Fly strains and procedures

The alleles used were shot/kak^V168, pot^P3, pot^P14, pot^P53 and pio^V132 (Walsh and Brown, 1998), pio^2L-22 and pio^2L-14 (Prout et al., 1997), and mys^XG43, dp^LV1 and sha^VAI51 [see FlyBase for references (http://flybase.bio.indiana.edu/)].

Microtubules were visualized using a pUASp::GFP65S- αTubulin84B insertion on the third chromosome (Grieder et al., 2000). Expression of upstream activation sequence (UAS) constructs was directed to the developing wing using the 69B (Brand and Perrimon, 1993) or CY2 (Queenan et al., 1997) Gal4 driver lines; 69B was also used for the embryonic epidermis.

Wing clones were introduced by heat shocking larvae in food vials for 2.5 hours in a 37°C water bath at 12-36 hours after hatching. To mark clones morphologically on chromosome arm 2R, the sha^VAI51 allele was introduced into FRT42D kak^V168 and FRT42D pio^V132 chromosomes. The MARCM (Lee and Luo, 1999) strategy was used to mark mutant clones positively on chromosome X and chromosome arm 2L.

Molecular characterization of piopio and papillote

The region containing piopio candidate region was narrowed to a 300 kb interval between the distal breakpoint of Df(2R)px² and the proximal breakpoint of Df(2R)bs^V100 (a deficiency recovered as a blistered allele) (Walsh and Brown, 1998) by PCR amplification of short test fragments from single embryos homozygous for either deficiency. PCR amplification of the predicted gene CG3541 (Adams et al., 2000) from genomic DNA of pio^V132/SM5 yielded a shorter, polymorphic fragment. Sequencing revealed that this was caused by a 1707 bp deletion removing the first two coding exons of the predicted gene (bases -512 to +1192 relative to the ATG). A 22 kb BamHI genomic fragment flanking CG3541 was cloned into a P-element vector and introduced into the genome to test for rescue. Sequencing of several cDNAs confirmed the predicted intron/exon structure and coding region of CG3541. However, the 3′ untranslated region (UTR) of pio exceeded the predictions by about 500 bp and there are several alternative splice forms of the 5′ untranslated region (UTR) (see submitted sequence, AY862157). The UAS-pioGFP construct was generated by PCR amplification of the pio open reading frame and 5′ UTR from cDNA and cloned between a UAS promoter and the green fluorescent protein (GFP) coding sequence. A chimeric transmembrane protein was generated by replacing the cytoplasmic tail of the Torso receptor tyrosine kinase (a type-I transmembrane protein) with that of Pio, as done previously with the integrin-β subunit cytoplasmic tail (Martin-Bermudo and Brown, 1999). The short Pio cytoplasmic tail was amplified and cloned into EcoRI/SpeI sites in the vector pWR UAS::torsoD. Sequences at the fusion junction are:

Pio: RTFAIAIAIAGLILMLAVVAAVLCIMARRSTKTVSNSGSSI-YS;

Torso: RGVLLSEGNMVKLVLFIIVPICCILMLCSLTFCRRNRSE-VQAL;

fusion: RGVLLSEGNMVKLVLFIIVPICCILMLCRILCIMARRST-KTVS.

The region containing pot was mapped by inclusion in Df(1)RA47 but not Df(1)KA7 or Df(1)N105. Sequence-tagged sites (STSs) derived from cosmid ends (Kimmerly et al., 1996) were mapped relative to the breakpoints of these deficiencies by single-embryo PCR, limiting the pot candidate region to a 73 kb interval between the STSs Dm0452 and Dm1866. From our phenotypic analysis, we expected an apical protein, and this interval includes a predicted gene [CG2467 (Adams et al., 2000)] encoding a protein with a signal peptide and a partial ZP domain. Sequencing of PCR products from heterozygous pot^P53/FM6 genomic DNA revealed a 13 bp deletion in the CG2467 gene, causing a frame shift after the last base encoding amino acid 686, which results in truncation of the mutant protein after an additional 15 unrelated residues. A 15kb SpeI/EclXI genomic fragment encompassing CG2467 was inserted into a P-element vector and introduced into the genome. To generate the Pot-CFP fusion construct, the cyan fluorescent protein (CFP) coding region (Clontech) was amplified and inserted into the pot genomic fragment using BamHI and EcoRI sites introduced beforehand by PCR. Sequencing of cDNAs confirmed the predicted exon-intron boundaries and the extent of the pot coding region, but the length of the 3′ UTR again exceeded the prediction (see submitted sequence, AY862156). Domain analysis was performed with SMART (Schultz et al., 1998) and PFAM (Bateman et al., 2002).

Immunohistochemistry and microscopy

Confocal images were collected using a Biorad Radiance 2000/Nikon E800 microscope with 40×/1.30 oil and 60×/1.40 oil objectives. Fluorophores used were either GFP and CFP or Alexa 568 and Alexa 488 dyes. GFP/CFP images were obtained directly from live embryos or unfixed pupal wings dissected from the pupal case and mounted in Vectashield or Voltalef oil. To perform whole-mount antibody staining on pupal wings, the wings were dissected, heat fixed in boiling PBS for 10 seconds, cut several times across the wing blade using microsurgical scissors and postfixed in 4% formaldehyde in PBS for 2 hours. After washing extensively, standard antibody-staining procedures were applied. To image wing morphology, Nomarski-interference-contrast images were collected at the wavelength of the excitation laser and overlaid with GFP images collected in parallel. Antibodies were used at the following dilutions: guinea-pig anti-Kakapo (courtesy of T. Volk, Weizmann Institute, Rehovot, Israel) 1:250; mouse anti-acetylated-α-tubulin (Sigma), 1:500; fluorescent secondary antisera (Vector Labs), 1:500. Ultrastructural analyses were carried out as described previously (Prokop et al., 1998a) on a Zeiss EM900 electron microscope. Figures were assembled in Photoshop (Adobe) and labelled with FreeHand (Macromedia).

Protein chromatography and detection

Wing lysates were generated by dissecting wings from transgenic and control pupae [either lacking any GFP or containing GFP alone expressed from the ubiquitin gene promoter (Luschnig et al., 2004)], shock freezing them on dry ice and grinding them in lysis buffer (PBS supplemented with 100 mM NaCl, 0.5% Triton X-100, 0.2% sodium dodecyl sulfate (SDS)] containing 1× concentrated Complete Protease Inhibitor Mix (Roche). Proteins from the supernatant were then separated on 10% or 12.5% polyacrylamide gels. Pio-GFP and Pot-CFP fusion proteins were detected using monoclonal anti-GFP antibody (Abcam; Vector Laboratories) and horseradish-peroxidase-coupled secondary antibodies (Jackson ImmunoResearch), followed by chemoluminescence detection (ECL; Pharmacia Biotech).

Molecular characterization of pio and pot

To discover the nature of the proteins encoded by pio and pot, we characterized them at the molecular level. Using deficiencies in the region, pio was mapped to a 300 kb region. As the mutant alleles were produced by X-rays, which often induce deletions, we scanned the genes in that interval for deletions by PCR. A deletion removing the first two coding exons of the predicted gene CG3541 (Adams et al., 2000) was found in the allele pio^V132 (Fig. 1A), suggesting that this gene corresponds to pio. To confirm this, a rescue construct containing 22 kb of genomic DNA encompassing the CG3541 transcription unit was constructed, introduced into the genome and tested for the ability to complement pio mutations. The construct fully rescued the lethality of pio mutations and therefore we conclude that CG3541 corresponds to pio.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

The wing-blister genes piopio and papillote encode ZP-domain-containing transmembrane proteins. The genes piopio (pio, A) and papillote (pot, B) are shown with their closely flanking genes and their encoded proteins. Untranslated regions are white-filled boxes and coding regions are black. Genomic rescue constructs were generated with the segment of DNA between the restriction sites: BamHI for piopio and SpeI and EclXI for papillote. The position of the 1.7 kb deletion in the pio^V132 mutant allele, which removes the signal peptide, is shown in A. The schematic of the proteins show signal peptides (SP) and transmembrane domains (TM) as black boxes, the zona pellucida (ZP) domains as grey boxes, and the insertion points of the fluorescent protein tags (GFP and CFP). For Piopio, the sequence surrounding the potential furin cleavage site (underlined) at the end of the ZP domain is shown, as is the conservation of the short cytoplasmic domain, with identical residues shaded. For Papillote, the position of the frame shift caused by the 13 bp deletion in the pot^P53 mutant allele, which causes a truncation of the protein after residue 686, is indicated. (C) Pio but not Pot is proteolytically processed. Extracts of pupal wings from wild type (wt) or animals expressing GFP, Pio-GFP or Pot-CFP were analysed by western blotting and probed with an anti-GFP antibody. Two distinct bands were detected for the Pio-GFP fusion protein, with apparent molecular weights as predicted for the full-length fusion protein lacking the signal peptide (76 kDa), and a C-terminal fragment fused to GFP generated by cleavage at the potential furin recognition site (39 kDa), distinct from GFP alone (27 kDa). By contrast, a single band was observed for Pot-CFP of the size expected for the full-length fusion protein (131 kDa).

Analysis of cDNAs identified in the expressed sequence tag (EST) project (Rubin et al., 2000) showed that different pio transcripts arise from alternative starts of transcription but that all transcripts produce the same protein. The sequence of the 462-amino-acid Pio protein shows that it is a transmembrane protein containing an extracellular ZP domain and a short cytoplasmic domain (Fig. 1A). ZP domains (Bork and Sander, 1992) were first identified in the proteins that compose the mouse zona pellucida and have since been found in several proteins secreted from the apical side of cells. Other ZP-domain proteins that have a similar structure to Pio, with a signal peptide and transmembrane domain, have been shown to be cleaved by furin-type proteases, thus secreting the extracellular domain (Litscher et al., 1999). The Pio protein contains a putative furin cleavage site (RPKR, residues 352-355) immediately following the ZP domain (amino acids 66-351), suggesting that its ZP domain might be secreted. The deletion in the pio^V132 allele removes the start of translation as well as the signal peptide and therefore this allele is a null mutation.

The pot gene was similarly mapped to a 73 kb region using overlapping deficiencies. Within the interval of genomic DNA containing pot, we found that CG2467 (Adams et al., 2000) contained a short deletion in the pot^P53 allele, which causes a frame shift. As described for pio, we confirmed that we had identified the right candidate by constructing and testing a rescue construct encompassing this gene (Fig. 1B), and therefore conclude that CG2467 corresponds to pot. EST analysis showed a single pattern of splicing for pot cDNA, which encodes a 963-amino-acid protein with an N-terminal signal peptide and region with weak homology to ZP domains (Fig. 1B). However, the Pot sequence contains an inserted transmembrane domain disrupting the region with ZP-domain homology, and the degree of sequence conservation drops after this insertion. We therefore conclude that Pot contains a highly divergent, partial ZP domain containing a well-conserved first third of this domain, immediately followed by a transmembrane domain. Because there are no predicted furin cleavage sites between the putative extracellular part of the ZP domain and the transmembrane helix, Pot is probably an integral membrane protein. Thus, both Pio and Pot share with Dp a ZP domain, which is linked to function in the aECM.

Both pio and pot are conserved in insects, with clear orthologues identifiable in the mosquito, honeybee and silkmoth (data not shown). Sequence similarity is particularly pronounced within the ZP domains of both proteins and, in the case of Pio, the short cytoplasmic tail (Fig. 1A). Although other animals have ZP-domain-containing proteins, there do not appear to be orthologues of pio and pot, suggesting either that they fulfil an insect-specific function or that their functional equivalents in other species cannot be detected by their primary sequence.

Apical distribution of Papillote and Piopio in the pupal wing

To analyse the subcellular localization of Pot protein, we generated a CFP-tagged version, by inserting the CFP coding sequence between the last amino acid of pot and the stop codon in a genomic rescue construct containing the transcription unit and flanking DNA with putative regulatory elements. This construct and its untagged version were able fully to rescue the lethality of all pot alleles, demonstrating that the CFP fusion does not impair function. Confocal microscopy revealed that in the developing wing Pot-CFP was localized to the enlarged apical surface of these cells, where it was evenly distributed (Fig. 2A). We did not detect expression of Pot-CFP at an earlier stage of wing development in the larval imaginal disc (not shown).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Subcellular localization of Pio and Pot. Distribution of fluorescent fusion proteins in late pupal wing (>70 hours after puparium formation), visualized by confocal microscopy. (A) Owing to the folding of the wing, this confocal section shows the localization of Pot-CFP around the circumference of the stellate apical surface (top) and an optical section through the two attached layer of cells (a cell in each layer is marked with a yellow arrow, with the arrowhead basal) (bottom). (bottom) Pot-CFP is uniformly localized along the apical surface (but not including the wing hair). (B,D) By contrast, at the same stage, a Pio-GFP fusion protein (green) is found in discrete dot-like structures, preferentially at the apical side (D) and near the circumference of the wing epithelial cells as visualized by differential interference contrast (right). (C) Tubulin-GFP shows broader but still discrete points of contact with the apical surface, also enriched at the cell periphery. The outline of one cell is drawn in white in B and C. (E) The basal position of integrin adhesive junctions is shown with integrin-linked kinase (ILK-GFP). (F) Schematic diagram showing the sections of the wing epithelia shown in A-E. Bar, 10 μm (applies to A-E).

Because the pio gene is substantially larger than pot, we used a cDNA to construct a version of Pio that was C-terminally tagged with GFP and expressed it with a UAS promoter (UAS::pio-GFP); the CY2 Gal4 driver was used to drive expression of Pio-GFP in the pupal wing. Like Pot, Pio was found at the apical surface of the wing epithelial cells (Fig. 2B,D), but in punctate structures rather than being evenly distributed. The dots are enriched preferentially along the apicolateral circumference of the cells (Fig. 2B). The microtubule transalar arrays have a similar distribution when viewed from the apical surface (Fig. 2C, Fig. 3). Neither Pio nor Pot was found at the basal integrin-mediated junctions, where proteins such as GFP-tagged integrin-linked kinase are localized (Fig. 2E) (Zervas et al., 2001). Thus, both Pio and Pot are found on the apical surface, where Dp is also thought to function.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Pio, but not other wing-blister gene products, is required for the transalar arrays of microtubule bundles in pupal wings. The distribution of tubulin in late pupal wings (60-70 hours after puparium formation) is shown in white or green in each panel except B, which shows an adult wing. All cells express tubulin-GFP in A,C-E. In a confocal section of a wild-type folded pupal wing (A), the prominent microtubule bundles spanning the wing cells can be seen. Pairs of attached cells are marked with yellow arrows, with the arrowheads basal. (B-E) Clones of mutant cells are marked with a mutation in shavenoid (sha), which disrupts the formation of wing hairs but does not cause blistering, as shown in a bright-field micrograph of the adult wing (B). In the absence of Pio, the microtubule bundles are not seen: (C) a section across the apical surface, combined with differential interference contrast, showing the absence of microtubules and wing hairs in the sha pio mutant clone; (D) the absence of the microtubule bundles in sha pio mutant cells (indicated with a white horizontal bracket); opposing wild-type cells also appear overcontracted. The microtubule-binding protein Shot, which localizes to both ends of the microtubule bundles as detected by antibody staining (F), is not required to maintain the bundles (E); this image combines tubulin-GFP fluorescence and differential interference contrast, and the white horizontal bracket shows the cells lacking Shot (sha shot). (G-I) The MARCM technique was used to express tubulin-GFP in just the mutant clones of cells. A control wild-type clone shows microtubule bundles in one layer of the wing (G). Clones of cells lacking Pot (H) or the integrin βPS subunit (I, mys) contain microtubule bundles. These bundles are disordered in the absence of integrins, presumably owing to the separation between the two cell layers (the other layer is not visible because it does not contain a clone). Bars, 10 μm.

The C-terminal CFP and GFP tags allowed us to assess the processing of Pio and Pot. As mentioned above, Pio, but not Pot, has a consensus recognition site for furin proteases, which are thought to cleave other ZP-domain-containing proteins. When PioGFP was examined by western blotting of pupal wing lysates, bands of a size consistent with both cleaved and uncleaved protein were detected at similar levels (Fig. 1C). In combination with the distribution of the GFP-tagged protein, this suggests that some Pio remains unprocessed and is associated with the punctate apical structures, and that some is cleaved. We cannot assess the distribution of the cleaved form of Pio-GFP, because it no longer contains the GFP moiety. By contrast, Pot-CFP does not appear to be cleaved at all (Fig. 1C), consistent with the lack of extracellular furin cleavage sites.

pio and pot mutant phenotypes in the wing

The apical localization of Piopio and Papillote, together with the fact that they contain ZP domains, suggested that they contribute to the aECM or its link to the epidermis. The wing blisters produced by clones pot or pio mutant wing cells might therefore be due to a separation of the aECM from the epithelial surface rather than a separation between the basal surfaces of the two epithelial sheets, as seen in integrin mutant clones (Brabant et al., 1996). To test this we used tubulin-GFP (Grieder et al., 2000) to label the two epithelial layers. Transalar microtubule bundles were readily detected in the intervein cells of late pupal wings of wild-type flies (Fig. 3A). We marked pio mutant clones with a mutant in shavenoid that eliminates the hair produced by each hair cell (Taylor et al., 1998) but does not by itself cause wing blisters (Fig. 3B). Although using the combination of tubulin-GFP and shavenoid failed to show the expected detachment between epithelia and aECM, it revealed a surprising phenotype for pio mutations.

We found that pio mutant cells completely lacked the transalar microtubule arrays (Fig. 3C,D). The loss of Pio and the microtubules resulted in an expansion of the mutant cell and a shortening of the wild-type cells in the opposing epithelial layer (Fig. 3D), suggesting that microtubules might be required for a balance in tension between the two layers. Despite the dramatic effect on the transalar arrays, the two epithelial sheets remained associated and the integrity of both epithelia was intact in the absence of Pio. This finding prompted us to examine microtubule organization in cells mutant for shot, a wing-blister gene that encodes a microtubule binding protein of the spectraplakin family (Röper et al., 2002). Shot is localized to apical and basal ends of the microtubule bundles in the embryonic tendon cells (Gregory and Brown, 1998; Strumpf and Volk, 1998) and we found that has a comparable distribution at the ends of the transalar microtubule bundles (Fig. 3F). In the absence of Shot in the embryo, muscle contraction ruptured the tendon cells and these cells had their microtubules detached from integrin junctions at the basal surface (Prokop et al., 1998b). However, we did not detect any change to the transalar arrays in shavenoid marked clones of shot mutant cells in the pupal wing, nor did the mutant cells rupture or detach from the opposing wild-type cells during pupal stages (Fig. 3E).

We examined additional wing-blister mutants using the MARCM technique (Lee and Luo, 1999) to mark the mutant clones, so that only mutant cells express tubulin-GFP. In this technique mitotic recombination results in clones of cells that are homozygous for a given mutant that have also lost a transgene ubiquitously expressing the transcriptional repressor Gal80. This relieves the Gal80-mediated repression of Gal4-driven expression of a marker under the control of the UAS promoter (in this case UAS::tubulin-GFP) in the mutant cell clone. Cells mutant for pot had apparently normal transalar arrays (Fig. 3G,H), as did cells mutant for dp, the third wing-blister gene encoding a ZP-domain protein (Wilkin et al., 2000) (data not shown). In clones of cells lacking the βPS integrin subunit (myospheroid mutant) (Brower and Jaffe, 1989), the separation between the cell layers was visible but the transalar microtubule bundles still extended along the apicobasal axis of the detached cells (Fig. 3I). Therefore, Pio appears to be uniquely required for the formation or maintenance of the transalar arrays. Because Pio has transmembrane and cytoplasmic domains, and some Pio remains uncleaved (Fig. 1), one hypothesis is that Pio recruits microtubule-nucleating machinery to apical junctions. This is consistent with the similarity between the punctate contacts of the microtubule bundles on the apical surface and the punctate distribution of Pio (Fig. 2B,C). We tested whether the short cytoplasmic domain of Pio could recruit microtubules on its own by fusing it to a heterologous transmembrane protein. However, expression of this chimeric protein in pupal wings did not perturb microtubule organization, nor was it able to rescue the loss of the transalar arrays in the pio mutant cells (data not shown).

As mentioned above, in the absence of integrins, the wing blister appears within the pupal wing (Brabant et al., 1996). By contrast, in the absence of Dp, Pio or Pot, the blister appears shortly after eclosion and we did not see any separation between the cell layers of the pupal wing. This demonstrates that the loss of these genes did not severely alter integrin function because, if it did, we would expect to see an early blister, as also found when the integrin-associated protein talin is removed (Brown et al., 2002). Unfortunately, we also did not see evidence for the expected detachment between the apical surface of the clone and the aECM. We therefore examined the phenotype in mutant embryos to test the role of these proteins in the formation or attachment of the aECM.

Detachment of apical ECM from the epidermis in pio, pot and dp mutant embryos

Mutations in each of the wing-blister genes encoding ZP-domain proteins are embryonically lethal when homozygous. By expressing tubulin-GFP in the epidermis, a separation between the epidermis and the cuticle was visible in all three mutant backgrounds (Fig. 4C,D, for pio only). In contrast to the wing, we did not see any specific failure in the assembly of microtubules in the tendon cells of the pio mutant embryos (data not shown; see ultrastructural analysis below), suggesting that Pio might specifically be required when microtubules are nucleated from the apical junctions in the pupal wing in the absence of centrosomes (Tucker et al., 1986).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

The absence of Pio causes cuticle detachment and tracheal defects. (A) In wild-type embryos at the end of embryogenesis, viewed by differential interference contrast, the dorsal trunk of the trachea appears smooth and rather straight. (B) Homozygous pio mutant embryos complete development but the tracheae appear twisted and broken (arrows). (C) In live wild-type embryos, the epidermis (marked with tubulin-GFP) closely follows the outline of the cuticle (viewed by differential interference contrast). (D) In homozygous pio mutant embryos, the cuticle detaches from the GFP-marked epidermis (arrow).

At the ultrastructural level, embryos homozygous for pio, pot or dp mutations displayed identical defects (Fig. 5). In each case, a separation of the aECM (cuticle) from the apical surface of the epidermis was observed. Defects were observed in the innermost layer of the cuticle, rather than in the cellular adhesive structures.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Ultrastructural analysis of mutant phenotypes. Schematic representations of wild-type cuticle (A) or muscle-attachment sites (B) of late Drosophila embryos (about 21 hours) and micrographs of these structures in wild-type and mutant embryos (C-F”). Symbols and abbreviations are used consistently throughout: aj or white arrow, apical junctions; at or curved white arrows, apical transverse filaments; ch, chitin layer; cu, trilaminar cuticulin layer; dz or black arrowheads, deposition zone; ed, epidermal cell; el, electron-lucent layer; en or asterisk, endocuticle; ep or black dot, epicuticle; es, extracellular space; mt or white arrowhead, microtubules; mtj or curved black arrow, myotendinous junction; mu, muscle; nu, nucleus; pe, fibrillar protein epicuticle; tf or white chevron, tonofilament; tn, tendon cell. [Nomenclature according to Kaznowski et al. (Kaznowski et al., 1985) and Tepass and Hartenstein (Tepass and Hartenstein, 1994).] The characteristic arrangement of the epicuticle (black dots) and the typical occurrence of the darker cuticular deposition zone (dz), which associates with epidermal microvilli (mv), can be seen in wild-type (A-C”), pio^V132 (D,D′), pot^P14 (E,E′) and dp^1v1 (F,F′) mutant embryos alike. Typical features of epidermal tendon cells (tn) are their pronounced basal junctions with muscles (curved black arrows) and apical junctions (white arrows) adhering to tonofilaments (white chevrons) that anchor in the cuticle. Apical and basal junctions of tendon cells are intracellularly connected via microtubules (white arrowheads). These characteristic features of wild-type embryos (B,C”) seem unaffected in pot^P14, pio^V132 and dp^1v1 mutant embryos (D“-F”), although they are difficult to find and interpret owing to epidermal detachment from the cuticle. The main defect of pot^P14, pio^V132 and dp^1v1 mutant embryos consists in a disruption of the chitin layer (ch) of the endocuticle (asterisks), leading to a separation of deposition zone and epicuticle (D-F,D′-F′), and failure of tonofilaments to anchor in the cuticle (D”,F”, white chevrons). Bar, 0.5 μm (left and right columns, C-F,C“-F”), 0.2 μm (middle column, C′-F′).

In wild-type embryos (Fig. 5C), the aECM, or cuticle, consists of a complex, multilayered, largely proteinaceous epicuticle that overlays the bilayered endocuticle (Fig. 5A,B). The endocuticle is composed of a thick, electron-light chitin layer [containing the chitin polysaccharide fibrils and associated proteins (Andersen, 1979)] and a thin, basal, darker layer called the deposition zone (Kaznowski et al., 1985). Contact between the epidermis and the cuticular deposition zone mostly occurs in form of microvilli on the apical surface of the epidermal cells. In each of the mutant embryos, the chitin layer of the endocuticle (Fig. 5D-F, asterisk) has disintegrated, especially towards its basal side. An electrondense layer, presumably corresponding to a defective deposition zone, was faint and discontinuous compared with the wild-type deposition zone, and its typical tight contact to the chitin layer above was lost (Fig. 5D-F′, black arrowheads). On the apical surface of the mutant epidermal cells, the typical microvilli can still be seen, often attached to the remaining fragments of the deposition zone. All elements of the external epicuticle appeared to be properly formed in the mutant embryos and remained in tight contact with the upper layers of the endocuticle. Thus, dp, pio and pot are required for the normal formation of the innermost layer of the aECM and its attachment to the epidermis.

Junctional structures within the epidermal cells appeared normal in dp, pio and pot mutant embryos. Because of the disruption of the transalar arrays in wing cells lacking Pio, we closely examined the apical connection of the microtubules in all three mutants, but they appeared normal. As in the wild type (Fig. 5C”), the apical and basal junctions of tendon cells in the mutant embryos were connected by large microtubule bundles (Fig. 5D“-F”, white arrowheads) and, at the apical side, long tonofilaments extended away into the extracellular space (Fig. 5D“-F”, white chevrons). Normally, tonofilaments anchor in the cuticle but, in all three mutant backgrounds, they appear to have been pulled out. Consistent with the proposed apical function of these proteins, the basal integrin-containing adhesive junctions of the tendon cells appeared normal (black curved arrows in Fig. 5C“-F”). In addition, the attachment of the basal surface of the rest of epidermal cells to the basement membrane [which is not dependent on integrins (Prokop et al., 1998a)], was also indistinguishable from the wild type (data not shown). In conclusion, the ultrastructural analysis has confirmed that the function of the ZP-domain-containing proteins Dumpy, Piopio and Papillote is restricted to the apical surface of the epidermal cells, where each of the proteins is required for the attachment to and/or assembly of the inner layer of the aECM.

Despite the similarities between the defects observed at the EM level, the embryonic mutant phenotypes of the three genes are not identical. The cuticle of homozygous pot embryos appeared faint after the internal tissues were dissolved with Hoyer's mountant (Walsh and Brown, 1998), suggesting that the aECM is not as resistant to this treatment, whereas the cuticle of pio or dp embryos appeared normal (data not shown). The normally straight main tracheal trunk appeared twisted and broken in places in pio (Fig. 4A,B) or dp mutant embryos, owing to problems in the intercalatory movements during tracheal morphogenesis (Jazwinska et al., 2003), but this phenotype was not found in pot mutant embryos (data not shown). These findings, combined with the observation that Pio is required for the formation of the transalar arrays whereas the others are not, indicated that these three proteins do not work together as a simple functional unit but that each has distinct functions.

Molecular functions of Pio, Pot and Dp

An extracellular site of function for the Pio, Pot and Dp proteins is consistent with their structure. All three possess a signal peptide and a ZP domain that is thought to mediate the formation of extracellular protein fibres or meshworks (Jovine et al., 2002). In addition, Pio was also molecularly characterized by a group studying tracheal morphogenesis in the Drosophila embryo (Jazwinska et al., 2003) and shown to be localized on the apical lumen of the trachea. Conceptual translation of the pio and dp open reading frames predicts proteins with a transmembrane domain and a short cytoplasmic tail, and thus these proteins might function as transmembrane receptors. However, transmembrane ZP-domain proteins can be post-translationally cleaved at recognition sites of furin-like endopeptidases, as shown for the mammalian ZP2 and ZP3 proteins (Litscher et al., 1999) and for the fly ZP-domain protein NompA, which is required for the adhesion of peripheral mechanoreceptors (Chung et al., 2001). Both Dp and Pio possess a putative cleavage site immediately following the ZP domain. We found that the Pio-GFP fusion protein was cleaved and that the size of the tagged C-terminal cleavage residue was consistent with cleavage of Pio at the furin site. Owing to the gigantic size of Dp, corresponding experiments are almost impossible, but it seems likely that similar proteolytic processing occurs. This would suggest that at least a proportion of Dp and Pio is cleaved and therefore released from the apical surface, where it could contribute to the aECM. The divergent ZP domain of Pot is interrupted by a predicted transmembrane domain and the putative furin cleavage site is not present in this protein, suggesting that Pot remains associated with the apical plasma membrane.

Ultrastructural analysis of the cuticle of embryos lacking Dp, Pio or Pot showed an identical defect, with the disruption of the cuticle limited to the chitinous endocuticle. The most-basal deposition zone appears faint and sometimes discontinuous, and layers within the endocuticle immediately above disintegrate, causing a gap to appear between the remainder of the cuticle and the underlying epidermis. Neither mutation affects assembly of the epicuticle or its attachment to the endocuticle. The unique faint cuticle phenotype of pot was not distinguishable at the EM level but fits with the even distribution of pot in the aECM of the wing. The EM phenotype of these three genes is clearly distinct from that seen in coracle mutations, in which the epicuticle splits away from the endocuticle (Lamb et al., 1998). Instead, the disruption of the endocuticle seen in pio, pot or dp is reminiscent of the defects in the tectorial membrane of mice carrying a targeted deletion in the ZP-domain-containing protein α-tectorin (Legan et al., 2000). In such mice, the tectorial membrane detaches from the apical side of the underlying neuroepithelium. In addition, the finely striated matrix surrounding the collagen fibrils of the tectorial membrane is lost (Legan et al., 2000), coinciding with an absence of immunoreactivity for the ECM proteins otogelin (Cohen-Salmon et al., 1997) and β-tectorin (Killick et al., 1995), which is structurally similar to Pio. In the zona pellucida itself, which gave rise to the name of the protein family, three ZP-domain-containing proteins form a three-dimensional meshwork, and it has been suggested that ZP1 cross-links fibres containing the ZP2 and ZP3 proteins (Greve and Wassarman, 1985; Jovine et al., 2002). A similar cooperative interaction between the three proteins could explain why mutations in pio, dp or pot cause indistinguishable defects in the association of the apical surface of the embryonic epidermis with the aECM or cuticle.

In addition to their shared function, each protein appears to provide unique functions. Dumpy is a gigantic protein, with hundreds of epidermal-growth-factor-like (EGF) repeats combined with additional repeated domains. These are likely to provide interaction sites for binding to other aECM components. Pot contains an unusual half-ZP domain, followed by a transmembrane domain. This could help to keep the aECM closely apposed to the plasma membrane. A role in the assembly of a rigid aECM is also plausible, because Pot-CFP was readily detected in pupal wings but not in imaginal discs within the larva, which contain a more diffuse aECM (Brower et al., 1987). Alternatively, it is possible that the hydrophobic region could form a tight interaction domain rather than a transmembrane domain, which induces dimers of Pot for secretion. Pio has the unique role of contributing to the formation or stability of the microtubule transalar arrays in the wing. Ultrastructural studies have shown that the apical junctions are associated with the microtubule nucleating activity required to form the transalar arrays (Mogensen and Tucker, 1987; Mogensen et al., 1989). These microtubules have a greater circumference than those nucleated in these cells at an earlier stage when a centrosome is still present (Tucker et al., 1986), suggesting that they are nucleated in a different way. Thus, Pio could play a role in the microtubule-organizing activity of the apical junctions.

The simplest way for Pio to contribute to microtubule organization is if a proportion of the protein remained an uncleaved transmembrane protein, and the cytoplasmic domain interacted with proteins that form the apical microtubule organising centre. Our efforts to test this by expressing a chimeric transmembrane protein containing the Pio cytoplasmic domain to see whether it perturbed microtubule organization were not successful. However, it should be noted that the conserved cytoplasmic domain is rich in residues that can be phosphorylated (Fig. 1A), so extracellular interactions might be essential to trigger modification of the cytoplasmic domain so that it can make the relevant intracellular interactions. Two findings support this model. First, C-terminally tagged Pio was readily detectable both by fluorescence and western blotting, whereas the cleaved C-terminal fragments from murine ZP2 and ZP3 were found to be rapidly cleared (Litscher et al., 1999; Qi et al., 2002). Second, Pio was found in discrete points on the apical surface, similar to the points where microtubules were concentrated, in contrast to Pot, which was evenly distributed over the apical surface. However, we cannot at present rule out an alternative model in which the extracellular domain of Pio is a ligand for an unknown transmembrane receptor that regulates microtubule function.

In addition to pio, pot and dp, the Drosophila genome contains another 14 genes encoding ZP-domain-containing proteins (PFAM entry PF00100) (Bateman et al., 2002). Two of these genes, miniature and dusky, have recently been shown to play a role in the development of the wing that is substantially different to the one described here (Roch et al., 2003). They are also likely to be components of the aECM but, in contrast to the wing-blister phenotype, the absence of these proteins leads to defects in the cell-shape changes that occur during pupal development. This phenotype highlights the role of the aECM in morphogenetic processes, as also seen with Pio and Dp during tracheal morphogenesis (Jazwinska et al., 2003). Five of the ZP-domain-containing proteins in the genome are closely related to NompA, in that they also contain extracellular PAN modules (Tordai et al., 1999). NompA is restricted in function to the nervous system, where it plays a role in linking the neuron to the aECM of the dendritic caps of the mechanoreceptors (Chung et al., 2001). Thus, although ZP-domain-containing proteins in Drosophila have consistent roles at the apical surface of the cell, a range of distinct functions are mediated by these proteins.

Mutations in pio are the first to be identified that lead to defects in the formation of the transalar arrays in Drosophila. We were surprised to find that loss of Shot, which contains a microtubule-binding domain, did not perturb the organization of the bundled microtubules. Therefore, further analysis of Pio function might lead to an understanding of how microtubules can be nucleated at plasma-membrane junctions independent of centrosomes, a process that is likely to be conserved across evolution.