The Structure, Location, and Function of Perlecan, a Prominent Pericellular Proteoglycan of Fetal, Postnatal, and Mature Hyaline Cartilages*

The aim of this study was to immunolocalize perlecan in human fetal, postnatal, and mature hyaline cartilages and to determine information on the structure and function of chondrocyte perlecan. Perlecan is a prominent component of human fetal (12-14 week) finger, toe, knee, and elbow cartilages; it was localized diffusely in the interterritorial extracellular matrix, densely in the pericellular matrix around chondrocytes, and to small blood vessels in the joint capsules and perichondrium. Aggrecan had a more intense distribution in the marginal regions of the joint rudiments and in para-articular structures. Perlecan also had a strong pericellular localization pattern in postnatal (2-7 month) and mature (55-64 year) femoral cartilages, whereas aggrecan had a prominent extracellular matrix distribution in these tissues. Western blotting identified multiple perlecan core protein species in extracts of the postnatal and mature cartilages, some of which were substituted with heparan sulfate and/or chondroitin sulfate and some were devoid of glycosaminoglycan substitution. Some perlecan core proteins were smaller than intact perlecan, suggesting that proteolytic processing or alternative splicing had occurred. Surface plasmon resonance and quartz crystal microbalance with dissipation experiments demonstrated that chondrocyte perlecan bound fibroblast growth factor (FGF)-1 and -9 less efficiently than endothelial cell perlecan. The latter perlecan supported the proliferation of Baf-32 cells transfected with FGFR3c equally well with FGF-1 and -9, whereas chondrocyte perlecan only supported Baf-32 cell proliferation with FGF-9. The function of perlecan therefore may not be universal but may vary with its cellular origin and presumably its structure.

The HS side chains of domain I of perlecan act as low affinity co-receptors for growth factors, such as FGF-1, -2, -7, and -9, whereas the core protein of perlecan may act as a receptor for FGF-7 (22)(23)(24)(25). These interactions are important for the correct presentation of the FGFs to FGF receptors (FGFRs) and their subsequent oligomerization and activation, which initiates cell signaling and subsequent down-line effects on cell proliferation and differentiation (22)(23)(24)(25)(26)(27). Binding of the FGFs to perlecan also protects them from proteolytic degradation in situ and increases their biological half-life (22,24,28,29). Connective tissue growth factor has also recently been shown to bind to the low density lipoprotein repeats within domain II of perlecan. Connective tissue growth factor modulates bone morphogenetic protein and transforming growth factor-␤ signaling and coordinates chondrogenesis and angiogenesis during skeletal development (30 -34).
The development of perlecan gene "knock-out" mice has illustrated the essential role that perlecan plays in cartilage * This work was supported by National Health and Medical Research Council of Australia Grant 211266, the Shriners of North America, seed grants from the Australian Arthritis Foundation, Rebecca Cooper Medical Research Foundation (to J. M.), and University of New South Wales and Australian Research Council Faculty of Engineering Infrastructure Research Grants LP0455407 and DP0557863 (to J. W.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1 To whom correspondence should be addressed: Level  development and skeletogenesis (35)(36)(37)(38). Homozygous gene knock-out mice, which survive to birth, display severe skeletal defects with short axial and limb bones, cleft palate, striking abnormalities in the growth plates of their long bones, and cardiovascular abnormalities. In normal growth plates, chondrocytes are arranged into zones of resting, proliferative, and hypertrophic cells, and the cellular activities within each of these zones and the transitions between them are regulated by key signaling molecules. In perlecan knock-out mice the proliferating zone is disorganized, biosynthetic activity in the hypertrophic zone enhanced, and ossification below the growth plate is more extensive laterally than longitudinally, resulting in short squat newborn mice (35,39). Mutations in the human perlecan gene (HSPG2) have been identified in dyssegmental dysplasia and the milder Schwartz-Jampel syndrome (chondrodystrophic myotonia) (3, 40 -42). Perlecan has previously been immunolocalized in a range of mouse tissues (43), however, mouse perlecan displays differences in structural organization compared with human perlecan and its localization could vary between the two species. Perlecan has been immunolocalized to the basement membranes of all human tissues tested (1) and human cartilage and chondrosarcoma, which are devoid of basement membranes (2). In addition, perlecan has been immunolocalized in human and bovine nasal cartilage (35,44), vertebral growth plate, and cartilaginous end plates of the ovine intervertebral disc, cartilaginous primordia of the fetal human spine, and ovine articular, growth plate, and meniscal cartilages with aging (45)(46)(47). To date, however, despite its obvious importance to tissue growth and development, perlecan has yet to be studied in detail in human cartilage.
The present study was undertaken to determine the distribution and structure of perlecan in human cartilage during fetal development, in the growing child, and in the mature adult. Unfortunately, insufficient human cartilage was available to isolate sufficient chondrocytes to undertake large scale cultures appropriate for the isolation of human chondrocyte perlecan for functional studies on growth factor binding. As an alternative, perlecan was purified from human endothelial cell and ovine chondrocyte cultures and used to study the interaction with FGF-1 and -9 and its consequences on cell proliferation.

EXPERIMENTAL PROCEDURES
Source of Antibodies-Histochoice fixative was purchased from Amresco (Solon, OH). Monoclonal antibodies to intact HS (mAb 10E4) and to the ⌬-HS stubs (mAb 3G10) generated by the action of heparitinase III were from Seikegaku Corp, Tokyo, Japan. Monoclonal antibody 7B5 to perlecan domain III was from Zymed Laboratories through Invitrogen. Monoclonal antibodies A76 and A74 to perlecan domains I and V were used as described previously (21,(45)(46)(47). Monoclonal antibody to the G 1 domain of aggrecan (clone 969D4D11) was purchased from BIOSOURCE Europe, Nivelle, Belgium. Novex HiMark protein molecular weight standards and other electrophoresis consumables were obtained from Invitrogen.
Preparation of Cartilage for Histology-Six 12-20-week-old human fetuses were obtained at termination of pregnancy following ethical approval by the Human Research Ethics Com-mittee of the Royal North Shore Hospital. Tissue specimens were fixed for 3 days in Histochoice® for perlecan immunolocalization or 10% (v/v) neutral buffered formalin for aggrecan histochemistry and briefly decalcified in 10% formic acid. Postnatal human femoral condylar cartilage was obtained with ethical approval at the time of autopsy from the pathology departments at l'Hôpital Ste-Justine and the Montreal General Hospital. Tissue specimens were fixed in 4% fresh paraformaldehyde in PBS, pH 7.2, overnight at 4°C. The fixed tissues were then dehydrated in graded alcohols and embedded in paraplast wax using standard histology procedures. Four to seven-m sections were cut in the coronal and sagittal vertical planes and mounted on glass slides, de-paraffinized in xylene (4 changes ϫ 2 min), and re-hydrated through graded ethanol washes (100 -70% (v/v)) to water.
Histochemistry-Formalin-fixed tissue sections were routinely stained with toluidine blue using fast green counterstain to visualize proteoglycans, and with hematoxylin and eosin to examine cellular morphology prior to selection of the most appropriate specimens for detailed immunohistochemical localizations.
Immunohistochemistry-Endogenous peroxidase activity was initially blocked by incubating the tissue sections with 3% H 2 O 2 . Histochoice-fixed tissue sections were pretreated with bovine testicular hyaluronidase (1000 units/ml) for 1 h at 37°C in phosphate buffer, pH 5.0, followed by three washes in 20 mM Tris-HCl, pH 7.2, containing 0.15 M NaCl (TBS). Paraformaldehyde-fixed sections were pretreated with chondroitinase ABC (0.25 units/ml) for 1 h at 37°C in 0.1 M Tris-HCl, 0.1 M sodium acetate, pH 7.5. Nonspecific binding was initially blocked by incubating the sections in 20% swine or goat serum for 30 min at room temperature. Incubations with primary antibodies to perlecan domains I (mAb A76), V (mAb A74), or III (mAb 7B5) and aggrecan G 1 domain (mAb 969D411) were performed for 1 h at room temperature or overnight at 4°C. Horseradish peroxidase or alkaline phosphatase-conjugated secondary antibodies were then added for 1 h. Color development was conducted with 0.05% 3,3Јdiaminobenzidene dihydrochloride and 0.03% H 2 O 2 in TBS, or with Nova RED substrates. Control sections were also prepared in which no primary antibody or an irrelevant isotype-matched primary antibody was substituted for the primary antibody of interest. Commercial isotype-matched mouse IgG (DAKO code X931), against Aspergillus niger glucose oxidase (an enzyme that is neither present nor inducible in mammalian tissues) was used for this step.
SDS-PAGE and Immunoblotting-Cartilage was diced into small pieces and extracted with 10 volumes of 4 M guanidinium chloride in the presence of proteinase inhibitors (48). Extracts were dialyzed into 50 mM Tris/HCl, pH 7.5, containing 10 mM CaCl 2 , and 20 l of the dialyzed extract was digested with heparitinase III (1 milliunit/ml) and/or chondroitinase ABC (10 milliunits/ml) and keratanase I (5 milliunits/ml) at 37°C for 3 h. Perlecan core proteins were then analyzed by SDS/PAGE on 4 -12% NuPAGE BisTris gels (Novex) run in MOPS SDS running buffer or 3-8% NuPAGE Tris acetate gels run in Tris acetate buffer following the manufacturer's instructions. Proteins were electroblotted onto nitrocellulose membranes (0.22 m) using NuPAGE transfer buffer supplemented with 10% methanol. MultiMark and HiMark protein molecular weight standards were also electrophoresed for calibration purposes. Monoclonal antibodies 7B5 (1:500 dilution), A76 and A74 (1/400 dilution), and 3G10 (1/500 dilution) were used as primary antibodies. A horseradish peroxidase-conjugated anti-mouse IgG was used as the secondary antibody, visualization of the immune complexes utilized an enhanced chemiluminescence (ECL) detection system (Amersham Biosciences) and subsequent exposure to x-ray film. Alkaline phosphatase-conjugated anti-mouse IgG secondary antibodies were also used with the colorimetric substrates nitro blue tetrazolium/5-bromo-4chloro-3-indolyl phosphate for detection in some cases.
Chondrocyte Culture and Immunopurification of Perlecan from Conditioned Media-Chondrocytes were isolated from ovine articular cartilage by sequential enzymatic digestion using Pronase, collagenase, and DNase-I as described (49), and established in monolayer culture in Dulbecco's modified Eagle's medium/Ham's F-12 media supplemented with 10% fetal calf serum, 2 mM L-glutamine and gentamicin (50 g/ml). The media was changed every 2 days and stored at Ϫ20°C until cell growth had reached confluence, when the cells were subpassaged into fresh 175-cm 2 canted neck flasks and media collected until all flasks had reached confluence. A total of 4 liter of conditioned medium was collected from the 1st and 2nd passage chondrocyte cultures. Perlecan was isolated from the chondrocyte-conditioned medium using a combination of anion exchange and immunoaffinity chromatography (50). Endothelial cell perlecan was also isolated in a similar manner as described previously (21,26,27).
Assessment of Perlecan Interaction with FGF-1 and FGF-9 Using Surface Plasmon Resonance-Immunopurified perlecan samples (100 g/ml) from the ovine chondrocyte and human umbilical artery endothelial cell cultures were biotinylated in 0.1 M NaHCO 3 buffer, pH 8.0, using a 50-fold molar excess of N-hydroxysuccinimido-biotin for 3 h at room temperature. Biotinylation was then terminated by the addition of an excess of 2 M NH 4 Cl solution and free biotin was removed from the sample by dialysis against PBS (26,27). Biotinylated perlecan samples (10 g/ml) in PBS were coupled to streptavidin-derivatized sensor chips at a flow rate of 5 l/min. A similar amount of each purified perlecan was immobilized to the streptavidin chips in each case as determined from the change in response units, which correlates with a change in mass of the BIAcore chip. BIAcore binding experiments with FGF-1 (200 nM) and FGF-9 (200 nM) were conducted in 0.01 M HEPES, 0.15 M NaCl, 5 mM MgCl 2 , 0.01% Tween 20, pH 7.4, at a flow rate of 20 l/min at 25°C using an injection volume of 50 l. The BIAcore kinetic function was used with a programmed dissociation time of 150 s where the perlecan surface was regenerated using a 30-s pulse of 1 M NaCl or heparin (100 g/ml). Sensograms were analyzed using BIAcore 2000 evaluation software 3.0. Sensograms were fitted with separate differential rate equations for the parts of the binding curve representing association and dissociation. Control experiments were also conducted with perlecan samples that had been pre-digested with heparitinase III to remove HS.

Assessment of Perlecan Interaction with FGF-1 and FGF-9
Using Quartz Crystal Microbalance with Dissipation Monitoring (QCM-D)-Perlecan samples (10 g/ml) in PBS were adsorbed onto gold-coated QCM-D quartz crystals (Q-Sense AB, Sweden) at a flow rate of 100 l/min for 10 min then, under static conditions, for 50 min followed by PBS rinses (100 l/min) for 10 min. 0.1% Bovine serum albumin in PBS (100 l/min) was exposed to the surface for 30 min to block nonspecific binding sites before rinsing with PBS (100 l/min) for 10 min. Binding of FGF-1 (100 ng/ml) and FGF-9 (100 ng/ml) was conducted in 0.1% bovine serum albumin in PBS at a flow rate of 100 l/min for 20 min with PBS rinses (100 l/min) for 10 min and heparin (10 g/ml) rinses at 100 l/min for 10 min to enable surface regeneration between growth factor injections. Experiments were performed in quadruplicate simultaneously (E4, Q-Sense AB, Sweden) at 37 Ϯ 0.1°C. The QCM-D (51) records frequency ( f ) and dissipation (D) changes during the experiment. f and D responses were analyzed using QTools (Q-Sense AB, Sweden) using the Voigt model to obtain adsorbed mass estimates. The Voigt model assumes that the adsorbed layer is of uniform thickness and density, conserves its shape, and does not flow (52).

Involvement of Perlecan/FGF Interactions in Cell
Proliferation-The ability of perlecan to support cell proliferation was tested in a cell based assay that measures the signaling of FGFs by their receptors (27,53). Baf-32 cells that do not express endogenous heparan sulfate proteoglycans and are transfected with the FGFR3iiic receptor isoform (kindly supplied by Prof. D. Ornitz, Washington University Medical School, St. Louis, MO) were maintained in RPMI 1640 medium supplemented with 10% fetal calf serum, 10% media from WEHI 3BD Ϫ cells (a kind gift from the Ludwig Institute, Melbourne), G418 (400 g/ml), and penicillin/streptomycin. Mitogenic assays were performed as outlined by Ornitz and Itoh (53), and Knox et al. (26,27). Briefly, Baf-32 cells were washed, resuspended in RPMI 1640 medium containing 10% fetal calf serum, and seeded into 96-well plates at a density of 6 ϫ 10 5 cells/50 l/well. The wells were then made up to a volume of 150 l with RPMI 1640 medium containing FGF-1, FGF-9, or heparin and purified perlecan samples to provide final concentrations of 5 nM for FGF-1 or -9, 2 g/ml for heparin, and 1.25 g/ml for the purified perlecans. The cells were then incubated at 37°C for 48 h under an atmosphere of 5% CO 2 in air and 98% humidity and 0.5 Ci of [ 3 H]thymidine in 20 l of medium was added per well and the cells incubated a further 6 h. The cells were then pelleted by centrifugation (1000 ϫ g for 10 min) and the supernatant solution flicked from the plates to waste. This procedure was repeated three times and then the cells were resuspended in PBS (100 l) and added to 2 ml of scintillation fluid and the samples counted on an automatic liquid scintillation counter. Control experiments were also conducted with perlecan samples that had been predigested with heparitinase III to remove HS chains.

RESULTS
Perlecan was immunolocalized in a number of cartilaginous fetal human tissues including the interphalangeal and metacarpophalangeal joints of the fingers and toes (Fig. 1); cartilaginous Perlecan in Human Cartilage DECEMBER 1, 2006 • VOLUME 281 • NUMBER 48 rudiments of the elbow (Fig. 2), and femoral condyle, patella, and tibial plateau of the fetal knee (Fig. 3) and postnatal adult knee articular cartilages (Fig. 4). Perlecan was a prominent extracellular matrix component of the finger and toe cartilaginous rudiments (Fig. 1, c and f-i) where it was a particularly prominent pericellular component of the hypertrophic growth plate chondrocytes (Fig. 1, f and h). Perlecan localizations in these tissues were similar to the distribution of total anionic proteoglycan staining with toluidine blue (Fig. 1a) although it was not localized strongly at the presumptive articulating joint surfaces (Fig. 1c). Perlecan was also strongly associated with small venules and arterioles in the synovial lining and perichondrial tissues of the fingers and toes (Fig. 1, g-i). Aggrecan was prominently localized predominantly at the margins of the cartilaginous rudiments in the fingers (Fig. 1d), and did not display as uniform a distribution throughout the cartilage rudiment as perlecan (Fig. 1c). Perlecan also had a widespread distribution in the cartilaginous rudiments of the fetal elbow (Fig. 2e) and knee (Fig. 3), was prominently associated with small blood vessels in the synovial lining tissues of the joint capsules (Fig. 3g) and perichondrium, and displayed diminished levels at the presumptive articulating joint surfaces where it was associated with prominent small vessel networks (Fig. 3, f and h).
Perlecan was also prominent in the pericellular environment of the chondrocytes of postnatal juvenile (Fig. 4a), adolescent (Fig. 4b), and adult (Fig. 4c) cartilage, but was deficient in the more remote intercellular matrix. This was particularly evident in the adult, where cell density is low. The discrete pericellular staining of perlecan was independent of the domain location of the antibody used for immunolocalization and contrasted with the FIGURE 1. Localization of anionic proteoglycan, toluidine blue-fast green stain (a) and immunolocalization of perlecan using anti-perlecan domain I mAb A76 (c) and of aggrecan using an anti-aggrecan G 1 domain 946D4D11 mAb (d) in vertical mid-sagittal sections of human fetal (14 week) fingers (a-e). Segments (b and e) are negative controls for the perlecan and aggrecan antibodies, respectively, segment b utilized an irrelevant, isotype and concentration matched IgG to the authentic aggrecan and perlecan primary antibodies used, whereas segment e was conducted with secondary detection reagents in the absence of primary antibody. Higher power views of perlecan localization in a 14-week-old human fetal finger are also depicted. Magnifications of the boxed areas in f are depicted in g (box 1), h (box 2), and i (box 3), respectively. Perlecan has a strong pericellular distribution around hypertrophic growth plate chondrocytes ( f ), but is also present in small blood vessels (arrows) in the synovial lining tissues (g), perichondrium (h and i), and growth plate (arrows). The asterisks in a-e with arrows depict the site of the nail-bed and facilitates the orientation of the specimen. Scale bars are 500 m in a-e and 100 m in f-i. more uniform intercellular staining observed for aggrecan, an elevation in pericellular aggrecan staining was also observed (Fig. 4d).
The high cell density of immature cartilage accounts for the ready visualization of perlecan in Western blotting of juvenile cartilage extracts (Fig. 5a) and contrasts with the chondrocytepoor adult cartilage samples. Analysis of the juvenile perlecan by SDS/PAGE and immunoblotting revealed the presence of both HS and CS substitution because readily resolved core proteins of discrete size could be visualized after either hepariti-nase or chondroitinase pretreatment (Fig. 5a). The largest sized core protein detected (ϳ470 kDa) was compatible with the intact perlecan core protein, whereas the smaller forms (200 -300 kDa) may be due to proteolytic processing or alternative splicing. When equivalent immunoblotting was performed using an antibody that recognizes the unsaturated oligosaccharide remaining on the core protein of HS-proteoglycans following heparitinase III treatment, it was apparent that the vast majority of the core proteins required both heparitinase and chondroitinase treatment for discrete resolution (Fig. 5b).
Western blotting of extracts of 3-and 65-year-old knee articular cartilage samples resolved on 3-8% NuPAGE gels using mAbs to domains I (mAb A76), III (mAb 7B5), and V (mAb A74) of perlecan identified a number of core protein species (Fig. 6a). Many of the perlecan core protein species were also detected using mAb 3G10 indicating the presence of HS substitution. Molecular weights were calculated from a calibration curve fitted to the standard protein relative migration data (Fig.  6b). Immunopurified human umbilical artery endothelial cell perlecan (HUAEC-perlecan) migrated as a 470-kDa core protein that was detected by mAbs A76, A74, and 3G10 in immunoblots. The molecular masses of the human cartilage perlecan core protein species ranged from 70 to 380 kDa, although their abundances were significantly lower (ϳ4-fold lower) in the  and perlecan (c, d, and f-h) in mid-saggital sections of a 12-week (a-c  and e) and 14-week-old (d and f-h) fetal human knee. The patella (p), femoral condyle (fc), and tibial plateau (tp) are indicated in the toluidine bluestained section a. Perlecan displayed a widespread distribution throughout the cartilaginous rudiments except at their margins (c, d, and f ), aggrecan displayed a more widespread distribution in the cartilaginous rudiments and was also relatively enriched at its margins where perlecan was deficient. b, a negative control slide using an irrelevant isotype and concentration matched IgG to that used in the aggrecan and perlecan localizations is depicted in e. Higher power views of the boxed areas depicted in d labeled 1-3 are displayed in segments f-h, respectively. These depict a prominent collection of perichondrial blood vessels (f, arrows) on the presumptive articulating surface of the femoral condyle (box 1 in d), an assortment of small blood vessels supplying the outer regions of the meniscus (box 2 in d) are also stained positively with perlecan domain I antibody (g, arrows) as are cartilaginous regions of the femoral condyle and tibial plateau. In contrast, the presumptive weight bearing contact regions (box 3 in d) between the femoral condyle and tibial plateau that have undergone joint cavitation contain cells of a flattened morphology that do not stain positively for perlecan (h, arrows). Scale bars are 500 m in a-e and 100 m elsewhere.  7B5 (a-c). Similar immunolocalizations were also obtained using anti-perlecan domain V (mAb A74) and anti-perlecan domain I (mAb A76) (results not shown). Aggrecan was also localized using monoclonal antibody 946D4D11 to the G 1 domain of the aggrecan core protein (d). Aggrecan staining is observed uniformly throughout the extracellular matrix and elevated pericellularly in contrast to the discrete pericellular localization of perlecan observed with all anti-perlecan antibodies. All panels are at the same magnification, scale bar 100 m. DECEMBER 1, 2006 • VOLUME 281 • NUMBER 48 older cartilage specimen. Non-dissociated high molecular mass perlecan greater than 400 kDa in size was also detected in the 65-year-old cartilage samples. Perlecan band 2 (215 kDa) was a major species in the 3-year-old cartilage specimen only and appeared derived from perlecan domain I because it was detected by mAb A76, it was not detected using mAb 3G10, thus it may represent a fragment of domain I devoid of the HS attachment region. Additional perlecan core protein species of molecular masses ranging from 127 to 176 kDa were identified by mAbs A76 and 3G10 indicating that these may be processed portions of domain I perlecan core protein.

Perlecan in Human Cartilage
Surface plasmon resonance studies indicated that HUAECperlecan bound significant amounts of both FGF-1 and FGF-9, whereas the chondrocyte perlecan bound well to FGF-9 but weakly to FGF-1 (Fig. 7). The association rate of both growth factors to HUAEC-perlecan was similar with equilibrium achieved instantaneously. The dissociation rate, however, was marginally slower for FGF-9, suggesting that the affinity for this growth factor might be slightly greater. The poor binding of FGF-1 to chondrocyte perlecan could not be explained by an absence of HS as the purified chondrocyte perlecan was reactive with the anti-HS mAb 10E4 in an enzyme-linked immunosorbent assay (data not shown). Thus the binding characteristics of endothelial and chondrocyte-derived perlecan are different. Control experiments conducted with perlecan samples that had been pre-digested with heparitinase III to remove HS gave essentially baseline sensograms indicating that growth factor binding was HS dependent for both perlecan samples (data not shown).
The quartz crystal microbalance studies confirmed the surface plasmon resonance results that indicated that endothelial perlecan bound growth factors FGF-1 and FGF-9 more efficiently than chondrocyte perlecan (see Fig. 8, a and b). How-ever, there was no significant difference between the amount of FGF-1 and FGF-9 binding to chondrocyte perlecan as was observed using surface plasmon resonance. Changes in dissipation measured by the QCM-D relate to changes in the viscoelasticity of the adsorbed layer. Dissipation measured for FGF-1 and FGF-9 binding to perlecan was normalized against the thickness of the adsorbed layer (see Fig. 8c). FGF-1 and FGF-9 binding to chondrocyte perlecan was significantly (p Ͻ 0.05) more dissipative than binding to endothelial perlecan suggesting that there are differences in growth factor binding between endothelial and chondrocyte-derived perlecan and that the chondrocyte perlecan binds the FGFs less efficiently.
HUAEC-perlecan supported the proliferation of Baf-32 cells using both FGF-1 and FGF-9. However, FGF-1 was poorly responsive with chondrocyte perlecan in this assay system, whereas cell proliferation with FGF-9 was similar to the level achieved with endothelial cell perlecan (Fig. 9). This supported the relative abilities of each perlecan sample to interact with these growth factors in surface plasmon resonance and QCM-D experiments (Figs. 7 and 8). Heparitinase-digested perlecan samples did not support cellular proliferation (data not shown), which is consistent with a lack of FGF-1 and -9 binding by the HS-free perlecans in plasmon resonance studies.

DISCUSSION
Perlecan was an abundant extracellular component of the primordial cartilaginous rudiments and a prominent pericellular component of hypertrophic growth plate chondrocytes in joint cartilages of the fingers, toes, elbow, and knee, which is in keeping with its role in chondrogenesis (2,29,34,36,38,44). Perlecan displayed an expression pattern that more closely correlated with the extent of the cartilaginous rudiments than did aggrecan suggesting that perlecan may represent a more reliable marker of chondrogenesis than aggrecan in developing cartilages and also that these proteoglycans may have differing functional roles in these early cartilages. Despite its predominant extracellular matrix localization pattern, aggrecan also displayed increased staining pericellularly in some of the cartilage specimens examined in this study. This is consistent with the enhanced pericellular distribution patterns demonstrated for newly synthesized aggrecan in cartilage explant cultures (54).
This study has also shown that the pericellular localization of perlecan persisted throughout life in human cartilage, and that perlecan may be present in forms that are devoid of GAG, possess either HS or CS, or are decorated with both HS and CS. The presence of both HS and CS substitution has previously been FIGURE 5. Western blot analysis of perlecan in cartilage extracts from individuals aged 7 months and 64 years. The extracts were pre-digested with combinations of heparitinase and chondroitinase then fractionated by SDS/PAGE on 4 -12% NuPAGE gradient gels and electroblotted onto nitrocellulose membranes. Perlecan was immunolocalized on the blots using anti-domain III monoclonal antibody 7B5 (a). The bands in the upper part of the gel (**) represent specific staining with monoclonal antibody 7B5, whereas the 81-kDa band is nonspecific (*) because it was visualized with the conjugated secondary antibody in the absence of primary antibody. Due to the lower abundance of perlecan core protein species in the 64-year-old sample, this segment of the blot was subjected to a longer development step to visualize the core protein species that are still faint but are indicated by arrows labeled i-iii (b). Replicate blots were also analyzed for ⌬-HS containing core protein species, ⌬-4,5 unsaturated uronate stub epitopes generated by heparitinase from the native HS chains were immunolocalized by monoclonal antibody 3G10.
described on perlecan isolated from the bovine rib growth plate (55). GAG attachment sites have been described in both domains I (7-12) and V (11,12,20) of perlecan, with three such sites being present in domains I and II in domain V, although it is not clear whether GAG substitution in domain V is common to all species, both domains may be substituted with either HS or CS. GAG-free forms of perlecan have also been described in other tissues (56). Both the absence of GAG and the type of GAG substitution could alter the functional properties of perlecan (26,27). At present it is not clear under what conditions the cell can alter its GAG substitution pattern, or whether all chondrocytes in a given tissue produce the same pattern of substitution.
It is also evident that perlecan core proteins of different sizes are present in the cartilage matrix. The human perlecan HSPG2 gene possesses 97 exons (GenBank accession number NT_004576) and there is therefore great potential for core protein size variation due to alternative splicing. Whereas there is some evidence for such splice variation (57), the most likely explanation of core protein size variation in the present study is FIGURE 6. Identification of core protein species by Western blotting using mAbs to perlecan domains I, III, and V (a) and estimation of their molecular weights relative to HiMark protein molecular weight standards (b). The blots presented are representative of triplicate blots. Aliquots of dialyzed 4 M guanidine HCl extracts of 3-and 65-year-old human articular cartilage were subjected to sequential digestions with chondroitinase ABC and keratanase followed by heparitinase III, dialyzed, and freeze dried, then electrophoresed in 3-8% NuPAGE Tris acetate gels and electroblotted to nitrocellulose for immunodetection of perlecan core protein species. The 3-year-old cartilage specimens were run in lanes 1, 3, 5, and 7 and the 65-year-old cartilage specimens in lanes 2, 4, 6, and 8, respectively, HUAEC perlecan was run in lanes 9 -11. Due to the relatively low abundance of perlecan in the 65-year-old cartilage samples, their sample loadings were double that of the 3-year-old specimens. Core protein species were identified using mAb A76 to perlecan domain I in lanes 1, 2, and 9; mAb 7B5 to perlecan domain III in lanes 3 and 4; mAb A74 to perlecan domain V in lanes 5, 6, 10, and mAb 3G10 to the ⌬-HS stub epitopes generated from pre-digestion of the HS side chains of perlecan by heparitinase in lanes 7, 8, and 11. Alkaline phosphatase-conjugated secondary antibodies were used to detect the immune complexes using the substrates nitro blue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate. Prestained HiMark protein molecular weight standards were also electrophoresed for size determinations on the perlecan core protein species. The migration positions of the protein standards and their known molecular weights were used to construct a third order polynomial line of best fit of log 10 M r versus R f described by the equation y ϭ Ϫ3.8703x 3 ϩ 7.1766x 2 Ϫ 5.1653x ϩ 6.4106 (R 2 ϭ 0.9922) (b). Calculated molecular mass values (kDa) for cartilage perlecan core protein species 1-10 depicted by arrowheads in a were, respectively, 381, 215, 176, 154, 140, 132, 127, 111, 80, and 69 kDa. FIGURE 7. Perlecan isolated from ovine chondrocytes and human umbilical artery endothelial cells was biotinylated and linked to a BIAcore streptavidin chip. FGF-1 (a) and FGF-9 (b) was passed over each chip and the amount of binding was monitored. Representative sensograms are depicted. Endothelial cell perlecan bound significant amounts of both growth factors, whereas the chondrocyte perlecan bound avidly to FGF-9 but very poorly to FGF-1. The association rate of both growth factors to endothelial cell perlecan was similar with equilibrium being achieved rapidly, whereas the dissociation rate was slower for FGF-9. The association rate of FGF-9 was slower for chondrocyte perlecan.
proteolytic processing. It is interesting to note that the sites of GAG substitution in perlecan domains I and V reside toward the two termini of the intact core protein. Whereas proteolytic processing at either end may remove sites for GAG substitution, retention of the opposite end would still permit substitution with either HS or CS, assuming that all potential sites are equally amenable to substitution. Several 200 -450-kDa perlecan core protein species were evident in the 7-month cartilage extracts in this study. Similar species have also been observed in embryonic chick cartilage (58). The abundance of these core protein species was considerably diminished in the mature cartilage specimens.
Several perlecan core protein species were identified in the human cartilage extracts from the 3-and 65-year-old individuals, these ranged in size from 70 to 381 kDa. A number of studies have demonstrated a wide range of perlecan core protein species of similar size ranges to those identified in the present study in endothelial (28), epithelial (59), and smooth muscle cells (60) and in bone marrow (61) and keratinocytes (62). Similar sized core protein species are also present in human follicular fluid (63). Furthermore, a 25-kDa C-terminal perlecan core protein fragment has been isolated from human urine in end stage renal failure (64) and a 90-kDa peptide from domain IV of perlecan (58) has been isolated and sequenced providing evidence that perlecan may be extensively processed in situ. Matrix metalloproteases and plasmin are capable of degrading perlecan (28), and have also been shown to be up-regulated in smooth muscle cells isolated from aortic aneurysm tissues (65) and in cancer (66).
The precise role of perlecan in cartilage remains unclear, although the severe influence of gene knock-out and mutation (3, 35-37, 39 -42) on cartilage development and function indicates the critical importance of this proteoglycan in the pericellular matrix. It has been shown that perlecan can interact with many components of the extracellular matrix via either core protein (15,67) or HS (18, 26 -28) interactions. Of particular FIGURE 8. Perlecan isolated from human umbilical artery endothelial cells (a) and ovine chondrocytes (b) was physisorbed onto gold QCM-D crystals and then blocked with 0.1% bovine serum albumin in PBS. FGF-1 and FGF-9 were exposed to each perlecan with heparin used to remove and regenerate the surface (as indicated by the arrow) between growth factor injections. Endothelial cell perlecan bound significant amounts of both growth factors more efficiently than chondrocyte perlecan. c, dissipation, normalized to thickness, for growth factor binding to perlecan shows that FGF-1 and FGF-9 binding to chondrocyte perlecan is more dissipative than to endothelial perlecan. * ϭ t test, p Ͻ 0.05 relative to chondrocyte perlecan. FIGURE 9. The endothelial cell and ovine chondrocyte perlecans were tested in an assay that measures cell proliferation induced by FGFs binding to their receptors. Baf-32 cells expressing the FGFR3iiic subtype were incubated with FGF-1 (a) or FGF-9 (b) either alone or in the presence of perlecan or heparin. Chondrocyte perlecan had no mitogenic effect on the presence of FGF-1 but had a positive mitogenic effect on the presence of FGF-9, whereas endothelial cell perlecan stimulated cell growth in the presence of either FGF-1 or 9. Key to lanes: 1, heparin ϩ FGF-1 or 9 control; 2, heparin control; 3, FGF-1 or 9; 4, endothelial cell perlecan ϩ FGF-1 or -9; 5, ovine chondrocyte perlecan ϩ FGF-1 or -9.
note are the interactions of the HS with basic FGF (26 -28) and PRELP (18). The former interaction may influence cell metabolism and the latter may provide a union between the pericellular matrix and the more remote intercellular matrix. Interestingly, neither basic FGF nor PRELP interacts with CS, which raises the question of whether all forms of perlecan fulfill the same role in the tissue. The present growth factor binding and cell proliferation studies on the interaction of perlecan with FGFs demonstrate that different perlecan structures do indeed vary in function (27). In the future, it will be important to establish whether the proportion of HS or CS varies spatially or temporally in specific cartilage sites and locations, and whether the different forms of perlecan have unique roles in growing and mature cartilages.