The plane facts of PCP in the CNS

Abstract

Planar Cell Polarity (PCP) pathways have been defined by their ability to direct the development of obviously polarized cellular architectures, such as the asymmetric localization of the stereociliary bundle in cochlear hair cells. Recent studies indicate that PCP pathways converge on the actin skeleton to regulate additional aspects of cell morphology that are not restricted to the plane of the epithelium. In the developing nervous system, these changes in the cytoskeleton are fundamental to neuronal migration, neuronal polarity, axon guidance and dendritic arborization, highlighting the importance of “planar polarity” genes for defining the shape of a neuron in all dimensions.

Introduction

One of the most striking features of the nervous system is the astonishing array of neurons with unique morphologies reflecting their specific functions. For example, sensory neurons in the dorsal root ganglia exhibit relatively simple pseudobipolar morphologies that allow them to transmit information rapidly from the periphery into the central nervous system, while Purkinje neurons develop highly complex dendritic trees that integrate information from many sources in the molecular layer of the cerebellum. Even the electrophysiological properties of a neuron are influenced by its overall dimensions and subcellular specializations. Although some progress has been made toward unraveling basic issues such as the decision to extend a single axon, we are still far from understanding how neuroblasts ultimately acquire the specific shapes that allow them to function properly within precisely wired networks of neurons.

Although neuronal morphologies are of particular interest to neuroscientists, most insights have instead come from cell biologists asking how any type of cell achieves and maintains its specialized architecture. Cells are patterned along two axes: the apical-basal axis (i.e. from top to bottom) and the planar cell polarity (PCP) axis (i.e. from front to back). Understanding how these axes are determined at the molecular level has been addressed by focusing on extreme examples of cell architecture, such as the protrusion of cilia only from the apical surface or of a wing hair only from the distal side of a cell. Hence, the key signaling pathways have been pieced together largely from studies in Drosophila, where changes in cellular morphology are easily identified in the wing, eye, or thorax (Wiggin et al., 2005; Yamanaka and Ohno, 2008; Zallen, 2007). Based both on genetic and biochemical interactions, three pathways have been described: the apical-basal determinants (i.e. atypical protein kinase C (aPKC), scribble, and discs large), the core PCP pathway, and the Fat pathway (Fig. 1).

Figure 1. PCP signaling pathways and possible points of interaction.

Two major PCP pathways have been defined: The Fat pathway (red) and the Core PCP pathway (blue). Fat signaling has global effects, possibly due to the production of additional signaling molecules such as the hypothetical Factor “X.” In vertebrates, Fat binds to two cytoskeletal regulators: Homer and Ena/Vasp, suggesting autonomous effects at the level of the cytoskeleton. Fat/Ds interactions are modulated by Fjx, which either acts as a secreted protein or in the Golgi to modify their extracellular domains. Interactions among the core PCP genes result in the asymmetric distribution of Vang and Pk to one side of the cell and Fz and Dvl to the other. Celsr is present on both sides. Protein localization is also influenced by cell-cell interactions, although it is not clear whether this involves direct binding. The apical-basal axis (green) is determined by mutually inhibitory interactions between the Par complex and the Scribble (Scrb) complex. Scribble can bind to Vang in vertebrates, where it is also required for PCP. One component of the Par complex, aPKC, has been shown to function downstream of a Wnt signaling pathway (purple). Although Wnts do not regulate polarity in Drosophila, non-canonical Wnt signaling is required for hair cell orientation and axon guidance. Vertebrate protein names are provided wherever possible. Proteins with only hypothetical roles in the pathways are shown in gray; unclear interactions are indicated with question marks. For example, a vertebrate homolog of Diego (Dgo) has not yet been investigated, nor is it clear whether Scrb and aPKC affect PCP signaling in the context of the Scrb and Par complexes respectively. The Hippo branch of the Fat pathway is not shown as there is as yet no evidence for a role in planar polarity. aPKC, atypical Protein Kinase C; Atr, Atrophin; Dlg, discs large; Ds, Dachsous; Dvl, Dishevelled orthologs; Fjx, Four-jointed ortholog 1; Fz, Frizzled orthologs; Lgl, lethal-giant larvae; Pk, Prickle ortholog; Scrb, Scribble ortholog; Vang, Van Gogh ortholog.

A role for the apical-basal determinants in the establishment of neuronal polarity is well accepted and has been reviewed extensively (Wiggin et al., 2005; Yamanaka and Ohno, 2008). In contrast, the importance of PCP and Fat signaling in the developing nervous system has received less attention, since circuit assembly events are rarely confined to the plane of the epithelium. However, just as genetics in flies has elucidated the basic principles of tissue polarity pathways, so have genetic studies in mice revealed a surprisingly broad array of functions for PCP genes in the developing nervous system. In fact, both the core PCP pathway and the Fat pathway affect much more than the planar organization of cells and are better thought of as signaling pathways that allow cells to make directed changes in the cytoskeleton that are coordinated both locally and globally.

How PCP signaling works is an area of active research, with many unanswered questions and lingering controversies, so the view presented here is simplified. A more detailed discussion of the underlying mechanisms of these two pathways is available in several excellent reviews dedicated to PCP signaling and tissue polarity (Fanto and McNeill, 2004; Saburi and McNeill, 2005; Wang and Nathans, 2007; Zallen, 2007).

The core PCP pathway

The main function of the core PCP pathway is to create molecular asymmetries that align neighboring cells. The core PCP pathway consists of Flamingo (Fmi), Frizzled (Fz), Dishevelled (Dsh), Van Gogh (Vang), Diego (Dgo) and Prickle (Pk). Initially, all proteins are uniformly distributed in the cell. Subsequently, the core proteins become localized to complementary domains, a molecular asymmetry that precedes the morphological asymmetry. For instance, in the fly wing, Fz and Dsh are enriched on the distal side of the cell, with crescents of Vang and Pk on the proximal side (Zallen, 2007). When the cell differentiates, the bristle emerges from the Fz-positive side of the cell. In mutants lacking one of the core PCP genes, the other PCP proteins fail to be properly localized and the bristle stays in the center of the cell, pointing in a random direction. Hence, cell polarity is determined by the asymmetric distribution of core PCP proteins.

A long standing question in the field has been how PCP proteins become asymmetrically localized. Proper protein distribution depends both on the activity of the PCP pathway inside the cell and on PCP-dependent interactions between neighboring cells. For instance, in chimeric animals, Fz-mutant cells induce changes in the distribution of Fz protein in neighboring wild-type cells, a phenomenon called domineering nonautonomy. Recent evidence from Drosophila indicates that Fmi is the key mediator for the cell-cell signaling that ensures that Fz and Vang are confined to opposite sides of the cell (Chen et al., 2008). Fmi is a homophilic cell adhesion molecule that is present on both sides of the cell. However, Fmi activity changes depending on whether or not Fz is also present. Hence, it is hypothesized that a Fz-bound Fmi molecule in one cell binds to a Fz-free Fmi molecule in the neighboring cell, causing bidirectional signaling that stabilizes Fz in one cell and Vang in the other. While the specific functions of other core PCP molecules remain unclear, proteins such as Pk and Dgo appear to amplify the initial bias in Fmi/Fz signaling and hence reinforce this molecular asymmetry, possibly by regulating the transport of Fz-containing vesicles along polarized microtubules (Shimada et al., 2006). Thus, PCP signaling acts both between cells (non-autonomously) to establish asymmetric distribution of proteins and within cells (autonomously) to reinforce and maintain this asymmetry.

The Fat pathway

The Fat pathway serves a more global role in tissue polarity (Saburi and McNeill, 2005), with mutations resulting in subtle PCP defects. In fat mutants, neighboring cells form whorls of cells with gradually shifting orientations (Ma et al., 2003). In contrast to mutations in core PCP genes, individual cells remain polarized, but this information is not effectively communicated across the population. Fat is a transmembrane protein with a huge extracellular domain and a cytoplasmic domain that is required for its ability to dictate tissue polarity (Matakatsu and Blair, 2006). Since Fat is uniformly expressed, spatial information appears to be provided by a gradient of Dachsous, another large cadherin that can bind to Fat (Matakatsu and Blair, 2004). In addition, the Golgi protein Four-jointed is expressed in an opposing gradient and modifies the activity of both Fat and Ds by phosphorylating serine and threonine residues in their extracellular domains (Ishikawa et al., 2008). Thus, Ds and Fj cooperate to determine where Fat is active and to what extent.

Although the details remain murky, genetic studies have defined two potential signaling pathways: the Atrophin-dependent PCP pathway (Fanto et al., 2003) and a tumor suppressor pathway that includes the STE-20 related Hippo kinase (Yin and Pan, 2007). For PCP, Fat both regulates asymmetric protein distribution and controls gene expression. First, Fat is required for the asymmetric distribution of an unusual myosin called Dachs (Mao et al., 2006; Rogulja et al., 2008). In addition, Fat induces changes in gene expression through regulation of the transcriptional repressor Atrophin (Fanto et al., 2003). One of the key target genes is Fj, providing one possible point of feedback in the Fat-Ds-Fj cassette. For tissue growth, Fat signals through Hippo and Warts kinases to determine whether or not the transcriptional repressor Yorkie can enter the cell nucleus (Yin and Pan, 2007). While this pathway has been studied mainly in light of its effects on cell growth, it is currently unclear whether there is a single Fat signaling pathway or whether the pathway forks downstream of Dachs, with one branch dedicated to PCP and the other to cell growth, akin to the canonical and non-canonical branches of the Wnt pathway. In fact, changes in PCP may in fact contribute to tumor overgrowth since mutations that suppress the growth effects in fat mutant flies can also suppress some PCP defects (Feng and Irvine, 2007). Hence, the Hippo signaling pathway may intersect with the Atrophin pathway to regulate events common to PCP and cell growth.

The nature of the relationship between the Fat system and the core PCP pathway remains controversial. One popular idea is that the Fat pathway induces a bias in Frizzled signaling and hence localization of core PCP components. Since Fat-mediated regulation of PCP requires a transcriptional regulator (Fanto et al., 2003), Fat may induce production of so-called Factor X, a signaling molecule proposed to act upstream of the core PCP pathway. This would provide an attractive way of converting a global gradient of Fat activity that is established by Ds into local biases in PCP activity that are reinforced by the Fz system. Alternatively, Fat could influence cell polarity autonomously through Dachs, which is asymmetrically localized and could serve as a scaffold that controls where other proteins are stabilized in the cell. Finally, there is also evidence that the Fat pathway acts parallel to and independent of the core PCP complex (Lawrence et al., 2007). For example, similar to what has been shown for the core PCP proteins, Fat and Ds may bind to each other in trans, enhancing localization of Fat on one side of the cell and Ds on the other. This could lead to asymmetric activation of Dachs and hence polarization of the cytoskeleton, independent of any change in the core PCP proteins. Thus, although Fat pathway genes are clearly required for the normal planar organization of cells, how this works at the molecular level remains a mystery.

Classic PCP signaling events in the developing nervous system

There is considerable evidence that the PCP pathways elucidated in flies act similarly in vertebrates. Investigations of vertebrate PCP pathways have focused on two examples of cells that are prominently polarized within the plane of the epithelium: progenitors undergoing convergent extension during neural tube closure and the hair cells of the inner ear (Fig. 2). These studies strongly support the presence of a conserved core PCP complex that has two main functions: to ensure asymmetric localization of other PCP proteins and to communicate this information to neighboring cells.

Figure 2. Cell Morphologies affected by PCP pathways.

(Top) The two most common examples of PCP are the proper alignment of hair cells in the inner ear and the directed movement of motile cells during convergent extension. In both cases, the cells exhibit obviously asymmetric morphologies within the plane of the epithelium (arrows). Similarly, PCP signaling enables apical daughter cells to move to the opposite side of the neural tube and prevents migrating neuroblasts from entering the neuroepithelium inappropriately. The table lists vertebrate genes involved in these classic polarity events, as well as their protein distribution and whether or not PCP phenotypes have been observed in mutant animals. nd, not determined; Y, yes; DKO, in double knock-outs only.

(Bottom) PCP genes are also required for aspects of neuronal morphology that are not restricted to a single plane. As they differentiate, neurons acquire polarized morphologies exemplified by the presence of multiple dendrites and a single axon (neuronal polarity). Subsequently, the axon is capped by a growth cone, which turns in response to cues during axon guidance. The dendrites branch to form arbors that vary greatly among cell types. Dendritic morphogenesis also depends on the activity of a subset of PCP genes. PCP genes implicated in each of these processes are listed beneath the relevant image. Evidence supporting a role for each gene in polarity, guidance, and dendrite development includes the distribution of the protein (P), manipulations in vitro (IV), and in some cases, knock-out phenotypes (KO). Many more players remain to be examined for potential functions in the developing CNS.

During neural tube development, cells undergo directed, convergent extension movements that both elongate the neuraxis and help the neural plate close to form the neural tube (Copp et al., 2003) (Fig. 2). This process clearly requires the activity of a conserved PCP pathway, as evidenced by the presence of severe neural tube defects in mouse mutants for frizzled (Fz) homologs (Wang et al., 2006b), the flamingo homolog Celsr1 (Curtin et al., 2003), the van gogh homologs Van Gogh like-1 and Van Gogh like-2 (Vangl1 and 2) (Kibar et al., 2001; Torban et al., 2008), and double knock-outs for dishevelled homologs (Dvl1 and 2) (Wang et al., 2006a). Heterozygous mice often have curly or looped tails, while homozygotes die with a completely open hindbrain and spinal cord, a condition called craniorachischisis.

In all of these mutants, neural tube defects are accompanied by a disorganization of the hair cells in the inner ear. Hair cells are prominent examples of planar polarity in the mammalian nervous system (Kelly and Chen, 2007). Protruding from the apical surface of each hair cell is a staircase array of actin-based stereocilia, with the tallest stereocilia asymmetrically positioned on one side of the cell. The mechanical deflection of the bundle towards the tallest stereocilia causes an electrical signal in the hair cell, emphasizing the importance of proper hair cell orientation for hearing and balance. Mutations in Vangl2, Celsr1, Fz3/6, and Dvl1/2 all cause hair cell polarity defects, with hair cells no longer sharing the same orientations as their neighbors.

At the cellular level, both neural tube closure and the alignment of hair cells depend on proper orientation of cells within the plane of the epithelium, analogous to the events elucidated in Drosophila. Moreover, as in flies, vertebrate PCP genes are required for asymmetric localization of the key PCP proteins (Deans et al., 2007; Montcouquiol et al., 2006; Wang et al., 2006b). For example, the Prickle ortholog Prickle like-2 is present in striking crescents on one side of hair cells, before the bundle begins to form its staircase structure. Pk2 crescents are opposite to crescents of Fz6 and disappear in Vangl2 mutants. Thus, the PCP pathway that was defined by genetic studies in Drosophila plays a prominent role in cell polarization in vertebrates.

A role for a Fat pathway in classic PCP processes in vertebrates has only recently been revealed. Of the four Fat orthologs, only the cytoplasmic domain of Fat4 shares significant homology with Drosophila Fat. Indeed, in Fat4 mutant mice, hair cells are mildly disorganized and the neural tube is broadened (Saburi et al., 2008). In addition, cell divisions in the elongating tubules of the kidney are misoriented, causing polycystic kidney disease. As in flies, Fat4 represses transcription of the sole four-jointed ortholog, Fjx1. In addition, Fat4;Fjx1 double mutant mice have enhanced kidney defects, consistent with the presence of a conserved pathway. However, the subtlety of defects in both Fat4 and Fjx1 mutant mice suggests that Fat4 could act together with other Fat genes, similar to the observed redundancies for mammalian Dvl and Fz genes. Fat4 does not appear to be required for proper localization of Vangl2 or Fz proteins, highlighting the unclear relationship between these two PCP pathways. There is circumstantial evidence for a Fat-Atrophin pathway, since Fat1 and Atrophin-2 mutant mice both suffer from neural tube defects, consistent with their co-expression in the notochord and floorplate (Ciani et al., 2003; Zoltewicz et al., 2003). However, neither mutant exhibits classic PCP defects or tumors, raising the possibility of novel cellular activities.

Despite abundant evidence that the fundamental features of PCP signaling pathways are conserved in vertebrates, important differences exist. One of the most striking modifications is the introduction of Wnt ligands. Since Wnt proteins bind to Fz receptors, there has been repeated speculation that a Wnt ligand might also act as a polarizing factor in flies that biases the asymmetric activation of Fz. However, PCP phenotypes do not occur in flies lacking any of the wingless genes or even in clones lacking all five Wnt ligands present in the wing disc, nor is the Wnt binding domain of Fz required for its PCP signaling capabilities (Chen et al., 2008). These observations fit with the results of computational models that demonstrate that interactions between PCP proteins are sufficient to generate polarized fields of cells in the absence of a polarizing morphogen (Amonlirdviman et al., 2005).

In contrast, Wnts are clearly involved in PCP signaling in vertebrates. For example, hair cells are misoriented in the cochlea of Wnt5a mutant mice (Qian et al., 2007) and both Wnt5 and Wnt11 are required for convergent extension in fish and frogs (Heisenberg et al., 2000; Kilian et al., 2003; Tada and Smith, 2000). The participation of Wnt ligands in vertebrate PCP signaling appears to have led to the recruitment of additional PCP molecules, such as the Cthrc1 secreted glycoprotein, which modulates the ability of Wnt ligands to selectively activate the PCP pathway upon binding to Fz and the Ror2 co-receptor (Yamamoto et al., 2008). While one can only speculate as to why Wnt ligands were added to PCP pathways in vertebrates, one possibility is that a secreted ligand is necessary for patterning large fields of cells that undergo much more extensive rearrangement than in Drosophila. Indeed, the cochlear phenotypes for both Wnt5a and Cthrc1 are mild compared to what occurs in Vangl2 mutants, consistent with the idea that the main function of the secreted Wnt ligand is to provide extra assurance for PCP interactions that are otherwise conserved. Likewise, the presence of other vertebrate-specific PCP molecules, such as Ptk7 (Lu et al., 2004), may reflect the need for more complex regulation of PCP in vertebrates, either by modifying Wnt ligand-specific aspects of signaling or by acting in new, as yet undefined pathways. Hence, elucidation of the molecular functions of Ptk7 may reveal still more differences in how PCP is controlled in vertebrates.

PCP signaling in vertebrates

functions outside the plane of the epithelium Although studies of mutant mice confirm the presence of conserved PCP pathways, the discovery of several unexpected phenotypes suggests that PCP signaling plays a broader role in neural development than anticipated. While early studies focused on the roles of PCP genes in classically defined polarization events, PCP pathways also affect other aspects of nervous system development that are not confined to the plane of the epithelium, including neuronal migration, neuronal polarity, axon guidance, and dendrite morphogenesis (Fig. 2). At the cellular level, the observed defects raise the intriguing possibility that PCP components sculpt cell morphologies in all dimensions.

Currently, our understanding of the role of PCP genes in CNS development is limited to little more than a list of phenotypes. However, several themes are emerging. First, individual genes seem to participate in multiple pathways, including cross-talk with other pathways such as the apical-basal determinants. Second, since PCP pathways both polarize individual cells and communicate this information to their neighbors, some phenotypes seem to be due to cell-autonomous changes in the structure of the cell, while other phenotypes are more related to a cell's inability to orient itself properly relative to its neighbors. Third, some PCP genes may have acquired new functions that exploit properties that are fundamental to PCP signaling, such as the ability to move in one direction along a microtubule. Indeed, a shared feature of the pathways is the ability to regulate the cytoskeleton, resulting in some cases in obvious changes in polarized architectures and in other cases in directed axon outgrowth and dendritic arborization. As more is learned about the protein interactions that underlie these distinct cellular events, we may find that planar polarity is simply an extreme example of a general mechanism that is used to generate cells of diverse architectures throughout the body.

Neuronal Migration

In both flies and vertebrates, one of the main functions of PCP signaling is to orient cells in a complex and changing environment. While cellular rearrangements have mostly been studied in the context of convergent extension, neuronal migration offers another excellent example of how cells move without losing track of their neighbors, no matter what plane the cells start out in. During early stages of neural development, post-mitotic neurons migrate away from proliferative zones to settle in their final location, which can be either nearby or quite far away. In the proliferative zone, dividing cells are polarized along the anterior-posterior axis, as revealed by the asymmetric localization of a Pk-GFP reporter protein in developing zebrafish (Ciruna et al., 2006). This localization is lost in Vangl2 mutant zebrafish. Moreover, in these mutants, apical cells fail to re-integrate into the neural epithelium after cell division and instead form ectopic clusters of cells and therefore neural tube defects. Remarkably, these effects are rescued by inhibition of cell division, indicating that the primary function of the PCP pathway in this context is to keep newly born neurons properly aligned after they divide.

Migrating motor neurons also make navigational errors in zebrafish mutant for frizzled or flamingo orthologs (Wada et al., 2006). The motor neurons normally take a tangential route under the surface of the pia and perpendicular to the neural epithelium. In frizzled3a and celsr2 mutant fish, the neurons extend aberrant radial processes and mistakenly integrate into the neural epithelium, thereby failing to reach their final destination. Thus, in both cases, PCP signaling appears to be necessary for keeping motile cells on the right track, either in the neuroepithelium as they divide or outside the neuroepithelium as they migrate. Although it remains to be seen whether PCP signaling plays a similar role during mammalian neurogenesis, it is easy to envision important roles in the regulation of symmetric vs. asymmetric cell divisions. It will be interesting to learn whether PCP proteins are asymmetrically localized in dividing precursors, and whether there are neurogenic defects in Vangl2 or Fz3/6 mutant mice that may be due to changes in the number of asymmetric cell divisions.

Convergent extension, hair cell orientation, and neural migration all depend on the cell-cell signaling aspects of PCP pathways. One way to think of this is that PCP proteins mark a subcellular domain and then keep this domain aligned with neighboring cells as they move. In this model, the asymmetric distribution of PCP proteins may force cells to adhere to each other only in one orientation. Once the first cell is in place, the orientation of every other cell will follow. This may explain why some cells are more affected by the loss of PCP signaling than others: cells that are surrounded by many other cells may be less likely to lose their way, while cells in looser environments may be more vulnerable. Consistent with this idea, hair cells at the edges of the organ of Corti are often more disoriented than those sandwiched in between (Montcouquiol et al., 2003; Wang et al., 2006b)

Axon Guidance

Another obvious polarized feature in neurons is the presence of a single axon that extends towards specific targets. Consistent with the fact that the axon is typically thought of as an apical domain of the neuron, the decision to extend a single axon involves components of the apical-basal polarity pathway, which includes Par3, Par6, and the atypical Protein Kinase C (Wiggin et al., 2005; Yamanaka and Ohno, 2008). Once specified, the axon can grow in any direction. Recent studies suggest that the PCP pathway intersects with the Par complex to affect not only the initial polarity of a neuron but also the subsequent guidance of its axon (Wolf et al., 2008; Zhang et al., 2007). These types of phenotypes are more consistent with the cell-autonomous functions of PCP molecules and raise the possibility that some PCP molecules have been recruited into new pathways used to remodel the cytoskeleton.

A role for PCP genes in axon guidance was first revealed when frizzled and flamingo were shown to be required for the formation of several major axon tracts. Indeed, mouse mutants for Fz3 and Celsr3 display remarkably similar guidance defects, including a loss of the anterior commissure and internal capsule, indicating that these two genes function in a similar pathway (Tissir et al., 2005; Wang et al., 2002). In vertebrates, Fz-mediated axon guidance is thought to occur in response to a Wnt cue (Lyuksyutova et al., 2003; Zou, 2004). For example, commissural neurons normally turn and grow rostrally towards the brain after crossing the ventral midline. Wnt4 is expressed in a gradient along the rostro-caudal axis and can reorient commissural axons in vitro. This effect requires Fz3, indicating that at least some of the tract defects in Fz3 mutant mice may be due to failed Wnt signaling.

Recent efforts to understand how Wnts regulates polarity and axon guidance have revealed an important convergence on the Par complex. Wnt5a induces multiple axons in cultured hippocampal neurons and this effect requires activity of the cytoplasmic effector protein Dishevelled (Dvl) (Zhang et al., 2007). Dvl, in turn, binds to aPKC, one of the key components of the Par complex. Morever, Wnt5a induces phosphorylation of aPKC, consistent with the idea that the Wnt pathway regulates early axon specification through regulation of the Par complex. A similar story is emerging for axon guidance, where aPKC is required for Wnt-mediated attraction and guidance of commissural axons (Wolf et al., 2008).

Understanding the role of PCP genes in axon guidance is confused by the fact that Wnt ligands act through multiple pathways to control diverse aspects of circuit assembly, including neuronal polarity, axon guidance, dendrite morphogenesis, terminal arborization and synaptogenesis (Salinas and Zou, 2008). Intriguingly, the same non-canonical Wnt5a ligand implicated in early axon specification also appears to regulate cochlear duct extension and the proper orientation of hair cells, highlighting the more typical PCP functions of this pathway (Qian et al., 2007). However, this distinction may be misleading since the so-called canonical ligand Wnt3a can also influence axon behavior, in this case acting through Adenomatous Polyposis Coli (APC) and β-catenin to alter the properties and organization of microtubules in the growth cone (Purro et al., 2008). Additional studies are needed to determine whether Wnt5a also acts through APC, or whether different classes of Wnt ligands exert their effects on the cytoskeleton through distinct pathways.

The use of PCP molecules for axon guidance is fundamentally different from what occurs during neural tube closure or in hair cells, and may not depend on the same pathways that operate during other polarization events. Extension of the axon occurs over a distance and requires significant addition of new membrane, while alignment of hair cells occurs locally and is not accompanied by a change in cell size. One major difference may be the use of the Wnt ligand, which can signal over a distance. Since other Wnt receptors, such as Ryk (Liu et al., 2005), are also involved in axon guidance, another possibility is that PCP signaling is accompanied by parallel activation of another Wnt pathway. Thus, the context of PCP activation may affect whether signaling results in local alignment or directed growth over a distance. Similarly, the response of the PCP complex to a secreted ligand in the extracellular matrix vs a transmembrane protein in the adjacent cell may generate distinct effects on the cytoskeleton.

Due to the wide variety of functions and downstream pathways demonstrated for both Wnt and PCP genes, the molecular basis of any one mutant phenotype must be interpreted with caution. For example, are the Fz and Celsr phenotypes indeed a reflection of Wnt-regulated axon guidance? In fact, while Wnts have been shown to influence axon guidance in vitro, these results have not yet been confirmed in vivo, raising the possibility that Fz and Celsr are regulated by an as yet unknown ligand. This idea is not without precedence, as the non-Wnt related ligand Norrin has already been shown to act through the Fz4 receptor (Xu et al., 2004). Similarly, it is not at all clear whether events downstream of Fz and Ceslr are conserved. Indeed, other PCP genes such as Vangl2 have not yet been shown to be required for axon guidance, raising the possibility that Fz/Ceslr act in a pathway that is distinct from the core PCP pathway. Analysis of mice lacking some of the intracellular players such as Pk2 will help to resolve this issue.

Dendrite Morphogenesis

One of the most obvious sources of variety in neuronal morphologies can be found in the dendritic arbors. However, in contrast to axon guidance, almost nothing is known about how these elaborate yet stereotyped branching patterns arise during development. One of the first genes shown to affect dendrite development is in fact Fmi, which was identified both in mutant screens for PCP defects and for abnormal dendrite tiling, i.e. the ability of neurons to elaborate a dendritic arbor that does not overlap with dendrites from neighboring neurons of the same type (Takeichi, 2007). In vertebrates, neural tube defects and misoriented hair cells occur in mice lacking the fmi homolog Celsr1, while Celsr2 and Celsr3 are proposed to play opposing roles in dendrite outgrowth (Curtin et al., 2003; Shima et al., 2007; Shima et al., 2004).

As for axon guidance, it is unclear how many of these effects reflect a PCP-like activity for Flamingo, since dendrite morphogenesis defects do not occur in fly frizzled mutants (Gao et al., 2000). This suggests that Celsr proteins may participate in an independent pathway for dendrite development that does not require Frizzled or other components of the core PCP pathway. Indeed, there are significant amino acid changes in the three Celsr family members that may allow Celsr proteins to bind to different effectors and hence activate distinct downstream pathways, only one of which is a PCP pathway (Shima et al., 2007). Celsr molecules might also function autonomously to mediate self-repulsion between neighboring dendrites in a single neuron, completely independent of any role in PCP. Given how little is known about dendrite development in general, elucidation of the binding properties of Celsr as well as the identification of downstream signaling molecules will provide an exciting entry point for understanding how neurons acquire their unique dendritic arbors.

The Fat pathway

In contrast to the core PCP pathway, much less is known about the Fat pathway or its functions. In flies, Fat proteins serve at least four cellular functions: cell proliferation, patterning of the axis of the wing, the formation of epithelial tubes, and tissue polarity (Tanoue and Takeichi, 2005). Whether these different cellular effects depend on activation of a single Fat pathway remains to be determined. Although Fat proteins are likely to act through multiple pathways to exert a wide range of effects in the developing nervous system, at least one of the functions is likely to be regulation of neuronal morphology.

Several observations indicate that Fat signaling can remodel the actin cytoskeleton. First, Fat regulation of PCP in flies requires asymmetric localization of the unconventional myosin Dachs (not to be confused with the ligand, Dachsous) (Mao et al., 2006; Rogulja et al., 2008). Second, in vertebrates, both Fat1 and Fat3 proteins exhibit distinct subcellular localizations, with Fat1 prominent in lamellipodia and filopodia (Moeller et al., 2004; Tanoue and Takeichi, 2004) and Fat3 enriched in mitral cell dendrites and in the processes that make up the inner plexiform layer of the retina (Nagae et al., 2007). Fat2 is restricted to the cerebellum, where the protein is localized to granule cell axons and has been proposed to regulate parallel fiber spacing here (Nakayama et al., 2002). In addition, Fat1 interacts with the scaffolding protein Homer and the actin binding proteins Mena and VASP via an EVH domain that is present only in Fat1 and Fat3 (Moeller et al., 2004; Schreiner et al., 2006; Tanoue and Takeichi, 2004). Consistent with these biochemical interactions, RNAi knockdown of Fat1 in vitro results in changes in the organization of the actin cytoskeleton as well as cell polarity defects, as revealed by the distribution of the Golgi complex (Moeller et al., 2004; Tanoue and Takeichi, 2004). There are no obvious vertebrate homologs to Dachs, making it difficult to test whether this branch of the Fat pathway is conserved. A challenge for the future is to determine whether an unconventional myosin is indeed involved in Fat signaling in vertebrates.

There is growing evidence that dendrites are an important site of action for the Fat signaling pathways. In mammals, a Fjx1-alkaline phosphatase fusion protein is efficiently secreted and binds to specific regions of the embryonic brain, leading to the proposal that Fjx1 is a ligand (Rock et al., 2005). Addition of secreted Fjx1 inhibits dendrite outgrowth and branching in cultured hippocampal neurons (Probst et al., 2007). In addition, Fjx1 mutant pyramidal neurons have simplified arbors and longer dendrites. How Fjx1 might mediate these effects is puzzling, as a recent study in flies revealed that the Four-jointed protein is a serine/threonine kinase that acts in the Golgi complex to phosphorylate residues in the Ds and Fat extracellular domains (Ishikawa et al., 2008). Since a secreted version of Four-jointed can also phosphorylate Fat (Ishikawa et al., 2008), the effects of Fjx1 on mammalian neurons in vitro could actually be due to changes in the phosphorylation state of Fat or Ds orthologs.

A key step toward understanding how the Fat pathway influences dendrite morphogenesis will be the identification of Fat and Ds orthologs that are relevant to the Fjx1 phenotype. For instance, does Fat act as a receptor in dendrites and if so, is a Ds ligand also involved? While gradients of Ds and Fj are important for Fat-mediated regulation of PCP, it is hard to envision why gradients would be useful for dendrite arborization or where these gradients might be. One alternative idea is that Ds and Fat orthologs act in the same cell to mediate interactions between neighboring dendrites. Intriguingly, the Hippo signaling pathway required for Fat-dependent cell growth has also been implicated in dendrite arborization and tiling (Emoto et al., 2006). This raises the exciting prospect that vertebrate Fats regulate dendritic morphogenesis through a conserved Hippo pathway.

Unraveling the events downstream of Fat will allow us to gain a more complete view of the potential functions of this pathway in the vertebrate nervous system. Fat molecules are remarkable not only for their large size but also for their diverse activities during development. For example, the four Fat orthologs have divergent cytoplasmic domains and may therefore activate distinct signaling cascades. Knowledge of how the four Fat genes differ will facilitate future interpretation of mutant phenotypes. This is underscored by the fact that Fat1 and Fat3 are expressed in hair cells (M. Deans and L. Goodrich, unpublished observation), but Fat4 appears to be the key regulator of hair cell polarity (Saburi et al., 2008).

What next?

As more and more roles for PCP genes emerge, the list of unresolved issues grows. With respect to the classic pathways, one of the most pressing questions is how this pathway is coordinated with the overall body axes of an animal: is a morphogen involved and if so, how might such a factor influence Fz and Fat activity? It will also be important to clarify the events that occur downstream of Fat and Fz in order to understand how PCP pathways influence cell morphology and orientation, particularly in the developing nervous system. For instance, during axon guidance and dendritic morphogenesis, asymmetrically localized PCP proteins might act cell-autonomously to direct localized changes in the actin cytoskeleton and hence oriented neurite outgrowth. Similarly, until we understand how Fat is activated (is Ds a ligand?) and what happens downstream (production of Factor X?), it will be difficult to uncover what happens at the level of the cytoskeleton.

Even as pathways emerge and mutants are created, understanding the role of PCP signaling in neural development will not be straightforward. During mutant analysis, it will be particularly important to distinguish the primary effects on neural development from secondary effects. For example, what may seem to be non-planar functions may instead be phenotypes that occur in response to more typical changes in planar polarity. One way to address these issues will be to study conditional knock-out phenotypes. For example, studies of Celsr3 conditional knock-out mice are consistent with the notion that Celsr3 acts as a homophilic adhesion molecule that mediates interactions between axons and guidepost cells during cortical development (Zhou et al., 2008). In addition, some CNS phenotypes may actually reflect additional functions for PCP genes outside of the traditional signaling pathways. Indeed, Fmi/Celsr both appear to participate in PCP and non-PCP pathways. Elucidation of the events both upstream and downstream of each gene will permit a more sophisticated interpretation of the phenotypes that have been observed to date.

Despite the many questions that persist, it is clear that both the core PCP and Fat pathways are going to become as familiar to neuroscientists as they are to cell biologists. As the downstream signaling events for both core PCP and Fat pathways are worked out, we move closer to understanding how molecular intersections between apical-basal and planar signaling pathways cooperate to achieve the exquisite architectures and precise connections so essential for the function of the nervous system.