Identification of the Heparan Sulfate Binding Sites in the Cellular Prion Protein

In this study we confirm direct interactions between recombinant PrP c and both heparin and HS and show that the binding of full-length recombinant GST::PrP c to heparin is weakened significantly in the presence of copper (II) ions. By competitive inhibition studies we reveal that dextran sulfate is a highly potent inhibitor of the PrP c -heparin interaction and show that 2- O sulfates of heparin are an essential component of the PrP c binding site (s). In a second series of experiments biosensor and ELISA analysis of GST-tagged recombinant peptides (covering the whole sequence of hamster PrP; 21, 49) and synthetic peptides have enabled us to identify three sequences in PrP with independent heparin / HS binding activity: amino acids 23-52, 53-93 and 110-128. At a higher concentration than that which affects heparin binding by intact PrP protein, copper (II) enhances the heparin-binding ability of isolated peptide 53-93, and in the presence of Cu (II) this peptide competes with full-length haPrP c for heparin binding, suggesting a central role in this interaction. The divergent actions of copper (II) ions on full-length PrP and on the peptide 53-93 are discussed. of conjugate (NA-9340, Amersham, for assays employing the polyclonal anti-GST) diluted 1:1000 in PBST. Colour was developed by addition of 100 µ l per well o -phenylene-diamine substrate solution (1 x 30mg OPD tablet (Sigma P-8412) per 75 µ l phosphate-citrate buffer (0.05M) pH 5.0 containing 0.03% sodium perborate (Sigma P-4922)). Absorbance was measured at 490nm after quenching the wells with 50 µ l 0.5M H 2 SO 4 . Results with monoclonal and polyclonal anti-GST antibodies were essentially identical. activity completely at 10 µ g/ml and inhibited only weakly at 100 µ g/ml. These results indicate an important role for 2- O -sulfate groups in the prion-heparin interaction. Interestingly, low concentrations of 2- O- desulfated heparin, like CS (Fig. 3c) appeared to promote PrP c binding to immobilised heparin. role for in direct HS the dependent of differing influence of Cu on the of and of the intact PrP serve illustrate the potential of over-reliance on peptide-derived data alone. sensitivity of PrP c -heparin binding to disruption with and polysulfate, proposed as prophylactic the demonstration of intimate role of 2- O sulfate in heparin recognition was also significant. Our data suggests that further investigation of the identified heparin-binding domains, and the potential specificity of carbohydrate binding sites in heparan sulfate, lead to further insights into the role of HS in the function of prion proteins.


Introduction
Considerable effort is being devoted to the identification of natural receptors for prion protein (PrP) both to refine our understanding of prion metabolism and to reveal potential targets for therapeutic intervention in the transmissible spongiform encephalopathies (TSEs). One group of potential co-factors are the glycosaminoglycans (GAGs) and in particular membrane proteins and extracellular matrix components elaborated with heparan sulfate (HS) sugar chains. Heparan sulfates (of which heparin is a heavily sulfated variant) are expressed on a wide variety of cell types, including those of neural origin, and modulate the activity of a wealth of cell-surface and extracellular signaling molecules such as growth factors and cytokines (1,2). In the TSE field it is known that HS proteoglycans (HSPGs) co-localize with the insoluble aggregates of the prion (PrP Sc ) that accumulate in the brain tissue of TSE affected animals (3)(4)(5). They also promote the formation of amyloid structures typical of Alzheimer's disease when co-injected into the brains of rats with A -protein (6). Furthermore, administration of select anionic compounds can restrict both tissue-specific accumulation of PrP Sc and the onset of neurodegenerative features in experimental models of prion disease. In this respect pentosan sulfate and DS500 (dextran sulfate with av. MW 500,000) are particularly effective (7)(8)(9). The basis for the anti-prion activity of these compounds is uncertain; one suggestion is that they compete with endogenous HSPGs for 5 significant amounts in the human brain and lacking residues amino-terminal to 111 or 112 (including the octapeptide-repeat), has no heparin-binding activity (37). Lastly, the PrP sequence aa53-93 is directly involved in an HSPG-dependent interaction between recombinant PrP and the 37kDa/67kDa laminin receptor precursor (LRP/LR) (21). Significantly, whereas Caughey reported heparin binding by PrP c that was independent of divalent ion concentration (14) a study by Brimacombe et al. describes heparin-sensitive binding of recombinant PrP to experimental nickel surfaces (15).
The only other region of PrP with supporting evidence for a role in GAG binding is a central, hydrophobic and amyloidogenic sequence between residues 106-126 (38,39). This sequence undoubtedly plays a major role in the biosynthesis of the protease-resistant form of PrP (PrP Sc ).
Cells expressing engineered variants of PrP c deleted for residues in this region of the molecule do not support the propagation of homologous PrP Sc , and PrP molecules deleted for this sequence cannot be converted to protease-resistant forms in vitro (40). Several groups have investigated the cytotoxic properties of the central region (41,42) and significantly a neurotoxic activity associated with a peptide corresponding to residues 106-126 is abrogated by soluble heparin and related GAGs (43). Though copper has been shown to affect the aggregation and neurotoxic properties of this region (24), the influence of copper ions on any interaction between this sequence and GAGs has yet to be reported.
Recombinant PrP has now been expressed and purified by several groups (44)(45)(46). These efforts have yielded proteins that resemble PrP c in that they fold into alpha-helical and -sheeted structures (20,46). They are also proteinase K sensitive (47), suggesting a lack of infectivity.

Biotinylation of heparin and HS for immobilisation
Three methods were chosen for biotin labeling of the heparins and HS used in this study: a.) GAGs were labeled by reaction of their aldehydic reducing groups using a method based on Nadkarni et al. (52) Briefly 50 nmol of saccharide was dissolved in 50 µl formamide containing 50 mM biocytin hydrazide and heated at 37 o C for 24 h. Heparin/ HS labeled in this way was used as substrates in occasional biosensor analysis of GST::haPrP c binding (e.g. Fig. 2b.) The second procedure is a modification of that described by Rahmoune et al. (53) and labels the free amino groups reported to occur occasionally along the length of heparin and HS. 30 µl of a 50 mM solution of biotin amidocaproate-NHS in DMSO was added to 50 nmol heparin / HS in dH 2 O.
The mixture was briefly mixed and left for 3 days at room temperature. This method was used to biotinylate heparin and HS for most biosensor analyses of GST::haPrP c and peptides. c.) A third procedure, used in ELISA studies of PrP and peptide binding is a modification of that of Lee and Conrad and also labels internally (54): 5 µmol heparin or HS was dissolved in 0.5 ml of sodium carbonate buffer pH 8.6 containing 15 µmol biotin amidocaproate 3-sulfo-NHS. The mixtures were shaken briefly to mix and left to stand at room temperature for 3 days. Free label and solvent was removed from all labelling reactions as follows: 5 volumes of pre-chilled ethanol were added to each tube and the sample stored at -20 o C for 30 minutes. The sample was next centrifuged (5 minutes, 13,000 rpm) and the ethanol decanted, chilled for a second time and recentrifuged. The precipitate from both stages was combined in 400µl dH 2 O and fractionated by gel filtration on three Hi-Trap columns (Pharmacia) arranged in series. Biotin-containing fractions were detected at 232 nm and the labeled GAG eluting in the void volume was retained.

Biosensor analysis of GST::PrP c and PrP peptide binding to heparin and HS
Biosensor analysis was performed on a Bia2000 instrument (BiaCore). Two channels of a recommendations (BiaCore). Briefly, the surface was treated with 50mM N-ethyl-N'- (3dimethylaminopropyl  GST::haPrP c was subsequently applied at 1µg/ml in PBS containing 3% BSA. After 2 hours incubation the plate was thoroughly washed (6 x in PBS/ 0.05% tween-20) and bound GST::haPrP c detected with polyclonal rabbit anti-GST/ anti-rabbit Ig (F(ab') 2 )-peroxidase conjugate as described above (conventional heparin-binding ELISA). In inhibition studies a selection of GAGs were included (100 µg/ml) during the GST::haPrP c incubation phase.

Recombinant hamster PrP c binds to immobilised heparin and HS
Surface plasmon resonance instruments are able to detect the mass changes that accompany the binding of soluble analyte to ligand immobilised on a detector surface. Initial experiments in this study investigated the binding of the GST::haPrP c fusion protein to immobilised heparin and HS. This treatment would be expected to have a strongly disruptive effect on heparin and HS directed binding and accounted for approximately 50% and 70% removal of residual GST::haPrP c from heparin and HS respectively.
Prion protein is known to bind copper (II) ions in vitro, and a linkage between divalent cation and polyanion binding activities has been demonstrated previously (15). This prompted us to undertake a second series of biosensor experiments in which GST::haPrP c was mixed with 10 µM copper (II) before injection onto heparin (Fig. 2a). This had three consequences: the absolute signal strength due to binding of PrP to heparin was reduced, the binding curve tended to plateau more quickly (i.e. tended towards equilibrium more rapidly) and bound protein was more were the most potent inhibitors of heparin binding (on a weight / volume basis). The heparins were substantially weaker and both the low molecular weight porcine intestinal heparin (LMW PIH) and two sources of HS were without effect. When copper (II) was added, the inhibitory activity of the GAGs/ polysaccharides was generally increased. Intact PIH (and also BLH, data not shown) approached DS and PPS in inhibitory effect and the binding of GST::haPrP to PIH could be weakly (but incompletely) disrupted by LMW PIH. These findings support the earlier evidence that Cu (II) reduces the interaction between full-length hamster PrP and heparin.
In a separate experiment human recombinant PrP c was allowed to bind to PI-heparin in the presence of a number of GAGs and modified heparins, added at 10 and 100 µg/ml (Fig. 4). Again BLH was a good inhibitor of heparin-binding, whereas PMHS had essentially no effect. Both

Binding of GST:: hamster PrP peptide fusions and synthetic PrP peptides (of human sequence) to immobilised heparin / HS
To identify regions of the PrP molecule with independent heparin / HS binding activity, recombinant GST::fusions of partial hamster PrP sequences collectively spanning the entire hamster prion sequence (21,44) were injected over a heparin-derivatised biosensor surface. When several independent batches of peptides were tested (in each case no more than 2 months after expression-purification), P2 (53-93) and P4n (110-128) consistently gave the most significant binding to heparin (Fig. 5a). P1, P3, P4, P5 and Px generally yielded weak biosensor responses, although significant binding of P1 and P5 to heparin was recorded in occasional batches. Note that the increase in signal resulting from P4n injection was not reversed by high salt, weak acid, alkaline or soluble heparin washes: procedures which should have disrupted all but the strongest electrostatic interactions. Peptides were tested in succession (as in Fig. 5a) and individually with no evidence of sample order-related enhancement or reduction of signal strength for any peptide.
In some experiments FGF-receptor (a known heparin-binding protein) was injected before and after a series of the PrP peptides and no change was observed in the extent of binding of this protein (data not shown).
Subsequent experiments revealed that the heparin-binding activity of both GST:: P2 (53-93) and GST:: P4n (110-128) were sensitive to copper (II) addition, elevating and suppressing biosensor response respectively. GST:: haPrP P2 (53-93) produced the most significant response when applied to the HS-derivatised biosensor surface but in contrast to its binding of heparin, recognition of HS was not influenced by Cu (II) (data not shown).
Although GST itself displays no heparin-binding activity in the buffer conditions chosen for GST:: haPrP/peptides (data not shown), it is conceivable that the GST portion of the fusion peptides could influence heparin binding activities of the attached PrP sequence. Indeed it may explain the inconsistent binding observed for some GST-linked peptides (e.g. GST::P5 (218-231).
For this reason synthetic PrP peptides containing no tag were also tested for binding to bovine lung heparin (Fig. 5b). In biosensor analysis low molecular weight entities such as small peptides that these three sequences are the strongest candidate heparin/ HS-binding regions in the prion molecule. As with the GSTtagged peptides the synthetic peptides were injected singly and in different order with no effect on the patterns of binding.
All GST-fusion peptides were tested for heparin binding by ELISA (synthetic peptides with no GST were not detectable in this method). Only GST:: P1 (23-52) bound heparin reproducibly, providing a signal that could be selectively inhibited (Fig. 6a). Note that the heparin-binding exhibited by P1 (23-52), like full-length PrP, was not competed by chondroitin sulfate. In order to explore the potential for intermolecular PrP-PrP interactions via heparin-binding sequences, GST::haPrP c was incubated in microtitre wells pre-coated with each of the synthetic peptides P1 (23-52), P2 (53-93) and P4n (110-128) (Fig. 6b) GAGs in a concentration dependent manner (Fig. 7a, b). Heparin recognition was successfully competed with soluble heparin (data not shown). Porcine intestinal HS, though a weak inhibitor of the binding of synthetic P2 to immobilised heparin (data not shown) was an effective competitor for the binding of P2 to an HS-derivatised surface (Fig. 7c). It was noted that P2  bound to the HS-derivatised surface in the presence of soluble PMHS was more readily eluted with the wash sequence (BLH, NaCl, NaOH) than P2 bound in the absence of competitor.

93)
As P2 (53-93) encompasses a region of PrP containing motifs for copper binding, biosensor experiments were conducted to assess any influence of copper availability on the heparin binding activity of this peptide. The biosensor response produced by 1 mg/mL P2 (53-93) was substantially enhanced (~ 7-fold) by the addition of 10µM Cu (II) (Fig. 8a). Intriguingly a second injection of P2 peptide (without added copper) applied immediately after one containing Cu (II) generated a higher signal (ca. 2-3 fold) than the initial injection of P2 onto a washed heparin surface. This result strongly suggested that a proportion of the enhancing activity of Cu (II) might be exercised as a complex with the heparin substrate. To investigate this further we compared the accumulation of synthetic P2 on the heparin surface after two pre-treatments: i.) after injection of Cu (II) alone and ii.) after injection of Cu (II) followed by EGTA (a potent chelator of copper) (Fig. 8b). Injection of a brief pulse of EGTA between injections of Cu (II) and P2 greatly reduced the enhancement possible by pre-treatment with Cu (II) alone. In a demonstration of the selectivity of this effect no such enhancement was recorded when Cu (II) was applied in advance P1 peptide (Fig. 8b) or P4n peptide (data not shown). To assess the ability of alternative metals to enhance heparin-binding by P2 (53-93), P2 (53-93) the peptide was injected in the absence of additional cation and also in the presence of 50µM copper (II), magnesium (II), nickel (II) or manganese (II) (Fig. 8c). None of the alternative cations replicated precisely the action of copper.
Neither magnesium (II) or manganese (II) affected binding significantly and although nickel (II) induced a larger absolute response, the rapid return of signal to baseline at completion of the contact period was indicative of a highly weak interaction. Since copper (II) had an inhibitory influence on binding of PrP to heparin in the range 0-2 µM (Fig. 3b), we decided to apply P2 peptide in a range of Cu (II) concentrations (Fig. 8d). The greatest enhancing activity was afforded at relatively high concentrations (>10 µM) whereas in the concentration range that inhibits full-length PrP binding to heparin, Cu (II) had no clear influence on synthetic P2 (53-93) peptide.

Competitive ELISA to determine the relevance of P1, P2 and P4n sequences in the binding of full-length GST::haPrP c to heparin
To access which sequences, singly or in combination, might contact heparin within the fully folded recombinant GST:: haPrP c molecule a competition ELISA was developed in which GST:: haPrP c was incubated on heparin in the presence of synthetic peptides P1, P2 and P4n (Fig. 9).
As the 2 o antibody in this ELISA recognises GST, only heparin-bound GST:: haPrP c was detectable. It was anticipated that sequences contributing to heparin-binding in native PrP would compete with PrP for heparin when presented as peptides in solution. In this experiment synthetic peptides were either pre-incubated on heparin in advance of GST:: ha PrP c or mixed directly with GST:: haPrP c . Experiments were also conducted both in the presence and absence of added Cu (II). Without Cu (II), none of the peptides affected the binding of GST:: haPrP to heparin ( Fig.   9a) but when copper (II) was supplemented at 50 µM (which predictably reduced the binding signal due to GST:: haPrP c ) co-incubation of GST:: haPrP c with synthetic peptide P2 (53-93) resulted in a significantly lower binding signal on both BLH and PIH substrates than GST:: haPrP c applied with no competitor (Fig. 9b).

Discussion
Prion proteins are placed firmly in the heparin-binding category of proteins. Not only have direct interactions been demonstrated with PrP (14,15), GAGs have been shown to influence PrP sc accumulation in both cell-culture and in vitro converting experiments (12-13, 17, 18) and to modulate PrP sc propagation and disease onset in animal models for scrapie (8,9). This study has addressed three areas of current interest: the structural features of GAGs that engender PrPbinding ability, the influence of metal ions on GAG-binding and the location of GAG binding domains within PrP.
By both biosensor and ELISA techniques we were able to show direct binding of PrP to heparin, thus confirming earlier work (14,15). We were also able to demonstrate a direct interaction between PrP and purified heparan sulfate. This complements previous studies in which PrP-HS binding was strongly implicated but not directly demonstrated (eg. 17,18,21). Two aspects of the interaction with heparin were of special interest. The ability of copper (II) ions to disrupt PrP binding was observed in several experiments (e.g. Figs. 2a&b, 3b-d, Fig. 9). The direct ELISA ( Fig. 3b) proved that this inhibition was only partial i.e. that a significant proportion of heparin affinity still remains even at high concentrations of copper. This suggests that copper alters the conformation of PrP rather than simply shielding binding sites in the heparin substrate. Another explanation which we cannot rule out is that the main effect of copper ions is disrupt oligomers/aggregates of PrP bound to heparin rather than inhibition of the PrP-heparin interaction directly, hence the incomplete inhibition. Nickel (II) and zinc (II), but not manganese (II) ions were also inhibitory, though not as strongly as copper. A related effect was first described by Brimacombe et al., who reported that recombinant PrP could be displaced from an experimental nickel biosensor surface by pulses of heparin (15). PrP loaded with copper and manganese assumes distinct conformations with differing levels of protease-resistance (25)(26)(27)(28)(29)(30). Our data suggests that one of the first consequences of metal-induced conformational changes might be an altered affinity for endogenous HSPGs. The reverse situation, that bioactive HSPGs might influence the uptake of metal ions by PrP is also a possibility.
The specificity of the PrP-heparin interaction was also investigated. The interaction of PrP with heparin (as detected by the ELISA) could be disrupted by soluble heparin, HS and by other sulfated polysaccharides (Fig. 3c, d and Fig. 4) but was refractory to inhibition by CS, confirming earlier work by others (14). This last observation on the non-activity of CS has a special significance as it suggests that presentation of sulfates is not in itself sufficient for prion-binding activity. A particularly interesting finding was the potent inhibitory activity of a fraction of dextran sulfate of average MW 8000 (DS8). On a weight / volume basis this preparation was found to be superior to heparin and at least as effective as pentosan polysulfate (PPS), a polyanion with well-documented anti-prion activities in tissue culture and animal models (36,7).
It would be of great interest to compare the inhibitory activity of DS8 with fractions of higher MW such as DS500 (average MW 500,000) which is a particularly effective inhibitor of PrP sc propagation in cell culture (13) and a potent anti-prion agent in animal studies of prion disease (8,9). Using this competition ELISA we also explored the contribution to prion binding of 2-and 6-O sulfate groupings, two of the three types of sulfate which elaborate heparin and HS (N-sulfates were not examined in this study). Removal of 2-O sulfates was found to significantly reduce the inhibitory activity of BLH in the competitive ELISA (Fig. 4), suggesting an important role in prion-ligation. FGF-2 and hepatocyte growth factor are other heparin-binding proteins with a particular requirement for 2-O sulfate (56).
The location of heparin-binding sites in the prion protein has not previously been determined with certainty. This study revealed three sites in recombinant PrP with independent affinity for GAGs: amino acids 23-52, 53-93 and 110-128. Whilst the first sequence  has no previously reported affinity for GAGs, there is some evidence for GAG binding activity at the other two sites (21,36,37,43). The biosensor response generated by P2 (53-93) on both heparin and HS surfaces was concentration-dependent and selectively inhibited and so is least likely to be artefactual. In  (Fig. 6a). However, since we were unable to demonstrate competition between this peptide and full-length GST::haPrP c for binding to heparin (Fig. 9a, b) we propose that this sequence is not a major heparin-binding site in intact PrP c . A more important association may be with sequences within PrP itself for we were able to demonstrate affinity for GST::haPrP c (Fig.   6b). Such an interaction may be a means by which the putative acceptor molecules for this sequence such as nucleic acids (55) and Hsp60-like chaperonins might influence PrP function.
In biosensor analysis P4n (110-128) bound heparin and HS both as GST::fusion and free peptide, but was not positive in the ELISA. This peptide particularly yielded biosensor signals that could not be reversed by salt washing or extremes of pH. A possible explanation for this unusual behaviour is extensive aggregate formation either in solution or in situ at the heparin surface especially as a related sequence (106-126) shows a well described propensity for fibril formation (38,39). Significantly the β−sheeted structures characteristic of peptide 106-126 in weakly acidic ionic buffers (conditions in which fibrils are also favoured) are evidently highly stable (resistant to 5% SDS or alkali to pH 12) (39). That P4n should show an affinity for heparin/HS was of interest as it has shown elsewhere that the cytotoxic properties of peptide 106-126 may be abrogated by GAGs (43). Again, as an excess of P4n (110-128) was not seen to interfere with the binding of full-length PrP to heparin this peptide is considered a weak candidate for a true physiological role in heparin/ HS binding by PrP The findings in this study support a prominent role for P2  in direct HS binding by PrP predicted from recent work demonstrating the strictly HSPG dependent binding of this peptide to mammalian cells (21). Importantly, the differing influence of Cu (II) ions on the heparin-binding properties of this peptide and of the intact PrP molecule serve to illustrate the potential hazards of over-reliance on peptide-derived data alone. The sensitivity of PrP c -heparin binding to disruption with dextran sulfate and pentosan polysulfate, both of which have been proposed as prophylactic agents, and the demonstration of intimate role of 2-O sulfate in heparin recognition was also significant. Our data suggests that further investigation of the identified heparin-binding domains, and the potential specificity of carbohydrate binding sites in heparan sulfate, will lead to further insights into the role of HS in the function of prion proteins.