Drosophila Furrowed/Selectin is a homophilic cell adhesion molecule stabilizing Frizzled and intercellular interactions during PCP establishment

Summary

Establishment of planar cell polarity (PCP) in a tissue requires coordination of directional signals from cell to cell. It is thought that this is mediated by the core PCP factors, which include cell adhesion molecules. Here, we demonstrate that furrowed, the Drosophila Selectin, is required for PCP generation. Disruption of PCP in furrowed-deficient flies results from a primary defect in Fz levels and cell adhesion. Furrowed localizes at/near apical junctions, largely co-localizing with Frizzled and Flamingo (Fmi). It physically interacts with and stabilizes Frizzled, and further, it mediates intercellular Frizzled-Van Gogh (Vang)/Strabismus interactions, similarly to Fmi. Furrowed does so through a homophilic cell adhesion role that is distinct from its known carbohydrate-binding function described during vertebrate blood-cell/endothelial cell interactions. Importantly, the carbohydrate function is dispensable for PCP establishment. In vivo studies suggest that Furrowed functions partially redundantly with Fmi, mediating intercellular Frizzled-Vang interactions between neighboring cells.

Introduction

Polarization of cells plays important roles in the maintenance of tissue integrity. Besides the characteristic apical-basal polarity, the most evident feature of epithelial cells, epithelial sheets often adopt a second type of polarity: Planar Cell Polarity (PCP; rev. in Goodrich and Strutt, 2011; Klein and Mlodzik, 2005; Lawrence et al., 2007; Seifert and Mlodzik, 2007; Wang and Nathans, 2007; Wu and Mlodzik, 2009; Zallen, 2007). PCP is the asymmetrical organization of cells within the plane of an epithelium. PCP of individual cells can propagate signal(s) to neighboring cells through intercellular communication and as a result generates a coordinated polarizing behavior across tissues.

PCP is best characterized in Drosophila as it is visible in all adult cuticular structures, incl. wings, thorax, or abdomen and compound eyes (Adler, 2002; Goodrich and Strutt, 2011; Lawrence et al., 2007; McNeill, 2010; Seifert and Mlodzik, 2007). In the wing, PCP controls the positioning of single actin-based hairs at distal verteces of each cell (Wong and Adler, 1993; Seifert and Mlodzik, 2007; Goodrich and Strutt, 2011). In the thorax, PCP ensures that sensory bristles and cellular hairs point posteriorly (Bellaiche et al., 2004; Krasnow and Adler, 1994). In the eye, PCP establishment regulates cell fate determination of two photoreceptor cells (Mlodzik, 1999; Strutt and Strutt, 1999): specification of R3-R4 establishes polarity of preclusters that will form adult ommatidia and also directs the ommitidial rotation to align them to form a mirror-image symmetry across the dorso-ventral midline (the equator; e.g. Goodrich and Strutt, 2011; Seifert and Mlodzik, 2007).

Two sets of conserved PCP factors have been identified via genetic analyses (rev. in Goodrich and Strutt, 2011; Lawrence et al., 2007; McNeill, 2010; Seifert and Mlodzik, 2007; Wu and Mlodzik, 2009; Zallen, 2007): (1) The Frizzled (Fz)/PCP group, which includes the transmembrane proteins Fz, Van Gogh (Vang, also known as Strabismus/Stbm), Flamingo (Fmi, a.k.a. Starry Night/Stan) and the cytosolic proteins Dishevelled (Dsh), Diego (Dgo) and Prickle (Pk); and (2) the Fat (Ft)/Dachsous (Ds) group comprising Ft, Ds, Four-jointed (Fj), Dachs (D) and Approximated (App). Proteins in the Ft/Ds signaling module regulate PCP through a set of distinct mechanisms and act independently of the Fz/PCP group (Casal et al., 2006; Lawrence et al., 2007; Matakatsu and Blair, 2008).

Asymmetric cellular distribution of Fz/PCP proteins is a discernable hallmark of PCP-type polarity establishment. In Drosophila this is evident in developing eyes, wings, and the notum/dorsal thorax (Axelrod, 2001; Bastock et al., 2003; Das et al., 2004; Gho et al., 1999; Strutt and Strutt, 2009; rev. in Seifert and Mlodzik, 2007; Strutt and Strutt, 2009). Establishment of cellular asymmetry is tightly linked to the propagation of PCP information to at least neighboring cells, if not across whole tissues. For example in wings, at approx. 30hr after puparium formation (APF), two complexes are detected on opposite sides of each cell along the proximal-distal (PD) axis: Fz, together with Dsh and Dgo, localizes to the distal edge, whereas Vang/Stbm and Pk localize to the proximal side. The 7-TM cadherin Fmi colocalizes with both the distal and proximal protein complexes, and forms homophilic interactions intercellularly to stabilize these complexes by interacting with Fz and Vang/Stbm individually in cis (Amonlirdviman et al., 2005; Chen et al., 2008; Devenport and Fuchs, 2008; Klein and Mlodzik, 2005; Lawrence et al., 2004; Le Garrec et al., 2006; Strutt, 2008; Strutt and Strutt, 2007). Genetic, clonal, biochemical, and cell culture analyses also suggested that Fz and Vang/Stbm physically interact in trans (Lawrence et al., 2004; Strutt and Strutt, 2008; Taylor et al., 1998; Wu and Mlodzik, 2008). It is believed that the formation of such Fz:Fmi/Vang:Fmi complexes allows cells to propagate/maintain polarization between cells (Chen et al., 2008; Klein and Mlodzik, 2005; Lawrence et al., 2004; Le Garrec et al., 2006 Strutt, 2008 #202). Importantly, the interactions among the core PCP factors and their asymmetric localization are conserved in vertebrate contexts of PCP establishment (Goodrich and Strutt, 2011; Wang and Nathans, 2007).

It has recently been shown that initial polarization of Fz/PCP complexes in Drosophila wings is evident at 5-6 hr APF and even already in late third instar discs, significantly earlier than the previously assumed 28-30 hr APF stage (Aigouy et al., 2010; Classen et al., 2005; Sagner et al., 2012; Strutt et al., 2011). At these early stages, the asymmetric protein complexes are oriented radially toward the wing margin. Subsequently, wing cells realign and remodel their PCP along the PD axis, through mechanisms of oriented cell division, cell elongation, cell rotation, and neighbor exchanges (Aigouy et al., 2010). How these rearrangement events are controlled and how they correlate with the establishment of late stage polarization is mechanistically undefined. It is likely that additional cell adhesion features and molecules are involved in these processes.

Here, we have identified Drosophila Furrowed (Fw) as a regulator of PCP establishment. Fw is the Drosophila homolog of the vertebrate Selectin family (Leshko-Lindsay and Corces, 1997). Vertebrate Selectins are single pass transmembrane proteins that mediate the “rolling” and initial attachment of leukocytes to endothelial cells through carbohydrate-protein interaction (Butcher, 1991; Kansas, 1996; Zarbock et al., 2011). Fw is most closely related to mammalian P-selectin, and previous analyses of fw mutants have identified defects in bristle morphogenesis (shorter bristles and hairs) and eye structure/mild eye disc overgrowth (Leshko-Lindsay and Corces, 1997).

We show that loss of fw causes characteristic PCP defects. Genetic analyses suggest a functional link between Fw and the Fz/PCP core proteins. We demonstrate that Fw can act as a cell adhesion molecule through homophilic interaction, independently of the carbohydrate recognition domain required for Selectin-adhesion in mammalian contexts. Clonal analyses and cell culture studies reveal that Fw promotes Fz stability. Fw is localized within the adherens junction region, and assays in S2R+ cells indicate that Fw forms a complex with Fz, which then can mediate the trans interaction between Fz and Vang/Stbm. Loss-of-function studies further suggest that Fw functions partially redundantly with Fmi in PCP. Our data suggest that Fw acts as a cell adhesion molecule, participating in PCP by mediating the interaction between Fz and Vang/Stbm during early pupal wing and thorax development, by affecting Fz stability and/or participating in the Fmi-mediated Fz-Vang interactions

Results

Identification of Furrowed as a PCP regulator

To identify Drosophila PCP regulators we conducted an EMS-based eyFLP/FRT screen on the X-chromosome, using ommatidial orientation in the eye as read-out. Flies carrying a rhodopsin1(rh1)-GFP transgene (expressed in outer photoreceptors R1-R6) were utilized to visualize the photoreceptor arrangement (Pichaud and Desplan, 2001). One screen hit, mutant “42-m9”, was isolated as a viable mutation producing PCP-type ommatidial defects, detected by abnormal rh1-GFP patterning (Supp. Fig. S1A-B). Adult eye sections of hemizygous 42-m9 mutant males, or mutant clones in females, confirmed PCP defects, including altered ommatidial rotation and occasional symmetrical clusters, with also mild R-cell loss (Fig. 1A,C and Fig. S1). Examination of the notum in 42-m9 hemizygous mutant males revealed strong PCP defects with misoriented bristles and cellular hairs (Fig. 1E,G), also bristles were often thicker and shorter in 42-m9 mutants (Fig. 1G). In wings, cellular hairs were misoriented in several regions, consistently in the area posterior to vein L5 (Fig. 1I,K-L). Besides the PCP defects, the patterning and morphology of 42-m9 mutant wings was otherwise comparable to wild type. Taken together, the phenotypic defects suggest that the 42-m9 mutation affects PCP establishment in general.

Figure 1.

fw is required for PCP establishment.

(A-D) Tangential eye sections of indicated genotypes, showing region flanking the equator (D/V-midline; anterior is left and dorsal up), bottom panels show schematics of ommatidial orientations. Black and red arrows: dorsal and ventral chirality; green arrows: symmetrical clusters; ommatidia with a loss of R-cells (precluding PCP scoring) are shown by black dots.

(A) Wild-type (w¹¹⁸), note mirror-image symmetry across equator.

(B) Knock-down of fw (via sevGal4, UAS-fw^IR @29°C): note occasional symmetrical ommatidia, rotation defects, and R-cell loss. (C) Hemizygous “42m9” (42/Y) males exhibit similar defects (cf. to B), as well as “42” mutant clones (see Fig. S1). (D) Defects in mutant “42m9” are fully rescued by Fw expression under tubulin promoter control (tub-FwEGFP), indicating that “42m9” is an allele of fw.

(E-J) Dorsal view of thorax (notum) of indicated genotypes. Anterior is up. (I-J) are SEM electron-micrograph high magnification views.

(E, I) Control notum (pnrGal4/+ at 25°C is phenotypically wild-type) with sensory bristles, and cuticular cells (I) oriented towards the posterior.

(F, J) pnrGal4>UAS-fw-IR: many bristles and cuticle cells (J) are misoriented in central notum (where pnrGal4 drives expression), and often also display “multiple cellular hairs” phenotype. Of note, some bristles are shorter and thicker, possibly reflecting cytoskeletal organization requirements.

(G) Notum of “42” hemizygous male exhibits similar, albeit stronger, defects as in (F).

(H) tub-FwEGFP fully rescues the defects observed in “42” mutant males, confirming that “42” is fw allele.

(L-N) High magnification of adult wing, posterior to vein L5 (see Fig. S1 for position of micrographs); proximal is to the left and anterior up.

(K) Wild-type (w¹¹¹⁸) wing: all hairs point distally.

(L) enGal4, UAS-fwIR wing with misorientated cellular hairs.

(M) Mutant “42m9” displays misoriented cellular hairs (with 90% penetrance, n=20). With respect to shape and size the wings appeared wild type.

(N) tub-FwEGFP transgene rescue of the “42”/Y wing phenotype (all wings looked wild-type; 100% rescue, n>20).

Scale bars: 5μm for A-D, 100μm for E-H, 10μm for I-N. See also Fig. S1.

Genetic mapping localized 42-m9 to the 10D6-11A1 region (determined via the DrosDel deficiency kit, as it failed to complement Df(1)ED7147), containing 40 genes including furrowed (fw). fw¹ failed to complement 42-m9, producing very similar “furrowed” defects in 42-m9/fw¹ females as compared to homozygous fw¹ females (not shown; Leshko-Lindsay and Corces, 1997). Consistently, using a transgenic RNAi approach to knockdown Fw (under sevenless[sev]-Gal4 control) during eye development caused very similar defects as those observed in 42-m9 eye clones (Fig. 1B-C). Eye PCP defects were also reproduced using fw-RNAi under eyeless-GAL4 and hairy-GAL4 control (not shown). Similarly, in the thorax and wing, reduction of Fw under pannier-GAL4 control (thorax, Fig. 1F,J, cf. 1I) or en-Gal4 (wing, Fig. 1L) led to orientation defects of bristles and cellular hairs, and multiple cellular hairs, very similar to defects observed in the 42-m9 mutant.

To confirm that fw is indeed mutated in 42-m9, we tested a transgene for rescue of the 42-m9 mutant: tub-FwEGFP (Fw expressed from the tubulin promoter) rescued all phenotypic defects associated with 42-m9 in hemizygous males (Fig. 1D,H and N, and not shown), demonstrating that 42-m9 is an allele of fw (henceforth referred to as fw⁴²). Sequencing of the fw⁴² chromosome revealed a mutation at Trp689 (TGG to TGA [stop codon] within the extracellular region of Fw), which is predicted to generate a truncated Fw protein (Suppl. Fig. S1). As fw⁴²/Df(1)ED7147 trans-heterozygotes displayed similar strength defects as fw⁴² homozygous females (not shown), we conclude that it is a strong (if not null) loss-of-function allele. Taken together, our data suggest that fw regulates PCP establishment.

fw encodes a single pass transmembrane protein localized near epithelial junctional complexes

fw encodes a single pass transmembrane protein of 1174 residues, with several conserved domains (based on the SMART program [Letunic et al., 2012; Schultz et al., 1998]), including the extracellular FTP (eel-Fucolectin Tachylectin-4 Pentaxrin-1 domain), CTL (C-type lectin/carbohydrate recognition domain), 11 CCP (complement control protein) domains (also known as SCR-short consensus repeats or SHUSHI repeats) and a short cytoplasmic tail (Fig. S1). The N-terminal CTL domains show strong homology to vertebrate selectins, with highest homology to P-selectin (Leshko-Lindsay and Corces, 1997). The vertebrate selectin family consists of L-, E- and P-selectin, each having characteristic extracellular regions composed of the Lectin domain, an EGF motif, and two to nine CCP repeats (Fig. S1; Butcher, 1991; Kansas, 1996).

Consistent with a proposed role in PCP establishment, Fw is expressed in imaginal discs during PCP establishment (Leshko-Lindsay and Corces, 1997; and not shown). We examined the subcellular distribution of Fw through the tub-FwEGFP transgene (FwEGFP expressed under tubulin promotor control), which fully rescued the mutant (Fig. 1D,H and L), and thus is both functional and expressed at functional levels. Fw is localized to the plasma membrane and mainly detected near/at subapical junctional complexes, partially overlapping with DE-cadherin (Fig. 2A-B,D). In the eye, PCP is established in third instar discs posterior to the morphogenetic furrow; within this area FwEGFP is enriched at membranes of R-cell precursors in developing ommatidial preclusters, colocalizing with DE-cadherin and the core PCP factor Fmi (Fig. 2A-A” and not shown,). Localization at membranes in R-cell precursors is similar to core PCP factors (Fz, Vang/Stbm, and Fmi; e.g. Das et al., 2002; Djiane et al., 2005). In developing wings, Fw is localized to subapical junctional membrane regions (again overlapping with DE-Cadherin and Fmi) during PCP establishment (examples of prepupae to late pupal stages shown in Fig. 2C,E-F). Whereas Fmi (and other core PCP factors) become asymmetrically enriched within the PCP axis once it becomes detectable (Axelrod, 2001; Das et al., 2002; Strutt et al., 2002; Strutt and Strutt, 2002, 2008; Usui et al., 1999), FwEGFP membrane localization remains more uniform within the subapical junctional region (Fig. 2; see Fig. legend for specific comments re. Fw localization studies). Fw localization in the developing thorax was very similar to wing regions (Fig. 2B-B”’).

Figure 2.

Fw localizes to subapical junctional regions.

Panels show confocal microscopy images of FwEGFP (green; expressed from tub-FwEGFP, which rescues the fw mutant), DE-cadherin (DE-cad) or phalloidin staining (labeling actin) in red, and anti-Fmi (blue).

(A-A”) 3rd instar eye disc posterior to morphogenetic furrow (MF), which is at the left. PCP is established in preclusters in rows 2-6 posterior to MF. FwEGFP is enriched at membranes of R8, R2/5 and R3/4 in preclusters (A-A’) and later accumulates generally at all R-cell membranes, partially colocalizing with DE-cad (A, A”).

(B-B”’) FwEGFP localization in 28hr APF thorax tissue. Anterior is up. FwEGFP is localized to cell boundaries, overlapping with DE-cad at subapical junctional complexes (DE-cad outlines junctional membranes; B, B”). Fw and Fmi also overlap, but FwEGFP is detected more uniformly around cells (B’), whereas Fmi is enriched asymmetrically along PCP axis (antero-posterior, arrow in B”’; arrowheads outline a few examples),

(C-C”’) 30hr APF pupal wings (proximal left and anterior up), phalloidin (red) marks growing actin hairs (C,C”). Fmi is polarized along PCP (P/D) axis (arrow in C”’; examples marked by arrowheads), FwEGFP is localized more evenly along cell boundaries (C’).

(D-D”) x/z-optical section of pupal wing from panel C with both wing layers (dorsal at top, ventral at bottom; A; apical; B: basal). FwEGFP is enriched at subapical junctional levels, overlapping with Fmi.

(E-F) FwEGFP localization during pupal wing patterning at 5hr APF (E) and 22hr APF (F; images at same magnification). FwEGFP is localized to membranes at all stages. Note that FwEGFP protein is expressed from exogenous promoter (tubulin) and at mildly higher level than endogenous protein; thus, although it rescued mutant, could be partially mislocalized due to protein levels. Scale bars: 10μm. See also Fig. S2.

fw suppresses Fz and enhances Vang/Stbm gain-of-function activities

The PCP defects in fw mutants resembled those of Fz/PCP core mutants, suggesting that fw could be involved within this molecular PCP cassette. Consistent with this hypothesis, fw interacted genetically with fz, suppressing the sev-Fz gain-of-function (GOF) eye phenotype (Fig. 3A-D): Fz is required for R3 fate determination and, accordingly, overexpression of Fz in R3 and R4 (under sev-enhancer control) frequently causes both cells to adopt the R3 fate (Fanto and Mlodzik, 1999; Tomlinson and Struhl, 1999). This effect was suppressed by removing one copy of fw (Fig. 3A,B,D), suggesting that Fw promotes Fz function. Removing fw completely (via hemizygous males) not only suppressed the sev-Fz PCP effect, but it shifted the effect towards canonical Wnt signaling associated defects, mainly R-cell loss (Fig. 3C). These data suggest that Fw helps to “focus” Fz activity to the PCP branch among the Wnt pathways. In contrast, removing a copy of fw mildly enhanced the sev-Vang/Stbm GOF defects (in which ommatidia often develop as R4/R4-type symmetrical clusters; Suppl. Fig. S3). Loss of fw did not dominantly affect GOF defects of other Fz-group core PCP genes, nor did it affect Fat and Ds defects in the equivalent assay (not shown).

Figure 3.

fw dominantly suppresses Fz gain-of-function defects.

(A-C) Eye sections near equator (see Fig. 1A for wt; arrows as in Fig. 1, anterior is left). sev-Fz induces symmetrical clusters (A), this is suppressed by fw−/+ (B; note reduction in green arrows); see also suppression by removing one gene copy using a deficiency (quantif. in D). This is further suppressed in fw/Y hemizygous males (C,D). sev-Fz in fw⁴²/Y background also showed GOF canonical Wg-signaling defects (seen by loss of R-cells, circles in schematic), suggesting that Fw can “channel” Fz activity towards PCP.

(D) Quantif. of genotypes in A-C: % of symm. ommatidia are shown (*p<0.01, **p<0.001 with student’s t-test; n=400-662 from 3-4 independent eyes; error bars are S.D.).

(E-H) Adult wings (proximal is left and anterior up): dorsal wing surfaces are shown. (E) Fz overexpression under dppGal4-control (dpp>Fz) causes non-autonomous hair reorientation outside dpp-expression, e.g. in boxed areas (orange) in E (GFP was coexpressed as a control) and (F). (F) dpp>UAS-Fz, UAS-fw-IR (RNAi knock-down of fw in cells where Fz is overexpressed) suppresses the Fz effect. Rosettes represent angle distribution of hair orientation in equivalent areas to orange boxes in E and F (p-value [with Kolmogorov-Smirnov calc’s, designed for changes in patterns] is **p=10⁻¹¹, n>100); dppGal4, UAS-fw-IR alone does not cause misorientation under these conditions. For ease of observation the second most frequent orientation angle sector was color coded in dark blue (note change of dark blue sector from around 60° in dpp>Fz to 0-15° in dpp>Fz, fw-IR). n>30 wings, suppression of the dpp>Fz effect in every wing coexpressing Fw-RNAi.

(G-H) Vang/Stbm overexpression (via nubGal4) caused mild misorientation defects (G), which were enhanced by fw/+ (H), quantitations as bars below respective genotypes (n=27 and 26, resp.). (G) represents “weak” phenotype (strongest seen in nubGal4>Vang), (H) represents “strong” phenotype (almost all fw/+; nubGal4>Vang wings). The interaction/enhancement was confirmed with a Df for the fw locus (not shown). Scale bars: 5μm A-C, 25μm E-H. See also Fig. S3.

In the wing, Fz overexpression causes non-autonomous effects, reorienting surrounding wild-type cells away from the Fz expression domain (Adler, 2002; Casal et al., 2006; Vinson and Adler, 1987; Wu and Mlodzik, 2008). For example, when expressed in a stripe along the antero-posterior compartment boundary (dpp domain), Fz reorients hairs of neighboring cells perpendicularly to the P-D axis, away from the dpp>UAS-Fz stripe. Knocking down fw within this domain while simultaneously overexpressing Fz, suppressed the Fz non-autonomous effect (Fig. 3E-F). As fw modified both autonomous and non-autonomous Fz effects, it may affect either Fz levels, activity, or membrane localization. Accordingly, co-overexpression of Fz and Fw within the dpp-stripe at temperatures where Fz alone causes only weak non-autonomous effects (18°C; Gal4 is temperature sensitive) enhances the Fz GOF effect (not shown). Consistent with the notion that Fw acts via Fz, defects of fz^P21 (null allele) are not enhanced in fw, fz double mutants, suggesting that fw function depends on fz activity. Moreover, fw LOF alleles dominantly enhanced a Vang/Stbm GOF wing phenotype: overexpression of Vang/Stbm under nubbin-GAL4 (throughout wing discs) generated patches of misoriented hairs, which was enhanced in fw/+ heterozygous backgrounds (Fig. 3G-H). These data suggest that Fw promotes Fz activity, possibly by affecting Fz levels or localization (see below).

Fw interacts molecularly with Fz and regulates Fz stability

Cellular polarization of core PCP factors in developing wings is seen as early as in late 3^rd instar and 5 hr APF stages (Aigouy et al., 2010; Classen et al., 2005; Sagner et al., 2012; Strutt et al., 2011). During early pupal stages, polarization of core PCP factors is detected in a radial axis (towards wing margin); subsequently membrane associated levels of core PCP factors are reduced (during cellular movements caused by the wing hinge contraction, between 18-26 hr APF) and robust core PCP asymmetries “reemerge” in the proximo-distal (P/D) axis at ~28 hrs APF (Aigouy et al., 2010). To explore the relationship between Fw and Fz, we induced fw LOF clones and analyzed Fz by immunostaining. We detected a reduction of Fz in fw⁴² mutant clones, most prominently at ~22 hr APF (Fig. 4A-B”), the stage when cellular contacts and shapes are rearranged (Aigouy et al., 2010). In later stage pupal wings (~32hr APF, when trichomes emerge), the effect on Fz was weaker, but asymmetric localization along the P/D axis was less focused in fw⁴² clones as compared to neighboring wild-type tissue (Fig. 4C-C”’ and 4D”; defects in PCP orientation as seen via Fz polarization were also affected; Fig. 4D-D’). The effect on Fz levels was specific, as other core PCP factors, Fmi or Vang for example, were not affected in fw LOF clones at any stage analyzed (Fig. 4E-F”’ and Fig. S4). In contrast to wing discs, we did not observe reduction of Fz staining in fw⁴² eye clones (Suppl. Fig. S4), which could be due to differences in PCP generation between tissues (see Discussion).

Figure 4.

Fw is required for Fz stability.

(A-E”’) Confocal images of immunostained pupal wings carrying fw⁴² clones at 22hr (A-B”’ and E-F”’) and 32hr APF (C-D”), stained with antibodies as indicated, Fz or Fmi in blue (monochrome in ’ and ” panels; fw mutant tissue marked by absence of β-Gal, green). Proximal is left, all images are at same magnification.

(A-A”’) 22hr APF, Fz membrane staining (blue, monochrome in A’) is reduced in fw⁴² clones. Phalloidin staining (red, monochrome in A”) labels F-actin and remains largely unchanged. (A”’) Quantified fluorescent levels; yellow line in A marking position of scan, and clone border in A”’.

(B-B”’) x/z-optical section of 22hr APF wing (equiv. to panel A). Fz levels are affected in fw⁴² mutant clones (staining as in A-A”’),

(C-C”’,D-D”) At 32hr APF, Fz membrane localization in fw⁴² clones is less regular than in surrounding wild-type tissue (only apical staining is shown). Phalloidin (red, monochr. in C”) marks prehairs. D-D”: PCP orientation angles in fw mutant cells are less regular than in wt (actual angles shown by yellow lines in each cell in D, summarized and analyzed in D’; n=226 for wt and 216 for fw⁴² (**p<10⁻⁵ by Kolmogorov-Smirnov). (D”) Quantif. of Fz levels (fluorescence intensity) across cell membranes in PCP axis in wt vs. fw⁴² cells, note reduced Fz levels in mutant (**p<0.0001).

(E-E”’) 22hr APF wing, Fmi in blue (monochr. in E’). Fmi localization is not affected in fw⁴² at any stage tested. (E”’) Quantif. fluorescent levels for Fmi and DE-cad (yellow line marks position of scan in E and clone border in E”’).

(F-F”’) x/z-optical microscopy section of 22hr APF wings (equiv. to E). Fmi and E-cad levels are not affected in fw⁴² clones (staining as in E-E”). Scale bars: 20μm. See also Fig. S4.

Together with the genetic interactions, the data suggest that Fw can regulate Fz stability or membrane association. Consistent with this notion, we detected increased Fz levels, when Fw was coexpressed with Fz in S2R+ cells (Fig. 5A) and in vivo in wing discs (Fig. 5D). This effect was dosage sensitive: increased levels of Fw led to increase in Fz levels (Fig. 5A) and specific, as other cell adhesion molecules, e.g. DE-cad, did not affect Fz levels (Fig. 5A). Also, the effect of Fw on Fz was specific to Fz and not observed with other PCP proteins, for example Vang and Fmi (Fig. 5C and not shown). These data are consistent with the in vivo LOF studies and support the notion that Fw functions to stabilize Fz at the membrane.

Figure 5.

Fw stabilizes and forms a complex with Fz.

(A-B) Western blot analyses of Fz (Fz-myc) levels from S2R⁺ cell lysates coexpressing FwEGFP, EGFP-CAAX, or DE-cadherin as controls (A). Increasing doses of FwEGFP and controls were cotransfected with Fz-myc and lysates immunoblotted. Note increasing doses of FwEGFP led to increase in Fz levels (quantif. below gel bands). (B) FwΔCCP2-EGFP (which fails in cell adhesion, Fig. 6), causes a comparable Fz stabilization to full-length FwEGFP (quantif. in bottom panel; experiments performed in triplicate; Fz levels were normalized to tubulin). Values from each lane were calculated using Image J package; *p<0.05, **p<0.001, student t-test.

(C) Vang levels are not affected by Fw cotransfection; experiment as described for Fz; changes are not significant (p with student t-test). Same results obtained for Fw and Fmi cotransfections with unchanged Fmi levels (not shown).

(D) Fw overexpression causes comparable increase of Fz levels in vivo (similar to S2 cells). Wing disc extracts from nubGal4/+ and nubGal4/UAS-Fw were analyzed for Fz (top blot), Tubulin (bottom blot), and DE-cad (not shown). Fz levels were normalized to Tubulin (experiment done in triplicate). Note Fz level increase in nubGal4/UAS-Fw (*p<0.001; student t-test). All error bars in A-D are S.D.

(E-G) Fw can co-immunoprecipitate (co-IP) Fz. (E) S2R⁺ cell lysates expressing Fz-myc, alone or cotransfected with either EGFP-CAAX or FwEGFP were IPed with anti-GFP and immunoprecipitates blotted with anti-myc to assay for Fz. Note specific co-IP of Fz (myc) with FwEGFP. Also note increase of Fz levels in bottom IB myc blot (cf. to A). WCE: whole cell lysate input.

(F) Cotransfection of FwEGFP with Fz-myc or DE-cad (control). Lysates were IPed with anti-GFP and blotted with anti-myc or anti-DE-cad. Note co-IP of Fz with FwEGFP.

(G) FwΔCCP2 can also co-IP Fz. Fz-myc was cotransfected with full length FwEGFP or FwΔCCP2-EGFP, IP’ed with anti-GFP and blotted with anti-myc. Note co-IP of Fz with FwΔCCP2-EGFP (albeit slightly weaker than with FwEGFP, positive control); EGFP-CAAX serves as negative control. FwΔCCP2-EGFP is less effective at co-IP of Fz possibly due to the fact that FwΔCCP2-EGFP is present at lower levels in plasma membrane (see Fig. 6).

We next examined whether Fw formed a complex with Fz using co-immunoprecipitation assays in S2R+ cells. From cell lysates coexpressing FwEGFP and Fz-myc, we detected Fz in the FwEGFP precipitate but not in the precipitate of EGFP-CAAX, a membrane tethered EGFP (Fig. 5E), suggesting that Fw and Fz are part of a complex. To confirm the specificity of the Fw-Fz interaction, we extended the assay to include DE-cadherin. DE-cadherin was not detected in FwEGFP precipitates, supporting that the Fw-Fz association is specific (Fig. 5F). Together with the protein stabilization and genetic data, we conclude that Fw affects PCP signaling through maintaining Fz levels by association with the Fz membrane complex(es).

Fw can function as a homophilic cell adhesion molecule

Vertebrate Selectins bind carbohydrate ligands to mediate the interaction between blood cells and endothelia (Kansas, 1996; Lasky, 1992), but Selectins have not been reported to bind other Selectins in trans through homophilic cell adhesion. Fw expression in Drosophila S2R+ cells resulted in their aggregation (Fig. 6A and S5), suggesting that Fw is capable of mediating homophilic binding/adhesion. Consistently, FwEGFP localized to cell contacts when both cells express Fw (Fig. 6A) and, likewise, FwEGFP expressing cells recruit Fw-HA (from cells expressing Fw-HA) to cell junction/contact areas (Fig. S5). Although Fw mediates cell adhesion in S2 cells, this is significantly weaker as compared to cadherins, e.g. DE-cad or Fmi (see quantif. of Fw mediated cell adhesion, Fig. S5).

Figure 6.

Fw mediates homophilic cell adhesion and intercellular Fz-Vang interactions in transfected cells.

All images are representative examples of transfected S2 cell populations as indicated.

(A) S2R⁺ cells transfected with full length FwEGFP: Note formation of cell adhesion bridges (enrichment at cell contact sites – arrowhead) in FwEGFP (green) transfected cells. Fixed cells were permeablized and counterstained with DAPI (blue) to label nuclei.

(B) S2R⁺ cells transfected with FwΔCCP2-EGFP. Note absence of cell adhesion bridges even when cells occasionally are in contact. See Fig. S5 for quantifications.

(C) Schematic of deletion constructs, indicating that CCP2 is essential for adhesion and PCP establishment, whereas carbohydrate binding C-type lectin is not required for these aspects or Fw function.

(D-E”’) Fw can mediate Fz-Vang/Stbm intercellular interaction. (D-D”’) Fz-myc and FwEGFP were cotransfected into one pool of S2R⁺ cells, and Vang-flag and FwEGFP were transfected into a second pool. After transfection, cells from both pools were washed and mixed, followed by immunostaining with anti-flag (Vang, red; also D’) and anti-myc (Fz, blue). Vang is stabilized at cell contacts with Fz expressing cells. FwEGFP (green, D”) is detected at membranes of all cell contacts. (D”’) For clarity, cells are outlined with white lines and labeled according to transfected DNA (see Fig. S5 for quantifs.). (E-E”’) As control, S2R⁺ cells transfected with Vang-flag (red, E’) and FwEGFP (green, E”), and mixed with cells transfected only with FwEGFP, and counterstained with DAPI. (E”’) For clarity, cells outlined with white lines and labeled according to transfected DNA (untransfected cells in mix outlined with dotted lines). Vang does not accumulate at membranes and is present in vesicular punctae, as seen previously (Strutt and Strutt, 2008), indicating that cell-cell contact is not sufficient for Vang membrane localization (Chen et al., 2008; Lawrence et al., 2004; Strutt and Strutt, 2008).

(F-F”’) FwΔCCP2, lacking cell adhesion potential, fails in the presence of Fz (blue) to recruit Vang (red, F’) to membranes of neighboring cells, indicating that Fw mediated cell-adhesion and Fz are both required for Vang membrane recruitment. (F”’) For clarity: cells outlined with white lines, labeled with transfected DNA. Scale bar: 5μm. See also Fig. S5.

To define which domain(s) of Fw mediate the homophilic interaction, we generated a series of truncations in its extracellular region. Cell adhesion studies revealed that Fw requires the CCP2 motif for adhesion (Fig. 6B-C). None of the other CCP repeats, nor the C-type lectin or FTP motifs (FwΔN) affected cell adhesion significantly (sum. in Fig. 6C). We next asked whether these requirements correlated with Fw’s in vivo function in PCP establishment and tested for phenotypic rescue of fw⁴² via transgenic flies expressing tub-FwΔCCP2 and tub-FwΔN. Unlike tub-FwEGFP (full length; Fig. 1), the tub-FwΔCCP2-EGFP transgene did not rescue fw⁴² (Suppl. Fig. S5), suggesting that the defects observed in fw mutant flies are caused, at least in part, by loss/reduction of cell adhesion. Western blotting confirmed that both full-length and FwΔCCP2 proteins were expressed in fw mutant flies. Although FwΔCCP2 was present with slightly lower abundance (compared to full-length Fw), two copies of tub-FwΔCCP2 also failed to rescue the mutant, arguing against protein level effects. In contrast, tub-FwΔN rescued all PCP defects of fw- (indistinguishably from tub-fwEGFP full length, Fig. 1) indicating that “sugar interactions” are not required for Fw’s PCP function (Suppl. Fig. S5; tub-FwΔN did not rescue the bristle shape defects and “furrowed” eye phenotype). Taken together, these data suggest that the CCP2 domain (and hence adhesion) is essential for in vivo PCP function of Fw, and that its carbohydrate binding C-type lectin is dispensable for PCP. Consistent with this conclusion, FwΔCCP2 is reduced at the membrane in vivo and cell culture assays (Fig. 6B and not shown).

In addition to its homophilic adhesion function, we tested whether the Fw CCP2 domain was required for the physical association of Fw with Fz. Although FwΔCCP2 could pull down Fz (Fig. 5G), it did so less effectively, suggesting that while the CCP2 motif is not essential for Fw-Fz interactions, it is augmented by its adhesive function/membrane localization. Consistent with these data, we observed increased Fz stabilization when FwΔCCP2 was co-transfected with Fz (Fig. 5B), suggesting that the cell adhesion function of Fw is at least in part separable from its effects on Fz association and stability.

fw is partially redundant with fmi and facilitates Fz-mediated Vang/Stbm membrane recruitment

For additional mechanistic insight into Fw function, we performed GOF studies. First, sev-enhancer driven expression of Fw, sev-FwHA (to mimic the sev-Fz overexpression in R3/R4 precursors during Fz-mediated R3 induction), induced PCP defects similar to sev-Fz, causing many clusters to adopt R3/R3-type symmetric arrangements (Fig. 7A, cf. to Fig. 3A; Fanto and Mlodzik, 1999; Tomlinson and Struhl, 1999). Consistent with previous data, this effect was dependent on fz levels as removal of one fz copy suppressed sev-Fw (Fig. 7B, quantif. in 7E). Whereas, dosage reduction in Vang/stbm did not have an effect (Fig. 7C,E), a strong dominant suppression of sev-Fw was observed by reducing fmi dosage (Fig. 7D).

Figure 7.

fw acts in a partially redundant manner with fmi.

(A-D) Eye sections around equator, genotypes as indicated. Dorsal is up and anterior left (schematic presentation in bottom panels as in Fig. 1).

(A) sev-Fw overexpression (2 copies) with classical PCP defects (sim. to sev-Fz) incl. symmetrical clusters (green arrows), chirality defects, and infrequent R-cell loss (black dots). (B) sev-Fw (2 copies); fz^P21/+: symmetrical cluster frequency is suppressed.

(C) sev-Fw (2 copies); Vang^stbm6/+: no significant effect as compared to sev-Fw.

(D) sev-Fw (2 copies); fmi^E59/+: strong suppression of sev-Fw PCP phenotype.

(E) Quantif. of (A-D) genotypes (n=850-1512 from 5-9 independent eyes; **p<0.001).

(F-N) fw acts partially redundantly with fmi. (F) Quantif. of eye defects of fmi^frz3 allele in wt or fw mutant backgrounds. Green bars: % of symmetrical ommatidia. Note enhancement of fmi^frz3 by reduction of fw (fw⁴²/+; fmi^frz3) or in fw hemizygous males (fw⁴²/Y; fmi^frz3); n=392-610, 3-4 indep. eyes (*p<0.05, **p<0.001 with student t-test). Error bars in E-F are S.D.

(G-I) Enhancement of fmi^frz3 wing phenotype in fw hemizygous males (cf. G-H). (I) Quantif. of wing interactions; note that the strong phenotypes (eye and wing) resemble but are not stronger than fmi null phenotype. See also Fig. S7.

(J-N) Enhancement of thorax PCP defects of fmi^frz3 (J) by fw dosage (L), and fw hemizygosity

(M); PCP distribution quantif. shown in lower panels (genotypes as indicated; N is wt control for comparison). fw⁴²/Y serves as control (K). fw dominantly enhances PCP defects of fmi^frz3 (cf. panels G and I) with strongest PCP defects present in fw, fmi^frz3 double homozygous mutants (*p<0.005 and ** p<10⁻⁴ with Kolmogorov-Smirnov, all n>100). Scale bars: 5μm A-D, 25μm G-H, 10μm I-N.

(O) Schematic model of proposed Fw function, forming homophilic cell adhesion intercellular bridges mediating a Fz-Vang intercellular interaction, similar to Fmi.

Fmi has been proposed to facilitate Fz-Vang/Stbm intercellular interactions via Fmi/Fz–Fmi/Vang intercellular bridges (Amonlirdviman et al., 2005; Chen et al., 2008; Klein and Mlodzik, 2005; Lawrence et al., 2004; Le Garrec et al., 2006; Strutt and Strutt, 2008). Fw, like Fmi, can promote homophilic adhesion (Fig. 6), and they both can immunoprecipitate Fz (Fig. 5A-C)(Chen et al., 2008). We thus hypothesized that fw might function (partially) redundantly with fmi. In support of this, fw dominantly enhanced a fmi LOF allele (fmi^frz3; Rawls and Wolff, 2003) in all tissues: eyes (Fig. 7F), wings (Fig. 7G-I; in addition to PCP defects the double mutant wings were less flat), and thorax (Fig. 7J-L). Moreover, a double homozygous fw⁴²; fmi^frz3 mutants displayed stronger PCP defects in the thorax than a fmi null allele (Fig. 7M and not shown), but not stronger than a fz null, as expected if both act through Fz. A synergistic effect was also observed in fw⁴²; fmi^frz3 double mutant eye scenarios (Fig. 7F). These data suggest that fw and fmi might function in parallel within the Fz-core PCP system (see below).

Combined with the genetic data above, we hypothesized that homophilic Fw cell adhesion might mediate the intercellular Fz-Vang/Stbm interaction (Strutt and Strutt, 2008; Wu and Mlodzik, 2008), as proposed for Fmi (Amonlirdviman et al., 2005; Chen et al., 2008; Klein and Mlodzik, 2005; Lawrence et al., 2004; Le Garrec et al., 2006; Strutt and Strutt, 2008). To test this, we transfected two individual pools of cells with Fw, one pool co-expressing Fz-myc and the other pool co-expressing Vang/Stbm-Flag, and asked whether Fw could facilitate Fz-mediated Vang/Stbm membrane recruitment, analogous to what has been proposed for Fmi (Strutt and Strutt, 2008). In cells expressing only Vang or in cells co-expressing Vang-Fw, most Vang protein is detected in cellular vesicular structures (Fig. 6E; Strutt and Strutt, 2008). In contrast, in the presence of Fw-Fz cells, Vang (in Fw-Vang cells) was recruited to/stabilized at membranes contacting Fw-Fz cells and localizing to such intercellular contact sites (Fig. 6D-D”’). Cells expressing only Fw were unable to recruit Vang to the membrane in Fw-Vang cells (not shown) and, similarly, FwΔCCP2 expressing cells (which do not mediate cell adhesion) failed to recruit Vang even in the presence of Fz in neighboring cells (Fig. 6F-F”’). This suggests that the Fw cell adhesion function can mediate the Fz-Vang intercellular interactions.

Discussion

Our data suggest that Fw serves as a homophilic cell adhesion molecule that physically interacts with and stabilizes Fz at membranes, facilitating Fz-Vang/Stbm intercellular interactions. We also conclude that Fw acts in a manner similar to that proposed for Fmi (Chen et al., 2008; Klein and Mlodzik, 2005; Lawrence et al., 2004; Le Garrec et al., 2006; Strutt and Strutt, 2008) and thus that Fw and Fmi may act in parallel (in a partially redundant manner) to facilitate Fz-Vang interactions.

Furrowed in PCP establishment

The function of Fw appears linked to that of Fz and Fmi, but the phenotypic strength of fw LOF is weaker than fz and fmi (except for the thorax). Mechanistic studies suggest that Fw is a homophilic cell adhesion factor and physically associates with and stabilizes Fz, promoting Fz PCP function. Similarly, the cell adhesion factor Fmi (Usui et al., 1999) can also associate with Fz and stabilizes it at the membrane (Chen et al., 2008). In vivo data suggest that fw and fmi function in parallel, partially redundantly, mediating intercellular Fz-Vang interactions as intercellular “bridges” (Fig. 7O). It is noteworthy that the double mutant phenotype of fw or fmi is not stronger than fz- itself, or in most cases stronger than the fmi null phenotype (except in the thorax where fw appears the more important of the two and the fmi null phenotype is not as strong as fz-).

In addition, Fw might affect Fz stability in a cell-adhesion independent manner as FwΔCCP2, with no cell adhesion capability, still stabilizes Fz when coexpressed. Thus our data suggest that Fw performs two separable mechanistic functions in PCP: (1) Fz stabilization via association with it (this might be of different importance in distinct tissues, e.g. more important in wing discs [thorax, wing] than eye discs), and (2) cell adhesion at junctional complexes, where it stabilizes Fz to facilitate intercellular Fz-Vang interactions (a function similar to Fmi [Chen et al., 2008]). Although the CCP2 domain is critical for cell adhesion, but not its Fz interaction, as Fz needs to be stabilized at cell junction complexes, the Fw effect on Fz in the absence of the CCP2 domain has no functional consequence. On both counts, Fmi is acting in a similar manner: it promotes Fz localization to subapical junctional membrane regions and overall affects Fz levels at the membrane (e.g. Das et al., 2004; Strutt and Strutt, 2008). fw⁻ has stronger LOF phenotypic defects in the thorax and wing, where overexpression of Fw shows no effects; in contrast, overexpression of Fw has strong effects in the eye, where LOF displays only a weak phenotype. It is likely that Fw levels are lower in the eye (hence the strong GOF PCP effect there) and Fmi largely serves the equivalent function(s) there.

In vivo and cell culture data suggest that Fw does not directly affect other core Fz-group PCP factors. The mild enhancement of Vang GOF defects is likely due to the effect of Fw on Fz, as Fz and Vang complexes antagonize each other intracellularly (rev. in Goodrich and Strutt, 2011; Seifert and Mlodzik, 2007). Fw does not have an apparent effect on Vang levels. Thus it appears that the phenotypic effects of Fw are mediated via its effects on Fz. Interestingly, PCP GOF effects of Fz in the eye are not only suppressed by the complete loss of fw, but are “misdirected” towards canonical Wg-signaling GOF defects, suggesting that Fw might contribute to Fz signaling specificity between the Wnt signaling branches.

Homophilic cell adhesion function of Fw

Fw is the sole Selectin in the Drosophila genome. In vertebrates, Selectins function as cell adhesion molecules via their carbohydrate binding C-type lectin domain (Zarbock et al., 2011), binding to glycolipids to mediate adhesion. This type of cell adhesion is prominent between leukocytes and endothelial cells, referred to as “rolling” under flow in blood vessels (Zarbock et al., 2011). The CCP repeats are thought to serve a structural function in this context, not mediating adhesion. In the context of PCP signaling, the CCP2 domain mediates direct homophilic adhesion between Selectin/Fw in neighboring cells. This is an unexpected result and reveals a role of Selectins in cell-adhesion.

The adhesive behavior of Fw is however significantly weaker than bonafide structural adhesion factors required for epithelial integrity like DE-cadherin. S2 cell based assays suggest that Fw provides about 1/5 the strength of DE-cad adhesion (determined by cell adhesion cluster size; see Suppl. Fig. S5). Accordingly, loss of function of fw does not affect epithelial integrity. In addition to our study on fw in PCP signaling, there are two additional defects associated with fw LOF alleles, (1) overgrowth in the retina (causing a “furrowed” appearance of the eye) and (2) a mild thickening/shortening or loss of sensory bristles (Leshko-Lindsay and Corces, 1997). Whereas the eye and bristle structure defects depend on both the CCP2 domain (cell adhesion) and the C-type lectin domain (sugar binding?), PCP establishment does not require the C-type lectin domain, suggesting that two Fw functions can be separated. As the mild overgrowth eye phenotype eye also depends on the CCP2 motif (and possibly on cell adhesion), fw could be considered a mild tissue specific tumor suppressor. Drosophila has an open circulation system and no blood vessels, thus it is unlikely that a glycolipid binding function, as established for vertebrate Selectins, is required in flies. Selectin knock-outs in the mouse or zebrafish models will provide a useful approach to address whether any of the vertebrate Selectins also function in PCP signaling.

Exp. Procedures

Fly strains and genetics

Flies were raised on standard medium, maintained at 25°C, unless otherwise indicated. Prepupae were collected and staged at 25°C for 5 h for prepupal wings and between 22-30 h for pupal wings.

fw⁴² was generated on FRT19A X-chromosome by EMS-based mutagenesis. The original chromosome was cleaned by recombination. y w fw⁴² clones were produced by mitotic recombination via the FLP/FRT system (Xu and Rubin, 1993) with ubxFLP in y w fw⁴² FRT19A/armlacZ FRT19A or with eyFLP (with rh1-GFP for the screen) in y w fw⁴² FRT19A/FRT19A backgrounds. GAL4/UAS system (Brand and Perrimon, 1993) was used for gene expression studies. The following lines were used: sev-GAL4, pnr-GAL4, dpp-GAL4, nub-GAL4, UAS-fw^RNAi (v39575, Bloomington Stock Center), and fmi^frz3 (Bloomington Stock Center).

Immunostaining and histology

Primary antibodies were: rabbit anti-Fz (1:300, gift from D. Strutt); rabbit anti-GFP (1:1000, Invitrogen); rabbit anti-myc (1: 500, Santa Cruz); mouse anti-myc (1:500, Santa Cruz); rat anti-DE-cadherin (1:20, DSHB); mouse anti-Flag M2 (1:1000, Sigma); mouse anti-Fmi (1:10, DSHB); rhodamine-phalloidin (1:500, Invitrogen); mouse anti-β-Gal (1:200, DSHB). Fluorescent secondary antibodies were from Jackson Laboratories. Confocal images were taken on a Zeiss 510 confocal laser-scanning microscope.

Eye sections were prepared as described (Gaengel and Mlodzik, 2008). For genetic interactions, eyes were sectioned near the equatorial region and only ommatidia with the correct photoreceptor number were scored for PCP defects.

Expression constructs and cell culture studies

please see Supplemental Information

Highlights.

1. Drosophila Selectin Furrowed (Fw) functions in planar cell polarity (PCP) signaling.

2. Fw promotes Frizzled (Fz) protein levels and stable membrane association.

3. Fw serves a homophilic cell adhesion function in PCP establishment.

4. Fw promotes formation of intercellular Frizzled-Vang interactions together with Fmi

E-TOC document

Chin et al. show that the Drosophila selectin Furrowed/Fw is required for planar cell polarity (PCP) establishment. Fw mediates PCP by promoting Frizzled membrane stabilization and intercellular interactions with Vang/Stbm. Fw also functions as a cell adhesion factor and is partially redundant with the PCP cadherin Flamingo.