R. Timpl, Max-Planck-Institut für Biochemie, Am Klopferspitz 18a, D-82152 Martinsried, Germany. Fax: 49 89 8578 2422, Tel.: + 49 89 8578 2440, E-mail: firstname.lastname@example.org
The C-terminal domain V of the basement membrane proteoglycan perlecan was previously shown to play a major role in extracellular matrix and cell interactions. A homologous sequence of 708 amino-acid residues from Drosophila has now been shown to be 33% identical to mouse perlecan domain V. It consists of three laminin G-type (LG) and epidermal growth factor-like (EG) modules but lacks the EG3 module and a link region found in mammalian perlecans. Recombinant production of Drosophila perlecan domain V in mammalian cells yielded a 100-kDa protein which was folded into a linear array of three globular LG domains. Unlike the mouse counterpart, domain V from Drosophila was not modified by glycosaminoglycans and endogenous proteolysis, due to the absence of the link region. It showed moderate affinities for heparin and sulfatides but did not bind to chick α-dystroglycan or to various mammalian basement membrane proteins. A single RGD sequence in LG3 of Drosophila domain V was also incapable of mediating cell adhesion. Production of a proteoglycan form of perlecan (≈ 450 kDa) in one Drosophila cell line could be demonstrated by immunoblotting with antibodies against Drosophila domain V. A strong expression was also found by in situ hybridization and immunohistology at various stages of embryonic development and expression was localized to several basement membrane zones. This indicates, as for mammalian species, a distinct role of perlecan during Drosophila development.
Several proteins of the extracellular matrix have been shown to mediate cellular interactions which are instrumental for the survival and function of epithelial and other cells in both vertebrates and invertebrates . This is reflected in the sharing of important components involved in these processes across distant phyla. In Drosophila melanogaster such components include two laminin isoforms, collagen IV and integrin receptors [2–4]. Unique components include the cell-adhesive tiggrin and the proteoglycan papilin, and a major mammalian basement membrane proteoglycan, perlecan, has not yet been characterized in Drosophila. An analogue of perlecan has, however, been identified in Caenorhabditis elegans (unc-52) and differs to some extent in the modular structure of the protein core . This perlecan was previously shown to be important for the connection of muscles to the body wall.
Mammalian perlecan consists of a 480-kDa protein core which is substituted by three to four heparan sulfate or chondroitin sulfate chains [6,7]. It is produced during early and late stages of embryonic development in many organs and is localized in basement membranes, vessel walls, cartilage matrix and some other extracellular spaces . Perlecan-deficiency in mice caused multiple defects and early lethality [9,10]. Major phenotypes were defective basement membranes in the heart muscle and brain, exencephaly, an impaired endochondral ossification and an unstable cartilage matrix. The complex phenotype presumably reflects a diverse array of binding epitopes for several cytokines, laminins, nidogens, fibulins, heparin, sulfatides, α-dystroglycan and β1 integrins located in different domains of the protein core or on the heparan sulfate chains . These binding sites are particularly abundant in domains IV and V of perlecan.
Domain V is located at the C-terminal end of the protein core and consists of a tandem array of three laminin G-type (LG) modules and four epidermal growth factor-like (EG) modules. This domain can bind to the laminin-nidogen complex, fibulin-2, sulfatides, β1 integrins and α-dystroglycan [12–14]. It also binds weakly to heparin and can be modified by glycosaminoglycan substitution and endogenous proteolysis. Mapping with smaller fragments showed the involvement of EG or LG modules in these interactions and the requirement of two LG modules for the efficient binding of sulfatides and α-dystroglycan . Similar binding sites on two LG modules of the laminin α1 and α2 chains have been mapped by site-directed mutagenesis to several nonconsecutive basic residues [13,15]. The data also demonstrated a considerable or partial overlap of the binding epitopes for heparin, sulfatides and α-dystroglycan. The recent elucidation of the crystal structure of the laminin α2LG5 module has now allowed a spatial interpretation of these epitopes and the correct alignment and modelling of other LG modules .
The potential of the Drosophila model for genetic studies led us to search for its perlecan analogue. Based on two cDNA clones, it was feasible to determine the sequence of Drosophila perlecan domain V and to obtain it in recombinant form. This demonstrated a similar modular arrangement to that in mammalian perlecan but only a limited conservation of binding properties for carbohydrate and protein ligands derived from vertebrates. Specific antibodies could be generated and were useful for studying perlecan expression in cell cultures and tissues of Drosophila embryos.
Materials and methods
Sources of proteins and other ligands
Several mammalian protein ligands were obtained from a basement membrane-producing mouse tumour or by recombinant production as described previously . Recombinant mouse fibulin-1C  and human BM-40  were purified as described. Chick α-dystroglycan was a gift from A. Brancaccio (Rome). Integrin αvβ3 was purified by RGD-chromatography from human placenta . Human plasma fibronectin (Behringwerke, Marburg, Germany) and bovine brain sulfatides (Sigma) were purchased.
Sequencing of cDNA clones
Two cDNA clones, GM01490 (2.0 kb) and HL02745 (1.7 kB), which overlapped by 1.5 kb, were used for sequencing and vector construction. These clones were generated by the Berkeley Drosophila Genome Project and were purchased from Genome Systems Inc. They were sequenced completely in both directions by cycle sequencing using BigDye Terminator Cycle Sequencing Ready Reaction Kit (ABI) and sequences were additionally confirmed by sequencing the expression vector.
Construction of expression vector and cell transfection
Amplification of cDNA by PCR with Takara Ex Taq polymerase (Takara Biomedicals) was performed according to the manufacturer's instructions. The 5′ and 3′ primers used with the GM01490 template were 5′-ACCCAAGCTTGCTAGCTGACTTCCCCAATATCCTG-3′ and 5′-GCGGATCCAATCGAGCTGC-3′, respectively, and introduced an NheI site at the 5′ end. The HL02745 template was amplified with the 5′ and 3′ primers 5′-GCGGATCCCGTGGTAGTGC-3′ and 5′-TTTTCCTTTTGCGGCCGCTCAGACGACTGGCGGTT-3′, respectively, which introduced a stop codon and an NotI site at the 3′ end. Both constructs shared a BamHI site and were restricted with BamHI, NheI and NotI then simultaneously ligated into the episomal expression vector pCEP-Pu adjacent to the BM-40 signal peptide . After verification of the sequence, the expression vector was used to transfect human 293-EBNA cells (Invitrogen) and serum-free medium was collected for protein purification according to established methods .
Protein purification and characterization
Serum-free culture medium (1 L) was passed at room temperature over a DEAE cellulose column (2.5 × 20 cm), equilibrated with 0.05 m Tris/HCl, pH 8.6, and eluted with a linear 0–0.6 m NaCl gradient (800 mL). The recombinant protein eluted at 0.13 m NaCl and was subsequently purified on a Superose 12 column (HR16/50, Pharmacia) equilibrated in 0.2 m ammonium acetate, pH 6.8. CD spectroscopy and electron microscopy were performed as described previously .
Ligand binding and cell adhesion assays
Affinity chromatography was performed on a 1-mL heparin HiTrap column (Pharmacia) equilibrated in 0.05 m Tris/HCl, pH 7.4. Solid-phase assays with plastic-immobilized ligands followed previously used methods . In the assays with sulfatides and α-dystroglycan, 1 mm CaCl2 was added to the buffer [13,15]. Cell adhesion assays were performed with Rugli and RN22 cells as described before .
Immunization of rabbits, affinity-purification of antibodies and ELISA titration followed standard protocols . Culture medium was separated by SDS/PAGE prior to immunoblotting .
Samples were hydrolysed (16 h, 110 °C) with 6 m or 3 m HCl for the determination of protein concentrations and hexosamine compositions, respectively, on a Biotronic LC3000 analyser. Edman degradation was carried out on a 473A sequencer (Applied Biosystems). SDS/PAGE followed standard protocols.
Digestions, preparation of embryonic extracts and immunoblotting
Conditioned medium (100 µL) of Drosophila Er1 cells was digested overnight with heparinase or heparitinase (Seikagaku Corporation, Tokyo, Japan) at 1.5 ng·µL−1 according to the manufacturer's instructions. Drosophila embryos (6–24 h) were homogenized in 6 m guanidine HCl, and incubated overnight in the same buffer. After centrifugation, the perlecan protein was still in the pellet, which was then subjected to extraction in protein electrophoresis sample buffer and boiled at 95 °C for 4 min. Samples of conditioned medium of Er1 and Kc1 cells and of boiled pellet were separated under reducing conditions by 6% SDS/PAGE. Following transfer onto nylon membranes, blots were probed with anti-perlecan Ig and detected with HRP conjugated secondary antibodies, followed by ECL chemiluminescence (Amersham).
Immunohistology and in situ hybridization
Drosophila embryos were staged according to Campos-Ortega and Hartenstein  and used for immunohistochemical staining  with the anti-perlecan serum diluted 1 : 3500. An anti-Drosophila laminin α3/5β1γ1 Ig (gift of J. Engel, Biozentrum, Basel, Switzerland) was used at 1 : 3000 dilution. The latter antibody showed the same staining specificity as described previously . In situ hybridization was performed with a digoxygenin-labelled GM01490 probe following standard procedures .
Amino-acid sequence of domain V of Drosophila perlecan
Two overlapping cDNA clones (GM10490, HL02745), together covering about 2.5kb, were used to determine the entire amino-acid sequence of Drosophila perlecan domain V (Fig. 1). The 708-residue sequence had an overall identity of 33% with mouse perlecan domain V (728 residues). Like the mouse sequence, Drosophila domain V consisted of three LG modules (LG1 to LG3), but had only three analogues of EG modules (EG1, EG2 and EG4) interrupting the LG tandems. Module EG3 was replaced in Drosophila by a 30-residue sequence containing seven prolines but no cysteines. A further difference was the lack of a 19-residue link region between EG4 and LG3 which in the mouse contains a single glycosaminoglycan attachment site and a protease-sensitive site . Yet Drosophila domain V has three potential SGE or SGD sequences suitable for glycosaminoglycan attachment and seven NXT/S sites for N-linked glycosylation. It also has a potential cell-adhesive RGD sequence at the beginning of the LG3 module, which is not conserved in mouse perlecan. The 5′ end of clone GM01490 encoded an additional 50 residues with 50% sequence identity to an Ig module from the adjacent domain IV of mouse perlecan.
Characterization of recombinant domain V of Drosophila perlecan
An episomal transfection vector was used to obtain Drosophila perlecan domain V in mammalian cells. This vector encoded the entire domain V but in addition an N-terminal APLA sequence derived from the BM-40 signal peptide  followed by the sequence DFPNILVV from domain IV. Immunoblotting of serum-free culture medium demonstrated a single 100-kDa band (Fig. 2A). Medium containing recombinant domain V of mouse perlecan showed the same band as well as endogenous proteolysis fragments of 70 kDa and 26 kDa and larger diffuse bands representing glycosaminoglycan-substituted forms . This demonstrated lack of these modifications in Drosophila domain V, in agreement with the lack of the link region involved in mouse perlecan.
Recombinant Drosophila perlecan domain V could be obtained in similar yields from mammalian cells (3–4 μg·mL−1) as described previously for mouse perlecan domain V . The purified product shifted from a 100-kDa band to a 90-kDa band when nonreduced, indicating the formation of internal disulfide bridges (Fig. 2B). It showed a single N-terminal APLADFPN sequence, consistent with the vector construct and a high purity. The product contained 15–16 residues of glucosamine and four galactosamines, indicating an incomplete substitution of the N-glycosylation sites and some O-glycosylation. These substitutions apparently account for the higher electrophoretic mobility of domain V when compared with a calculated molecular mass of 78 kDa. Electron microscopy of Drosophila domain V demonstrated predominantly a row of three globular domains separated by small gaps, similar to the mouse domain V (Fig. 3). These globular domains probably represent the LG modules. Correct folding of Drosophila domain V was also indicated by CD spectroscopy, which showed ≈ 50% β structure (data not shown). It was also possible to obtain specific rabbit antibodies against Drosophila domain V (ELISA titre 1 : 20.000) which did not cross-react with mouse domain V, reflecting their low sequence identity.
Binding properties of Drosophila perlecan domain V
Mouse perlecan domain V was previously shown to be a cell-adhesive substrate and to bind to the α-dystroglycan receptor, several extracellular matrix proteins, sulfatides and weakly to heparin [12,14]. A weak affinity for heparin was also found for Drosophila domain V, which could be displaced from a heparin affinity column with 0.15 m NaCl. Domain V of both species also showed a comparable binding to immobilized sulfatides, half maximal at 10 nm(Fig. 4A). However, no binding was observed between Drosophila domain V and chick α-dystroglycan, in contrast with the strong binding of mouse domain V (Fig. 4B). Drosophila domain V also showed no interaction in solid-phase assays with nidogen-1, laminin-1-nidogen complex, fibulin-1, fibulin-2, fibronectin and BM-40 of either mouse or human origin.
Mouse perlecan and its domain V, but not Drosophila domain V, were shown to be an adhesive substrates for rat glioma Rugli cells (Fig. 5). The same pattern was also observed with rat Schwannoma RN22 cells, which are known to bind to the RGD sequences of vitronectin . In solid-phase assays with immobilized human αvβ3 integrin, fibronectin and vitronectin bound strongly, as shown previously , with 0.1–0.2 nm required for half-maximal binding, whereas no interaction was observed with Drosophila domain V up to 1 µm.
Tissue expression of Drosophila perlecan RNA
Using digoxygenin-labelled probes, the spatial expression of Drosophila perlecan was examined. RNA transcripts were first detected in syncytial blastoderm (Fig. 6A), where uniform staining was observed. During later stages of embryogenesis and until stage 14, only low levels of uniform expression were observed (data not shown). Only at stage 15 (Fig. 6B–D) was prominent staining observed in the visceral mesoderm of the gut and in cardiac cells (Fig. 6B,C), and also in the fat body (Fig. 6D). Similar expression patterns were found for Drosophila laminin α3/5 chain . No staining was observed in haemocytes, which usually synthesize many proteins of the Drosophila extracellular matrix. At stage 16, staining in visceral mesodermal and cardial cells still persisted, although the overall levels of ubiquitous expression increased (Fig. 6F). During this stage, particularly strong expression was found in cardiac cells, but not on pericardial cells, where the transcripts seemed to accumulate more towards the midline (Fig. 6E, arrow). During postembryonic development, transcription of perlecan is also readily detectable in imaginal discs. Particularly strong expression was found in eye and antennal discs (Fig. 6G), where groups of cells at the morphogenetic furrow, in the presumptive Oc region and in the PalD region show strong staining (arrows). Strong staining was also observed in the leg discs in parts of concentric rings (Fig. 6H).
Expression of the perlecan protein in Drosophila
To assess the nature and expression of the perlecan protein, the polyclonal antiserum against domain V was used for immunoblotting and on whole mount embryos. Blotting of conditioned medium of Drosophila Kc1 cells showed a single 450 kDa band (Fig. 7, lane 1), while another cell line, Er1, produced a broad smear of lower mobility (Fig. 7, lane 2), suggestive of post-translational modifications. To assess the nature of these modifications, we subjected conditioned medium from Er1 cells to heparinase or heparitinase digestion, followed by immunoblotting. Both treatments changed the diffuse band of higher molecular weight into a distinct band of ≈ 450 kDa (Fig. 7, lanes 4 and 5). This indicated that perlecan is substituted with heparan sulfate side chains, although the degree of modification seems not to be as extensive as that for mouse perlecan [6,7]. A similar broad band could also be detected in SDS extracts of embryonic Drosophila tissue (Fig. 7, lane 6)
The perlecan protein is first detected in tissues at stage 15 in a quite ubiquitous manner, with particular accumulation around the central nervous system (CNS), the visceral mesoderm and the hindgut (Fig. 8A). Within these locations, it appears that perlecan accumulates in the basement membrane structures surrounding the tissues. At stage 16, the pattern remains essentially the same, except that on the dorsal side the cardiac cells show an accumulation of perlecan (Fig. 8B,F). During stage 16, a particular deposition was also observed in the basement membranes covering the channels of the CNS (Fig. 8C, arrows and Fig. 8D), on dorsal median cells (Fig. 8F) and on dorsal muscle attachment sites (Fig. 8E). As the pattern was reminiscent of the expression of Drosophila laminin (composition α3/5β1γ1 ), we compared the two patterns and found an overlapping localization in the basement membranes surrounding the CNS channel basement membrane and in the heart (Fig. 8G vs. Fig. 8H). This suggests that both proteins are integral members of certain basement membrane structures of Drosophila embryos.
The sequence analysis clearly identified Drosophila perlecan domain V as an analogue of the same mouse perlecan structure, with 33% sequence identity and the module arrangement LG1EG1EG2LG2EG4LG3. The mouse EG3 module was replaced in Drosophila by a short proline-rich segment and a linking segment between EG4 and LG3 was also absent. This link contains a glycosaminoglycan attachment site, which is partially occupied in tissue-derived mouse and human perlecan, and its occupation was shown to inhibit cell adhesion and α-dystroglycan binding to domain V [12,14]. The complete sequence of C. elegans perlecan has recently been elucidated ; it demonstrated a similar modular arrangement for domain V as in Drosophila but had a low sequence identity (25%). This sequence also showed a replacement of the EG3 module by a 170-residue segment rich in threonine and glutamic acid and lacked the glycosaminoglycan attachment site of the link region.
A crystal structure has recently been established for the laminin α2LG5 module and showed a β-sandwich providing a hydrophobic core and various loops on the top and bottom of the core responsible for ligand binding . The study also demonstrated ligation of a single calcium ion and the importance of two aspartates in this ligation. These residues are conserved in many LG modules. It was furthermore speculated that calcium ligation plays an essential role in the interaction with the cellular receptor α-dystroglycan. An alignment of the mouse and Drosophila perlecan LG modules showed a good conservation of hydrophobic amino acids in the β strands, certain critical glycines and two cysteines involved in a disulfide bridge (Fig. 9). The conservation of calcium-ligating aspartates (or asparagines) was complete for mouse domain V but only existent for the LG1 module in the Drosophila domain V. This is of particular interest as α-dystroglycan binding to mouse perlecan domain V requires at least two LG modules  and depends on calcium-binding.
Recombinant Drosophila domain V was properly folded, as demonstrated by electron microscopy and a high content of β structure, which is typical for LG and EG modules. It was therefore a useful ligand to compare binding properties established for mouse perlecan domain V with the aim of selecting appropriate residues for site-directed mutagenesis. The studies were restricted to ligands obtained from vertebrate sources and showed a comparable binding of mouse and Drosophila perlecan domain V to heparin but not to several protein ligands. Binding of heparin to mouse domain V involved some of the five arginines in the EG2 module , but only two of them are conserved in Drosophila (Fig. 1). It is therefore possible that basic residues present in the H–I and J–K loops of the LG structure (Fig. 9), which are close in space and also contain critical residues for heparin binding to laminin α1LG4 and α2LG5 modules [13,15], are replacing the EG-binding epitope in Drosophila. The lack of binding of domain V to various mammalian basement membrane proteins, including laminin-1 and nidogen-1, indicates a divergent evolution of potential binding epitopes. Nevertheless, this information will now be useful for designing mutations in nonconserved loop regions (Fig. 9) in order to locate such binding epitopes on mouse perlecan domain V. Lack of conservation of binding epitopes seems, however, not to be always the case as mouse nidogen-1 binds to Drosophila laminin of the chain composition α3/α5β1γ1. This is explained by nearly identical binding epitopes being present in mouse and Drosophilaγ1 chains .
A further distinct binding could be demonstrated between Drosophila domain V and bovine brain sulfatides and was comparable to that reported previously for LG modules of mouse perlecan and laminin α1 and α2 chains [13–15]. Site-directed mutagenesis assigned the laminin binding epitopes to several basic residues in the F–G and H–I loops [13,15,16] and thus to regions which also have a basic character in perlecan domain V of Drosophila although not in that of mouse (Fig. 9). Here, therefore, we should again anticipate an evolutionary change in perlecan binding epitopes. Unlike mouse perlecan and laminins, Drosophila perlecan domain V did not bind to the dystroglycan receptor, which plays a major role in cell–matrix interactions of vertebrates . As discussed above, this could reflect the absence of two calcium-binding sites , or the divergence of dystroglycan receptors in invertebrates.
Drosophila laminin was previously shown to promote attachment, spreading and differentiation of various Drosophila cells mediated by integrins but failed to be a substrate for vertebrate cells [3,4,32]. Here we show that Drosophila perlecan domain V failed to bind two rat cell lines which are known to adhere to mouse perlecan and its domain V through β1 integrins [12,22]. The LG3 module of Drosophila domain V contains a potentially adhesive RGD sequence which is not conserved in mouse (Fig. 1 and Fig. 9). No binding to this RGD was observed in solid-phase assays with purified human αvβ3 integrin which binds avidly to such sequences of vitronectin and fibronectin. This could be explained by the location of the RGD in a short loop between β strands A and B of Drosophila (Fig. 9), a site which is not well accessible in the α2LG5 structure . Confirmation of this possibility will require studies with DrosophilaαPS2βPS integrin that binds readily to RGD sequences in the Drosophila substrates tiggrin, laminin α3/α5β1γ1 and ten-m [33,34].
The data with Drosophila cell cultures and tissue extracts demonstrated a perlecan protein core of at least 450 kDa, which can be substituted by heparan sulfate side chains. This large size agrees with genomic sequence data available from the Berkeley Drosophila Genome Project, which are still incomplete for the Ig modules. As for C. elegans perlecan , they demonstrate a good conservation of domains II–V, whereas domain I is not present and is apparently replaced by an extension of domain II with yet another 19 low-density-lipoprotein receptor type A modules and an Ig module (M. Schneider and S. Baumgartner, unpublished data). As domain I of vertebrates contains three major SGD attachment sequences for heparan sulfate [35,36] and, unlike in vertebrates , no additional substitution could be detected in the recombinant Drosophila domain V, it remains open as to which part of Drosophila perlecan is actually modified by heparan sulfate. A possible candidate is a single SGD sequence found in one IG module of domain IV. There is also another RGD sequence in Drosophila perlecan domain III which needs to be examined for cell adhesion activity.
At least two other heparan sulfate-containing proteins have been described so far in Drosophila, including a syndecan  and a glypican referred to as dally . Both proteins appear to be involved in growth factor signalling, and the heparan sulfate side chains were shown to be particularly important for proper signalling with different growth factors. Likewise, it could be envisaged that Drosophila perlecan, like its vertebrate analogues [39,40], will also interact with the fibroblast growth factor and its receptor signalling pathway, which appear to be conserved in evolution.
A second major aim of this work was to obtain suitable reagents for studying the expression of Drosophila perlecan during embryonic development.
There was an abundant expression of perlecan mRNA and protein in Drosophila embryos and its localization partially overlapped that of laminin α3/α5β1γ1, which can be found integrated in most basement membranes of the embryo. Mouse perlecan was shown previously to bind to laminins either by its heparan sulfate chains or through the protein core domain IV, the latter interaction being mediated by nidogen-1 [11–14,41,42]. It is likely that similar mechanisms exist in Drosophila, thus making the insect basement membranes comparable to those of vertebrates. Other basement membrane components or corresponding cellular receptors exist in Drosophila as well, such as dystroglycan (M. Schneider and S. Baumgartner, unpublished data), nidogen-1 (Drosophila Flybase) and BM-40 (E. Kohfeldt, S. Baumgartner and R. Timpl, unpublished data). At least dystroglycan shares the CNS and cardiac cell basement membrane staining with both laminin and perlecan (M. Schneider and S. Baumgartner, unpublished data). In this context, it is worth noting that the defects seen in perlecan-deficient mice [9,10], including neuronal ectopias, exencephaly and defective basement membranes ensheathing heart muscle, correlates well with the expression of Drosophila perlecan in brain and CNS and in cardiac cells, suggesting a conserved function of perlecan across distant phyla. This hypothesis can be tested once Drosophila perlecan mutants are available for analysis.
We are grateful for the technical assistance of M. Reiter, M. Mogren and S. Da Rocha and for the help of T. Sasaki, A. Brancaccio and K. Mann. This work was supported by a Swedish NFR Grant 409005501-5 to S. Baumgartner.