Comparative analysis of amyloid-beta chemical structure and amyloid plaque morphology of transgenic mouse and Alzheimer's disease brains.

We have undertaken an integrated chemical and morphological comparison of the amyloid-beta (Abeta) molecules and the amyloid plaques present in the brains of APP23 transgenic (tg) mice and human Alzheimer's disease (AD) patients. Despite an apparent overall structural resemblance to AD pathology, our detailed chemical analyses revealed that although the amyloid plaques characteristic of AD contain cores that are highly resistant to chemical and physical disruption, the tg mice produced amyloid cores that were completely soluble in buffers containing SDS. Abeta chemical alterations account for the extreme stability of AD plaque core amyloid. The corresponding lack of post-translational modifications such as N-terminal degradation, isomerization, racemization, pyroglutamyl formation, oxidation, and covalently linked dimers in tg mouse Abeta provides an explanation for the differences in solubility between human AD and the APP23 tg mouse plaques. We hypothesize either that insufficient time is available for Abeta structural modifications or that the complex species-specific environment of the human disease is not precisely replicated in the tg mice. The appraisal of therapeutic agents or protocols in these animal models must be judged in the context of the lack of complete equivalence between the transgenic mouse plaques and the human AD lesions.

Alzheimer's disease (AD) 1 is a progressive neurodegenerative disorder characterized by the presence of extracellular amyloid plaques composed principally of amyloid-␤ (A␤) surrounded by dystrophic neurites (1). This association and the realization that the basis of certain early-onset familial forms of AD seems to be the enhanced production of one or more A␤ peptides have led to the hypothesis that A␤ is intimately involved in the AD pathogenic process (2). A promising experimental approach to unraveling the role(s) of A␤ in AD pathology has been the construction and characterization of transgenic mice that overexpress the amyloid precursor protein (APP) (3)(4)(5)(6)(7)(8)(9)(10)(11)(12). Several transgenic mouse lines have been described that produce A␤ deposits that accumulate in an agedependent fashion and morphologically resemble the senile plaques characteristic of human AD (3,6,8,12,37,38).
The APP23 transgenic (tg) mice contain an APP751 cDNA with the Swedish familial AD mutation under the control of the neuron-specific Thy-1 promoter and express this human gene at levels 7-fold greater than endogenous murine APP (12). Longitudinal studies of these mice have revealed that extracellular amyloid deposits become evident as the APP23 tg mice age. These deposits exhibit, at their earliest appearance, the Congo red birefringence characteristic of the dense core plaques of human AD (12). A gradual progression from a diffuse deposit to a dense plaque is not a feature of the APP23 tg mouse pathology, paralleling our previous finding (13) that the diffuse amyloid deposits of AD do not represent a precursor developmental stage of senile plaques. A transgenic mouse model system that faithfully mimics every aspect of AD has not been developed. The APP23 tg mouse model reproduces some of the neuropathological changes associated with AD such as amyloid plaques with a core of amyloid, neuritic alterations, and neuron loss (15) as well as astrogliosis, microglial activation, and deposition of a cerebrovascular amyloid (14). Although an exact analog of AD is not at hand and may not ultimately be attainable, the understanding, management, and mitigation of AD may be facilitated through the use of the available animal models. Even though the exact role of senile plaques remains undefined, massive amyloid deposition is clearly an AD hallmark and is hypothesized to be seminal in the pathophysiology of this disease. Because there have been no detailed biochemical studies of the amyloid present in the APP23 tg mouse, we have undertaken an integrated chemical and morphological study to examine and compare the biochemical and biophysical properties of the transgenic mouse amyloid plaques with those of AD patients.

MATERIALS AND METHODS
Preparation of Formic Acid-extracted Brain Lysates-The generation of APP23 tg mice has been previously described (12). These mice express the human APP751 cDNA with the Swedish double mutation under the control of the neuron-specific mouse Thy-1 promoter fragment. The cerebral cortices, hippocampi, and olfactory bulbs, which contained the highest concentration of amyloid plaques in the transgenic (APP23) mice (12,15), representing ϳ165 mg of tissue/animal, were carefully dissected from five male and five female animals at room temperature. The average age of the animals was 22.5 months (range of 22.1-22.7 months). The tissue from each brain was finely minced and immediately homogenized in 3.0 ml of 90% (v/v) glass-distilled formic acid (GDFA) using a glass Dounce homogenizer at room temperature. Samples of 1.5 ml were loaded into 2-ml thick-walled polyallomer centrifuge tubes. The acid-insoluble material was separated by centrifugation in a Sorvall TST 60.4 rotor at 217,000 ϫ g for 1 h at 4°C. The clear supernatant was carefully collected, avoiding the surface layer of lipid and the small acid-insoluble material deposited in the bottom of the tube.
Collection of the Brain Lysate 3-8-kDa Mass Range Molecules-Samples (500 l) of the clear formic acid-extracted whole brain (brain tissue and its vascular network) lysates were fractionated by sizeexclusion fast protein liquid chromatography on a 1 ϫ 30-cm Superose 12 column (Amersham Pharmacia Biotech, Uppsala, Sweden) equilibrated and developed with 80% GDFA. The Superose 12 column fractions were calibrated using a set of proteins of known molecular mass and with the reverse-sequence A␤ peptide residues 40 to 1. The peptides were loaded into a 500-l injection loop and separated at a flow rate of 15 ml/h at room temperature, with the chromatographic eluate absorbance monitored at 280 nm. Fractions of 500 l were collected into polypropylene tubes every 2 min; those containing the 3-8-kDa molecules (corresponding to a retention time interval of 50 -66 min) were pooled into a single tube, and 20 l of 10% (w/v) betaine was added. The formic acid was removed by vacuum centrifugation (Savant Instruments, Inc., Farmingdale, NY).
Purification of Vascular Amyloid-To isolate the amyloid deposited in the walls of the cerebral blood vessels, the complete cerebral hemispheres from two male and two female APP23 tg mice were each sectioned into three coronal portions. The brain tissue was gently stirred for 24 h in 300 ml of 50 mM Tris-HCl (pH 7.5) containing 2% SDS and 2 mM EDTA at room temperature. The SDS-insoluble tufts of blood vessels consisting of the extracellular matrix with attached insoluble vascular amyloid were collected by filtration (20-m nylon mesh) and washed with Tris buffer. The insoluble vascular amyloid was solubilized and extracted with 3 ml of 80% GDFA and centrifuged at 217,000 ϫ g for 1 h at 4°C, and the supernatant was submitted in aliquots of 500 l to fast protein liquid chromatography as described above for the fractionation of brain lysate samples.
Purification of A␤ Peptides from the 3-8-kDa Range Pool-The A␤ peptides present in the 3-8-kDa size-exclusion fractions were separated by reverse-phase HPLC on a 4.6 ϫ 150-mm Source 5RPC column (Amersham Pharmacia Biotech) using a 500-l injection loop. Solvent A was 0.1% (v/v) 30% ammonium hydroxide in water adjusted with GDFA to pH 9.0, and solvent B was 60% (v/v) acetonitrile and 40% (v/v) solvent A. The chromatography was developed at room temperature with a linear gradient from 25 to 50% solvent B in 100 min at a flow rate of 1.0 ml/min, with absorbance monitored at 214 nm. The 3-8-kDa fractions obtained from four separate fast protein liquid chromatography runs of brain cleared lysate samples were pooled to provide the material for each reverse-phase HPLC separation. To render the peptides soluble, 20 l of 2 N NaOH was added to each of the polypropylene tubes. After pooling, an additional 20 l of 6 N NaOH was added, followed by 375 l of base-line solvent (25% solvent A and 75% solvent B) and a final addition of 25 l of 6 N NaOH. The A␤-(1-40) and A␤-(1-42) chromatographic retention times were established using the corresponding synthetic peptides obtained from California Peptide Inc. (Napa Valley, CA), which had been further purified in our laboratory by HPLC prior to their use as size markers.
Tryptic Hydrolysis and Separation of A␤ Peptides-The A␤ peptides separated by reverse-phase HPLC were lyophilized, solubilized in 1 ml of 80% (v/v) GDFA, and dialyzed (Spectrapor No. 6, M r 1000 cutoff) against two changes of distilled water and three changes of 100 mM ammonium bicarbonate. Once equilibrium was attained, the A␤ peptides were digested with L-1-tosylamido-2-phenylethyl chloromethyl ketone-treated trypsin (10 g/specimen; Worthington) for 16 h at 37°C. The resulting tryptic peptides were lyophilized, dissolved in 500 l of 0.1% (v/v) trifluoroacetic acid in water, thoroughly mixed and centrifuged at 12,000 ϫ g for 10 min to remove any insoluble material. The insoluble tryptic core, consisting mainly of the hydrophobic A␤ C-terminal peptide residues 29 -42, was dissolved in 300 l of 80% GDFA and, after the addition of a small crystal of CNBr, incubated for 16 h at room temperature. The soluble tryptic peptides and the CNBr-cleaved peptides were separated by HPLC at room temperature on a Spherisorb ODS2 C 18 reverse-phase column (4 ϫ 250 mm, 5-m beads; LKB, Bromma, Sweden). The chromatography was developed with a 90-min linear gradient of 0 -20% acetonitrile in 0.1% trifluoroacetic acid, followed by a second linear gradient of 20 -60% acetonitrile in 0.1% trifluoroacetic acid for 30 min using flow rates of 0.7 ml/min. Absorb-ance was monitored at 214 nm. Alternatively, the tryptic peptides were separated on a Source 5RPC column (4.6 ϫ 150 mm, 5-m beads) using a linear gradient of 0 -60% solvent B under the conditions described for the purification of A␤ peptides.
Automatic Amino Acid Analysis-The tryptic and CNBr-derived peptides were submitted to acid hydrolysis on a vapor-phase system Work Station (Waters) in 6 N HCl and 1% (w/v) phenol at 150°C for 100 min. After removal of the acid by vacuum centrifugation, the amino acid compositions of the peptide hydrolysates were determined at 570 and 440 nm using an automatic injector sampler-HPLC system (Thermo Separation Products, Fremont, CA) and a post-column ninhydrin reaction system (Pickering Laboratories, Inc., Mountain View, CA). The amino acid separations were performed on a sodium cation-exchange column (4 ϫ 150 mm, 5-m beads; Pickering Laboratories, Inc.) using the reagents and programs provided by the manufacturer.
Protein Sequencing-HPLC-purified samples were dissolved in 50 l of 50% (v/v) aqueous acetonitrile containing 0.1% trifluoroacetic acid. For the major HPLC peaks, ϳ30 l of the sample was dried on a fiberglass peptide disc (Beckman Coulter, Inc., Fullerton, CA) and used for amino acid sequence determination. Sequencing was performed on a Porton 2090E gas-phase protein sequencer (Beckman Coulter, Inc.) equipped with an on-line Hewlett-Packard 1090L HPLC apparatus. The remainder of the sample was used for mass spectrometry analyses.
Mass Spectrometry Analyses of the Trypsin-digested A␤ Peptides-Mass spectrometry samples were prepared by mixing 5 l of the dissolved HPLC-separated peptides with a saturated solution of ␣-cyano-4-hydroxycinnamic acid (also dissolved in 50% aqueous acetonitrile containing 0.1% trifluoroacetic acid), and aliquots of ϳ1 l were dried on the ends of stainless steel sample pins. For precise mass determinations, bovine insulin (mass of 5734.5 Da for the singly protonated molecular ion) was added to the mixture of sample and matrix and used as an internal standard. Mass spectra were obtained using a Vestec Lasertec Research mass spectrometer (PE Biosystems, Framingham, MA) operated in the positive ion mode with an accelerating voltage of 23 kV. This instrument incorporates a Laser Science VSL-337ND nitrogen laser that provides 3-ns pulses of 337-nm light of variable intensity. Data were collected using a TDS 520 digital storage oscilloscope (Tektronix, Beaverton, OR) and analyzed on a personal computer using LabCalc software (Galactic Industries Corp., Salem, NH). Each of the mass spectra reported represents the average of 128 shots.
Morphological Characterization of Amyloid Plaques-Cerebral hemispheres from transgenic and control mice as well as cerebral cortices from human AD brains were fixed for 48 h in 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4), dehydrated by passage through an alcohol and xylene solution series, and embedded in paraffin wax. Sections of 5 m were taken on a rotary microtome and stained with hematoxylin and eosin for general structure, thioflavin S for amyloid, and the Campbell-Switzer silver stain for diffuse amyloid plaques. Stained sections were viewed by bright-field light microscopy (hematoxylin-eosin; Campbell-Switzer-stained sections), fluorescence microscopy, and confocal scanning laser microscopy (thioflavin S-stained sections). Human and mouse sections were stained under identical conditions, in the same batch.

APP23 Transgenic Mice Produce More Soluble Amyloid
Plaques-One of the most striking characteristics of the amyloid cores present in AD patient brains is their extreme resistance to denaturing agents and detergents such as urea, guanidine hydrochloride, and SDS. During our initial attempt to enrich the amyloid cores from the APP23 tg mouse cerebral cortices using a 2% SDS solution, we noticed that in contrast to the human AD tissue samples, the parenchymal amyloid cores and fibrils completely disappeared. Extended high-speed centrifugation (250,000 ϫ g for 3 h) of the SDS supernatant failed to sediment any dispersed amyloid fibrils in the animal sample. These observations suggested that the plaque cores present in the APP23 transgenic animals were substantially more soluble than those in the AD brain. In contrast to the parenchymal deposits, the amyloid deposited around the blood vessels in the APP23 tg mice after SDS treatment remained intact and, as in the case of the AD brain, firmly attached to the vascular walls. We believe that the SDS insolubility characteristic of the vascular amyloid may in part be related to a stronger association between A␤ and the molecules of the extracellular matrix that required formic acid exposure for complete extraction.
Our experiments also revealed that a large proportion of the transgenic mouse parenchymal A␤ was freely soluble in detergent-free aqueous buffers. The APP23 tg mouse brains were homogenized at 4°C in an aqueous buffer (20 mM Tris-HCl (pH 7.4)) containing a mixture of protease inhibitors (16), followed by centrifugation (100,000 ϫ g at 1 h). The soluble A␤ present in the resulting supernatants was quantified by europium immunoassay (26). At 22 months of age, the total water-soluble A␤ and the total water-insoluble A␤ represented averages of 56.4 and 750 g/g of brain tissue, respectively.
Quantification and Chemical Characterization of the Total A␤ Present in APP23 Transgenic Mouse Brain Tissue-The increased water solubility of the APP23 tg mouse amyloid in comparison with its human AD counterpart suggested that fundamental structural and compositional differences exist between the amyloid plaques formed in transgenic mice and humans. Cortical areas, hippocampi, and olfactory bulbs were carefully dissected from the brain; and to account for the total amyloid present in the brain parenchyma and vascular walls, the tissue was acid-lysed and chromatographed in GDFA. Four discrete peaks were resolved on a size-exclusion column (Fig.  1), of which only the 3-8-kDa range molecules included A␤. These fractions were pooled; their volumes were reduced by vacuum centrifugation; and they were subjected to chromatography on a Source 5RPC reverse-phase column. A polystyrene/ divinylbenzene reverse-phase column was employed in these experiments because, in our experience, A␤-(1-42) exhibits better solubility at alkaline pH (9.0) than at an acid pH (2.0). Of the several chromatographic conditions investigated, a linear gradient of 25-50% acetonitrile gave the best resolution and eliminated most of the contaminating molecules either as unbound material or in the Ͼ50% acetonitrile eluate. A total of 10 discrete fractions were recovered (Fig. 2). The two major fractions had retention times identical to those of the synthetic A␤ markers A␤-(1-40) and A␤- . All the fractions were submitted to mass spectrometry (Table I) Both peptides also demonstrated the complete A␤ N-terminal amino acid sequence of DAEFRHD. The atomic masses of the molecules in the smaller peaks 2 and 5 corresponded to those of A␤-(1-37) and A␤- . We suggest that the A␤ peptides ending in residues 39, 38, and 37 in the tg mice represent soluble products of A␤ degradation. This suggestion is based on our recent analysis of the soluble A␤ peptides from the AD brain, which also show a series of shorter C-terminal peptides. 2 Alternatively, the A␤ peptides ending at residues 39, 38, and 37 could represent by-products of ␥-secretase-like activity. The loss of C-terminal residues greatly enhances the solubility of the A␤ peptides.
The minor fraction 4 was identified as A␤-(1-42) with an oxidized Met 35 . The smaller peaks 6 and 10 were designated as formylated forms of A␤-(1-38) and A␤-(1-42) since their molecular masses were those of the corresponding peptides increased by 28 mass units. This was also the situation with a smaller mass spectrometry signal observed within peak 9, which was identified as formylated A␤- . The minute amounts of formylated A␤ are probably artifactual modifications resulting from the use of formic acid during the extraction and chromatographic procedures. The minor peak 8 had the same atomic mass as peak 7 (A␤-(1-40)). Based on previous experience with chromatographic separations of A␤, we designated this fraction as A␤ with D-Asp either at position 7 or 23 (16,18), which imparts a slightly delayed retention time upon reverse-phase chromatography. Under the same chromatographic conditions, the AD A␤ peptide residues 17-42 or P3 fraction, a component of the diffuse amyloid deposits, has a retention time of ϳ70 min. No peaks were evident in this area in the case of the APP23 tg mouse amyloid.
Finally, the mass spectra of peak 9 also revealed, in addition to its major component (A␤- (1-42)), a small signal with an atomic mass of 3152. (calculated mass of 3151.7 Da). This suggestion is supported by the amino acid sequence data indicating a very small amount of protein, ϳ3-4% of the main peptide, starting with the sequence EVHHQ (A␤ residues [11][12][13][14][15]. The finding of this minor peptide is interesting since in vitro experiments have suggested that the ␤-secretase, besides generating A␤ peptides starting at Asp 1 , is also capable of cleaving between Tyr 10 and Glu 11 of the A␤ sequence (40,41). The presence of small quantities of A␤ starting at residue 11 in the brains of AD patients and in the APP23 tg mice also supports this contention (31,43,44). In addition, mass spectrometry analysis of non-transfected and ␤-APP-transfected mouse neuroblastoma cells has revealed the presence of A␤ peptides starting at residue 11 in the former, but not in the latter (42). Furthermore, ϳ40 A␤-related peptides were detected in the conditioned medium, probably resulting from the action of amino-, carboxy-, and endopeptidases (42). To detect any post-translational A␤ modifications that were not evident using mass spectrometry, such as isomerization of Asp at positions 1 and 7, known to occur in the human A␤ peptides (16), we digested the Source 5RPC chromatography peaks 7 (A␤-(1-40)) and 9 (A␤-(1-42)) with trypsin. The resulting peptides were separated either by C 18 reverse-phase or by 5RPC reverse-phase HPLC (Figs. 3 and 4) and acidhydrolyzed, and their amino acid sequences were determined. In contrast to the complex tryptic digest pattern obtained from the human AD A␤ peptide (which revealed up to 14 distinct peptides due to degradation of the N-terminal domain; the presence of L-iso-Asp, D-iso-Asp, L-Asp, and D-Asp at positions 1 and 7; and cyclization of Glu at position 3 (16,19)), the tg mouse A␤ tryptic digest map pattern was comparatively simple. The APP23 tg mice (peak 7) yielded only four tryptic digest peptides (Fig. 3), which corresponded to residues 1-5 (DAEFR), 6 -16 (HDSGYEVHHQK), 17-28 (LVFFAEDVGSNK), and 29 -40 (GAIIGLMVGGVV) of the A␤ peptide. The retention times of these tryptic peptides matched those of the tryptic peptides derived from synthetic A␤-(1-40) employed as calibration standards (Fig. 3). The C-terminal tryptic peptide of A␤-(1-42) produced in the tg mice, which precipitated as an insoluble core after tryptic digestion, was further cleaved by CNBr at Met 35 . This procedure generated two heptapeptides corresponding to A␤ residues 29 -35 (Gly-Ala-Ile-Ile-Gly-Leu-Hse) and residues 36 -42 (Val-Gly-Gly-Val-Val-Ile-Ala). These two peptides were separated by C 18 reverse-phase chromatography as previously described (Ref. 16 and data not shown). Chromatography of tryptic peptides at pH 9.0 yielded the expected pattern of four peptides (Fig. 4) and did not reveal the presence of any chemical modifications or N-terminal degradation.
Characterization of the SDS-insoluble Vascular Amyloid-Separation of the A␤ peptides present in the cerebrovascular walls by reverse-phase HPLC on a Source 5RPC column produced only three fractions corresponding to A␤-(1-40 ox), A␤-(1-40), and A␤- . Mass spectrometry revealed that the relative masses of these peptides were 4345.8, 4329.8, and 4517.0 Da, respectively, which is in agreement with their theoretical expected values.
Morphological Characteristics of APP23 Transgenic Mouse and Human Amyloid Plaques-As previously reported (12), senile plaques were present in the brains of the APP23 tg mice and were readily apparent with silver and thioflavin S stains in FIG. 2. 5RPC reverse-phase HPLC separation of amyloid peptides from APP23 tg mice. Subjecting the Superose 12 size-exclusion column eluate peak C (see Fig. 1) to reverse-phase chromatography resulted in 10 discrete factions (fractions 1-10). These fractions were submitted to mass spectrometry analysis and peptide sequencing. The identification of each fraction is shown in Table I. The lower chromatography trace is that of the calibration standards, synthetic A␤-(1-40) (peak A) and A␤-(1-42) (peak B), subjected to the same experimental conditions, which had identical retention times as fractions 7 and 9 from the tg mice. AUFS, absorbance units at full scale. mature animals (14-and 20-month-old animals). Most of the plaques were composed of compact amyloid, which fluoresced brightly with the thioflavin S stain (Fig. 5A), indicating the presence of a ␤-pleated sheet conformation. Old animals also displayed small numbers (Ͻ10%) of plaques (Fig. 5G, arrow), which have been termed "diffuse" deposits in humans (20 -22). The internal structure of the compact plaques was most apparent after thioflavin S staining, which showed that they were composed of bundles of filaments radiating outward from a central core (Fig. 5, A and C). These plaques ranged up to 200 m in diameter, averaging ϳ80 -120 m. In comparison, human AD compact plaques (Fig. 5, B and D) are classified as "burned-out" or "naked core" plaques and classical plaques (20 -22). Burned-out plaques consist simply of a central amyloid core ϳ10 -20 m in diameter (data not shown), whereas classical plaques have a central core surrounded by an empty region with a halo of fluorescent material, with overall diameters averaging ϳ60 -80 m. Human classical plaques, like the tg mouse plaques, possess wisps of fluorescent material radiating from a central core region (Fig. 5, C and D). In the tg mice, these wisps were clearly composed of filament bundles, whereas in the human, they had no discernible internal structure. The Campbell-Switzer stain revealed additional differences between the tg mouse and human AD plaques. Mouse compact plaques took up very little of the stain, appearing light brown in color (Fig. 5, E and G), unlike human compact plaques, in which the core and halo regions were both stained intensely black (Fig. 5F). Many mouse compact plaques had a delicate halo of black fibrillar material, like that seen in surrounding FIG. 3. C 18 reverse-phase HPLC profiles of tryptic peptides derived from the 5RPC reverse-phase HPLC peak 7 as shown in Fig. 2. Four A␤ tryptic peptides corresponding to residues 1-5 (peak 1), 6 -16 (peak 2), 17-28 (peak 3), and 29 -40 (peak 4) were analyzed. The lower chromatography trace is that of the four tryptic peptides obtained from tryptic digestion of synthetic A␤- . The retention times of peaks A-D obtained from synthetic A␤ coincide with those of peaks 1-4 obtained from tg mice. All tg mouse tryptic peptide peaks were acid-hydrolyzed, and their composition was determined by automatic amino acid analysis. AUFS, absorbance units at full scale.
FIG. 4. 5RPC reverse-phase HPLC analysis of tryptic peptides derived from the 5RPC reverse-phase HPLC peak 9 shown in Fig. 2. The peptides designated as A-D corresponded to A␤ peptides residues 1-5, 6 -16, 17-28, and 29 -42, respectively. The low yield of the C-terminal peptide is due to its hydrophobic nature and precipitation after tryptic digestion. Tryptic hydrolysis was also carried out on A␤ peptide residues 1-38 (peak 3 in Fig. 2). Separation of the resulting peptides using the 5RPC column revealed that the C-terminal tryptic peptides (residues 29 -38) had a retention time of ϳ50 min, which is indicated by the dashed line and designated as peptide E. AUFS, absorbance units at full scale. diffuse plaques (Fig. 5E, arrow), whereas others, however, did not (Fig. 5G, asterisk). Classical human AD plaques stained with thioflavin S possessed a relatively empty zone between the intensely stained core and halo (Fig. 5F), whereas no such empty region was observed in mouse plaques. Diffuse plaques, which were somewhat apparent with the thioflavin S stain, were, as with human diffuse plaques, more readily visualized with the Campbell-Switzer stain (Fig. 5G, arrow). These resembled human diffuse plaques (Fig. 5H, arrow) in their irregular shape and "cotton wool" appearance, although they lacked the fine granular material commonly observed in the diffuse plaques of the AD brain.

DISCUSSION
Amyloid plaque cores isolated from AD patients are entirely resistant to proteolytic degradation (23,24), and only a limited number of powerful chaotropic agents are capable of completely dissolving the human amyloid fibrils. Chemical characterization of the separated AD amyloid plaque components has revealed the presence of a complex mixture of A␤, glycoproteins, and glycolipids, of which the latter two represent ϳ20% of the total amyloid core mass (24,25). The amyloid core insolubility is partially due to the presence of these ancillary molecules (28). 3 In addition, the A␤ peptides isolated from the human brain have numerous post-translational modifications, are extensively degraded at the N terminus, and contain a high proportion of cross-linked A␤ molecules resulting from intermolecular linkages between tyrosyl residues. 4 Of the total amount of human A␤ solubilized from the amyloid cores by formic acid, monomeric, dimeric, and trimeric/tetrameric A␤ molecules represent an average of 55, 25, and 20%, of the total amyloid, respectively. All these factors that contribute to the AD amyloid insolubility and resistance to proteolytic degradation are apparently absent in the APP tg mice.
Several transgenic mouse models that reproduce certain aspects of AD have been engineered (3-12, 37, 38). The APP23 transgenic mice develop an intense amyloid deposition in the cerebral vasculature and produce Congo red/thioflavin S birefringent extracellular plaques with associated microglia in the brain parenchyma (14,27). Our experiments have revealed that the condensed amyloid plaque cores deposited in an agespecific pattern in the transgenic APP23 mice are not precise reproductions of those observed in AD, but differ in fundamental chemical and morphological aspects from the human senile plaques. The majority of the amyloid plaques in the APP23 tg mice consist of compact amyloid cores that morphologically resemble human plaque cores, but are much larger and lack the clear zone and halo present in the classical human plaques. Furthermore, in the APP23 mice, there are relatively few diffuse amyloid deposits, which are characteristically abundant in the AD brain.
The increased solubility of the A␤ peptides in the APP23 tg mice relative to that in AD brains is clearly reflected by the amounts of these peptides extracted by detergent-free aqueous buffer per unit of tissue. The averaged yield (n ϭ 6, mean age of 22 months) of total water-soluble A␤, representing both A␤-(1-40) and A␤- , in the tg mice was 56.4 g/g of brain.
In contrast, the average (n ϭ 8, mean age of 81 years) watersoluble total A␤ recovered from AD brains, using an identical extraction procedure, amounted to 48 ng/g of cortex (17). This is a Ͼ1000-fold increase in the amount of water-soluble A␤ per unit weight in the APP23 tg mice over that obtained from the AD brain.
Chemical analyses of the A␤ isolated from the APP23 tg mice demonstrated that most of the A␤ molecules initiated at Asp 1 in both A␤-(1-40) and A␤- . Only a minute proportion of A␤-  was detected in the tg mice. In the human, the N-terminal region of brain parenchymal A␤ is extensively degraded by aminopeptidases, ␤-secretase(s), or both. As reported by our laboratory and other investigators, in the human amyloid plaque cores, the A␤ molecules initiate at residues 2, 3, 4, 5, 6, 8, 9, 10, and 17 (13, 16, 18, 19, 29 -32). In AD, only a small proportion of A␤ starts at position 1 in the form of L-Asp in the amyloid plaques (ϳ10%), and this fraction is higher in the vascular amyloid (ϳ65%). In human amyloid plaques, ϳ50% of the amyloid starts at position 3 in the form of pyro Glu, whereas in the vascular amyloid, this form accounts for only 11% (19); no A␤ molecules with pyro Glu were detected in the APP23 tg mice. In the human A␤ isolated from neuritic plaques and vascular walls, the N-terminal iso-Asp accounts for ϳ20 and 6% of the total A␤, respectively, whereas none of this form was detected in the APP23 tg mice. With respect to iso-Asp at position 7, ϳ75% of the A␤ in the human amyloid plaques carries this isomerization. This ␤-shift is present in lesser quantities in the vascular amyloid (ϳ20%). Apparently, the APP23 tg mice are free from this ␤-shift between the peptide bond of Asp 7 -Ser 8 that drastically alters the conformation of the ␣-carbon backbone of the A␤ peptide. Finally, only a reduced proportion (ϳ10%) of the APP23 tg mouse A␤ appears to be oxidized at Met 35 . An important difference between the tg mice and human experimental subjects is the unavoidable longer post-mortem delay in sample processing for the latter that may account for some of the changes present in the amyloid chemistry between the two species. The presence and extent of potential time-dependent post-mortem artifacts affecting amyloid chemistry are currently under investigation.
It is likely that the structural alterations of the A␤ peptides in the human neuritic plaque cores account for their remarkable physical stability as well as for the inability of the phagocytic cells to remove them from the brain during the course of AD progression. We have previously demonstrated that Nterminal truncation of the A␤ molecules results in the net loss of polar residues, thereby increasing the overall hydrophobicity of the peptides and rendering them significantly more waterinsoluble and resistant to enzymatic degradation (33). This tenet is clearly illustrated by the shorter A␤-(17-42) peptide, which, even at low concentrations, avidly precipitates into amorphous aggregates and is abundant in the diffuse amyloid deposits present in the AD brain and Down's syndrome (13,36). It is also well established that isomerization of Asp residues results in a greater resistance to and even inhibition of proteolytic degradation in the A␤ peptides as well as in other molecules (33,34). Oxidation of Met 35 to its sulfoxide form has also been reported to increase the stability of the A␤ peptides (33). The generation of pyroglutamyl at position 3 of A␤ blocks N-terminal degradation of these peptides by aminopeptidases and increases the hydrophobicity of A␤ (32,33). Finally, the irreversible oligomerization of A␤ due to stable dimers and trimer/tetramers hinders enzymatic degradation (33). The average ratio among the four main A␤ fractions found in the APP23 tg mice, A␤-(1-40 ox), A␤-(1-38), A␤-(1-40), and A␤- , which in this study represent the contribution of both vascular and parenchymal A␤, was 12:9:50:29. These values contrast with the human neuritic plaque amyloid cores, in which A␤-(n-40) and A␤-(n-42) have an average ratio of 15:85. The relative ratio of the tg mouse vascular A␤-(1-40 ox), A␤-(1-40), and A␤-(1-42) peptides was 7:90:3. These data suggest that in the APP23 tg mice, the major contributor to the total amount of A␤-(1-40) is the vascular amyloid and that most of the brain's A␤-(1-42) originates from the cores of parenchymal amyloid.
The values for the human vascular amyloid are more variable since the A␤-(n-40) peptides have a range of 40 -90%. The larger the amount of vascular amyloid deposited in AD, the greater the amount of A␤-(n-40) peptide in the vessels, reaching a peak in those individuals with apoE-⑀4/⑀4 genotype (39). Furthermore, our studies have revealed two patterns of vascular amyloid deposition in the apoE-⑀4 AD cases. In the first, there is an overwhelming number of amyloid deposits affecting the form of spherical cores intimately attached to the basal lamina of the capillary network, producing an ornate pattern that resembles "pussy willows" (see Fig. 3 (C and D) in Ref. 29). In the second, there is an abundance of arteriolar and small artery (up to 500 m in diameter) amyloid deposits that resemble those observed in the APP23 tg mice. The dominance of A␤-(1-40) over A␤-  in this strain may reflect the heavy arterial amyloid deposition, an outstanding characteristic of the APP23 tg mice (14). 5 One important factor that may explain the almost complete absence of structural changes in the A␤ isolated from the APP23 tg mice is that the alterations commonly observed in the human brain are thought to be time-dependent. In the transgenic animals, amyloid is produced at levels that overwhelm the capacity of any endogenous clearance mechanisms and that lead to a vastly accelerated depositional process. In the human brain, the process of amyloid accumulation begins several decades prior to the onset of clinical symptoms and continues to occur until the death of the patient. The life span of the APP23 tg mice is just over 2 years, which may not be sufficient to manifest the full spectrum of structural A␤ changes observed in terminal AD (15). An additional possibility is that the structural modifications observed in AD do not represent merely a passive and random accumulation of time-dependent changes in A␤ structure, but are actually age-and perhaps speciesspecific (35), conceivably representing alterations in the expression and activity of key enzymes. It is possible that the processing required to create authentic AD plaques cannot occur in transgenic animals because either the necessary enzyme homologs are not present or the elevated pace of amyloid deposition simply precludes the prerequisite maturational reactions.
Morphological comparison of tg mouse and human senile plaques revealed some similarities and some differences. Both the mouse and human possess compact and diffuse plaques (20 -22). In AD, the diffuse amyloid exhibits a blend of short fibrils and aggregates of a fine granular material. 6 Biochemical analysis of these granular collections revealed the presence of the A␤ sequence of residues 17-42 (13), a peptide apparently absent in the tg mice. The mouse compact plaques display a fibrillar internal structure that is not visible in human compact plaques. These plaque structures are significantly larger in the tg mice, being on average about twice the diameter of human classical plaques.
Amyloid plaques may be characterized by either the presence or absence of "dystrophic" neurites. In this respect, it has been reported that compact plaques in these tg mice are associated with dystrophic neurites expressing phosphorylated tau epitopes (12), but without recreation of the paired helical filaments seen in human classical plaques. Another important difference between the plaques observed in the APP23 tg mice and the AD brain is that in the tg mice, the plaques are morphologically homogeneous, whereas in the human AD brains, a large variety of subtypes are typically observable. Other ␤-APP tg mouse strains have been constructed, including some in which there is an accelerated A␤ deposition resulting from the coexpression of mutant presenilin-1 and amyloid precursor proteins (37,38). The precise degree of similarity of the amyloid deposits in these animals to those in AD awaits experimental investigation. Our studies indicate that the amyloid fibrils deposited in the brain of the APP23 tg mice are chemically and morphologically distinct from the ones accumulated in the AD brain. In conclusion, the appraisal of therapeutic agents or protocols directed toward inhibiting or reversing plaque formation in these animals must be judged in the context of this lack of complete equivalence between the transgenic mouse plaques and the AD lesions.