Planar cell polarization requires Widerborst, a B′ regulatory subunit of protein phosphatase 2A

Coordination of planar polarity within groups of cells is crucial for organismal development, from gastrulation (Heisenberg et al., 2000; Wallingford et al., 2000; Wallingford et al., 2001) through to the differentiation of complex tissues (Eaton, 1997; Mlodzik, 2000; Adler and Lee, 2001). The alignment of hairs and bristles formed by the cuticle-secreting epithelium of insects is a particularly striking example of planar polarization that has been well studied genetically (Adler, 1992; Adler and Lee, 2001). In the Drosophila wing epithelium, each cell polarizes its actin and microtubule cytoskeleton within the plane of the epithelium to generate a single distally directed membrane outgrowth that becomes the wing hair (Wong and Adler, 1993; Eaton et al., 1996; Eaton, 1997; Turner and Adler, 1998). Global coordination of hair orientation is controlled by the tissue polarity proteins, including Frizzled, which localize asymmetrically to form distinct proximal and distal cortical domains at the junctional region of each cell (Strutt, 2001; Shimada, 2001; Usui, 1999; Axelrod, 2001; Feiguin, 2000). Although two of these proteins, Frizzled and Flamingo, are transmembrane proteins that could potentially recognize a ‘polarization’ ligand, no such ligand has been identified, and the signals and cell biological mechanisms that orient the polarity of the proximodistal domains are unknown.

Many genes with important roles in cell polarization may cause organismal or cell lethality when mutated, precluding their identification as tissue polarity genes in loss-of-function screens. We therefore performed an overexpression screen to identify new genes involved in tissue polarization (Feiguin et al., 2000). We address the function of widerborst, a B′ regulatory subunit of PP2A identified in this screen. Our data suggest that widerborst plays roles in both cortical polarization and hair outgrowth in the Drosophila wing. They further suggest that widerborst is part of a conserved planar polarization mechanism that also operates during vertebrate gastrulation.

Drosophila EP3559 males were treated with EMS and crossed to either tubulinGAL4/TM3Sb or ApGAL4/CyO virgin females. Emerging non-Sb progeny were examined for duplicated thoracic bristles and apparent revertants were recrossed to tubulinGAL4/TM3Sb or ApGAL4/CyO. We recovered five EP3559 chromosomes that repeatedly failed to produce a bristle duplication phenotype. Sequencing revealed point mutations within the Wdb transcript for four of these.

Three different dominant negative mts-expressing constructs were produced by cloning N-terminally truncated subfragments generated by PCR into pUAST via EcoRI and XhoI linkers. The truncated proteins began at amino acids 165, 181 and 214. Dominant negative wdb was produced by cloning an N-terminally truncated PCR subfragment of wdb into pUAST via EcoRI and XhoI linkers. The truncated protein begins at amino acid 94.

Flies were dried and stored in a desiccator under vacuum. To prepare for scanning electron microscopy, dried wings were mounted on metal knobs with double-sided adhesive tape. The knobs were placed in a sputter coated (Balzers) and coated with bold (twice for 1 minute at 0.1 Torr, 100 V and 15 Ma). Scanning electron microscopy was performed with a Zeiss Novascan-30.

Standard image processing was performed using Adobe Photoshop 6.0 and Adobe Illustrator 9.01 (Adobe Systems) Deconvolution of images of anti-tubulin-stained wings was performed on 15 0.2 μm serial confocal sections with Huygens software.

The multiple sequence alignment of the selected PP2A B′ sequences was constructed using ClustalX with standard gap parameters and the BLOSUM-matrix series.

The phylogenetic tree was constructed with ClustalX, whereby gap positions were excluded and correction for multiple substitutions was turned on. The tree was bootstrapped with 1000 trials. The fungal sequences were used as an outgroup to root the phylogenetic tree. The branch length indicated in the upper right hand corner corresponds to 0.05 amino acid substitutions.

Mouse anti-Flamingo and rat anti-Dishevelled were kindly provided by Tadashi Uemura (Usui et al., 1999; Shimada et al., 2001). Guinea pig anti-Coracle was a gift from Rick Fehon (Fehon et al., 1994). Mouse anti-α and -β tubulin monoclonals were purchased from Sigma (Cat. No. T5168 and T4026). Rabbit anti-Wdb was raised to the C-terminal peptide PATTNAKIKQDKADN by Sigma-Genosys (Cambridgeshire, UK). The antibody was affinity purified using the immunizing peptide and used at a dilution of 1:300. The antibody was judged to be specific in immunofluorescence studies because its staining was elevated in cells overexpressing wdb and because staining disappeared in clones mutant for the IP allele of wdb. The antibody did not detect any specific bands when used on western blots.

Pupae were dissected in Graces tissue culture medium (Sigma) at room temperature on 3 cm plastic tissue culture dishes. During the dissection process, it was crucial that the wings not be damaged or stretched; we have found that stretching the wing disturbs the organization of the planar microtubule webs, on which Wdb is localized. In many cases where wings had been stressed, for example, by attempting to make an opening in the distal cuticle, microtubules collapsed to one side of the cells over large regions of the wing, along with Wdb protein. This sometimes led to apparent misorientation of Wdb staining. Distal localization of Widerborst was observed when wings were fixed with cold methanol, which efficiently preserves the structure of the planar microtubule webs, but not when they were fixed with paraformaldehyde, which does not (data not shown).

We dissected pupae as follows: for 18-30 hour pupae, the whole pupa was removed from the pupal case. The head and abdomen were then pierced, and the internal fat droplets removed. The head and abdomen were then removed gently, taking care not to exert force on the wings. Pupae were then opened along the dorsal thorax, and the cuticle covering the wing hinge region was broken, exposing the underlying epithelium at the proximal end of the wing. No attempt was made to remove the cuticle coving the wing itself. These carcasses were placed into 3 cm petri dishes containing 100% methanol precooled to 4°C and left on ice for 2 minutes. The dish containing the wings was then transferred to a metal plate at –20°C and left for a further 18 minutes. The dish was then placed on ice and the wings were removed to another 3 cm dish containing 90% methanol in water and left on ice for 3 minutes. Wings were then transferred to dishes containing 50% methanol, where they were separated from the rest of the carcass and detached from the surrounding cuticle.

To obtain prepupal wings, we held on to the anterior tip of the prepupa with number 5 forceps and gently severed the pupal case approximately one third of distance from the anterior end. Wings usually remained attached to the head region and they were transferred to cold methanol (as described above) without being directly touched.

Wings were then accumulated in PBS + 0.1% Triton X-100 (PBT) in one well of a 24-well tissue culture plate. After rinsing twice with PBT, the wings were blocked for 10 minutes in PBT + 0.5% FCS, then incubated overnight with relevant antibodies.

In situ hybridization with digoxigenin-incorporated antisense RNA probes was as described (Heisenberg et al., 1996). RNA and morpholino antisense oligonucleotides against zebrafish wdb1 and wdb2 were injected as described (Heisenberg et al., 2001). Control-morpholino oligonucleotides (Gene Tools) and morpholinos against various other genes (axin1, sqt, cyc) were injected to determine the specificity of the obtained phenotypes.

To identify new genes involved in planar polarization, we used an EP modular misexpression screen (Rorth, 1996; Feiguin et al., 2000). We induced expression of 2500 different EP lines in the developing wing and thorax, and first selected those lines that induced ectopic or apolar bristles in the thorax. We chose for further study those lines that also produced defects in hair polarity or number. Expression from EP3559 caused bristle duplication in the thorax and induced many cells throughout the wing to form multiple rather than single hairs (compare Fig. 1A with 1B), suggesting that the gene activated by EP3559 might regulate planar polarization.

To identify the gene overexpressed by EP3559, we cloned and sequenced the adjacent DNA and identified corresponding ESTs in the BDGP (Berkeley Drosophila Genome Project) database; these were derived from gene CG5643 in the GadFly database. Sequencing revealed no differences in the coding potential of the cDNAs, but suggested that alternative splicing gives rise to transcripts differing in their 5′ untranslated sequence (submitted to the GadFly genome annotation database). Three different cDNAs were placed under the control of a GAL4UAS promoter and used to generate transgenic flies. After inducing expression in the wing, we observed that only the cDNA with the longest 5′ untranslated region, LD2456, caused multiple wing hair formation (Fig. 1C). This shows that CG5643 indeed represents the gene causing the multiple wing hair phenotype, and suggests that the 5′ untranslated region may play a part in its post-transcriptional regulation.

We named the gene from which LD2456 is derived widerborst, which, in German, means something stubborn or recalcitrant (derived from wider, meaning against, and borst, meaning bristle). It encodes a B′ regulatory subunit of protein phosphatase 2A (PP2A), an enzyme that is conserved from yeast to mammals. PP2A is a holoenzyme that consists of a catalytic (C) subunit, an A regulatory subunit and one of a large family of B, B′ or B′′ subunits. The latter subunits are thought to regulate the activity of the C subunit and provide substrate specificity. In metazoans, the B′ subunits have diverged into two related subclasses (Fig. 1D). The central regions of these proteins are strongly conserved, but they differ at their N and C termini (Fig. 1E). The protein encoded by widerborst is more closely related to the human α, β and ε subunits (62-66% identity) than to the β or γ subunits (52-59% identity). Its sequence suggests that wdb might influence tissue polarization by regulating PP2A activity with respect to specific targets.

To ask whether Wdb was necessary for tissue polarization, we sought loss-of-function wdb mutations. After treating EP3559 males with EMS, we crossed them to flies containing either tubulinGAL4 or apterousGAL4 and screened for chromosomes that could no longer produce the bristle phenotypes characteristic of wdb over-expression. Five different ‘revertant’ chromosomes were isolated in this fashion; four of the five (wdb¹², wdb¹⁴, wdb^IP, and wdb^dw) fell into one complementation group. Each mutation was homozygous lethal, and either lethal or semi-lethal with other wdb alleles.

We analyzed these possible wdb mutants in three different ways. First, we looked for mutations in the wdb transcribed region. Second, we quantified the extent to which overexpression of the different alleles could cause multiple wing hair formation. Finally, using an antibody to the Wdb C terminus, we measured Wdb protein accumulation caused by patchedGAL4-driven expression from the EP element. These data are summarized in Table 1 and Fig. 1E.

From these analyses, we concluded that wdb¹⁴ is closest to a null; it contains a nonsense mutation at codon 150 that removes two-thirds of the protein, including most of the highly conserved region. Consistent with this, the C-terminal antibody fails to detect overexpressed Wdb¹⁴ protein, and the mutation completely suppresses the formation of multiple wing hairs caused by overexpression.

Wdb^IP harbors a stop codon at position 332 and encodes a C-terminally truncated protein that is significantly longer than Wdb¹⁴ and contains most of the central, conserved region. Wdb^IP may retain some function, because wdb^dw/wdb^IP escapers have weaker wing phenotypes than wdb^dw/wdb¹⁴ escapers (see below). As expected, the truncated protein encoded by wdb^IP is not recognized by the C-terminal antibody (Table 1).

Wdb^dw suffered a nonsense mutation at position 9, and thus might potentially encode the shortest protein. However, expression of wdb^dw from the EP insertion produced protein that could still be detected with the C-terminal antibody. This suggests that translational reinitiation probably occurs at a downstream methionine, resulting in an N-terminally truncated protein. Despite its inability to produce duplicated bristles, wdb^dw expression caused almost as many multiple wing hairs as did the wild type gene, suggesting that the mutation is mild with respect to tissue polarity.

Wdb¹² is mutated in the donor splice site that follows the first coding exon. As the C-terminal antibody still detects protein expressed from wdb¹², albeit at reduced levels, we presume either that splicing still occurs to some extent or that an alternative translational initiation site is used. Its mild suppression of the multiple wing hair phenotype confirms that wdb¹² is a weak allele.

As the wdb mutations were homozygous lethal, we first examined their effects on hair formation by placing them on FRT chromosomes and generating homozygous clones by mitotic recombination. We were unable to recover clones of the null mutation, wdb¹⁴, suggesting that cells that have no Wdb activity do not survive. We did recover clones of the partial loss-of-function mutations wdb^dw, wdb^IP and wdb¹², but weaker alleles had no effect on hair formation or polarity (data not shown). Wdb^IP clones, however, caused outgrowths in the wing and eye, suggestive of ectopic Wingless signaling (M. H., unpublished).

Because clonal analysis failed to resolve whether Widerborst was involved in polarization or hair formation, we examined hairs in wings from wdb^dw/wdb¹⁴ and wdb^dw/wdb^IP flies. Although most flies of these genotypes die before emerging, we were able to analyze wings from the occasional escapers. These wings contained bald patches on the wing blade where hair formation had either failed, or was compromised (Fig. 1F). The stunted hairs observed in these regions were sometimes of altered polarity (arrows). These defects occurred most often on the ventral wing blade between veins 3 and 4, a region in which hairs are normally somewhat shorter and finer than in the rest of the wing. This part of the wing may be more sensitive to conditions that discourage hair formation. Clones of wdb^IP or wdb^dw that fell outside of this small region would not be expected to have any effect. Loss of hairs was more severe in wdb^dw/wdb¹⁴ than in wdb^dw/wdb^IP wings. Overall, the phenotypes given by wdb transheterozygotes suggest that endogenous Wdb positively regulates hair formation and may impinge on hair polarity.

Although incomplete loss of function interfered with hair formation in some regions of the wing, we suspected that the null phenotype of Wdb might be more severe. As cell lethality prevented us from analyzing hairs formed by cells totally missing Wdb activity, we designed a dominant-negative version of the protein analogous to one described for a vertebrate B′ γ1 regulatory subunit. N-terminal truncation of this γ subunit has been shown to dominantly inhibit dephosphorylation of paxillin by PP2A (Ito et al., 2000). We hoped that a similar truncation of Wdb might dominantly inhibit dephosphorylation of Wdb targets. We expressed a Wdb construct from which the first 93 amino acids had been deleted along the AP compartment boundary under the control of ptcGAL4, and examined the resulting wings by scanning electron microscopy. Hairs that formed in the region of the wing expressing N-terminally truncated Wdb were often stunted and sometimes absent, as might be expected from the phenotype observed for partial loss of Wdb function (cells expressing the dominant negative are shaded red in Fig. 1G,H). Furthermore, the stunted hairs produced by Wdb dn-expressing cells were often of abnormal or even reversed polarity, similar to those occasionally produced by Wdb transheterozygotes. The polarity defects are probably not a secondary consequence of faulty hair outgrowth, as cells expressing dnWdb also exhibited polarity alterations when hair morphology was otherwise normal (Fig. 1H). The effects of dominant-negative expression also resembled those of mild loss of Wdb function in that they were more severe on the ventral side of the wing. Hairs in the patched expression domain on the dorsal side of the wing were most often wild type in appearance, although no obvious differences in expression level were observed. These results show that dnWdb expression produces defects qualitatively similar to, through stronger than, those produced by mild wdb loss of function.

To confirm that the defects produced by dominant-negative Wdb were due to inhibition of Wdb rather than another activity, we asked whether co-expressed wild-type Wdb could reduce their severity. Co-expression of the wild-type protein strongly suppressed the defects in hair orientation and length (data not shown), indicating that the molecular functions disrupted by dominant negative Wdb are ones that can actually be performed by Wdb itself. These results do not rule out the possibility that dnWdb also interferes with a second protein that acts redundantly with Wdb.

Interestingly, hairs near the Wdb dominant negative expression domain were occasionally stunted or missing (Fig. 1G, arrows), suggesting that dnWdb expression can non-autonomously disrupt hair formation.

B′ regulatory subunits are thought to act by modulating the activity of the catalytic subunit of PP2A with respect to specific protein targets. To address the effect of Widerborst on the activity of the catalytic subunit, we tested whether reducing the level of the C subunit could modify the wdb overexpression phenotype. Fig. 2A shows that transheterozygosity for two different strong alleles of microtubule star (mts), the gene that encodes the Drosophila PP2A C subunit, greatly reduces the number of cells forming multiple hairs in wdb-overexpressing wings. These data suggest that Wdb overexpression exerts its effects by positively regulating the enzymatic activity of the catalytic subunit of PP2A.

To confirm that PP2A activity could regulate hair formation, we examined the effect of overexpressing the catalytic subunit. Fig. 2B shows a wing in which mts was expressed along the compartment boundary under the control of the patched promoter. Strikingly, cells that overexpress mts can construct as many as 20 wing hairs that cover their apical surface and point in all directions. In those cells that make fewer wing hairs, it is also clear that hair polarity is disturbed. These data indicate that unregulated activity of the catalytic subunit enlarges the region of the cell capable of hair formation and interferes with normal polarization.

To determine whether endogenous PP2A activity was required for either hair outgrowth or polarization, we wanted to examine the effect of loss of PP2A function. As null alleles of mts are cell lethal (Wassarman et al., 1996), we approached the problem by expressing dominant negative versions of the protein. Truncations of the catalytic subunit that remove the active site have been shown behave as dominant negatives based on their ability to interact non-productively with A and B subunits (Evans et al., 1999). We constructed similar truncations of mts and expressed them under the control of the patched promoter during wing development. Wings derived from these flies are devoid of hairs over a significant region of the AP compartment boundary (Fig. 2C). In regions where hairs could be seen, they were often of abnormal orientation (Fig. 2D). Taken together, these data show that PP2A activity is necessary both for hair formation and hair orientation.

Because regulation of the PP2A catalytic subunit by Widerborst affects hair formation, we wondered how activation might normally be restricted to the distal side of the cell. To address this question, we examined the subcellular localization of Wdb shortly before and during hair formation. Between 26 and 28 hours after puparium formation (apf) the tissue polarity genes Fz, Dsh, Fmi and Dgo are expressed at high levels at the proximodistal cortex (Usui et al., 1999; Feiguin et al., 2000; Axelrod, 2001; Shimada et al., 2001; Strutt, 2001) (Fig. 3A,B), but hairs will not form for another 3 to 5 hours. At this stage, Wdb protein is distributed on predominantly cytoplasmic punctate structures that are most abundant distally (Fig. 3B,C). Co-staining with antibodies to tubulin revealed that these spots of Wdb at least partially overlap with a subset of microtubules. Many microtubules in wing epithelial cells are organized in a meshwork lying parallel to the epithelial plane at the level of apical junctions (Eaton et al., 1996). Fig. 3D shows a projection of the microtubules comprising this ‘planar web’. Widerborst is enriched on microtubule bundles on the distal side of this web; colocalization is seen most clearly in single sections (Fig. 3G-I). During hair formation, Wdb polarity is lost, although it still associates with microtubules in general, including those in the outgrowing hair (data not shown). The enrichment of Wdb on distal microtubules suggests that PP2A is locally activated there and that microtubule associated proteins may be dephosphorylated in a polarized fashion before hair formation.

To identify the time at which the Wdb polarization mechanism is active, we examined Wdb localization at earlier stages of wing development. In third instar wing discs, and in early pre-pupal discs, Wdb was not polarized (data not shown). By 8 hours app, after the everted wing pouch had flattened significantly, Wdb became enriched on the microtubules to one side of each cell; surprisingly, its orientation at this stage was predominantly proximal with a slight anterior slant, rather than distal (Fig. 4A-C). By 18-20 hours apf, Wdb had relocalized distally (Fig. 4D,E). Flamingo, by contrast, was polarized only slightly, if at all (Fig. 4A). These results show that epithelial cells develop a polarized planar axis, reflected by Wdb localization, in pre-pupal wings. The polarity of the axis is initially proximal, but reverses direction sometime between 8 and 18 hours after puparium formation. The timing of distal Wdb reorientation suggests that it either precedes or is coincident with the onset of cortical domain polarization.

The subcellular localization of Wdb suggests that it does not polarize by binding directly to components of the cortical tissue polarity complex. To examine whether Wdb localization depends on the cortical domains at all, we stained dsh¹ mutant wings for Wdb. In dsh mutant cells, both Frizzled and Flamingo fail to polarize along the proximal distal axis (Usui et al., 1999; Strutt, 2001). Strikingly, Wdb is well polarized in these wings (Fig. 5A). This observation suggests that polarized proximodistal cortical domains are not necessary for Wdb polarization.

To further examine the role of the cortical domains, we examined Wdb polarization in wings mutant for starry night/flamingo. In Flamingo mutant clones, both Frizzled and Diego not only fail to polarize, but do not localize the cortex at all (Feiguin et al., 2000; Strutt, 2001). Nevertheless, Wdb is normally polarized in stan³ mutant wings (Fig. 5B). Taken together, these data suggest that Wdb polarization does not require formation of polarized cortical domains.

To ask whether Wdb would respond to dominant reorientation of proximodistal cortical domains, we examined its localization in wings that expressed Frizzled under the control of the patched promoter (Fig. 5C-F). In these wings, Flamingo, Diego and Dsh (Usui et al., 1999; Feiguin et al., 2000; Axelrod, 2001; Shimada et al., 2001) (Fig. 6D) relocalize perpendicular to the normal proximodistal axis. The polarization of Wdb, however, is unchanged (Fig. 5C,E,F). These data show that the proximodistal domains formed by tissue polarity proteins are neither necessary nor sufficient to specify the axis of Wdb polarization.

To ask whether cortical polarization might depend on Widerborst, we examined the localization of Fmi and Dsh in cells expressing dominant-negative Wdb along the AP compartment boundary. Fig. 5G-L shows dnWdb-expressing wings that have been stained with an antibody to Widerborst, and either Fmi (Fig. 5H,I) or Dsh (Fig. 5K,L). The cells that express dnWdb are identifiable by their elevated staining with the Widerborst antibody (Fig. 5G,H,J,K). In these wings, it is clear that expression of dnWdb causes both Fmi and Dsh to accumulate uniformly around the cortex at high levels (Fig. 5H,I,K,L). By contrast, Dsh and Fmi are normally polarized in most of the cells that do not express dominant negative Wdb. These data indicate that Wdb activity is required for normal cortical polarity.

Interestingly, the expression of dnWdb appears to inhibit the distal accumulation of endogenous wild type Wdb in the cells adjacent to the dnWdb-expressing domain (Fig. 5G,J). This observation is consistent with the non-autonomous effects on hair formation observed in the adult wings of these flies (Fig. 1G). Furthermore, Dsh and Fmi are often depolarized in cells up to five cell diameters away from the dnWdb-expression domain (Fig. 5H,K). This might either be due directly to lower levels of endogenous Widerborst, or to non-autonomous propagation of cortical depolarization in the dnWdb-expression domain.

To ask whether Widerborst was required for the cytoskeletal integrity of wing epithelial cells, we examined the effect of dominant negative expression on actin, microtubules and Coracle. Cortical organization of both filamentous actin and Coracle appears essentially normal in dnWdb-expressing cells (Fig. 6A,B,C,F), suggesting that failure to polarize Flamingo and Dishevelled distribution does not result from gross defects in the subcortical actin cytoskeleton. By contrast, the organization of the planar microtubule web was perturbed by expression of dominant negative Wdb. Although dnWdb-expressing cells accumulate microtubules to at least normal levels, their ordered, web-like structure is not maintained (Fig. 6D-F). This suggests that Wdb normally directs the dephosphorylation of a protein that is important for microtubule organization and that the structure of the planar web may contribute to the development of cortical polarity.

The PP2AB′ regulatory subunits of the α/ε family are highly conserved and homologous genes are present from worms to humans. To ask whether the function of widerborst has been conserved in different cellular processes that require cell polarization, we examined whether it played a role in the regulation of gastrulation movements in zebrafish. One of the main cellular rearrangements during gastrulation in zebrafish and Xenopus is convergent extension. In convergent extension, cells move to the dorsal side of the gastrula to redistribute there along the forming anteroposterior body axis. Convergent extension movements are driven by mediolateral cell intercalations that require prior mediolateral polarization of cells. In recent studies, it has been shown that convergent extension movements depend on some of the same proteins that are responsible for organizing planar polarity in Drosophila, including zebrafish homologs of dishevelled and strabismus/Van Gogh (Heisenberg et al., 2000; Wallingford et al., 2000; Park and Moon, 2001).

We identified two zebrafish widerborst homologs (wdb1 and wdb2), that both clearly fell into the α/ε subclass of B′ regulatory subunits (Fig. 1D); the proteins they encode are 63 and 64% identical to Dm and Wdb, respectively. Zebrafish wdb1 and wdb2 are maternally provided and expressed during all stages of gastrulation (data not shown). To address their role(s) in convergent extension, we injected wdb1 and wdb2 morpholino antisense oligonucleotides alone and in combination into one-cell-stage embryos and examined the resulting phenotypes at bud stage (Table 2). Embryos injected with low concentrations of wdb1 and wdb2 morpholinos exhibited a shortened and broadened body axis at the end of gastrulation as monitored by in situ hybridization using notail (notochord), dlx3 (anterior edge of neural plate) and hgg (prochordal plate) as markers (Fig. 7A-D). This phenotype suggests that the embryos are defective in convergent extension movements.

Embryos injected with higher concentrations of the same morpholinos were strongly dorsalized as seen by an expanded domain of gsc expression in the presumptive shield anlage at the onset of gastrulation (Fig. 7E,F). These data suggest that zebrafish Wdb regulates dorsoventral axis formation, as well as convergent extension, but that convergent extension is more sensitive to the dose of Wdb.

For both phenotypes, injections of zebrafish wdb1 and wdb2 morpholinos alone caused similar although sometimes weaker phenotypes than injections of a combination of both. These data may suggest that, although Wdb1 and Wdb2 perform the same functions, sufficient protein levels are only attained when both genes are expressed. Alternatively, they may have non-redundant functions in both dorsoventral axis formation and gastrulation.

As the Drosophila and zebrafish Widerborst proteins show a high degree of sequence conservation, we also tested whether injection of the Drosophila dominant-negative widerborst and PP2A-C subunit constructs interfered with convergent extension movements in zebrafish. Injections of both dominant-negative constructs caused reduced convergent extension movements while dorsoventral patterning was largely unaffected, indicating that the Drosophila PP2A B′ α/ε subunit fulfills similar functions in zebrafish and Drosophila (Table 2).

The observation that tissue polarity gene products organize proximal and distal cortical domains in the wing epithelium was the first step towards a cell biological understanding of tissue polarization. Nevertheless many aspects of tissue polarization remain mysterious. In particular, the nature of the signal that initiates polarization of the cortical domains and defines their axis is not known. Furthermore, it is clear that additional, frizzled-independent, mechanisms operate in the wing to polarize the cytoskeleton during wing hair formation and to align the axes of hair polarity between cells. Up until to now, there have been few clues as to the nature of these processes. In this work, we identify a B′ regulatory subunit of PP2A, Widerborst, that is required both for cortical polarization and for hair outgrowth.

Widerborst is unique in that it does not colocalize with other tissue polarity proteins at the cell cortex. Instead, as cortical polarization is beginning (18-24 hours apf), it is found on microtubules on the distal side of each wing epithelial cell. Furthermore, it localizes there before obvious organization of proximodistal cortical domains, and its polarization is independent of them. Strikingly, at earlier developmental stages (7-9 hours apf), Wdb polarity is not distal but proximal. These dynamic shifts in Wdb polarity and their independence from previously described tissue polarity genes suggest the existence of a novel polarization mechanism.

How might Wdb operate to specify cortical polarity? When Wdb activity is reduced, components of the cortical domains like Dsh and Fmi accumulate uniformly around the cell cortex at high levels. By contrast, disruption of Frizzled signaling interferes with the accumulation of Dsh and Fmi at the cell cortex (Usui et al., 1999; Axelrod, 2001; Shimada et al., 2001). This suggests that Wdb is not required to activate Frizzled signaling, but rather is important for making it asymmetric.

Our genetic data indicate that Wdb exerts its activity by activating the catalytic subunit of PP2A with respect to specific substrates, and the localization of Wdb suggests that it does so on the distal side of the planar microtubule web. Which proteins might be targeted for dephosphorylation by Widerborst? One possibility is Dishevelled. Heterozygosity for wdb strongly suppressed the mwh phenotype of dsh¹ suggesting that, during tissue polarization, these two proteins act antagonistically (M. H., unpublished). Dishevelled cortical localization correlates with hyperphosphorylation (Yanagawa et al., 1995; Axelrod, 2001), and the cortical localization of Dsh is certainly expanded in Wdb dominant-negative expressing cells. Supporting this possibility, two-hybrid experiments have indicated that Dishevelled can physically interact with a Xenopus B′ regulatory subunit (Ratcliffe et al., 2000). If Wdb normally acted by antagonizing Dsh, then the dominant-negative might overactivate Frizzled signaling and cause defects in tissue polarity. This model is not easily reconcilable with a role for the distal localization of Wdb; one might naïvely expect an antagonist of Frizzled signaling to accumulate proximally instead of distally. Nevertheless, although the early distal localization of Wdb is suggestive, we have not proven that it is relevant to cortical polarization; for example, Wdb might have a role in transducing the Frizzled signal, for which distal localization is not required.

What might be the importance of Wdb binding to the distal microtubule web? Binding to the cytoskeleton might simply allow stable distal localization of an otherwise diffusible cytosolic molecule. More interestingly, this association raises the possibility that Widerborst directs the dephosphorylation of a microtubule-associated protein. Consistent with this idea, the structure of the planar microtubule web is disrupted by dnWdb expression. PP2A activity is important for the accumulation of stable microtubules (Gurland and Gundersen, 1993), presumably through its effects on the phosphorylation state of MAPs (Sontag et al., 1996; Gong et al., 2000a; Gong et al., 2000b). Microtubule stability can affect the binding of microtubule motor proteins (Gagnon et al., 1996; Liao and Gundersen, 1998) and can contribute to polarized protein delivery (Pous et al., 1998). In the wing, microtubules have been suggested to play important roles in hair polarity; depending on the time at which it is added, vinblastine treatment of pupal wings causes either failure of hair outgrowth or the formation of multiple wing hairs (Turner and Adler, 1998). Polarized dephosphorylation of MAPs within the planar microtubule web might bias the transport of vesicles containing components of the proximodistal cortical domains. At later stages, it might also help direct transport of components of the hair formation machinery to the distal side of the cell, or promote the stability of microtubules in the outgrowing hair. This model for Widerborst action could provide a single explanation for its effects on hair outgrowth and on cortical polarity. Identification of the relevant Widerborst substrate(s) should greatly advance our understanding of the cell biology of tissue polarization.

Our data also support other studies indicating that B′ α/ε regulatory subunits antagonize the classical Wnt signaling pathway. Experiments in Xenopus embryos and tissue culture cells have shown that increasing the level of a B′ α subunit inhibits Wnt signaling and causes ventralization (Li et al., 1995; Seeling et al., 1999; Ratcliffe et al., 2000). Consistent with this, our experiments show that reducing Wdb levels causes dorsalization of zebrafish embryos. Although Wdb, like Frizzled and Dishevelled, is a shared component of both planar polarization and classical Wnt signaling pathways, it probably has different functions in each; during classical Wnt signaling, the B′ α is thought to act downstream of Dishevelled, forming part of a β-catenin degradation complex (Seeling et al., 1999; Li et al., 2001) that plays no role in planar polarity signaling.

The observation that widerborst is needed both for distal polarization of Drosophila wing hairs and for convergent extension movements during zebrafish gastrulation points to a conserved role for Wdb in regulating tissue polarity in development. Furthermore, it provides additional evidence supporting the conjecture that components of the planar polarization pathway in Drosophila are also used to control cell polarity and movement during vertebrate gastrulation (Heisenberg et al., 2000; Tada and Smith, 2000; Wallingford et al., 2000). To date, the evidence for this is based on analysis of various dsh constructs and, more recently, on the analysis of vang/stbm and rhoA during vertebrate gastrulation (Habas et al., 2001; Park and Moon, 2001). Our identification of Wdb as another shared component provides further evidence that this signaling cascade is indeed conserved between Drosophila and vertebrates. Additional experiments will have to address the precise function(s) of vertebrate wdb homologs and where wdb acts in the genetic pathway regulating vertebrate gastrulation movements.

We gratefully acknowledge Bianca Habermann for assistance with bioinformatics, Jens Rietdorf and Arshad Desai for help with deconvolution, and Tadashi Uemura and Rick Fehon for providing antibodies. Arshad Desai, Christian Dahmann, Tony Hyman and Elly Tanaka provided helpful comments on the manuscript. Part of this work was performed at the EMBL in Heidelberg.