Planar polarization of the denticle field in the Drosophila embryo: Roles for Myosin II (Zipper) and Fringe

Abstract

Epithelial planar cell polarity (PCP) allows epithelial cells to coordinate their development to that of the tissue in which they reside. The mechanisms that impart PCP as well as effectors that execute the polarizing instructions are being sought in many tissues. We report that the epidermal epithelium of Drosophila embryos exhibits PCP. Cells of the prospective denticle field, but not the adjacent smooth field, align precisely. This requires Myosin II (zipper) function, and we find that Myosin II is enriched in a bipolar manner, across the parasegment, on both smooth and denticle field cells during denticle field alignment. This implies that actomyosin contractility, in combination with denticle-field-specific effectors, helps execute the cell rearrangements involved. In addition to this parasegment-wide polarity, prospective denticle field cells express an asymmetry, uniquely recognizing one cell edge over others as these cells uniquely position their actin-based protrusions (ABPs; which comprise each denticle) at their posterior edge. Cells of the prospective smooth field appear to be lacking proper effectors to elicit this unipolar response. Lastly, we identify fringe function as a necessary effector for high fidelity placement of ABPs and show that Myosin II (zipper) activity is necessary for ABP placement and shaping as well.

Introduction

It has become appreciated in recent years that cells within an epithelium exhibit a property termed planar cell polarity, or PCP, which refers not to apical–basal polarity of epithelial cells, but polarization within the plane of the epithelium. This axis of polarization allows epithelial cells to coordinate their development to that of the tissue in which they are ensconced (reviewed in Adler, 2002; Strutt, 2003). The phenomenon has been observed in many tissues in the fly, including directional rotation of each ommatidial unit in the eye, orientation of sensory bristles on the abdomen, thorax and legs and hairs on the wing. In vertebrates, PCP has been implicated in hair and feather orientation, sensory stereocilia of the inner ear, stereotypic beating of epithelial cilia and the gastrulation movements of convergence–extension (Adler et al., 2000; Tilney and Saunders, 1983; Wallingford et al., 2000) (reviewed in Adler, 2002; Strutt, 2003). The common theme linking these disparate morphogenetic events is the organization and dynamic control of actin-rich structures at cell interfaces (reviewed in Adler, 2002; Klein and Mlodzik, 2005).

For example, cells in the developing wing of Drosophila can distinguish their proximal from distal vertex. Distal cell edges are where the actin-based prehair extension initiates. The current model for prehair initiation invokes a set of core PCP proteins, with frizzled (fz) and disheveled (dsh) asymmetrically positioned at the distal vertex and a distinct molecular complex located proximally (Axelrod, 2001; Shimada et al., 2001; Strutt, 2001; Tree et al., 2002). Mutation of any of the core PCP components causes initiation and extension of the actin prehair from a central position on the cell surface (reviewed in Adler, 2002) (Krasnow and Adler, 1994; Shimada et al., 2001; Wong and Adler, 1993). In current models, these core PCP proteins play some role both in generating the polarity and, through a feedback mechanism, amplifying and sustaining it, though even in the same organism (Drosophila) there are striking differences, say, comparing wing hair cells and ommatidia (Winter et al., 2001; Yang et al., 2002). Whether this process is under the influence of short or long-range organizing interactions is not clear in any tissue.

Since dynamic actin-based structures are often a key target within cells of the tissue exhibiting PCP, it is perhaps not surprising that actin-modulatory proteins have also been implicated in this process. These include nonmuscle myosins and Jun and Rho kinases as suggested downstream effectors (Eaton, 1997; Eaton et al., 1996; Fanto et al., 2000; Strutt et al., 1997; Weber et al., 2000; Winter et al., 2001). For example, in wing cells, proteins such as these are thought to govern properties such as the number of prehair initiation sites (Winter et al., 2001). Still, a clear understanding of how PCP information is translated into the necessary cell biological outputs is lacking.

More recently, the number of processes during development known to rely on the generation of PCP has increased. Convergence and extension gastrulation movements in the fly have been shown to rely on cell polarization, and this is not established by the core PCP proteins previously characterized (Bertet et al., 2004; Zallen and Wieschaus, 2004). Rather, local cell interactions, mediated by transcription factor deployment at the blastoderm, appear to engage the heavy chain of nonmuscle Myosin II (zipper, zip), its regulatory light chain (spaghetti squash, sqh) and at least some members of the Par complex (Par3; bazooka) to police the orderly cell intercalations necessary for axial elongation (Bertet et al., 2004; Zallen and Wieschaus, 2004). The realization that this tissue exhibits polarity during convergence and extension, together with the identification of some of the effector components involved, has broadened our appreciation for the possible mechanisms involved in planar polarization. It also expands the possibilities of identifying components and principles both common and unique to the establishment and action of PCP in various tissues and across species.

Here, we focus on the embryonic epidermis in Drosophila. A major attraction for us is that there exists a reasonably thorough understanding of the signaling interactions that play out during the patterning of this epithelium, starting from the deployment of a transcription factor cascade at blastoderm, through the detailed signaling interactions acting among cells locally near the time of cell differentiation (Alexandre et al., 1999; Gritzan et al., 1999; Wiellette and McGinnis, 1999) (reviewed in DiNardo et al., 1994; Hatini and DiNardo, 2001a; Sanson et al., 1999). Early on, the body plan is divided into parasegments, and cells at the anterior and posterior border of each produce Hh and Wg, respectively, and these two signals cooperate to generate the differentiated pattern (Alexandre et al., 1999; Bejsovec and Martinez-Arias, 1991; Bejsovec and Wieschaus, 1993; Bokor and DiNardo, 1996; Dougan and DiNardo, 1992; Gritzan et al., 1999; Heemskerk and DiNardo, 1994).

The pattern is etched in the cuticle that is elaborated apically from these epithelial cells. At the time of cell differentiation, a ventral parasegment has roughly 15 rows of cells, and the cuticle is either undecorated (smooth), emanating from about eight rows of hexagonally packed cells, or decorated with denticles, emanating from the remaining seven rows, which are more rectilinear in arrangement (see Fig. 1) (Lohs-Schardin et al., 1979; Walters et al., 2005). Each cell row within a denticle belt elaborates one to three denticles with a characteristic directional hook at its tip. Each denticle is an actin-based cellular protrusion from the apical epithelial cell surface (Dickinson and Thatcher, 1997; Hillman and Lesnick, 1970) (reviewed in Martinez Arias, 1993), and an indelible mold of the extension is created when cuticle is deposited and hardened.

Fig. 1.

The prospective denticle field is polarized. (A) Wild-type cuticle preparation highlights the approximately seven rows of denticles (numbered 1–7) in phase contrast. Rows 1 and 4 point anteriorly; rows 2 and 3 point posteriorly; row 5 denticles are larger and point posteriorly; row 6 and 7 denticles are more closely spaced, smaller and have no obvious orientation. Note the vertical (dorso-ventral) alignment of each row. (B, C) Apical, en face view of ventral epithelium (stage 14 embryo) showing anti-phosphotyrosine (red) and Moesin-GFP (green; sGMCA3.1) double label. (B) Merge; note the rectilinear contours of the aligned denticle field cells (D-belt) compared to the more hexagonal packing among smooth field cells. Top and middle arrows highlight cell edges along one row that are aligned. Bottom arrow highlights an early sensory cell within the smooth field. That the protrusions are at cell edges is more easily appreciated in the magnified views: (C) Moesin-Abd-GFP; (C′) anti-phosphotyrosine; (C″) merge. (D) ABPs map to posterior cell edges. (D) Engrailed-GAL4×UAS-GFP (green in Merge panel D″). The arrow points to the posterior Engrailed cell, which secretes denticles of the first row (Dougan and DiNardo, 1992). (D′) Anti-phosphotyrosine highlights forming protrusions, as well as cell contours (red in Merge panel D″). Note that denticles of the first row are at the posterior edge of the Engrailed cell. The denticles of following cell rows are each at their cell’s posterior edge. Anterior is to the left in all figures, unless noted otherwise. Scale bar in A = 10 μm, B = 30 μm, C and D = 7 μm.

The actin reorganization events associated with denticle formation have analogs in most epithelia. At the core of microvilli, bristles and stereocilia are a parallel actin bundle, composed of filaments of uniform polarity cross-linked by proteins, such as Villin, Espin and Fascin (reviewed in DeRosier and Tilney, 2000). Because the plasma membrane of the cell conforms to the actin bundles, the bundles are the scaffolds that must determine the dimensions of these fingerlike cellular protrusions. However, just how actin bundle dimensions, i.e., shaping, are controlled is quite unclear.

During the stages leading up to cell differentiation, Hh and Wg help generate the pattern through the subdivision of the parasegment into smaller signaling territories. One such territory is a focus for the production of the ligand, Spitz, which activates the EGF receptor. This territory helps accomplish the first task: the decision whether or not to make a denticle. Competition between Wg and Spitz signaling defines whether a cell makes smooth (Wg-activated) or denticulate (Egf-R-activated) cuticle (O’Keefe et al., 1997; Szüts et al., 1997). That competition results in the expression of a master regulator for denticle formation, the transcription factor Shaven baby (Svb/Ovo) (Payre et al., 1999). Svb/Ovo is expressed only in cells slated to make denticles; it is activated by EGF-R and repressed by Wg signaling. svb mutants lack most denticles (Wieschaus et al., 1984), and ectopic Svb expression in the midst of cells slated to make smooth cuticle redirects them to make denticles. In addition to a role in establishing the Spitz territory, Hh and Wg signaling also set the anterior and posterior boundaries for another signaling territory — that comprised of Serrate, the ligand for Notch activation. We recently defined a crucial role for Serrate–Notch signaling in adjusting the breadth of EGF-R activation, which in turn affects the extent of the Svb expression domain and, thus, the width of the denticle field (Walters et al., 2005).

Here, we describe two cell biological properties of the epidermal epithelium of Drosophila embryos. First, cells comprising each denticle row align precisely, such that their anterior edges line up and their posterior edges line up. Alignment requires Myosin II function, and we find that MyoII is enriched at anterior and posterior cell edges of all cells during the stages during which denticle field cells align. This implies that actomyosin contractility, in combination with other denticle field-specific effectors, helps execute the cell rearrangements involved. Second, the actin-based protrusions (ABPs) formed within denticle field cells come to be positioned uniformly at each cell’s posterior edge. Thus, prospective denticle field cells express an asymmetry, uniquely recognizing one cell edge (the posterior) over others. We also show that cells of the prospective smooth field appear to be lacking proper effectors to elicit such a unipolar response. Lastly, we identify fringe and MyoII function as necessary effectors for high fidelity placement of ABPs.

Materials and methods

Fly strains and embryo derivations

Presumptive null mutations were used for zip, fz and fng: zip¹ (FBal0018862), fz^P21 (FBal0004937), fz^H51 (FBal0004931) (Jones et al., 1996); fng¹³ (FBal0034611) and fng⁸⁰ (FBal0034617). The fng stocks were made yw and balanced over TM6 B Tb P{w+; y+} for cuticle analysis or TM3 Sb P{w+; Ubx-LacZ} for gene expression analysis. sGMCA3.1 is a transgenic fly line where the squash promoter is used to express a GFP-tagged actin binding domain of Moesin (Kiehart et al., 2000); P{w+; UAS-eGFP}, P{w+; Ptc-GAL4} and P{w+; En-GAL4} were from the Bloomington stock center; P{w+; UAS-SlamHA} was from J. Zallen (Lecuit et al., 2002); P{w+: UAS-Svb/Ovo} was from F. Payre (Delon et al., 2003). A stock containing the phosphomimetic variant of Squash, the regulatory light chain of MyoII, employed a transgene using its own promoter and containing two glutamic acid (E) missense substitutions (E20 E21): sqh^AX3/FM7; P{w+; Sqh-E20E21 (FBal0122886 (Winter et al., 2001); Squash-GFP was also expressed off its own promoter in transgenic flies: sqh [AX3]/FM7; P{w+; Sqh-Sqh-GFP} (FBal0119137) (Morimoto et al., 1996; Sisson et al., 1999).

Embryos maternally and zygotically deficient for all frizzled activity were generated by deriving fz^P21/fz^H51 heteroallelic females and mating these to fz^H51/TM3 P{w+; Ubx-LacZ}males. Half of the resulting embryos are both maternally and zygotically deficient for frizzled activity. For antibody stains, these were unambiguously identified by the absence of Ubx-LacZ expression. The LacZ-expressing sibs (zygotically fz+) were used as wild-type controls for comparison. Embryos deficient for zip (MyoII) function were identified among progeny from zip[1]/CyO parents as those exhibiting severely reduced anti-myosin heavy chain staining. As reported previously (Young et al., 1993), by the stage when ABPs are being initiated, which is about the time of initiation of dorsal closure, much of the maternally deposited Zip protein has been depleted, though there is some residual and variable amount remaining.

Ectopic induction of Svb/Ovo

Embryos ectopically expressing Svb/Ovo in individual cells or cell groups were produced by the flip-out technique (Struhl and Basler, 1993). Parents of the genotypes yw P{ry+: HS-Flp}; P{w+: UAS-Svb/Ovo}/MKRS and w; P{w+; Act5C>y+>GAL4} P{w+; UAS-GFP}/CyO were placed in cages at 18°C, and overnight egg collections were heat-shocked for 1 h in a 37°C incubator by overlaying the apple juice agar collection plate with pre-warmed S27 halocarbon oil. Collection plates were then incubated at 29°C (for more robust GAL4 activation) and processed for fixation 8 h later (high concentration formaldehyde-short time, see below) and stained with anti-GFP, anti-phosphotyrosine and phalloidin. Ptc-GAL4 was also used to drive ectopic Svb/Ovo from appropriate crosses by collecting embryos at 29°C overnight and processing for stains as above.

Drosophila Rho kinase inhibitor injections

Wild-type (w¹¹¹⁸) embryos or those expressing a phosphomimetic form of the regulatory light chain (sqh^AX3/FM7; P{w+; Sqh-E20E21}), which confers resistance to the inhibitor, were dechorionated, aligned for injection on a glued coverslip, dehydrated appropriately and overlaid with halocarbon oil. Embryos were injected from the anterior to avoid non-specific damage of more posterior segments where denticle pattern would be scored. For the experiment of Fig. 5, in sequential sessions, embryos were either dehydrated but not injected, injected with vehicle only or with increasing concentrations of the D-Rok-specific inhibitor Y-27632 (Sigma-Aldrich) as indicated (Bertet et al., 2004). An approximately 50-fold dilution was assumed upon injection. The injected embryos were then aged at 25°C for appropriate lengths of time, for either cuticle pattern analysis (Fig. 5I), or to be processed for fixation and staining. In the latter instance, in order to recover embryos, after ageing, the coverslip was tilted to drain off as much halocarbon oil as possible, and the embryos were rinsed gently off the glued coverslip and into a watch glass by a gentle stream of heptane. They were fixed with 4% formaldehyde in PBS:heptane for 15 min and then rinsed with heptane, which was then removed and the embryos picked up again onto a glued coverslip. A tungsten needle was used to remove their vitelline membrane, and the freed embryos were then processed for antibody/phalloidin staining as outlined below. The cuticle preparations were scored by one of us (J.W.W.) and also scored blind by S.D.

Fig. 5.

Inhibition of DRho kinase phenocopies zip mutants. (A) Cuticle; injected with buffer alone, phase contrast. (B, C) Two examples of resulting denticle pattern defects observed after 120 μM injection of the Rok inhibitor. (B) This example highlights shaping defects as some denticles of row 1 and row 4 exhibit ambiguous shaping or reversals hook (arrows). We tallied these defects collectively as shaping defects and also tallied separately reversals in denticle hooking (see histograms, below). This example also shows some ectopic as well as missing denticles. (C) This example highlights irregularities in alignment along rows, as well as missing denticles and reversals of shaping. (D, E) Embryo mock-injected at stage 12 and then aged and processed at stage 15; Phosphotyrosine (green) and phalloidin (red). Robust ABPs are evident, and cells are aligned (arrows). (E) Magnified view of D, showing protrusions positioned at posterior cell edges. The non-homogeneities in phosphotyrosine staining are due to the initial deposition of cuticle which inhibits antibody penetration. (F, G) Two examples of embryos injected with 120 μM Rok inhibitor at stage 12 and then aged and processed at stage 15; phalloidin staining results on stage 15. Note irregularities in size, shaping and placement of protrusions. The inhibitor dramatically affects the accumulation of phosphotyrosine epitopes, so we cannot use this to reveal cell outlines. Outlines are visible from the phalloidin stain, acquired at a higher gain and presented in grayscale. The normal rectilinear alignment of cells within the denticle field is compromised by Rok inhibition. (H) Magnified view of G shows that several protrusions are misplaced to anterior edges or mid-face of cells (arrows). (I, J) Histograms representing quantification of Rok inhibitor experiments. The recipient embryos were either w¹¹¹⁸ (w−) or embryos expressing Sqh E20 E21, and “Mock” were injected with buffer alone, while inhibitor concentrations are as indicted. (I) The number of larval cuticles scored as wild-type for each condition is as follows: for w recipients, buffer-injected, 34 (39 total scored); at 80 μM inhibitor, 2 (28 total scored); 100 μM, 2 (29 total scored); 120 μM, 3 (27 total scored); 240 μM, 0 (40 total scored). For Sqh {E20,E21} recipients, buffer-injected, 43 (44 total scored); at 80 μM, 15 (32 total scored); 100 μM, 22 (42 total scored); 120 μM, 12 (56 total scored); 240 μM, 0 (40 total scored). (J) Fraction of cuticles scored having either aggregate shaping defects or reversals (tallied separately). Scale bar = 10 μm in panels A, D, F, G; 7 μm in B, C; 3 μm in E, H.

Cuticle preparation, immunohistochemistry and in situ hybridization

Embryos were collected on apple juice agar plates, aged for the appropriate time, and either processed to visualize cuticle pattern by phase-contrast microscopy (van der Meer, 1977), or fixed and processed for immunofluorescence. In some cases, dechorionated embryos were processed using a standard fixation protocol in a 1:1 mixture of heptane and 4% formaldehyde, 1× PBS for 15–25 min at room temperature. The formaldehyde in this case was diluted from a 16% stock (Electron Microscopy Sciences) that had been aliquoted and stored at 80°C. Alternatively, and for most stains reported here, we used a high concentration–short fixation, where dechorionated embryos were introduced into vials containing a 1:1 mixture of heptane and 37% formaldehyde for 4–5 min (Teodoro and O’Farrell, 2003). For either fix, embryos were rinsed with heptane to clear excess formaldehyde and pipetted onto a glass slide. Excess heptane was wicked away, and the embryos were picked up onto a strip of double stick tape affixed to a coverslip, placed face up in a watch glass and covered with PBS. A tungsten needle was used to poke the embryos out of their vitelline membranes. This avoided a methanol treatment step normally used in en masse devittelinization and consequently preserved the phalloidin binding sites on filamentous actin. For anti-Zipper, we also tried a third fixation protocol, which involved a short heating cycle, where dechorionated embryos were immersed in a hot Triton X-100 (0.03%)–NaCl (0.4%) solution (TNS) for 30 s and then quickly cooled by addition of excess chilled TNS. Vitelline membranes were removed by shaking the embryos vigorously in a mixture of heptane and methanol (1:1). This fixation procedure appears to highlight the localization of proteins such as Armadillo at the zonula adherens at the expense of other cellular pools (Miller et al., 1989). We noted that anti-Zipper immunohistochemistry can appear different with different fixations, which is why we used live Sqh-GFP for several experiments.

The following antibodies (and dilutions) were used for 2 h at room temperature, unless otherwise noted: rabbit anti-myosin heavy chain (Zipper; 1:500, a gift from Dan Kiehart), rabbit anti-betagalactosidase (1:2000, Molecular Probes), or chick anti-betagalactosidase (1:1000, Abcam); rabbit-anti-GFP (1:2000, Molecular probes); anti-phosphotyrosine (1:500, Upstate Cell Signaling, cat. 06–427). Secondary antibodies, used at 1:400 for 1 h, were conjugated to Alexa^™ (Molecular probes) or Cy3 and Cy5 dyes (Jackson Labs). Dye-coupled phalloidin was used at 1:200 (Alexa-350, Bodipy- or Rhodamine-coupled, Molecular Probes). Stained embryos were mounted in Prolong Gold (Molecular probes), and images were obtained using either a laser scanning confocal microscope (LSM 510), or structured illumination (Zeiss Apotome), and assembled in Adobe photoshop.

Embryo live imaging

Dechorionated, staged embryos, expressing either sGMCA3.1 or Sqh-GFP, were mounted with their ventral side on a glued cover slip, covered with S700 weight halocarbon oil and inverted onto a special slide fabricated to hold an oxygen-permeable Teflon membrane (YSI inc.). Images were obtained using a 63×, 1.3 NAwater (LSCM) immersion objective lens for sGMCA3.1, and a 40×, 0.75 NA lens (Apotome) for Sqh-GFP.

Quantification of actin accumulation

For actin accumulation comparisons, the embryos were collected, fixed and stained in the same tube. A Rhodamine–phalloidin conjugate was used to label actin and also marked cell outlines. Images were obtained by Z-section using a Zeiss Apotome, and appropriate sections of the apical face of the epidermis for stages early 13, late 13 and 14 were gathered for analysis. Three cells of three different animals were divided into six domains progressing from A to P. Mean pixel intensities for each domain type (1–6) were tallied and the process repeated for each relevant stage.

Results

The ventral epidermis exhibits planar cell polarity

To examine the ABPs directly, we stained stage 14 embryos expressing a GFP-tagged form of the actin binding domain of Moesin (sGMCA3.1; (Kiehart et al., 2000) with anti-GFP and doubly labeled using anti-phosphotyrosine, which parallels cortical actin enrichment and therefore nicely outlines cells. Note first the difference in apical cell contours comparing denticle field cells (labeled “D-belt”) to those of the smooth field. Denticle field cells are close-packed rectangles, where the length of their dorsal and ventral edges is shorter than their elongate anterior and posterior edges (Figs. 1B, C′). It is also apparent that anterior edges of cells along each prospective denticle row are aligned, as are their posterior edges (Fig. 1B, top and middle arrows). Contrast this to the cells within the prospective smooth field, which exhibit hexagonal packing (Fig. 1B, bottom arrow denotes an early sensory cell). Thus, denticle field cells are aligned in a manner distinct from prospective smooth cells.

Simple inspection also showed that each cell within the prospective denticle field produced one to three ABPs, although the number depends on the specific denticle row and the position of the particular cell relative to the ventral midline (J.W.W. and S.D., unpublished). A closer examination revealed that every actin protrusion was located at or near an A or P cell edge (Figs. 1C′, C″). Given the alignment of cell edges along each row, the ABPs within each row were also aligned (Fig. 1C, green). This is in agreement with aligned denticles observed in the differentiated cuticle pattern (Fig. 1A).

To distinguish whether the actin protrusions were positioned specifically at the anterior, the posterior or perhaps at either edge of a given cell, we took advantage of the fact that the first row of denticles is formed from a row of Engrailed-expressing cells (Dougan and DiNardo, 1992). The posterior row of Engrailed cells were identified in Engrailed GAL4 UAS-GFP embryos (Fig. 1D). Anti-phosphotyrosine staining showed that the protrusion was positioned at the posterior edge of this cell (Fig. 1D′; red in merge, D″). The positioning of each subsequent row of ABPs follows from this as those within row 2, 3, etc. are each at the posterior edges of their respective cell rows. We will refer to this positioning as unipolar asymmetry. Recently, reported live imaging of the ABPs also supports this contention (Price et al., 2006). We therefore conclude that denticle field cells can discriminate their posterior from their anterior edges since they exhibit unipolar placement of ABPs. It has also been shown that ABPs emerge at an angle, extending posteriorly out over the next cell (Dickinson and Thatcher, 1997; Price et al., 2006). Taken together, these data demonstrate that the prospective denticle field epithelium exhibits planar cell polarity (PCP).

ABPs induced in prospective smooth cells are not placed correctly

Since the prospective denticle field is clearly polarized, we wondered if smooth field cells were similarly polarized but simply did not express a marker, such as the ABPs, that revealed that polarity. To test this, we induced the formation of ABPs among prospective smooth cells by ectopically expressing the transcription factor svb/ovo in small groups of cells in the ventral epidermis (Figs. 2A–I) and marked these cells by co-expression of GFP (Figs. 2,A, B, E, F and I; see Materials and methods). Expression of svb/ovo is necessary and sufficient to induce the formation of ABPs (Payre et al., 1999), so by expressing svb/ovo in the smooth field, we can visualize the localization of these protrusions within the cell. When svb/ovo is induced in the smooth field, the ectopic ABPs did not preferentially localize to the posterior edge of cells. Instead, they showed a stochastic dispersal around the apical surface of the cell (Figs. 2B–E and F–I). Arrows in Figs. 2B–E highlight locations of ectopic and misplaced ABPs. The top arrow in Figs. 2B–E shows an ABP in the center of the cell, while the bottom arrow shows an ABP at the anterior edge of the cell. Both anti-phosphotyrosine and phalloidin stains label these misplaced ABPs, indicating that phospho-epitopes as well as actin are present. The middle arrow in Fig. 2B highlights a sensory cell. Out of the thirty-eight ectopic denticles scored, 63% were mis-positioned (nine on an anterior edge, fifteen placed centrally) and only 37% were localized to the posterior edge of the cell.

Fig. 2.

Unipolar placement of actin-based protrusions is a local property of denticle field cells. Ventral epithelium with F-actin visualized using phalloidin (red in merge panels) and cell outlines with anti-phosphotyrosine (white in merge panels). svb/ovo-positive cells were indicated by staining with anti-GFP (green in merge panels). All embryos are approximately stage 15. (A) HS-Flp Act5c>y+>-GAL4 UAS-GFP; UAS-svb/ovo embryo showing ectopic actin protrusions in svb/ovo-positive smooth field cells. (B–E) Higher magnification of the area outlined in panel A. B, merge; C, phosphotyrosine; D, actin; and E, GFP, marking ectopic Svb/Ovo expression. Top and bottom arrows indicate ABPs located mid-face and on the anterior cell edge, respectively. Other misplaced protrusions are also apparent. Middle arrow in panel B highlights an early sensory cell. (F–I) Another HS-Flp Act5c>y+>-GAL4 UAS-GFP; UAS-svb/ovo example, shown only at high magnification with ovo-positive cells exhibiting ectopic actin protrusions. Stains are as in panels B–E series. Arrows indicate actin protrusions, where upper arrow indicates one placed correctly at a posterior cell edge, while the lower protrusion is located mid-face. (J) Ptc-GAL4; UAS-svb/ovo/UAS-GFP embryo showing numerous ectopic actin protrusions in svb/ovo-positive smooth field cells. Note that these protrusions show no preference for cell boundaries. (K–N) Higher magnification of the area outlined in panel J. Arrows indicate two actin protrusions emerging from middle of cell, stains are as in panels B–E series. Scale bar = 30 μm in low magnification images and 20 μm in high magnification images.

It is also possible that the stochastic ABP localization observed in the svb/ovo-positive cells was not due to the lack of polarity effectors in the smooth field, but instead a simple issue of developmental timing. Since we were using a heat-shock-driven recombinase to induce svb/ovo expression, we did not have precise control over the timing of the recombination event or svb/ovo induction. We sought to remedy this concern by ectopically expressing svb/ovo using Ptc-GAL4 (Figs. 2J–N). Since patched is expressed well before cell fate specification, any ectopic ABPs that are present should have had ample time to localize to the posterior cell edge. However, even when svb/ovo is expressed at this early time point, ABPs in the smooth field showed no preference for the posterior edge of cells and remained stochastically positioned (Figs. 2K–N). These data strongly suggest that smooth cells do not possess a latent ability to place ABPs with the unipolar asymmetry seen among prospective denticle field cells. At the minimum, an effector of unipolar asymmetry must be active only among prospective denticle field cells. Alternatively, unipolar asymmetry is only established late and is restricted to the prospective denticle field (see Discussion). To investigate the unipolar asymmetry further, we next describe the phenomena of ABP formation and eventual placement at the posterior edge of denticle field cells.

The timing of cell alignment and formation of ABPs among denticle field cells

Denticle field cells are aligned and exhibit posterior ABPs at late stages (Stage 14). Since the denticle field is specified between late stage 11 and stage 12 (Alexandre et al., 1999; Gritzan et al., 1999; O’Keefe et al., 1997; Payre et al., 1999; Szüts et al., 1997), we monitored cell alignment and ABP formation from stage 11 onward. At late stage 12, there was little alignment among cells within each parasegment. However, cell contours evolved through stage 14. Note the more aligned rectilinear cell profiles of denticle field cells in Figs. 3C and D compared with A and B.

Fig. 3.

The evolution of actin-based protrusions from a cortical web to the cell’s posterior edge. Apical view of ventral epithelium, examining embryos of increasing age stained for phalloidin. (A) Stage 11/12 embryo; no difference between cells of prospective smooth or denticle field in levels of actin at the apical cell face (shown) or phosphotyrosine (data not shown). Cell contours are just beginning to adopt their distinct alignment within prospective denticle field (bracket). Note the purse-string like accumulation of actin along one such alignment front (Arrow). (B) Early stage 13 embryo; rectilinear organization of prospective denticle field cells is more obvious (bracket). There is also an increase in actin over the apical face of the prospective denticle field cells as a web or meshwork of actin appears. (C) Late stage 13 embryo; actin coalesces as protrusions begin to emerge. Arrows point to several examples where protrusions are off posterior edge of the cell. Phosphotyrosine epitopes exhibit a parallel accumulation (data not shown). (D) Late stage 14 embryo; most protrusions are clearly at the posterior cell edge (eventually all will be). (E) Graph of actin intensity across apical face of cells. Intensity of Rhodamine-labeled phalloidin was measured in six areas (from A to P) in several cells using different animals for embryonic stages Early 13, Late 13 and Stage 14. Intensity was graphed for areas: A edge, 2, 3, 4, 5 and P edge. Note that, during the evolution of ABPs, the placement of actin coalesces changes from being distributed across the cells in Early and Late stage 13 embryos to the posterior edge at stage 14. Scale bar = 10 μm.

Turning to the development of ABPs, at late stage 12, there was little or no enhancement in actin accumulation at the apical surface of the prospective denticle field cells (Fig. 3A bracket) compared to cells within the smooth field. Early in stage 13, a diffuse actin meshwork appeared at the apical surface of prospective denticle cells (Fig. 3B). By late stage 13, the apically enriched actin had coalesced into several patches that appear to represent nascent protrusions (Fig. 3C, arrows). Surprisingly, these nascent protrusions were often located away from the posterior edge of a cell (Fig. 3C). Only later, during stage 14, were ABPs more uniformly at or near each cell’s posterior edge (Fig. 3D). Note that we can observe more fully pointed and curved ABPs at stage 15 (e.g., Fig. 2A).

Quantification of actin accumulation within slices across the apical face of a cell throughout these developmental stages supported the notion of a progression to the posterior edge (Fig. 3E; see Materials and methods). We measured the intensity of Rhodamine–phalloidin-labeled actin by dividing a prospective denticle field cell into six domains (progressing from A to P) and recording the pixel intensities in each of the domains at the apical face of that cell. We repeated the process for 3 cells in 3 different animals at early stage 13, late stage 13 and stage 14. The results showed that, during early and late stage 13, actin was not biased to any cell edge. At stage 14, actin was enriched dramatically at the posterior edge. We conclude that actin first accumulated stochastically on the apical face of the cell and then is later positioned at the posterior edge. We next tested a core PCP component, frizzled (fz), for its potential role in establishing or maintaining unipolar asymmetry among prospective denticle cells.

frizzled plays a minor role in unipolar asymmetry

The frizzled gene is important for polarizing the hairs on the wing and in the abdomen and for ommatidial orientation. We first examined denticle cuticle pattern in the progeny of fz^H51/fz^P21 mothers crossed with fz^H51/TM3 Ubx-LacZ fathers. Most larval cuticles appeared normal, though half of these were expected to be null for fz function. Two of thirty cuticles analyzed did exhibit patterning errors, but these were fusions among segments (data not shown). This might have been caused by a partially penetrant deficit in canonical Wingless signaling since fz is also used in this pathway (where it is redundant with Dfrizzled2). We also examined the ABPs directly in fz-deficient embryos; these were unambiguously identified by the lack of Ubx-LacZ expression (see Materials and methods). All fz-deficient embryos exhibited normal cell alignment of denticle field cells, and most mutants (six of eight) exhibited normal posterior placement of ABPs. Only two mutants exhibited misplaced protrusions, and in these embryos, the defects were restricted to row 1. Price et al. reported fully penetrant, but similarly restricted defects (to rows 1 and 2) for the core PCP mutants fz, strabismus, flamingo and even more weakly so for the PCP-specific mutant of disheveled (dsh¹; (Price et al., 2006). Collectively, the defects observed implicate core PCP genes in denticle field polarization, but the restriction of these defects solely to anterior-most rows suggests a more minor role than expected in executing unipolar asymmetry.

Myosin II mutants affect cell alignment as well as ABP placement and shaping

In selecting other candidate genes for polarization, we noted that, at earlier stages, this epithelium is polarized for convergence and extension movements. That is, during convergence and extension, epidermal cells intercalate with adjacent cells, located either dorsally or ventrally, causing the tissue to elongate along the A–P axis (Campos-Ortega and Hartenstein, 1985; Irvine and Wieschaus, 1994). This intercalation is brought about by a polarized remodeling of cell junctions that is driven by the selective enrichment and activity of Myosin II along A and P cell edges relative to their D and V edges (Bertet et al., 2004; Zallen and Wieschaus, 2004). This led us to test for a role of the nonmuscle Myosin II heavy chain (zip) and its regulatory light chain (sqh) in denticle field patterning.

Larvae zygotically deficient for zip exhibited variable and striking cuticle pattern defects. A few mutants had mild shaping defects, including reversals, but had a relatively normal number of denticles per field (compare Figs. 4A and C, arrows in C highlight reversals). Other mutants were very severe with an accordion-like appearance to the cuticle (Fig. 4E), which is most likely due to the severe dorsal closure defects present in these mutants. This most severe phenotype prevented analysis of the denticles as the cuticle preparations do not expand enough to reveal the denticles well. However, most zip mutant larvae were moderate in severity, and denticle defects were analyzable (Fig. 4D). In these cases, the normally trapezoid-shaped denticle field was ovoid in appearance, and denticle rows meandered rather than being aligned properly. In support of this, cell contours of zip mutants were irregular compared to cells within the denticle field of stage 15 sibling controls (Fig. 4G compared to similarly staged siblings B and B′ above, arrow in B′ highlights aligned cell edges of rows 4/5). In addition, ABPs were less focused, and often not positioned at a cell edge (Fig. 4, compare circles in B′ and G). Finally, there were strong shaping defects as denticles were less broad at their base and more elongate and wavy (circles in D and F). While these data suggest that the Myosin II heavy chain is important for cell alignment, as well as for placement and shaping of ABPs, another possibility is that the defects we observed were secondary to those incurred by the earlier requirement for zip during convergence and extension.

Fig. 4.

zip mutants exhibit striking defects in the shaping and positioning of protrusions, as well as in cell alignment. (A) Wild-type cuticle, phase contrast. (B, B′) Stage 15 sibling control (zip/CyO), stained for phalloidin (B; red in merge) and phosphotyrosine (B and B′, green in merge). Note the characteristic coordinate rectilinear alignment of cell contours within prospective denticle field (arrow in panel B′) and the positioning of protrusions to the posterior cell edge (circle in panel B). (C–E) The range of zip mutant cuticle phenotypes. (C) Note the shaping defects as well as less well-aligned rows of denticles. This zip phenotype was occasionally observed. (D) Moderate zip phenotype exhibited by most mutants (just under a quarter of larvae). The denticle field is reduced, shaping defects are obvious and some denticles are shortened. (E) Occasional zip larvae exhibited no detectable denticles. (F, G) Ventral epidermis of stage 15 homozygous zip mutant embryos, identified by reduced Anti-Myosin heavy chain (Zip) stain (Young et al., 1993). These images are magnified 3-fold compared to panels B, B′ to highlight shaping and cell contour defects. (F) Actin accumulation (phalloidin) reveals striking shaping defects as denticles were less broad at their base, more elongate and wavy (compare to panels B, B′). (G) Actin protrusions were often not positioned at the posterior edge of denticle field cells. Cell contours were much more irregular compared to sibling denticle field cells (compare to similarly staged panels B, B′). Scale bar = 10 μm in panels A–E, H; 3.5 μm in F, G.

Functional role of MyoII during denticle field patterning

To test whether MyoII acts directly during denticle field patterning, we inhibited MyoII function in wild-type embryos after convergence and extension had occurred. Myosin II is activated by Rho kinase (Rok) phosphorylation of its regulatory light chain, encoded by sqh (Amano et al., 1996, 1997; Jordan and Pabo, 1988; Karess et al., 1991; Kimura et al., 1996). Therefore, to inhibit MyoII, we injected a Rho kinase inhibitor Y-27632, which has been widely used, including during convergence and extension in the fly (Bertet et al., 2004). We examined the resultant cuticle pattern of injected larvae as well as cell alignment and ABP formation and placement in embryos. A preliminary set of inhibitor injections into Sqh-GFP-expressing Stage 14 embryos showed a striking loss of localized Sqh (data not shown), similar to that observed by Bertet et al. (2004). Encouraged by this, we executed a series of controlled experiments on earlier-stage embryos.

The inhibitor was injected at increasing concentrations into wild-type embryos just prior to cell alignment and ABP production (stage 12), and in the first sets of experiments, embryos were allowed to develop until cuticle differentiation. Cuticles were first tallied for the fraction with normal denticle patterning (representative wild-type-appearing cuticle in Fig. 5A, where “Mock” is buffer alone). Additionally, cuticles were scored for the fraction having either general shaping defects (denticles less broad at their base and more elongate and wavy than normal) or for shape reversals (i.e., normally anterior-pointing denticles pointing to the posterior) (representative cuticles Fig. 5C, circle Fig. 5B arrows respectively). Both of these phenotypes were observed in zip mutants (refer back to Figs. 4C, D and F). For the injection series, cuticles were scored as having shaping and reversal defects only if there were multiple instances present in more than one denticle belt in the same animal. Note that, when reversals and misshaping were present, they often appeared in the same larva. One control consisted of buffer-only injections, while a second control consisted of injecting inhibitor into transgenic embryos expressing a variant Sqh protein that contains two phosphomimetic substitutions, Sqh {E20, E21}. These substitutions make Sqh less susceptible to Rok inhibition and are used to confirm the specificity of the inhibitor (Bertet et al., 2004; Winter et al., 2001).

We first tallied the fraction of larval cuticles that appeared normal (WT); i.e., those exhibiting neither shaping defect (Fig. 5I). Embryos injected with buffer only were predominantly normal in appearance with neatly aligned rows and well-shaped ABPs, as expected. In contrast, inhibitor injections significantly reduced the fraction of normal larval cuticles (Fig. 5I), and cuticles with reversals and shaping defects increased at each concentration increase (Fig. 5J). Notably, compared to the injection of wild-type embryos, injection of Sqh{E20,E21} embryos showed an increased fraction of normal cuticles (Fig. 5I) and a decreased fraction with shaping defects for each inhibitor concentration tested (except for the highest, 240 μM; Fig. 5J). Since the phosphomimetic Sqh embryos were not fully resistant to Y-27632-induced cuticle defects, there may be some Sqh-independent effects. However, the relative resistance of Sqh{E20,E21} argues that the MyoII regulatory light chain is the target of the inhibitor in our experiments, as expected (Bertet et al., 2004; Winter et al., 2001). Thus, these results suggest that MyoII activation is indeed necessary for ABP shaping during denticle field patterning.

To test directly whether MyoII was necessary for cell alignment and ABP positioning, we aged a subset of inhibitor-injected wild-type embryos until stage 15 and then processed them to visualize actin and cell outlines. Embryos injected with buffer alone had normally aligned cells and ABPs appropriately positioned at the posterior edge of cells (Figs. 5D and E, “Mock”). In contrast, sibling embryos injected with inhibitor displayed defects similar to zip mutants (compare zip mutants in Figs. 4F and G with inhibitor-injected embryos in Figs. 5F–H). These embryos had defects in cell alignments (Figs. 5F and G and data not shown), as well as mis-positioned ABPs at both anterior edges and the center faces of the prospective denticle field cells (Figs. 5G–H, top arrows in H indicates ABP at the anterior edge and bottom arrow indicates ABP at the center of cell, dashed line indicates cell outline). Taken together with our analysis of the zip mutant phenotype, we conclude that Myosin II activity is required for cell alignment and for the placement and shaping of ABPs.

Myosin II accumulates preferentially along anterior and posterior cell boundaries

We next investigated the subcellular distribution of MyoII subsequent to convergence and extension and during denticle field patterning, using anti-Zip stains (data not shown) and live-imaged Sqh-GFP transgenic flies (Karess et al., 1991) (Figs. 6A–E). It has been shown recently that MyoII exhibits uniform apical membrane accumulation prior to convergence and extension (during stage 6), but then during convergence and extension (stage 8) MyoII shifts to a bipolar distribution, enriched along anterior and posterior cell edges (Bertet et al., 2004; Zallen and Wieschaus, 2004) (Fig. 6A, arrow) and reduced along D/V edges (Fig. 6A, arrowhead). Our analysis of stages after convergence and extension revealed further changes in MyoII localization. First, we found that MyoII lost its A–P edge preference by stage 11 and accumulated uniformly at the perimeter of cells (Fig. 6B). However, beginning at stage 12, MyoII again became bipolar, enriched along A and P cell edges, becoming almost exclusively so by stage 13 (Figs. 6C, D; below, we demonstrate that MyoII is on both A and P edges and thus will refer to this as bipolar accumulation). Note also the extended portions of Sqh-GFP accumulation on edges along aligned cells (Figs. 5C and D, arrows). Contrast this to the relatively low level accumulation on dorso-ventral edges along these same cell rows. The redeployment of Sqh-GFP to a bipolar distribution at the onset of denticle field patterning suggests a role for MyoII activity in cell alignments (see Discussion). Finally, from stage 13 on, we noted that MyoII began to co-localize with the ABPs (Figs. 5D, E). By late stage 14, MyoII appeared most strongly on ABPs, though still bipolar with regard to cell edges (Fig. 5E). The enrichment of MyoII on ABPs as they emerge is consistent with a direct role for Myosin II in ABP positioning and shaping, as also suggested by our Rok inhibitor experiments.

Fig. 6.

Redeployment of Myosin II to anterior and posterior cell edges during denticle field patterning. Snapshots from live imaging of ventral epidermis in embryos expressing Sqh-GFP. (A) Stage 8 embryo, during convergence extension movements. Note the polarized accumulation of Sqh-GFP to A–P cell edges (arrow) as there is reduced accumulation on D–V cell edges (such that these are less visible in the micrograph, arrowhead) (Bertet et al., 2004; Zallen and Wieschaus, 2004). Asterisks in panels A and B highlight signal that is an aggregation artifact of the Sqh-GFP fusion protein, we do not see these aggregates when using an antibody to Zipper. (B) Stage 11 embryo, prior to denticle field patterning. Note that Sqh-GFP distributes roughly homogenously around most apical cell contours. (C–E) Embryos during denticle field patterning. (C) Stage 13, note the general enrichment along A/P cell edges (arrows) at the expense of D–Vedges. Some D–Vedges (arrowheads) do show Sqh-GFP to levels equivalent to that at A–P edges. There is also some accumulation on the cell face, likely matching the accumulation of actin in a meshwork at the apical cortex (see Fig. 3B). (D) Early stage 14. Accumulation of Sqh-GFP along A–P edges is still quite apparent (arrows), as well as on nascent actin protrusion (arrowhead). (E) Late stage 14 embryo. Sqh-GFP continues to appear enriched along A–P cell edges and definitively along protrusions. There is also an encirclement of the protrusion at its base. A = anterior, P = posterior, D = dorsal, V = ventral; scale bar = 25 μm.

Given that MyoII is bipolar in cells, during convergence and extension, we wondered if this was the case during denticle field patterning. Light microscopy does not have the resolution required to distinguish whether Myosin II is enriched to anterior, posterior, or both cell edges since all cells express these proteins. To surmount this limitation, we expressed an epitope-tagged form of the Zip binding protein Slam in specific rows of cells using UAS-Slam-HA transgenic flies (Lecuit et al., 2002; Zallen and Wieschaus, 2004). Using either the Ptc- or the En-GAL4 driver, Slam-HA was enriched on A/P versus D/V edges of cells (Figs. 7B, C). This was consistent with that observed in live Sqh-GFP embryos (Fig. 6, and data not shown). Since the extant GAL4 drivers are not restricted to a single row of cells, for each experiment, the informative cell interfaces are only those at the boundaries of the expression domain, where a Slam-HA expressing cell abuts a non-expressing cell. If there is enrichment here, the edge contributed by the expressing cell is the edge where MyoII is enriched. By using the Ptc-Gal4 driver, we can examine Slam-HA accumulation in the anterior edge of row 2 cells, which make row 2 denticles (red asterisk in cartoon Fig. 7A, and micrograph Fig. 7B, arrows). The anterior edge showed Slam-HA accumulation when Ptc-Gal4 was used, suggesting that MyoII accumulates at A edges. Similarly, by using the En-Gal4 driver, we can examine Slam-HA accumulation in the posterior En cells, which make row 1 denticles (blue asterisk in cartoon Fig. 7A and micrograph Fig. 7C, arrows). We observed Slam-HA accumulation at P edges of these cells. Assuming that what is true for the row 1 and row 2 cells holds across the denticle field, we can conclude that MyoII is bipolar within prospective denticle field cells.

Fig. 7.

Zipper/Myosin heavy chain accumulates at both A and P cell edges within denticle field cells. (A) A cartoon cross-section of the ventral epithelium, showing gene expression domains and the correspondence to the prospective smooth or denticle field. Black or white brackets A–C indicate one parasegmental unit. Denticle rows 1 through 7 are indicated as are Wingless-, Engrailed-, Rhomboid- and Serrate-expressing cells (W, E, R, S). The domain of expression of Ptc-GAL4 UAS-SlamHA (red) and En-GAL4 UAS-SlamHA (blue) is indicated by bars. (B) Ptc-GAL4 UAS-SlamHA embryo at stage 14 stained with anti-HA (green; Bʺ) and phalloidin (red; B′). The relevant interface to consider for this experiment is the one between row 2 cells, which express SlamHA (indicated by red asterisk), and the previous row 1 cells, which do not. It is unambiguous that Slam, and, thus, MyoII are enriched at the A edge of row 2 cells (Arrows). (C) En-GAL4 UAS-SlamHA embryo at stage 14 stained with anti-HA. It is obvious that Slam is enriched at cell interfaces. The relevant one to consider for this experiment is the one between row 1 cells (indicated by blue asterisk), which express SlamHA, and row 2 cells, which do not. It is unambiguous that Slam, and, thus, MyoII are enriched at the P edge of row 1 cells (Arrows). We conclude that MyoII accumulates in a bipolar manner on A and P cell edges within prospective denticle field cells. Scale bar = 20 μm.

Our Slam-HA experiments also suggest polarization of MyoII among prospective smooth cells. Slam-HA is enriched on the A edge of the first En cell (a smooth cell) (Fig. 7C, arrowhead) and on the P edge of the prospective smooth cell row just anterior to the first En cell (Fig. 7B″, large arrowhead). Assuming we can extend these observations to other cells of the smooth field, then MyoII appears to be bipolar in smooth cells, as well as within the denticle field.

We conclude that MyoII undergoes several periods of dynamic re-localization within this epithelium. MyoII becomes bipolar during convergence and extension (Bertet et al., 2004; Zallen and Wieschaus, 2004), redistributing homogenously thereafter, only to redeploy in a bipolar manner. Although this last phase correlates with the onset of cell alignment and ABP formation within the prospective denticle field, the late-stage bipolar phase encompasses the whole parasegment, including the prospective smooth cells, just as does the earlier convergence and extension phase. Thus, the whole epithelium exhibits (bipolar) planar polarity (see Discussion). However, we have found that prospective denticle field cells exhibited two behaviors distinct from the smooth field cells (cell alignment and posterior placement of ABPs). This suggests that there is a second layer of polarization (unipolar asymmetry) imparted solely to denticle field cells. Alternatively, the prospective denticle field exhibits a unique response to the parasegment-wide polarization, for which the prospective smooth field is not programmed. We reasoned that a candidate factor for conferring either a second layer of polarization or eliciting a unique response among denticle field cells might be expressed at late stages and within the denticle field. This turned our attention to fringe, which fits both of these criteria.

fringe plays a role in unipolar ABP placement

fringe (fng) is a modulator of Notch signaling and is known to attenuate Serrate–Notch interactions in various tissues (Bruckner et al., 2000; Irvine, 1999; Irvine and Wieschaus, 1994). We previously showed that Serrate is expressed in cell rows 5–7 of the prospective denticle field at embryonic stages 12 to 14 (Walters et al., 2005). We showed that fringe RNA expression is roughly coincident, though temporally delayed, with Serrate expression within the denticle field, and further, that it attenuates Serrate–Notch interactions in this tissue (Walters et al., 2005). We also reported that fringe mutant embryos exhibited numerous, poorly shaped small denticles; these were interspersed among denticles of size and shape appropriate for their row. Further inspection of cuticle preparations now suggested to us that denticle rows were misaligned (Fig. 8C row 1) and also contained denticle reversals (Fig. 8C arrows, compare to 8A, WT), suggesting potential polarity and shaping defects. To investigate this directly, we compared phosphotyrosine-stained late stage 14 embryos that were either wild-type (n = 10), fng heterozygotes (fng/TM3-UbxLacZ, which are phenotypically wild-type; n = 11) or fng homozygotes (n = 9) for defects (representative animals, Figs. 8B and D–F). All fng homozygous mutants exhibited positioning defects in their ABPs, consistent with a role for modulating PCP activity. In these embryos, the poorly shaped ABPs of variable size are interspersed with the well-shaped protrusions of rows 1–5 (Figs. 8D–F, arrows and circles). The poorly shaped protrusions are inappropriately placed at either the anterior edge of the cell (arrows D–F) or the center (circles D–F). The well-shaped protrusions in the same cells were correctly placed at the posterior edge, similar to wild-type. Reversal phenotypes were difficult to score due to the small size of the protrusions. We conclude that fng plays a role as a local effector of PCP within the prospective denticle field.

Fig. 8.

Fringe mutants affect the placement of actin-based protrusions. (A) fng/+ sibling control (fng¹³/TM3 Ubx-LacZ) cuticle, phase contrast. Note normal alignment along rows and proper shaping of individual denticles in their row-specific manner. (B) Stage 14 sibling control (fng¹³/TM3 Ubx-LacZ), stained with anti-phosphotyrosine; (C) fng mutant cuticle. Note the ectopic denticles, some of which are highlighted by arrows. Note also the less regular alignment along rows. (D–F) Three examples of stage 14 fng mutant embryos, stained for phosphotyrosine. Arrows and circles highlight some of the protrusions that are misplaced, being located at anterior cell edges, or mid-face, rather than at the posterior cell edge. Scale bar = 10 μm.

Discussion

Our results demonstrate that the epidermal epithelium of Drosophila embryos exhibits planar cell polarity. We show that intricately shaped ABPs are uniquely placed on the posterior edges of a subset of cells within each parasegment, the prospective denticle field cells. We also demonstrate that Myosin II and Fringe are important effectors for this unipolar asymmetry. Myosin II also plays a role in shaping these protrusions, as well as in the organized cell alignment that develops among denticle field cells. This work establishes a new experimental paradigm for the study of diverse mechanisms involved in epithelial PCP.

Unipolar asymmetry in the prospective denticle field

Denticle field cells elaborate ABPs, and this has allowed us to establish that these cells are polarized in the plane of the epithelium. Forcing smooth field cells to elaborate ABPs failed to reveal any latent unipolarity within this portion of the parasegment. Nevertheless, our evidence supports the proposition that both fields are polarized, albeit in different ways.

We document the bipolar enrichment of Myosin II on prospective smooth and denticle field cells. This suggests that the whole ventral epithelium exhibits bipolar PCP, and all cells can discriminate their A/P from their D/V edges. Since this bipolar enrichment emerges during stage 12, it is possible that this reflects de novo establishment of polarity within the epithelium. However, at earlier stages, cells of this epithelium exhibit a strikingly similar bipolar distribution of Myosin II then used for convergence extension. Thus, the bipolar redeployment of Myosin II might reflect a memory of that earlier polarization. This possibility is particularly compelling given that Myosin II orchestrates the rearrangements of cell junctions necessary for convergence extension (Bertet et al., 2004; Zallen and Wieschaus, 2004). Perhaps Myosin II is being engaged similarly at the later stages, and the re-emergence of a bipolar preference is a precondition to accomplish the necessary junctional reorganization for cell alignments. What signals direct this re-emergence are not yet known. In addition, since Myosin II is bipolar among both prospective denticle and smooth cells, though the latter do not align, there must either be cues or effectors specific to the denticle field that initiate the cell alignment process.

Note also that the bipolar enrichment of Myosin II yields no clues as to how the denticle field cells uniquely identify their “P” cell edge and faithfully position the ABPs to this edge. One possibility is that only the prospective denticle field cells have proper effectors to transmute the global bipolarity discussed above into asymmetric unipolarity. Since it is difficult to imagine how this might occur, a second possibility is that unipolarity is imparted locally, only across the prospective denticle field. The failure to observe proper ABP placement after ectopic expression of svb/ovo in the prospective smooth field supports this possibility. If unipolarity is imparted locally, then this also places constraints on the timing of the signals for unipolarity. The epithelium is sorted into smooth and denticle fields only late in development as a consequence of the antagonism between Wingless signaling and EGF receptor activation (O’Keefe et al., 1997; Szüts et al., 1997). This results in the establishment of the domain of expression for Svb (Delon et al., 2003; Payre et al., 1999; Walters et al., 2005). Thus, the denticle field is only established after this time, and we favor the idea that unipolarity is assigned after this stage. We believe that our analysis of Fringe supports this contention: fringe comes to be expressed within denticle field cells only after this stage (Walters et al., 2005) and mutation of fringe interferes with unipolarity across the whole denticle field. This identifies fringe as, at the minimum, an effector of denticle field unipolarity.

Among a set of core genes involved in the establishment and maintenance of polarity in other tissues is frizzled (reviewed in Adler, 2002; Strutt, 2003), and we were surprised that it did not play a major role in denticle field polarization. Only two of eight embryos exhibited any defects, and in these, only row 1 cells appeared affected. Yet, we do not rule out a role for frizzled signaling in denticle field unipolarity due to possible redundancy with DFrizzled 2 (see below). In fact, recent work by Price et al. implicated several core PCP genes in denticle field unipolarity (Price et al., 2006). Mutation of frizzled, dsh, flamingo or strabismus led to mild defects restricted to rows 1 and 2 reminiscent of what we report here for the minority of fz embryos. In addition, they report enrichment for flamingo, fz and dsh on the edges of prospective denticle field cells. In aggregate, these data are very suggestive for a role of core PCP genes in ABP polarity. However, the restriction of the phenotype to the very anterior rows of the denticle field leaves open the possibility that these genes do not play as major of a role as they do in, for example, the wing. In addition, while Price et al. show that Fz and Dsh are enriched on certain cell edges, it will be of interest to know whether the enrichment is uniquely to one edge of cells, as it is in wing cells (Axelrod, 2001; Strutt, 2001). For example, our data show that Zipper and Squash-GFP are enriched to cell edges but that these are likely both A and P edges of each cell. In denticle field cells, if Fz and Dsh exhibit such bipolar enrichment, rather than the unipolar asymmetry observed in wing cells, then their role in the denticle field would be quite distinct from that currently proposed for wing cells (Axelrod, 2001; Shimada et al., 2001; Strutt, 2001).

It will be difficult to establish definitively whether Fz or Dsh play more extensive roles in denticle field polarity, especially as our data strongly suggest that effectors for this polarity are likely established late, after the epithelium is sorted into smooth versus denticle field cells. It would not be possible to easily interpret the removal of all function for dsh and frizzled (which likely would entail the removal of both frizzled and Dfrizzled2) because both proteins play earlier and essential roles in Wingless signal transduction. Removing both maternal and zygotic dsh (or fz Dfz2) function might well lead to polarity phenotypes, but these may be secondary to earlier deficits in Wg signaling. This is especially the case as Wg (and Hedgehog as well) plays a major role in establishing the denticle versus smooth field and subdividing parasegments into smaller signaling territories as well as establishing the responses of the cells to those signals (Alexandre et al., 1999; Gritzan et al., 1999; Hatini and DiNardo, 2001b; Wiellette and McGinnis, 1999). These considerations also raise a caution in drawing the conclusion from Wg or Hh null mutants that these pathways play any direct role in polarization.

Cell alignment in the prospective denticle field

During denticle field patterning, while the smooth cell field remains largely hexagonally packed, striking alignment occurs along each row of denticle field cells. This rearrangement requires cooperative interactions among cells along a file and likely involves directed modulation of adherens junctions among these epithelial cells to cause the straightening observed. One way to coordinate such a cellular response would be to recruit the cortical actomyosin network. Our data suggest this to be the case. Coincident with the emergence of this alignment, actin, Zipper, and Squash become enriched and aligned from cell to cell along a row. The placement and timing of this enrichment suggest that the cell shape change observed is driven by the action of Myosin II. We suggest that adhesion among cells along a row couples actomyosin function in the cortex of one cell to that in the next cell in a manner that straightens initially irregular edges between cells and, thus, aligns the cell rows. This may be similar to other examples of coordinate changes within cell fields, such as convergence extension and the “actin purse string” utilized during dorsal closure (Bertet et al., 2004; Kiehart et al., 2000; Zallen and Wieschaus, 2004). Our inhibition of Rho kinase activity prior to the cell alignments lends support to this idea as we find that denticle field cells did not align, but, rather, appeared less rectilinear, and similar to that observed in zip mutant embryos. Thus, partial inhibition of MyoII function, which likely leads to decreased actomyosin contractility, precludes the cellular alignment normally observed. It will be interesting to investigate the dynamics of junctional reorganization during the alignment process (Bertet et al., 2004).

Initiation and posterior placement of actin-based protrusions

The phenomena that ABPs are placed at a specific cell edge have been described for wing prehairs and now the ventral epidermis. The fact that Rok and MyoII are involved in the actin organization events was also described for wing cells, though the particulars differ (Winter et al., 2001). Winter et al. show that selective inhibition of Drok increases the number of prehair initiation sites, but polarity (an overall distal vector) is mostly maintained. We found that fewer ABP initiations were present after inhibition of Drok; and we also found more extensive involvement as placement, cell alignment and shaping were affected. Two additional features contrast our paradigm from that of wing prehair as it is currently understood: the number of ABPs per cell and progress of actin accumulation to a cell edge.

While wing cells make only one ABP per cell, a ventral epidermal cell will make up to three. In fact, increases in the number of actin hairs have been a basis for isolating mutants in maintenance or effector genes for PCP in wing tissue (reviewed in Adler, 2002). This difference may not be surprising given that wing and ventral epidermal cell contours differ. The exaggerated rectilinear shape and dramatic alignment we describe for ventral epidermal cells may hold the key to this difference. One can imagine that actin-based accumulations at cell edges may become separated as the length of that edge increases. Interestingly, it seems that the most elongated cells in the prospective denticle fields have more protrusions (unpublished observation, J.W.W. and S.D.) Perhaps some additional support for this idea comes from the fact that larger cells in the wing will tend to produce more prehairs (Adler et al., 2000).

A second feature distinguishing the ventral epidermis from current descriptions of the wing is that ABPs in the ventral epidermis emerge as apical enrichments of actin, only later to coalesce and localize to the posterior edge (Figs. 7C and F). It will be interesting to see if this is also the case in the wing. Among denticle field cells, while the mechanism for the final placement remains to be identified, two models can be envisioned: actin coalesces and these budding protrusions diffuse randomly over the apical cortex but can be captured or stabilized solely at the posterior edge; or actin coalescence(s) are directed to a posterior edge. Recent work by Price et al. suggests that actin coalescence(s) “condense” to the posterior edge (Price et al., 2006), implying directed movement. It will be valuable to distinguish between these two models as a way to narrow potential players involved in denticle field PCP.

Whichever model is true, it is clear that Myosin II is required genetically for final posterior edge placement. What is not yet clear is whether this requirement is direct or reflects a requirement for proper cell alignment to place the protrusion posteriorly. We do not believe this to be the case. First, we note that Zipper protein continually associates with the protrusions during localization (and through shaping, see below). Second, fringe mutants suggest that cell alignment can be separated from posterior placement, so there is no reason why Zipper cannot be utilized for both processes. Denticle field cells in fng mutants largely maintain their narrow A–P and elongated D–V contours. Thus, fng mutant cells are competent to carry out instructions to change their overall shape and align, but fail to implement instructions to place actin protrusions faithfully at an edge. This implies that the control of cell contours and of proper placement of ABPs are separable. Our working hypothesis is that actomyosin contraction is necessary for mobilization or capture uniquely to the posterior edge of denticle field cells. The two models suggested above provide a framework for exploring this hypothesis.

We believe that fringe plays a role as an effector of unipolar asymmetry, as in fringe mutants, we observed defective ABP placement in most denticle rows and in all mutant embryos. The involvement of fringe in posterior edge placement of ABPs, of course, implicates Notch signaling in this process. However, we have not yet been able to test for roles for Notch or its various signaling components as Notch plays several earlier roles in ventral patterning, including one in specifying the width of the prospective denticle field itself (Walters et al., 2005). Intriguingly, recent work supports the existence of a non-transcriptional branch of the Notch pathway in the wing that influences F-actin (Major and Irvine, 2005). This may provide a toehold for investigating the role of fng and Notch signaling as effectors of unipolar asymmetry.

Shaping of actin-based protrusions

In the wild-type cuticle pattern, denticles are aligned along rows. We have shown that this is clearly a result of cell alignment among denticle field cells and posterior placement of the ABPs that prefigure the denticles. While important questions remain to be addressed on each of these points, denticles are not only shaped intricately, but those shapes are coordinated along rows such that the direction of the hook changes between certain rows. That aspect of denticle patterning remains elusive.

The actin reorganization events associated with denticle formation have analogs in most epithelia. At the core of microvilli, bristles and stereocilia are a parallel actin bundle, composed of filaments of uniform polarity cross-linked by proteins, such as Villin, Espin and Fascin (DeRosier and Tilney, 2000). However, just how actin bundle dimensions are controlled is quite unclear and a very active line of research. Among the many actin-based extensions, bristles and hairs undergo shaping as they taper towards their ends. It is not understood what generates this change in shape nor how it is regulated. We believe the ventral epidermis is a tractable paradigm for elucidating the shaping of the denticles in our system, allowing us to define the signals that cause shaping to occur.

We show that Myosin II plays a role in shaping and have observed its localization to ABPs before edge placement. Selective inhibition of Myosin II after convergence extension leads to many ABPs that have a longer, wavy appearance than mock-injected siblings. Additionally, Price et al. observes “donut-like” shapes at the base of actin accumulations before elongation, and we occasionally see suggestive donut shapes in Sqh-GFP (Fig. 6E) or Anti-Zipper stains, though these appear sensitive to fixation conditions (J.W.W. and S.D., unpublished observations). These observations suggest an additional role for Zipper in perhaps “corralling” protrusive determinants into coordinated shapes. Donut-like shapes that Price et al. observed for actin are reminiscent of the localization of harmonin, an F-actin bundling protein that interacts with Myosin VIIa and is required in patterning stereocilia (Boeda et al., 2002; Price et al., 2006).

We suggest that signals are deployed in the denticle field that choreograph denticle shaping and coordinate those shapes along cell rows. Once the denticle field is specified, svb/ovo expression and function are clearly necessary and sufficient to impart the instructions to a cell to make an ABP. While it has been assumed that Svb/Ovo can “…initiate the entire process of cytoskeletal remodeling…” (Delon et al., 2003), we show here that Svb/Ovo-initiated protrusions within the smooth cell field are neither positioned posteriorly nor shaped well. Thus, our data suggest strongly that the genes induced by the transcription factor Svb/Ovo must collaborate with intradenticle field factors and signals to carry out these processes properly. One hypothesis is that the interfaces between adjacent signaling territories lead to row-specific denticle shaping (Alexandre et al., 1999; Struhl et al., 1997). Interestingly, Price et al. nicely describe the associations of actin-modulatory proteins at relatively late stages in ABP organization. The activity of such proteins may well be influenced by local signaling, and their coordination with locally deployed determinants may be one way to properly shape the protrusion. Indeed, investigation of links between fng, other local factors and PCP will add new depth to our understanding of this already fruitful system.