The Collagen-binding A-domains of Integrins α1β1 and α2β1Recognize the Same Specific Amino Acid Sequence, GFOGER, in Native (Triple-helical) Collagens*

We have previously assigned an integrin α2β1-recognition site in collagen I to the sequence, GFOGERGVEGPOGPA (O = Hyp), corresponding to residues 502–516 of the α1(I) chain and located in the fragment α1(I)CB3 (Knight, C. G., Morton, L. F., Onley, D. J., Peachey, A. R., Messent, A. J., Smethurst, P. A., Tuckwell, D. S., Farndale, R. W., and Barnes, M. J. (1998) J. Biol. Chem. 273, 33287–33294). In this study, we show that recognition is entirely contained within the six-residue sequence GFOGER. This sequence, when in triple-helical conformation, readily supports α2β1-dependent cell adhesion and exhibits divalent cation-dependent binding of isolated α2β1 and recombinant α2A-domain, being at least as active as the parent collagen. Replacement of E by D causes loss of recognition. The same sequence binds integrin α1 A-domain and supports integrin α1β1-mediated cell adhesion. Triple-helical GFOGER completely inhibits α2 A-domain binding to collagens I and IV and α2β1-dependent adhesion of platelets and HT 1080 cells to these collagens. It also fully inhibits α1 A-domain binding to collagen I and strongly inhibits α1β1-mediated adhesion of Rugli cells to this collagen but has little effect on either α1 A-domain binding or adhesion of Rugli cells to collagen IV. We conclude that the sequence GFOGER represents a high-affinity binding site in collagens I and IV for α2β1 and in collagen I for α1β1. Other high-affinity sites in collagen IV mediate its recognition of α1β1.

The integrins are important receptors mediating both cellcell contact and cellular recognition of the extracellular matrix. They are heterodimers comprising an ␣ and a ␤ chain and are classified according to the identity of the latter (1). Integrin recognition sequences have been identified in a number of matrix proteins. RGDX 1 (where X is one of several possible amino acids) occurs in a wide variety of adhesive glycoproteins and recognizes several of the integrins. In fibronectin, for example, RGD recognizes a number of integrins, including ␣ 5 ␤ 1 , ␣ V ␤ 3 , and ␣ IIb ␤ 3 (2, 3).
Integrins ␣ 1 ␤ 1 and ␣ 2 ␤ 1 are the major integrin collagen receptors (4 -6). Each recognizes a variety of collagens, including collagen I, the most abundant and widely distributed of all the collagens. Recognition of collagen IV by integrin ␣ 1 ␤ 1 has been reported to involve an aspartyl residue at position 461 in the ␣ 1 (IV) collagen chain and an arginyl residue at the same residue position in the ␣ 2 (IV) chain (7).
Integrin ␣ 2 ␤ 1 plays an essential role in platelet adhesion to collagens in the blood vessel wall under flow conditions (8). This adhesion depends on collagen being in the triple-helical conformation (9) and is important in hemostasis, but it may also be a crucial initiator of thrombosis. Fragmentation of collagen I has indicated the presence of several ␣ 2 ␤ 1 -binding sites throughout the molecule recognized by platelets (9). In particular, fragment ␣ 1 (I)CB3 2 is as good as the parent collagen in supporting ␣ 2 ␤ 1 -mediated platelet adhesion (9,10). We synthesized this fragment as seven overlapping triple-helical peptides and measured their ability to mediate ␣ 2 ␤ 1 -dependent cell adhesion and to bind isolated ␣ 2 ␤ 1 and the A-domain derived from the ␣ 2 subunit, which is known to be essential for the recognition of collagen by ␣ 2 ␤ 1 (11)(12)(13)(14). On this basis, we identified the sequence, GFOGERGVEGPOGPA, corresponding to residues 502-516 of the collagen I ␣ 1 (I) chain, as an ␣ 2 ␤ 1 binding locus in ␣ 1 (I)CB3 (15,16). Here we report further that ␣ 2 ␤ 1 recognition resides totally within the sequence GFOGER, that the glutamyl residue cannot be replaced by an aspartyl residue, that recognition of this sequence is entirely dependent upon the presence of a triple-helical conformation, and that the same sequence is recognized by the ␣ 1 A-domain of integrin ␣ 1 ␤ 1 . Moreover, integrin ␣ 2 ␤ 1 -mediated platelet and other cell adhesion to collagen can be completely inhibited by triplehelical GFOGER.

EXPERIMENTAL PROCEDURES
Materials-Collagen type I, for use in cell adhesion studies and solid-phase binding assays, was purified from bovine skin, following limited pepsin digestion, as described previously (9,17). Collagen type IV from human placenta was from Sigma-Aldrich Co. Ltd., Poole, Dorset, UK.
For the experiment, cells were harvested with trypsin/EDTA, fetal bovine serum was added at 1:1 v/v, and cells were recovered by centrifugation. After four washes with Dulbecco's phosphate-buffered saline (Ca 2ϩand Mg 2ϩ -free; PBS), the pellet was suspended at a concentration of 0.3 ϫ 10 6 cells/ml in adhesion buffer (TBS plus glucose, 0.9 g/l) containing 1 mM Mg 2ϩ or 2 mM EDTA, as necessary. Immulon-2 multiwell plates were coated with collagen or peptide, normally with 100 l of a 10 g/ml solution in 0.01 M acetic acid for 1 h at 20°C, blocked with BSA (100 l of a 1 mg/ml solution in PBS), and then washed four times with PBS. 100 l of cell suspension were then added and adhesion measured at 20°C (to maintain the peptides in a triple-helical conformation) at times specified later (see "Results"). Unattached cells were counted using a Coulter Counter (model ZF) and adhesion was calculated as the number of adherent cells expressed as a percentage of the total cell count. Assays were undertaken in triplicate, and the data were expressed as the mean Ϯ S.D. BSA-coated wells were used to determine nonspecific background adhesion. HT 1080 cells attached rapidly to collagen, reaching a maximum value of about 80% within 30 min. Rugli cells initially attached rapidly but only reached a maximum adhesion, typically about 65%, after 90 min. Results of experiments are expressed on the basis of a value of 100% for collagen. Data presented are representative of three similar experiments. Cells were preincubated with antibody or peptide, when testing for inhibition, for 15 min.
Platelet Adhesion-Adhesion was determined colorimetrically (19). Washed platelets, from platelet-rich plasma as described previously (20), were suspended routinely at 1.0 -1.5 ϫ 10 7 /ml in adhesion buffer, TBS containing 0.1% BSA (Sigma A7638), and the suspension rested for 30 min prior to use. Mg 2ϩ or EDTA were added to 2 mM as required. When testing mAbs or peptides for inhibitory activity, the platelet suspension was preincubated with antibody for 15 min. Immulon-2 96-well plates were coated with collagen or peptide, routinely at 10 g/ml, and then blocked with BSA (15). 50 l of platelet suspension were added to each well, and plates were incubated for 60 min at 20°C (to ensure retention of triple-helical conformation).Unbound platelets were then discarded, and the wells were washed three times with 200 l of adhesion buffer. 150 l of lysis buffer (0.1 M citrate, pH 5.4, containing 0.1% Triton X-100 and 5 mM p-nitrophenyl phosphate) was added to each well. Reaction was terminated after 60 min by addition of 100 l of 2 M NaOH, and plates were read at 405 nm using an automated plate reader (Emax; Molecular Devices). Assays were made in triplicate and the results expressed as the mean Ϯ S.D., relative to a value of 1.0 for collagen. Data are representative of three repeat experiments. In a typical experiment, a platelet concentration of 1.25 ϫ 10 7 /ml gave an A 405 ϳ1 with collagen I as substrate. This corresponds to adhesion of about 15% when expressed as a fraction of the number of cells applied.
Integrin ␣ 2 ␤ 1 -The integrin was extracted from human platelet membranes and purified by affinity chromatography on collagen-Sepharose as described previously (15). Homogeneity was established by polyacrylamide gel electrophoresis, and identification as ␣ 2 ␤ 1 was by immunoprecipitation and Western blotting (15). Protein concentration was determined with a Micro BCA Protein Assay reagent Kit (Pierce and Warriner (UK) Ltd., Chester, UK). The protein was biotinylated using an Amersham Pharmacia Biotech ECL biotinylation module, according to the manufacturer's instructions. The suitability of the biotinylated product for use in solid-phase assays was demonstrated in our earlier work (15,16).
Recombinant Integrin ␣ 1 and ␣ 2 A-Domains-The production of recombinant human ␣ 1 and ␣ 2 A-domains and their isolation as A-domain glutathione S-transferase (GST) fusion proteins has been described previously (13,21). The suitability of these materials for use in solidphase binding assays has been established in our earlier work (13,15,16,21).
Solid-phase Binding Assays-Assays were performed as described previously (13,15,16,21). Briefly, 96-well Maxisorp plates (Nunc) were coated with collagen or peptide at 10 g/ml for 1 h, blocked for 30 min with 200 l of 50 mg/ml BSA (Sigma A4503) in TBS, and then washed three times with TBS containing 1 mM Mg 2ϩ and 1 mg/ml BSA (Sigma A7638). Ligand dissolved in 100 l of TBS containing 1 mg/ml BSA (A7638) and 2 mM Mg 2ϩ or 5 mM EDTA as required, either biotinylated ␣ 2 ␤ 1 (1 g/ml) or A-domain fusion protein (5 g/ml), was applied to wells. Where required, antibody or peptide, as detailed later, was added to ligand solutions 15 min prior to their application to wells. Plates were incubated for 90 min at 20°C, then washed three times as above. Bound biotinylated ␣ 2 ␤ 1 was detected with streptavidin-horseradish peroxidase (1:1500 in TBS), bound A-domain fusion protein with horseradish peroxidase-linked anti-GST antibody (Sigma) at 1:500 in TBS. Color was subsequently developed using a TMB substrate kit (Pierce) according to the instructions of the manufacturer and plates read at 450 nm with the E-max plate reader. Assays were performed in triplicate and results expressed as the mean Ϯ S.D., relative to a value of 1.0 for collagen. Data as shown are representative of at least three repeat experiments.
Peptides-Synthesis of the seven overlapping peptides CB3(I)-1 to -7, based on the sequence of the collagen type I fragment ␣1(I)CB3, together with peptide CB3(I)-5/6-GAR containing the overlap sequence (GFOGERGVEGPOGPA) between CB3(I)-5 and -6, except that the glutamyl residue in the triplet GER has been replaced by an alanyl residue, has been described by us earlier (15). The peptide CB3(I)-5/6-GPP (formerly designated 5/6-HYP2) containing the overlap sequence within repeat GPPs, rather than repeat GPOs as in the above peptides, has also been described in our earlier work (16). Peptides made in the current study are listed in Table I. Sequences of interest were incorporated within repeat GPP triplets (to ensure triple-helicity) rather than repeat GPO triplets because the platelet collagen receptor glycoprotein VI does not recognize the GPP sequence (16), and therefore platelet adhesion would not be complicated by the occurrence of glycoprotein VI-mediated adhesion. Peptides were synthesized as C-terminal amides on TentaGel R RAM resin in a PerSeptive Biosystems 9050 Plus Pep-Synthesizer exactly as described in our earlier studies (15,16). Peptides were purified by reverse-phase high performance liquid chromatography on a column of Vydac 219TP101522 using a linear gradient of 5-45% acetonitrile in water containing 0.1% trifluoroacetic acid. Fractions containing homogeneous product were identified by analytical high performance liquid chromatography on a column of Vydac 219TP54, pooled, and freeze-dried. All peptides were found to be of the correct theoretical mass by mass spectrometry. The triple-helical stability of each peptide was assessed by polarimetry as described previously. The melting temperature (T m ) was calculated by fitting a theoretical melting equation to the melting curve by nonlinear regression. 3 Values are given in Table I.

RESULTS
Identification of GFOGER as an ␣ 2 ␤ 1 -recognition Sequence-In our previous study, we showed that platelet and HT 1080 cell adhesion to the collagen I sequence GFOGERGVEG-POGPA, incorporated within repeat GPP triplets to ensure a triple-helical conformation (peptide CB3(I)-5/6-GPP, Table I), was divalent cation-dependent and totally mediated by ␣ 2 ␤ 1 . The sequence bound both isolated ␣ 2 ␤ 1 and recombinant ␣ 2 A-domain, and binding was inhibited by both EDTA and anti-␣ 2 mAbs. The GER triplet appeared essential for activity because replacement of the glutamyl or arginyl residue by alanine eliminated recognition of ␣ 2 ␤ 1 (15,16). In the present study, we have examined the effect of removal from the C terminus of CB3(I)-5/6-GPP of GPA (peptide desGPA-GPP in Table I), GPOGPA (desGPOGPA-GPP, Table I), and GVEG-POGPA (peptide GFOGER-GPP, Table I) and of GFO from the N terminus (peptide desGFO-GPP, Table I). Removal of the C-terminal sequences had no significant effect on the level of adhesion of platelets or HT1080 cells nor on the extent of binding of the isolated intact ␣ 2 ␤ 1 integrin or the recombinant ␣ 2 A-domain. However, removal of GFO caused a marked loss of activity in every case (Fig. 1). This indicates that ␣ 2 ␤ 1 recognition resides entirely within the sequence GFOGER. Like the parent peptide CB3(I)-5/6-GPP, peptide GFOGER-GPP supported platelet adhesion and exhibited ␣ 2 ␤ 1 binding as good as that to collagen I, whereas binding of the ␣ 2 A-domain was consistently higher than that to the collagen (Figs. 1 and 2). Platelet adhesion and the binding of ␣ 2 ␤ 1 and ␣ 2 A-domain, as for collagen, was divalent cation-dependent and strongly inhib-ited by the anti-␣ 2 mAb 6F1 ( Fig. 2; data for ␣ 2 ␤ 1 not shown). HT 1080 cell adhesion to GFOGER-GPP, comparable with that to collagen (Fig. 1), was also totally divalent cation-and ␣ 2 ␤ 1dependent (data not shown).
Recognition of ␣ 1 ␤ 1 -The A-domain of the ␣ 1 subunit of integrin ␣ 1 ␤ 1 is known to be essential for ␣ 1 ␤ 1 binding to collagens, including collagen I (21,22), just as the ␣ 2 A-domain is required for ␣ 2 ␤ 1 collagen binding. In this study, we have tested ␣ 1 A-domain binding to the seven overlapping peptides CB3(I)-1 to -7 described in our earlier work based on the collagen I fragment ␣ 1 (I)CB3 (15). Sequences of these peptides are given in Table II. The ␣ 1 A-domain displayed precisely the same specificity of binding as observed previously for the ␣ 2 A-domain (15). Good binding, as good as to collagen, was observed with peptides CB3(I)-5 and -6, but little if any with any of the other peptides (Fig. 3). As previously shown for the ␣ 2 A-domain, ␣ 1 A-domain binding could be attributed to the sequence GFOGERGVEGPOGPA representing the overlap between peptides CB3(I)-5 and -6 because the peptide CB3(I)-5/ 6-GPP (containing the overlap sequence) bound the ␣1 A-domain at a level consistently higher than that to collagen (Figs. 4 and 5). The location of the ␣1 A-domain binding activity within the overlap sequence to GFOGER was established by demonstrating binding to peptides desGPA-GPP, desGPOGPA-GPP, and GFOGER-GPP comparable with that to CB3(I)-5/6-GPP, but much reduced binding to desGFO-GPP (Fig. 4). As with the ␣ 2 A-domain, binding of ␣ 1 A-domain to peptide CB3(I)-5/6-GPP or peptide GFOGER-GPP, as to collagen (21,23), was divalent cation-dependent and was inhibited by antibody directed to the A-domain (Figs. 4, 5, and 6). Binding of the ␣ 1 A-domain to either peptide, while blocked by anti-␣ 1 A-domain, was unaffected by anti-␣ 2 A-domain mAb. Conversely, binding of ␣ 2 A-domain was inhibited by anti-␣ 2 A-domain mAb, not by the anti-␣ 1 A-domain antibody. Results for CB3(I)-5/6-GPP are shown in Fig. 5.
Confirmation of the recognition of GFOGER by ␣ 1 A-domain was obtained with Rugli cells which express the integrin ␣ 1 ␤ 1 that mediates their adhesion to collagen (24). Testing the ␣1(I)CB3-based peptides, CB3(I)-1 to -7, we found, as for ␣ 2 ␤ 1mediated cell adhesion (15,16), preferential adhesion to CB3(I)-5 and -6 that was divalent cation-dependent and oc- Peptide CB3(I)-5/6-GPP containing the 15-mer ␣ 2 ␤ 1 -recognition site, shown in bold, corresponding to residues 502-516 of the ␣1(I) chain of type I collagen has been described by us previously (16). Other peptides listed were synthesized in the present study. The sequence of interest (shown in bold) was incorporated within repeat GPP triplets, as shown, to ensure triple helicity. A cysteine was incorporated at either end to allow cross-linking as desired.

Peptide
Sequence T m CB3(I)-5/6-GPP GPC(GPP) 5 GFOGERGVEGPOGPA(GPP) 5  curred at a level comparable with that to collagen (Fig. 3). Good adhesion, as good as to collagen, occurred to CB3(I)-5/6-GPP, confirming the presence of an ␣ 1 ␤ 1 -recognition site in the overlap sequence, GFOGERGVEGPOGPA (Fig. 7). Adhesion to collagen was completely blocked by the anti-␣1 subunit mAb Ha31/8 (tested at 15 g/ml), and that to CB3(I)-5/6-GPP was inhibited by 80% (data not shown). The location of the ␣ 1 ␤ 1 recognition site to GFOGER was substantiated on the basis that adhesion to GFOGER-GPP was as good as that to collagen or CB3(I)-5/6-GPP (Fig. 6), and the relatively poor adhesion to desGFO-GPP, and to CB3(I)-5/6-GAR in which the glutamyl residue in the GER triplet has been replaced by an alanyl residue (Fig. 7). Adhesion of Rugli cells to GFOGER-GPP, as to collagen, was divalent cation-dependent and inhibited by the anti-␣1 mAb Ha31/8 (data not shown).
Structural Specificity of GFOGER for ␣ 1 ␤ 1 and ␣ 2 ␤ 1 Recognition-The conservative replacement of the glutamyl residue in GFOGER-GPP by an aspartyl residue caused a marked loss of integrin recognition. GFOGDR-GPP (see Table I) failed to support platelet adhesion ( Fig. 2A)  (data not shown), nor was the peptide able to bind ␣ 2 ␤ 1 (data not shown) or ␣ 2 or ␣ 1 A-domain (Figs. 2B and 6). Rugli cell adhesion to the peptide was only around one-quarter of that to collagen I (data not shown). On the other hand, replacement of arginine by lysine (peptide GFOGEK-GPP; Table I) caused only a partial loss (about 50%) of cell adhesion ( Fig. 2A) and integrin binding (data not shown) although A-domain binding was largely eliminated (see Figs. 2B and 6). Requirement for the Triple-helical Conformation-Peptide GFOGER-GAP (Table I) contains the integrin A-domain recognition sequence GFOGER in repeat GAP rather than GPP triplets. As anticipated, polarimetry indicated the absence of any triple-helical structure. In marked contrast to GFOGER-GPP, the peptide failed to support adhesion of platelets ( Fig.   2A) or HT 1080 cells (data not shown) and exhibited complete absence of binding of isolated ␣ 2 ␤ 1 (data not shown) or ␣ 1 (Fig.  6) and ␣ 2 A-domains (Fig. 2B), emphasizing the crucial role of the triple-helical conformation for recognition of ␣ 1 ␤ 1 and ␣ 2 ␤ 1 A-domains by collagen.
Inhibition of Cell Adhesion by Peptide GFOGER-GPP-GFOGER-GPP was a potent inhibitor of platelet adhesion to collagens I and IV with an IC 50 for collagen I of approx. 75 g/ml or 7 M; M r 11,112 (Fig. 8A), and of ␣ 2 ␤ 1 and ␣ 2 A-domain binding to these collagens (IC 50 ϳ30 g/ml or 3 M; Fig. 8, B-D), confirming the identity of GFOGER as an ␣ 2 ␤ 1 integrin recognition site and establishing the crucial importance of this sequence as a platelet ␣ 2 ␤ 1 -binding locus in collagens I and IV. The peptide also fully inhibited ␣ 2 ␤ 1 -mediated adhesion at 30 min of HT 1080 cells to collagens I and IV with an IC 50 of ϳ450 g/ml or 40 M (data not shown). GFOGER-GAP, tested up to 3 mg/ml (300 M; M r 10,176) exhibited no inhibitory activity, confirming the essential requirement of the triple-helical conformation for recognition of GFOGER (Fig. 8, A and B).
GFOGER-GPP also totally inhibited ␣ 1 A-domain binding to collagen I (Fig. 8C), and in accord with this, ␣ 1 ␤ 1 -dependent adhesion (at 30 min) of Rugli cells to this collagen was strongly inhibited (80%) by the peptide (IC 50 around 40 M; data not shown). By contrast, GFOGER-GPP had little effect on ␣ 1 Adomain binding to collagen IV (Fig. 8D) or the attachment of Rugli cells, even when tested up to 3 mg/ml (275 M). DISCUSSION We previously identified the sequence GFOGERGVEG-POGPA, residues 502-516 of the ␣1(I) chain of collagen I, as an integrin ␣ 2 ␤ 1 recognition sequence (15,16). Here we show that recognition resides within the six-residue sequence GFOGER. There appears to be an absolute requirement for the glutamyl residue because even the conservative replacement with an aspartyl causes a complete loss of recognition. The requirement for a glutamyl residue is of particular interest in view of crys-

FIG. 8. Inhibition by GFOGER-GPP.
A, inhibition by GFOGER-GPP of platelet adhesion to collagen I (Ⅺ) and collagen IV(ࡗ) and lack of inhibition (to collagen I) by GFOGER-GAP (E). B, inhibition of ␣ 2 ␤ 1 binding to collagen I by GFOGER-GPP (Ⅺ) and lack of inhibition by GFOGER-GAP (ࡗ). C, inhibition by GFOGER-GPP of ␣ 1 A-domain (Ⅺ) and ␣ 2 A-domain (ࡗ) binding to collagen I. D, inhibition by GFOGER-GPP of ␣ 1 A-domain (ࡗ) and ␣ 2 A-domain (Ⅺ) binding to collagen IV. Each point represents the mean of three determinations. Missing error bars were too close to show. tallographic evidence that predicts recognition by the ␣ 2 ␤ 1 A-domain of a glutamate sidechain in collagen (25). In the peptide desGFO-GPP, GFO is in effect replaced by GPP (see Table I), and this is shown here to be inactive. We have also found that a peptide akin to desGFO-GPP, but in which the repeat GPPs are replaced by repeat GPOs, so that in effect GFO becomes GPO, is also inactive. 4 This indicates that it is the identity of the residue in the X position in the GXY triplet, namely the phenylalanyl residue in the GFO triplet, that is crucial for recognition.
Our studies here have shown that the sequence GFOGER is recognized equally well by both ␣ 1 and ␣ 2 A-domains and that the sequence can support both ␣ 1 ␤ 1 -and ␣ 2 ␤ 1 -dependent cell adhesion. In each case, recognition requires the same structural features. In particular, the glutamyl residue is essential for recognition of both A-domains. It is of interest that the ␣ 1 A-domain has a crystal structure very similar to that of the ␣ 2 A-domain (26). Recognition of GFOGER by two different integrins, ␣ 1 ␤ 1 and ␣ 2 ␤ 1 , is perhaps akin to the recognition of the same RGD sequence in fibronectin by at least eight different integrins, including ␣ 5 ␤ 1 , ␣ V ␤ 1 , ␣ V ␤ 3 , and ␣ IIb ␤ 3 (2, 3).
Interestingly, substitution of R by K in GFOGER led to a loss of recognition of the ␣ 2 A-domain, but ␣ 2 ␤ 1 -mediated cell adhesion was only reduced by around one-half. The reason for this is unclear but may suggest that structural requirements for recognition of the isolated A-domain are more stringent than those required for recognition of the domain within the intact integrin located in the cell membrane.
Triple-helical GFOGER is a potent inhibitor of ␣ 2 ␤ 1 -mediated adhesion of platelets and HT1080 cells to collagen I and fully inhibits binding of the isolated ␣ 2 ␤ 1 integrin and the recombinant ␣ 2 A-domain to this collagen. The same holds true for collagen IV, and it is of interest that the major cell-binding domain of collagen IV possessing both ␣ 1 ␤ 1 -and ␣ 2 ␤ 1 -binding sites (7) contains a GFOGER sequence in the ␣ 1 (IV) chain (see residues 405-410 in Ref. 7;Ref. 27). The collagen I fragment ␣ 1 (I)CB3 containing the GFOGER sequence under consideration here supports ␣ 2 ␤ 1 -mediated platelet adhesion as well as the parent collagen and better than other collagen I-derived fragments (9,10). Our results indicate that GFOGER is a major ␣ 2 ␤ 1 recognition site in collagens and is responsible for their interaction with cells via the integrin ␣ 2 ␤ 1 . However, some collagen I fragments, for example, bovine ␣ 1 (I)CB7 and ␣ 1 (I)CB8, support some ␣ 2 ␤ 1 -mediated platelet adhesion (9) despite the absence of the sequence GFOGER (28). Furthermore, ␣ 2 ␤ 1 -dependent cell adhesion to bovine collagen III is totally inhibited by the peptide GFOGER-GPP 4 although the GFOGER sequence is not present in collagen III (28). Presumably, other sequences are able in some measure to support ␣ 2 ␤ 1 -mediated cell adhesion.
Binding of ␣ 1 A-domain to collagen I, like that of the ␣ 2 A-domain, is fully inhibited by triple-helical GFOGER-GPP, and ␣ 1 ␤ 1 -dependent adhesion of Rugli cells to this collagen is mostly prevented, suggesting that GFOGER (or sequences of similar affinity) play a major role in mediating ␣ 1 ␤ 1 -dependent cell adhesion to collagen I. However, ␣ 1 A-domain binding to collagen IV is only relatively poorly inhibited by GFOGER-GPP, which is in accord with our finding that Rugli cell adhesion to collagen IV is not inhibited by GFOGER-GPP. This indicates that a sequence of higher affinity than GFOGER must mediate ␣ 1 ␤ 1 -dependent cell adhesion to collagen IV, despite the presence of GFOGER, and this accords with the data of others (7) that ␣ 1 ␤ 1 -binding to collagen IV involves the residues aspartyl 461 in the ␣ 1 (IV) chain and arginyl 461 in the ␣ 2 (IV) chain.
In summary, we find that ␣ 1 and ␣ 2 A-domains each recognize the sequence GFOGER, and our data are consistent with the proposal that GFOGER is responsible in large part for ␣ 2 ␤ 1 -dependent cell recognition by collagens I and IV and, conceivably, other collagens. We find too that the sequence may mediate ␣ 1 ␤ 1 -dependent cell adhesion to collagen I but plays no significant role in such adhesion to collagen IV.