CLAC binds to amyloid beta peptides through the positively charged amino acid cluster within the collagenous domain 1 and inhibits formation of amyloid fibrils.

CLAC (collagenous Alzheimer amyloid plaque component) is a proteolytic fragment derived from a novel membrane-bound collagen, CLAC-P/collagen type XXV, that deposits in senile plaques associated with amyloid beta peptides (Abeta) in the brains of patients with Alzheimer's disease. We previously showed that CLAC binds to the fibrillized form of Abeta in vitro, although the mechanism and the subdomains that mediate interaction of CLAC with Abeta as well as the effect of binding of CLAC on amyloid fibril formation remain unknown. Here we show that the collagenous domain 1 of CLAC, which is rich in positively charged amino acid residues, mediates its interaction with Abeta and that this binding is mediated by an electrostatic interaction and requires formation of the triple helix structure of CLAC. The soluble form of CLAC purified from the media of cells transfected with CLAC-P inhibited fibrillization of Abeta in vitro, especially in its elongation phase. These results suggest the anti-amyloidogenic roles of CLAC in the pathophysiology of Alzheimer's disease.

Alzheimer's disease (AD) 1 is an elderly onset neurodegenerative disease causing dementia that is pathologically characterized by a massive accumulation of amyloid deposits in senile plaques or cerebrovasculature and of tau-rich neurofibrillary lesions (1). The principal component of the amyloid deposits is amyloid ␤ peptide (A␤), which is proteolytically produced from a transmembrane precursor, ␤APP, as 38 -43-amino acid fragments (2,3). A␤ secreted from cells exhibits a heterogeneity in its C-terminal extent; A␤42 ending at the 42nd residue, which comprises ϳ10% of A␤40 in total secreted A␤, aggregates much faster than A␤40 (4) and deposits initially as diffuse-type plaques in AD or Down's syndrome brains (5). Moreover, missense mutations in ␤APP, presenilin 1 and presenilin 2 genes linked to early onset familial AD increase the production of A␤42 (6 -8). These findings collectively suggest that aggregation and deposition of A␤ are closely linked to the pathogenesis of AD.
A number of non-A␤ proteinaceous components, e.g. apolipoprotein E (apoE), ␣1-antichymotrypsin, ␣ 2 -macroglobulin, complement component C1q, amyloid P component, have been identified associated with senile plaque amyloids (9), some of which have been shown to affect fibrillization of A␤ (10). Genetic polymorphism of apoE, especially the ⑀4 genotype, is known as the major genetic risk factor of AD (11), and accumulating evidence from experimental studies suggests that apoE may affect the fibrillization and deposition of A␤, thereby leading to AD; ablation of apoE gene reduced A␤ deposition in transgenic mice overexpressing human ␤APP, whereas further transgenic supplementation of human apoE ⑀4 gene accelerated deposition of ␤-sheet-rich A␤ (12). In vitro studies, however, showed that apoE also has an inhibitory effect on fibrillization of A␤ (13,14), implicating apoE in various aspects of ␤-amyloid formation in AD. These previous results emphasize the compelling need for the analysis of effects of other non-A␤ components of senile plaques on the fibrillization and deposition of A␤.
We have recently identified a novel collagenous protein as a component of senile plaque amyloid and designated it as CLAC (collagenous Alzheimer amyloid plaque component) (15). CLAC is derived from its precursor CLAC-P/type XXV collagen, a member of collagenous transmembrane proteins (16), specifically expressed in neurons. We also showed that the ectodomain of CLAC-P is shed by furin convertase and secreted as a soluble form of CLAC (sCLAC) (15). Immunohistochemical analysis of postmortem human brains revealed a unique pattern of CLAC deposition in AD and Down's syndrome brains (17); CLAC was negative in A␤42-positive, A␤40-negative pure diffuse plaques that are considered to be the earliest form of A␤ deposition, whereas a fraction of primitive-type senile plaques became CLAC-positive at a relatively early stage of plaque maturation as the fibrillization of A␤ progresses. Notably, mature plaques that appear at later stages were A␤40/thioflavin S-positive but remained CLAC-negative (17). These findings suggested the possibility that CLAC may play an inhibitory role in the deposition of senile plaque amyloid by binding to early A␤ deposits and preventing further incorporation of A␤ peptides to develop ␤-sheet-rich, mature senile plaques.
To test this hypothesis by in vitro experiments, we studied the effects of sCLAC on A␤ fibrillization in vitro and examined the mechanism as well as the subdomains that mediate interaction of sCLAC with A␤. We found that sCLAC in its proper triple helix form binds to A␤ fibrils through the positive charge cluster within the collagenous domain 1 and that binding of sCLAC inhibits fibrillization of A␤ in the elongation phase of amyloid fibril formation, providing further support to the inhibitory role of CLAC in ␤-amyloidogenesis.
Cell Culture and Transfection-HEK293 cells were cultured as described (8). Stable cell lines were generated by transfection of cDNAs using Lipofectamine 2000 reagent (Invitrogen) as described (15).
In Vitro Binding Assay-The in vitro binding assay for sCLAC and fibrillized A␤ or heparin immobilized on a solid phase was performed as described (15). Briefly, a 50-l aliquot of 100 g/ml heparin, bovine serum albumin (as a negative control), and prefibrillized synthetic A␤-(1-42) (Peptide Institute, Inc.) was allowed to dry on wells of enzyme-linked immunosorbent assay plates, blocked for 1 h with Block Ace (Snow Brand, Sapporo, Japan), and washed with PBS containing 0.05% Tween 20. Microplate wells were then incubated with conditioned media of HEK293 cells stably transfected with wild-type or mutant CLAC-P cDNAs for 1 h, washed with PBS containing 0.05% Tween 20, and then reacted with anti-NC4 antibody for 1 h. After incubation with a horseradish peroxidase-tagged secondary antibody, bound sCLAC was quantitated by development using a TMB microwell system (KPL, Inc.). The effects of ionic strength on the interaction between fibrillized A␤ and wild-type or mutant sCLAC were assayed as follows; conditioned media containing sCLAC were dialyzed against 10 mM phosphate buffer followed by the addition of NaCl to the final concentrations of 0 to 1 M and then loaded onto microtiter plates precoated with fibrillized A␤. For the competition assays 200-l aliquots of conditioned media were preincubated with 10 g of prefibrillized A␤ (or A␤ fragment peptides) at 4°C for 8 h and then loaded onto microtiter plates that were precoated with fibrillized A␤-(1-42).
Trypsin Digestion Assay-Trypsin digestion assays for the determination of the formation and preservation of the collagenous triple-helix structures were performed according to a previously described method (18). Briefly, conditioned media containing wild-type or mutant sCLAC were dialyzed against 0.1 M Tris-HCl (pH 7.4) with 0.4 M NaCl buffer, heated at 20 -60°C for 5 min, and treated with 0.15 mg/ml trypsin at 20°C for 2 min. The trypsin digests were subsequently inactivated by the addition of SDS sample buffer and boiling for 10 min followed by immunoblotting with anti-NC3 and anti-NC4 antibodies.
Purification of sCLAC-Procedures for purification of sCLAC are summarized in the flow chart of Fig. 6A. Briefly, conditioned media of HEK293 cells stably expressing human CLAC-P were initially applied to a DEAE-cellulose column (DE52, Whatman), and the flow-through fractions were then applied to a heparin-Sepharose column (Amersham Biosciences) and eluted with phosphate buffer containing 1 M NaCl. The eluates were dialyzed against PBS, concentrated, and further separated by reverse phase HPLC on an Aquapore RP300 column with a linear gradient of 0 -64% acetonitrile in 0.1% trifluoroacetic acid. Purified sCLAC in acetonitrile solution was evaporated and dissolved in 1 mM HCl at 0.1 mg/ml.
Negative Stain Electron Microscopy-Samples were spread on 400mesh collodion-coated grids and then negatively stained with 2% phosphotungstic acid (pH 7.0) and viewed in an electron microscope (JEOL 1200EXII) as described (19). Negative-stain immunoelectron microscopy was performed as described (19). Briefly, samples were spread on grids and blocked by 1% gelatin containing PBS for 30 min and then incubated with anti-CLAC-P antibody for 1 h. After washing, the grids were incubated by anti-rabbit IgG secondary antibody conjugated with 10-nm colloidal gold for 1 h and negatively stained and viewed by electron microscopy as above.

RESULTS
Collagenous Triple-helix Structure of CLAC Is Indispensable to Its Interaction with Fibrillized A␤-CLAC-P is a transmembrane protein with a type II orientation (Fig. 1A). We previously reported that CLAC-P is cleaved at Arg-112-Glu-113 by furin convertase, thereby liberating the extracellular domain as a soluble/secreted form of CLAC-P (sCLAC) (15). CLAC-P harbors three Gly-X-Y collagenous repeat motifs and forms a homotrimeric structure under non-reducing conditions (15). We examined whether sCLAC retains the triple-helix structure by a trypsin digestion assay. Native collagens with a triple-helix structure are resistant to digestion by authentic proteases including trypsin or pepsin, whereas heat-denatured collagen at Ͼ40°C is susceptible to these proteases (18,20). We heated conditioned media of HEK293 cells stably transfected with human CLAC-P containing sCLAC at various temperatures between 25-60°C and then digested them by trypsin and found that sCLAC remained resistant to trypsin digestion by preheating at Ͻ45°C, whereas it became labile at Ͼ50°C (Fig.  1B). We previously reported that sCLAC is bound specifically to fibrillized form of A␤ by an in vitro solid-phase A␤ binding assay (15). To examine whether the triple-helix structure of sCLAC is required for the binding between sCLAC and fibrillized A␤, we preheated conditioned media containing sCLAC at 25-60°C and subjected them to the in vitro A␤ binding assay. sCLAC preheated at Ͻ45°C was bound to A␤, whereas sCLAC denatured at Ͼ50°C failed to interact with fibrillized A␤ (Fig.  1C). This strongly suggested that the preservation of the collagenous triple-helix structure of sCLAC is the prerequisite for its interaction with A␤ fibrils.
CLAC-P harbors two highly conserved, ␣-helical coiled-coil domains that are composed of repeats of heptad sequences (abcdefg; hydrophobic amino acid residues occur at positions a and d) that are shown to be important for the assembly of a set of membrane-bound collagens including type XIII collagen, a homologue of CLAC-P/type XXV collagen (21), within the NC1 and NC3 domains ( Fig. 2A). Moreover, it has been shown that type XIII collagen fails to form a triple helix and becomes misfolded when the NC1 coiled-coil domain is deleted (22). We generated three mutant (mt) CLAC-P cDNAs (NC1ccmt, NC3ccmt, and NC1/NC3ccmt) in which the hydrophobic amino acid residues within the coiled-coil domains are replaced with lysine and stably expressed them in HEK293 cells. NC1ccmt CLAC-P was moderately sensitive to trypsin digestion, suggesting a partial impairment in the triple-helix structure, whereas the NC3ccmt and especially the NC1/NC3ccmt, CLAC-P, were highly sensitive to trypsin digestion without prior denaturation, indicating the lack of the triple-helix formation (Fig. 2B). We then tested the relationship between triple-helix formation of CLAC-P and sCLAC/A␤ binding by the in vitro A␤ binding assay. Binding of fibrillized A␤ relative to that with wild-type (wt) CLAC-P was slightly reduced in NC1ccmt CLAC-P, whereas it was significantly impaired with NC3ccmt CLAC-P and completely lost with NC1/NC3ccmt sCLAC polypeptides (Fig. 2C). These results clearly indicate that the collagenous triple-helix structure of sCLAC is indispensable to its interaction with A␤ fibrils.
The Binding between Fibrillized A␤ and sCLAC Is Mediated by Electrostatic Interaction and Is Competed by Heparin-We previously reported that the interaction between fibrillized A␤ and sCLAC is blocked by a high concentration of NaCl (15). To examine the effects of ionic strength in more detail, we dialyzed conditioned media containing sCLAC against 10 mM phosphate buffer followed by the addition of NaCl at concentrations of 0 -1 M and subjected them to the in vitro A␤ binding assay and found that the presence of high concentrations of NaCl inhibited the interaction between fibrillized A␤ and sCLAC (IC 50 ϭ 0.73 M) (Fig. 3A). The possibility that A␤ immobilized on wells were detached by a high concentration of NaCl was excluded by probing the plates by an anti-A␤ antibody (data not shown). These data further confirm that sCLAC is bound to fibrillized A␤ through an electrostatic interaction.
We next investigated the interaction between sCLAC and heparin. sCLAC harbors one of the predicted binding consensus motifs for heparin (XBBXBX or XBBBXXBX, where B and X are basic and hydrophobic residues, respectively) (23) at 182 IKRRLIKG 189 in the NC2 domain. In addition, sCLAC has four positively charged amino acid cluster regions within its collagenous domains (Fig. 4A), where positively charged amino acid residues are frequently located at X and/or Y positions in the Gly-X-Y collagenous repeats. These domains are predicted to constitute a cluster of positive charges when CLAC-P or sCLAC form a native triple-helix structure. Such positive charge cluster has been implicated in the binding between collagen and heparin (24,25). Taken together with the previous findings that type XIII and XXIII collagens bind to heparin (26,27), we suspected if one or the other of the four positively charged amino acid cluster regions of CLAC-P may serve as binding domains for heparin. We examined the binding of sCLAC and heparin by an in vitro binding assay, coating heparin onto microwell plates and found that sCLAC and heparin bind each other (Fig. 3B) and that the interaction is completely inhibited by the addition of 1 M NaCl (data not shown). These results suggested that sCLAC also was bound to heparin by an electrostatic interaction. To examine if heparin competes for the binding domain of sCLAC with fibrillized A␤, we added different concentrations of heparin (0 -1 mg/ml) to the conditioned media containing sCLAC in the in vitro A␤ binding assay and confirmed that heparin inhibits the binding between sCLAC and fibrillized A␤ in a dose-dependent manner (Fig.  3C). These results strongly suggest that fibrillized A␤ and heparin interacts with an identical subdomain in sCLAC.

The Positively Charged Amino Acid Cluster in the COL1
Domain Is Crucial for the Interaction of sCLAC with Fibrillized A␤-We found that the preservation of the collagenous triplehelix structure of sCLAC is required for its interaction with fibrillized A␤, suggesting that A␤ fibrils recognize the threedimensional structure, not simply the primary structure, of sCLAC. Furthermore, the heparin binding region of sCLAC seems to overlap with that for A␤ fibrils. These findings led us to hypothesize that one or more of the positively charged amino acid cluster regions within the collagenous domain may especially be important for the interaction of sCLAC with fibrillized A␤ (Fig. 4A). We then constructed five mt CLAC-P cDNAs (COL1mt, COL2mt, COL3-1mt, COL3-2mt, and COL1/3-2 double mt) in which the positively charged amino acid residues in the X and/or Y positions of the Gly-X-Y collagenous repeats were replaced with proline, the latter being the most common residue at the X or Y positions in collagens and contributing to the stability of the triple-helix structure. We stably expressed these mt CLAC-P cDNAs in HEK293 cells and confirmed that the mt sCLAC proteins also were secreted into culture media (Fig. 4B), forming an intact triple-helix structure as determined by trypsin digestion (Fig. 4C). COL1mt, COL3-2mt, and COL1/3-2 double mt CLAC-P polypeptides migrated slower than wt CLAC-P, presumably reflecting slight changes in charge or structure. In addition, the immunoreactivities of COL3-1mt, COL3-2mt, and COL1/3-2 double mt polypeptides to anti-NC3 or anti-NC4 antibodies were decreased by ϳ50 -75%, whereas those for COL1mt or COL2mt were preserved at similar levels to that for wt CLAC-P, possibly due to proximity of epitope locations and the mutated regions (data not shown).
Considering these differences in the reactivity of detector antibodies in the in vitro A␤ binding assay, we have chosen to compare the binding affinities of mt sCLAC to fibrillized A␤ as the relative strength of the ionic interactions in each mutant. To this end we dialyzed conditioned media containing wt or mt sCLAC against 10 mM phosphate buffer and then added NaCl to final concentrations of 0 -1 M and subjected them to the in vitro A␤ binding assay. The interaction between wt sCLAC and A␤ was inhibited by the addition of Ͼ0.5 M of NaCl (Fig. 3, A  and D) and COL2mt, COL3-1mt, or COL3-2mt sCLAC showed almost similar inhibition profiles of A␤ binding to that with wt Note that all mt sCLAC exhibited resistance to trypsin in a similar manner to wt sCLAC, suggesting proper trimer formation of mt sCLAC polypeptides. D, in vitro binding assays of wt or mt sCLAC with fibrillized A␤. Binding of wt, COL1mt, COL2mt, COL3-1mt, COL3-2mt, or COL1/3-2 double mt sCLAC with fibrillized A␤ at different NaCl concentrations were monitored by the in vitro assay and plotted as ratios relative to the optical densities of immunoreactions for each mt sCLAC in the absence of NaCl as 1.0. Mean Ϯ S.D. of optical densities in four independent experiments are shown. E, comparison of the binding of wt and COL1mt sCLAC with fibrillized A␤ at 131 mM NaCl (normalized by the levels of sCLAC in media). The amount of COL1mt sCLAC bound to fibrillized A␤ as detected by anti-NC4 antibody was ϳ20% that of wt sCLAC. The optical density with wt sCLAC was normalized to 1.0, and the mean Ϯ S.D. in three independent experiments are shown. sCLAC (Fig. 4D). In sharp contrast, the interactions of COL1mt as well as the COL1/3-2 double mt sCLAC with fibrillized A␤ were inhibited by lower concentrations of NaCl compared with that of wt sCLAC (Fig. 4D). We further compared the absolute levels of sCLAC binding with fibrillized A␤ in wt and COL1mt sCLAC, which were confirmed to react at similar intensities with anti-NC3 or NC4 CLAC-P antibodies, and demonstrated that the binding of COL1mt sCLAC with A␤ is decreased (Fig. 4E). These data strongly suggested that the positively charged amino acid cluster region in the COL1 domain plays a crucial role in the binding to fibrillized A␤.
Fibril Formation of A␤ Is Required for Its Interaction with sCLAC-We previously showed that CLAC-P and sCLAC specifically bind to a fibrillized form of A␤ (15). To locate the binding domain to sCLAC within A␤, we used a series of synthetic A␤ fragment peptides (Fig. 5A). We preincubated A␤ fragment peptides to induce fibrillization and preabsorbed the conditioned media containing sCLAC with the peptides. Thus preabsorbed media were subjected to the in vitro A␤ binding assay coating A␤-(1-42) on the solid phase. Preabsorption by A␤-(1-42) or A␤-(1-40) inhibited the interaction of residual sCLAC with fibrillized A␤ to a similar extent, whereas the interaction was not significantly inhibited by preabsorption with A␤-(1-28), -(1-16), -(17-42) or A␤-(12-28) (Fig. 5B). To evaluate the extent of fibril formation of A␤ fragment peptides, we quantitatively examined the ␤-sheet contents by thioflavin T (thioT) fluorescence, a fluorescent small molecule that binds to ␤-sheeted structures (28), and found that A␤-(1-42), A␤-(1-40), or A␤-  show increased thioT fluorescence reflecting formation of ␤-sheet rich fibrils, whereas other fragment peptides do not (Fig. 5C). Formation of fibrils was confirmed by negative-stain electron microscopy (data not shown). These results indicate that the fibril formation of A␤ is the prerequisite for interaction of A␤ with sCLAC. Furthermore, the binding domain of A␤ with sCLAC was estimated to be located within the N-terminal 16 amino acid residues because fibrillized A␤-(17-42) did not bind sCLAC.
Purification of sCLAC and Binding to Fibrillized A␤-To gain insights into the effects of sCLAC on A␤ fibril formation, we purified sCLAC from conditioned media of HEK293 cells stably expressing CLAC-P. Conditioned media were sequentially purified by a DEAE-cellulose column, heparin-Sepharose column, and reverse-phase HPLC (Fig. 6A). sCLAC was purified to a near homogeneity in fractions after separation by reverse-phase HPLC as determined by protein silver staining (Fig. 6B). To test whether purified sCLAC retained affinity with fibrillized A␤, we examined the binding of purified sCLAC to fibrillized A␤ by the in vitro A␤ binding assay. Purified sCLAC showed affinity to bind fibrillized A␤ at ϳ40% of those in crude conditioned media (Fig. 6C), and negative-stain immunoelectron microscopy showed that purified sCLAC directly binds to prefibrillized A␤ fibrils (Fig. 6D).
sCLAC Inhibits Fibrillization of A␤-  in Vitro-To verify the effects of sCLAC on fibrillization of A␤, we incubated A␤-(1-42) (at 0.1 mg/ml) alone or together with purified sCLAC or ␣1mG at a molar ratio of A␤-(1-42)/sCLAC or/␣1mG at 100/1 and examined the extent and time course of A␤ fibrillization by thioT fluorescence. Purified sCLAC decreased fibrillization of A␤-(1-42), whereas ␣1mG had no effects on its fibrillization (Fig. 7A). Electron microscopy showed the formation of A␤ fibrils by incubation of A␤-(1-42) for 8 h (Fig. 7B), whereas coincubation of A␤-(1-42) with purified sCLAC abolished formation of fibrils at the same incubation period (Fig.  7B). To rule out the possibility that sCLAC masked the thioT binding site on A␤ fibrils, we examined the thioT fluorescence of A␤ fibrils incubated with purified sCLAC at molar ratios (A␤/sCLAC) of 1000/1, 250/1, 100/1, 50/1, or 25/1 and confirmed that all samples showed equal levels of fluorescence and contained CLAC-positive fibrils as revealed by immunoelectron microscopy (data not shown), excluding the possibility of masking thioT binding sites in this experimental setting. These results suggested that sCLAC has an ability to inhibit fibrillization of A␤ in vitro.
sCLAC Inhibits Fibrillization of A␤ in Its Elongation Phase-The process of A␤ fibrillization consists of two phases, i.e. nucleation and elongation (Fig. 8A) (4); the nucleation phase represents the conversion of nascent A␤ monomer to an unstructured conformation followed by the formation of fibril seeds. Once fibril seeds are formed, A␤ monomer is rapidly assembled into the seeds or fibrils, eliciting the elongation phase. Thus, the time lag before the formation of fibrils as determined by thioT fluorescence represents the speed of nucleation, whereas the speed of subsequent fibril formation reflects elongation. To investigate whether sCLAC affects either of these two phases of amyloid fibril formation, we incubated A␤-(1-40) (at 0.25 mg/ml) alone or together with purified sCLAC, apoE, or ␣1mG. The use of A␤-(1-40), which fibrillizes slower than A␤-(1-42), enabled us to evaluate the effects of coexisting proteins on the nucleation phase, the latter being represented by the time lag before fibril formation. It has been shown that apoE inhibits A␤ fibril formation chiefly in the nucleation phase in vitro (13,14). Purified sCLAC and apoE inhibited the final extent of fibril formation of A␤-(1-40), whereas ␣1mG did not affect this process (Fig. 8B). However, during the initial phase of A␤ fibrillization (0 -4 h), the speed of fibrillization of A␤-(1-40) was similar by coincubation with purified sCLAC or ␣1mG, whereas apoE inhibited the initial rise in A␤ fibril formation, delaying the start of fibrillization by ϳ24 h (Fig. 8B). This suggested that sCLAC inhibits A␤ fibril formation in the elongation phase. To further verify this idea, we incubated A␤-(1-40) (0.1 mg/ml) with prefibrillized/sonication-disrupted A␤-(1-42) as artificial seeds (molar ratio of A␤-(1-40)/A␤-(1-42) seed: 100/1). The addition of a small amount of prefibrillized A␤-(1-42) strongly promoted the incorporation of A␤-(1-40) into fibrils, bypassing the nucleation phase (4). In the presence of A␤-(1-42) seeds, apoE or ␣1mG no longer affected fibril elongation of A␤-(1-40), whereas purified sCLAC inhibited the elongation of A␤ fibrils. Taken together, we con- cluded that sCLAC has an inhibitory effect on fibrillization of A␤ in its elongation phase. DISCUSSION In this present study we examined the binding of sCLAC to A␤ in vitro and showed the following. 1) Formation and preservation of the collagenous triple-helix structure of sCLAC is indispensable to its interaction with A␤; 2) binding of sCLAC with A␤ is mediated by an electrostatic interaction through the positively charged amino acid cluster region in the COL1 domain of sCLAC; 3) fibril formation of A␤ is the prerequisite for its interaction with sCLAC; 4) binding of sCLAC inhibits A␤ fibril formation in the elongation phase. These data provide mechanistic explanation for the selective binding of CLAC to A␤ deposits in AD brains and strongly implicate the inhibitory role of CLAC in ␤-amyloid formation in AD brains.
We showed that sCLAC is bound to fibrillized A␤ through the positively charged amino acid cluster region in the COL1 domain. Neutralization of the positive charges by proline substitution, suppression of trimer formation by mutation of the coiled-coil domain that is essential to trimer formation, and heat denaturation of sCLAC to disrupt the triple-helix structure altogether abolished binding of CLAC to A␤, underscoring the importance of the positive charge clusters orderly displayed on the surface of the triple helix. Positively charged amino acid clusters within the collagen regions have been identified as binding sites for a variety of ligands, including A␤ fibrils, in a couple of collagen-related molecules. Type I scavenger receptor class A, known as one of the cell surface receptors for microglial phagocytosis of A␤ fibrils (29), harbors its ligand recognition domain at the collagenous 332 GPKGQKGEK 340 sequence located at the C terminus, in which positively charged amino acids are clustered at the 3rd residues of the collagen repeat sequences. Cells expressing mutant type I scavenger receptor class A substituted at these three basic amino acid residues by Ala failed to bind and degrade its major ligand, oxidized low density lipoprotein (30). Complement component C1q, known as one of the amyloid associating proteins in senile plaques of AD brains, has been shown to interact with A␤ through the positively charged amino acid cluster 14 AGRPGRRGRPGLK 26 within the N-terminal collagenous domain of the C1q A chain (31), whereas C1q B or C chains that lack basic amino acid clusters do not bind A␤. These previous findings are in agreement with our present data and provide further supports to our view that the positively charged amino acid cluster within collagenous domain 1 of CLAC, which is exceptionally rich in positively charged amino acid residues, serves as an optimal binding site for A␤. This is consistent with our preliminary immunohistochemical data that ectodomain fragments of type XIII or type XXIII collagens do not appear to be associated with senile plaque amyloid. 2 Recently Soderberg et al. (32) reported that another positively charged amino acid-rich domain of CLAC, LIKRRLIK, within the non-collagenous domain 2, is the major binding site for A␤ (32); they deleted or replaced these BSA, bovine serum albumin. D, binding of purified sCLAC to A␤ fibrils as revealed by negative-stain immunoelectron microscopy using anti-NC3 and a secondary antibody tagged with 10-nm colloidal gold. Scale bar ϭ 50 nm amino acid sequences with the homologous region of type XIII collagen that contains smaller number of basic amino acids compared with CLAC and found a reduced A␤ binding. This domain seemed to bind to A␤ irrespective of the triple-helix formation of CLAC because synthetic peptides mimicking these sequences were also bound to A␤. Moreover, they showed that deletion of the basic amino acid cluster within COL1 domain, i.e. KRGKRGRR, which corresponds to the domain we investigated in this present study, did not affect A␤ binding. The reason for this apparent discrepancy between these two studies is not clear at present. One possibility would be that a large deletion within the collagen domain, associated with a duplication of glycine residue flanking the deleted portion, might have caused a change in the structure of sCLAC, rendering other positive charge clusters more accessible to A␤ fibrils. In this regard, our strategy not to alter the relative length of each collagenous/non-collagenous domain but neutralize the charges by amino acid substitutions to proline would better preserve the native structure of sCLAC. Another difference is the addition of Myc/His tags in the aforementioned study, which might have caused alterations in structure or charge states. In any event, it remains possible that there are multiple potential A␤ binding sites across the collagenous and non-collagenous domains in CLAC.
We also showed that sCLAC bound fibrillized synthetic A␤-(1-42) and A␤-(1-40) but not A␤-(17-42) despite that the latter formed thioT-positive filaments. This suggests that sCLAC may recognize the N-terminal third of A␤ in its fibrillized form. Indeed, the N-terminal 11 amino acids of A␤ harbor four negatively charged amino acid residues (i.e. DAEFRHDS-GYE) that may interact with positive charges on CLAC. Our recent immunohistochemical analysis of postmortem brains from patients with AD or Down's syndrome showed that CLAC is a relatively early component of senile plaque amyloid coexisting with A␤42, whereas deposition of A␤40 occurs chiefly in CLAC-negative senile plaques at a later stage (17). This apparently suggested that CLAC may have different affinities to A␤42 and A␤40. However, our preabsorption analysis showed that sCLAC binds to A␤-(1-42) and A␤-(1-40) at similar affinities in vitro. It is most plausible that the packing density or structure of A␤ fibrils in A␤40-positive plaques may be different from those in A␤40-negative ones in vivo, hampering bind- ing of CLAC in A␤40-and thioflavin S-positive SP, whereas in vitro generated A␤-(1-40) and A␤-(1-42) fibrils harbor similar structures in the CLAC binding regions and, thus, exhibit comparable affinities to sCLAC.
We showed by in vitro assays that coincubation of soluble A␤ with sCLAC inhibited A␤ fibril formation and that this inhibitory effect was chiefly directed to the elongation phase of amyloid fibrils. This process was monitored by thioT fluorescence assay, although a parallel negative-stain electron microscopic analysis confirmed that the decrease in thioT reactivity was not due to masking of thioT binding site by binding of CLAC, because the amount of fibrils generated during incubation was actually reduced. Although the mechanism of this inhibition is still unclear, the effect showed a sharp contrast with that of apoE, which inhibited the seed formation, whereas the addition of prefibrillized seeds overrode this inhibitory effect. It is most plausible to speculate that sCLAC avidly binds to and obstructs the docking sites on A␤ fibrils or protofibrils that serve as seeds for fibril propagation to which monomeric A␤ is successively incorporated, thus inhibiting elongation of A␤ fibrils. Further in vitro studies to examine the binding of sCLAC to A␤ polymers in various states, i.e. oligomers, protofibrils, and fibrils, will be needed.
Our present in vitro data strongly suggest that CLAC may serve as an inhibitory factor for A␤ fibrillization that slows down the deposition of ␤-amyloid in AD brains. However, the possibility that CLAC binding may have multiple downstream effects that variously influence A␤ deposition should be reserved. For example, collagens harboring triple-helix structure are known to be resistant against conventional proteinases, suggesting that CLAC deposited in senile plaques may confer amyloid deposits resistance to proteolytic degradation or microglial phagocytosis. Indeed, we have shown that A␤ immunoreactivities in CLAC-positive senile plaques are more resistant to proteinase K digestion compared with those in CLACnegative ones (33). Considering the discrepancies in the effects of apoE on A␤ fibrillization and deposition that were previously documented in vitro (inhibitory (13,14)) and in vivo (promoting (12)), however, further in vivo studies, especially cross-breeding experiments of transgenic mice overexpressing ␤APP and CLAC-P, will be mandatory. This will provide important clues to the elucidation of the pathological roles of CLAC in the deposition of ␤-amyloid and AD as well as of its therapeutic potentials.