Post-translational Modifications of (cid:97) 5 (cid:98) 1 Integrin by Glycosaminoglycan Chains

Cell-fibronectin interactions, mediated through sev- eral different receptors, have been implicated in a wide variety of cellular properties. Among the cell surface receptors for fibronectin, integrins are the best characterized, particularly the prototype (cid:97) 5 (cid:98) 1 integrin. Using [ 125 I]iodine cell surface labeling or metabolic radiolabel-ing with sodium [ 35 S]sulfate, we identified (cid:97) 5 (cid:98) 1 integrin as the only sulfated integrin among (cid:98) 1 integrin het- erodimers expressed by the human melanoma cell line Mel-85. This facultative sulfation was confirmed not only by immunoprecipitation reactions using specific monoclonal antibodies but also by fibronectin affinity chromatography, two-dimensional electrophoresis, and [ 35 S]Sulfate-labeled Mel-85 cell extracts were immunoprecipitated using the same mono- or polyclonal antibodies as above. Immunopre- cipitates were analyzed by 7.5% SDS-PAGE (40) followed by electrotransference onto nitrocellulose membranes (41) and exposure to x-ray films (Kodak, Rochester, NY). The same procedure was used to study HCT-8 and MG-63 cell extracts with an anti- (cid:97) 5 integrin antibody. Ionic, Chaotropic, and Denaturating Experimental Conditions— The [ 35 S]sulfate-labeled immunoprecipitates obtained with anti- (cid:97) 5 integrin monoclonal antibody were incubated with 4 M NaCl, 4 M MgCl 2 , 6 M guanidine HCl, and 8 M urea for 2 h at 37 °C. Mixtures were then boiled for 5 min, and (cid:97) 5 (cid:98) 1 integrin was separated from Sepharose beads by centrifugation for 1 min at 13,000 (cid:51) g . Supernatants were dialyzed against water, concentrated in a speed vaccum concentrator, subjected to 7.5% SDS-PAGE under nonreducing conditions, and electrotrans- ferred onto nitrocellulose membranes that were then exposed to x-ray films at room temperature for 10 days. Fibronectin Affinity Cromatography— Fibronectin-affinity chromatography of [ 35 S]sulfate-labeled Mel-85 cell extract was performed using purified human plasma fibronectin coupled to CNBr-activated Sepharose (Pharmacia Biotech Inc.) as detailed elsewhere (30). Gel Electrophoresis, Purification of (cid:97) 5 and (cid:98) 1 Integrin Subunits, and Immunoblotting— SDS gel electrophoresis was performed as described

Proteoglycans are complex molecules formed by a core protein to which one or more glycosaminoglycan (GAG) 1 chains are linked. This basic definition, although true, hides the molecular complexity shown by these molecules. They encompass an exceptionally large range of structures involving different core proteins, different classes of GAGs, and different numbers and lengths of individual GAG chains. Other post-translation modifications such as N-and O-glycosylation increase the complexity of these molecules (for review see Refs. 1 and 2).
The biological functions of proteoglycans are numerous. They have been involved in several biological effects (1,(3)(4)(5), such as extracellular matrix (ECM) assembly (6) and cell surface-ECM receptors for growth factors and hormones (2,5,7) or have had a role in biological processes such as cell-cell recognition (8) and control of cell growth (9). The fact that several ECM proteins, such as fibronectin (10), laminin (11), thrombospondin (12), vitronectin (13), type IV collagen (14), and tenascin (15), have GAG binding sites adds credence to the postulated multiple roles of proteoglycans. Supporting the idea of proteoglycans as ECM receptors, syndecan type I binds fibronectin, thrombospondin, collagens (5), and tenascin (16); the heparan sulfate proteoglycan of Schwann cells binds laminin (17); a cell surface chondroitin sulfate proteoglycan is apparently involved in cell adhesion to laminin (18); and a cell surface phosphatidyl inositol-anchored heparan sulfate proteoglycan mediates melanoma cell adhesion to fibronectin (19). Strong corroboration for these proteoglycan-ECM interactions comes from the presence of a heparan sulfate proteoglycan that co-localizes with ␤ 1 integrins as a widespread component of focal adhesion (20).
Among the several ECM molecules that bind proteoglycans, the role of fibronectin should be emphasized not only because of its GAG binding domains but also because of the adhesive properties conferred to this molecule by these domains together with the RGD cell-binding fragment (21)(22)(23). Cells devoid of proteoglycans or bearing proteoglycans with altered GAG chains have a reduced capacity of adhesion to fibronectin and have a defective focal adhesion plaque formation in response to this molecule (24,25).
The best studied receptors for fibronectin that bear adhesiveness and focal adhesion plaque formation are integrins that are ␣/␤ heterodimers widely expressed by almost all animal cells (26,27). Integrins represent good examples of how post-translational modifications can alter the structure of a molecule, thus modulating its biological activity. Integrin glycosylations represent a kind of regulation by which a wide variety of these receptors have their specificity and affinity modulated in several cell lines (28 -30). However, the versatility of cells to modulate the binding properties of integrins is not restricted to glycosylation. Integrin functions can be modulated by acylation of membrane lipid (31), by divalent metal binding (32), and, for the cytoplasmic domain, by tyrosine phosphorylation, which is the best understood example of this type of biological modification, especially in leukocytes and platelets (27,33).
In the present study, we characterize ␣ 5 ␤ 1 integrin as a part-time proteoglycan containing both heparan and chon-droitin sulfate, which per se could affect cell adhesion to both fibronectin RGD and GAG binding domains.

EXPERIMENTAL PROCEDURES
Reagents and Antibodies-Human fibronectin was purified from fresh plasma (obtained from Hospital A. C. Camargo, Sã o Paulo, Brazil) by gelatin affinity chromatography as described (34). Monoclonal antibody A-1A5 that recognizes the ␤ 1 integrin subunit (35) and B-5G10 that reacts with the ␣ 4 integrin subunit (36)  Immunoprecipitation Reactions-Cell surface expression of ␤ 1 integrin heterodimers in Mel-85 cells was probed through immunoprecipitation reactions of cells that were surface labeled (using [ 125 I]iodine) by the lactoperoxidase-H 2 O 2 method as described previously (39). After washing, cells were solubilized by lysis buffer (50 mM Tris-HCl, pH 7.3, 1% Triton X-100, 50 mM NaCl, 5 mM CaCl 2 , 5 mM MgCl 2 , 1 mM phenylmethanesulfonyl fluoride, and 2 g/ml of aprotinin) for 15 min at 4°C. The extract was clarified by centrifugation for 10 min at 13,000 -g, and the supernatant was preincubated with either normal mouse or rabbit serum followed by precipitation with protein A-Sepharose (Sigma). Mel-85 extract (at the same mass of protein, 1 mg) was incubated respectively with antibodies against different integrin subunits (as shown above), and for B-5G10 (an IgG 1 molecule), rabbit IgG was preincubated against mouse IgG followed by protein A-Sepharose. Affinity beads were washed with lysis buffer, and the immunoprecipitates were eluted by boiling for 5 min with Laemmli buffer.
[ 35 S]Sulfate-labeled Mel-85 cell extracts were immunoprecipitated using the same mono-or polyclonal antibodies as above. Immunoprecipitates were analyzed by 7.5% SDS-PAGE (40) followed by electrotransference onto nitrocellulose membranes (41) and exposure to x-ray films (Kodak, Rochester, NY). The same procedure was used to study HCT-8 and MG-63 cell extracts with an anti-␣ 5 integrin antibody.
Ionic, Chaotropic, and Denaturating Experimental Conditions-The [ 35 S]sulfate-labeled immunoprecipitates obtained with anti-␣ 5 integrin monoclonal antibody were incubated with 4 M NaCl, 4 M MgCl 2 , 6 M guanidine HCl, and 8 M urea for 2 h at 37°C. Mixtures were then boiled for 5 min, and ␣ 5 ␤ 1 integrin was separated from Sepharose beads by centrifugation for 1 min at 13,000 ϫ g. Supernatants were dialyzed against water, concentrated in a speed vaccum concentrator, subjected to 7.5% SDS-PAGE under nonreducing conditions, and electrotransferred onto nitrocellulose membranes that were then exposed to x-ray films at room temperature for 10 days.
Gel Electrophoresis, Purification of ␣ 5 and ␤ 1 Integrin Subunits, and Immunoblotting-SDS gel electrophoresis was performed as described (40). Samples under reducing or nonreducing conditions were analyzed on 5 or 7.5% polyacrylamide gels, and proteins were transferred overnight to nitrocellulose filters as described (41). Molecular mass markers (myosin, 205 kDa; ␤-galactosidase, 116 kDa; and phosphorylase b, 98 kDa; albumin, 67 kDa) were purchased from Sigma. For two-dimensional electrophoresis, samples were separated in the first dimension by isoelectric focusing (42) using an ampholyte gradient (pH 4.0 -6.5, six parts and pH 3.0 -10.0, four parts, Pharmacia) followed by 7.5% SDS-PAGE in nonreducing conditions. Glycosaminoglycan analysis was performed using agarose gel electrophoresis in 0.05 M 1,3-diaminopropane acetate buffer pH 9.0 (Aldrich). After the electrophoretic run, compounds were precipitated in the gel using 0.1% Cetavlon for 2 h at room temperature (43). After drying, the gel was stained with toluidine blue and exposed to x-ray films (X-Omat, Kodak) for 10 days at room temperature. GAG standards used were heparan sulfate from bovine pancreas (44), dermatan sulfate from pig skin, and chondroitin sulfate from shark cartilage (Seikagaku, Kogyo Co., Tokyo, Japan).
To study the specific pattern of glycosylation of ␣ 5 and ␤ 1 integrin subunits, [ 35 S]sulfate-labeled Mel-85 cell lysate was immunoprecipitated using a monoclonal antibody against the ␣ 5 integrin subunit as already described, and the precipitate was submitted to a preparative 7.5% SDS-PAGE under nonreducing conditions using prestained ␤-galactosidase (116 kDa) that comigrates with the ␤ 1 integrin subunit as a standard. Autoradiography of separated ␣ 5 and ␤ 1 subunits was done in identical conditions as above and used as a guide. Gel pieces were then cut off in the positions of separated ␣ 5 and ␤ 1 subunits, and proteins were excised and eluted from the gel by incubation in 50 mM Tris-HCl, pH 7.3, containing 0.1% Triton X-100 overnight at 4°C. The mixtures were then filtered through 0.45-m filters (Nalgene, Rochester, NY) to remove polyacrylamide, and the solutions containing extracted proteins were dialyzed against water and concentrated 20-fold. Purified ␣ 5 and ␤ 1 integrin subunits were then submitted to ␤-elimination reaction to release free GAG chains, which were incubated with chondroitinase ABC, heparitinases I and II, or a mixture of these enzymes (see below), and the digests were analyzed by agarose gels.
Immunoblotting reactions using Rb3847 (a rabbit polyclonal antibody that only reacts with the ␤ 1 integrin subunit but not with the denatured ␣ 5 integrin subunit) and a monoclonal antibody against chondroitin sulfate chains (CS-56) were performed as described previously (30).
Chemical ␤-Elimination and Enzyme Digestions-The GAG chains from [ 35 S]sulfate-labeled ␣ 5 ␤ 1 integrin, purified by immunoprecipitation using a monoclonal antibody against the ␣ 5 integrin subunit, were liberated by digestion of the protein core using excess proteinase-K (50 g; Sigma) at 58°C overnight or by a ␤-elimination reaction (treatment overnight at 37°C with 0.1 M NaOH in the presence of 2 M NaBH 4 ; Sigma). The products obtained were analyzed by agarose gel electro-2 K. Yamada, personal communication.

FIG. 1.
There is a sulfated ␣␤ 1 integrin in Mel-85 cells. 35 S-Labeled Mel-85 cell extract (lane 1) was immunoprecipitated by a specific monoclonal antibody to the ␤ 1 integrin subunit (lanes 3 and 5). As a negative control, the same extract was precipitated by normal mouse serum (lanes 2 and 4). Immunoprecipitates were separated by 7.5% SDS-PAGE, transferred onto a nitrocellulose membrane that was exposed to an x-ray film (lanes 2 and 3), or exposed to a rabbit polyclonal antiserum against the ␤ 1 integrin subunit (lanes 4 and 5). Open arrow, pre-␤ 1 integrin subunit; closed arrow, ␤ 1 integrin subunit; arrowhead, ␣ subunit. Molecular mass markers are on the left.

Sodium [ 35 S] Sulfate
Labeling of ␣␤ 1 Dimer Integrin-Because integrins are substrates for several different post-translational modifications, we decided to determine whether they could function as substrates for sulfation. We decided to address this question using [ 35 S]sulfate labeling of the cells, immunoprecipitation, and blotting experiments. As shown in Fig.  1, Mel-85 cells in culture efficiently incorporate [ 35 S]sulfate. The cell lysate was submitted to immunoprecipitation with a monoclonal antibody against the ␤ 1 integrin subunit, and a [ 35 S]sulfated ␣␤ 1 integrin molecule dimer was detected. This suggests that ␤ 1 integrin is a substrate for post-translational sulfation.
␣ 5 ␤ 1 Is the Only Sulfated Integrin in Mel-85 Cells-After the demonstration that ␣␤ 1 dimer integrin is a sulfated molecule, our next experimental procedure was to identify the ␣ subunit complementing the ␤ 1 subunit in this particular integrin heterodimer. To investigate this, Mel-85 cells were surface radiolabeled with [ 125 I]iodine by the lactoperoxidase method or metabolically labeled with [ 35 S]sulfate. Both cell lysates were immunoprecipitated with antibodies against different integrin subunits. We can see in the [ 125 I]iodine-labeled immunoprecipitates ( Fig. 2A) the presence of ␤ 1 , ␣ 2 , ␣ 3 , ␣ 4 , ␣ 5 , ␣ 7 , and probably ␣ 1 subunit, a 200-kDa signal ( Fig. 2A, lane 1) that could be precipitated with the ␤ 1 subunit. Neither cell flow cytometry nor immunoprecipitation showed detectable levels of the ␣ 6 integrin subunit in Mel-85 cells (data not shown). Interestingly, Fig. 2B shows that only ␣ 5 and the corresponding ␤ 1 subunit are sulfated. These findings suggested that ␣ 5 ␤ 1 integrin is a facultative sulfated ␤ 1 integrin molecule because none of the other ␤ 1 integrin molecules incorporated [ 35 S]sulfate.
It is known that after reduction of the disulfide bonds by ␤-mercaptoethanol (chemical reduction), the ␣ 5 integrin subunit comigrates with the ␤ 1 subunit (36). The immunoprecipitates obtained using monoclonal antibodies to ␤ 1 and ␣ 5 integrin subunits or a polyclonal antibody against the ␣ 5 ␤ 1 integrin dimer from a [ 35 S]sulfate-labeled Mel-85 cell lysate were then subject to chemical reduction. As shown in Fig. 2C, after chemical reduction the immunoprecipitates reveal just one band in the gel, confirming that ␣ 5 ␤ 1 integrin is a sulfated molecule. It is also known that the ␣ 5 ␤ 1 integrin has fibronectin as the only ECM ligand (27,46). Fig. 2D shows that after elution from a fibronectin affinity chromatography ␣ 5 ␤ 1 integrin is detected as a sulfated molecule.
␣ 5 ␤ 1 Integrin Is the Only Sulfated Molecule; No Other Sulfated Molecule Co-precipitates with Integrin-Data in the literature describe proteoglycans as ␣ 5 ␤ 1 integrin-associated molecules that complement the requirements involved in cell adhesion to fibronectin. In addition, in melanoma cells a heparan sulfate proteoglycan of 150/175 kDa has been described to bind fibronectin (19,22,47,48). We cannot therefore discard the possibility of a physical association between a third molecule that comigrates with ␣␤ integrin subunits, masking the sulfated signals in the autoradiograms. To rule out this possibility the same [ 35 S]sulfate-labeled Mel-85 cell extract was again immunoprecipitated by specific monoclonal antibodies to ␣ 5 and ␤ 1 integrin subunits and now submitted to a twodimensional electrophoresis (Fig. 3, A and B). We can observe that the immunoprecipitation reactions using either anti-␤ 1 antibody or anti-␣ 5 antibody show only a sulfated signal of ␣ 5 ␤ 1 dimer.
To corroborate the findings described above and demonstrate that the sulfate groups in ␣ 5 ␤ 1 integrin are covalently linked and not adsorbed to this molecule or to the beads during immunoprecipitation, this integrin was immunoprecipitated from a [ 35  , and a polyclonal antibody against the ␣ 7 subunit (lanes 6). Immunoprecipitates were separated by 7.5% SDS-PAGE and transferred onto nitrocellulose membranes that were exposed to x-ray films. C, detergent extracts from Mel-85 cells metabolically labeled with [ 35 S]sulfate were immunoprecipitated by monoclonal antibodies against the ␣ 5 integrin subunit (lane 1), ␤ 1 integrin subunit (lane 3) and by a rabbit polyclonal antibody against ␣ 5 ␤ 1 integrin (lane 5) or normal mouse serum (lanes 2 and 4) and normal rabbit serum (lane 6) as negative controls. The immunoprecipitates were reduced by a ␤-mercaptoethanol containing buffer, separated by 7.5% SDS-PAGE, and electrotransferred to a nitrocellulose membrane that was exposed to an x-ray film. D, cell lysate from [ 35 S]sulfate-labeled Mel-85 cells were chromatographed on a bovine serum albumin-Sepharose column ( lanes  1 and 3) or a fibronectin-Sepharose column (lanes 2 and 4). EDTAeluted materials were separated by 7.5% SDS-PAGE and transferred onto a nitrocellulose membrane that was exposed to an x-ray film ( lanes  1 and 2), or the membrane was treated with a specific rabbit polyclonal antiserum against the ␤ 1 integrin subunit (lanes 3 and 4). The arrow represents the ␤ 1 integrin subunit, and the arrowhead represents the ␣ 5 integrin subunit. Molecular mass markers are on the left. submitted to 7.5% SDS-PAGE, and transferred onto nitrocellulose filters, which were then exposed to an x-ray film (Fig.  3C). These results show that the ␣ 5 ␤ 1 integrin bears covalently linked sulfate groups.
␣ 5 ␤ 1 Integrin Is a Hybrid Proteoglycan-Proteoglycans represent the best characterized sulfated molecules containing serine-linked sulfated GAG chains as a result of post-translational modifications of the protein core (2,7,10). To determine the site of sulfate substitution in the ␣ 5 ␤ 1 integrin dimer, [ 35 S]sulfate-labeled Mel-85 cell lysate was immunoprecipitated using antibody against the ␣ 5 subunit and subjected to ␤-elimination. This procedure cleaves GAG chains from the protein core. As shown in Fig. 4A (lane 2), ␣ 5 ␤ 1 integrin after ␤-elimination showed two sulfated bands that comigrate electrophoretically with chondroitin and heparan sulfate standards. An identical result was obtained when [ 35 S]sulfated ␣ 5 ␤ 1 integrin was submitted to proteinase-K digestion (a serine protease of broad specificity) (Fig. 4A, lane 3). These experiments suggest that ␣ 5 ␤ 1 integrin is a hybrid chondroitin/heparan sulfate proteoglycan. This was further investigated by degradation of the ␤-eliminated material from ␣ 5 ␤ 1 integrin with specific enzymes that degrade chondroitin sulfate (chondroitinase ABC) and heparan sulfate (heparitinases type I and II). We can see that under these conditions, both sulfated bands were completely degraded by the specific enzymes (Fig. 4B), demonstrating that ␣ 5 ␤ 1 integrin is in fact a hybrid chondroitin/heparan sulfate proteoglycan.
Both ␣ 5 and ␤ 1 Integrin Subunits Have Chondroitin and Heparan Sulfate Chains-Because in Mel-85 cells ␣ 5 and ␤ 1 subunits of integrin contain sulfate, our next goal was to de-termine the specific pattern of glycosylation, that is, to which subunit chondroitin and heparan sulfates are linked. Two complementary approaches based on immunologic specificities were used. Fist, ␣ 5 ␤ 1 integrin was immunoprecipitated from a 35 S-labeled Mel-85 cell extract as described above. After separation by polyacrylamide gel electrophoresis, the immunoprecipitate was exposed to an x-ray film, blotted, and reacted with a monoclonal antibody specific for chondroitin sulfate chains (Fig. 5A). We can see that both integrin subunits bear chondroitin sulfate chains. Also, Mel-85 cell lysate was immunoprecipitated with a anti-chondroitin sulfate monoclonal antibody and blotted with a polyclonal antiserum against the ␤ 1 integrin subunit (Fig. 5B). Interestingly, we can see that only the 116-kDa ␤ 1 integrin subunit, which corresponds to the completely glycosylated form, has chondroitin sulfate chains. In contrast, the pre-␤ 1 integrin chain (100 kDa), which corresponds to a high mannose structure, has no chondroitin sulfate chains.
Pre-␤ 1 integrin shows the glycosylation profile of a protein that has not crossed the Golgi. Because synthesis of GAG occurs in the Golgi, the absence of chondroitin sulfate in the pre-␤ 1 integrin should be expected (49). This finding was also substantiated by results shown in Fig. 1 in which the pre-␤ 1 integrin subunit, although coprecipitated, is not sulfated. These results demonstrate that the integrin dimer ␣ 5 ␤ 1 is a proteoglycan and that in Mel-85 cells both integrin subunits have covalently linked chondrotin sulfate chains.
As a second approach we have used successive immunoprecipitation reactions to isolate ␣ 5 ␤ 1 integrin from a 35 S-labeled After elution and dialysis, precipitates were separated by 7.5% SDS-PAGE and transferred onto nitrocellulose strips that were exposed to x-ray films. Molecular mass standards are on the left. Lane 2 represents ␤-eliminated material before enzymic digestion, and lane 1 represents glycosaminoglycan standards as above. The incubation mixtures were analyzed by agarose gel electrophoresis and were dried and exposed to x-ray film.
Mel-85 cell lysate. The purified ␣ 5 ␤ 1 integrin was submitted to preparative polyacrylamide gel electrophoresis. Using an autoradiogram of the gel and prestained molecular mass standards as guides, separated ␣ 5 and ␤ 1 subunits were removed from the gel and subjected to ␤-elimination to obtain GAG free chains. These chains were analyzed by agarose gel electrophoresis before and after treatment with chondroitinase ABC, heparitinases, and a mixture of these enzymes (Fig. 5C). The result obtained from this last set of experiments support the concept that ␣ 5 ␤ 1 integrin is a proteoglycan. It shows that both subunits contain chondroitin sulfate and heparan sulfate, thereby demonstrating that ␣ 5 ␤ 1 integrin is a hybrid chondroitin/heparan sulfate proteoglycan.
Sulfate Incorporation in ␣ 5 Integrins Is a Conserved Phenomenon-Because all experiments described so far were performed using the human melanoma cell line Mel-85, which has a neuro-ectodermic origin, we decided to investigate whether this ␣ 5 ␤ 1 integrin post-translational modification was also present in cell lines of endodermic (HCT-8 cells) and mesodermic (MG-63 cells) origin. Cells were labeled with [ 35 S]sulfate, and lysates were immunoprecipitated with monoclonal antibodies against the ␣ 5 integrin subunit and analyzed by SDS-PAGE followed by electroblotting onto nitrocellulose. A polyclonal antibody against the ␤ 1 integrin subunit (Fig. 6A) was used to detect the integrin. We can see that both cells display the [ 35 S]sulfate incorporation into the ␣ 5 integrin subunit. The results indicate that GAG substitution of ␣ 5 ␤ 1 integrin has been conserved, suggesting its biological significance. However, in the case of the MG-63 cells, only the ␣ 5 subunit was labeled; the ␤ 1 subunit found in MG-63 cells did not contain sulfate. To corroborate the results described in Fig. 6A and provide more evidence that the glycosylation of integrin ␣ 5 ␤ 1 integrin as a proteoglycan is maintained in different tissues, we submitted [ 35 S]sulfate-labeled ␣ 5 ␤ 1 integrin obtained from HCT-8 and MG-63 cell extracts to a ␤-elimination reaction. Products were analyzed by agarose gel electrophoresis. As shown in Fig. 6B, we can see that heparan and chondroitin sulfates are present in ␣ 5 ␤ 1 integrins of endodermic and mesodermic origins, as observed for neuro-ectodermic cells. The results not only confirm the conservative proteoglycan nature of ␣ 5 ␤ 1 integrin from different origins but also indicate a similar glycosylation pattern.

DISCUSSION
Working with the human melanoma cell line Mel-85, we have described ␣ 5 ␤ 1 integrin as a hybrid chondroitin/heparan sulfate proteoglycan. Based on immunoprecipitation reactions from cell lysates that were cell surface labeled with [ 125 I]iodine or metabolically labeled with [ 35 S]sulfate, we were able to detect ␣ 5 ␤ 1 integrin as the only sulfated integrin compared with other ␣ (s) ␤ 1 heterodimers present in Mel-85 cells. Sulfation of ␣ 5 ␤ 1 integrin was confirmed not only by immunological methods but also by fibronectin affinity chromatography, twodimensional electrophoresis, and reduction of disulfide bonds of the ␣ 5 ␤ 1 heterodimer leading to comigration of both ␣ 5 and ␤ 1 integrin subunits, characteristic of this integrin as described (36). Based on different procedures such as chemical deglycosylation by ␤-elimination, proteinase-K digestion, immunological methods, and susceptibility to chondroitinase ABC and heparitinases, we were able to confirm this integrin as a proteoglycan. These results raise the important question of which mechanisms determine ␣ 5 ␤ 1 as the only sulfated integrin. Why are ␤ 1 subunits not sulfated in other ␣␤ 1 heterodimers? The existence of alternative splicing for the ␤ 1 integrin subunit as described (50) (reviewed in Ref. 27) could explain in part such differences. However, because glycosylation of cell surface proteoglycans is restricted to extracellular domains (2) and the ␤ 1 integrin subunit has only alternatively spliced cytoplasmic domains (27) this mechanism does not explain our findings. Oligomerization of ␣␤ integrin heterodimers is an event that occurs during transit through the endoplasmic reticulum (28) and precedes glycosaminoglycan biosynthesis, which occurs during transit through the Golgi (2). Perhaps the best explanation for the part-time proteoglycan nature of ␣ 5 ␤ 1 integrin is that the conformation of the heterodimer exposes the serine residues that are acceptors for the GAG chains, which does not happen with other ␣␤ 1 heterodimers. This conformational hypothesis is consistent with the lack of a consensus sequence for proteoglycan biosynthesis initiation (1,2) and by the experiments performed with decorin, a proteoglycan in which the primary structure of the protein core surrounding the sugar acceptor serine residue can be changed without appreciable modification in the glycosaminoglycan (51). Considering the findings described above, it is possible to assume that the same conformational folding of ␣ 5 ␤ 1 that makes this integrin capable of recognizing the RGD peptide only in fibronectin among several other ECM molecules (52) is also responsible for a specific GAG synthesis that complements the molecular requirements involved in the interaction of this integrin with fibronectin.
The possibility that other integrins can also be sulfated is not ruled out by the present study because we could not detect ␣ 6 ␤ 1 integrin in Mel-85 cells. This is an integrin that binds laminin, a molecule with a GAG binding domain spatially close to the E8 domain corresponding to the ␣ 6 ␤ 1 integrin binding site (53). Furthermore, the structural relationship of the ␣ 5 chain with ␣ IIb and ␣ v (reviewed in Ref. 27) could suggest other ␣␤ integrin heterodimers as putative acceptors for GAG addition. Interestingly, Hayashi, Madri, and Yurchenco (54) have shown that endothelial cell interaction with the basement membrane proteoglycan (perlecan) occurs between the core protein of perlecan and ␤ 1 and ␤ 3 integrins, an interaction partially RGDindependent and modulated by GAGs. The ␤ 1 integrin heterodimer involved in this adhesion resembles the ␣ 5 ␤ 1 molecule and the ␤ 3 integrin, the ␣ v ␤ 3 vitronectin receptor (54).
The present study suggests for the first time that integrins such as ␣ 5 ␤ 1 may have two extracellular binding sites that play a role in fibronectin binding. Previous studies have implicated a specific involvement of the heparin binding site of fibronectin with cell adhesion. These data were based on the fact that the purified fibronectin fragment containing only the heparin binding domain without the RGD peptide can promote adhesion in several different cell models (21,23). Because our work describes ␣ 5 ␤ 1 integrin as a part-time proteoglycan compared with other ␣␤ 1 dimers, we can postulate that the fibronectin-␣ 5 ␤ 1 integrin interaction, which occurs primarily through the RGD peptide in fibronectin, is complemented and stabilized by the secondary interactions of ␣ 5 ␤ 1 chondroitin or heparan sulfate chains with the fibronectin heparin binding domains. The possibility that ␣ 5 ␤ 1 integrin, an integrin that binds only fibronectin, has chondroitin and heparan sulfate chains interacting with the fibronectin heparin binding domains is suggested by the facts that during ECM assembly the fibronectin heparin binding domain can also bind chondroitin sulfate or dermatan sulfate proteoglycans (10) and that soluble proteoglycans can inhibit cell adhesion to fibronectin (reviewed in Ref. 22) and by the existence of nonintegrin fibronectin receptors like CD44 (a chondroitin sulfate proteoglycan) and a heparan sulfate proteoglycan (19,48) as well as by the recent finding that monoclonal antibodies raised against the fibronectin heparin binding domain (Hep II/IIICS) inhibit cell adhesion and also partially inhibit integrin binding to fibronectin (55). A model is postulated in which the RGD and heparin binding sites in fibronectin, although linearly separated, are spatially close due to fibronectin folding. It is thus possible to assume that cell surface proteoglycans and integrin cooperativity during cell adhesion can really be achieved in the case of ␣ 5 ␤ 1 integrin by two binding sites in the integrin molecule that bind RGD peptide and GAG binding domains in fibronectin.  4) were immunoprecipitated using a anti-␣ 5 integrin monoclonal antibody. Following 7.5% SDS-PAGE separation and electrotransference onto nitrocellulose membrane, materials were exposed to x-ray films (lanes 1 and 3) or reacted with a rabbit polyclonal anti-␤ 1 integrin subunit (lanes 2 and 4). The arrowhead points to the ␤ 1 integrin position, and the arrow shows the ␣ 5 subunit. Molecular mass markers are on the left. B, ␣ 5 ␤ 1 integrin obtained from [ 35 S]sulfate-labeled extracts of HCT-8 cells or MG-63 cells were subjected to ␤-elimination to obtain the free GAG chains. These chains were subjected to agarose gel electrophoresis, and the gel was dried an exposed to an x-ray film as described in the legend to