INTRODUCTION

Serum amyloid A proteins (apoSAAs)¹ are encoded by an ancient, multigene family, which has been conserved for at least 200 million years (1). In mice four genes are actively expressed, producing one constitutive isoform (apoSAA4), one intermediate acute-phase isoform (apoSAA3), and two archetypical acute-phase isoforms (apoSAA1 and -2) (2). The latter two are composed of 103 residues differing by 9 substitutions (3) and make up about 95% of the plasma apoSAA, associated mainly (90%) with high density lipoprotein (HDL) (4, 5). During the acute phase of the inflammatory response, their synthesis is induced in the liver by cytokines (interleukin-1, interleukin-6, and tumor necrosis factor) with peak transcription rates attained within 3 h (6, 7), which leads to an increase in plasma concentration of up to 1000-fold (to approximately 1 mg/ml) within 18-24 h, declining to less than 50 µg/ml by 48 h (8). Since their production is stimulated by infection or tissue injury, it is widely accepted that apoSAA enhances host survival by either neutralizing the infectious agent or contributing to the repair process (2, 9).

ApoSAA and at least 16 other unrelated, normally nonfibrillar proteins are known to be precursors of different amyloid deposits. Amyloid is a generic term describing pathological accumulations of fibrillar proteinaceous deposits, which are associated with a number of disorders including Alzheimer's disease, rheumatoid arthritis, chronic hemodialysis, diabetes, and cancer (10, 11). These amyloid fibrils invade primarily the extracellular space of organs, where they can disrupt tissue architecture and function. Although the fibrillar proteins may differ in different diseases, amyloids have common tinctorial, ultrastructural, and compositional features.

All amyloids are composed of nonbranching fibrils (7-10 nm) with a crossed -pleated sheet conformation, which stain with Congo Red and exhibit red/green birefringence under polarized light. In many cases mutations in precursor proteins (e.g.. transthyretin, -protein precursor) have been linked to accelerated deposition, but such mutations are not a prerequisite. In addition, abnormal proteolysis has also been thought to precede fibril formation, but again in many cases this has been shown to be unnecessary. However, it has been recognized that the basement membrane (BM) matrix is disrupted near amyloid deposits (12, 13, 14), and in those amyloids that have been investigated, one or more BM components appear to codeposit with amyloid (15, 16, 17). Where analyzed, the accumulation of specific BM protein mRNAs coincided with amyloid deposition in affected tissues implying that the BM components found in amyloid deposits do not originate from preexisting BM, but from de novo synthesis (18, 19, 20). Once secreted it has been demonstrated both in vivo and in vitro, that BM components can spontaneously assemble to generate a BM matrix (21, 22), a process that seems to be disrupted in and around amyloid deposits.

How apoSAA becomes deposited as an amyloid is still largely unknown. During reactive (secondary) amyloidosis, a complication of chronic inflammation, apoSAA is deposited in the spleen, kidney, and liver, where it is processed leaving the amino half to two-thirds of its sequence behind in the fibril (23). In mice only apoSAA2 serves as the amyloid precursor, and its amino-terminal sequence (1-15 residues), which differs from apoSAA1 by two residues, can form fibrils in vitro, albeit only at very low pH (24). For humans, six allelic apoSAA variants, produced by three SAA genes, show no difference in sequence at the amino end and may explain why they all have amyloid forming potential (2).

Significant advances in understanding the mechanism of amyloidogenesis have emerged from immunohistochemical studies which have demonstrated that BM components codeposit with amyloid (16, 25). In addition, heparan sulfate (HS), a glycosaminoglycan covalently linked to the BM-type proteoglycan, perlecan, has been shown to cause an increase in -sheet structure (prevalent in amyloid) exclusively in apoSAA2 (26) and not with apoSAA1, nor with apoSAA_CE/J isolated from an amyloid-resistant mouse strain (23). For the Alzheimer's amyloid peptide (A), HS was found to enhance fibrillogenesis in vitro (27). Recently, HS analogs (anionic polysulfates and polysulfonates) have been shown to reduce AA amyloid accumulation in vivo and also interfere with HS-induced A fibril formation in vitro (28). These observations support the hypothesis that specific interactions between amyloid precursor proteins and BM components, particularly perlecan, may facilitate amyloid formation and cause the concurrent disruption in BM structure. As a first step to test this hypothesis, we investigated apoSAA's binding potential for the major BM components, collagen type IV (C-IV), entactin (also called nidogen), laminin, and perlecan.

Employing an ELISA technique proven very effective in the characterization of interactions involving different BM components (29, 30, 31), we have detected and characterized a saturable, high affinity, association between laminin (laminin-1 from Engelbreth-Holm-Swarm tumor) and murine apoSAA preparations containing apoSAA1 and apoSAA2. Laminin-1 is the prototype of the laminin family of BM glycoproteins composed of three different subunits (1, 1, and 1) arranged into a multidomain cruciform structure with three amino-terminal short arms and one carboxyl-terminal long arm. Entactin, a BM protein that functions to cross-link the BM and is normally complexed with laminin (32), could inhibit laminin-apoSAA binding. In addition, apoSAA, like entactin, had no effect on the Ca²⁺-dependent polymerization of laminin, an integral process in BM assembly (22, 33, 34, 35). These observations suggest that the apoSAA binding site(s) may map close to the entactin one identified on the laminin -chain (36), and are clear of the amino-terminal globular domains required for polymerization (35). Perlecan binding of apoSAA was not detected, but this does not exclude the possibility of a low affinity association.

The data presented suggest that apoSAA interacts in situ, with laminin which is free of entactin, and possibly in concert with HS facilitates apoSAA fibrillogenesis. We could find no evidence that the coincidental disturbance in BM organization observed with AA amyloidosis was caused by this interaction. Since the avidity of the laminin-apoSAA interaction is maximal under physiological conditions (pH, ionic strength) with "wild-type" proteins, we also postulate that it contributes to the normal acute phase response, possibly mediating some of the functions recently proposed for apoSAA.

MATERIALS AND METHODS

Plasma apoSAA concentrations were experimentally elevated in CD1 mice by a subcutaneous injection of 0.5 ml of 2% (w/v) AgNO₃ (37), resulting in a sterile abscess and an acute inflammatory state. After 18-20 h, mice were sacrificed by CO₂ narcosis and exsanguinated by cardiac puncture preventing clotting with a small amount of 7% EDTA. High density lipoprotein containing apoSAA (HDL_SAA) was isolated from plasma by sequential density flotation (38). The density of the plasma was adjusted to 1.063 g/ml with KBr or NaBr and centrifuged at 175,000 × g for 18 h in a 70.1Ti rotor (Beckman) at 4 °C. The top layer containing very low density lipoprotein/low density lipoprotein was removed and discarded. The pooled infranatants were adjusted to a density of 1.21 g/ml and re-centrifuged at 250,000 × g for 48 h at 4 °C. The top layer (HDL_SAA) was aspirated, pooled, dialyzed against 0.15 M NaCl, 0.1% (w/v) EDTA, pH 6.4 (2 × 1 liter), for 18 h.

ApoSAA was purified from HDL_SAA by dialysis against 10% formic acid, pH 2.0 (2 liters), for 18 h at 4 °C, followed by gel filtration on a Sephacryl-S-100HR column (2.5 cm × 110 cm) eluted in the same buffer at 25 ml/h. The first major peak contained mainly apoA-I/A-II/C, and the second contained apoSAA1 and apoSAA2, which was collected, quick-frozen in liquid N₂, lyophilized, and stored at 20 °C.

AA peptides were purified from amyloid-laden mouse spleens by the method of Skinner et al., 1983 (39). All protein concentrations were determined by the DC protein assay (Bio-Rad). Lipid free bovine serum albumin (BSA) was used as the reference protein. Purity of the apolipoproteins and AA peptides was evaluated by SDS-urea-polyacrylamide gel electrophoresis (PAGE) stained with Coomassie Blue R-250 (40).

BM components were purified from Engelbreth-Holm-Swarm (EHS) mouse sarcoma propagated in nonlathyritic mice (Swiss Webster; Charles River Laboratories, Montreal, Quebec, Canada), harvested at 2-4 cm, frozen in N₂(l), and stored at 70 °C. Purification of laminin (laminin-1 isoform) and entactin was carried out as described by Ancsin and Kisilevsky (41). Tumor was homogenized and washed twice in 3.4 M NaCl, 50 mM Tris, 2 mM EDTA, 1 mM N-ethylmaleimide, 1 mM phenylmethylsulfonyl fluoride, pH 7.5, centrifuged at 10,000 × g for 15 min, discarding the supernatant on each occasion. The residue was extracted in the same buffer but with 0.5 M NaCl, for 4-8 h, then centrifuged as above. The supernatant was precipitated with 30% saturated ammonium sulfate and the precipitate redissolved and dialyzed against 50 mM Tris, 150 mM NaCl, pH 7.5 (TBS). NaCl was increased to 1.7 M, precipitating collagen type IV, which was removed by centrifugation. The supernatant was applied to a Bio-Gel A5m (2.5 cm × 120 cm) gel filtration column eluted with TBS. Fractions containing the laminin-entactin complex were pooled, concentrated, and dialyzed against TBS-2 M guanidine HCl. The dialysate was applied to a Sephacryl-S400HR (2.5 cm × 140 cm) column eluted with the same buffer. Laminin (850 kDa) (42) and entactin (148 kDa) (43, 44) fractions were pooled, dialyzed against TBS, concentrated, frozen in N₂(l), and stored at 70 °C. Purification steps were monitored by SDS-PAGE (45). Perlecan was purified as described previously (46). C-IV and fibronectin were purchased from Sigma.

An enzyme-linked immunosorbent assay technique was carried out to study the interaction between BM proteins, apoSAA, and AA peptides. Rabbit anti-perlecan was generated against EHS perlecan. Anti-laminin and anti-C-IV were purchased from Sigma, and anti-entactin from Upstate Biotechnologies Inc. Polystyrene microtiter plates (Immulon 4, Dyntech Laboratories) were coated with 100 µl of fibronectin (0.5 µg/ml), entactin (0.2 µg/ml), or collagen type IV (0.5 µg/ml) in 20 mM NaHCO₃, pH 9.6. For apoSAA (0.8 µg/ml) and AA peptides (0.8 µg/ml), 4 M urea was included in the coating buffer to disaggregate soluble complexes. After overnight incubation at 4 °C, the plates were washed with 20 mM Tris-HCl, 150 mM NaCl, pH 7.5 (TBS), then incubated with 1% BSA in TBS (150 µl) for 2 h at 37 °C to block residual hydrophobic surfaces. Plates were washed with TBS containing 0.05% (w/v) Tween 20 (TTBS), incubated with ligands at different concentrations in the same buffer, and left overnight at 4 °C to allow maximum binding. Different additives were included in the binding buffer as stated under "Results." Plates were washed again in TTBS and incubated with mouse anti-laminin IgG (Sigma) diluted 1:750 in TTBS, 0.1% BSA for 2 h at 37 °C, rewashed, then incubated in the same way with goat anti-rabbit IgG conjugated with alkaline phosphatase (Boehringer Mannheim) at a 1:500 dilution. After washing, bound IgG was detected by the addition of an alkaline substrate solution containing 2 mg/ml p-nitrophenyl phosphate, 0.1 mM ZnCl₂, 1 mM MgCl₂, and 100 mM glycine, pH 10.0. Plates were left at room temperature for 15-30 min, and the reaction was stopped with 50 µl of 2 M NaOH. The absorbance due to the released p-nitrophenol was measured at 405 nm with a Titertek Multiscan/MCC 340 (Flow Laboratories). Controls using BSA-coated wells were included in all experiments to which binding was negligible. Additional controls included the incubation of laminin with different binding buffers such as TTBS, TTBS plus 0.3 M NaCl, TTBS plus 2 M urea, TTBS plus 10 mM N-ethylmaleimide, and TTBS plus 1 mM dithiothreitol. These conditions had no significant effect on the sensitivity of the antibody detection of laminin. The amount of bound ligand was determined by subtracting the absorbance of wells coated only with BSA from those coated with the test protein. For direct quantitation, laminin standards were precoated onto wells on the same plates as the test proteins in order to generate standard curves. Coating efficiency of laminin, apoSAA, and AA peptides on the microtiter plates was 90-100% based on the amount of residual protein remaining in the coating buffer after incubation. Binding data was analyzed as described previously (47, 48), with a nonlinear curve fit program (SigmaPlot, Jandel Scientific) using Equation 1 for a one-binding site model with nonspecific binding or Equation 2 for a two binding-site model with nonspecific binding, where S is the proportionality constant for nonspecific binding and L is the laminin concentration.

(Eq. 1)

(Eq. 2)

In all cases the data fit the one-site model the best, and nonspecific binding was very low (S < 10¹⁰).

Polymerization of laminin was assayed as described by Yurchenco et al. (34). Laminin (0.3 mg/ml), initially centrifuged to remove aggregates, was incubated for 4 h at 37 °C in TBS with 1 mM CaCl₂ or 15 µM ZnCl₂ or CaCl₂/ZnCl₂. A 10 and 20 M excess over laminin of apoSAA (43 µg/ml) and HDL_SAA (670 µg/ml), respectively, was included in a series of tubes to test their effect on laminin polymerization. Molarity was based on a M_r = 850,000 for laminin (42), M_r(average) = 12,200 for apoSAAs (49) (apoSAA1 = 12,600, apoSAA2 = 11,800) and M_r = 200,000 with a protein composition of 50% for HDL/apoSAA (49, 50). After incubation, samples were centrifuged for 15 min, at 12,000 × g, and the polymerized fraction was calculated by subtracting the supernatant concentration from the total.

RESULTS

All BM components used in these binding studies were extracted from mouse EHS tumor as described previously, prepared by us or purchased from scientific companies (see "Materials and Methods"). Proteins were evaluated by SDS-PAGE (45) prior to use (data not shown). ApoSAA was isolated from HDL_SAA lipoprotein particles in inflamed mice, which contained the constitutive apolipoproteins, apoA-I (23 kDa), apoA-II (8720 kDa), apo-C (3500 kDa), plus the apoSAA isoforms apoSAA4 (14 kDa), apoSAA1 (12.6 kDa), and apoSAA2 (11.8 kDa) as seen by SDS-urea-PAGE (Fig. 1). Formic acid-treated HDL_SAA was resolved into two peaks on gel filtration. The first peak contained most of the apolipoproteins (apoA-I, -II, and -C) and lipid eluting as a high molecular weight HDL remnants. The second peak was composed of apoSAA1 and apoSAA2 and was used in the binding experiments described herein (referred to as apoSAA for simplicity).

A preliminary screen of the major the BM components revealed that laminin had the highest binding activity (K_d  2 nM) for apoSAA preparations (Table I). Laminin also bound the AA peptides isolated from AA amyloid laden spleens with similar affinity. Little or no binding was detected between apoSAA and the laminin-entactin complex, entactin, or perlecan. C-IV bound with a K_d of about 14 nM. The negative result with laminin-entactin complex suggested that entactin blocked laminin-apoSAA binding sites, which was confirmed by competition assays described below.

Table I. Summary of binding activities for the major basement membrane components Ligand Protein/peptide-coated Kda nM Laminin ApoSAA 2.2 ± 0.9 (n = 8) AA peptides 1.9 ± 1.3 (n = 3) Collagen type IV ApoSAA 13.9 ± 1.7 (n = 3) Laminin-entactin ApoSAA NBb (n = 3) Entactin ApoSAA NB (n = 4) Perlecan ApoSAA NB (n = 3) Laminin Entactin 1.8 ± 1.6 (n = 12) Laminin-entactin Collagen type IV 2.5 ± 1.6 (n = 3) Perlecan (HSPG) Fibronectin 1.6 ± 0.1 (n = 3) a  Dissociation constants (Kd) are shown as the mean and standard deviation calculated for n experiments. Ligand binding to immobilized (coated) protein was assayed by ELISA and the data analyzed with a nonlinear curve fit program using an equation for a one-binding site model. b  Binding very low or not detected.

Laminin-entactin, laminin-entactin-C-IV, and perlecan-fibronectin binding experiments were included as controls testing both the normal binding activities of these BM components, as well as the accuracy of the assay. The dissociation constants observed (K_d = 1.6-2.5 nM) were all within expected values previously reported (29, 30, 31). For some experiments the binding maxima were also measured (see below), although the precise number of binding sites per molecule could not be obtained with this assay because neither the orientation nor the availability of binding sites could be determined.

Since laminin showed the highest binding activity for apoSAA, we focused our attention on the physicochemical nature of this interaction. Denaturation by heat had dramatically different effects on the two proteins (Fig. 2). Laminin binding activity was completely destroyed by heat denaturation (100 °C, 5 min), while surprisingly, a similar treatment for apoSAA (in 20 mM bicarbonate, 4 M urea, pH 9.6) appeared to increase affinity 3-fold and binding maxima by a third (relative absorbance, nanogram amounts not determined). Similar results were obtained with AA peptides but with a more modest improvement in both affinity and binding maxima (Fig. 2B).

Laminin-apoSAA binding activity was also influenced by a number of common metals (Fig. 3, Table II). At their respective normal plasma concentrations, ZnCl₂ was found to enhance laminin binding activity the most (K_d = 1.8 nM, B_max = 20.1 ng), CuCl₂ and CdCl₂ promoted binding to a lesser degree, while MgCl₂ and CaCl₂ appeared to have an inhibitory effect (compared with EDTA; K_d = 7.7 nM, B_max = 9.9 ng). The inhibition by CaCl₂ was not due to the promotion of Ca²⁺-dependent polymerization of laminin since both the temperature (4 °C) and laminin concentrations (up to 16 µg/ml) in this assay were well below the 37 °C and 0.1 mg/ml critical laminin concentration required for polymerization to take place (34). Trivalent metals such as Al³⁺ and Fe³⁺ were found to cause a nonspecific increase in laminin binding and are probably not important for this interaction (data not shown).

Table II. Summary of laminin:apoSAA binding activity under different conditions Additives Kd Bmax nM ng Nonea 3.1 ± 1.8b 4.5 ± 1.5 (n = 6) EDTA (5 mM) 7.7 ± 4.2 9.9 ± 3.3 (n = 6) ZnCl2 (15 µM) 1.8 ± 0.6 20.1 ± 5.4 (n = 8) NaCl (0.3 M) NBc <1 (n = 4) CaCl2 (2 mM) 12.5 ± 5.6 4.0 ± 2.7 (n = 3) CdCl2 (15 µM) 4.1 ± 2.2 8.5 ± 1.5 (n = 4) CuCl2 (15 µM) 3.8 ± 0.9 6.8 ± 2.4 (n = 2) MgCl2 (5 mM) NB <1 (n = 3) Heparin (1 µg/ml)/ZnCl2 (15 µM) NB <1 (n = 3) Urea (2 M)/ZnCl2 (15 µM) NB <1 (n = 3) DTTd (1 mM)/ZnCl2 (15 µM) 3.4 ± 2.7 1.4 ± 0.9 (n = 2) NEM (10 mM)/ZnCl2 (15 µM) 1.5 ± 0.7 4.3 ± 2.0 (n = 3) Entactin (5 M excess over laminin)/ZnCl2 (15 µM) 5.9 ± 0.5 10.8 ± 1.8 (n = 2) a  Incubation buffer only: 20 mM Tris, 150 mM NaCl, 0.05% Tween, pH 7.5. b  Mean and range or standard deviation calculated for n = 2 or n  3 experiments, respectively. c  Unable to calculate due to very low binding. d  DTT, dithiothreitol.

Binding activity was affected by the pH of the binding buffer (Fig. 4) with a change of as little as 0.5 pH units from pH 7.5 being enough to lower specific binding activity (Fig. 4). Neutral to acidic pH levels had the greatest influence on binding maxima, B_max = 4.6 ng and 3.1 ng at pH 7.0 and pH 6.0, respectively, while affinity was affected most at alkaline pH levels: K_d = 5.0 nM and 8.3 nM at pH 8.0 and pH 9.0, respectively. Nonspecific binding to BSA was not changed over the pH range tested (data not shown). Hence laminin-apoSAA binding appeared to be surface charge-dependent, although protein conformation may have also been affected.

A number of compounds were also discovered to affect laminin-apoSAA binding activity (Fig. 5, Table II). Chemical denaturation with urea (2 M) prevented binding, confirming that the interaction was protein conformation-dependent. Increasing the NaCl concentration to 0.3 M also significantly reduced binding consistent with the binding sites having an ionic nature. Heparin blocked laminin-apoSAA binding most likely by steric hindrance, occupying either one or more heparin binding sites on laminin (51, 52), or the putative glycosaminoglycan binding site on apoSAA (residues 82-87) (53).

Alkylation with N-ethylmaleimide (NEM) without reduction of disulfide bonds also reduced the binding maxima by 79% without significantly affecting the K_d. This result indicated that the apoSAA binding site(s) on laminin contains one or more sulfhydryl groups (apoSAA has none). Laminin in fact has 42 Cys-rich repeats found on the amino-terminal ends of its three subunits, of which 12 contained nested zinc finger consensus sequences (41, 54, 55) (Fig. 6).

Since the BM appears to be disorganized near AA amyloid deposits, we investigated the possibility that laminin-apoSAA binding can disrupt key laminin interactions important for BM assembly. Two such laminin interactions were investigated: one involving self-assembly forming homopolymers (21, 22), and the other a high affinity interaction with entactin, another BM protein important for the cross-linking of laminin and collagen type IV polymer networks (30, 56). Laminin polymerized in a Ca²⁺-dependent manner as reported previously (33, 34, 35) unaffected by ZnCl₂ (41) (Fig. 7). Neither apoSAA nor the HDL_SAA particle at 10 and 20 M excess of laminin, respectively, inhibited laminin polymerization. However, apoSAA and entactin could act as mutual competitors for laminin binding, suggesting their respective binding sites colocalized (Fig. 8). Even at the very low concentrations (picomolar) used in the ELISA, purified laminin-entactin complex appeared to remained intact with little or no laminin-apoSAA binding detected (Table I). When entactin was included at a 5 M excess over laminin, binding to apoSAA was reduced probably by a mixture of competitive and noncompetitive inhibition since both affinity and binding maxima were significantly reduced (Fig. 8B, Table II).

DISCUSSION

ApoSAA was first discovered in the serum because of its cross-reactivity with antisera against AA peptides isolated from AA amyloid (57, 58), and its deposition as amyloid appears to be an aberrant consequence of the acute phase inflammatory response. Despite more than 20 years of research since its discovery, the mechanism by which soluble apoSAA becomes deposited as an insoluble fibrillar structure has remained unknown. However, based mostly on work with animal models, some pieces of the amyloid puzzle have been revealed. In 1985, Snow et al. (15) identified the major glycosaminoglycan in amyloid deposits as heparan sulfate (HS), which was subsequently shown to be linked to the protein core of the BM-type proteoglycan, perlecan (15, 16, 17). The significance of this find was further corroborated when of a number of different glycosaminoglycans tested, HS alone increased the -sheet content for apoSAA2, the amyloid precursor, but not for apoSAA1 nor apoSAA_CE/J, both nonamyloidogenic proteins (23, 26). Later, it was observed that in fact all the major BM components, including perlecan, codeposited with amyloid.

From these results a hypothesis was put forth that one or more BM components may be involved in nucleation events leading to fibril formation. Such a mechanism would probably involve physical interactions with apoSAA, and after testing the major BM components, a saturable, high affinity association was detected between laminin (laminin-1 isoform), and an apoSAA preparation containing apoSAA1 and 2. This interaction involved a single class of binding sites, which appeared to be conformation-dependent, ionic in nature, and significantly enhanced by Zn²⁺. Heat or urea denaturation rendered laminin inactive. Boiling of apoSAA in 4 M urea, however, caused the binding activity to increase, indicating that the binding site on apoSAA may be a relatively short, peptide sequence made more accessible on denaturation of apoSAA.

The ionic nature of the interaction became apparent when the NaCl concentration was increased or when the pH deviated from pH 7.5, both resulting in decreased binding. The binding maxima was particularly affected by lowering the pH to 7.0 and 6.0. In this pH range the imidazole ring of His (pK_a  6.0-7.0) may have become protonated, gaining a positive charge and contributing to the reduction in binding. Additionally, protonation of Cys sulfur atoms would have reduced their ability to form a zinc finger (discussed below). As the pH was increased to pH 8.0 and 9.0, binding affinity was reduced without significantly affecting binding maxima. At these pH levels, the Cys₄-Zn²⁺ structure is probably maintained, but a change in the ionization state of Lys and Tyr side chains (pK_a = 10.8 and 10.1, respectively) may have taken place, causing the reduction in binding activity. These data suggest that the laminin and/or apoSAA binding sites contain Cys and His residues. However, one should bear in mind that, the binding experiments were performed using intact proteins, and the ionization state for the mentioned residues is influenced by their unique microenvironments; hence, predicting the impact of these residues based solely on the theoretical pK_a of their individual side chains is speculative. In addition, the binding sites on laminin and apoSAA are likely affected simultaneously, and potentially in different ways, further complicating the interpretation of the binding data.

Since the region containing the first 15 residues of apoSAA is hydrophobic, and the binding activity for the AA peptides that are missing COOH-terminal 27-45 residues was similar to that for apoSAA, it is plausible that the laminin binding site is located between residues 15 and 76, a region in which ionizable residues are well represented. Experiments are presently being carried out to map the binding site within this region. Preliminary results suggest that apoSAA1 and apoSAA2 have common laminin binding sequences since both bound laminin equally well.²

In addition to being conformation-dependent, the apoSAA binding site(s) on laminin may also require Zn²⁺ and free sulfhydryls, as demonstrated by the stimulatory and inhibitory effects of ZnCl₂ and NEM, respectively. Similar effects were observed for laminin binding of PP, the Alzheimer's amyloid precursor protein (48) and more recently, for laminin-entactin binding (41). Incubation with a large excess of NEM (10 mM) was found to reduced the B_max of laminin-apoSAA binding by 79%. The amino acid sequence in and around the remaining binding sites may have protected them from modification by NEM. Complete alkylation of laminin (reduce disulfide bonds prior to alkylation) could not be tested, since reduction alone destroyed binding activity virtually completely (Table II). Investigation of the sequence for the entactin binding site, located on the laminin -chain (36), revealed that it contained a zinc finger consensus sequence (41). In fact, 12 potential zinc finger motifs within as many Cys-rich repeats could be found distributed on the laminin short arms, and laminin-bound zinc (up to 8 mol/mol of laminin) has been detected (41).

Like entactin, apoSAA had no effect on laminin polymerization, a Ca²⁺-dependent process requiring the amino-terminal globular domains distal to the Cys-rich repeats (33, 34, 35). In addition, the ability of apoSAA and entactin to act as mutual competitors for laminin binding indicated that their binding sites may map close together. When the laminin-apoSAA binding activity was assayed with entactin as the competitor, both the affinity and binding maxima were reduced, reflecting either allosteric or a mixture of steric and allosteric inhibition. Therefore, one or more apoSAA binding sites must map some distance from the entactin site. However, we cannot exclude the possibility that one binding site also overlaps with the entactin one which would contribute to the reduction in B_max observed. By occupying its binding site of no more than 58 residues on the laminin -chain (36), entactin as a relatively large rod-shaped protein (148 kDa) probably exerts its blocking effect over a wide region of the  and -chain short arms. There are six and three zinc finger-containing repeats on the  and -chains, respectively, which might contain an apoSAA binding site or sites. Taken altogether the data suggest that one or more, Zn²⁺-coordinated, zinc finger-like sequences may represent the actual apoSAA binding sites, or at least contribute to them significantly.

Based on their codeposition during amyloidosis (16) and saturable, high affinity association in vitro, it is likely that laminin-apoSAA binding takes place at locations of amyloid deposition. Previous reports have provided strong evidence that perlecan through its HS side chains promotes fibrillogenesis (26), possibly by attaching to a putative GAG binding site (residues 82-87) on apoSAA2 (53). High affinity binding has been reported between perlecan and PP (47); however, perlecan-apoSAA binding was not detected in this study. The avidity of this interaction may be too low to have been measured by ELISA but supports a role for laminin in apoSAA fibrillogenesis. Specifically, this role may be to sequester apoSAA from plasma causing its local concentrations to increase and/or provide a surface in close proximity to HS, on which nucleation events could take place. Laminin may also contribute directly to changes in apoSAA conformation favoring fibril formation although this has not been investigated.

The aberrant synthesis of BM components may be a contributing factor in amyloidogenesis. In brains of subjects with Alzheimer's, a significant increase in laminin mRNA has been reported (18). In mice experimentally induced to develop AA amyloid, increased levels of C-IV, perlecan, laminin -chain, and entactin mRNAs are detected at 18-24 h, which precedes the histological accumulation of AA-fibril and BM components (19, 20).³ Under normal circumstances, newly secreted BM components would spontaneously assemble through specific interactions to generate the BM matrix (21, 22) but at sites of amyloid depositions this process appears to be blocked. We suspected that apoSAA might interfere with normal BM assembly by inhibiting laminin's BM forming potential. However, neither HDL_SAA nor purified apoSAA affected laminin polymerization, an integral step in BM assembly (21, 22). In addition, apoSAA had little or no binding activity for laminin-entactin complexes, the form in which laminin is secreted and incorporated into BM by endothelial and other cell types (59, 60, 61). These complexes are very stable (K_d  1 nM) and require chaotropic agents such as guanidine HCl to be dissociated (32).

Hence it is unlikely that apoSAA could displace entactin from laminin, and the demonstrated ability of apoSAA to inhibit entactin binding of laminin may simply reflect the close proximity of their binding sites and not a dynamic exchange contributing to a disturbance in BM structure. The possibility that apoSAA may be influencing BM assembly through C-IV, for which it had a lower binding affinity, was not addressed in this study. However, laminin and C-IV homopolymer networks are believed to form independently (34), and a laminin-rich BM could be expected to form in the absence of C-IV polymerization. To our knowledge, widespread disruption in BM metabolism during inflammation has not been noted and BM homeostasis is probably not directly affected by apoSAA. Why BM synthesis would be associated with amyloid deposition in not known. The form in which laminin is secreted is also unknown, but our results suggest some laminin would have to be secreted free of entactin to allow apoSAA access to bind. This supports the idea that a disturbance in BM metabolism is linked, or even an underlying factor in amyloidogenesis.

The importance of the laminin-apoSAA interaction may not be solely pathological. Indeed the avidity of the laminin-apoSAA association suggests that apoSAA's function may be at the level of the BM. Additionally, because the apoSAA binding activity favors free laminin over the laminin-entactin complex, it is possible that abnormal or damaged BM may be the target of apoSAA. The rapid and transient appearance of apoSAA early in the injury/repair cascade is consistent with its binding to such BM soon after the injury has occurred, and probably not to newly synthesized BM which would appear later in the repair process, after most of the apoSAA is cleared from the plasma (62). Cellular release of proteases from damaged cells and activated neutrophils at injury foci would cause local damage to BMs, potentially exposing apoSAA binding sites on laminin. Entactin is known to be the most susceptible of all the BM proteins to proteolysis (63, 64). In contrast, the laminin short arms containing the Cys-rich repeats are remarkably protease-resistant (65, 66). Hence, it is conceivable that more laminin would be available for apoSAA binding at sites of injury.

Although the transcriptional and post-transcriptional regulation of acute phase apoSAA synthesis is well understood (67, 68, 69), the actual function of the proteins remains confused. Because a significant pool of apoSAA is located on HDL, which appears to enhance HDL binding to macrophages, it has been postulated that apoSAA focuses reverse-cholesterol transport (cholesterol transport from periphery to liver) to sites of injury, thereby optimizing removal of cholesterol from such sites (70). However, apoSAA has also recently been shown to increase secretory phospholipase A₂ activity, which the authors suggested would lead to an increase in the ratio of HDL cholesterol-phospholipid, thereby promoting cholesterol delivery to sites of injury (71). Either scenario could be accommodated by apoSAA if it functioned to target HDL_SAA particles to exposed laminin at sites of injury. Platelets use a similar strategy in response to injured blood vessels adhering to exposed BM via C-IV/laminin receptors (72, 73).

Some of the other functions proposed for apoSAA are not related to cholesterol metabolism and imply apoSAA is sequestered on HDL, possibly in an inactive state. These include inhibition of lymphocyte antibody response (74), the oxidative response in neutrophils (75), platelet activation (76, 77), and interleukin-1 and tumor necrosis factor-induced fever (78). ApoSAA has also been reported to block lymphocyte and metastasized tumor cell adhesion to BM (79), and to have monocyte and leukocyte chemoattracting activity (80, 81), both of which could be mediated through laminin-apoSAA binding. In the latter proposal apoSAA's chemotactic activity would be activated after release from the HDL, possibly by proteolysis, at or near sites of injury. This is analogous to the situation with complement factor C5, which is activated only after enzymatic cleavage by C5 convertase releasing the C5a peptide a chemoattractant and C5b a component of the membrane attack complex (82). As an amphipathic protein apoSAA contains putative lipid, calcium (83), glycosaminoglycan binding (53), and protein kinase C phosphorylation sites (84), plus a highly conserved domain unique to apoSAAs (2), and it is conceivable that these domains may contribute to different functions during inflammation.