Planar polarity in the Drosophila eye: a multifaceted view of signaling specificity and cross‐talk

Marek Mlodzik

Author Affiliations

  1. Marek Mlodzik*,1
  1. 1 EMBL, Developmental Biology Programme, Meyerhofstrasse 1, D‐69117, Heidelberg, Germany
  1. *Corresponding author. E-mail: mlodzik{at}
View Full Text


Functional tissues not only polarize their epithelia in the apical–basolateral axis, but also often display a polarity within the plane of the epithelium. In Drosophila, all adult structures are derived from epithelia called imaginal discs and display planar polarization; the eye and the wing are particularly well suited for analysis. Studies of their polarization have identified several conserved genes that regulate both nuclear signaling and cytoskeletal architecture. In particular, the Frizzled (Fz) receptor has been identified as a key component of polarity establishment in all tissues. The Fz signaling pathway and associated events are beginning to be unraveled, shedding light on a novel Wnt/Fz signaling cascade.


In multicellular organisms, epithelia form highly organized structures. Epithelial apical–basolateral polarity enables the tissues to perform functions such as vectorial transport of fluid (e.g. in the kidney) or directed secretion of specialized components to either their apical or basal side. However, the function of some tissues requires an additional axis of polarity within an epithelium, namely uniform polarity of single cells or multicellular units within the plane of the epithelium, commonly referred to as (epithelial) planar polarity. Examples include aspects of skin development in vertebrates, e.g. the ordered appearance of scales in fish or feathers in birds. Planar polarization is also evident in internal organs, including the inner ear epithelium, where the stereocilia bundles are aligned for normal sensitivity to sound, and the oviduct, with the cilia allowing directional transport of an egg. Other representative examples include the bristles and exocuticle in insects. These are particularly well suited to the study of the development of planar polarity (Figure 1). In addition, in insect compound eyes, planar polarity is evident in the uniform ordered appearance of the facets or ommatidia (Figure 1; see below). It is an intriguing question as to how cells that are hundreds of cell diameters apart adopt the same polarity in the plane.

Figure 1.

Planar polarity in Drosophila. A schematic drawing of planar polarity in an epithelial sheet is shown in (A). Planar polarity is perpendicular to apical–basal polarity. Examples of polarized tissues are shown for the wing (B), dorsal thorax (C) and eye (D). Note that the axis of polarity is different in the three examples, and that each tissue displays different apects of polarity, e.g. single cells are polarized in the wing as shown by the ordered appearance of the hairs, whereas groups of cells are reflecting polarization in other tissues.

The determination of apical–basolateral polarity is a direct consequence of an interaction of a cell with the environment, e.g. mediated by interaction with the basal extracellular matrix, or by general cell adhesion properties (reviewed in Eaton and Simons, 1995). The establishment of polarity within the plane is more complex as there are no obvious permanent clues between the cellular environments in the anterior–posterior, proximal–distal or dorsal–ventral axes. Therefore, it is thought that the signal(s) that control planar polarization operate over a long range. Moreover, the cellular planar polarity response is very diverse, ranging from the organization of cytoskeletal elements in single cells to the organization of multicellular units within themselves and their surrounding environment.

The recent identification of the molecules involved in the generation of planar polarity in Drosophila has increased our understanding of the molecular mechanisms involved and shed some light on the similarities and differences of planar polarization in distinct tissues. In this review, I summarize the current knowledge of the signaling pathway(s) and molecular mechanisms involved in planar polarity generation in Drosophila, with focus on the eye, and compare these with other tissues and other organisms.

Planar polarity in Drosophila

Epithelial planar polarity (also called tissue polarity) in Drosophila is most obvious in the wings, dorsal thorax and eyes (Figure 1). In the wing, each cell orients itself proximally to distally, generating a hair at the distal vertex. On the thorax, like most parts of the adult body, orientation of cells is in the antero‐posterior axis, easily apparent in the orientation of bristles, the peripheral nervous system sensory organs. In the eye, planar polarity is reflected in the regular arrangement of the ommatidia in the dorso‐ventral axis (see below for more details). Planar polarity mutations result in random orientation of hairs, bristles or ommatidia (Adler, 1992; Gubb, 1993; Theisen et al., 1994; Zheng et al., 1995; Strutt et al., 1997).

The phenotypes of the planar polarity mutants in different tissues (one cell in the wing versus multicellular units as bristles and ommatidia) underline a profound difference in the mechanistic aspects of polarity generation. In the wing, every cell adjusts its polarity more or less independently and reorganizes its cytoskeleton to form an actin spike at its distal vertex (Figure 1B). Thus, it has been suggested that the main requirement for planar polarity signaling is to affect cytoskeletal elements directly and lead to their rearrangement, with no requirements for a transcriptional response (reviewed in Eaton, 1997). In such a scenario, the asymmetric subcellular localization (or enrichment) of a component would suffice for the process. However, in the eye (ommatidia) or the thorax (sensory bristles), groups of cells make a coordinated decision. Thus, in this context, planar polarity signaling affects the organization of groups of cells, and thus the primary response is likely to include transcriptional activation and intercellular communication. These differences might reflect different cellular programs downstream of a general ‘polarity signaling pathway’.

Several genes that are required for planar polarity generation in all tissues have been identified genetically: frizzled (fz), dishevelled (dsh), prickle‐spiny legs (pk‐sple), rhoA, strabismus (also called Van Gogh) and flamingo (Adler, 1992; Theisen et al., 1994; Zheng et al., 1995; Strutt et al., 1997; Taylor et al., 1998; Wolff and Rubin, 1998; Gubb et al., 1999; Usui et al., 1999). In addition, several genes have been identified that are required for polarity generation in specific tissues. For example, fuzzy, inturned and multiple wing hairs (mwh) only affect polarity in the wing (Adler et al., 1994), whereas mutations in nemo and roulette appear only to affect ommatidial polarity (Choi and Benzer, 1994). Based on these observations, it was suggested that the genes affecting polarity in all tissues might be components of a common signaling pathway, responsible for reading and relaying a common polarity signal (see below), whereas genes such as fuzzy or nemo act as tissue‐specific effectors of this signaling pathway.

Similarly to the tissue‐specific effectors, the source of the polarizing signal is tissue/organ specific. Whereas the presumptive signal is thought to polarize the wing (or other appendages) from proximal to distal, the main body shows polarization in the antero‐posterior axis. The eye is polarized from the center to the poles (see below for details). Thus, in each tissue, the generation of planar polarity can be subdivided into three steps: (i) establishment of the source of a polarizing signal; (ii) reception and interpretation of the signal in single cells or groups of cells; and (iii) organization of the cells in response to the signal. Whereas the first and third steps are tissue specific, the second aspect (reading the polarity signal) appears to share many factors and is probably a generally conserved mechanism or signaling pathway. Although the polarity signal is not yet known, its presumed receptor, Frizzled, is required in all tissues (Adler, 1992).

Planar polarity in the eye

The Drosophila eye is polarized in a spectacular way. Polarity is reflected in the mirror image arrangement of ommatidia (or facets) of opposite chiral forms across the dorso‐ventral midline, the equator. This pattern is generated when immature ommatidia (each developing pre‐cluster consists of five cells at the time of polarity establishment) rotate 90° towards the equator (Figure 2) (Wolff and Ready, 1993; Reifegerste and Moses, 1999). Chirality is evident in mature ommatidia, as the R3 and R4 photoreceptors are positioned asymmetrically at the tip of the ommatidial trapezoids (Figure 2C). Initially, in developing ommatidial pre‐clusters, the R3/R4 precursor pair is arranged symmetrically in the eye imaginal discs, with the R3 precursor being closer to the equator than R4 (Figure 2A). In mutants such as fz and dsh, both chirality and the direction and degree of rotation become random (Gubb, 1993; Theisen et al., 1994; Zheng et al., 1995). In addition, in these mutants, some ommatidia remain symmetrical, giving rise to either V‐ or U‐shaped adult ommatidia with non‐chiral R3/R3 or R4/R4 photoreceptor pairs (Figure 2C). Genetic analysis of planar polarity mutants has demonstrated that the development of the R3/R4 pair, in particular the assignment of cell identity within the pair, is critical for establishing chirality and direction of ommatidial rotation (Zheng et al., 1995; Fanto et al., 1998; Cooper and Bray, 1999; Fanto and Mlodzik, 1999).

Figure 2.

Establishment of planar polarity in the eye. (A) Partial view of a developing eye imaginal disc demonstrating the regularity of polarity establishment. Ommatidial clusters are marked with anti‐Elav (green; labeling all photoreceptors as they join the cluster) and the seven‐up expression pattern (red). Seven‐up is expressed initially in the R3/R4 pair (see left side of panel) and later also in R1/R6 at weaker levels. Anterior is left. The morphogenetic furrow is on the left side just outside the field shown. Orientation of some ommatidial clusters is highlighted with an arrow in the upper, dorsal half; the equator is marked by a white line. (B) Cartoon of the logic of polarity generation during eye development. Initially, the ommatidial clusters are organized in the A/P axis and are symmetrical. Subsequently, they rotate 90° towards the equator and at the end symmetry is broken and chirality is evident by the difference in the R3/R4 cells. (C) Schematic presentation of chiral organization of mature ommatidia (compare with Figure 1D). In addition to the two chiral forms, symmetrical clusters with R3/R3 or R4/R4 pairs can be found in mutants affecting this process.

The polarity features mentioned above suggested that the polarizing signal in eye development originates at the equator, the dorso‐ventral midline. How is the equator set up? The dorso‐ventral midline is established by a complicated interplay of several signaling molecules and pathways. These include: (i) Wg emanating from the poles of the eye disc and its signaling cascade (including the arrow gene, a novel component of this pathway; Heberlein et al., 1998; Wehrli and Tomlinson, 1998); (ii) the homeobox genes of the iroquois–mirror complex that restrict fringe (fng) expression to the ventral half and the subsequent Fng‐mediated Notch activation at the Fng expression boundary (McNeill et al., 1997; Cho and Choi, 1998; Dominguez and de Celis, 1998; Papayannopoulos et al., 1998); and (iii) signaling of unpaired and the Hopscotch/JAK pathway (Luo et al., 1999; Zeidler et al., 1999). It is thought that in the second larval instar a combination of these pathways defines the D/V midline, which then serves as the equator, from which later the polarizing signal, often referred to as factor X, emanates and polarizes the eye field (reviewed in Blair, 1999; Reifegerste and Moses, 1999; Strutt and Strutt, 1999). Factor X then acts through Frizzled behind the morphogenetic furrow to establish polarity.

The Frizzled receptor is at the top of the planar polarity pathway

Many of the components for planar polarity establishment have been identified genetically, the most prominent being fz. Molecular cloning of fz suggested that it might function as a receptor: its primary sequence predicted a seven‐pass transmembrane receptor‐like molecule (Vinson et al., 1989). Recently, several Fz‐like proteins have been found in all animal species analyzed and they have been shown biochemically to act as Wg/Wnt receptors (Bhanot et al., 1996). Genetics has confirmed that they are required in the Wg signaling pathway (Bhat, 1998; Cadigan et al., 1998; Kennerdell and Carthew, 1998; Mueller et al., 1999).

Of the other known planar polarity genes, dsh is the best characterized (Klingensmith et al., 1994; Theisen et al., 1994). Genetic epistasis analysis has placed dsh downstream of fz (Krasnow et al., 1995; Strutt et al., 1997). The dsh gene encodes a 70 kDa cytoplasmic protein and, although it shares no similarities to proteins with known functions (Klingensmith et al., 1994; Theisen et al., 1994; Yanagawa et al., 1995), it contains three domains that are conserved in other proteins. Homologs of Dsh have been identified in many organisms, ranging from nematodes to humans. All Dsh proteins share three highly conserved domains: a DIX domain (found in Dishevelled and Axin), a central PDZ domain (thought to mediate protein–protein interactions and present in many proteins) and a C‐terminal DEP (named after Dishevelled, Egl‐10 and pleckstrin) domain (reviewed in Boutros and Mlodzik, 1999). All these domains have been implicated as protein–protein interaction modules, and thus Dsh might serve as an adaptor molecule.

Since Fz receptors and Dsh proteins are essential components of Wnt signaling, the initial expectation was that the planar polarity pathway and the canonical Wg/Wnt pathway (reviewed in Cadigan and Nusse, 1997) are the same (Tomlinson et al., 1997). However, in contrast to dsh, the other known components of Wg signaling do not display planar polarity defects (Axelrod et al., 1998; Boutros et al., 1998). These observations indicated that a distinct pathway acts downstream of Fz/Dsh in planar polarity establishment (Figure 3).

Figure 3.

The planar polarity signaling cascade. Schematic presentation of the information flow and its comparison with canonical Wnt signaling (in gray) is shown. The two pathways split (probably) at the level of Dsh. Different domains within Dsh are responsible for the specific activation of the two pathways. In the embryo, however, both Fz and Fz2 are activated by Wg and signal through the Armadillo/β‐catenin/Wnt pathway.

A planar polarity signaling pathway is emerging

Recently, a combination of genetic and biochemical studies has demonstrated that the planar polarity pathway downstream of Fz and Dsh consists of the small GTPase Rho and Misshapen (Msn, related to the yeast pheromone signaling STE20 kinase) (Strutt et al., 1997; Paricio et al., 1999). Genetic interactions suggest that a JNK (Jun N‐terminal kinase)‐type MAP kinase module acts downstream of Msn (Boutros et al., 1998; Paricio et al., 1999). None of these factors is involved in Wg signaling, and thus a signaling cascade that is distinct from the standard Wg/Wnt pathway is emerging (Figure 3). As with fz and dsh, both RhoA and msn are required for the generation of planar polarity in all tissues analyzed (Strutt et al., 1997; Paricio et al., 1999).

The involvement of and requirements for other components that have been implicated in this pathway, either by genetic interactions or biochemically, remain to be elucidated. Both Drosophila Dsh and its human homologs have been shown to act as potent activators of JNK signaling (Boutros et al., 1998; Li et al., 1999). Although mutations in components of the JNK cascade dominantly suppress gain‐of‐function genotypes of fz and dsh, they do not show planar polarity phenotypes in simple loss‐of‐function analyses, suggesting that their function is redundant (Boutros et al., 1998; Paricio et al., 1999), and thus additional experiments will be necessary to elucidate their exact role in polarity establishment.

Despite the identification of these new components of planar polarity signaling, it remains unclear how these components are linked molecularly. Although Fz recruits Dsh to the membrane (Axelrod et al., 1998), there is no direct molecular interaction between Fz and Dsh. Similarly, it is not known what factor(s) Dsh binds to in order to activate its downstream planar polarity effectors RhoA and/or Msn/STE20.

Dsh as a branchpoint in Frizzled signaling pathways

The observation that the canonical Wg/Wnt pathway and Fz/polarity signaling differ downstream of Dsh (Figure 3) suggested that Dsh might act differently on the pathway‐specific effectors. The mechanism by which Dsh routes information into the different intracellular pathways appears to lie in its domain composition.

Structure–function analysis and the molecular nature of a planar polarity‐specific dsh allele have shown that Dsh uses different domains in a modular fashion to activate distinct downstream effector cascades: the DIX domain is essential for Arm stabilization in Wg signaling and dispensable for planar polarity; whereas the DEP domain is essential for planar polarity and activation of the JNK kinase, and dispensable for Wg signaling (Yanagawa et al., 1995; Axelrod et al., 1998; Boutros et al., 1998). Interestingly, the analysis of domain requirements has identified the DEP domain as necessary and sufficient for membrane localization (Axelrod et al., 1998). Although the overlapping Dsh domain requirements for membrane localization and polarity signaling suggest a causal relationship, the role of membrane recruitment of Dsh is currently not known (for more details on Dsh, see Boutros and Mlodzik, 1999).

Are alternative Wnt signaling pathways used in other contexts?

Besides the difference(s) between the canonical Wnt pathway and planar polarity signaling as discussed above, evidence for ‘alternative’ Wnt signaling pathways is accumulating in several organisms. Vertebrate Wnts have been categorized in different groups according to their ability to transform mammary epithelial cells and their potential to induce secondary axes in frog development (Moon et al., 1993; Wong et al., 1994). In addition, although several Wnts are expressed during chick limb development, only some are able to signal through the canonical pathway, whereas Wnt7a induces a set of marker genes independently of β‐catenin and TCF/Lef (Kengaku et al., 1998). Additional genetic evidence for Wnt pathways independent of Arm/β‐catenin and TCF/Lef comes from genetic requirements in Caenorhabditis elegans. Whereas mom‐2/Wnt, mom‐5/Fz, WRM‐1/β‐catenin (Arm) and pop‐1/TCF are all required for the polarization of the EMS cell, only mom‐2 and mom‐5 are involved in the orientation of the mitotic spindle of the ABar cell (Herman et al., 1995; Rocheleau et al., 1997; Thorpe et al., 1997).

In addition to the pathway described above, recent experiments have suggested that XWnt5a and several Fz family receptors, which do not cause axis bifurcation in Xenopus, can signal through calcium‐dependent protein kinase C and heterotrimeric G‐proteins (Slusarski et al., 1997; Sheldahl et al., 1999). It is unclear whether this pathway overlaps or is independent of the components of planar polarity signaling.

The generation of cell polarity via Fz/Dsh signaling seems to be conserved between nematodes and flies. During bristle development in Drosophila, sensory organ precursor (SOP) cells undergo two rounds of asymmetric cell divisions. Fz and Dsh are also required for the orientation of the spindle apparatus and asymmetric distribution of the intracellular components in SOPs (Gho and Schweisguth, 1998). A similar asymmetric cell division of the ABar cell is Wnt/Fz dependent in C.elegans. Thus, a transcription‐independent, conserved pathway for inducing asymmetries in cells may exist in many (if not all) organisms. These findings could link epithelial planar polarity signaling to mechanisms whereby cells establish (and inherit) asymmetries.

How does a group of cells respond to the polarization signal?

How does Fz/planar polarity signaling generate a difference within a group of cells, such as the ommatidial cluster in the eye? Genetic manipulation has demonstrated that Fz activity in the R3/R4 photoreceptor pair is critical for polarity establishment; in particular, it was shown that the cell that has Fz activity (or more of it) will become the R3 cell, giving an ommatidium the respective chirality (Zheng et al., 1995; Fanto and Mlodzik, 1999). However, the R3 and R4 precursors are direct neighbors within a pre‐cluster and can be >100 cells away from the presumptive signal source (for ommatida far from the equator), and thus the signaling difference of the Fz pathway must be small. It is difficult to imagine how a 100% reliable read‐out is generated between the R3/R4 cells. In addition, in fz and dsh mutants, most ommatidia still remain chiral, with R3/R4 being specified in these clusters (albeit at random).

These observations suggested that there could be a secondary signal acting downstream of Fz that is involved in R3/R4 specification. Recently, the Notch pathway has been implicated in the generation of chirality, and Notch signaling was shown to specify R4 (Cooper and Bray, 1999; Fanto and Mlodzik, 1999). Expression analysis of the Notch ligand, Delta, demonstrates that Delta is a transcriptional target of Fz signaling in the R3 precursor. Delta amplifies the Fz signal by activating Notch signaling in the neighboring R4 precursor, locking the binary R3/R4 cell fate and chirality decision in place. This two‐tiered mechanism explains how a small initial difference in Fz signaling between R3 and R4 reliably generates the correct decision. This observation also explains why chirality, albeit stochastically, is still present in fz and dsh null mutants.

Could this two‐tiered Fz/Notch activation be a general mechanism for polarity establishment in multicellular units? The asymmetric activation of Notch signaling in response to a graded signal is possibly used also in the generation of polarity in the multicellular feather buds in vertebrates, where polarized expression of Notch pathway components has been reported (Chen et al., 1997). Although it is not known whether Fz signaling is involved in this context, the usage of Fz and Notch in the same processes is observed in several other tissues (e.g. bristle patterning and wing development). The potentially antagonistic or synergistic Fz–Notch interactions have led to speculations that Notch is also involved in Wg signal transmission (Couso and Martinez Arias, 1994). The observation that Dsh can interact directly with Notch as shown in a yeast two‐hybrid assay (Axelrod et al., 1996) suggests that the pathway interactions are complex, and possibly involve quenching of pathway components.

In addition, the recent identification of the flamingo (fmi) gene suggests that another signaling pathway acts downstream of Frizzled as an effector or modulator of the Fz signal (Usui et al., 1999). Fmi is a seven‐pass transmembrane receptor‐like protein of the cadherin superfamily. Strikingly, Fmi is localized differentially at cell–cell boundaries along the proximal–distal axis in the wing in a fz‐dependent manner (Usui et al., 1999). Like fz, fmi mutants affect all aspects of planar polarity, including bristle polarity on the notum and ommatidial polarity in the eye (Lu et al., 1999; Usui et al., 1999). Thus, fmi appears to be a common component of polarity establishment and might be required to attenuate or modulate Fz signaling in general (Usui et al., 1999). Interestingly, another cadherin superfamily member, dachsous, has been shown to affect wing hair polarity as well (Adler et al., 1998).

Other components of planar polarity establishment

Several other genes involved in general planar polarity establishment have been identified that do not (yet) fit into the Fz or any other pathway. The strabismus (stbm) gene shows the same phenotypic features as fz and dsh (Taylor et al., 1998; Wolff and Rubin, 1998). However, on close inspection its requirement in the eye is opposite to that of fz and dsh; stbm is required in R4 and not in the R3 cell. Molecular features of stbm are not informative as it shares no functional homology with other proteins (Wolff and Rubin, 1998). The requirement in R4 might indicate that the role of Stbm is to antagonize Fz/Dsh signaling. No genetic or molecular interactions between Stbm and components of the Fz pathway have been reported, however, and its molecular function remains elusive.

Similarly, the common planar polarity gene pk‐sple (Gubb et al., 1999) does not show informative genetic interactions (Strutt et al., 1997). The Prickle protein contains three LIM domains and might thus be involved in protein–protein interactions, possibly serving a scaffold function (Gubb et al., 1999). However, its role in planar polarity establishment and/or its relation to Fz signaling remain obscure due to the lack of molecular or informative genetic interactions (Gubb et al., 1999). In addition, Rac1 and Cdc42 have been implicated in the process of wing hair formation by the use of dominant‐negative isoforms (Eaton et al., 1996). It is not yet clear whether these are required for planar polarity in general.

The eye‐specific effector nemo encodes a distinct member of the MAPK family. nemo is required for the execution of ommatidial rotation, but not chirality establishment or the direction of rotation (Choi and Benzer, 1994). Interestingly, a Nemo‐like kinase (NLK) has recently been shown to antagonize Wnt signaling at the level of the TCF transcription factor in C.elegans and mammalian cell culture assays (Ishitani et al., 1999; Meneghini et al., 1999; Rocheleau et al., 1999). Although it is not known whether nemo itself has any Wg‐related phenotypes in Drosophila, or if its fuction in rotation is linked in any way to Wg signaling, this observation is intriguing, as it might suggest that the planar polarity and Wg pathway interconnect again at a different level.

Concluding remarks

The recent past has provided new insights into planar polarity/Fz signaling and has identified a novel route of Wnt/Fz signaling. The bifurcation of the planar polarity pathway to signal to nuclear versus cytoskeletal elements has not yet been addressed, although down to the level of Msn the pathway appears the same in the wing and the eye (Paricio et al., 1999). Despite the progress, there are still many gaps to be filled in the understanding of Fz signaling pathways. Both for planar polarity and for the canonical Wnt pathway, there are still missing components, and the way in which the known factors are interconnected is often unknown. The detailed study and understanding of planar polarity signaling and establishment will not only help to understand polarity generation in general, but will also be instrumental in dissecting the specificity of Wnt signaling and cross‐talk with other pathways.

Polarity mutants do not only exist in Drosophila. In vertebrates, for example, mutations in myosin VIIA cause deafness in mice and humans (Gibson et al., 1995; Weil et al., 1995). Interestingly, the stereocilia in the ear epithelium of these mutant mice have random polarity, suggesting that this is the primary cause of deafness (Self et al., 1998). Thus, the study of planar polarity establishment should also have several medical implications.


I thank all the members of my laboratory and Suzanne Eaton for sharing their enthusiasm and ideas, and colleagues at the EMBL for many stimulating discussions. Thanks to Ursula Weber for the photograph shown in Figure 2A, and M.Boutros, J.Curtiss, M.Fanto and U.Weber for helpful comments on the manuscript. Apologies to colleagues whose work has not been cited in full due to space limitations.


View Abstract