PCP Signaling between Migrating Neurons and their Planar-Polarized Neuroepithelial Environment Controls Filopodial Dynamics and Directional Migration

The planar cell polarity (PCP) pathway is a cell-contact mediated mechanism for transmitting polarity information between neighboring cells. PCP “core components” (Vangl, Fz, Pk, Dsh, and Celsr) are essential for a number of cell migratory events including the posterior migration of facial branchiomotor neurons (FBMNs) in the plane of the hindbrain neuroepithelium in zebrafish and mice. While the mechanism by which PCP signaling polarizes static epithelial cells is well understood, how PCP signaling controls highly dynamic processes like neuronal migration remains an important outstanding question given that PCP components have been implicated in a range of directed cell movements, particularly during vertebrate development. Here, by systematically disrupting PCP signaling in a rhombomere-restricted manner we show that PCP signaling is required both within FBMNs and the hindbrain rhombomere 4 environment at the time when they initiate their migration. Correspondingly, we demonstrate planar polarized localization of PCP core components Vangl2 and Fzd3a in the hindbrain neuroepithelium, and transient localization of Vangl2 at the tips of retracting FBMN filopodia. Using high-resolution timelapse imaging of FBMNs in genetic chimeras we uncover opposing cell-autonomous and non-cell-autonomous functions for Fzd3a and Vangl2 in regulating FBMN protrusive activity. Within FBMNs, Fzd3a is required to stabilize filopodia while Vangl2 has an antagonistic, destabilizing role. However, in the migratory environment Fzd3a acts to destabilize FBMN filopodia while Vangl2 has a stabilizing role. Together, our findings suggest a model in which PCP signaling between the planar polarized neuroepithelial environment and FBMNs directs migration by the selective stabilization of FBMN filopodia.


Introduction
The Planar Cell Polarity (PCP) signaling pathway is best understood as a cell contact dependent mechanism for generating and maintaining polarity in the plane of an epithelium [1,2]. Its function was first described in the static epithelial cells of the fly where the molecular asymmetry of "core" PCP proteins results in the morphological asymmetry of a single actin-rich hair at the distal side of each wing cell [3][4][5]. Subsequently, planar polarity established by the core pathway has been shown to be a characteristic of many epithelial tissues in vertebrates and invertebrates alike [6][7][8][9][10]. The core PCP pathway is comprised of two protein complexes that localize to distinct cell membranes. In the fly wing, the transmembrane protein Frizzled (Fz) is confined to distal apical cell junctions along with the cytosolic proteins Disheveled (Dsh) and Diego (Dgo), while the transmembrane protein Van Gogh (Vang) (Strabismus(Stbm)) and the cytosolic protein Prickle (Pk) are proximally localized. This molecular asymmetry of PCP promotes actin polymerization at the distal side of the cell, downstream of Fz and Dsh [11][12][13]. While the factors that initially polarize PCP components are context dependent [14], the asymmetric localization of PCP proteins is maintained within polarized cells via intracellular destabilizing interactions between the Vang complex and the Fz complex [15,16]. This polarization of PCP proteins is coordinated between cells by the formation of intercellular stabilizing interactions between Vang and Fz complexes across cell junctions [17][18][19][20][21]. In spite of the antagonistic roles of Vang and Fz complexes, loss of function of any core PCP component results in a loss of polarity.
The PCP pathway drives the stereotyped tangential migration of facial branchiomotor neurons (FBMNs) in the vertebrate hindbrain. FBMNs are a subset of cranial branchiomotor neurons that originate ventrally in rhombomere (r)4 and undergo a posterior migration to r6 where they form the facial motor nucleus, whose axons exit the hindbrain in r4 and innervate muscles derived from the second branchial arch [45,47]. Forward genetic screens in the zebrafish have identified multiple core PCP components (Vangl2, Pk1b, Fzd3a, Celsr2 and Scribble) as being required for FBMN migration [31,[48][49][50][51]; this PCP requirement has also been shown for mouse FBMN migration [52][53][54]. Unlike the cell migrations mentioned above, screens have failed to identify a role for Wnts or other chemotactic cues. Although it is clear that many components of the PCP pathway are required for tangential FBMN migration, how these components regulate this highly dynamic process is unknown.
As a first step in answering this question we defined the cell types participating in PCP signaling during FBMN migration, as previous studies using a range of approaches have yielded conflicting results [31,48,49,51,55]. Using the Gal4/UAS system to systematically disrupt PCP in a cell-type and rhombomere-specific manner, we demonstrate the dual requirement for PCP within FBMNs and the planar-polarized r4 neuroepithelial environment in which they arise, and identify reciprocal PCP-dependent interactions between FBMNs and the planarpolarized floorplate as being sufficient, though not required, to promote migration. Since cell migration results from the contact-dependent stabilization of cellular protrusions and PCP signaling is known to regulate actin dynamics, we examined the protrusive activity of single FBMNs using high-resolution single-cell time-lapse microscopy in chimeric embryos and demonstrate opposing functions for the PCP core components Fzd3a and Vangl2 in regulating FBMN filopodial protrusive activity in vivo. Within FBMNs we show that Fzd3a is required to stabilize filopodia while Vangl2 has an antagonistic, destabilizing role. However, in the migratory environment we show that Fzd3a is required to destabilize filopodia while Vangl2 has a stabilizing role. In spite of having antagonistic roles at the cellular level, Vangl2 and Fzd3a mutants have the same FBMN migration phenotype. These findings are thus reminiscent of the intracellular antagonistic versus intercellular stabilizing roles that core PCP proteins perform in stably polarized epithelia. Consistent with a role for Vangl2 in regulating filopodial dynamics, we show that Vangl2 localizes transiently to the tips of retracting FBMN filopodia; consistent with a role for Vangl2 and Fzd3a in the microenvironment, we show planar polarized localization of these proteins in the adjacent floorplate. Together, our findings support a model in which canonical interactions between PCP components within FBMNs and between the FBMNs and their planar polarized neuroepithelial environment promotes migration via the selective stabilization of FBMN filopodia.

PCP signaling within FBMNs is required for their migration
Initial chimeric analyses suggested that the PCP components Vangl2, Fzd3a, Celsr2 and Scrib primarily act non-cell-autonomously to regulate FBMN migration [31,48,49]. An additional cell-autonomous role for Vangl2 and Scrib in FBMN migration has been demonstrated [51], but refuted by others [55]. To determine whether PCP signaling is required cell-autonomously within FBMNs for their migration, we expressed a dominant negative (DN) form of the PCP core component Dvl specifically in branchiomotor neurons using the islet-1 (isl1) CREST enhancer (Fig 1A, 1B and 1C) [56]. Dvl is the branching point between multiple Wnt signaling pathways, and the overexpression of its individual domains exert pathway-specific DN properties [57]. Work in multiple vertebrate systems has demonstrated that Xdd1 and Dvl-DEP, two truncated forms of Dvl, act as PCP-specific DNs [25][26][27].
In previous studies mRNA injection of these DNs failed to disrupt zebrafish FBMN migration [31,54]. We reasoned that this could be due to decreased DN mRNA levels or activity by the time of FBMN migration. To stably express DN forms of Dvl in FBMNs we raised stable Tg  isl1:Dvl-DEP-GFP) zebrafish in which FBMNs express Dvl-DEP-GFP. In wild type embryos, FBMNs fully migrate to r6 by 48 hours post fertilization (Fig 1B). However Dvl-DEP-GFP expressing FBMNs largely fail to migrate, with 31/35 of Tg(isl1:Dvl-DEP-GFP) embryos displaying FBMN migration defects where most FBMNs (>75%) remain in in r4 ( Fig 1C). This demonstrates that PCP signaling within FBMNs is required for their migration.
To further confirm this, and to test specifically whether the core transmembrane PCP component Fzd3a, like Vangl2 [51], is required within FBMNs for migration, we used chimeric analysis to assess the ability of fzd3a rw689 mutant FBMNs to migrate in a normal planar polarized neuroepithelium. In these experiments we prevented host FBMN migration using a pk1b morpholino since it is well known that migrating FBMNs can carry other FBMNs with them independent of PCP signaling, complicating the interpretation of chimeras [51,58]. pk1b morphants precisely phenocopy pk1b mutants in which FBMNs fail to migrate even though the surrounding neuroepithelium can support wild type FBMN migration [50,51,59]. While 70.9% of wild type FBMNs migrate out of r4 in a pk1b morphant environment, only 19% of fzd3a mutant FBMNs do so (S1 Fig). This suggests that Fzd3a is required within FBMNs for migration.
Taken together, the disruption of migration due to FBMN-restricted DN expression, our chimeric analysis of fzd3a -/and previous chimeric analysis of vangl2 -/-FBMNs [51] and the failure of FBMN migration after FBMN-specific disruption of Vangl2 in the mouse confirms a FBMN-autonomous requirement for PCP signaling in migration.

FBMN migration requires PCP signaling non-autonomously in r4
While these data support a cell-autonomous requirement for PCP signaling in FBMN migration, PCP signaling in FBMNs is not sufficient for their migration. Indeed, a non-autonomous requirement for PCP signaling in FBMN migration has been well established in chimeras in which wild type FBMNs are unable to migrate in vangl2, fzd3a, celsr2 or scrib mutant hosts [31,48,49,51]. Since PCP is a cell-contact mediated signaling pathway in which the same transmembrane protein components are required in both contacting cells [2], an attractive hypothesis is that FBMNs receive PCP cues from cells in their environment that promote or direct their migration. Thus we sought to determine where PCP signaling is required in the FBMN migratory path for migration.
To block PCP signaling in distinct compartments of the hindbrain, we used the Gal4/UAS system to drive rhombomere-restricted expression of Tg(UAS:Xdd1-GFP) as well as a C-terminally truncated Fzd3a Tg(UAS:Fzd3aΔC-GFP) that lacks its cytoplasmic region, which has been shown to function as a potent PCP DN tool in zebrafish [49]. We used Tg(egr2b:KalTA4) to drive expression throughout r3 and r5 starting at 12 hpf [63] and Tg(hoxb1a:Gal4) [64] to drive expression throughout r4 starting at 10 hpf ( Fig 1A). Expression of Xdd1-GFP or Fzd3aΔC-GFP along the migration path in the r5 neuroepithelium does not affect migration (Fig 1G and 1H). In contrast, r4-restricted expression of Xdd1-GFP or FzdΔC-GFP completely blocks FBMN migration (Fig 1I and 1J). This suggests that PCP signaling is required at the onset, but not throughout the course of FBMN migration. However we note that r5 expression of Xdd1-GFP or Fzd3aΔC-GFP using egr2b:KalTA4 comes on slightly later than r4 expression using hoxb1a:Gal4 (12 hpf compared to 10 hpf), so the caveat remains that PCP signaling is not fully disrupted in r5 at the time of migration with the available tools.
It was not surprising that FBMNs fail to migrate in Tg(hoxb1a:Gal4); Tg(UAS:DN-GFP) embryos given that FBMNs arise in r4, and thus express hoxb1a throughout their early development, and we had already shown a cell-autonomous requirement for PCP signaling within FBMNs. To assess whether PCP signaling plays a role in the r4 neuroepithelium outside of FBMNs, we transplanted wild type Tg(isl1:mRFP) donor FBMNs into the presumptive ventral hindbrain of Tg(hoxb1a:Gal4); Tg(UAS:Xdd1-GFP) embryos and assessed the positions of donor-derived FBMNs at 48 hpf. In control hosts, 87% (328/378) of wild type donor-derived FBMNs migrated out of r4. In contrast, in hosts expressing Xdd1-GFP in r4, only 17% (33/190) of donor-derived wild type FBMNs migrate out of r4 (Fig 1K, 1L and 1M). Thus, expression of Xdd1-GFP throughout r4 significantly hinders wild type FBMNs from initiating migration (p<0.0001, χ 2 = 207.8) (Fig 1M). This demonstrates that there is a non-autonomous requirement for PCP signaling for FBMN migration in r4.

Environmental PCP signaling is required for the migration of post-mitotic FBMNs
Having established that PCP signaling is required both within FBMNs and their r4 neuroepithelial environment for migration to occur; we asked when this signaling is required. PCP signaling polarizes neuroepithelial progenitors before FBMNs differentiate [7,65,66]. It is possible that this early neuroepithelial polarity is maintained in FBMNs to orient their initial migration. Alternatively, PCP signaling active at the time of migration initiation may promote migration. We reasoned that in the former case a planar polarized environment would not be required for migration after FBMNs had differentiated while in the latter case PCP function in the r4 environment would continue to be essential for migration.
To determine when PCP signaling is required for FBMN migration, we transplanted a small number (1-5) of pre-migratory but post-mitotic FBMNs directly from r4 of a Tg(isl1:GFP) donor into r4 of a stage-matched wild type or vangl2 m209 mutant Tg(isl1:mRFP) host (Fig 2A). During this extraction procedure, transplanted cells round up and become separated in the transplant pipette and are unlikely to retain polarity information. Nevertheless, 28% (48/174) of surviving post-mitotic FBMNs transplanted into a wild type host r4 migrated to r6 (Fig 2B  and 2D). To rule out the possibility that the transplanted FBMNs are simply being carried by migrating host FBMNs, we transplanted post-mitotic FBMNs into a pk1b mutant host, which has normal neuroepithelial PCP but no host FBMN migration; in this environment 11%, (17/ 152) of transplanted FBMNs migrated (S2 Fig). This suggests that host neurons do contribute to migration [51,58], but that post-mitotic transplanted FBMNs can migrate without contribution from migrating host neurons. Importantly, 0% (0/73) of FBMNs migrated to r6 after being transplanted into a vangl2 mutant host (Fig 2C and 2D). Together, these results suggest that post-mitotic FBMNs engage PCP signaling as they initiate their migration out of r4.

Floorplate PCP is not required for FBMN migration
FBMNs migrate in the ventral neural tube adjacent to the floorplate (Fig 1A, [49,67]) making the floorplate a potential source of PCP signaling for FBMN migration. A recent report found that floorplate expression of Vangl2 is both necessary and sufficient for FBMN migration [55]. Here, to investigate whether PCP signaling in the floorplate is required for FBMN migration, we generated a Tg  With the caveat that this effect on floorplate planar polarity was scored after FBMN migration was complete (48 hpf) rather than at the onset of migration (18-24 hpf), DN expression in the floorplate in the floorplate had no effect on FBMN migration (Fig 3A and 3B). To confirm this, we specifically knocked Vangl2 out in the mouse floorplate using the floxed Vangl2 allele described above [61] and Shh-driven Cre recombinase [68]. We found that Cre-mediated deletion of Vangl2 in the mouse floorplate does not disrupt FBMN migration (S4 Fig). These results suggest that PCP signaling in the floorplate is not required for FBMN migration, and point to the possibility that loss of PCP in the floorplate can be compensated for by other planar polarized cells in the r4 neuroepithelial environment.
Floorplate PCP could nevertheless be sufficient to rescue FBMN migration as has been suggested [55]. We tested the sufficiency of Vangl2 in the floorplate for FBMN migration in two ways. We expressed a GFP-Vangl2 fusion protein specifically in the floorplate of vangl2 mutants and wild type siblings using stable Tg(shh:Gal4) driver and Tg(UAS:GFP-Vangl2) transgenic lines (vangl2 m209/m209 ; Tg(shh:Gal4); Tg(UAS:GFP-Vangl2)). Although GFP-Vangl2 was expressed broadly in the floorplate in these otherwise mutant embryos starting at 14 hpf, and exhibits planar-polarized localization (S3A Fig and see below), it neither disrupted FBMN migration in a wild type embryo nor rescued migration in a vangl2 mutant embryo (Fig 3C and  3D). Since a caveat of this experiment is that Vangl2 over-expression can itself disrupt planar polarity, we used targeted transplantation of wild type cells into the floorplate of vangl2 mutants to test whether floorplate Vangl2 is sufficient to rescue FBMN migration. We never observed rescue of host FBMN migration in vangl2 mutant Tg(isl1:mRFP) hosts with wild type donor-derived cells in the hindbrain floorplate ( Fig 3E). This includes 9 cases with 10 or more wild type floorplate cells in rhombomere 4. This is contrary to the findings of Sittarmane et al. (2013) [55] who found that a single wild type floorplate cell in r4 of a vangl2 mutant could rescue FBMN migration. Together, our findings suggest that Vangl2 function in the floorplate is not sufficient for FBMN migration.
In these transplant experiments we noted that FBMNs as well as floorplate cells differentiate from donor-derived cells. This is not unexpected, given the close proximity of floorplate and branchiomotor neuron progenitors in the early embryo [70]. Interestingly, we observed that unlike the mutant host FBMNs, wild type donor-derived FBMNs sometimes migrate (Fig 3F), and their ability to do so correlates with the number of wild type cells in the hindbrain floor plate (R 2 = 0.244; p = 0.005). We conclude that Vangl2 function in the floorplate is not sufficient for FBMN migration, but that Vangl2 function in the floor plate can support the migration of vangl2 + FBMNs in an otherwise vangl2 mutant neuroepithelium. Taken together, we conclude that the floorplate can serve as a source of PCP signals for FBMN migration, but other cells in the r4 environment, which are also planar polarized (see below) can compensate for the loss of normal floorplate PCP signaling.

Vangl2 localization in the migratory environment
Thus far, we have shown that PCP signaling in FBMNs and their immediate neuroepithelial/ floorplate r4 environment can drive migration. The localization of core PCP components is known to be crucial for many PCP mediated processes [2,22]. Therefore, to better understand how PCP signaling might be used in neuronal migration we asked where PCP proteins localize within FBMNs and in their neuroepithelial microenvironment. Using a polyclonal antibody against zebrafish Vangl2, we observed localization of Vangl2 to cell membranes throughout the In the r4 floorplate, we noted a 1.6-fold enrichment of Vangl2 protein at anterior/posterior membranes of floorplate cells compared to their lateral membranes (Fig 4A and 4B). Co-staining with ZO1 shows that this staining is sub-apical, at the level of the tight junctions (Fig 4A'). In order to distinguish anterior from posterior membrane localization we mosaically expressed GFP-Vangl2 in the floorplate so we could visualize Vangl2 localization in isolated floorplate cells. This revealed that Vangl2 is specifically enriched at the anterior subapical membrane (Fig 4C and 4E). The normalized mean fluorescent intensity ratio of GFP-Vangl2 at the anterior membrane versus the posterior membrane in expressing floorplate cells is 2.2 (std. deviation 0.9; N = 29 cells in 8 embryos). Conversely, Fzd3a-GFP is enriched at the posterior membrane ( Fig 4D). These findings for PCP protein localization in the floorplate are consistent both with the requirement for PCP core components in the posterior localization of the floor plate primary cilium [6], and with a conserved deployment of PCP core components in vertebrate and invertebrate epithelia.
While the regular organization of floorplate cells makes it easy to visualize their planar polarization, our findings suggest that the primary source of environmental PCP signaling in FBMN migration comes from neuroepithelial progenitor cells outside of the floorplate. Previous studies demonstrated a planar polarization of GFP-Pk and GFP-Vangl2 in neuroepithelial progenitor cells during zebrafish neurulation [7,65], and of endogenous Vangl2 in the Xenopus neural plate [66]. We asked whether neuroepithelial progenitor cells display planar polarization of Vangl2 in r4 at the time of FBMN migration. Using the Tg(hoxb1a:Gal4) line we mosaically expressed GFP-Vangl2 and observed a subtle but significant asymmetry of GFP-Vangl2 to the anterior sub-apical side of r4 neural progenitors. While GFP-Vangl2 polarization is subtle and not detectable in all expressing neuroepithelial progenitors, in blinded experiments we were able to correctly guess the A-P orientation of embryos based exclusively on GFP-Vangl2 localization in r4 progenitors in 18/23 mosaically expressing embryos (p = 0.004 that 18/23 correct guesses were due to chance alone). The normalized mean fluorescent intensity ratio of GFP-Vangl2 at the anterior membrane versus the posterior membrane in cells where asymmetry is detectable is 1.82 (std. deviation 0.47, N = 17 embryos, 23 cells.) ( Fig 4F). Thus both the r4 neuroepithelium and floorplate exhibit planar polarized Vangl2 localization at the time of FBMN migration.

Vangl2 is enriched at the tips of retracting FBMN filopodia
We next sought to determine where Vangl2 localizes in migrating FBMNs. Endogenous Vangl2 in FBMN membranes and the membranes of surrounding cells could not be resolved using the anti-Vangl2 antibody and, unlike static floorplate cells and neuroepithelial progenitors, FBMNS are highly dynamic, extending primarily filopodia-like protrusions as they migrate [51,71]. Reasoning that Vangl2 localization would be similarly dynamic, we mosaically expressed GFP-Vangl2 in FBMNs and visualized localization using spinning disc time-lapse imaging. We found that GFP-Vangl2 localizes throughout the membrane as well as in putative cytoplasmic vesicles, as is predicted for a transmembrane protein (Fig 5A). However, in addition to its membrane localization, we observe transient enrichment of GFP-Vangl2 at the tips of a subset filopodia immediately preceding filopodia retraction (Fig 5A'-5C and S1 Movie). Before enrichment the mean fluorescent intensity ratio of GFP at the filopodia tip versus the filopodia base is approximately 1 (0.99 ± 0.01), as is the case for mRFP (background membrane marker) (0.92 ± 0.02). During the enrichment event, this ratio for GFP-Vangl2 increased to 1.31 ± 0.05 while the ratio for mRFP remained close to 1 (0.97 ± 0.02) (Fig 5B). Since the ratio for mRFP remained close to 1, this suggests that the enrichment of GFP-Vangl2 correlates with increased Vangl2 protein levels at filopodia tips and not simply condensation of the membrane due to retraction. Furthermore, as described below, addition of exogenous GFP-Vangl2 in FBMNs results in a reduced filopodial lifetime which is opposite to the effect of loss of Vangl2 in FBMNs. This suggests that exogenous GFP-Vangl2 is functioning in FBMNs and that this observed localization of Vangl2 at the tips of filopodia is correlated with retraction. This enrichment of GFP-Vangl2 in filopodia never lasted for more than one time-point (images were taken at 30-45 second intervals) and was only detected in a subset of filopodia (N = 11/84 filopodia on 8 neurons in 7 embryos); it is likely that due to the transient nature of enrichment events and the constraints of our imaging rate we failed to observe many enrichment events. Importantly, however, the enrichment events we captured invariably preceded filopodial retraction; filopodia never extended further after an enrichment event (Fig 5C). Consequently, we infer that Vangl2 may function in FBMN filopodia to signal retraction events.
Vangl2 and Fzd3a function cell-autonomously to regulate FBMN filopodial activity in an antagonistic manner Our findings that PCP signaling is required within FBMNs for migration, and that Vangl2 localizes transiently to the tips of retracting filopodia, suggested the possibility that PCP signaling influences filopodial dynamics in migrating neurons in vivo. In order to determine the cellular basis of FBMN migration defects in PCP mutants, we imaged the protrusive dynamics of single mutant FBMNs at high resolution in vivo. Previous studies have described membrane protrusions in fixed or live embryos expressing cytoplasmic GFP or membrane-RFP in bulk FBMNs at low time resolution, however the overlap between FBMNs allows only a subset of protrusions to be visualized and their dynamics could only be inferred from distant time points [31,55,58,71]. To visualize the protrusive activity of single FBMNs at high time resolution, we utilized cell transplantation to generate embryos in which one or a few FBMNs express membrane-localized teal fluorescent protein (Tg(isl1:mTFP)), and imaged protrusion dynamics of single FBMNs at 30-second intervals, the shortest interval at which we could acquire comprehensive z-stacks on our instruments. We focused on the function of Vangl2 and Fzd3a, the mutually antagonistic transmembrane core components, whose localized activity is both the hallmark and the driver of classical epithelial planar polarity [2].
Time-lapse imaging of FBMN membranes and f-actin dynamics revealed that filopodia are the prevalent protrusion type in FBMNs (Fig 6A-6E  To characterize protrusive membrane dynamics we quantified filopodial lifetime (number of seconds each filopodium is present during a 15 minute time-lapse period) and filopodial maximum length (the greatest length during the lifetime of filopodia lasting 90 seconds or longer) of Tg(isl1:mTFP) FBMNs. Wild type FBMNs generate filopodia with an average lifetime of 224.5 ± 18.66 (SEM) seconds and an average maximum length of 3.6 ± 0.3 μm (Fig 6A, 6F and 6G and S3 Movie). This filopodial lifetime is comparable to that observed in other vertebrate cells both in vivo and in culture [72][73][74][75]. When compared to wild type, FBMNs in vangl2 mutant embryos have much more stable filopodia with a longer average lifetime of 537.3 ± 81.78 seconds (Fig 6F, S4 Movie; p = 0.0059). Filopodia of these vangl2 mutant FBMNs also reach a greater average maximum length of 6.4 ± 1.1 μm (Fig 6G; p = 0.0406). We saw a similar trend when we used microinjection rather than transplantation to mosaically express mTFP in FBMNs in wild type and vangl2 mutant embryos. These results suggest that Vangl2 is required to destabilize FBMN membrane protrusions. Previous studies have demonstrated that FBMNs in vangl2 mutants move more slowly than wild type FBMNs and in random directions, which is consistent with Vangl2 being required to destabilize protrusions [31].
Since Vangl2 is required within FBMNs and their r4 microenvironment for migration, we sought to determine where Vangl2 functions to regulate filopodia dynamics. To determine the cell-autonomous function of Vangl2, we transplanted vangl2 mutant FBMNs into a wild type host. Donor embryos carried the Tg(isl1:mTFP) transgene to visualize FBMNs and contained rhodamine dextran to track other donor-derived cells so we could ensure that donor-derived FBMNs were in fact in a genetically chimeric environment (S7 Fig). We found that vangl2 mutant FBMNs in a wild type environment have longer, more stable filopodia with a mean lifetime of 432.0 ± 55.65 seconds and a maximum length of 6.8 ± 0.8 μm, similar to vangl2 mutant FBMNs in a vangl2 mutant host (Fig 6B, 6F and 6G and S5 Movie; p = 0.005 and p = 0.0078 respectively). To further test if Vangl2 functions cell-autonomously to control FBMN protrusive dynamics, we mosaically expressed GFP-Vangl2 in wild type FBMNs. FBMNs expressing GFP-Vangl2 in wild type embryos have less stable filopodia compared to wild type FBMNs, with an average lifetime of 123.4 ± 14.27 seconds (S1 Movie; N = 6 embryos, 7 neurons, 42 filopodia, p = 0.0013). Together, these loss-and gain-of-function findings suggest that Vangl2 functions within FBMNs to destabilize filopodia, since filopodia are affected in vangl2 mutant and GFP-Vangl2-expressing FBMNs regardless of the genotype of cells in their microenvironment.
Fzd3a, like Vangl2, is required cell-autonomously and cell non-autonomously for FBMN migration (S1 Fig, [49]). To determine whether Fzd3a has a cell-autonomous role in FBMN protrusive activity, we transplanted fzd3a mutant FBMNs into a wild type host. We found that filopodia of fzd3a mutant FBMNs are significantly less stable than filopodia of wild type neurons, with a mean lifetime of 163.3 ± 8.006 seconds (Fig 6C and 6F and S6 movie; p = 0.0092). However, mean maximum filopodia length (3.4 ± 0.4 μm) was not significantly different than that of wild type FBMNs (Fig 6G). This suggests that Fzd3a normally functions within FBMNs to stabilize filopodia protrusions, consistent with a conserved role for Fzd in actin polymerization [11][12][13]. Taken together, our results suggest that Vangl2 and Fzd3a function antagonistically within FBMNs to regulate filopodial stability.
Given that we observed this cell-autonomous function for Vangl2 and Fzd3a in regulating FBMN membrane protrusions, we asked whether these proteins regulate FBMN protrusive dynamics independently of cells in the migratory environment. To address this question, we analyzed the protrusive dynamics of isolated FBMNs in primary culture. We found that cultured FBMNs display altered filopodial dynamics compared to FBMNs in vivo. Cultured wild type FBMNs have a mean lifetime of 537.5 ± 32.28 seconds (during a 600 second time-lapse) and an average maximum length of 6.0 ± 0.5 μm (S9 Fig). Furthermore, there is no difference in filopodial dynamics between cultured wild type and cultured vangl2 mutant FBMNs (S9 Fig). This suggested to us that the cell-autonomous functions we observe for Vangl2 and Fzd3a in vivo depend on interactions with cells in the migratory environment.  65 filopodia. Graphs represent data as mean ± SEM. *p<0.05, **p<0.01 compared to WT!WT control; n.s., not significant. Significance was determined using an unpaired, two-tail t-test with Welch's correction. doi:10.1371/journal.pgen.1005934.g006

Vangl2 and Fzd3a have opposing cell non-autonomous functions in regulating FBMN filopodial activity
Since Fzd3a and Vangl2 are also required non-autonomously for FBMN migration and since FBMN protrusive dynamics depend on cells in the migratory environment, we asked whether cells in the FBMN environment influence FBMN protrusive activity in a PCP-dependent manner. In order to assess the role of Vangl2 in the environment, we imaged protrusion dynamics of wild type FBMNs in a vangl2 mutant environment. Interestingly we found that wild type FBMNs have less stable filopodia in a vangl2 mutant environment compared to a wild type environment, with a mean lifetime of 143.3 ± 18.28 seconds (Fig 6D and 6F and S7 Movie; p = 0.0056). The decrease in the average filopodia lifetime of wild type FBMNs in a vangl2 mutant environment is largely due to these neurons having a larger proportion of filopodia present for only one (30 seconds) or two (60 seconds) time-points (S8 Fig). The mean average length however was not different between wild-type FBMNs in a wild type environment and wild type FBMNs in a vangl2 mutant environment (Fig 6G) (3.1 ± 0.2 μm). Removing Fzd3a from the migratory environment had the opposite effect on FBMN filopodia. Wild type neurons in a fzd3a mutant environment generate dramatically more stable filopodia compared to those in a wild type environment, with a mean lifetime of 363.8 ± 51.12 seconds (Fig 6E and 6F and S8 Movie; p = 0.0273). Together our results suggest that the core PCP components Vangl2 and Fzd3a antagonize each other's activity to control filopodial dynamics during neuronal migration in vivo and they do so by functioning both within FBMNs and in cells in their migratory environment.

Discussion
Based on our findings, we propose a model for the role of PCP signaling in FBMN migration in which canonical interactions between the transmembrane PCP core components Vangl2 and Fzd3a regulate filopodial dynamics, thereby signaling and/or regulating adhesion for directional migration (Fig 7). We note that this model for filopodial dynamics is based on genetics and that our work does not elucidate the molecular nature of these interactions, which remain controversial even in the context of epithelial polarity [17,76]. This model is consistent with 1) a dual cell-autonomous and cell-non-autonomous requirement for PCP core components, specifically for the transmembrane components Vangl2 and Fzd3a, in FBMNs and their rhombomere 4 environment for migration; (this work, [31,49,51]); 2) the cytoskeletal and conserved molecular planar polarization of the r4 neuroepithelial environment including the floorplate (this work, [6]); 3) the ability of the planar polarized floorplate to promote the migration of wild type but not mutant FBMNs; 4) the localization of Vangl2 to retracting FBMN filopodial tips; 5) the antagonistic cell-autonomous roles of Fzd3a and Vangl2 in FBMN filopodial stability and 6) the opposite roles of Fzd3a and Vangl2 in the FBMN environment on FBMN filopodial stability.

Mutual antagonism of Vangl2 and Fzd3a
Our in vivo observations of filopodial dynamics in genetic chimeras demonstrate an antagonistic intracellular relationship between Vangl2 and Fzd3a in migrating FBMNs that regulates the stability of filopodium-like protrusions. While occurring in the context of a highly dynamic structure, this antagonistic relationship of Vangl2 and Fzd3a is reminiscent of the situation in stably polarized epithelia, where mutual intracellular antagonism between Fzd and Vangl complexes sets up polarized actin dynamics within the cell, with Fzd activating actin polymerization distally and Vangl suppressing it proximally [11-13, 77, 78]. This conserved interaction between Fzd promoting and Vangl suppressing actin growth may be common to other migratory cells. In metastatic breast cancer cells induced by stromal Wnt11-containing exosomes, Fzd6 and Vangl1 exhibit mutually exclusive localizations, with Fzd6 on the leading edge of cell protrusions and Vangl1 on non-protrusive cell surfaces, and knock-down of either protein decreases cell motility [44]. Similarly in migrating leukemia cells, Dvl-3 (part of the Fzd complex) localizes to the leading edge while Vangl2 localizes to the trailing edge [79]. During mesodermal and neuroectodermal convergence, mediolaterally oriented cell surfaces exhibit increased actomyosin contractility [33,80] that correlates with the asymmetric localization of PCP components Dvl and Pk (part of the Vangl complex) [7,81], suggestive of a conserved intracellular antagonism of these complexes mediating actin dynamics. In contrast, in commissural growth cones, Vangl2 promotes Fzd3-dependent outgrowth induced by diffusible Wnt5a by antagonizing a non-canonical inhibitory interaction between Dvl1 and Fzd3 identified in that context [41]. These examples show that core PCP components localize to discrete domains of moving cells and we have shown in vivo for the first time that this results in opposing effects on filopodial stability.

The role of filopodia in FBMN cell migration
Filopodia are commonly associated with the promotion of directed cell migration, although in some instances, axons and cells can achieve proper targeting and guidance without filopodia [82][83][84]. Due to their dynamics and long thin architecture, filopodia are capable of probing a Model of PCP regulation of directed neuron migration. Based on the filopodial dynamics and migratory behaviors of FBMNs we observed in genetic chimeras, and the localization of Vangl2 and Fzd3a we observed in FBMNs and the cells of their migratory environment, we suggest a model in which antagonistic interactions between Vangl2 and Fzd3a mediate the observed effects on FBMN filopodial dynamics and through them, directional neuron migration. Within FBMNs, Vangl2 (green) localizes to filopodial tips and destabilizes them while Fzd3a (magenta) has the opposite, stabilizing effect. In the planar-polarized cells of the migratory environment Vangl2 serves to stabilize filopodia while Fzd3a destabilizes them. In light of the known intracellular and intercellular interactions between Vangl and Fzd that underlie epithelial planar polarization, we hypothesize that interactions between Fzd3a and Vangl2 complexes destabilize one another intracellularly while they promote one another's effects on the actin cytoskeleton when they interact across cell membranes. Whether these interactions provide directional cues for migration remains to be discovered. wide area around cells, and they can contain receptors for diverse diffusible and membranebound signals and extracellular matrix molecules [85]. Thus, it is thought that the primary function of filopodia is as "antennae" that cells use to sense their microenvironment to orient directed cell migration [86]. Indeed it has been demonstrated that elimination of filopodia in axon growth cones does not impair axon outgrowth, but instead impairs growth cone turning in response to environmental cues [87][88][89]. This sensing role for filopodia has also been demonstrated in cell migration [84,90]. In addition to a sensing role, filopodia are thought to contribute directly to cell motility, as cells lacking filopodia tend to migrate more slowly due to the absence of filopodial adhesion molecules which could induce traction and also through force generated by actin streaming in filopodia [82,[90][91][92][93]. In the context of FBMN migration, filopodia extend in all directions from neurons when they initiate their migration, and we see no bias in the orientation of the filopodia that are affected in PCP mutants. We hypothesize that filopodia act as sensors of asymmetrically localized cell-surface PCP components on the neuroepithelial cells through which they are migrating and that this sensing fine tunes filopodium dynamics such that these filopodia can promote migration by acting as force generators or appropriately sensing other, as-yet unidentified environmental cues. In other migrating cells, several effectors have been identified as possible links between PCP signaling and cytoskeletal regulation [33,94,95]. While our work does not elucidate how those signals are transduced to the filopodial actin cytoskeleton in FBMNs, our previous work identified the WAVE-homology domain containing actin regulator Nhsl1 as localizing to FBMN filopodia and being required cell-autonomously for FBMN migration [51,96]; we hypothesize that PCP signals may be transduced to the actin cytoskeleton in FBMNs via Nhsl1.

Fzd3a and Vangl2 function in the migratory environment
A more surprising finding than opposing cell autonomous roles for Fzd3a and Vangl2 in FBMN filopodial dynamics and migration is that the same PCP components function in the FBMN environment to influence filopodial dynamics but in the opposite way: Fzd3a in the environment destabilizes filopodia while Vangl2 in the environment stabilizes them. These non-autonomous functions for Fzd3a and Vangl2 in filopodial dynamics correlate with their non-autonomous roles in FBMN migration [31,49]. Again, this is reminiscent of classical planar-polarity, where localized Fzd activity depends on the presence of Vangl in adjacent cells in the epithelium and vice versa; this is the mechanism by which PCP is coordinated across an epithelium [17][18][19][20][21]. We hypothesize that the cell-autonomous activities of Fzd3a and Vangl2 are activated in different filopodia when they contact Vangl2 and Fzd3a domains of neuroepithelial cells in the r4 environment (Fig 7), with consequences on the actin dynamics regulating filopodial stability, leading to changes in signaling and/or adhesion. We have shown here that Vangl2 and Fzd3a exhibit planar polarized localization in the r4 neuroepithelium and floorplate at the time of FBMN migration. In PCP mutants, this polarized information is absent and/or cannot be correctly interpreted by filopodia resulting in a failure of directional cell migration. We note that the cell-autonomous filopodial phenotypes appear to be dominant, since in constitutive mutants filopodia have the cell-autonomous phenotype (long and stable in vangl2 mutants; unstable in fzd3a mutants). Together our findings suggest that conserved intracellular and intercellular interactions between PCP core components can have divergent effects on actin dynamics and consequently on cell behaviors.
While the similar effects on filopodial dynamics when Vangl2 is depleted from FBMNs and when Fzd3a is depleted from their environment suggest that the two proteins are working together, environmental PCP may also influence filopodia dynamics of FBMNs through an indirect mechanism. For instance, core PCP proteins have been implicated in the trafficking and regulation of membrane levels of cadherins in fly and in vertebrate epithelial cells [97][98][99]. Therefore, Vangl2 and Fzd3a in the migratory environment may modulate FBMN filopodia dynamics by regulating N-cadherin levels at the surface of neuroepithelial cells. Another potential mechanism by which PCP in the migratory environment may regulate FBMN filopodial dynamics is through regulation of membrane type-1 matrix metalloproteinase (MMP14), which are known to degrade extracellular matrix proteins. During zebrafish gastrulation, an increase in Mmp14 activity was observed in vangl2 mutant embryos [100]. Thus, the decreased FBMN filopodial stability observed when Vangl2 is absent in the migratory environment could be due to decreased extracellular matrix.
Which cells in the environment of FBMNs are the source of PCP cues for filopodial dynamics and migration? We have shown that disruption of PCP signaling in the r4 environment prevents FBMN migration, demonstrating that PCP signaling is required to initiate directional migration. It was recently reported that Vangl2 expression even in a single cell in the r4 floorplate is sufficient to rescue FBMN migration in a vangl2 mutant [55]. In contrast, our results show that floorplate Vangl2 is neither required nor sufficient for FBMN migration.  [101,102].
We did, however, note that the presence of wild type cells in the floorplate could partially rescue the migration of wild type FBMNs in an otherwise vangl2 mutant embryo. This suggests that bidirectional PCP signaling between the planar-polarized floorplate and the FBMNs can promote migration. However this rescue was incomplete, indicating that other planar polarized cells in the r4 environment normally contribute to the pro-migratory environment. Consistent with this hypothesis, we found that disrupting the planar polarization of the floorplate alone in both fish and mouse embryos was insufficient to prevent FBMN migration (Fig 3 and S4 Fig). We conclude that the planar polarization of the entire r4 environment surrounding the FBMNs is required to effectively initiate migration. We were unable to confirm this by rescuing FBMN migration in a vangl2 mutant with r4-restricted expression of GFP-Vangl2, likely because the over-expression of PCP components disrupts planar polarity as efficiently as their loss [28,31,103,104].
Our study provides new insights into the role of the planar cell polarity pathway in neuronal migration by identifying when and where PCP signaling is required and how it affects the dynamic cell behaviors of migrating neurons in vivo. Our data suggests that a planar polarized hindbrain rhombomere 4 neuroepithelium serves to promote FBMN migration through the selective stabilization and destabilization of FBMN filopodia using conserved intra-and intercellular interactions between the PCP components Vangl2 and Fzd3a. Whether neuroepithelial planar polarity directs posterior migration or simply enables it, and through what effectors PCP signaling regulates filopodial dynamics in vivo are important questions to be answered in future work. and has letters of assurance on file with OLAW. The IACUC routinely evaluates the Fred Hutchinson animal facilities and programs to assure compliance with federal, state, local, and institution laws, regulations, and policies. The OLAW Assurance number is A3226-01.
To generate Tg(shh:Gal4VP16)fh445, the ar-B enhancer element of zebrafish sonic hedgehog (shh) [108,109] was amplified from a plasmid (gift from Uwe Strähle). For the Gal4 lines, the shh and isl1 enhancers were inserted upstream of the gata2 minimal promoter element [110]. The Xdd1 and full-length Xenopus Dvl are described in Sokol et al. (1996) [25]. Transgenic elements were cloned using the Gateway (Life Technologies) system using the primer sequences listed in S1 Table. Final DNA constructs were assembled in the pDESTpBHR4R3 plasmid (gift from the Brockerhoff Lab) or the CG5 Tol2 expression vector [107]. Transgenic embryos were generated by Tol2 transposase RNA co-injection with each plasmid at the single cell stage [111].

Mouse lines and husbandry
All mice were maintained at Fred Hutchinson Cancer Research Center under Institutional Animal Care and Use Committee approved guidelines. For general colony maintenance, all mouse lines were crossed into the C57BL/6J background (The Jackson Laboratory strain 00064). The Vangl2 Loxp and Vangl2 ΔTM lines were a gift from the Deans laboratory [61], the Isl1Cre (Isl tm1 (cre)Sev ) line was a gift from the Evans laboratory [62] and the Shh:gfp-cre (Shh tm1(EGFP/cre)Cjt ) line was purchased from The Jackson Laboratory (strain 005622).

Cell transplantation
Chimeric embryos were generated by transplantation at the blastula or gastrula stage as previously described [51,112]. To track transplanted cells, donor embryos carrying the Tg(isl1:GFP) rw0, Tg(isl1:mRFP)fh1 or Tg(isl1:mTFP)fh350 transgene were injected with 1% cascade blue-dextran or rhodamine-dextran (for live imaging) and 1% biotin-dextran (for imaging after fixation) (10,000 mw, Life Technologies). Host embryos were then processed and imaged for all donor-derived cells, donor-derived FBMNs or floorplate cells, and host FBMNs. Host and donor embryo genotypes were identified either by observing body axis elongation defects (for vangl2 mutant hosts), by examining FBMN location at 48 hpf or by genotyping (for fzd3a mutant hosts).
For transplantation of post-mitotic FBMNs, cascade blue-dextran labeled donor embryos and unlabeled host embryos were mounted in agar on coverslips at the 15-somite stage. The head of each animal was exposed by careful removal of agar with insect pins, and a hole was cut in the skin overlying the forebrain to enable entry of a thin (10 μm diameter) transplant pipette. Pre-migratory FBMNs (visualized by isl1:GFP or isl1:mRFP expression) were removed from r4 of a donor embryo and transplanted into r4 of a host embryo using a Zeiss AxioSkop fixed-stage microscope fitted with a 40X long working-distance water-immersion ("dipping") lens. During this process some non-isl1:GFP/mRFP-expressing neuroepithelial progenitor cells were inevitably co-transplanted but these usually died shortly after transplantation; any surviving donor-derived cells that were not FBMNs were detected by the presence of the cascade blue dye. Due to the disruptive approach, which removes nascent axons and other processes, even wild type FBMNs transplanted into a wild type environment do not migrate as well as FBMNs transplanted at earlier stages.

Imaging and data analysis
Imaging was performed using a Zeiss 700 confocal microscope or a Zeiss spinning disc microscope with a QuantEM EMCCD camera for live time-lapse imaging. For timelapse imaging, Z-stack images at 1μm steps were captured every 30 seconds for 15 minutes for in vivo time-lapse images and every 5 seconds for 10 minutes for cultured neurons. Filopodia were defined as long thin protrusions, less than 0.2 μm in diameter and more than 0.75 μm in length, measured from the cell body margin to the protrusion tip. In vivo filopodia lengths, lifetimes and fluorescent intensities of mRFP and GFP-Vangl2 were quantified using Zeiss Zen 2012 software. For cultured neurons, filopodium quantification was performed semi-automatically using Imaris FilamentTracer software (http://www.bitplane.com/imaris/filamenttracer). Mean anti-Vangl2 fluorescent intensity for all cell membranes were measured in user-drawn regions of interest using Zeiss Zen 2011 software or ImageJ's "Plot Profile" tool. Anterior/posterior GFP-Vangl2 fluorescent intensity ratios for each cell were normalized by dividing this value by the anterior/ posterior ZO-1 fluorescent intensity ratio. Graphs were generated and statistics were computed using GraphPad Prism software. All statistical analyses were performed using a 95% confidence interval. In most cases significance was determined using an unpaired, two-tail t-test with Welch's correction. For the anti-Vangl2 staining quantification significance was determined using a paired two-tail t-test with Welch's correction. Differences in FBMN distributions were analyzed using a Chi-square test where the distribution of FBMNs in control animals served as the expected frequencies or null hypothesis to determine if the observed frequencies were significantly different. Circular plots were generated using Oriana 4 software.  Table. Primers used in the creation of transgenic constructs. (PDF)