type notochord cells 2.01±0.59, 226 cells, five embryos versus nocodazole-treated 1.97±0.59, 179 cells, seven embryos). Regarding Pk-GFP distribution, the cells were classified as previously described (Yin et al., 2008): spots limited to the anterior membrane, spots or uniform labeling not limited to the anterior membrane or cytosolic labeling. Surprisingly, the anterior enrichment of Pk-GFP puncta was maintained or decreased slightly (Fig. 5D) nocodazole-treated (n=7) compared with control (n=7) embryos. These results support the notion that intact MTs are required for initial clustering and anterior localization of Pk, either directly or by making membranes competent to accumulate Pk, perhaps through cell-cell adhesion or anterior membrane identity. Once Pk accumulation at the cell anterior is established, MTs are not needed for its maintenance. We propose that MTs are required to establish, but not to maintain, the anterior localization of Pk during C&E. Together, our observations reveal reciprocal functional interactions between MTs and Wnt/PCP signaling in the process of cell polarization during vertebrate gastrulation.

Here, we have analyzed the intracellular position of the MTOC during zebrafish gastrulation. MTOC is in vigorous, often back and forth, motion in gastrula cells (Fig. 1G), similar to observations from cultured cells (Piel et al., 2000). Despite this dynamic nature, analyses of populations of MTOCs showed that they have preferred locations within the cell at distinct stages of gastrulation. Our observations reveal that intracellular positions of MTOC are polarized along two axes. First, by midgastrulation, MTOCs are polarized along a superficial-deep axis through the ectoderm and mesoderm. Second, MTOCs become polarized within the planes of the ectoderm and mesoderm; this planar polarization of MTOCs emerges between mid and late gastrulation.

By midgastrulation, the germ layers are well separated into a superficial ectodermal sheet and deeper mesendoderm layer (Warga and Kimmel, 1990; Concha and Adams, 1998). At this stage, MTOCs display a biased position along the superficial-deep axis of the ectoderm (Fig. 2), but no obvious polarity with respect to the direction of cell migration or body axes (Fig. 3). In the ectoderm, MTOCs were found enriched near the surface of the embryo, suggesting that ectodermal cells orient their presumptive apical surfaces towards the outside and their basal ones towards the inside of the embryo. By contrast, in the mesoderm, MTOCs were found toward the interior of the germ layer (Fig. 2). We speculate that this asymmetric distribution of MTOCs in the germ layers at midgastrulation reflects an early step toward a well-organized apical-basacellular axis.

Several observations suggest that germ layers gradually acquire polarity along the apical-basal axes during zebrafish gastrulation. Epithelia typically form a cellular sheet bounded basally by extracellular matrix (ECM), with basally positioned nuclei and apically located centrosomes or cilia and tight junctions (Thiery, 2002; Barrios et al., 2003). Loosely apposed ectodermal cells condense into a sheet at the onset of zebrafish gastrulation (Concha and Adams, 1998). During internalization, lateral mesoderm disperses into single cells (Warga and Kimmel, 1990), which gradually reaggregate as they converge dorsally (Warga and Kimmel, 1990; Jessen et al., 2002), whereas dorsal mesoderm remains cohesive through dynamic cell-cell adhesion (Montero et al., 2005). Both germ layers show regulated adhesion; they dynamically modulate E-cadherin expression and acquire gap and tight junctions (Essner et al., 1996; Montero et al., 2005; Speirs et al., 2010). A secreted extracellular coat of fibronectin and laminin accumulates at the putative basal sides of the ectoderm and mesoderm, starting as discontinuous patches during gastrulation and becoming continuous by the end of gastrulation (Davidson et al., 2004; Latimer and Jessen, 2009). During segmentation stages, apical-basal polarity becomes overt by asymmetric distribution of classical markers. Asymmetric centrosome and cilia position in the somitic mesoderm has been observed as early as 14 hpf (10 somite stage) (Barrios et al., 2003; Borovina et al., 2010). Pard3 becomes apically localized by early segmentation stages (six or seven somites, 12 hpf) in cells of the neural tube (Tawk et al., 2007; Hong et al., 2010), accompanied by the tight junction marker ZO-1 (five- and 10-somite stage, 11.4 and 14 hpf) (Hong and Brewster, 2006; Yang et al., 2009) and polarity component aPKC (12 somites, 15 hpf) (Hong and Brewster, 2006). Furthermore, orientation of the MT cytoskeleton after 12 hpf (Hong et al., 2010) reveals the maturing apicobasal polarity in the neural epithelium. We suggest MTOC positioning is the earliest indicator of the nascent apical-basal polarity in ectoderm and mesoderm, and trace its origin to before mid-gastrulation.

The second type of polarity, a developing planar cell polarity, was exhibited by biased MTOC positions relative to the body axes in cells undergoing C&E movements. The centrosome organizes the MT cytoskeleton during gastrulation (Hong et al., 2010) (D.S.S. and L.S.-K., unpublished), as well as the Golgi complex, the proteasome and the cilium (Badano et al., 2005). Asymmetry in intracellular MTOC position indicates the location of these organelles and suggests possible planar polarity in their positions. During midgastrulation (8 hpf), the C&E movements of lateral mesodermal cells are thought to be Wnt/PCP independent (Jessen et al., 2002; Sepich and Solnica-Krezel, 2005). At this time, MTOCs exhibited a random or uniform distribution relative to the body axes (Fig. 3). By late gastrulation (after 9 hpf), dorsal ectoderm and mesoderm cells engage in fast C&E movements and display polarized phenotypes mediated by the Wnt/PCP pathway (Sepich et al., 2000; Jessen et al., 2002; Myers et al., 2002; Roszko et al., 2009). Our observations revealed MTOCs were non-randomly oriented relative to the body axes in these cells (Fig. 3), becoming biased towards the posterior and dorsal sides of cells. During the next hour, the positions of MTOC became biased either posterior-medial or posterior-lateral of the nucleus (Fig. 3). At this time, polarized radial intercalations and mediolateral intercalations are observed in the paraxial mesoderm (Yin et al., 2008). In cultured cells, MTOC and/or Golgi position are influenced by physical constraints of the cellular environment, cellular polarity and adhesive contacts. Many, but not all, cultured cell types migrate with a forward-biased MTOC (Yvon et al., 2002). However, the direction of migration of converging gastrula cells is not a good match for MTOC position until late gastrulation, when both MTOC and cell movements align in a dorsal direction. We hypothesize that changes in MTOC orientation may reflect changes in these polarized cell movements. Specifically, medial or lateral positioned MTOC may reflect medial or lateral cell intercalation, respectively. Ongoing time-lapse analyses will directly address this hypothesis. Earlier on, the average MTOC position is about 90° away from the movement direction. Thus, migration direction may have only a partial influence on MTOC positioning in cells engaged in C&E.

Spatial constraints play a role in MTOC position in cultured cells migrating over slender lines of fibronectin (Pouthas et al., 2008). These cells take a dramatically elongated shape with the MTOC trailing behind the nucleus. The MTOC/Golgi also takes a position that is rearward in the leading cells of the lateral line primordium in zebrafish embryos (Pouthas et al., 2008). Cells in the late gastrula elongate less than spindle-shaped cultured cells, but even this elongation might constrain MTOC position. Adhesion between a cell and its neighbors mediated by E- or N-cadherin (Desai et al., 2009; Dupin et al., 2009) or adhesion to substratum, i.e. fibronectin-coated micropatterns of different shapes, has also been shown to control MTOC and nuclear positions (Etienne-Manneville and Hall, 2001; Bornens, 2008). The cell-cell and cell-matrix adhesion could be involved in MTOC position during gastrulation, and testing this is an important future direction for research.

The timing of biased MTOC orientation indicated Wnt/PCP signaling could be a regulator of MTOC position. Wnt signaling, acting through Wnt5a and Dvl, works in parallel to Cdc42 to reorient MTOC in cultured cells (Schlessinger et al., 2007). Additionally, the co-receptor Ror2, and downstream effectors JNK and aPKCζ are involved (Nomachi et al., 2008). Our data are consistent with a role for Wnt/PCP signaling in influencing MTOC position during gastrulation. When Wnt/PCP signaling was disrupted in zebrafish gastrulae, MTOC position relative to the body axes was randomized, or assumed a weak anterior bias. We conclude that two components of Wnt/PCP signaling are required for cellular asymmetry visible in the orientation of the MTOC. Surprisingly, our preliminary studies suggest normal MTOC distributions in trilobite/vangl2 zygotic mutants, implying Vangl2 is not essential for this aspect of MT polarity (D.S.S., Roszko, I. and L.S.-K., unpublished). Finally, whether Wnt/PCP signaling can effect the distribution of growing tips of MTs has not been tested in this system. Another report (Shindo et al., 2008) shows that distribution of growing MT tips is unaffected by loss of Wnt/PCP signaling in Xenopus notochord.

Previous studies have suggested that C&E gastrulation movements are adversely affected by loss of intact MTs (Strahle and Jesuthasan, 1993; Solnica-Krezel and Driever, 1994). Similarly, in Xenopus, MTs are involved in initiation of gastrulation (Lane and Keller, 1997; Lee and Harland, 2007), and later MTs maintain an organized actin cytoskeleton to allow cell contacts and protrusions in the involuting mesoderm (Kwan and Kirschner, 2005). Our studies show reciprocal interactions between Wnt/PCP signaling and the MT cytoskeleton during gastrulation. As discussed above, Wnt/PCP biased intracellular position of MTOC. Conversely, our MT disruption experiments at mid- or late gastrulation revealed that MTs are needed for the initial clustering and localization of Drosophila Prickle-GFP at the anterior cell membrane (and, thus, for a step in AP polarization of cells engaged in C&E). Disruption of MTs by nocodazole treatment at late gastrulation caused only a few cells to round and lose anterior Pk spots; most cells remained elongated with anteriorly accumulated Pk-GFP. Therefore, MTs are needed for the establishment, but not for maintenance, of the anterior Pk-GFP localization and mediolateral cell elongation. Kwan and Kirschner (Kwan and Kirschner, 2005) observed two distinct time-dependent responses of actin to MT depolymerization. Brief exposure to nocodazole strongly reduced the number and extent of lamellipodial protrusions. Long exposure caused cell rounding and loss of cell-cell contacts. Therefore, MTs may be needed to maintain cell contacts, which in turn may be required for anterior identity to be established. Alternatively, the loss of Pk puncta resulting from early treatment with nocodazole may show a direct role for the MT cytoskeleton in the transport or clustering of Pk at the anterior membrane, as in asymmetric transport of Fz observed in Drosophila wing epithelia (Shimada et al., 2006).

The mesoderm has intrinsic AP pattern that is interpreted by Wnt/PCP signaling to guide C&E (Ninomiya et al., 2004). For example, alignment of cell division (Concha and Adams, 1998; Gong et al., 2004) (Table 3) and polarized directed migration are oriented relative to the AP axis and require intact PCP signaling. Polarized planar and radial intercalations contribute to axis elongation by preferentially separating anterior and posterior neighbors in a Wnt/PCP-dependent manner (Yin et al., 2008). AP patterning is also revealed by asymmetric localization of Wnt/PCP proteins and cellular organelles, with Pk-GFP localizing to the anterior (Ciruna et al., 2006) and Dvl-GFP to the posterior edges of cells engaged in C&E (Yin et al., 2009). Recently, studies in zebrafish and mouse showed that centrosomes and cilia in the floor plate of the neural tube, and in the asymmetry organs (Kupffer's vesicle and node) during segmentation stages, are found at the posterior side of the cell in a Wnt/PCP-dependent fashion (Borovina et al., 2010; Hashimoto et al., 2010).

Our studies also reveal posterior bias of MTOC localization that is first detected at late gastrulation, coincident with AP polarization of cells engaged in C&E (Topczewski et al., 2001; Yin et al., 2008). It is tempting to speculate that the posterior bias of MTOC we observe during gastrulation is a precursor of posterior localization of cilia found at later stages of development. Hence, the AP cell polarization established by Wnt/PCP signaling during gastrulation is maintained by tissues at later stages of embryogenesis. We propose a model of reciprocal Wnt/PCP and MT interactions in which Wnt /PCP signaling shapes the MT cytoskeleton by biasing the intracellular position of the centrosome and possibly dependent organelles. In turn, Wnt/PCP signaling requires MT function so it can respond to global AP positional information by enriching Wnt/PCP components at anterior or posterior cell edges and mediate polarized cell movement behaviors underlying C&E.

We thank Christina Speirs, Dan Carlin and Isabelle Roszko for comments, and the following researchers for providing reagents: Masa Tada for Dvl clones, Kristin Kwan and Adrian Salic for the eGFP-Xcentrin clone; and Brian Ciruna and Andreas Jenny for the Drosophila Pk-GFP clone. Experiments were performed in facilities funded by Vanderbilt University Academic Venture Capital Fund and VUMC Cell Imaging Core Facility (NIH grant 1S10RR015682). This work was supported by NIH R01 (GM55101) and a grant from Human Frontiers in Science Program. Deposited in PMC for release after 12 months.