Pericardin, a Drosophila type IV collagen-like protein is involved in the morphogenesis and maintenance of the heart epithelium during dorsal ectoderm closure

Cell interactions between the extracellular matrix (ECM) and neighboring cells play crucial roles in the regulation of cell behavior and fate. From the very first stages of embryogenesis, these interactions are a prerequisite for the full expression of the determination-differentiation properties of cells and for tissue organization. Beyond the obvious scaffolding functions of the ECM in cell adhesion, migration and tissue morphogenesis, the matrix is also responsible for transmitting environmental cues to cells that affect essentially all aspects of the life of the cell (for a review, see Geiger et al., 2001).

In vertebrate heart development, the process by which the two lateral precardiac regions fuse along their lengths following a lateral to ventral folding is the result of an interplay between the splanchnic mesoderm and the underlying endoderm (Molkentin et al., 1997). Interactions of substrate adhesion molecules, such as those of integrins expressed by heart precursor cells with fibronectin and laminin in the extracellular matrix, are considered to be instrumental for cardiogenesis (Collo et al., 1995; Kalman et al., 1995). Likewise, both soluble morphogens such as BMP4 (TGFβ family) and extracellular matrix proteins such as fibronectin may be key actors in directing the migration of cardiomyocytes during heart development in amphibians and birds (Linask and Lash, 1988a; Linask and Lash, 1988b; Sugi and Lough, 1995)

The Drosophila heart, or dorsal vessel, is a hemolymph pumping organ made up of a limited number of cells and cell types (Ruggendorff et al., 1994) (reviewed by Bodmer and Frasch, 1999). It consists of a double row of cardial cells that express muscle-specific proteins, coalesce to form the heart tube enclosing a lumen and are the contractile cells of the heart. Located at the dorsal midline, the heart is flanked on either side by several types of pericardial cells. These cells are loosely associated with cardial cells and do not express muscle proteins. The precise roles pericardial cells play during heart development as well as their physiological function have remained, until now, largely unknown (Rizki, 1978). Anteriorly, the pericardial cells terminate in the lymph glands, which are bilaterally symmetric cell clusters and, at the most anterior end, the dorsal vessel is surrounded by the ring gland, an endocrine organ of complex origin. The heart tube is covered by a network of extracellular matrix components (Ruggendorff et al., 1994) among which some are localized in specialized areas of its surface (Zaffran et al., 1995). The heart tube is affixed to the underside of the dorsal body wall by seven segmentally repeated pairs of alary muscles.

The development of the Drosophila heart begins with the specification of cardial precursor cells that originate from the dorsal crest of the mesoderm monolayer and involves the action of the homeobox gene tinman (Azpiazu and Frasch, 1993; Bodmer, 1993). This initial subdivision of the mesoderm requires additional inductive signals from the overlying dorsal ectoderm to become fated to form cardial and/or pericardial cells (for a review, see Bodmer and Frasch, 1999). In stage 11 embryos, these precursor cells are metamerically organized in clusters of mesenchymal cells. During germband retraction (late stage 11 and stage 12), the cells acquire a polarity in a typical mesenchyme-epithelium transition and reorganize their shape to form a continuous epithelial layer on each side of the dorsal opening (Ruggendorff et al., 1994; Zaffran et al., 1995; Frémion et al., 1999). Later in the process of dorsal closure, the two rows of cardial cells, together with the pericardial cells, which are attached to the basal membrane of cardial cells, migrate dorsally and fuse at the dorsal midline to form the heart tube enclosing a lumen.

Once the heart has formed in late embryogenesis, the dorsal vessel shows clear structural differences along its length (Rizki, 1978; Bate, 1993). The posterior part constitutes the heart proper. This region is broader than the rest of the heart tube and consists of three segments. It contains the pacemaker activity and three segmental pairs of ostiae or valves that allow the lateral entry of hemolymph in the heart. Anterior to the heart is a narrower section termed the ‘aorta’ that encompasses four segments and is devoid of ostiae (Rizki, 1978; Molina and Cripps, 2001) (M. A., unpublished).

Despite its apparent simple structure and function, the Drosophila dorsal vessel shares several similarities with the early-stage hearts of vertebrates (reviewed by Bodmer and Frasch, 1999). Homologues of Drosophila genes function during vertebrate cardiogenesis, suggesting a conservation of molecular mechanisms in the formation of these essential circulatory organs. Early in vertebrate development, uncommitted splanchnic mesoderm residing on each lateral half of the developing embryo becomes specified to a cardiogenic fate by diffusible factors released from the underlying endoderm. Once specified, the cardiogenic precursors reorganize by a mesenchyme-epithelium transition and migrate along the anterior intestinal portal (AIP) to converge at the midline of the embryo where they form the cardiac crescent that folds ventrally, resulting in the fusion of the cardiac primordia and formation of the linear heart tube. The tube, subsequently, undergoes looping morphogenesis, which does not occur in Drosophila (Molkentin et al., 1997).

Adhesion molecules and ECM interactions are presumably involved in the main steps – cell specification, mesenchyme-epithelium transition, acquisition of cell polarity, migration, maintenance of structure – that are necessary to build up the cardial epithelium and to form the linear heart tube. During the early stages of Drosophila heart development, ECM molecules (laminin, collagen IV) are not expressed at detectable levels. When the heart lumen starts to form, adhesion molecules become integrated in the apical and basal extracellular matrix that underlines both sides of the tube (Tepass and Hartenstein, 1994). Laminin participation in the maintenance of the ultrastructure of the heart and a role for the PS integrins in the movement and migration of the pericardial cells via laminin as a ligand have been proposed (Yarnitzky and Volk, 1995; Stark et al., 1997; Martin et al., 1999).

In this work, we have focused our analyses on the function that extracellular matrix-mediated adhesion could play in the coordinated dorsal migration of the heart cells and that of the overlying ectoderm during dorsal closure (see Fig. 1). A novel collagen-like component of the Drosophila extracellular matrix, recognized by a monoclonal antibody, EC11 (Zaffran et al., 1995), is expressed in pericardial cells and is highly concentrated in the dorsolateral part of the heart, a region in close contact with the ectoderm. We have named the protein recognized by the antibody, Pericardin (Prc). Pericardin is a good candidate to participate in the movement of the heart and ectodermal cells. Our results suggest that, during dorsal closure and migration of the cardial epithelium, Prc could serve as a link to coordinate the movements of the two cell populations, and that modification of its expression could result in a concomitant disruption of the heart tube.

Wild-type embryos were from the Oregon R strain. The Df(3L)vin6 (Akam et al., 1978) deficiency in the region of the prc gene and the puc^E69 allele (Martin-Blanco et al., 1998) were obtained from the Bloomington Drosophila Stock Center.

Standard molecular biology methods were used (Sambrook et al., 1989). Three EST clones CK 02611 (1.2 kb), CK 01593 (0.95 kb) and CK 02594 (0,9 kb) were obtained from the Berkeley Drosophila Genome Project (BDGP) (Kopcynski et al., 1998) and were used to screen a Canton-S 12-24 hour embryonic cDNA library (Brown and Kafatos, 1988). No clone could be identified when using the EST 02594 as probe and the EST expression profile described by BDGP could not be reproduced. That EST, therefore, was not considered as relevant and was not used further. The two other ESTs identified a cDNA clone – prcV2 – that was sequenced on both strands (Genome Express, Grenoble, France). High density filters of P1 clones from the P1 Drosophila library were purchased from Genome Systems (St Louis, MO). Hybridization of the filters with the EST clones was carried out as described in ‘P Drosophila filter overview’ (provided by Genome Systems). The P1 clone DS 00169 (BDGP) was obtained from Dr Ashburner’s laboratory

A 4.5 kb genomic fragment (prc4.5) located upstream of the 5′-end of prcV2 (see Fig. 4A) was PCR amplified from DS 00169 by using as primers T7 and a specific primer: 5′-CCGATTTGCTT-CCGATCGCG-3′, complementary to the 5′-end of prcV2 and containing the first PvuI restriction site of the cDNA, subcloned in pGEM-T Easy (Promega, France) and finally cloned in the NotI site of the pCaSpeR-AUG-β-gal polylinker (Thummel et al., 1988).

The cDNA prcV2 was excised from pNB40 by BglII digestion and inserted into BglII-cut pUAST (Brand and Perrimon, 1993). This construct was injected in flies and independent UAS-prc lines were established and crossed to engrailed-GAL4 flies (a generous gift from Dr Gallet).

GenBank Accession Number for prc is AF203342.

prc4.5 was amplified with a modified T7 primer that included an additional NotI site in its 5′ end, purified, digested with PvuI, and ligated in a 1/1 molar ratio to PvuI-NotI cut and purified prcV2. The presence of a full-length minigene after subcloning in pBlueScript (Stratagene), was assessed according to three criteria: (1) restriction enzyme mapping, (2) PCR amplification with primers scattered randomly on the total length of the minigene and (3) Southern blotting with probes from different regions of the minigene. It was finally inserted into NotI cut CaSpeR 4 (Pirotta, 1988).

The construct MN-prc was injected with the Δ (2-3) helper plasmid (Robertson et al., 1988) in yw embryos to generate transgenic flies by standard methods (Rubin and Spradling, 1982). For rescue experiments, flies of genotypes MN-prc; Df(3L)vin6/+ were constructed and crossed.

As described by Kennerdell and Carthew (Kennerdell and Carthew, 2000), two prc PCR amplified products [1025 bp from prcV2 3′-region (3446-4471); see Fig. 4A] were made with differing ends. One product had an EcoRI site at the 3446 end and a SfiI site at the 4471 end (first pair of primers: a and b, 5′-CGGAATTCGGACAACCTGGAATAGGCGG-3′ and 5′-GGCCAAGATGGCCGGATTGTGCAGCACCATGGT-3′; top strand sequence, GGCCATCTT-GGCC). The second product had a XbaI site at the 3446 end and a different SfiI site at the 4471 end (second pair of related primers: a′ similar to a with a different 5′ end, 5′-CGTCTAGA….; b′ similar to b with a different 5′ end, 5′-GGCCTTCTC…..; top strand sequence CCGGAAGAGCCGG). Underlined sequences indicate the central non palindromic core of each site. After digestion with SfiI, the two amplified products were ligated, the dimers purified and subcloned in EcoRI-XbaI cut pBlueScript (Stratagene). The resulting SfiI site created in the dimer has a central non palindromic ATCTC sequence. Finally, the dimers were subcloned in EcoRI-XbaI cut pUAST (UAS-IR prc) (Brand and Perrimon, 1993). All the ligated products were transformed into the DH5α strain of E. coli to maximize the stability of the inverted repeats (Kennerdell and Carthew, 2000). The spacer between the two repeats was only 13 bp long, which could explain the somewhat inefficient rate of cloning of the inverted repeat in E. coli.

A GAL4-driver gene specific for pericardial cells was constructed by using the prc4.5 genomic fragment. The GAL4 coding sequence excised from pGATB (Brand and Perrimon, 1993) and prc4.5 recovered from pCaSpeR-AUG-β-gal were inserted sequentially into CaSpeR4 (prc-GAL4).

According to the protocol described by Frémion et al. (Frémion et al., 1999), DIG-labeled DNA probes were used for whole-mount in situ hybridization and fixed staged embryos were stained with primary and secondary antibodies as follows: mouse or preadsorbed rabbit anti-β-galactosidase (Promega and Cappel, respectively) 1:1000; rabbit anti-Tinman (Azpiazu and Frasch, 1993) 1:800, preadsorbed; mouse anti-Pericardin (EC11) (Zaffran et al., 1995) 1:2; rabbit anti-Mef2 (Nguyen et al., 1994) 1:1000, preadsorbed; rabbit anti-Oddskipped (Ward and Skeath, 2000) 1:1000, preadsorbed; anti-α-Spectrin (Lee et al., 1993) 1:500, preadsorbed; mouse anti-Nrt (Piovant and Léna, 1988) 1:500. Affinity-purified secondary antibodies were either coupled to alkaline phosphatase or to biotin (Jackson Immuno Research Laboratories) and used at a 1:1000 dilution or were either Alexa-488 or Alexa-594 conjugated (Molecular Probes) and used at a 1:500 dilution. In some cases, the signal was amplified with the aid of a ‘Tyramide Signal Amplification’ kit (NEN life sciences). The stained embryos were mounted in Geltol medium (Immunotech, France) or, when fluorescent, in Vectashield (Vector Laboratories) for further observation under an Axiophot Zeiss microscope or a LSM 410 Zeiss confocal microscope.

Stage 17 embryos and first instar larvae were fixed as described (Berryman and Rodewald, 1990) and embedded in LR Gold resin (TAAB Laboratories equipment). Ultrathin sections were incubated with undiluted anti-Prc antibody overnight at 4°C and then, with 10 nm gold-labeled goat anti-mouse IgG (1:25) (Aurion) for 1 hour at room temperature. The sections were post-fixed in 2% glutaraldehyde and observed under a Leo 912 electron microscope.

Ten to 16 hours old embryos were homogenized in PBS buffer containing 1 mM EDTA, 2 M urea, 1.5% Triton X-100 and protease inhibitors. The proteins were incubated for 2 hours at 4°C with the EC11 monoclonal antibody coupled with dimethylpimelimidate (Pierce chemicals) to protein A-Sepharose 4B (Pharmacia, Uppsala). The bound antigen(s) was eluted with diethylamine (pH 11.5) and further purified by preparative electrophoresis on 5% polyacrylamide gels and western blotting. The nitrocellulose-bound antigen was submitted to trypsin digestion, and the peptides were purified by HPLC and sequenced. These experiments were carried out in Dr K. Williams’ laboratory (Yale University, New Haven). Owing to the high molecular weight of the protein, the peptides were difficult to purify. Three of them were sequenced but only one has yielded, with reasonable confidence, a eight amino acid stretch whose sequence (NFQSTYYTK) can be found in Prc. The two other sequences had too many ambiguities to be considered.

During dorsal closure, the migration of the two rows of epithelial cardioblasts that will join to form the dorsal vessel underneath the dorsalmost ectodermal leading edge (LE) cells is coupled to that of the dorsal ectoderm (Fig. 1). Cell-ECM interactions through local receptor mediated signaling between mesoderm and ectoderm have been suggested as a mechanism to ensure such a coordinated movement. The cardioblasts and the attached pericardial cells are close enough to the overlying ectoderm (Ruggendorff et al., 1994) to engage efficiently in these interactions via the extracellular matrix network that surrounds the heart tube (Zaffran et al., 1995).

Fig. 2 illustrates the relative positions of the dorsalmost ectodermal cells that express the puckered (puc) gene (Ring and Martinez-Arias, 1993; Martin-Blanco et al., 1998) and the heart tube. During dorsal closure (Fig. 2A), the two rows of cells expressing Prc on their surface followed the same direction as the lacZ-expressing cells (puc cells) but with a slight asynchrony in their movement (Fig. 2B) that did not persist after closure because the ectodermal cells are, by then, perfectly aligned with the cardioblasts (Fig. 2C). These results suggest that the interaction of the heart tube with the dorsal ectoderm involved cells that are in a more lateral position than the LE cells, in agreement with previous observations by Ruggendorff et al. (Ruggendorff et al., 1994) (see also Fig. 3E,D).

At stage 11, Decapentaplegic (Dpp), a member of the TGFβ superfamily, is expressed in the Puc-LE cells. Puc negatively regulates dpp expression through the Jun kinase (JNK) pathway that is involved in dorsal closure. Whereas dpp provides an effector of dorsal closure, puc encodes a regulatory element that controls the amount of signaling through the pathway (Martin-Blanco et al., 1998). In puc^E69 embryos, the JNK pathway remains constitutively activated resulting in Dpp overexpression in the ectodermal cells and in an abnormal expansion of Puc expression to several rows of cells (Fig. 2D) (Martin-Blanco et al., 1998). In homozygous puc^E69 mutants, dorsal closure takes place but the LE cells do not differentiate properly. Under these conditions, the two rows of cardioblasts are no longer able to migrate and to join at the dorsal midline to form the heart tube (Fig. 2D-F) resulting in a cardia bifida phenotype as described for MesP1 GATA4 mutant mice (Molkentin et al., 1997; Saga et al., 1999) and for various zebrafish mutants (Stainier, 2001). As shown in Fig. 2D,E, the cardial epithelium is confined to the boundary between puc mutant cells and more lateral ectodermal cells, as if the two rows of cardioblasts were repelled from the Puc-expressing cells. This observation indicates that a change in the fate specification of LE cells had resulted in a concomitant alteration of the coordinated movements of the dorsolateral ectodermal cells and cardioblasts cells during dorsal closure.

As mentioned above, the interaction taking place between dorsal ectoderm and the heart epithelium to coordinate cell movement during dorsal closure probably involves components of the extracellular matrix (Fig. 1). The spatiotemporal expression profile of Pericardin suggests that it could coordinate these movements (Zaffran et al., 1995). Prc is clearly detectable in early stage 13 embryos, its concentration increases in late-stage embryos and remains abundant in adults (Zaffran et al., 1995).

During dorsal closure and after the heart tube has finally formed, Prc is located around the periphery of the pericardial cells and outlines the basal surface of the cardioblast epithelium (Fig. 3A,B). As depicted in Fig. 3C, in embryos double labeled for Prc and α-Spectrin, a specific marker for the basal lateral membrane of epithelial cells, Prc appears particularly abundant at the boundary between the cardioblasts and the dorsal ectoderm. A better illustration of Prc localization was gained from sections focused either on the heart tube (Fig. 3E) or on the ectoderm (Fig. 3D). In that latter view, after the dorsal ectoderm has fused at the dorsal midline, Prc expression is concentrated in a position that corresponds to a layer of cells situated in a more lateral position (second or third row of ectodermal cells) than dorsal ectodermal cells.

Electron microscopy observation of a stage 17 embryo (Fig. 3F,G) in which the heart lumen is still very small and the ECM in a nascent state, shows that the expression of Prc is detectable only in the extracellular space around the pericardial cells and on the basal surface of cardiomyocytes. At the first instar larval stage (Fig. 3H,I), the lumen has considerably enlarged but Prc is totally absent from luminal ECM, while its concentration increases concomitantly with the maturation of the basal ECM, thus confirming the disymmetric distribution of Prc.

In conclusion, the specific location of Prc in close proximity to the dorsal ectoderm and around the heart tube could be consistent with its participation in the proper migration of the two layers of cells and/or in consolidating morphogenesis of the heart once formed.

In a search of Drosophila databases for ESTs that displayed expression profiles resembling those described above for Prc (in BDGP, 3000 individual cDNAs screened by in situ hybridization, 19 expressed in the dorsal vessel), three such ESTs were noticed that labeled pericardial cells and oenocytes, another site of Prc expression. For reasons mentioned in the Materials and Methods, one of them was not considered as relevant. The two other ESTs identified two overlapping DS phages on high density filters of P1 clones that allowed the cytological mapping of the gene to 68 E2-E3 (Fig. 4A).

Screening of a cDNA library (Brown and Kafatos, 1988) with the two ESTs led to the recovery of an almost full length cDNA – prcV2 (5.6 kb) – that hybridized to a single 5.8 kb transcript on northern blots (data not shown), appeared in 8- to 12-hour-old embryos and increased in abundance in older embryos. No transcript was detected in preblastoderm stage embryos, thus indicating strict zygotic expression of the gene.

Sequence analysis of the genomic region upstream of the first exon (Fig. 4A) identified several putative promoter sites and a TATA box that positioned the transcriptional start in that region. This was further supported, as will be discussed below (see also Fig. 5E,F), by the observation that the prc4.5 fragment, when inserted upstream of a lacZ reporter gene, is able to direct β-gal expression in a pattern superimposable on that described for the staining with the anti-Prc antibody.

The prcV2 cDNA contains a long open reading frame of 1729 amino acids (Fig. 4B) with a calculated molecular mass of 165 kDa, consistent with size estimates from western blots revealed by the anti-Prc antibody (Zaffran et al., 1995).

The presumptive initiation codon has been ascribed to the first in-frame ATG (Fig. 4A) that was preceded by an in-frame TAA stop codon. The initiating methionine is followed by a sequence (20 amino acids) containing structural features characteristic for a signal sequence (Fig. 4B) (von Heijne, 1986). No other long hydrophobic regions indicative of transmembrane spanning segments were found, suggesting that this protein is a secreted protein.

The sequence of a particular stretch of eight amino acids in the C-terminal region was identical to that in a peptide obtained by trypsin-mediated cleavage of the EC11-immunoprecipitated and purified Prc antigen (see Materials and Methods) (Fig. 4B). Prc is the unique protein in Drosophila that contains this motif (Blast analysis, pattern search), thus confirming that the cloned cDNA probably codes a protein corresponding to the antigen recognized by anti-Prc (EC11).

An ARG-GLY-ASP (RGD) sequence, a well known potential cell attachment site mostly found in extracellular matrix components (for a review, see Hynes, 1992), is present within the extreme C-terminal moiety. In addition, an ARG-GLU-LYS-ARG tetrapeptide corresponding to the Furin consensus sequence for efficient cleavage of its substrates (Molloy et al., 1999) resides in the N-terminal extremity of Prc. Furin is a serine proteinase that can regulate the composition and, thereby, the function of extracellular matrix by processing matrix components and/or by activating metalloproteinases (Molloy et al., 1999).

Several characteristics raised the possibility that Prc could be a type IV collagen-like protein. Similar to type IV collagens, Prc has a high content of Gly (29%), Gln (13%) and Pro (10%) residues, its transcript is large in size (5.8 kb) and it is organized in three characteristic domains that include a leader peptide and short N-terminal non-collagenous segment, a long collagen-like domain and a C-terminal non-collagenous domain (NCI) (Vuorio and de Crombrugghe, 1990). The amino acid sequence of the collagen-like domain displayed high scores of homology (about 35% identity) with type IV collagen α-chains from various species, including Drosophila, Ascaris, C. elegans and human. The typical (Gly-X-Y)_n triplet repeat in which X and Y can be any amino acid but X is often proline, and which is responsible for the triple helical structure of collagens, is present in that domain. The length of the repeat is especially short as n never exceeds five and interruptions are very frequent. One function of these discontinuities observed in other collagens, such as type IV (Vuorio and de Crombrugghe, 1990) or type XVIII (Oh et al., 1994) collagens, is to provide flexibility between triple helical regions.

In this same domain, the (Gly-X-Y)_n triplet repeat is preceded by another repeat of the type (AGQPGYGXQZGIGGQTG)_n, where n=26. Interruptions that varied in composition and length could be noticed flanking either side of 14 consecutive strictly identical repeats. The sequence of the repeat is unique in Drosophila and within the whole phylum.

Search for predicted secondary structures with the aid of a PSIPRED Program (Jones, 1999) revealed no striking differences between the Prc collagen-like domain and those of Drosophila or human type IV collagens, suggesting that, in spite of its atypical repeats, Prc may have conserved the potential for a triple helical structure.

Only two putative N-linked glycosylation sites were detected, indicating that Prc is a poorly glycosylated protein as is the case for other collagens and as was previously inferred from in vitro deglycosylation experiments that did not modify the electrophoretic migration of the protein (Zaffran et al., 1995).

The short N-terminal and long C-terminal domains displayed no homologies with any known proteins, including the NCI domain of type IV collagens.

Prc may, therefore, constitute a new extracellular matrix component possessing characteristic hallmarks of basement membrane proteins.

The expression of the prc transcript is shown in Fig. 5A,B. It specifically appears at stage 13, at the onset of dorsal closure, in the two rows of pericardial cells that are attached to the surface of the cardial cells (Fig. 5A). In the mature embryonic heart, the prc transcript is very abundant in the pericardial cells and in the ring gland (Fig. 5B). prc is also expressed in oenocytes and in small cells in the anterior part of the embryo that seem to arise from the ring gland and whose identity has not been determined (Fig. 5B). This pattern of expression is superimposable on that of the protein. As depicted in Fig. 5C,D,L, the anti-Prc antibody detects an epitope expressed in the pericardial cells, the ring gland and in the oenocytes as well, thus confirming that the EST probably encodes the protein identified by EC11.

In a transformed line of flies (prclacZ) carrying the prc4.5 genomic fragment inserted upstream of the lacZ reporter gene (see Fig. 4A), lacZ activity (Fig. 5E,F) was detected in the same cells as that of the prc mRNA, indicating that prc4.5 contains all the regulatory elements necessary for prc expression.

Co-expression of Odd and Prc is observed in all the Odd-pericardial cells (Ward and Skeath, 2000). These Odd-expressing pericardial cells are large cells particularly easy to recognize in third instar larvae and are considered to be ‘classical’ pericardial cells (Rizki, 1978; Ward and Skeath, 2000) (Fig. 5G). These cells have been shown to express the Mab3 epitope (Yarnitsky and Volk, 1995; Ward and Skeath, 2000), whose expression pattern is identical to that of EC11, suggesting that the two epitopes may be shared by the Prc protein, although no attempt has been made to formally demonstrate this point. Prc is also expressed in Tin-positive pericardial cells (Jagla et al., 1997) in the heart region of the dorsal vessel (Fig. 5H) that do not express Odd (compare Fig. 5G with 5H). It has been verified that the absence of anterior signals did not result from sectioning through different planes along the anteroposterior axis. prc could also be detected, although more weakly, around the cells in seven segmentally repeated clusters that were cardioblasts, as judged from their staining with anti-Mef2 (data not shown). They correspond to the seven-up (svp)-expressing cardioblasts (Gajewski et al., 2000; Ward and Skeath, 2000; Lo and Frasch, 2001) (Fig. 5I,J; see also Fig. 2), which are the precursors of the larval ostiae (Molina and Cripps, 2001). At the anterior end of the aorta, Prc is expressed in the ring gland while Odd is expressed in the lymph glands (Fig. 5K).

Finally, as shown in Fig. 5M, when the prc cDNA was driven under UAS control in a distinct pattern, such as stripes with engrailed-GAL4, EC11 immunoreactivity detected an epitope ectopically expressed in stripes in addition to its endogenous pattern, exemplified by the labeling of oenocytes. These observations strongly support that fact that the protein encoded by the prcV2 cDNA is identical to the antigen recognized by the EC11 antibody.

Further characterization of Prc function required a detailed description of a mutant phenotype. Df(3L)vin6 (Akam et al., 1978) uncovers the prc gene and abolishes the expressions of the prc transcript and of the protein (not shown). This deficiency is, however, large and uncovers several other genes. However, available lethal alleles in the region maintained prc expression and no P-element insertion close enough to the prc locus has been identified. To circumvent the absence of mutants in the prc gene, we have chosen to use RNA interference (RNAi) to trigger degradation with double-stranded RNA (dsRNA) of the mRNA bearing the same sequence. This process was first developed for use in invertebrates and later, in vertebrates (for a review, see Baulcomb, 1999). Methods have been devised in Drosophila and in C. elegans that use a heat-inducible promoter (Lam and Thummel, 2000; Tavernarakis et al., 2000) or the GAL4/UAS system to express extended hairpin-loop RNA in a controlled temporal and spatial pattern and flexibly induce inhibition at any time of the life cycle or in a specific tissue (Kennerdell and Carthew, 2000; Martinek and Young, 2000; Piccin et al., 2001).

We have attempted hairpin-loop RNA interference of prc mRNA with a prc- GAL4 driver (see Materials and Methods) aimed at a specific reduction of the function of prc in pericardial cells. The inverted repeat sequence was chosen in the C-terminal domain of Prc in a region that was completely free of the N-terminal repeated motifs (see Fig. 4A) and that shared no homologies with any other Drosophila protein. The length of a single repeat was close to 1 kb as recommended by Fire et al. (Fire et al., 1998).

As shown in Fig. 6B, dsRNA interference of prc mRNA resulted in a significant decrease in protein expression (compare Fig. 6B with 6A). The partially silenced embryos display a clear disorganization of the cardial epithelium during its migration in concert with the dorsal ectoderm (Fig. 6B). In many places, the alignment of the cardial cells is interrupted by holes from which they are absent (compare Fig. 6A with 6B) and cells accumulate in small clusters around the holes along the rows of cardioblasts (Fig. 6B). As shown in Fig. 6C,D, disorganization of the pericardial cells also results in interruptions in the dorsal vessel and, as in the case of the cardioblasts, in repeated clustering of the cells, even though their total number did not seem to be modified. The interference was not total and ∼50-60% of embryos were silenced. Each affected embryo displaying a reduced expression of Prc exhibited all the phenotypes described above with somewhat variable numbers of interruptions and, in some instances, interruptions in only one side of the tube.

The phenotypes obtained by RNA interference of prc were similar to those provoked by a total lack of prc expression and of EC11 immunoreactivity (data not shown) in vin6-deficient embryos, at least with regard to the heart tube morphogenesis (Fig. 6E,F). The cardial epithelium formed properly but fell apart with holes and clustering of cardial cells during its coordinated migration with the dorsal ectoderm. The phenotypes in deficient embryos were not much stronger than in silenced embryos. The incomplete penetrance of the phenotype and the absence of EC11 immunoreactivity could suggest that other molecules may play a redundant role with Prc in the formation of the heart tube. However, the comparable phenotypes strongly support the idea that the inverted repeat inactivation of the prc gene is specific. In vin6-deficient embryos, which are grossly perturbed in their development, only a small percentage of individuals survived to the stage of complete closure of their dorsal ectoderm and presented severe defects in their muscles. Generally, the size and the number of cells were different from those in a wild-type animal. In particular, the number of pericardial cells was reduced and the cells were larger (Fig. 6G,H).

As a consequence, the heart phenotypes of vin6-deficient embryos could only be partially rescued by a MN-prc minigene. Prc expression was restored although to a lesser extent than in a wild-type animal (Fig. 6I,J), probably because of the reduced number of pericardial cells. The cardial epithelium is less frequently disrupted, especially in the places where Prc is expressed (Fig. 6I,J). It is thus reasonable to postulate that the cardial phenotypes obtained both by RNA interference and in the vin6 deficiency are primarily due to a reduced expression of Pericardin. In addition, the recovery (even partial) of EC11 immunoreactivity in rescued embryos reinforces the hypothesis that prc encodes the protein identified by EC11.

A more detailed analysis of the cardial phenotypes in silenced and prc loss-of-function (vin6) embryos is depicted in Fig. 7 to gain a better understanding of the function of Prc in morphogenesis of the heart epithelium. At the onset of dorsal closure (Fig. 7A), the alignment of cardial cells appears to be as regular, as in a wild-type embryo. As dorsal closure progresses, the epithelium becomes disorganized, as if the interaction with dorsal ectoderm has loosened and it eventually collapsed (Fig. 7B). A concomitant disorganization of dorsal muscles but not of the entire somatic mesoderm should be noted. This disorganization might result from an indirect effect of the disruptions in the attachment of the bilateral primordia of the dorsal vessel to dorsal ectodermal cells and does not necessarily imply a function of Prc in muscle morphogenesis. In rare occasions in which the mutant embryos completed the closure of their heart tubes, the correct organization of the cardial cells in two rows was often not respected and, in several positions, the shape of the cells was distorted and their polarity seemed incorrect (Fig. 7D). It should be noted that, in these cases, the cells have detached from the dorsal ectoderm, which could explain the loss of structure.

A novel protein, Pericardin, which is expressed in pericardial cells and is a component of the extracellular matrix, could mediate the crosstalk between the dorsal ectoderm and cardioblasts required to insure their coordinated movement during dorsal closure.

The homology of Prc with type IV collagen α-chains, including those of the two Drosophila collagens encoded by Dcg1 (Cecchini et al., 1987) and viking (Yasothornsrikul et al., 1997) was primarily due to its high content of Gly, Gln and Pro residues. The typical (Gly-X-Y)_n collagen repeat is, however, highly divergent and another atypical repeat is present in several copies. This observation challenged the classification of Prc as a type IV collagen, even though its primary structure could be divided into the three conventional collagen domains, its predicted secondary structure was compatible with a characteristic triple helical folding and it contained only two N-linked glycosylation sites (Vuorio and de Crombrugghe, 1990). We propose, therefore, to call Prc a type IV collagen-like extracellular matrix component.

As no specific mutants for Prc could be generated in spite of several unsuccessful attempts at creating P-element induced alleles, dsRNA interference of Prc expression (Kennerdell and Carthew, 2000) was attempted by using a prc-GAL4 driver that led to hypomorphic phenotypes. Prc expression was efficiently and sufficiently reduced to provoke alterations in the heart epithelium. However, it was not completely abolished and a small amount of protein might have already been synthesized, because of the late activation of UAS-IR prc. In vin6-deficient embryos, which have totally lost Prc function, the alterations in the cardial epithelium were similar to those obtained by RNA interference. In addition, an architecture of the heart tube resembling that in a wild-type embryo could be almost fully rebuilt by expressing a prc minigene in vin6-deficient embryos. All these observations suggest that the lack of Prc synthesis was the primary cause of the defects.

Prc does not participate in the specification of the cardial cells precursors or in the mesenchyme-epithelium transition because it is synthesized after these two events take place. Furthermore, in vin6-deficient and in partially silenced prc RNAi embryos, the cardial epithelium appeared normal at the onset of dorsal closure, thus excluding a role of the protein in the acquisition of cell polarity. In places where Prc concentration had been locally diminished, the interaction between cardial epithelium and dorsal ectoderm might have been interrupted while it remained unchanged elsewhere, causing distorsions during dorsal ectoderm migration that led to breaks in the heart tube. The cardial and pericardial cells no longer affixed to the ectoderm fell apart and formed clusters on either side of the breaks. The function of Prc in maintaining the structural integrity of the heart tube and in coordinating the migration of the dorsal ectoderm and of the cardial epithelium might, however, not be the unique actor in these processes. Because, in the absence of that function, some cardial cells conserve their epithelial polarity and migrate in concert with the dorsal ectoderm, even in vin6-deficient embryos, additional partners whose functions could be partially redundant with Prc are likely to be present.

Drosophila mutants with a dorsal hole that are affected in dorsal closure, are also affected in the migration of the two rows of epithelial cardioblasts and pericardial cells. This has hampered the study, independently, of the two movements, and alteration in heart morphogenesis may be due indirectly to defects in the dorsal ectoderm. Mutants in the puc gene proceed to a complete dorsal closure even though the puc-expressing cells have acquired a more lateral cell fate (for a review, see Noselli, 1998). In puc mutants, we have shown that it was possible to uncouple dorsal closure from cardial epithelium migration: the two rows of cardial epithelial cells and the attached pericardial cells did not join at the dorsal midline and a cardia bifida phenotype was obtained. This defect in heart tube formation, resulting from a change in LE cells identity, may reflect defects in the morphogenetic signals that are normally required for the specific interaction between dorsal ectoderm and cardial epithelium.

The precise nature of these signals has not been elucidated, but extracellular matrix proteins, such as laminin or fibronectin, and their receptors, such as integrins, may be efficient partners in cell-substrate adhesion during dorsal closure. Expression of integrin molecules in flies is suggestive of their role in movements of tissues. For example, in myospheroid (mys) mutant embryos (the mys gene encodes one of the two integrin β subunits), the pericardial cells appear to dissociate, migrate randomly and be sparse (Stark et al., 1997). Likewise, in scab mutant embryos (scab encodes an integrin α subunit, which is expressed at the edge of the epidermis and in cells of the dorsal vessel) mislocalization of the pericardial cells and reduced numbers of these cells have been observed (Stark et al., 1997). Both mutants were identified as the result of a failure in dorsal closure, suggesting that the two integrin subunits are involved in that latter process as well. Laminin could be a ligand for these integrins although its late expression in the cardiac tube precludes a role in the migration of the cardial epithelium (Yarnitzky and Volk, 1995).

Pericardin localized expression on the basal surface of the cardioblasts and around the pericardial cells, in the extracellular matrix and in close proximity to the dorsal ectoderm, as well as the disorganization of the cardial epithelium when Pericardin activity was reduced or abolished, strongly favor the hypothesis that Prc participates in a link between the two layers of cells. Such an interaction seems necessary to maintain the structure of the bilateral heart epithelium and to coordinate its migration in concert with the dorsal ectoderm during dorsal closure. No tools are yet available to examine the role for Prc in the maintenance, once formed, of the structural integrity of the cardiac tube because the phenotypes associated to a loss of Prc function appear during the migration step before the tube is completed.

The early events in heart formation have been extremely well conserved in invertebrates and vertebrates. The heart origin can be traced back to two primordia located on either side of the vertical axis that marks the embryo center. As development progresses, the lateral wings of the precardial splanchnic mesoderm with the underlying endoderm fold inward ventrally to converge at the midline of the embryo and develop the centrally located single primitive heart tube. The yolk sac, which is contiguous with the underlying endoderm, is pulled from its lateral position to a ventral position. Several mutations in vertebrates, caused by a fundamental loss in ventral folding, seem to affect the migration of heart cells to the midline and cause two heart tubes to form resulting in a cardia bifida phenotype. In zebrafish, among the eight cardia bifida mutations that have been reported and grouped in three categories, based on the process they control, the miles apart gene, for example, seems to control the migration of the myocardial cells to the midline but not their differentiation (for a review, see Driever, 2000; Stainier, 2001). Likewise, mouse Mesp1 mutant cells seem to be slower than wild-type cells at leaving the primitive streak and reaching the anterior-lateral regions of the embryo (Saga et al., 1999). Homozygous Gata4-null mice most notably lacked a primitive heart tube. The embryos developed splanchnic mesoderm that differentiated into primitive cardial myocytes. The two promyocardial primordia failed to migrate ventrally but, instead, remained lateral and generated two independent heart tubes. The overlying endoderm and somatic mesoderm never moved ventrally, so that the amnion remained dorsal and did not surround the embryo (Kuo et al., 1997; Molkentin et al., 1997). Wild-type endoderm abrogates the ventral developmental defects associated with GATA4 deficiency, including heart tube formation (Narita et al., 1997). These observations point towards a crucial role for GATA4 in regulating the rostral to caudal and lateral to ventral folding of the embryo that is needed for normal cardiac morphogenesis (Kuo et al., 1997).

All these examples underline the importance of a relationship between endoderm (or ectoderm in invertebrates) and mesoderm in the morphogenesis of the primitive linear heart tube. We propose that extracellular matrix components such as Pericardin serve, at least partially, as partners in the relationship.

We thank the Bloomington Stock Center for stock flies and Dr Ashburner’s laboratory for the DS 00169 P1 clone (BDGP). We thank Drs Kiehart, Piovant, Frasch, Nguyen and Skeath for their gifts of some of the antibodies, and Dr Gallet for his gift of the engrailed-GAL4 line. We are deeply indebted to Jean-Paul Chauvin for his helpful assistance in the electron microscopy experiments, and to Dr Williams for the preparation and sequencing of the EC11 tryptic peptides. We acknowledge S. Long and F. Graziani for their technical assistance in the food preparation for the maintenance of the stock flies. This work was supported by the Centre National de la Recherche Scientifique and by grants from Association Française contre les Myopathies (AFM), Ligue Régionale contre le Cancer (LRCC) and Association pour la Recherche contre le Cancer (ARC), which also supported A. C.