The biochemical analyses of Aβ peptides were performed on GM from temporal lobe including the entorhinal cortex (case 1) and on parietal and temporal cortex (case 2). The postmortem delay time before freezing of the tissue was approximately 4 hours for case 1 and 12 hours for case 2. Brain sections were subjected to Campbell-Switzer silver stain¹⁵ and 1.0% aqueous thioflavine-S for 10 minutes. To assess the distribution and relative number of senile and diffuse plaques in the cerebral cortex, brain sections were immunocytochemically stained. This step also permitted the estimation of the amount of tissue required to perform subsequent biochemical experiments. Briefly, brain sections were deparaffinized and treated with 80% formic acid for 20 minutes. Sections were incubated overnight with anti-Aβ antibodies 6E10 and 4G8 diluted 1:1000 (Signet Laboratories, Dedham, MA). Anti-mouse IgG was used as a secondary antibody (1:10,000), and then developed with diaminobenzidine in 0.5% nickel ammonium sulfate. Hematoxylin was used as a counterstain.

The water-soluble material represents soluble interstitial fluid Aβ, Aβ monomers and low molecular weight oligomers loosely attached to amyloid plaques, and soluble Aβ free in the cytosol.¹⁶ The leptomeninges covering the cerebral cortex were separated and stored at −80°C. Twenty grams of cerebral cortex was finely minced with a razor blade, suspended in 150 ml of ammonium bicarbonate buffer (0.1 mol/L NH₄HCO₃, 5 mmol/L ethylenediamine tetraacetic acid, and 0.01% sodium azide), with three tablets of complete enzyme inhibitors (Roche, Mannheim, Germany), and allowed to stand at 4°C for 48 hours. The brain cortex was homogenized with a 40-ml capacity Tenbroeck glass homogenizer (six manual gentle strokes) and centrifuged at 150,000 × g for 3 hours (Beckman SW28 rotor) in a Beckman Optima LE-80K ultracentrifuge (Beckman, Fullerton, CA). The water-soluble fraction was lyophilized and analyzed by fast protein liquid chromatography (FPLC) (Amersham Biosciences, Piscataway, NJ) and high performance liquid chromatography (HPLC) (Thermo Separation Products, Schaumburg, IL).

The water-insoluble pellets were dispersed and stirred for 12 hours in 50 ml of 0.1 mol/L NH₄HCO₃. Fifty milliliters of 30% SDS was added (final SDS concentration 15%) and stirred for an additional 14 hours at room temperature. To remove the SDS-insoluble tufts of parenchymal blood vessels, the suspension was filtered through a series of 300-, 45-, and 25-μm nylon meshes. The SDS-containing filtrate was centrifuged at 300,000 × g for 5 hours at 20°C in an SW41 Ti Beckman rotor. The resulting pellets were suspended in 60 ml of 0.1 mol/L ammonium bicarbonate, pH 10.0 (titrated with NH₄OH), and after 24 hours of continuous stirring, centrifuged at 300,000 × g for 5 hours at 20°C. The pellets were stored, and the supernatant was lyophilized and analyzed by FPLC and HPLC.

The leptomeninges were washed 10 times with 500 ml of cold water to remove blood cells by hypotonic shock. Vessels larger than 600 μm in diameter were eliminated, and the smaller vessels were finely minced and washed three times (5 minutes at 1500 × g) with 0.1 mol/L Tris-HCl and 3 mmol/L CaCl_2, pH 8.0. Vessels were suspended in the wash buffer amended with 0.3 mg/ml Clostridium histolyticum collagenase CL3 (Worthington Biochemicals Corp., Lakewood, NJ), and 10 μg/ml DNase I (Sigma-Aldrich, St. Louis, MO) was added to create a final suspension of 100 ml, which was incubated in a shaker at 37°C for 3 hours. The preparation was then filtered through a 150-μm nylon mesh, and the filtrate was centrifuged at 3000 × g for 10 minutes. The supernatant was eliminated, and the pellet was washed with water, suspended in water, and centrifuged at 3000 × g. The pellet, containing sheets of leptomeningeal vascular amyloid, was dissolved in 4 ml of glass distilled formic acid (GDFA) with the aid of a glass homogenizer, left at room temperature for 30 minutes, and centrifuged at 500,000 × g for 10 minutes in a TL 100 rotor (Beckman Optima TLX ultracentrifuge). The supernatant was lyophilized and analyzed by FPLC and HPLC.

The 300-, 45-, and 25-μm nylon mesh-retained vessel tufts from the cortical SDS-insoluble amyloid preparation were immediately shaken in 500 ml of distilled water to suspend the retained vessels. The material was allowed to settle, and the clarified liquid was decanted with the remaining volume centrifuged at 1500 × g for 10 minutes. The pellets containing the cortical blood vessels were washed twice with distilled water to remove remaining SDS and were suspended in 0.1 mol/L Tris-HCl and 3 mmol/L CaCl₂, pH 8.0. Next, 0.3 mg/ml collagenase CL3 (Worthington) and 10 μg/ml DNase I (Sigma) were added, and samples were incubated in a shaker at 37°C for 3 hours. The preparation was filtered through a 40-μm nylon mesh, and the filtrate fraction was centrifuged at 3000 × g. The pellet was treated with 4 ml of GDFA, spun at 500,000 × g as described above, and analyzed by FPLC and HPLC.

This fraction represents the most insoluble of the amyloid aggregates and is composed primarily of Aβ 17–42 peptides recovered after high-speed centrifugation of the SDS/alkaline insoluble pellets. This material was dissolved in 80% GDFA with the aid of a glass homogenizer, allowed to stand at room temperature for 1 hour, parceled into 0.7-ml aliquots in thick-walled polyallomer tubes, and centrifuged at 500,000 × g for 15 minutes in a Beckman Optima TLA-ultracentrifuge using a 120.2 rotor. The dark brown pellets containing formic acid-insoluble material mostly composed of lipofuscin remnants were discarded, and the clear supernatant was submitted to FPLC-GDFA chromatography. Column eluate samples containing amyloid fraction P3 were pooled, reduced in volume by Speed-Vac centrifugation, and subjected to size-exclusion HPLC chromatography on a GPC-100 column, equilibrated, and developed with 80% GDFA. The GPC-100 column fraction containing the P3 was refined further by chromatography on a reverse-phase Zorbax SB-C8 column as described below.

The WM was separated from the adjacent GM to avoid contamination from cortical amyloid. The WM (3 g) was finely minced and thoroughly homogenized in 18 ml of 80% GDFA. The homogenate was centrifuged at 250,000 × g for 1 hour using 11-ml polyallomer tubes in a Beckman SW41 rotor. The fatty material collected at the top of the tube and the minute bottom pellets were discarded. The transparent supernatant was brought to 20 ml by the addition of GDFA and centrifuged under the same conditions. The clear supernatant was submitted to FPLC, and further purification of Aβ peptides was achieved by reverse-phase chromatography on a C8 column Zorbax RX-C8 (Mac-Mod, Chadds Ford, PA) (see below).

The lyophilized water-soluble fraction, the SDS-insoluble fraction, leptomeningeal-vascular amyloid and cortico-vascular Aβ, the P3 fraction, and the WM Aβ were all independently submitted to size-exclusion chromatography using a Superose 12 column (1 × 30 cm; General Electric, Uppsala, Sweden) equilibrated with 80% GDFA. In the case of the water-soluble and SDS-insoluble fractions, aliquots of 40 mg each of the lyophilized material were dissolved in 500 μl of 80% GDFA and chromatographed at 15 ml/h with column output monitored by UV absorbance at 280 nm. The 2- to 8-kd fraction was collected, acid was removed, and volume was reduced to 50 μl by Speed-Vac centrifugation and stored in a ultra-low freezer (−80°C). In some instances to further refine the separation of the Aβ peptides from higher and lower molecular mass flanking contaminants, the FPLC Superose 12 chromatographic step was repeated twice.

The FPLC fractions containing 2- to 8-kd molecules were further purified by reverse-phase HPLC. Five FPLC runs were pooled and concentrated to 50 μl, 200 μl of 90% GDFA was added, and the final volume was adjusted to 500 μl with distilled water. The specimens were loaded onto a reverse-phase Zorbax SB-C8 column (4.6 × 250 mm; Mac-Mod) equilibrated with 20% acetonitrile/water and 0.1% trifluoroacetic acid (TFA). Separations were performed using a linear (20 to 40%) acetonitrile gradient developed for a period of 60 minutes at 80°C at 1 ml/minute flow rate. The column eluate was monitored at 214-nm UV absorbance. Alternatively, the Aβ peptides were further purified in a semipreparative reverse-phase column Zorbax RX-C8 (9.4 × 250 mm; Mac-Mod). The chromatography was developed with a linear gradient formed by 20% acetonitrile/water, 0.1% TFA and 60% acetonitrile/water, 0.1% TFA, at 1.5 ml/minute at room temperature. The chromatography was monitored at 214 nm. The FPLC-enriched diffuse plaque Aβ-related peptides were purified by size-exclusion chromatography on a silica-based Syncropack GPC-100 column (5-μm particle; 100-Å pore; 500 × 7.8 mm; Synchrom, Inc., Lafayette, IN), equilibrated, and developed at room temperature with 80% GDFA at a flow rate of 0.5 ml/minute. The chromatography was monitored at 280 nm. All HPLC columns were calibrated to establish the retention times for Aβ tetramer, Aβ trimer, Aβ dimer, Aβ monomer, and Aβ-P3 (residues 17–42 species). The Aβ monomers were synthesized by California Peptide Research, Inc. (Napa, CA) and further purified in our laboratory by size-exclusion FPLC (Superdex 75) and reverse-phase HPLC (C8). The Aβ oligomers were generated by dialyzing synthetic monomer Aβ 1–42, dissolved in 80% GDFA (3 mg/ml SpectraPor 6, 1000 d cut-off) against 4 L of 50 mmol/L Tris-HCl buffer, pH 7.5, followed by incubation at 37°C for a period of 8 weeks. Before the opening of the dialysis bag, the Aβ was dialyzed for 24 hours against 0.1 mol/L ammonium bicarbonate followed by lyophilization.¹⁷

The HPLC-purified Aβ peptides were dissolved in 50 μl of 50% (v/v) acetonitrile and 0.1% TFA. Approximately 30 μl of the sample was dried on a fiberglass disk (Beckman Coulter, Fullerton, CA). Sequencing was performed on a Porton 2090E gas-phase protein sequencer (Beckman Coulter) attached to an on-line Hewlett-Packard 1090L HPLC.

All preparations were performed at 4°C. One hundred milligrams of tissue was homogenized in 800 μl of 5 mol/L guanidine HCl and 50 mmol/L Tris-HCl, pH 8.0, with an electric grinder. The homogenates were shaken for 3 hours and then centrifuged in a 50.4-Ti rotor (Beckman) for 30 minutes at 60,000 × g. The supernatants were collected and stored at −80°C. Tissue extractions were also performed in 50 mmol/L Tris-HCl, pH 8.0, following the above protocol. The ELISA Aβ 40 (Immuno-Biological Laboratories, Minneapolis, MN) and Aβ 42 (Innogenetics, Belgium, Germany) kits were used according to the manufacturers’ instructions.

Matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF) was performed as described in detail by Esh et al.¹⁸ The complete mass spectrometry data are presented as Supplemental Tables 1 to 4 (http://ajp.amjpathol.org).

It is evident that the amyloid deposits in the cerebral cortex of AN-1792-vaccinated human cases focally disappeared, suggesting that polyclonal anti-Aβ antibodies crossed the blood-brain barrier. These antibodies were induced using synthetic fibrillar Aβ 1–42 as an antigen, and the resulting antibodies apparently disassembled fibrillar amyloid plaques. In case 1, there were patchy areas in which the diffuse and compact amyloid deposits were totally absent, although abundant accumulations of amyloid were present in the cortical vessels (Figure 1, C, D, E, and H). In this study, the term “diffuse amyloid deposits” describes plaques that are a conglomerate of nonfibrillar Aβ peptides organized into amorphous aggregates. They stain faintly with thioflavine-S (Figure 1J) or thioflavine-T, are negative for Congo red birefringence, and are very visible and stain intensely with silver dyes such as Campbell-Switzer (Figure 1, A and B). In other areas of the cortex, there were focally decreased densities of compact core plaques as revealed by histochemical stains and the 6E10 antibody probing, with remaining diffuse amyloid deposits clearly demonstrated by the 4G8 antibody (Figure 1, F and G), respectively. NFT density remained high in regions with focally depleted plaques. Earlier neuropathological reports of NFTs in AN-1792 vaccinated cases^13,14 found no apparent changes in their abundance and distribution. Previous biochemical studies have revealed that the diffuse deposits of Aβ in sporadic AD are mainly composed of a mixture of Aβ 17–42 and Aβ 1–42.¹⁹ In some areas, the remaining amyloid plaques exhibited a peculiar morphology, consisting of a clearly defined halo of amyloid-positive material and diminutive centrally situated punctiform core. Between these two structures, there was a clear area devoid of Aβ immunoreactivity (Figure 1I). Concerning case 2, histological sections of the frontal lobe stained with Campbell-Switzer and thioflavine-S revealed a complete absence of amyloid core plaques and diffuse plaques (Figure 2, A and B) in agreement with previously published data.¹⁴ However, although the parietal and temporal cortex demonstrated sparse core amyloid plaques, diffuse plaques were abundant (Figure 2, C and D). Histological sections of case 1 showed an extensive deposition of Aβ in the cortical vessels and in the leptomeningeal vasculature (Figure 1, D and H; Figure 2, E and F). However, they were less pronounced in case 2.

Before the general analytical preparations to separate the soluble and insoluble fractions, the amount of Aβ peptides present in the GM and WM of case 1 and case 2 and those in sporadic AD and age-matched non-demented (ND) subjects were quantified by ELISA (Table 1). Homogenization of the brain tissue with 5 mol/L guanidine hydrochloride (GHCl) yielded 868 and 186 μg/g total Aβ for cases 1 and 2, respectively. In comparison, the average amount of Aβ for neuropathologically confirmed AD cases (n = 31) was 406 μg/g tissue and for age-matched ND controls (n = 22) was 221 μg/g tissue. Interestingly, the water-soluble Aβ obtained after a gentle homogenization of finely minced cerebral cortex in aqueous buffer was 61 μg/g for case 1 and 13.5 μg/g for case 2. These values were in sharp contrast with those obtained from AD (n = 7) and ND (n = 4), which both amounted to 3.9 μg/g tissue. In both vaccinated cases, the levels of Aβ in the WM, extracted by 5 mol/L GHCl, were 96.8 μg/g for case 1 and 16.9 μg/g for case 2. In the GM and WM of case 1, Aβ n-40 and n-42 were equally represented in the GHCl extract. In case 2, the Aβ n-40-to-n-42 ratio differed, being 2.8:1 in the GM and 1.8:1 in the WM. Most of the water-soluble Aβ corresponded to the n-40 isoform in both of the cases.

We developed techniques that permitted the independent separation and purification of the Aβ peptides present in the walls of leptomeningeal vessels and the cortico-vascular vessels. This result was achieved by the combined use of collagenase digestion of the extracellular matrix and GDFA extraction/solubilization of Aβ. Both tissue fractions were independently submitted to FPLC for Aβ enrichment (Figure 3, A and B). On subsequent HPLC, the leptomeningeal amyloid produced three amyloid-containing fractions, whereas the parenchymo-vascular amyloid produced seven fractions as depicted in Figure 4, A and B. Because we obtained a limited amount of purified vascular Aβ, a decision was made to investigate these fractions by mass spectrometry to maximize the amount of information at the peptide structural level. A large number of peptides were detected by MALDI-TOF. In the case of the leptomeningeal vessels, 21 Aβ potential peptides were observed that started at positions 1 through 5 and 7 through 10 and ending at positions 37 through 42 (Supplemental Table 1). In the case of the cortico-vascular vessels, a total of 26 likely Aβ-related peptides were identified with N termini at positions 1 through 5 and 7 to 9 and 11 and with C termini at positions 32 and 37 through 42 (Supplemental Table 1). In some instances, these peptides were modified by formylation at hydroxyl groups or by modification of Met to Met-sulfone or Met-sulfoxide. Peptides with N-terminal pyroglutamyl were also observed at positions Glu3 or Glu11 of the Aβ sequence. In sporadic AD, the leptomeningeal amyloid is mainly composed of Aβ 1–40 with a lesser, variable amount of Aβ 1–42. Characteristically in sporadic AD, the leptomeningeal amyloid exhibits very limited N-terminal degradations and posttranslational modifications that, in contrast, are commonly observed in the GM amyloid plaques.^20–22 These limited modifications are probably due to the constraints imposed by the biochemical constitution of the vascular environment, in which endothelial cells, pericytes, and myocytes predominate. However, in the vaccinated individuals, the composition of the leptomeningeal and cortico-vascular Aβ is more reminiscent of that of the typical AD amyloid plaques with more N-terminally degraded Aβ peptides possessing shorter C termini. These observations suggest a mobilization from the soluble cortical pool to the vascular deposits.^12,13 In support of this contention are the recent observations of Boche et al,²³ who reported a larger number of cortical vessels containing Aβ n-40 and Aβ n-42 in AN-1792-immunized individuals. The elevated Aβ n-42 probably originated from the senile plaques, where it represents the major Aβ component. The apparent increase in amyloid angiopathy was also accompanied by a larger number of microhemorrhages than those observed in nonimmunized AD controls.²³ The composition of the Aβ present in the walls of the cortico-vascular amyloid in sporadic AD is intermediate between the plaque core and leptomeningeal amyloid, with a lesser degree of N-terminal degradations and posttranslational modifications than those observed in the plaque core amyloid.²⁰

The 2- to 8-kd water-soluble fractions separated by FPLC/GDFA (Figure 3C) submitted to reverse-phase chromatography (Zorbax RX-C8 column) yielded six fractions containing Aβ peptides (Figure 5A). Western blots demonstrated the presence of more dimers than monomers. Scanning densitometry showed a total ratio of Aβ 40 monomers to Aβ 40 dimers of 2:98, whereas the ratio of Aβ 42 monomers to Aβ 42 dimers was 17:83 (Figure 5A). MALDI-TOF suggested the presence of 87 Aβ-related peptides. Because our analyses account for peptides present in the soluble brain homogenate, they are also likely to detect peptides derived from Aβ degradation by amino-, carboxy-, and endo-peptidases, including those carrying posttranslationally modified amino acids by isomerization and racemization as well. Amino acid modifications due to naturally occurring or artifactual formylation, oxidation, and cyclization were also detected in some Aβ peptides. We identified Aβ-related peptides with N termini commencing from residues 1 through 27 (with the exclusion of those starting at residues 10 and 25). Most of these peptides (59%) had their N termini at residues 1 through 9 and 11, indicating β-secretase digestion and probably amino-peptidase degradation. The C termini, on the other hand, included peptides ending at residues 14 through 43 (with the exception of those ending at positions 16, 19, and 32). However, the majority of these peptides (53%) had their C termini at positions 36 through 43, suggesting processing by the γ-secretase. Three Aβ-related peptides: 17–36, 17–41, and 17–43, may have resulted from α- and γ-secretase hydrolysis (Supplemental Table 2).

In both Aβ-vaccinated cases under study, there was a remarkable reduction in the number of SDS-insoluble amyloid cores, which are abundant in the SDS-insoluble fraction obtained from the cerebral cortex in the vast majority of neuropathologically severe sporadic AD cases.²⁴ However, in addition to the scarce amyloid cores, this fraction must have contained a substantial quantity of fibrillar water-insoluble Aβ, which originated from disassembled cortical amyloid cores that were solubilized by the alkaline treatment and remained in the supernatant after ultracentrifugation. After FPLC separation (Figure 3D), these Aβ peptides were submitted to reverse-phase HPLC chromatography (Figure 5B). Western blots indicated the presence of Aβ in 9 of the 12 fractions recovered with a preponderance of dimers over monomers. The ratio of Aβ 40 monomers to Aβ 40 dimers was 0:100 and that of Aβ 42 monomers to Aβ 42 dimers was 13:87 (Figure 5B). However, the more insoluble Aβ n-42 fraction revealed on the Western blot a small amount of trimers, tetramers, and larger oligomers up to ~30 kd (Figure 5B). The ratios of Aβ monomers to Aβ dimers in the water-soluble fraction and SDS-insoluble fraction were very similar probably because they are, to some extent, derived from the same pool of insoluble plaque fibrillar amyloid. Mass spectrometry (MALDI-TOF) analysis of 13 chromatographic fractions suggested the presence of 100 Aβ-related peptides that could originate from Aβ itself, APP, and its secretase by-products (Supplemental Table 3). The chromatographic fractions contained Aβ peptides with N termini from position 1 through 23 (except for those starting at positions 9 and 21). The C termini encompassed peptides ending from positions 17 through 43 (excluding those ending at positions 18, 20, and 24). These peptides were present in their pristine condition and in their modified forms by formylation and oxidation. N-terminal pyroglutamyl was detected at positions 3 and 11. As in the case of the SDS-insoluble (fibrillar Aβ) of sporadic AD,²¹ the SDS-insoluble fraction of the vaccinated cases under study also contained abundant Aβ peptides with posttranslational modifications (isomerization and racemization) and extensively degraded N termini.

This material represents the most insoluble of all Aβ brain parenchymal peptides and is recovered by high-speed centrifugation as the insoluble pellet that remains after water, SDS, and alkaline extractions. Staining by thioflavine-S revealed homogeneous amorphous conglomerates that resemble a flock of cotton wool (Figure 2G). The GDFA-solubilized P3 fraction Aβ peptides were submitted to size-exclusion chromatography on a GPC-100 column, which yielded a single large peak and with lesser minor peaks (Figure 4C). The larger peak was subsequently resolved by reverse-phase chromatography (Zorbax SB-C8) into six well-separated peaks (Figure 4D). On automatic amino acid sequencing, all fractions, with the exception of peak 4, yielded the peptide Leu-Val-Phe corresponding to residues 17-18-19 of Aβ. Further automatic Edman degradation analyses performed on peak 1 revealed the amino acid sequence of Leu-Val-Phe-Phe-Ala-Glu-Asp-Val-Gly (Aβ residues 17 through 25). The molecular weight of 2409.4 d on the mass spectrometry (MS) reflectron mode confirmed the identity of these peptides as Aβ 17–42, with an oxidized Met 35. Interestingly, peaks 4 and 6 also contained peptides with a molecular mass of 991.73, which may correspond to the Aβ peptide residues 3 to 10 with an N-terminal pyroglutamyl residue. In addition, peak 5 also showed a peptide with a mass of 3437.68 d, which may correspond to the Aβ peptide amino acid residues 11 to 43. The different retention times may represent isoforms containing iso-Asp at position 23 and possibly 27 that do not change the molecular mass because they are the results of β-shifts and/or also due to racemization of Asp and Ser residues as previously observed on reverse-phase chromatography.^21,25,26

White matter lesions are present in two-thirds of patients with AD. They represent severe alterations in protein, lipid, and cholesterol metabolism; dysmyelination; demyelination; loss of axons and astrogliosis compounded by a reduced number of microvessels; concurrent ischemia/hypoxia; and interstitial fluid retention with dilation of periarterial spaces.^49,64 In elderly ND individuals, the WM contains a small pool of soluble Aβ that probably originates in the GM and gradually diffuses into the WM.⁶⁴ In the absence of visible WM Aβ deposits, such as those observed in the cerebral cortex of AD patients, it can be assumed that the WM Aβ peptides are either soluble oligomers or associated with structural molecules such as myelin sheets that may inhibit Aβ fibrillization. In addition, an amplified Aβ-mediated WM neuroinflammatory response would increase the complement-mediated production of membrane attack complexes that will further destroy myelin and axonal integrity.⁶⁵ Experimentally, Aβ peptides induce severe oligodendrocyte damage by interfering with the sphingomyelinase-ceramide pathway⁶⁶ as well as alterations in axonal cytoskeleton manifested by loss of neurofilament immunoreactivity.⁶⁷ Both guanidine and formic acid extractions recovered a substantial amount of WM Aβ n-40 and Aβ n-42. In sporadic AD WM, the total average soluble Aβ is 2.2 μg/g tissue⁶⁴ whereas the WM of immunized cases 1 and 2 contained 97 and 17 μg/g tissue of soluble total Aβ, representing 44 and 8 times more Aβ peptide load, respectively. The higher amount of total WM Aβ (~ 5.8-fold) of case 1 over case 2 may be due to the postvaccination aseptic encephalitis that gravely affected the WM or to differences in prevaccination Aβ levels. The substantial increase in WM amyloid burden suggests that this compartment acted as the ultimate “sink” for Aβ released from disrupted senile plaques that could not escape through the vasculature. The ultimate consequences of a shifted amyloid tissue burden are unknown.