The HHQK Domain of β-Amyloid Provides a Structural Basis for the Immunopathology of Alzheimer’s Disease*

The β-amyloid peptide 1–42 (Aβ1–42), a major component of neuritic and core plaques found in Alzheimer’s disease, activates microglia to kill neurons. Selective modifications of amino acids near the N terminus of Aβ showed that residues 13–16, the HHQK domain, bind to microglial cells. This same cluster of basic amino acids is also known as a domain with high binding affinity for heparan sulfate. Both Aβ binding to microglia and Aβ induction of microglial killing of neurons were sensitive to heparitinase cleavage and to competition with heparan sulfate, suggesting membrane-associated heparan sulfate mediated plaque-microglia interactions through the HHQK domain. Importantly, small peptides containing HHQK inhibited Aβ1–42 cell binding as well as plaque induction of neurotoxicity in human microglia. In vivo experiments confirmed that the HHQK peptide reduces rat brain inflammation elicited after infusion of Aβ peptides or implantation of native plaque fragments. Strategies which exploit HHQK-like agents may offer a specific therapy to block plaque-induced microgliosis and, in this way, slow the neuronal loss and dementia of Alzheimer’s disease.

Alzheimer's disease (AD) 1 is a neurodegenerative dementia associated with loss of neurons and the appearance of reactive glia (1,2). The neuropathological hallmarks of this disorder include neuritic and core senile plaques, which are complex aggregations of proteins composed largely of a 42-residue peptide, ␤-amyloid (A␤1-42) (3). A preponderance of evidence indicates that a chronic imbalance in the production and clearance of A␤ initiates pathological responses, which include neuritic and synaptic abnormalities, neurofibrillary tangles, and loss of neurotransmitters (4). Although A␤ peptides found within senile plaques are widely believed to damage the nervous system (4), the mechanisms by which these molecules actually drive AD pathology remain uncertain. One pathway for A␤-induced neuron damage may involve inflammatory cells, for it has long been recognized that reactive microglia are closely associated with neuritic and core plaques (5)(6)(7)(8). Since reactive microglia release such bioactive agents as proteolytic enzymes, cytokines, free radicals, and nitric oxide (9 -11), the immunopathology of AD is likely to involve microglial release of cellular poisons. Recent studies also show that exposure to neuritic and core plaques stimulated cultured human microglia to secrete a potent neurotoxic amine that was found in the AD brain (7). Moreover, A␤1-42, the most abundant component of neuritic and core plaques (12) served as the major plaque-derived signal to evoke neuron-killing by microglia (13). As described here, A␤13-16 (HHQK) is necessary for an initial microglial interaction with plaque through a cell-surface binding site involving heparan sulfate. Small peptides containing HHQK block plaque induction of neurotoxic microglia in vitro and reduce inflammation in vivo. Strategies exploiting HHQK-like agents to suppress plaque-microglia interactions offer a selective means to prevent AD immunopathology.

EXPERIMENTAL PROCEDURES
Cell Culture-Rat microglia were isolated from newborn animals using the method of Giulian and Baker (14), with recovery of a Ͼ98% homogenous population monitored by binding of the fluorescent probe, acetylated low density lipoprotein labeled with 1,1Ј-dioctadecyl-3,3,3Ј,3Ј-tetramethylindocarbocyanine (DiI-Ac-LDL). Human microglia (Ͼ98% homogeneity) were obtained from adult frontal cortical tissue isolated within a 6-h postmortem interval as described earlier (7) and grown in chemically defined medium. Cultured neurons prepared from embryonic rat hippocampus (13) consisted of process-bearing neurons (10 -20% of total cell population) atop a bed of astroglia (Ͼ70% of the cells) containing microglia (5-10% of the cells). In order to eliminate microglia, cultures were exposed to saporin (a ribosome-inactivating protein) coupled to acetylated LDL (Ac-LDL) (13). Saporin-Ac-LDL selectively bound to scavenger receptors and at 10 g/ml, after 18 h, reduced microglial numbers to Ͻ0.1% of the total population, with no effect on viability of either neurons or astroglia. After 14 days in vitro, cultures (with a final concentration of 0.6% fetal bovine serum) were exposed to test substances in the presence or absence of exogenous microglia for 72 h. Human microglia (250,000/chamber) were placed in Millicell chambers (0.4-m filter insert with a 12-mm diameter; Millipore, Bedford, MA) atop cultures of rat hippocampal cultures depleted of microglia. All neuronal cultures were fixed in 3% paraformaldehyde at room temperature for 6 h and immunostained by overnight incubation with a mixture of anti-neurofilament antibodies (SMI-311, 1:150; RT-97, 1:150; Sternberger Monoclonals, Inc.) plus anti-MAP-2 (Boehringer Mannheim, 184959; 1:200) at 4°C in the presence of 2% horse serum and 0.3% Triton X-100. Data were expressed as percent mean survival in relation to parallel untreated control cultures, after scoring eight randomly selected fields for each of three coverslips from at least three independent experiments.
Preparation of Plaques, Proteins, and Peptides-Native amyloid protein aggregates were isolated from AD neocortical gray matter laden with neuritic and core plaques using discontinuous sucrose gradients, with recovery of neuritic/core plaque fragments from the 1.4/1.7 M sucrose interface (3). Native A␤ monomers (a mixture of A␤1-42 and 1-40) were extracted from these plaque fragments and separated by gel filtration (peak 5 as described in ref. 13). All synthetic A␤ peptides (California Peptide; Napa, CA) were dispersed in phosphate buffered saline (pH 7.2; microtip sonication; Branson Sonifier 250) Ͼ7 days prior to use. Under these conditions, synthetic A␤ peptides form fibrils and aggregates (12,15). Chemical modifications of A␤1-42 were carried out at room temperature for 2 h, and targeted amino acids were modified either at Arg 5 using 2.5 mM cyclohexanedione (16) in 50 mM sodium borate buffer (pH 8.9); at Tyr 10 using 20 M tetranitromethane (17) in 50 mM Tris buffer (pH 8.0); at Lys 16 and Lys 28 using 10 mM ethyl acetimidate (18) in 200 mM triethylamine-HCl buffer (pH 10.0); or at His 6 , His 13 , and His 14 using 1 mM diethylpyrocarbonate (DEPC) (19) in 100 mM potassium phosphate buffer (pH 6.8). Decarbethoxylation (19) involved the addition of 1.5 M hydroxylamine (pH 7.0) at room temperature overnight. Glutamine residues were enzymatically modified using 0.026 unit/ml transglutaminase (protein-glutamine:amine ␥-glutamyltransferase, EC 2.3.2.13; from Sigma) in 100 mM Tris buffer (pH 7.4) containing 200 M ethylamine at 37°C for 2 h (20). In all cases, samples were concentrated and washed using ultrafiltration (Centricon 2; recovery of modified peptides Ͼ80%). Biological activities of modified A␤1-42 peptides were tested at Ն1 moles/liter in neurotoxicity assays. Structural modifications of A␤ peptides were confirmed by reverse phase-high performance liquid chromatography (Zorbax 300SB-C3 or Zorbax SB-C18 (4.6 mm ϫ 25 cm; MAC-MOD Analytical, Inc., Chadds Ford, PA) at 70°C) and mass spectroscopy (Sha-Nan Lin, Ph.D. Laboratory of Analytical Chemistry, University of Utah). Matrix-assisted laser-desorption/ionization mass spectrometry using time of Binding Assays-Fluoresbrite carboxylate YG microspheres (1.0-m diameters; 0.5 ml of a 2.5% suspension; Polyscience Inc., Warrington, PA) were activated with 1% carbodiimide for 4 h at room temperature. Washed spheres were resuspended in 0.2 M borate buffer (pH 8.5) in the presence of 300 g of A␤1-42 (Spheres A␤1-42 ) or 400 g of bovine serum albumin (Spheres BSA ) in 6% Me 2 SO. After overnight mixing at room temperature, the spheres were washed extensively and blocked by 1 M glycine (pH 8.0) for 30 min. Isolated microglia (1,250/mm 2 ) adherent to 13-mm glass coverslips in 24-well culture plates were mixed with 250,000 Spheres A␤1-42 in the presence or absence of A␤ peptides or glycosylaminoglycans (chondroitin sulfate and heparan sulfate; Sigma) at 37°C. Mild trypsinization of microglia (5000 units/ml; Sigma) was carried out at 37°C for 60 min. Prior to binding, trypsin was inactivated by the addition of soybean trypsin inhibitor (500 g/ml; T9003 Sigma) at a ratio of 1 mg of inhibitor to 10,000 units of enzymatic activity, and followed by multiple washings of cell cultures with medium containing 10% fetal calf serum. Enzyme-treated control cultures included microglia that were incubated for 60 min with trypsin immediately inactivated by trypsin inhibitor prior to addition to cultures. Microglia were also exposed to heparitinase (heparin lyase, EC 4.2.2.8, from ICN; two consecutive treatments each of 0.01 unit/ml for 60 min) or chondroitinase (chondroitin ABC lyase, EC 4.3.3.4, from Sigma; two consecutive treatments each of 0.02 unit/ml for 60 min). The glycosylaminoglycan synthesis inhibitor, 4-methylumbelliferyl-␤-D-xyloside (␤-D-xyloside) was acquired from Sigma. Binding assays were carried out at 37°C for 4 h in chemically defined culture medium after which coverslips were dipped 10 times in phosphate buffer and fixed with 4% buffered paraformaldehyde. Microsphere adherence to cells was scored at ϫ 200 magnification with phase/fluorescence microscopy. Data, expressed as mean percent inhibition of sphere binding, was calculated as (1 Ϫ (total number of spheres per field of treated group Ϫ background binding)/ (total number of spheres per field of untreated control cultures Ϫ background binding)) ϫ 100%. Background binding was defined as nonspecific adherence of unmodified Fluoresbrite microspheres exposed to sister cultures under identical conditions. All values were based upon five coverslips from at least three independent experiments.

FIG. 1. Microglial activation requires a specific A␤ domain. Panel A,
testing of various A␤ peptides in a neurotoxicity assay using rat hippocampal cultures supplemented with microglia showed a 70 -80% killing of neurons after exposure for 72 h to human A␤1-40, or A␤1-42. Elimination of microglia from the cultures prevented neuron death. The pattern of neuron killing by synthetic peptides was similar to that elicited by native A␤ monomers purified from plaques. Interestingly, rodent A␤1-40 (Gly 5 , Phe 10 , and Arg 13 ) did not activate microglia. The A␤ peptides containing either the N terminus of the peptide (A␤1-11, A␤1-16, A␤13-16, and A␤1-28) or C terminus (A␤17-42, A␤25-35) alone were also inactive. Panel B, the capacity of A␤1-42 (1 mol/liter) to activate microglia was examined after modification of the N-terminal region by chemical or enzymatic methods. Altering residues in the 13 to 16 domain blocked the A␤1-42 induction of neurotoxic microglia. Cyclohexanedione (CHD) modification of Arg 5 ; tetranitromethane (TNM) modification of Tyr 10 ; DEPC modification of His 6 , His 13 , His 14 with hydroxylamine used to reverse the DEPC effect; transglutaminase (TNG) modification of Gln 15 ; ethyl acetimidate (EAM) modification of Lys 16 and Lys 28 ; A␤1-42 Gln13Gln14 represents a synthetic peptide with replacements of His 13 and His 14 . mmol/liter) peptide, of A␤13-16 (1 mmol/liter) peptide, or of artificial cerebral spinal fluid alone. Solutions of synthetic human A␤1-42 (1 mmol/liter) peptide were mixed with 2 volumes of A␤1-5 (1 mmol/liter), 2 volumes of A␤13-16 (1 mmol/liter), or 2 volumes of cerebral spinal fluid alone. Plaque suspensions (4 l, total volume) using a 22-gauge catheter or A␤ solutions (1.5 l, total volume) using a 33-gauge syringe were deposited bilaterally into the neocortex at 4.0 mm posterior to the bregma, 3.5 mm lateral from the midline, and 1.3 mm below the surface of the brain at a rate of 0.25 l/min. Animals were sacrificed 7 days after injection with 0.8 ml/100 g of body weight of the anesthetic mixture, followed by cardiac perfusion with 50 ml of phosphate-buffered saline (pH 7.4) containing 25 units/ml heparin. Brains were then cut by razor blades into 2-mm thick sections using a coronal rat brain matrix (Ted Pella Inc., Redding, CA) and incubated in cell culture medium containing 250 ng/ml DiI-Ac-LDL for 12 h at 37°C (21). Slices were then fixed with 4% paraformaldehyde in 0.1 M phosphate buffer, pH 7.4, placed in 30% sucrose for 24 h, frozen, and cut into 32-m thick sections. Each brain provided 50 -60 serial sections. Labeled cells were viewed by fluorescence microscopy (523-557-nm excitation/595-nm emission).

Specific Regions of A␤ Are Required for Inducing Microglial
Toxicity-As reported earlier (13), native A␤ monomers isolated from senile plaques induced microglia to release neuron killing factors (Fig. 1A). Although the synthetic forms of human A␤1-40 and A␤1-42 were also potent stimuli for neurotoxic microglia, rodent A␤1-40 was not (Fig. 1A). Since the only differences between the human and rodent peptides are found at residues 5, 10, and 13 (Arg 5 3 Gly; Tyr 10 3 Phe; His 13 3 Arg), it appeared that the N-terminal portion of human A␤ held some primary structure necessary to stimulate microglia. Accordingly, we altered full length human A␤1-42 to mimic the rodent form. Cyclohexanedione modification of the residue Arg 5 or tetranitromethane (TNM) modification of Tyr 10 had no effect upon A␤ activation of microglia (Fig. 1B), suggesting that neither residues 5 nor 10 are essential for cell stimulation. In contrast, diethylpyrocarbonate (DEPC) modifications of His 6 , His 13 , and His 14 , eliminated neuron killing. Reversal of the DEPC modifications by hydroxylamine restored A␤ as a stimulus, pointing to a need for the His residues. The inability of synthetic peptide A␤1-42 Gln13Gln14 to elicit neurotoxic responses ( Fig. 1B) focused attention upon His 13 and His 14 . Exploring residues neighboring His 13 and His 14 , we found that both transglutaminase (TNG) cross-linking of Gln 15 to ethylamine or acetimidination of Lys 16 and Lys 28 by ethyl acetimidate also blocked A␤1-42 induction of neurotoxic microglia (Fig. 1B). Although residues His 13 , His 14 , Gln 15 , and Lys 16 (the HHQK domain within A␤) were required for induction of neurotoxic microglia, the HHQK peptide by itself did not trigger neuron killing (Fig. 1A). It was possible, therefore, that the HHQK domain served as a binding site allowing full length A␤1-42 to interact with microglia. This hypothesis is supported by the fact that A␤17-42 did not activate microglia while A␤10 -42 elicited neuron killing similar to that of A␤1-42 ( Fig. 2A).

HHQK Binding to Microglia Utilizes Membrane-associated
Heparan Sulfate-To delineate further the role of the HHQK domain during A␤ binding to microglia, it was first necessary to produce a stable ␤-amyloid target for cell-based assays. Because A␤ peptides can rapidly aggregate when mixed with cell culture medium and precipitate from solution (12,13,22), we coupled peptides to 1-m diameter fluorescent microspheres. These peptide-coupled spheres, or "artificial plaques," were then added in concentrations of 250,000/well to homogeneous populations of cultured microglia (1,250 cells/mm 2 ) at 37°C; the binding of the spheres to cells was monitored by fluorescence microscopy. In general, spheres coupled to peptides containing the HHQK domain adhered to microglia far more rapidly than did spheres coated with agents lacking the domain. For example, after 4 h of incubation, 250 Ϯ 11 Spheres A␤1-42 per mm 2 were bound to cells compared with 20 Ϯ 8 Spheres A␤17-42 or 35 Ϯ 10 Spheres BSA . The importance of the HHQK domain as a recognition site is evident in photomicrographs which show that far more Spheres A␤10 -42 bound to microglia than did Spheres A␤17-42 (Fig. 3). Microglial avidity for HHQK-containing artificial plaques persisted throughout long term incubations of 72 h in microglia-neuron co-cultures. Moreover, binding of Spheres A␤1-42 to microglia was markedly reduced by mild trypsinization of intact cells (Fig. 4A), suggesting involvement of membrane surface proteins. Importantly, artificial plaques elicited patterns of microglial activity identical to those elicited by native plaques and by unbound peptides, since Spheres A␤1-42 and Spheres A␤10 -42 brought about neuron killing while Spheres A␤17-42 did not (Fig. 2B).
The HHQK domain represents an unusual cluster of basic amino acids that is thought to exist on the hydrophilic surface of A␤ fibrils (22). The availability of this cluster to cell surfaces is of further interest since HHQK also functions as a charged site which binds to heparan sulfate or heparan sulfate-containing glycosaminoglycans (23)(24)(25). To investigate the role of heparan sulfate and HHQK binding to cells, we modified the surface membranes of microglia by enzymatic cleavage. Heparitinase eliminated Spheres A␤1-42 binding (Fig. 4A) to microglia while treatment with chondroitin sulfatase had no effect. The selective sensitivity to heparitinase suggested the need for microglia-associated heparan sulfate in the recognition of HHQK. Moreover, when microglia were incubated with excess amounts of soluble heparan sulfate, chondroitin sulfate, or dextran sulfate, only heparan sulfate blocked Spheres A␤1-42 binding to cells (Fig. 4B). To determine if heparan sulfate was needed for induction of neurotoxic behavior, we next reduced levels of glycosylaminoglycans in long term co-cultures and tested for neuron killing induced by plaques. Heparitinase treatment alone caused a partial reduction in neurotoxicity (Fig. 4C). Because microglia synthesized glycosylaminoglycans during the 72-h assay, we both cleaved cell-surface heparan sulfate and then blocked further glycosylaminoglycan production by addition of 1 mM ␤-D-xyloside (26). Heparitinase plus ␤-D-xylose prevented neuronal killing (Fig. 4C), while chondroitinase, with or without ␤-D-xyloside, showed only a small effect. These observations suggested that the binding of cellassociated heparan sulfate to the HHQK domain of A␤ was necessary for plaque activation of neuron-killing in microglia.
Since the interaction between A␤1-42 and microglia involved cell-surface recognition of the HHQK domain, agents that compete with HHQK recognition might reduce A␤1-42 avidity for microglia. Accordingly, synthetic peptides derived from A␤1-42 were added to homogeneous cultures of microglia in the presence of Spheres A␤1-42 . As shown in Fig. 5, peptides containing the HHQK domain (A␤10 -16, A␤10 -20, A␤12-28, or human A␤1-40) effectively suppressed Sphere A␤1-42 binding to microglia, while peptides lacking HHQK (A␤1-5, A␤17-42, rodent A␤1-40, A␤25-35, or human A␤1-42 Gln13Gln14 ) did not. Further testing of other derivatives from human A␤1-42 confirmed that only those peptides containing HHQK influenced A␤1-42 binding to cells (Table I). Thus, it appeared that competition by HHQK for a binding site found within full length human A␤, as well as competition by soluble glycosylaminoglycan for cell-surface heparan sulfate, inhibited microglial interactions with ␤-amyloid.
Small Peptides Containing The HHQK Domain Suppress Induction of Neurotoxic Microglia-Since peptides containing the HHQK domain blocked A␤1-42 adherence to microglia, they might also impede A␤1-42 induction of microglial neurotoxicity. To test this possibility, various peptides (each at 10 The effects of A␤ peptides upon plaque/microglial or A␤1-42/microglial interactions were measured as inhibition of Spheres A␤1-42 binding to microglia, the blocking of neurotoxicity induced by A␤1-42, and the suppression of neurotoxicity induced by plaques. As shown, peptides which prevent binding to microspheres are the same peptides which suppress toxicity induced either by A␤1-42 or plaque fragments. All blocking peptides contain the HHQK domain. mol/liter), were added to neuron cultures containing microglia in the presence of human A␤1-42 (1 mol/liter). Only A␤1-28, A␤1-16, A␤10 -20, A␤10 -16, or A␤13-16 prevented neuron killing (Table I; Fig. 6A) elicited either by A␤1-42 in solution or by Spheres A␤1-42 . Increasing concentrations of blocking pep-tides, as shown for A␤10 -16 (Fig. 6B), reduced both the number of Spheres A␤1-42 binding to cells and the degree of neuron killing. The plateau of induced neurotoxicity suggested further that A␤1-42 binding sites, which participated in microglial activation, had become saturated (Fig. 6B). Dose-response curves (Fig. 6C) showed two distinct patterns of inhibition, with A␤1-16, A␤10 -16, and A␤13-16 the more potent blocking agents (exhibiting ED 50 values of about 30 nmol/liter) and A␤13-20 and A␤1-28 less potent (ED 50 values of 250 nmol/ liter). These differences in blocking efficacy might be due to the presence of hydrophobic residues 17-20 (Leu-Val-Phe-Phe) in the less potent blocking peptides. It should also be noted that none of the blocking peptides prevented induction of neurotoxic microglia (27) following exposure to zymosan (yeast-wall) particles or to lipopolysaccharide (Fig. 6D, LPS). Such data indicate that HHQK suppresses only microglial toxicity resulting from A␤ exposure and does not inhibit cell killing capacity in general.
Although HHQK-containing peptides inhibited A␤1-42 activation of microglia, it remained uncertain whether such protective effects would occur when cells were exposed to native plaques, which consist of complex mixtures of brain proteins including A␤ peptides (13). To test this possibility, rat or human microglia were placed in cell chambers and mixed with neuritic/core plaque fragments recovered at autopsy from AD cortical tissues. Peptides A␤1-16, A␤13-20, A␤10 -16, and A␤13-16 each blocked neurotoxic activity of either rat (Fig. 7A) or human (Fig. 7B) cells stimulated by plaques, while peptides lacking the HHQK domain (A␤1-5, A␤1-11, A␤17-42, A␤25-35, or A␤36 -42) had no effect. Thus, the presences of small peptides containing HHQK not only impair A␤1-42 recognition by microglia but also plaque induction of neurotoxic cells.
If human A␤1-42 activates rat microglia in culture, it follows that placement of A␤1-42 into the rodent brain should elicit a similar response. Penetrating injury inflicted by a small needle to the neocortex of adult rat will induce a reactive microgliosis that dissipates within 4 days (21). Synthetic human A␤1-42 or fragments of native plaques infused into the neocortex will produce a reactive microgliosis extending beyond the interval brought about by simple needle trauma (28). Accordingly, infusions of A␤ peptides or native plaque fragments were used to assess the immunosuppressive effects of HHQK upon ␤-amyloid-dependent inflammation in vivo. Seven days FIG. 4. Properties of the microglial A␤ binding site. Panel A, enzymatic treatments of microglia alter A␤ binding to cells. Spheres A␤1-42 were incubated with microglia for 4 h following pretreatment of cells with trypsin (5000 units/ml at 37°C for 60 min followed by inactivation with soybean trypsin inhibitor), with heparitinase (heparin lyase, EC 4.2.2.8; two consecutive treatments each of 0.01 unit/ml for 60 min), or with chondroitinase (chondroitin ABC lyase, EC 4.3.3.4; two consecutive treatments each of 0.02 unit/ml for 60 min). Binding by Spheres A␤1-42 to microglia was reduced by trypsin and heparitinase, but not chondroitinase. Panel B, competition with ligands again suggested the involvement of a heparan sulfate-containing site on microglia with reduction of binding Spheres A␤1-42 in the presence of heparan sulfate (10 mol/liter) but not by dextran sulfate (100 mol/liter) or chondroitin sulfate (10 mol/liter). Panel C, plaque induction of neurotoxicity in microglia involves a heparan sulfate-containing site. Microglia mixed with hippocampal neurons were treated with combinations of ␤-D-xyloside (1 mM), heparitinase (0.02 unit/ml), or chondroitinase (0.04 unit/ml), and then exposed to plaques. Enzyme treatments alone, particularly that of heparitinase, brought on some reduction in neurotoxic activity during the 72-h neurotoxicity assay; however, a combination of both enzymatic degradation of heparan sulfate plus competitive blockade of glycosylation by ␤-D-xyloside completely eliminated plaque activation of microglia.
FIG. 5. Small peptides containing the HHQK domain inhibit A␤ binding to microglia. A␤1-42 was coupled to fluorescent microspheres and the Spheres A␤1-42 monitored for specific binding to microglia after 4 h at 37°C in the presence of peptides (all at 10 mol/liter). Only peptides containing residues HHQK (A␤10 -16, A␤10 -20, A␤1-40, A␤12-28) were able to block binding of Spheres A␤1-42 . Nonspecific binding for this assay was measured as the number of unmodified spheres that adhered to sister control cultures.
after infusion of 0.5 nmol of A␤1-42 into rat neocortex, injection sites contained a prominent number of reactive mononuclear phagocytes (Fig. 8A). This local inflammation was markedly reduced, however, if 1.0 nmol of A␤13-16 is co-injected into the same site (Fig. 8B). Importantly, co-injection of 1.0 nmol of A␤1-5, a control peptide lacking HHQK, did not suppress reactive cells (Fig. 8C). Moreover, neuritic/core plaque fragments placed within the neocortex of rats (Fig. 9A) elicited an inflammatory response that was also attenuated by 1.0 nmol of A␤13-16 (Fig. 9B). In this way, both in vitro and in vivo experiments confirm that HHQK-containing peptides reduce microglial responses brought on by plaques or by A␤1-42. Moreover, the specificity of HHQK peptide blocking of A␤induced neuron killing in vitro (Fig. 6C) suggests that such protective effects are restricted to plaque interactions underlying the immunopathology of AD. DISCUSSION Gliosis is an important feature of AD pathology and is closely associated with senile plaques and neuronal injury. Histological study shows, for example, that nearly all neuritic and core plaques are surrounded by clusters of reactive microglia (7,29).
In vitro studies confirm that quiescent microglia found in normal brain will become reactive when placed in contact with isolated neuritic/core plaque fragments, as demonstrated by changes in morphology and release of neurotoxins (7). Such cell culture observations support the idea that immune responses contribute to the neuronal pathology of AD. Further study has shown that plaque-activated microglia produce a neurotoxic amine which can also be recovered from brain tissues heavily laden with clusters of reactive microglia (7). Moreover, this toxin was found to destroy hippocampal pyramidal cells. Since the hippocampus participates in memory and cognition (2), it is reasonable to implicate microglia-derived toxins as destroyers of cognitive function in AD patients (13).
To delineate early events of AD immunopathology, we examined plaque activation of cultured microglia. Fractionation studies of native neuritic and core plaques have shown that A␤ peptides, and not such plaque-associated components as ␣ 1antichymotrypsin or apolipoprotein E, were the principal stimulants which drove microglial reactivity (13). Dose response curves revealed that full-length human A␤1-42 and A␤1-40 were the most effective stimuli for neurotoxic microglia while The numbers of specifically-bound Spheres A␤1-42 were obtained by subtracting the background number of unmodified spheres which adhered to sister cultures. In both conditions, 250,000 spheres were initially added to hippocampal cultures in the presence or absence of blocking peptide. Panel C, dose curves showed a greater blocking capacity for those peptides containing residues within the 1-16 hydrophilic portion of A␤. Addition of more hydrophobic segments (beyond residue 16) diminish the ability of peptide to block A␤1-42 interactions with microglia. Panel D, specificity of peptide blocking effect on activation of neurotoxic microglia was shown by the ability of HHQK-containing peptides (all at 10 mol/liter) to prevent induction of neuron-killing in microglia induced by A␤1-42 (1 mol/liter) but not by zymosan (2 l of suspension/well) or lipopolysaccharide (LPS) (1 g/ml). rodent A␤1-40 (Arg 5 3 Gly; Tyr 10 3 Phe; His 13 3 Arg) was inactive (13). As described here, chemical modifications of synthetic peptides uncovered the significance of human A␤ primary structure by showing that A␤13-16 (the HHQK domain) was necessary for induction of neurotoxic microglia (Table I).
These observations led us to study the binding of A␤ peptides to microglia. Such experiments were hampered by the tendency of A␤1-42 to rapidly form aggregates and fibrils as it was mixed with cell culture medium. In order to provide a stable reagent for quantitative assays, we coupled A␤1-42 to microspheres to create artificial plaques suitable for binding studies. Such Spheres A␤1-42 elicited neurotoxicity, as did native plaques. Excess concentrations of small A␤ peptides containing the HHQK domain prevented microglial binding to full length A␤1-42 and eliminated induction of microglial neurotoxicity. Moreover, mild trypsinization of microglia suggested that surface proteins  were involved in this binding interaction. Further work showed A␤ induction of neurotoxicity required heparan sulfate (sensitivity to heparinase and ␤-D-xyloside blockade) associated with microglial surfaces for an HHQK binding site. Since other investigators have previously identified residues 12-17 of A␤ (VHHQKL) as a binding domain for heparan sulfate (23)(24)(25), we suggest that membrane-associated heparan sulfate plays an important role in the immunopathology of AD by promoting plaque accessibility to brain inflammatory cells.
Current thinking holds that A␤, the major constituent of neuritic and core plaques, participates in the neuron destroying events that cause the dementia of AD. Some laboratories consider A␤ peptides as direct cell poisons (15), while others view these peptides as only one part of a more complex pathogenic process (30). Observations using different in vitro models support either viewpoint. For example, A␤25-35, a synthetic peptide, will damage a variety of cells including tumor cells, fibroblasts, and astrocytes when applied in mol/liter concentrations (13,31,32). This nonspecific cytotoxic action of A␤25-35, however, is clearly different from the indirect neuron-killing mediated by A␤ activation of brain inflammatory cells described here. First, the immune-mediated process requires the presence of microglia, while the action of A␤25-35 is direct and independent of glia (13). Second, drug inhibition studies show the involvement of the neuronal NMDA receptors in microglia-derived toxicity (7,13), while free radical production appears necessary for the action A␤25-35 (33). Third, A␤1-42 will activate neurotoxic microglia at 10 nmol/liter or less ( Fig.  2A), while A␤25-35 kills cells at Ն30 mol/liter (13). And fourth, the microglia-derived toxicity is selective to neurons, while A␤25-35 acts indiscriminately upon a wide range of cell types. Since the validity of any disease model will be judged by its approximation of AD pathogenesis, it should be noted that A␤25-35 has not been found to exist as a natural product either in AD brain or in cultured cells (34 -36).
Immune suppression therapies have been suggested to limit neuronal damage brought about by microglia during stroke, trauma, and AD (7,21,(37)(38)(39). We believe that immune-mediated neuron killing, as mimicked by the culture system used here, has many features common to cellular events occurring in AD brain, including a specificity of neuron destruction, selective activation by neuritic/core plaques (versus diffuse plaques), and a high sensitivity to naturally occurring human A␤ peptides (versus rodent peptides). Importantly, this microglia-dependent toxicity model allows further opportunity to search out key steps in the immunopathology of AD and to identify immune suppression strategies applicable to the treatment of dementia. Data presented here form the basis for a novel treatment strategy to suppress microgliosis in AD. Small peptides with the HHQK domain compete with plaques for binding to microglia and in this way block plaque induction of neuronkilling behavior. Importantly, HHQK blockade of neuron-killing has specificity for A␤ and did not prevent microglial activation by other immuno-stimulants such as zymosan and lipopolysaccharide. Such a selective targeting of plaque-microglia interactions is desirable, for it would spare systemic re-sponses to other immune challenges. Agents that mimic the HHQK domain of ␤-amyloid plaques might block immunemediated pathology unique to Alzheimer's disease.