R. Timpl, Max-Planck-Institut für Biochemie, Am Klopferspitz 18A, D-82152 Martinsried, Germany. Tel. + 49-(0)89-8478-2440, Fax: + 49-(0)89-8578-2422.
Domain IV of mouse perlecan, which consists of 14 immunoglobulin superfamily (IG) modules, was prepared from recombinant human cell culture medium in the form of two fragments, IV-1 (IG2–9, 100 kDa) and IV-2 (IG10–15, 66 kDa). Both fragments bound to a heparin column, being eluted at ionic strengths either below (IV-2) or above (IV-1) physiological level, and could thus be readily purified. Electron microscopy demonstrated an elongated shape (20–25 nm), and folding into a native structure was indicated by immunological assay and CD spectroscopy. Solid-phase and surface plasmon resonance assays demonstrated strong binding of fragment IV-1 to fibronectin, nidogen-1, nidogen-2 and the laminin-1–nidogen-1 complex, with Kd values in the range 4–17 n m. The latter binding apparently occurs through nidogen-1, as shown by the formation of ternary complexes. Only moderate binding was observed for fibulin-2 and collagen IV and none for fibulin-1 and BM-40. Fragment IV-2 showed a more restricted pattern of binding, with only weaker binding to fibronectin and fibulin-2. None of these activities could be demonstrated for recombinant fragments corresponding to the N-terminal perlecan domains I to III. This indicates a special role for domain IV in the integration of perlecan into basement membranes and other extracellular structures via protein–protein interactions.
Several large proteoglycans, including perlecan [1, 2], agrin  and bamacan , have been identified as abundant constituents of basement membranes. They are integrated through multiple interactions with other basement membrane proteins and can also bind to cellular receptors such as integrins and α-dystroglycan [3, 5–10]. They also control glomerular filtration, cytokine release and proteolysis, mainly through their heparan sulfate side chains [1, 11–13], and in this way make major contributions to the regulation of angiogenesis, wound healing and tumor invasion. Perlecan was originally isolated from a mouse tumor [14, 15] and from fibroblasts  and shown to consist of a 400-kDa core protein substituted by heparan and/or chondroitin sulfate side chains. A complex structure was also indicated from its cDNA sequence [17–19], which demonstrated the presence of seven distinct extracellular protein modules arranged in five predicted domains referred to as I to V [1, 2, 17, 20].
The structure–function relationships of perlecan have recently been studied using various binding assays. Its glycosaminoglycan side chains were shown to mediate interactions with fibroblast growth factor-2 [12, 13], laminin-1, collagen IV and fibronectin [5, 21]. The perlecan core protein, however, has binding epitopes for the basement membrane proteins nidogen-1, nidogen-2 [5, 22] and fibulin-2 , as well as for Alzheimer’s β-amyloid protein [24, 25], fibronectin , transthyretin  and platelet-derived growth factor . The core protein also seems to be involved in perlecan self-assembly , while data on its role in integrin-mediated cell adhesion remain complex and controversial [6, 9, 30–32].
The large size and the need for denaturing conditions during purification of tissue-derived perlecan  make its core protein unsuitable as a substrate for precise binding and mapping studies. This problem was recently circumvented by the recombinant production of individual perlecan domains in mammalian or insect cells including domains I and II [13, 33, 34], III [31, 32, 35] and V . This has facilitated the location of sites for glycosaminoglycan attachment in domain I [34, 36] and a high-affinity binding site for platelet-derived growth factor in domain III-2 . Domain V showed a complex binding repertoire, binding to β1 integrins, heparin, nidogen-1 and fibulin-2, and could also be modified by glycosaminoglycans .
In the present study, we completed our previous analysis of the structure and function of perlecan domains [9, 32–35] by producing domain IV as two recombinant fragments. Domain IV of mouse perlecan  consists of a tandem array of 14 modules of the immunoglobulin (IG) superfamily, most closely related to the C2-type [37, 38]. The N-terminal fragment IV-1 showed high-affinity binding to the two known nidogen isoforms, laminin-1–nidogen-1 complex (LN) and fibronectin, and binding to heparin at physiological ionic strength. The C-terminal fragment IV-2 shared only a few of these binding properties with a lower affinity. Thus domain IV of perlecan, like domain V, seems to make major contributions to the functional repertoire of this proteoglycan.
Materials and methods
Sources of proteins
Perlecan , laminin-1 and the LN complex  were purified from the mouse Engelbreth–Holm–Swarm tumor. Mouse nidogen-1 , human nidogen-1  and nidogen-2 , mouse fibulin-1C  and fibulin-2  and human BM-40  were obtained as recombinant products from mammalian cells. Human plasma fibronectin was of commercial origin [Behringwerke , Marsburg, Germany]. Recombinant perlecan fragments IA and II [33, 34], III-1 to III-3 [32, 35] and V  were prepared as described. The proteolytic fragments of mouse laminin-1 were prepared by standard protocols [39, 44, 45]. Treatment of recombinant proteins with 6 m guanidine or by reduction and alkylation in 6 m guanidine followed previous procedures .
Construction of expression vectors
Mouse perlecan cDNA clone 12  was used to make a construct encoding the C-terminal part of domain IV (amino acid residues 2435–2982). After digestion of clone 12 with BamHI, the isolated and religated 5.4-kb fragment, which now contained a unique PstI site, was digested with PstI (position 9511) and XhoI (MCS) and ligated with a PstI/XhoI adapter (5′-GGTGCCAGAGCGATAGC and 5′-TCGAGCTATCGCTCTGGCACCTGCA) to introduce an artificial stop codon (in bold). Restriction with BsrGI (position 8876) and XhoI released a 652-bp fragment which was purified and inserted into BsrGI/XhoI digested original clone 12. A 1.62-kb fragment was released from this product by digestion with ApaI (positions 7876 and 7905) and XhoI. This fragment was inserted into an NheI/XhoI-digested pCEP-Pu/BMs vector  using a second adapter (5′-CTAGCTACGGTGTCTGTGCTCCCCGAGGGCC and 5′-CTCGGGGAGCACAGACACCGTAG) to connect the NheI and ApaI sites. This yielded the episomal transfection construct pCEP/PG IV-2, which encoded the immunoglobulin-like repeats IG10 to IG15  and two additional amino acid residues of domain V at the C-terminus in-frame with the upstream BM-40 signal peptide .
The cDNA sequence encoding amino acid residues 1676–2434 of mouse perlecan domain IV was prepared using cDNA clones BPG7 and clone 72 [17, 48]. Two overlapping subfragments were generated by PCR using Vent DNA polymerase (New England Biolabs) according to the supplier’s instructions. Primer 1 (5′-GGTCTAGAGCTAGCTCTGGAGGTTCAGATCCATC), which introduced a novel NheI site at the 5′ end, and primer 2 (5′-GCCTGTTGTCCCGGCTGCACGG) were used to amplify a 5′ fragment (635 bp) from clone 72. The 3′ fragment (1714 bp) was generated from BPG 7 using primer 3 (5′-CCCTCCACAGCTCACCGTGCAGCC) and primer 4 (5′-CTCGGGGAGCACAGACACCGTAG). A mixture of these two subfragments was denatured, annealed and used as template to obtain a full-length (2.31-kb) cDNA by 10 cycles of PCR without terminal primers. After addition of primer 1 and primer 4, the full-length fragment was amplified by another 25 cycles of PCR, purified (Qiagen , GmbH, Germany) and digested with ApaI and NheI. The digested fragment was isolated and inserted into NheI/XhoI sites of pCEP-Pu/BMs, using a third adapter (5′-CCCTTAGC and 5′-TCGAGCTAAGGGGGCC) which provided a stop codon (bold letters) and connected the ApaI and XhoI sites. The resulting product, pCEP/PG IV-1, encoded immunoglobulin-like repeats IG2 to IG9 downstream of the BM-40 signal peptide.
The correct in-frame insertion of these products into the pCEP-Pu/BM vector and the correct sequence of PCR-generated fragments was confirmed by cycle sequencing on a 373A DNA sequencer (Applied Biosystems). This led to a correction of the sequence of module IG9 by insertion of a G 3′ to nucleotide position 7782 and deletion of T at position 7814, which changed the amino acid sequence at positions 2402–2412 from APGPATMEPTA  to GTRPSNHGAYR, the latter being almost identical with the corresponding human perlecan IG16 sequence [18, 19]. In addition, a CG to GC conversion at nucleotide positions 9154/9155 in IV-2 led to a Thr to Ser change.
Cell transfection and medium collection
Each of the two expression vectors were used to transfect  human embryonic kidney cells 293-EBNA (Invitrogen). Transfected cells were selected with puromycin as previously described [22, 46]. Serum-free culture medium was collected for 1–2 days from confluent cultures for protein purification, and 1 m mN-ethylmaleimide, 1 m m phenylmethanesulfonyl fluoride and 5 m m EDTA were immediately added to the conditioned medium containing fragment IV-1 to prevent endogenous proteolysis. The same protease inhibitors were used at half the concentrations for the first chromatographic step during purification of both fragments.
Purification of recombinant fragments IV-1 and IV-2
Medium (about 0.5 L) containing fragment IV-1 was dialyzed against 0.02 m Tris/HCl, pH 7.4, containing 0.15 m NaCl and passed at 4 °C over a heparin–Sepharose column (2.5 × 20 cm; Pharmacia) equilibrated in the same buffer. Bound proteins were eluted with a linear gradient of 0.15–1.15 m NaCl (300 mL). The fragment was eluted in the range 0.3–0.8 m NaCl and was precipitated with 80% saturated ammonium sulfate overnight at 4 °C. The precipitate was dissolved in 0.02 m Tris/HCl, pH 7.4, containing 0.15 m NaCl and passed in the same buffer over a Superose 12 column (HR 16/50; Pharmacia). Medium containing fragment IV-2 (1 L) was dialyzed against 0.05 m Tris/HCl, pH 8.6, and passed over a DEAE-cellulose column (2.5 cm × 20 cm) which was eluted with a linear 0–0.6 m NaCl gradient (600 mL). Alternatively, chromatography on heparin–Sepharose in 0.02 m Tris/HCl, pH 7.4, and elution with a 0–0.6 m NaCl (400 mL) was used. Final purification was achieved on a Superose 12 column equilibrated in 0.2 m ammonium acetate, pH 6.8.
Solid-phase binding assays with one ligand immobilized to plastic wells (10 µg·mL−1; 50 µL) were carried out as previously described . Serial dilutions of soluble ligands (50 µL·well−1) in 0.05 m Tris/HCl, pH 7.4, containing 0.15 m NaCl and 1% BSA were incubated for 5 h at room temperature. Binding was detected with specific rabbit antisera, which were diluted to yield an A490 of 1.5–2.2 in a regular ELISA. Binding assays with fibulins were performed in the presence of 2 m m CaCl2[23, 42]. The formation of ternary complexes mediated by nidogens was determined as previously described [40, 41]. Analytical affinity chromatography was performed on a 1-mL heparin HiTrap column (Pharmacia) which was equilibrated in 0.05 m Tris/HCl, pH 7.4, and eluted with a linear 0–0.6 m NaCl gradient.
Surface plasmon resonance assays were performed with BIAcore equipment (BIAcore AB, Uppsala, Sweden) with various ligands covalently immobilized on to a CM-5 sensor chip . Binding assays were performed in neutral buffer under conditions that avoid mass transport problems [28, 51]. Kinetic constants were calculated by non-linear fitting of the association and dissociation curves according to the 1 : 1 model, following the manufacturer’s instructions (BIAevaluation software version 3.0).
Analytical and miscellaneous methods
Amino acid and hexosamine compositions were determined on a Biotronic LC3000 analyzer after hydrolysis (16 h, 110 °C) of the samples with 6 m or 3 m HCl, respectively. Edman degradation of electroblotted protein bands was performed in a Procise sequencer (Applied Biosystems) following the manufacturer’s instructions. Electrophoresis in SDS/5–22% polyacrylamide gels followed standard protocols. CD spectra were recorded and evaluated as previously described . Rotary shadowing of proteins and electron microscopy  and immunological assays  followed standard protocols.
Recombinant production of perlecan domain IV
Episomal expression vectors were used to prepare domain IV as two distinct fragments containing either IG modules 2–9 (IV-1) or 10–15 (IV-2) with their sequence borders identified in Table 1. Production of both recombinant fragments could be demonstrated by electrophoresis of serum-free culture medium obtained from the transfected embryonic kidney cells and both fragments could be purified in two chromatographic steps (see Materials and Methods) in reasonable yields (1–7 mg·L−1 of medium). Protease inhibitors were required after medium collection or at the first chromatographic step, however, in order to avoid extensive endogenous proteolysis, particularly in the case of fragment IV-1. Possible modification by glycosaminoglycans was examined by carbazol assay [9, 34] as well as by immunoblots with specific antisera (see below). These analyses failed to demonstrate any significant substitution in fragments IV-1 and IV-2.
Table 1. Size and other properties of recombinant mouse perlecan fragments IV-1 and IV-2. The molecular mass calculated from the sequence is compared with that found by electrophoresis. The sequence positions are taken from Noonan et al. . In the N-terminal sequences, the starting APLA motif derived from the BM-40 signal peptide is omitted. Hexosamine contents are expressed as residues per molecule and shown as mean ± SD from three or four determinations.
IG2 to IG9
IG10 to IG15
3.9 ± 0.8
3.2 ± 0.6
2.9 ± 0.8
The purified fragments moved mainly as single bands on electrophoresis ( Fig. 1) and showed a single N-terminal sequence ( Table 1) corresponding to that expected from the expression vector after signal peptide cleavage. Reduction of disulfide bonds decreased the electrophoretic mobility to an apparent molecular mass of 100 kDa for fragment IV-1 and 66 kDa for fragment IV-2 ( Fig. 1), suggesting the opening of internal disulfide bonds. As the calculated molecular masses were smaller ( Table 1), this indicated post-translational modification of the fragments, as was in fact shown by hexosamine analysis ( Table 1). The values are consistent with full occupation of the single N-linked acceptor site in fragment IV-2, while fragment IV-1 showed only partial occupation of its three potential acceptor sites.
Structural and immunochemical characterization
Electron microscopy of rotary shadowed specimens revealed a limited polymorphism in the shape of the two fragments ( Fig. 2). The most abundant particles were elongated with a length of 20–25 nm, occasionally giving the impression of a row of small globular domains. A few other particles, however, showed a compact globular structure. It is possible that this variability is an artefact of adsorption to the mica discs. Far-UV CD spectroscopy revealed a distinct ellipticity profile with a minimum at 217 nm and a maximum at 202 nm, as shown for fragment IV-2 in Fig. 3. Secondary-structure calculations indicated only a small amount of α helix (4%) but about 45% β structure, which is consistent with the known folding patterns of IG modules [37, 38].
Rabbit antisera were raised against each fragment and showed a high titer in ELISA and radioimmunoassays (about 1 : 104) against the fragment used for immunization and a slightly weaker reaction with tissue-derived perlecan. Cross-reaction of the antiserum against IV-1 with fragment IV-2 was negligible and vice versa. Sensitive and specific radioimmunoinhibition assays were established for each fragment, as shown for IV-1 ( Fig. 4), and were used for immunochemical comparisons. The activity of perlecan was twofold lower than that of fragment IV-1 and a similar reduction in affinity was observed after treatment of IV-1 with guanidine. Complete reduction and alkylation of disulfide bonds in fragment IV-1 almost completely abolished its inhibitory capacity, indicating that the immunological epitopes are conformation-dependent. No inhibition was observed with fragment IV-2 or any other recombinant fragment of perlecan. When fragment IV-2 was used in place of fragment IV-1 with the antiserum against fragment IV-2, very similar inhibition profiles were observed except that fragment IV-2 and not IV-1 was now the strongest inhibitor (data not shown).
Binding to heparin
Affinity chromatography of culture medium on heparin–Sepharose indicated that both fragments IV-1 and IV-2 have the capacity to bind but with different strengths (see Materials and Methods). This binding was examined on an analytical scale using purified fragments on a heparin HiTrap column equilibrated at low ionic strength (0.05 m Tris/HCl, pH 7.4). Both fragments bound to the column (> 95%) and could be eluted in narrow peaks by the 0–0.6 m NaCl gradient. Maximal elution was achieved at 0.18 ± 0.01 m NaCl for fragment IV-1 and at 0.11 ± 0.01 m NaCl for fragment IV-2. Biotinylated heparin–albumin was used in a second solid-phase binding assay at physiological ionic strength ( Fig. 5). This substrate showed distinct binding to the heparin-binding laminin-1 fragment E3 (half-maximal at 4 n m) and a 10-fold lower affinity for perlecan fragment IV-1. No binding was detected with fragment IV-2 or perlecan. The latter observation could either indicate inactivation of the binding site during purification under denaturing conditions  or reflect electrostatic repulsion during the binding reaction due to the heparan sulfate chains present on perlecan.
Binding to basement membrane proteins in solid-phase assays
Previous studies have shown that perlecan binds to laminin-1, nidogens, collagen IV, fibronectin and fibulin-2, through either its core protein or heparan sulfate chains [5, 22, 23]. We have now used the same assays to examine the potential binding properties of fragments IV-1 and IV-2, which were used as immobilized ( Table 2) or soluble ligands in solid-phase assays. Several of the ligands showed typical binding profiles, as illustrated for nidogen-1 and nidogen-2 in Fig. 6, and the concentrations of soluble ligands required for half-maximal binding are shown in Table 2. Nidogen-1, nidogen-2 and LN complex bound strongly to immobilized fragment IV-1, a more moderate binding was observed with fibulin-2 and fibronectin, and no binding was seen with fibulin-1C and BM-40. When the same tests were carried out with immobilized fragment IV-2, only fibulin-2 showed significant binding activity. Binding tests were also performed with fragments IV-1 and IV-2 as soluble ligands (data not shown). This confirmed the strong binding of nidogens (0.5–2 n m) and the LN complex (8 n m) to fragment IV-1 but failed to show any binding activity for fragment IV-2. To demonstrate the specificity of the observed interactions, we also used the N-terminal recombinant perlecan fragments IA, II, III-1, III-2 and III-3, described previously, as immobilized ligands. No binding was observed with soluble nidogen-1, nidogen-2, LN complex, fibronectin and fibulin-2 (data not shown), with the exception of a moderate binding of fragment IA to the LN complex. This is presumably mediated by its heparan sulfate chains, which can bind to the fragment E3 structure of laminin-1 [5, 21].
Table 2. Binding of soluble extracellular matrix ligands to immobilized perlecan fragments IV-1 and IV-2 in solid-phase assays. Values are based on two to five determinations and are shown with their error range. LN refers to the complex between laminin-1 and nidogen-1. An asterisk denotes complete absence of binding at 300 n m.
Concentration (n m) required for half-maximal binding to
3.4 ± 2
0.8 ± 0.2
0.3 ± 0.1
0.4 ± 0.2
9 ± 6
18 ± 10
12 ± 6
As the binding activities of fragment IV-1 for nidogen-1, nidogen-2 and the LN complex were similar in the solid-phase and surface plasmon resonance assays (see below), this suggested that these interactions may all occur through closely related binding epitopes. This was examined for nidogen-1 and nidogen-2 in a competitive binding assay, where the binding of one isoform was measured in the presence of a fixed amount of the second isoform. This clearly demonstrated that the two nidogens competed for binding to fragment IV-1 ( Fig. 6A) and perlecan ( Fig. 6B). Nidogen-2 appeared to be the stronger competitor as it produced a larger reduction in binding at the point where both nidogens were at equimolar concentration (33 n m).
The lack of binding of fragment IV-1 to laminin-1 ( Table 2, Fig. 7A) indicated that its strong interaction with the LN complex was mediated through its nidogen-1 moiety. This possibility was examined by ternary complex formation, which demonstrated that addition of an excess of nidogen-1 to immobilized laminin-1 promoted significant binding of soluble fragment IV-1 and perlecan. The same effect was achieved with nidogen-2 ( Fig. 7), strongly supporting the possibility that nidogens can act as a bridge between laminin-1 and the perlecan fragment IV-1 structure. However, we cannot exclude the possibility that additional binding sites that may exist on laminin-1 itself have been inactivated during its dissociation from nidogen-1 in 2 m guanidine. Yet several proteolytic laminin-1 fragments (E1XNd, P1, E3, E4, E80, which cover about 80% of its structure  and were not exposed to guanidine, failed to demonstrate significant binding (data not shown).
Analysis of binding affinities by surface plasmon resonance assays
A kinetic assay with BIAcore equipment was used to confirm the solid-phase binding data and to estimate equilibrium dissociation constants. Typical association and dissociation curves could be obtained for several extracellular ligands with soluble fragments IV-1 and IV-2 (not shown) demonstrating variable levels of binding ( Table 3). For fragment IV-1, high-affinity binding (Kd = 8–14 n m) could be detected with the two nidogen isoforms, the LN complex and fibronectin, while lower affinities were found for fibulin-2 and collagen IV. Fragment IV-2 showed a more restricted binding repertoire, binding only to fibronectin, fibulin-2 and the LN complex with Kd values in the range 30–300 n m. No binding of fragment IV-2 was observed for nidogens and collagen IV. Neither of the perlecan fragments bound to fibulin-1C or BM-40.
Table 3. Binding of perlecan fragments IV-1 and IV-2 to various extracellular matrix ligands in surface plasmon resonance assays. Laminin-1 was used as a complex with nidogen-1 (LN) and nidogen-1 was biotinylated and coupled to a streptavidin chip. Binding assays to BM-40, nidogen-1 and fibulin-2 were carried out in the presence of 2 m m CaCl2. Most binding data are mean values of three or four independent experiments and shown by the error range for the Kd values.
Soluble ligand (µm)
ka ( m–1·s–1)
Kd (n m)
No or only questionable binding with fibulin-1C and BM-40.
No or only questionable binding with LN complex, nidogens, fibulin-1C and BM-40.
Surface plasmon resonance assays were also performed with immobilized perlecan fragments, to check the reproducibility of the above data and to examine some additional ligands ( Table 3). Very similar Kd values (4–17 n m) were obtained for the interactions of fragment IV-1 with nidogen-1, LN complex and fibronectin, while binding to fibulin-2 was about eightfold weaker. The binding affinity of fragment IV-2 for fibronectin and fibulin-2 also showed about a fourfold difference. This could reflect the fact that fibronectin and fibulin-2 are potentially bivalent ligands which could undergo co-operative binding and hamper a precise analysis by surface plasmon resonance.
Owing to its large size and low solubility , the core protein of tissue-derived perlecan could only be used as an immobilized ligand ( Table 3). It showed distinct binding to nidogen-1 (Kd = 3 n m) and fibulin-2 (Kd = 78 n m), with affinities in reasonable agreement with those observed for fragments IV-1 and IV-2. No reproducible binding could be observed with fibronectin, however. Perlecan domain V is the only recombinant fragment that has previously been examined for nidogen-1 binding and showed a Kd of 56–78 n m. We therefore also tested recombinant fragments II, III-1, III-2 and III-3, which comprise the N-terminal half of the core protein ( Fig. 8) as immobilized ligands, but failed to detect any binding of nidogen-1. This suggests that perlecan domains IV-1 and -V may possess the only binding sites for nidogen-1.
Perlecan is one of the largest extracellular matrix proteins with a complex modular structure and potential binding repertoire for other extracellular ligands ( Fig. 8). The large central domain IV of perlecan was examined in the present study in order to further our understanding of its potential functions. A major binding site was identified for nidogens, which have previously been postulated to mediate a second strong connection between perlecan and laminins independent of the heparan sulfate chains . Domain IV was obtained in the form of two fragments, IV-1 and IV-2, consisting of eight and six contiguous IG modules, respectively ( Fig. 8). They were produced in good yields and could be readily purified. This demonstrated that the fragments represent autonomous folding units, as previously shown for other recombinant fragments comprising the mouse perlecan domains I, II [33, 34], III-1 to III-3 [32, 35] and V , which could all be produced on a similar preparative scale. Preliminary experiments aiming to obtain the entire domain IV as a single recombinant piece have so far failed, however, because of low production and a low resistance to endogenous proteolysis (M. Hopf and R. Timpl, unpublished observations).
Fragments IV-1 and IV-2 were modified by O- and/or N-glycosylation but not by glycosaminoglycans, the latter apparently being restricted to perlecan fragments I and V [9, 34]. They were both obtained as monomers, despite an extra cysteine in the IG2 and IG13 modules which are not required to stabilize the IG fold. This is of some interest, as these additional cysteines have been postulated to form an intermodular disulfide bond and thus force the entire domain IV into a large globular domain . Electron microscopy of fragments IV-1 and IV-2 showed either an elongated or collapsed globular shape and an analysis of their hydrodynamic parameters is therefore required to determine which form dominates in solution. A similar electron-microscopic polymorphism has also been described for the N-terminal domain of the collagen α3(VI) chain, which consists of eight globular VWA modules . Such observations could therefore be a frequent phenomenon for large tandem arrays of globular modules. The proper folding of the IG modules in the recombinant fragments IV-1 and IV-2 was indicated by CD spectroscopy, formation of disulfide bonds and by analysis of conformation-dependent immunological epitopes that are shared by tissue-derived perlecan. The approximately twofold reduction in antibody-binding activity of perlecan is probably caused by its treatment with 6 m guanidine which is required for its extraction and to prevent proteolysis  and could be reproduced by similar treatment of the recombinant fragments.
Previous solid-phase binding assays with perlecan have demonstrated interactions of the LN complex, nidogen-1, nidogen-2 and fibulin-2 with the core protein [5, 22, 23] while binding of collagen IV, laminin-1 and fibronectin was mediated through the heparan sulfate chains [5, 21]. Here we could show that fragment IV-1 binds with high affinity (Kd = 4–14 n m) to the LN complex and both nidogen isoforms, indicating that it provides the major binding epitope measured in the previous interaction studies with perlecan. A distinct binding activity could also be detected for fibronectin in most assays, in agreement with other data showing fibronectin binding to the core protein of perlecan obtained from a fibroblast matrix under non-denaturating conditions . The level of binding to fibulin-2 was more moderate and showed some variability in different assays. Fragment IV-1 was also recently shown to bind to platelet-derived growth factor with a Kd of 34 n m. In contrast, fragment IV-2 was a poor ligand in most of these assays, but may have moderate to weak affinities for fibulin-2, fibronectin and the LN complex. The only other recombinant perlecan fragment with a distinct affinity for nidogen-1 (Kd = 56–78 n m) is domain V . Recombinant fragments comprising domains I, II and III showed no significant binding to nidogens, as shown in the present study.
Fragments IV-1 and IV-2 also bound to heparin, which has previously only been shown for perlecan domain V  as well as the entire core protein . Heparin binding to fragment IV-1 but not IV-2 could be demonstrated at physiological salt concentrations, indicating that this may also occur in vivo. A second binding site for heparin which exists in perlecan fragment V requires a similar salt concentration (0.2 m NaCl) for dissociation . These interactions may reflect binding properties for heparan sulfate proteoglycans, which could be either cellular receptors or other matrix proteins, or may also include perlecan self-assembly  or the formation of internal associations, as discussed for perlecan domain V . A good candidate for heparin binding within fragment IV-1 is the rather basic IG5 module, which has a calculated pI of 10.5, while all other IG modules have a pI in the range 7.2–8.4. Preliminary data with smaller recombinant fragments of the perlecan IV-1 structure support this prediction (M. Hopf & R. Timpl, unpublished observations). Previous studies have also demonstrated heparin inhibition of fibronectin binding to the perlecan core protein, which could not be fully explained . As the binding epitopes for both ligands are localized within the perlecan IV domain, as shown here, we assume that this inhibition may occur through steric hindrance.
Nidogen-1 was previously shown to bind by two different sites to the perlecan core protein . Whether these correspond to the two binding sites of different affinities present on fragments IV-1 and V of perlecan remains to be established. The more recently discovered nidogen-2 binds with similar strength to perlecan and collagen IV, but with about 100-fold lower strength to laminin-1 than nidogen-1 . High-affinity binding of nidogen-1 (Kd = 0.2–0.5 n m) was localized to a single LE module of the laminin γ1 chain  and the binding epitope has been defined by X-ray crystallography and site-directed mutagenesis [56, 57]. Here we demonstrate that nidogen-1 and nidogen-2 bind to either identical or closely related epitopes on the perlecan structure so far not identified, and this binding could provide a strong ternary linkage to the laminin γ1 chain, which is shared by most of the laminin isoforms discovered so far. Antibodies that interrupt the laminin-γ1–nidogen interaction were also shown to interfere with basement assembly and branching morphogenesis in embryonic organ culture [58, 59]. Thus it will be of interest to see whether antibodies that block nidogen binding to perlecan domain IV produce similar phenotypes.
Perlecan domain IV apparently has a more complex binding repertoire than any of the other perlecan domains ( Fig. 8). In this context, it is of interest that human perlecan has an insertion of seven IG modules between the position of two exons encoding the IG5 module of mouse perlecan, presumably as the result of alternative mRNA splicing [18, 19, 60]. This might either increase binding complexity or modulate binding activities present in the shorter variant. Evidence exists for a similar variant in mouse perlecan . It also now seems feasible to locate the binding sites for heparin and nidogens to individual IG modules using the technologies described here, possibly paving the way for a precise epitope analysis as previously reported for the laminin γ1 chain [56, 57].
We thank Mrs Mischa Reiter and Mrs Hanna Wiedemann for excellent technical assistance and Dr Karlheinz Mann for sequencing. The study was supported by a grant from the Deutsche Forschungsgemeinschaft [SFB 266].