Cross-reactive immunodeterminants on Streptococcus sanguis and collagen. Predicting a structural motif of platelet-interactive domains.

Cross-reactive immunodeterminants on a fibril-associated surface antigen of Streptococcus sanguis and types I and III collagen participate in the induction of aggregation of human platelets. To further understand the basis for this apparent molecular mimicry, antitype-specific collagen antibodies, anti-KPGEPGPK (an analogue of platelet-interactive domains on collagen) and a panel of KPGEPGPK-like synthetic peptides were used as probes. When collagen or S. sanguis cells were pretreated with the anti-collagen antisera, the induction of aggregation of platelet-rich plasma was greatly delayed or abrogated. These anti-collagen antibodies also neutralized KPGEPGPK and purified S. sanguis platelet-interactive antigens as inhibitors of S. sanguis or collagen-induced aggregation of platelets in plasma. In immunoblot analyses, these anti-collagen antibodies reacted with S. sanguis platelet-interactive antigens. Additionally, antisera against the platelet-interactive antigen of S. sanguis selectively reacted with undigested type I collagen and with fragments CB3 and CB6 of cyanogen bromide-treated type I collagen. Finally, when platelets were pretreated with synthetic peptides containing specific amino acid substitutions within the KPGEPGPK sequence, the time to onset of platelet-rich plasma aggregation by both agonists was altered. The hierarchical pattern of responses of platelets to these peptides and predictions of the structural changes produced by simulated insertions of each peptide into the CB4 sequence of type III collagen suggested conformational requirements for interactions with platelets. Thus, these data show that cross-reactive immunodeterminants of S. sanguis and collagen induce platelet aggregation. The platelet-interactive domains are predicted to be characterized by a structural motif with the consensus sequence X-P-G-E-P/Q-G-P-X.

Cross-reactive immunodeterminants on a fibril-associated surface antigen of Streptococcus sanguis and types I and I11 collagen participate in the induction of aggregation of human platelets. To further understand t h e basis for this apparent molecular mimicry, antitype-specific collagen antibodies, anti-KPGEPGPK (an analogue of platelet-interactive domains on collagen) and a panel of KPGEPGPK-like synthetic peptides were used as probes. When collagen or S. sanguis cells were pretreated with the anti-collagen antisera, the induction of aggregation of platelet-rich plasma was greatly delayed or abrogated. These anti-collagen antibodies also neutralized KPGEPGPK and purified S. sanguis platelet-interactive antigens as inhibitors of S. sanguis or collagen-induced aggregation of platelets in plasma. In immunoblot analyses, these anti-collagen antibodies reacted with S. sanguis platelet-interactive antigens. Additionally, antisera against the plateletinteractive antigen of s. sanguis selectively reacted with undigested type I collagen and with fragments CB3 and CB6 of cyanogen bromide-treated type I collagen. Finally, when platelets were pretreated with synthetic peptides containing specific amino acid substitutions within the KPGEPGPK sequence, the time to onset of platelet-rich plasma aggregation by both agonists was altered. The hierarchical pattern of responses of platelets to these peptides and predictions of the structural changes produced by simulated insertions of each peptide into the CB4 sequence of type I11 collagen suggested conformational requirements for interactions with platelets. Thus, these data show that cross-reactive immunodeterminants of S. sanguis and collagen induce platelet aggregation. The platelet-interactive domains are predicted to be characterized by a structural motif with the consensus sequence X-P-G-E-P/Q-G-P-X.
Proteinaceous fibrils on the surface of cells of Streptococcus sanguis induce human platelets in plasma to aggregate in vitro (1)(2)(3). Platelet-adhering (Adh) and aggregation-inducing (Agg) phenotypic variants occur across strains (4). The Agg+ phenotype is associated with a platelet-interactive immuno-* This work was supported in part by National Institutes of Health Grants DE05501 and DE08590 and individual Dentist-Scientist Award DE00224. 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.
$ To whom correspondence should be addressed Dept. of Preventive Sciences, 17-164c Moos HS Tower, 515 Delaware St., SE, Minneapolis, MN  determinant located in a 23-kDa segment of a 65-kDa cell wall-bound protein fragment (5). This immunodeterminant is also identified on exported and released precursor proteins of 111 and 115 kDa, respectively, from cultured protoplasts (6). When disassembled from the surface fibrils on the cell wall, these proteins or protein fragments do not act as agonists, but as inhibitory haptens for S. sanguis and collagen-induced platelet aggregation (2,7,8).
Monospecific antibodies against the S. sanguis plateletinteractive immunodeterminant (class I1 antigen) react with cells of S. sanguis and types I or I11 collagen to inhibit induction of platelet aggregation (7). These antibodies are specifically neutralized by purified platelet-interactive proteins or fragments from S. sanguis and a synthetic peptide (KPGEPGPK),' which also acts as an inhibitory hapten and was patterned from platelet-interactive domains on collagen (7,8).
To confirm the molecular mimicry of collagen, antibodies against type-specific collagens and anti-KPGEPGPK were reacted with platelet-interactive immunodeterminants of S. sanguis. In addition, antibodies specific for the S. sanguis antigen were reacted with cyanogen bromide digests of type I collagen. A panel of peptides similar to KPGEPGPK was synthesized with selected amino acid substitutions. The specificity of cross-reactivity was explored using these congener octapeptides as inhibitors of S. sanguis-induced platelet aggregation. Structural predictions based upon the hierarchical pattern of responses suggest common structural features of the platelet-interactive domain of the class I1 antigen of S. sanguis and collagen.

EXPERIMENTAL PROCEDURES
Preparation of Antibody Probes-Monospecific rabbit anti-types I (titer 1:200, by ELISA), I11 (titer 1:250), and IV (titer 1:150) collagen were obtained from Chemicon International (El Segundo, CA) and tested to verify specificity (data not shown). Rabbit antiserum against S. sanguis I 133-79 whole cells (titer 1:128, by immunodiffusion), 23-kDa fragment of the platelet interactive antigen (titer 1:32), or nonspecific hyperimmune serum were prepared in female New &aland White rabbits as described previously (2,3). Monospecific antibody (titer 1%) against KPGEPGPK was prepared as above using a bovine serum albumin conjugate as immunogen and an ovalbumin with carbodiimide (9). Briefly, BSA or ovalbumin (30 mg) and conjugate for screening. Conjugates were prepared by cross-linking KPGEPGPK (15 mg) were dissolved in 1 ml of distilled H20, stirred, and l-ethyl-5-(3-dimethylaminopropyl)-carbodiimide HCl was added slowly. The peptide-carrier conjugate was purified by gel filtration The abbreviations used are: KPGEPGPK, NH,-Lys-Pro-Gly-Glu-Pro-Gly-Pro-Lys-COOH; PRP, platelet-rich plasma; ELISA, enzyme-linked immunosorbent assay; BSA, bovine serum albumin; HPLC, high performance liquid chromatography; SDS, sodium dodecyl sulfate; PBS, phosphate-buffered saline. and confirmed by amino acid analysis. For use in immunoblot analyses, IgG was prepared from rabbit antisera by precipitation with saturated ammonium sulfate (lo), followed by ion exchange chroma-Synthetic Peptides-The platelet-interactive collagen-like octapeptide KPGEPGPK (7), originally described by Legrand et al. (8), was synthesized by the Microchemical Facility of the University of Minnesota. Purity was confirmed by amino acid analysis and HPLC. A panel of 24 additional congener peptides, with selected amino acid substitutions, was synthesized by the Microchemical Facility on a parallel peptide synthesizer (model 350, Advanced Chem Tech, Louisville, KY) and purity-confirmed by HPLC.
Platelet Aggregometry-Platelet aggregometry was performed as described (1,2) with freshly isolated citrated human platelet-rich plasma (PRP) from a single healthy medication-free donor according to procedures approved by the Committee on the Use of Human Subjects in Research of the University of Minnesota. Fifty microliters of soluble type I collagen (0.05 mg/ml, final concentration), prepared as described previously (7), or standardized suspensions of cells of S. sanguis (2 X 10scells/ml,final concentration) were added to stirred (1200 rpm), warmed (37 'C) PRP (0.45 ml adjusted to 2 X 10' platelets/ml) in a recording aggregometer. The lag time to onset and extent of platelet aggregation were measured.
Platekt Adhesion-The platelet-bacterial adhesion assay was performed as described previously (1,7). In some experiments, cells of S. sanguis I 133-79 were pretreated for 10 min at 37 "C with rabbit antibodies against type-specific collagens.
Immunoblots-To demonstrate reaction with anti-collagen antisera, minimal tryptic digest and purified platelet-interactive antigens were prepared from cells of S. sanguis and subjected to electrophoresis on 5-10% gradient SDS-polyacrylamide gels (5, 11). Unstained gels were electroeluted onto nitrocellulose, quenched with hemoglobin, and incubated with specific rabbit antisera (5). In some experiments the antisera had been neutralized with the 23-kDa fragment of the S. sanguis platelet-interactive protein. The binding of rabbit IgG antibodies to specific antigens was detected by incubation with goat antirabbit IgG, which was conjugated to alkaline phosphatase and reacted with 5-bromo-4-chloro-3-indoyl phosphate in the presence of nitro blue tetrazolium (12).
To further demonstrate the immunologic cross-reaction of collagen and purified S. sanguis platelet-interactive antigen, cyanogen bromide fragments of type I collagen were prepared as described previously (13). Briefly, type I collagen was dissolved in formic acid (70%), reacted with cyanogen bromide (w/w, 2:1, CNBr:collagen) at 30 "C for 4 h, diluted 10-fold with distilled HzO, and lyophilized. The resulting fragments were subjected to electrophoresis on 10-15% graaient SDS-polyacrylamide gels (5, ll), electroeluted (5), quenched, incubated with rabbit IgG antibodies specific for the 23-kDa fragment of the S. sanguis platelet interactive antigen, and stained as described above.
Peptide Structure Predictions-Based on the inhibitory activity of the congeners of KPGEPGPK, a hierarchy of activity was developed. The KPGEPGPK sequence within CB4 of collagen I11 was then replaced with each congener sequence, and predictions of secondary structures were performed using MELPROT (Biochemistry Department, Melbourne University; Ref. 14), which incorporated the predictive algorithms of Chou and Fasman (15,16), Hopp and Woods (17), Kyte and Doolittle (18), and Karplus and Schulz (19). tography (5).

Effects o n Platelet
Interactions-To establish that collagen and cells of S. sanguis share platelet-interactive immunodeterminants, these agonists were pretreated with antibodies against type-specific collagens. The anti-types I, 111, and IV collagen and anti-KPGEPGPK antisera each delayed the onset or abrogated the platelet response to soluble type I collagen or S. sanguis (Table I). When aggregation occurred, the rate and extent were unaffected (data not shown). S. sanguis and collagen retained their ability to induce aggregation of platelets in plasma when pretreated with unrelated hyperimmune rabbit serum or PBS. Pretreatment of cells of S. sanguis with anti-collagen antibodies did not affect adhesion to platelets (data not shown).
S. sanguis Antigens React with Anti-collagen Aatibodies-Anti-types I and 111 collagen antisera (0.1 ml) were then pretreated with 0.3 nmol of either the 23-kDa fragment of platelet-interactive antigen or collagen-like octapeptide. When these mixtures were incubated with cells of S. sanguis or collagen, the ability of the anti-collagen antibodies to abrogate platelet aggregation was neutralized ( (4) 19.7 f 3.9 (4) >20 (3) >20 (3)

TABLE I1
Znteractions between antisera and soluble proteinlpeptide preparations Antibody preparations (0.1 ml) were preincubated as described in the text with 0.3 nmol of KPGEPGPK or the S. sanguis 23-kDa platelet-interactive peptide. 0.05 ml of the mixture was then used to pretreat S. sanguis cells or soluble type I collagen. Values are means * S.D.; number of experiments in parentheses.
Anti-collagen antisera were reacted with antigens from cells of S. sanguis in immunoblots (Fig. 1). Anti-types I, 111, and IV collagen antisera each reacted (panels 2-4) with the 23-(lane B ) and 65 (lane C)-kDa forms of platelet-interactive antigen from purified preparations or in a minimal tryptic digest of whole cells (lane A ) . As expected, these antisera also reacted (data not shown) with the untrypsinized 115-kDa form of platelet-interactive antigen obtained from protoplast culture fluid (6). Pretreatment of these antisera with the 23-kDa fragment of the platelet-interactive antigen neutralized reactions in immunoblots with the 23-and 65-kDa forms of the S. sanguis antigens (panels [6][7][8]. Similarly, a single precipitin resolved in immunodiffusion reaction between the anti-collagen sera and the 65-kDa antigen and failed to appear when preceded by absorption with the 23 kDa fragment (data

Neutralization of the platelet-aggregation inhibitory effects of antitype I and anti-type III collagen antisera by the 23-kDa fragment of the S. sanguis datelet-interactive antigen and KPGEPGPK
Neutralization concentration Antisera Agonist s.

FIG. 1. Immunoblot demonstrates cross-reactions between
platelet-interactive tryptic peptide fragments of S. sanguis and anti-collagen antisera. All samples contained 10 pg of protein, were solubilized in 1% (w/v) SDS sample buffer, subjected to electrophoresis on 5-10% gradient gels, and immunoblotted as described under "Experimental Procedures." Lane A in each panel contains crude S. sanguis tryptic peptides; lanes B and C, the 23-kDa fragment and the 65-kDa platelet-interactive protein fragment, respectively. Panels I and 5 were stained for total protein with Aurodye colloidal gold, panels 2-4 were stained after reaction with rabbit IgG antibodies t o types I, 111, and IV collagen, respectively. Panels 6-8 were stained similarly after pretreatment of the antisera with the 23-kDa S. sanguis antigen.
not shown). Rabbit anti-octapeptide(-BSA) showed a reaction of identity in immunodiffusion with equimolar concentration of KPGEPGPK-ovalbumin (Fig. 2, well 1 ), the 23-(well 2 ) and 65-kDa (well 3) fragments, and the 115-kDa plateletinteractive protein (well 4 ) , but was unreactive with ovalbumin alone (data not shown). Similar molar concentrations of each were also required to inhibit platelet aggregation in response to cells of S. sanguis or collagen (Table IV). Inhibition was apparently specific, since ADP-induced platelet aggregation was unaffected by pretreatment with any of the four platelet-interactive preparations.

Immunodeterminants from Cyanogen Bromide Fragments of Type I Collagen React with Monospecific Polyclonal IgG
Antibodies Specific for the S. sanguis Platelet-interactive Antigen (Fig. 3)"This antisera reacted with purified 115-kDa S. In their cyanogen bromide digest, bands which correspond to CB3 and CB6 of type I collagen were reactive.

Effects of Collagen-like Peptides on Platelet Interactions-T o learn more about the structural requirements
of KPGEPGPK, a panel of similar peptides, with selected amino acid substitutions, were prepared. The S. sanguis-induced platelet aggregation responses were then compared after pretreatment (10 min, 37 "C) of platelets with peptides a t concentrations corresponding to the ID50 of the octapeptide (30 nM). As shown in Table V, inhibition was actually enhanced when the amino-terminal residue was altered and the carboxyl 7 residues left intact. Replacing the charged amino acid a t residue 4 with a neutral amino acid significantly reduced the inhibitory activity. In addition, substitutions which led to the reduction of the P-turn potential of residues 5 and 7 signifi-   cantly reduced the inhibitory activity of the peptide. Reduced &turn potential of residue 2 also decreased the inhibitory activity, but to a lesser degree.

DISCUSSION
Antibodies against platelet-interactive immunodeterminants on the cell surface of the oral bacterium S. sanguis were shown previously (7) to be cross-reactive with domains of similar function on types I and I11 collagen. It is now clear that antibodies against types I, 111, and IV collagen react with the platelet-interactive immunodeterminant on S. sanguis. This collagen cross-reactive immunodeterminant is found within a 23-kDa portion of a 65-kDa cell wall-derived fragment of a fimbrial protein (5) and also within a 115-kDa precursor released into the culture medium of protoplasts (6). This immunodeterminant is functional, since reaction of cells of S. sanguis with anti-collagen antibodies inhibits induction of platelet aggregation. Furthermore, antibodies against the octapeptide, KPGEPGPK, which mimics the platelet-interactive domain on cyanogen bromide fragment 4 (CB4) of type I11 collagen (7), also react with the S. sanguis determinant.
Antibodies against types I or I11 collagen are neutralized as inhibitors of S. sanguis or collagen induction of platelet aggregation by similar molar concentrations of KPGEPGPK or each purified molecular form of the streptococcal antigen. Since each form of the platelet-interactive determinant also shows a reaction of identity and similar R p in immunodiffusion reaction with antibodies against KPGEPGPK conjugated t o a protein carrier, a single structural or conformational analogue of the collagen peptide is probably contained per molecule of the streptococcal antigen.
We have now demonstrated that bovine collagen fragments al(I)-CB3 and al(I)-CB6 are also immunologically cross-reactive with the platelet-interactive domain on S. sanguis.
Also cross-reactive, the platelet-interactive fragment cyl(II1)-CB4 includes residues 479-486 from which KPGEPGPK was derived (Table VI). Highly conserved sequences were identified in residues 480-485 of bovine al(I)-CB3 and in residues 948-953 of bovine cyl(I)-CB6. The 6-amino acid sequence of cyl(I)-CB6 is identical to residues 480-485 of cyl(III)-CB4, except for a glutamine (residue 951) replacing proline (residue 483). More recently, platelet recognition sites have been reported in human types I and I11 collagen, specifically cyl(1)-CB7 and -CB8 and al(III)-CB3 and -CB4 (22). The structural motif common to these platelet-interactive domains in bovine and human types I and I11 collagen (21, 22) and S. sanguis was modeled by establishing a hierarchy of activity shown by members of a panel of congener peptides. Using MELPROT, each peptide sequence was inserted into cyl(III)-CB4, and the changes in structural predictions were compared with the activity with platelets. These data suggest that the platelet-interactive site requires a negatively charged amino acid surrounded by two regions of amino acids with Pturn potential. This interruption in the cy-helix of collagen may facilitate interactions with platelets. Amino acid substitutions that altered the predicted structure reduced the inhibitory activity of the peptide, a criterion for loss of platelet interactivity. Although the spacing of basic residues (lysine and arginine) was suggested to be a structural requirement for induction of platelet aggregation, this criterion does not apply to all platelet-interactive domains of human collagen (e.g. al(1)-CB8). When the cyanogen bromide fragments of human collagens were surveyed for domains that fit the predictive requirements, several domains were identified that fit our criteria (Table VI). All known platelet aggregationassociated sites do, therefore, fit the predicted structural motif of two regions with P-turn potentials that surround an acidic amino acid within a 7-residue portion of polypeptide. Therefore, the sequence Pro-Gly-Glu-(Pro/Gln)-Gly-Pro within the polypeptide may form the minimal structural motif for the cross-reactive platelet-interactive domains of collagen and S. sanguis. This S. sanguis immunodeterminant, which shows immunochemical and functional mimicry for collagen, must have access to a signal-transducing receptor on platelets to induce P R P aggregation. Since glycoprotein Ia-deficient platelets fail to aggregate in response to cells of S. sanguis and collagen (24), a common receptor may exist. Soluble and particulate collagens induce platelets to aggregate, but the response pathways differ (7). Although particulate type I collagen induces platelet aggregation within seconds (7,25,26), with differences in the dose affecting the rate and extent, soluble collagen and S. sanguis show only a dose-dependent effect on lag time to onset of aggregation. Glycoprotein Ia-IIa appears to interact directly with collagen in plasma, contributing to the transduction of signal involved in modulating the lag time (27). Indeed, the lag time to onset of aggregation may reflect the signal transduction mechanism of platelets in the absence of higher avidity adhesion events.
Strong adhesive interactions between glycoprotein Ia-IIa and collagen appear to require Mg2' and are demonstrable predominately in the absence of plasma (27). In these conditions, isolated platelet Ia-IIa (cy&) integrin receptor complex selectively binds the cyl(I)-CBS peptide of collagen (28). When Ca2+ and additional plasma constituents are present, other collagen domains may bind glycoprotein Ia-IIa. Indeed, the platelet aggregation domain of type I11 collagen is active in plasma and in the presence of Ca". Since this domain and a related immunodeterminant on type I collagen are cross-

-
Identification for each peptide.
As described under "Experimental Procedures." Dash designates a negative charged amino acid residue at position 4, 0 designates a neutral amino acid residue * Boldface designates substitution of residue compared with octapeptide sequence Al.
at position 4. e Blank spaces designate residues which lost or had reduced predicted @-turn potential. Bovine collagen al(I)-CB3 479-486 K P G E q G V P al(I)-CBG 947-954 s P G E q G P S al(III)-CB4 479-486 K P G E P G P K Human collagen al(I)-CB7 623-630 d r G E P G P p al(I)-CB8 191-198 v r G E P G P p al(III)-CB3 29-36 p P G E P G q a al(III)-CBI 479-486 K P G E P G P K a Platelet recognition sites reported in cyanogen bromide fragments of bovine (20, 21) and human (22) types I and 111 collagen.
' Residue numbers (23) identifying sequences which fit the predicted structural requirement of two regions with 8-turn potential surrounding an acidic amino acid within a 6-8-residue portion of the polypeptide.
Capital letters indicate residues which are identical with KPGEPGPK, the known platelet-interactive domain within al(II1)-CB4 of bovine collagen.
reactive with the platelet-interactive antigen on S. sanguis, platelet glycoprotein Ia, perhaps in complex with IIa, may bind collagen in physiological conditions to transduce signal for induction of aggregation, rather than promote adhesion.
Recently, Staatz et al. (29) have suggested that the tetrapeptide DGEA corresponding to residues 435-438 within al(I)-CB3, may be the recognition sequence for the a& integrin in their platelet-collagen adhesion assay. They also suggested that a peptide fragment of collagen, which contained residues 478-486 (KPGEPGPK), did not inhibit the binding of platelets to collagen substrate even at a peptide to platelet ratio 107-fold greater than that used in our analyses. Although peptide-peptide self-association may interfere with binding to the platelet receptor, Morton et al. (27) have suggested that platelet adherence to collagen may involve sites in the collagen molecule distinct from those associated with aggregation. Therefore, the DGEA-and KPGEPGPKlike peptides may represent two different domains within al(I)-CB3 that interact with collagen. Indeed, platelets may have several proteins that promote their adhesion to collagen (30), including glycoprotein IIb-IIIa (31) and glycoprotein IV (32, 33). Likewise, collagens have multiple domains that are adhesive for platelets (34)(35)(36)(37)(38)(39)(40). It must be learned if the several candidate platelet proteins that bind collagen and S. sanguis have discrete functions in physiological conditions.