PTK7 is essential for polarized cell motility and convergent extension during mouse gastrulation

The morphogenetic movement of convergent extension (CE) plays key roles in early vertebrate morphogenesis. It extends the presumptive anterior-posterior(A/P) body axis during gastrulation and neurulation, and functions in other aspects of morphogenesis, such as blastopore closure in amphibians and primitive streak elongation in birds(Keller et al., 2000; Glickman et al., 2003; Voiculescu et al., 2007). Failure of CE results in gastrulation defects, notably failure of blastopore closure in amphibians (Schechtman,1942; Keller,1981; Keller,1984), failure of A/P body axis elongation in frogs(Keller, 2002; Keller and Shook, 2008), fish(Myers et al., 2002; Glickman et al., 2003) and mice (Lu et al., 2004; Wang et al., 2006; Ybot-Gonzalez et al., 2007),and failure of neural tube closure in amphibians(Jacobson and Gordon, 1976; Wallingford and Harland, 2001; Goto and Keller, 2002) and mice (Lu et al., 2004; Wang et al., 2006; Ybot-Gonzalez et al., 2007). This central element of early morphogenesis is thus of great biomedical importance, particularly in understanding neural tube defects.

In amphibians, CE of the mesodermal and posterior neural tissues during gastrulation and neurulation drives early body axis elongation(Keller and Danilchik, 1988). Mediolateral intercalation of cells (MIB), driven by mediolateral bi-polarization of mesodermal cell motility(Keller and Tibbetts, 1989; Shih and Keller, 1992a; Shih and Keller, 1992b) and by medially directed cell motility in posterior neural tissues(Keller et al., 1992a; Elul and Keller, 2000; Ezin et al., 2003), produces a narrower, longer array of cells. In the fish, similar MIB(Kimmel et al., 1994; Myers et al., 2002) is driven by polarized motility in the notochord(Glickman et al., 2003) and dorsally directed migration in the lateral mesoderm(Trinkaus, 1998; Jessen et al., 2002; Lin et al., 2005). MIB appears to underlie CE during gastrulation and neurulation in birds(Schoenwolf and Alvarez, 1989; Alvarez and Schoenwolf, 1991; Schoenwolf and Alvarez, 1992)and mice (Smith et al., 1994; Sulik et al., 1994; Wang et al., 2006; Ybot-Gonzalez et al., 2007),where labeling cells shows cell intercalation of the notochord and neural plate (Sulik et al., 1994; Wang et al., 2006; Ybot-Gonzalez et al., 2007),although the underlying cell motility is unknown.

Non-canonical Wnt/planar cell polarity (PCP) signaling appears essential for polarization of the cell behaviors underlying CE in both the frog(Tada and Smith, 2000; Wallingford et al., 2000; Wallingford and Harland, 2001; Goto and Keller, 2002) and the zebrafish (Heisenberg et al.,2000; Sepich et al.,2000; Jessen et al.,2002). The Wnt/PCP pathway is similar to the DrosophilaPCP pathway, which is essential for polarizing cells in the plane of the epithelium of the wing and the eye (Adler,2002; Mlodzik,2002). The core components of the Drosophila pathway include the atypical cadherin-like molecule, Flamingo (also known as Starry night), transmembrane proteins Frizzled and Van Gogh (also known as Strabismus), and cytoplasmic signaling proteins Dishevelled, Diego and Prickle. In Xenopus and zebrafish, perturbation of Dishevelled (Dsh)(Wallingford et al., 2000; Tada et al., 2002), Van Gogh(Darken et al., 2002; Goto and Keller, 2002),Prickle (Veeman et al., 2003),and Wnt5a and Wnt11 (Moon et al.,1993; Heisenberg et al.,2000; Smith et al.,2000) result in failure of CE and the underlying MIB, and in the case of Dsh and Van Gogh, failure of the underlying polarized protrusive activity (Wallingford and Harland,2001; Voiculescu et al.,2007).

Mutations of non-canonical Wnt/PCP pathway genes in the mouse, including Van Gogh (Vangl1/Vangl2 - Mouse Genome Informatics)(Kibar et al., 2001; Murdoch et al., 2001; Montcouquiol et al., 2003; Torban et al., 2008), flamingo (Celsr) (Curtin et al., 2003), and dishevelled (Dvl1/Dvl2)(Wang et al., 2006) lead to short body axes, wide somitic mesoderm and neural plate, and an open neural tube. Mutations in the human gene for Vangl1 are also associated with neural tube defects (Kibar et al.,2007). Mutations in the vertebrate genes Ptk7 and Scrb1 (Scrib - Mouse Genome Informatics), which are not core PCP genes in Drosophila, lead to similar defects and interact genetically with Wnt/PCP genes(Montcouquiol et al., 2003; Lu et al., 2004). These observations suggest that axial extension in the mouse may also occur by MIB,regulated by mechanisms similar to those in frog and fish.

Despite these conserved features, the mechanisms underlying CE in the mammal cannot be extrapolated from frogs, fish or flies, for several reasons. First, several different tissue- and stage-specific types of mediolaterally polarized cell behavior occur during vertebrate CE, and these differ between frogs and fish (Keller et al.,2000; Glickman et al.,2003; Lin et al.,2005). Second, in frogs and fish the active, force-producing cell intercalations occur by polarized protrusive activity among mesenchymal cells,whereas the cell intercalation observed in Drosophila occurs among epithelial cells, not by polarized protrusive activity but by polarized apical junctional remodeling (Keller,2002; Blankenship et al.,2006). In the mouse, unlike frogs and fish, both epithelial(notochordal, neural) and mesenchymal (somitic) tissues undergo CE. Finally,the planar polarity of cell intercalation during CE of the Drosophilagermband does not require the PCP pathway but instead requires A/P stripes of gene expression (Zallen and Wieschaus,2004), whereas CE of the Drosophila hindgut requires unpaired (Upd, also known as outstretched) and STAT(Johansen et al., 2002; Lengyel and Iwaki, 2002). In contrast to its role in vertebrate CE, the Drosophila PCP pathway regulates the polarity of terminally differentiated structures, including the hairs in the wing, bristles on the body and ommatidial cell differentiation(Adler, 2002; Mlodzik, 2002)

Here we show via high-resolution imaging of living embryos that mouse mesodermal CE occurs by MIB, and that the first event in this process is mediolateral elongation and alignment of cells, followed by mediolaterally polarized cell motility and intercalation. Lack of PTK7 leads to failure of the initial polarization and to subsequent failure of CE. These results describe for the first time the cellular mechanisms underlying mammalian CE,and show that some aspects are unique and some are shared with other model systems. This characterization will guide future studies of cell polarity during CE and body axis formation.

Animal-use protocols were reviewed and approved by the University of Virginia Institutional Animal Care and Use Committee, and were in compliance with PHS and USDA guidelines for laboratory animal welfare. Mice of strain Tg(GFPX)4Nagy/J containing an X chromosome-integrated EGFP transgene(Hadjantonakis et al., 2001)were obtained from The Jackson Laboratory (Bar Harbor, ME) and from Dr Terry Magnuson (University of North Carolina, Chapel Hill, NC, USA), and maintained as a heterozygous stock. For labeled Ptk7 mutant mice,Tg(GFPX)4Nagy/J males were crossed with female XST87 mice, which carry a disrupted allele of the Ptk7 gene(Lu et al., 2004). Females carrying both mutant Ptk7 and EGFP-X alleles were backcrossed to XST87 males and the offspring were intercrossed to generate Ptk7^mu/mu EGFP-X^+/- embryos.

Embryos were dissected at 7.5 and 8.5 days post-coitum and those expressing EGFP were staged and oriented posterior side down on glass bottom culture dishes (see Fig. S1 in the supplementary material), secured with Nitex mesh and cultured in whole embryo culture medium [50% rat serum (Harlan Bioproducts for Science, Indianapolis, IN, USA) and 50% Eagle Plus medium(Stephens et al., 1995; Martin and Sutherland, 2001)]in a PDMI-2 stage heater (Harvard Apparatus, Holliston, MA, USA) at 37°C,with 5% CO₂. Sibling embryos in static culture in an incubator served as controls for normal development.

For whole-mount immunostaining, embryos were fixed for 15 minutes in 3.7%formaldehyde, washed in PBS, permeabilized in 0.1% Triton X-100, blocked in 10% FBS/5% BSA overnight at 4°C, then incubated with rhodamine-phalloidin(Sigma, St Louis, MO, USA) and antibodies against brachyury (Santa Cruz Biotechnology, Santa Cruz, CA, USA) and either fibronectin (Millipore,Billerica, MA, USA) or PTK7 (Lu et al.,2004) for 1 hour at room temperature (RT). They were then washed in PBS, incubated with Alexa 350 anti-goat and Alexa 488 anti-rabbit antibodies (1 hour, RT), washed in PBS, mounted and analyzed by multiphoton microscopy at 740 nm excitation. All secondary antibodies were obtained from Invitrogen (Carlsbad, CA, USA). For sectional analysis, embryos were fixed as described above, then washed in PBS, infiltrated with a solution of cold water fish gelatin [15% cold water fish gelatin, 15% sucrose in PBS(Fagotto and Gumbiner, 1994)],embedded, frozen in liquid nitrogen and serially sectioned in the sagittal plane at 7 μm thickness.

EGFP-positive and immunostained embryos were examined using a BioRad Radiance 2100 laser scanning confocal/multiphoton microscope, consisting of: a Nikon TE300 epifluorescence microscope; a 100W mercury arc lamp; a BioRad Radiance 2100 confocal/multiphoton scanning system (Carl Zeiss Microimaging,Thornwood, NY, USA); an Argon Ion laser, a HeNe 543 laser; and a 10W Verdi pumped tunable, from 700-1000 nm (model 900 Mira, www.coherentinc.com),mode-locked ultrafast (78 MHz), pulsed (<150 femtosecond) laser. The system was equipped with a laser spectrum analyzer (Model E201; Mirion Technologies,San Ramon, CA, USA) to monitor the excitation wavelength, and a power meter to measure the laser power at the specimen plane (Model SSIM-VIS & IR;Coherent, Santa Clara, CA, USA) (Chen and Periasamy, 2004). Excitation wavelengths of 488 nm for EGFP and 740 nm for multiphoton imaging were used. Plan Fluor 20× NA 0.75 multi-immersion, Plan Apo 40× NA 1.0 oil and C Apo 40× NA 1.15 water immersion objectives were used for laser scanning confocal image acquisition. LaserSharp2000 software was used to acquire both confocal and multiphoton images using the three internal detectors and also for the transmission detector. Scanning confocal image stacks were taken at 4-minute intervals for 20× time-lapse movies, and at 2-minute intervals for 40× time-lapse movies. Each stack consisted of approximately 17 or 12 sections, respectively, each 4 μm apart. Multiphoton image stacks consisted of approximately 110 sections, 1 or 0.2 μm apart.

Confocal and multiphoton data were analyzed using Volocity and Openlab(Improvision, Waltham, MA, USA) and ImageJ (NIH, Bethesda, MD, USA) software. The images were divided into three regions relative to the posterior notochord(PNC) (Blum et al., 2007) and notochordal plate, with the midline defined by a line through the middle of the PNC and along the notochordal plate. Region 1 is posterior to the PNC and contains cells in the primitive streak (anterior primitive streak), region 2 is lateral to the PNC and contains posterior presomitic mesoderm, and region 3 is lateral to the notochord and posterior to the last-formed somite, and contains anterior presomitic mesoderm. Cells from each region in seven different embryos were traced, and the length of the longest axis and its longest perpendicular axis, and the angle of the longest axis relative to the midline of the image were determined. The angle measurement was corrected for any deviation of the embryonic A/P axis from the image midline. These data were used to calculate the aspect (length to width) ratio and angle of orientation relative to the A/P axis for each cell examined. Repeated measures models were used to compare aspect ratios and distortion measurements of cells across Ptk7 mutant and wild-type (WT) embryos. These models allow for correlations between cells measured from the same embryo. F-tests in these models were used to make specific comparisons on how measurements differed by section in WT and Ptk7 mutant embryos.

Protrusive activity was analyzed for 10-29 cells in each defined sector of the embryo using OpenLab (Improvision). Each protrusion observed over a 4-hour period was marked using the cell-counter tool. The angular distribution of the observed protrusions was determined by counting the number of protrusions that fell into each of 12 sectors (of 30°) of a circle around the cell perimeter, the center of which corresponded to the center of the cell. The 90°/270° axis was aligned with the A/P axis of the embryo. The average distribution of protrusions was determined by combining observations for all cells in each defined sector surrounding the PNC, and were analyzed as described previously (Elul and Keller,2000; Ezin et al.,2003; Ezin et al.,2006), using a χ² test to determine whether the distribution of protrusive activity was significantly different from random and the 2ai test (Baschelet,1981; Zar, 1998)to determine whether they were monopolar or bipolar(Ezin et al., 2003). We computed `preferred directions' of protrusion in a given section of the embryo by comparing the frequency of observed protrusion directions by a cell to that expected by chance, and permutation tests were used to assign a probability to the observed difference. If the probability in one direction was less than 0.05, the cell was designated as preferring one direction or `monopolar'. If the probability in two directions was less than 0.05, and the preferred directions were approximately 180° apart, the cell was designated as`bipolar'. In all other cases, the cell was designated as having no preferred direction, or `multipolar'. χ² tests were used to compare frequency of cells labeled as `monopolar', `bipolar' and `multipolar' across sections.

Radial movement of cells was assessed by counting, during a 2-hour time-lapse sequence, the number of cells that disappeared from or appeared into a 12-16 μm thick region (3-4 summed confocal z-sections)located 12 μm deep to the surface endoderm layer. Between 30 and 34 measurements were taken on regions 2 and 3 of eight WT embryos, and 24 measurements were taken on regions 2 and 3 of six Ptk7 mutant embryos. The measurements were averaged, and analyzed for statistical significance by ANOVA followed by the Student-Newman-Keuls test.

The thickness of the presomitic mesoderm was examined in WT and Ptk7 mutant embryos by the direct measurement of sagittal sections of embryos. The first three sections on either side of the notochord (total of six sections) were selected for each of three embryos of each genotype (WT and Ptk7 mutant), and the region posterior to the last segmented somite was measured in each section. The measurements were averaged and evaluated for statistical significance by ANOVA.

Morphology of WT and Ptk7 mutant embryos at late streak, headfold,and 5-7 somites was compared by laser scanning confocal microscopy of fluorescent marker staining. In WT embryos, the primitive streak forms in the proximal posterior region of the epiblast, and elongates until it reaches the distal tip of the embryo at the late streak stage(Fig. 1A). Elevation of the neural folds and invagination of the foregut pocket follow(Fig. 1A), along with condensation of the somites as pairs of rectangular, nearly square blocks of tissue along both sides of the neural tube(Fig. 1B, S). The elongation of the primitive streak culminates in the formation, at the late streak/early headfold stage, of an oval, bilayered structure composed of columnar epithelial cells located at the anterior end of the primitive streak(Fig. 1B, PS). This structure has commonly been designated as the node, but more recent evidence has corrected its designation to the posterior notochord (PNC)(Blum et al., 2007), which is continuous anteriorly with the notochordal plate(Fig. 1B, not) and floor plate of the neural tube (Sulik et al.,1994). The region surrounding the PNC consists of presumptive somitic cells (Fig. 1B, PreS)(Kinder et al., 1999).

Homozygous Ptk7 mutant embryos were morphologically indistinguishable from WT prior to early somite stages, with normal formation of the primitive streak, head folds and foregut(Fig. 1A). By the 5-somite stage, however, Ptk7 mutants were shorter and wider than WT or heterozygous littermates (Lu et al.,2004), and the PNC and notochordal plate were significantly wider than in WT embryos (Fig. 1B and Fig. 2; see Fig. S2 in the supplementary material). The somites of Ptk7 mutant embryos were shorter in the A/P dimension, wider mediolaterally(Fig. 1B) and compressed into a limited space along the presumptive A/P axis, with the last somite pair formed directly adjacent to the PNC (S) (Fig. 1B). This is substantially more posterior to the position of the last somite pair in WT littermates, making the presomitic mesoderm very short in the A/P axis, and very wide mediolaterally. Somites in Ptk7 mutant embryos did epithelialize and cavitate, but they either formed several,separated cavities or one mediolaterally elongated cavity (asterisks, Fig. 1B,C; see Fig. S2 in the supplementary material).

To examine CE in paraxial mesoderm, we mapped tissue shape changes in three regions: region 1, consisting of the primitive streak region posterior to the PNC; region 2, the presomitic mesoderm lateral to the PNC; and region 3, the presomitic region anterior to the PNC and posterior to the somites. Tissue distortions were mapped by tracking the movements of individual mesodermal cells lying at the intersects of grid patterns over a 4-hour period in time-lapse movies of WT embryos (Fig. 3A-C) and Ptk7 mutant embryos(Fig. 3D-F). In WT and heterozygous embryos, regions 2 and 3 progressively lengthened and narrowed over 4 hours, whereas region 1 remained relatively unchanged(Fig. 3A-C). By contrast, in Ptk7 mutant embryos all three regions retained their initial shape and instead increased in area more-or-less uniformly in all directions(Fig. 3D-F). Thus CE of the presomitic mesoderm is impaired in embryos that lack Ptk7.

Cell shapes differed between the three regions defined above. In region 1 the mean aspect ratio was 1.4, whereas mean aspect ratios were significantly higher in regions 2 and 3, at 2.4 and 2.6, respectively(Fig. 4A). WT presomitic cells also became oriented in region 2. In region 1 the long axes of the cells were randomly oriented relative to the midline(Fig. 4B), but cells in region 2 were oriented with their longest axis at a ∼60° angle relative to the midline (Fig. 4B), whereas in region 3 the orientation of the longest axes of these cells was essentially perpendicular to the midline (Fig. 4B).

By contrast, mesoderm cells in Ptk7 mutant embryos showed no shape change between regions 1, 2 and 3 (Fig. 4A). Aspect ratios for Ptk7 mutant cells in all regions were similar to those of region 1 in WT embryos(Fig. 4A). Ptk7 mutant cells also failed to become oriented; the long axes of the cells in regions 2 and 3 remained randomly oriented (Fig. 4B, Ptk7 mutant). The failure of cell elongation and orientation in regions 2 and 3, regions that also fail to display CE in Ptk7 mutant embryos, strongly suggests that the polarization of cells as they exit the primitive streak is an essential element in the development of CE behavior.

To determine whether polarized protrusive activity and mediolateral intercalation behavior (MIB) underlies CE in mouse mesoderm, we examined protrusive activity and displacement of mesodermal cells in the regions described above. Individual cells were tracked through 5-hour time-lapse projections (12 μm depth) (Fig. 5), and protrusions made by these cells were marked, binned and summed in rose diagrams representing the angular distributions of their frequencies (Fig. 6). The overall rate of protrusive activity was not significantly different between WT(42.54 protrusions/hour/cell) and Ptk7 mutant cells (43.62 protrusions/hour/cell). Cells in region 1 of WT embryos moved by directed migration with high persistence (infrequent change of direction), in an anterior and lateral direction (Fig. 5A; see Movie 1 in the supplementary material). These cells showed multipolar protrusive activity directed nonrandomly in lateral and anterior directions, but were also significantly more likely to exhibit monopolar protrusive activity (Fig. 6A). Thus the behavior of WT cells in region 1 is a canonical directed migration in which the dominant direction of protrusive activity coincides with the direction of cell translocation.

By contrast, cells in regions 2 and 3 moved very differently. At the outset in region 2, the cells showed decreased persistence and reversed their direction in the mediolateral axis relatively often, creating a zigzag pattern of local movement characterized by large lateral and medial excursions and a large number of neighbor changes (Fig. 5B; see Movie 2 in the supplementary material). Cells in region 2 also exhibited multipolar protrusive activity, but were significantly less likely to exhibit monopolar protrusive activity than those in region 1(Fig. 6A). Further anteriorly in region 3, cells showed local, conservative neighbor changes (cell intercalation), without the medial and lateral excursions characteristic of cells in region 2, with persistent anterior displacement and separation along the A/P axis (Fig. 5C; see Movie 3 in the supplementary material). These cells were strongly bipolar in their protrusive activity, and this activity was oriented mediolaterally(Fig. 6). Although they maintained a consistent, anterior-directed displacement, it was produced by mediolaterally biased protrusive activity, mediolateral intercalation and convergence leading to anterior extension, rather than by anteriorly directed migration. In region 3, the behavior of cells is a canonical MIB, where the dominant direction of protrusive activity is perpendicular to the direction of translocation.

In Ptk7 mutant embryos, cells in region 1 showed the same type of anteriorly and laterally directed displacement of high persistence seen in WT embryos, through anteriorly and laterally biased multipolar protrusive activity (Figs 5 and 6). However, this behavior continued unaltered in the compressed regions 2 and 3 lateral to the PNC and posterior to the last segmented somite, and there were no changes in the angular distribution of protrusions between these regions(Fig. 6). All cells examined in the Ptk7 mutant equivalent of region 3 exhibited monopolar or multipolar protrusive activity, similar to region 1(Fig. 6). The cells in regions 2 and 3 continued to move using anteriorly and laterally directed migration,and never exhibited the regional behaviors seen in the WT of either the large medial and lateral excursions (region 2) or the final phase of highly mediolaterally polarized protrusive activity (region 3)(Fig. 5D; see Movie 4 in the supplementary material), nor do they undergo mediolateral cell intercalation(Fig. 5E). To determine whether there were any earlier differences in protrusive activity of mesodermal cells,we examined cells within the primitive streak of WT and Ptk7 mutant embryos at E7.5. At this stage the displacement and the protrusive behavior of cells in Ptk7 mutant embryos were indistinguishable from those of WT embryos (Fig. 6).

Radial intercalation (intercalation perpendicular to the tissue plane) is often paired with mediolateral intercalation (within the tissue plane) to regulate the three-dimensional shape change of the tissue(Keller et al., 2008; Yin et al., 2008). We examined the rate of radial intercalation by determining the number of cells per unit time that disappeared (negative radial intercalation) or appeared (positive radial intercalation) into a 12-16 μm region located 12 μm deep to the surface endoderm. In WT embryos, cells disappeared predominantly in the middle and at the lateral edges of both regions 2 and 3 at an average rate of about six cells per 2-hour period, and appeared at an average rate of less than one cell per 2-hour period, suggesting that the tissue is thickening by generating new, intermediate layers of fewer cells per layer(Fig. 7). In surprising contrast, whereas Ptk7 mutant embryos showed similar numbers of cells disappearing from the observed layer in regions 2 and 3, they had a rate of appearance of almost six cells per 2-hour period(Fig. 7). Thus, whereas WT presomitic mesoderm has a net negative radial intercalation, Ptk7mutant presomitic mesoderm has no net change in the number of cells within the plane of the observed region, which should lead to a thinner tissue of greater area in the mutant embryo.

To examine this issue, we measured the thickness of the presomitic mesoderm at three separate points within the first 25 μm on either side of the notochord (total of six measurements per embryo, three embryos each of WT and Ptk7 mutant). The average thickness of the presomitic mesoderm in the Ptk7 mutant embryo (75.3 μm) was indeed significantly less than that of the WT (112.8 μm) (Fig. 8).

Activation of the non-canonical Wnt/PCP pathway is thought to involve localization of DVL to the plasma membrane(Wallingford et al., 2000),and this is supported by findings that DVL localization is altered in PCP mutants (Yin et al., 2008) and after impairment of PTK7 function(Shnitsar and Borchers, 2008). However, staining of embryos with antibodies against DVL2, which, of the three DVL proteins in the mouse, plays a dominant role in neural tube closure(Wang et al., 2006; Etheridge et al., 2008),revealed no differences in the localization of DVL2 to the plasma membrane between WT and Ptk7 mutant embryos in the axial or paraxial mesoderm,or in the primitive streak (see Fig. S3 in the supplementary material). Nor were any differences seen in the localization of actin, myosin IIa, or myosin IIb in the presomitic mesoderm of the Ptk7 mutant or WT embryos (see Fig. S3 in the supplementary material). Thus PTK7 does not appear to affect gross aspects of DVL2 distribution or of cytoskeletal organization.

The results presented here characterize for the first time the normal pattern of mesodermal cell behavior during CE of the mouse body axis. Prospective somitic mesoderm cells initially exhibit multipolar, biased protrusive activity resulting in migration laterally and anteriorly from the anterior primitive streak, but then elongate and aligns with the presumptive mediolateral axis of the tissue and polarized their protrusive activity mediolaterally, leading to a high degree of local mediolateral intercalation and CE. Throughout this progression the cells are displaced anteriorly;initially anterior displacement are driven directly by anteriorly directed protrusive activity, and subsequently are driven indirectly by mediolaterally polarized protrusive activity and MIB.

MIB in the mouse increases progressively along the A/P axis, and the major transition in polarized cell behavior occurs lateral to the PNC as the cells leave the streak, paralleling the post-involution transition to MIB in the frog. It results in CE of the tissue, and occurs in the same relative location and orientation with respect to the notochord and last-segmented somites(Wilson et al., 1989; Shih and Keller, 1992b). However, MIB in the mouse mesoderm follows an epithelial to mesenchymal transition (EMT) in the streak, and anterior- and lateral-directed cell migration, behaviors not seen prior to MIB in the frog. In the zebrafish,presumptive somitic mesoderm undergoes a post-involution, dorsally directed,Gα12/13-dependent migration that is followed by a Wnt/PCP-dependent MIB and CE (Sepich et al., 2000; Lin et al., 2005; Sepich et al., 2005; Yin et al., 2008).

Unique to mouse MIB are the significant elongation and alignment of cells prior to the onset of strongly biased protrusive activity. In the frog, the mediolateral bias of protrusive activity occurs simultaneously with mediolateral cell elongation and alignment(Shih and Keller, 1992a; Shih and Keller, 1992b), and the causal relationship between the two has never been determined. Extreme cell elongation occurs in explants in which intercalation is mechanically frustrated, arguing that polarized traction on other cells or extracellular matrix puts the cells under mediolateral tension, thus driving elongation(Keller et al., 1992b; Keller and Shook, 2008), but the initial elongation may still be due to internally driven cell shape change rather than polarized traction.

Additionally, mouse somitic mesoderm shows a phase of `zigzag' movement, in which cell movement is characterized by unusually large medial and lateral excursions with an overall anterior trajectory, a behavior not seen in Xenopus mesoderm. However, Xenopus deep neural plate cells express an abnormal, bipolar mode of intercalation under experimental conditions in which their bipolarity is statistically balanced over time, but tends to be unbalanced in the short term, resulting in excursions first one way and then the other (Elul et al.,1997; Elul and Keller,2000; Keller et al.,2008). We interpret the zigzag behavior in the mouse as being a similar temporary imbalance, as the cells reorient their protrusive activity from the predominantly multipolar/monopolar distribution of the primitive streak to the definitive bipolar distribution driving MIB.

The roles that these cell behaviors play in driving axial extension are clarified in the Ptk7 mutant embryo. In the absence of PTK7, the anteriorly directed migration of the somitic mesoderm is unaffected, whereas the strong MIB and CE seen in region 3 of the WT do not occur in the Ptk7 mutant. This suggests that the initial migration behavior,unlike the dorsally directed migration in fish mesoderm, does not contribute to CE, and that MIB is the primary cellular mechanism of mouse mesodermal cell intercalation and CE. Whether the anterior-directed migration of mouse paraxial mesoderm is mechanistically similar to the Gα12/13-mediated dorsal migration of zebrafish cells (Lin et al., 2005), or to the FGF-dependent migration of chick mesodermal cells (Yang et al.,2002), remains to be seen.

The failure of polarized cellular behaviors and MIB in the absence of PTK7 leads to a shorter, wider body axis, similar to Vangl2 or Dvl1 Dvl2 mutant mouse embryos, in which defects in neural tube and notochord elongation and CE have been described(Wang et al., 2006; Ybot-Gonzalez et al., 2007). Vangl2 and Ptk7 do interact genetically, as compound heterozygotes exhibit a more severe phenotype than either single heterozygote(Lu et al., 2004). However,the phenotype of the compound heterozygote is also unlike that of either single homozygote, suggesting that the two genes may act in separate pathways that both affect axial extension and neural tube closure. In this context it is interesting that the Ptk7 mutation appears to have a stronger effect on paraxial mesoderm cell behavior than either the Vangl2mutation or the Dvl1 Dvl2 double knockout. Embryos homozygous for the Ptk7 mutation have obvious defects in axial elongation and somite morphology by the 5-7 somite stage, whereas Vangl2 mutant embryos exhibit normal axial elongation during this period(Gerrelli and Copp, 1997), and Dvl1 Dvl2 double mutants have a shortened axis but have comparatively normal somite morphology (Wang et al.,2006). This may be due to compensating functions of VANGL1 and DVL3 proteins in the paraxial mesoderm, or it may be that PTK7 has a unique activity in the paraxial mesoderm. Such is the case in the zebrafish for the Gα12/13 proteins, which function earlier than the core Wnt/PCP proteins in a pathway parallel to the Wnt/PCP pathway, but affect cell behavior and tissue morphogenesis similarly to the core Wnt/PCP proteins(Lin et al., 2005).

The lack of PTK7 has a dramatic effect on axial morphogenesis, but a discrete effect on individual cell behavior: PTK7 specifically affects the polarization of the presomitic mesoderm relative to the midline of the embryo,and the subsequent onset of MIB. The directed migration and protrusive behavior of cells within the primitive streak is unaffected, and the prospective somitic mesoderm cells exhibit normal directed migration as they leave the streak. The lack of PTK7 does not affect cell fate or somitogenesis;somites condense and epithelialize similarly to WT, albeit with a different(wider) morphology owing to the lack of CE. These findings argue that PTK7 is a specific regulator of cell polarity rather than a general regulator of cell motility or cell fate.

The molecular mechanisms underlying this polarization process are not yet understood. PTK7 is a pseudokinase and thus is unlikely to be signaling directly by phosphorylation (Boudeau et al., 2006). A recent study implicated the kinase domain of PTK7 in the formation of a membrane-bound complex with Dsh and Frizzled 7 in the frog(Shnitsar and Borchers, 2008),whereby activation of the PCP pathway involves membrane localization of Dsh(Wallingford and Harland,2002). However, DVL does not have a polarized distribution in mouse neuroepithelium, nor is its membrane localization affected by mutations in Vangl2 (Wang et al.,2006). We also find that DVL2 distribution is not polarized in mouse mesodermal cells, and its membrane localization is not affected by PTK7. The absence of gross changes in the actomyosin cytoskeleton of PTK7 mutants reported here also suggests that PTK7 has a very specific function in polarity. Clearly much additional work is necessary to understand the molecular mechanisms regulating Wnt/PCP signaling during axial elongation, the potential mechanistic differences between species and the molecular mechanisms underlying cell polarization by PTK7.

The abnormal thinning of the paraxial mesoderm in the Ptk7 mutant suggests that forces generated by radial intercalation normally oppose the tendency for convergence forces to thicken, as well as elongate, the tissue. In the frog, several lines of evidence suggest that tissue convergence, driven by mediolateral intercalation, produces compressive forces both along the A/P axis to produce extension, and in the radial axis to produce thickening(Keller and Tibbetts, 1989; Wilson and Keller, 1991; Keller et al., 2008). Such thickening is opposed by a tissue-thinning force generated by radial intercalation, and its magnitude regulates the relative amounts of extension and thickening that result from convergence(Keller and Tibbetts, 1989; Wilson and Keller, 1991; Keller et al., 2008). Alternatively, A/P polarization of radial intercalation can lead synergistically to increased extension, as shown in a recent study of the paraxial mesoderm of the zebrafish (Yin et al., 2008). Our results show that in the mouse the balance of radial movement in the presomitic mesoderm is negative, thus the tissue normally undergoes substantial radial thickening at the same time as it is converging and extending. In the absence of PTK7 there is no net radial movement of presomitic mesoderm cells, leading to abnormal thinning and lateral expansion of the tissue. As a result, the presomitic mesoderm and somites are thinner in the dorsal-ventral axis, broader in the mediolateral axis, and shorter in the A/P axis. Unfortunately, owing to the nature of the EGFP expression in the embryos used in this study we could not observe the relative position of all of the cells surrounding those that were moving in the radial direction, and thus were unable to determine whether the radial intercalation was polarized in either the A/P or mediolateral axis, and this remains as an important question to be addressed.

In summary, mesodermal CE in the mouse embryo is driven by polarized cell protrusive activity, the polarity of which is established through a PTK7-dependent event that occurs as the cells leave the primitive streak. Without this polarization the mesoderm cells fail to polarize their protrusive activity and develop MIB, and continue instead to undergo directed migration laterally and anteriorly. Concomitant, unopposed radial intercalation further broadens the somitic mesoderm field, and the resulting somites are mis-shapen in all axes as well as malpositioned along the A/P axis, leading to a shortened body axis posterior to the midbrain.