Mass Spectrometry of Purified Amyloid f? Protein in Alzheimer's Disease*

The amyloid @-protein (GB) that is progressively de- posited in Alzheimer's disease (AD) arises from proteolysis of the integral membrane protein, &amyloid precursor protein (BAPP). Although AB formation ap-pears to play a seminal role in AD, only a few studies have examined the chemical structure of AB purified from brain, and there are discrepancies among the findings. We describe a new method for the rapid ex-traction and purification of AB that minimizes artifac- tual proteolysis. AB purified by two-dimensional re-verse-phase HPLC was analyzed by combined amino acid sequencing and mass spectrometry after digestion with a lysylendopeptidase. The major AB peptide in the cerebral cortex of all five AD brains examined was aspartic acid 1 to valine 40. A minor species beginning at glutamic acid 3 but blocked by conversion to pyroglutamate was also found in all cases. A species ending at threonine 43 was detected, varying from -5 to 25% of total AB COOH-terminal fragments. Peptides ending with valine 39, isoleucine 41, or alanine 42 were not detected, except for one brain with a minor peptide ending at valine 39.


I
The amyloid @-protein ( G B ) that is progressively deposited in Alzheimer's disease (AD) arises from proteolysis of the integral membrane protein, &amyloid precursor protein (BAPP). Although AB formation appears to play a seminal role in AD, only a few studies have examined the chemical structure of AB purified from brain, and there are discrepancies among the findings. We describe a new method for the rapid extraction and purification of AB that minimizes artifactual proteolysis. AB purified by two-dimensional reverse-phase HPLC was analyzed by combined amino acid sequencing and mass spectrometry after digestion with a lysylendopeptidase. The major AB peptide in the cerebral cortex of all five AD brains examined was aspartic acid 1 to valine 40. A minor species beginning a t glutamic acid 3 but blocked by conversion to pyroglutamate was also found in all cases. A species ending a t threonine 43 was detected, varying from -5 to 25% of total AB COOH-terminal fragments. Peptides ending with valine 39, isoleucine 41, or alanine 42 were not detected, except for one brain with a minor peptide ending at valine 39. Our findings suggest that AB"40 is the major species of &protein in AD cerebral cortex. AB"40 and A@"43 peptides could arise independently from BAPP, or AB1-43 could be the initial excised fragment, followed by digestion to yield AB1-' ' . These analyses of native AB in AD brain recommend the use of synthetic A@1-40 peptide to model amyloid fibrillogenesis and toxicity in uitro.
Alzheimer's disease (AD)' is the most common cause of progressive intellectual failure in aged humans. The filamentous lesions that define AD occur within neurons (neurofibrillary tangles), in extracellular cerebral deposits (amyloid plaques), and in meningeal and cerebral blood vessels (amyloid angiopathy) (1). A peptide with a molecular weight of -4,000, designated the amyloid @ (A@) protein, is the subunit of vascular and plaque amyloid filaments in individuals with AD (2-9), trisomy 21 (Down's syndrome) (10) hereditary * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The abbreviations used are: AD, Alzheimer's disease; A@, amyloid 0-protein; @APP, @-amyloid precursor protein; HCHWA, hereditary cerebral hemorrhage with amyloidosis; SDS, sodium dodecyl sulfate; HPLC, high performance liquid chromatography; API, Achromobacter protease I. cerebral hemorrhage with amyloidosis (HCHWA)-Dutch type (11,12), and normal brain aging (13). The cloning of cDNAs encoding the A@ protein demonstrated that it was a fragment of a large integral membrane polypeptide, the @-amyloid precursor protein (PAPP), the gene for which is located on the long arm of human chromosome 21 (14-17). Elucidation of the exon-intron structure of the @APP gene revealed that the A@ protein is derived from portions of two exons and must therefore arise by proteolysis of the precursor polypeptide (18).
Several lines of evidence indicate that progressive cerebral deposition of A@ plays a seminal role in the pathogenesis of Alzheimer's disease and can precede cognitive symptoms by years or decades (reviewed in Ref. 1). Importantly, missense mutations at residue 717 of the 770-amino acid isoform of BAPP have been found in affected members of at least nine families with autosomal dominant Alzheimer's disease (19)(20)(21)(22)(23). In addition, a mutation at residue 693 has been identified as the cause of HCHWA-Dutch (24). There is great interest in understanding the proteolytic processing of BAPP and identifying the proteases responsible for cleaving at the NHz and COOH termini of the AB region. However, only a few studies have examined the chemical structure of the native A@ protein purified from human brain, and there are discrepancies among the reported findings. Sequencing of A@ isolated from meningeal vascular amyloid has consistently yielded an aspartic acid (position 672 of @APP7.,J as the amino terminus (2,(7)(8)(9)(10)12). However, there has been disagreement about the sequence of compacted amyloid plaque cores. One laboratory reported ragged amino termini from cores isolated using a protocol that includes overnight digestion with pepsin, with the major species beginning with the phenylalanine at position 4 of AB (3). Another laboratory could not obtain NHz-terminal sequence from cores purified in a sodium dodecyl sulfate-containing buffer without protease treatment and postulated a blocked amino terminus in compacted plaque amyloid (4). Similarly, two other laboratories obtained no interpretable NHz-terminal sequences from plaque cores isolated by other methods (6,9). The precise carboxyl terminus of AB has likewise been in dispute. Two groups reported the valine at AB position 40 as the COOH terminus of meningovascular amyloid in Alzheimer's disease (7,9), and another laboratory identified valine 39 as the last residue of meningovascular amyloid in AD (8) and in HCHWA-Dutch (12). The laboratory that reported the NHz-terminal sequencing of amyloid plaque cores (3) identified alanine 42 or threonine 43 as the COOH terminus (15).
In view of the potential therapeutic importance of inhibiting the proteolytic cleavages that liberate the NH2 and COOH termini of A@ in Alzheimer's disease, the precise chemical nature of the native peptide must be established. We report a new method for the rapid isolation and purification of intact AB protein that minimizes the likelihood of artifactual proteolysis during the preparation. The purified A@ was analyzed by both amino acid sequencing and mass spectrometry, the latter method allowing precise determination of the lengths of the cerebral A@ proteins. Our results explain and resolve several previous discrepancies about the nature of A@ protein in Alzheimer's disease.

MATERIALS AND METHODS
Purification of A@ Protein from Cerebral Cortex-A@ protein was prepared from frozen (-80 "C) cerebral cortex of 15 histopathologically confirmed cases of AD. Mini-scale preparations (30-50 mg wet weight starting cortex) followed by immunoblotting with an antibody to AB""" (25) revealed that 5 of the 15 brains had high contents of AB, and these were used for the purification and quantitative analysis described here. Frozen cortex (1-5 g) was homogenized in a buffer containing 10% SDS, 150 mM NaCl, 50 mM Tris-HC1, pH 7.6, and the protease inhibitors leupeptin (1 pg/ml), pepstatin (0.1 pg/ml ), phenylmethylsulfonyl fluoride (0.5 mM), DFP (0.1 mM), and l-chloro-:Ltosylamido-7-amino-2-heptanone (I pglml). The homogenate was centrifuged a t 100,000 X g for 40 min (20 "C). The pellet was rehomogenized in the same buffer and sedimented again. This pellet was washed with 2% SDS, 150 mM NaC1,50 mM Tris-HC1, pH 7.6, and recentrifuged. The washed pellet was extracted in 100% formic acid (3) for 60 min (20 "C). After centrifugation at 100,000 X g for 20 min, the A@ protein was obtained in the formic acid-soluble fraction. This supernatant was chromatographed on a C4 reverse-phase HPLC column (Baker, Inc.) according to our previous method for purifying fragments of ubiquitin from Alzheimer paired helical filament preparations (26). The A@-immunoreactive peak was then rechromatographed on the same column but using a different elution gradient (see "Results"). The final purified A@ was confirmed as such by SDSpolyacrylamide gel electrophoresis (Coomassie Blue staining) and by immunoblotting with a rabbit polyclonal antibody raised to synthetic AB"".
Immunocytochemical staining with anti-A@'" was conducted on slices of cortex taken at autopsy directly from the fresh brain specimen before freezing the latter for the biochemical experiments. These slices, which were thus immediately adjacent to the tissue used to purify A@, were fixed briefly (3 h) in neutral buffered formalin, sectioned on a Vibratome, and immunostained conventionally (25).
Pyroglutamate Aminopeptidase Reaction-After mass spectrometric analysis (see "Results") of fraction B from the above C4 column, the nitrocellulose target was extracted 3 times with 50% isopropanol. The resultant extract was treated with pyroglutamylpeptidase (Lpyroglutamyl peptide hydrolase, EC 3.4.19.3) (0.4 milliunits, Nakarai Tesque Co.) in 10 mM sodium phosphate buffer, pH 7.8.1 mM EDTA and 10 mM dithiothreitol for 5 h (37 "C). The reaction was stopped by addition of formic acid, and the fraction was applied to a C4 reverse-phase HPLC column. A new fraction that eluted with a shorter retention time than the original fraction B was collected and subjected to amino-terminal sequencing.
Protein Sequencing-HPLC-purified A/3 protein and its API-generated fragments were sequenced on an Applied Biosystems 477A/ 120A protein sequenator. Some fragments were sequenced after analysis by mass spectrometry (see "Results").
Mass Spectrometry-The purified A@ protein or its API-generated fragments were applied to nitrocellulose targets, dried by flushing with nitrogen gas, and rinsed with 0.1% trifluoroacetic acid. Plasma desorption mass spectrometry was performed on a Bio Ion 20 Biopolymer Mass Analyzer (Applied Biosystems). Atomic masses were determined after calibration with hydrogen (H) and nitrous oxide (NO) in each spectrum. The theoretical molecular weights of the peptide fragments were calculated as their singly protonated molecular ions. An acceleration voltage of 15 kV was used. The channel resolution was 0.66 and 1.5 atomic mass units at m/z 1,000 and 5,000, respectively.

RESULTS
Mass Spectrometry of Purified AB Protein-AB protein was purified from AD cerebral cortex by two sequential reversephase HPLC runs (Fig. 1). The first run used a linear gradient of 0-80% isopropanol/acetonitrile (7:3) in 0.1% trifluoroacetic acid. The fraction containing AB protein (Fig. lA, arrow), as judged by Western blotting with anti-A@'42. was then rechromatographed using a gradient of 0-80% acetonitrile in 0.1% trifluoroacetic acid (Fig, 1B). The purified A@ protein from the second HPLC run was then analyzed by mass spectrometry as described under "Materials and Methods" (Fig. 2). An unusually broad peak was observed that distributed asymmetrically around a mass of 4356.7, suggesting that AP protein was not homogeneous. Protein sequencing of this fraction from the second HPLC run yielded Asp-Ala-Glu-Phe-Arg-His as the first 6 six amino acids. Taken together, these data suggest that the major molecular species of AP protein was a 40-residue peptide beginning with Asp-1 and ending with Val-40. We next attempted to obtain proteolytic fragments of AB protein in order to determine formally the precise chemical structure of its amino and carboxyl termini.
NH2-terminal Sequencing of Lysylendopeptidase-generated Fragments of AB Protein-Purified AB protein was digested as previously reported (26, 27) with the protease API, which specifically cleaves peptide bonds between lysine and the adjacent COOH-terminal amino acid. Since AB has 2 lysines in its sequence (positions 16 and 28), at least three fragments were expected B17-28, and Bzg-40 (or p29-42'43 if one assumes the chemical structure of AB is Asp-1 to Ala-42 or Thr-43; Ref. 15). The resultant fragments were separated by reversephase HPLC, as described under "Materials and Methods." As seen in Fig. 3, four peaks were in fact identified; these were designated fractions A, B, C, and D.
The NHz-terminal sequence of fraction A was shown to be Asp-Ala-Glu-Phe-Arg. Fraction B did not show any significant sequence signals, suggesting that this peptide contained an unknown NHz-terminal blocking group (see below). Fraction C began with the sequence Leu-Val-Phe-Phe-Ala-Glu-Asp, corresponding to the central portion of AB (B17-2s). Fraction D began with the sequence Gly-Ala-Ile-Ile-Gly, thus corresponding to the carboxyl-terminal fragment of AB.

Mass Spectrometry of the Lysylendopeptidase-generated
Fragments of AB Protein-Mass spectra of the lysylendopeptidase-derived A@ fragments (fractions A-D) are shown in Fig. 4. A mass of 1956.4 was observed for fraction A (Fig. 4A).
This value agreed well with the theoretical molecular weight of 1956 for the peptide Asp-1 to Lys-16. Taken together with the protein sequencing data for this fraction, fraction A was concluded to be the amino-terminal fragment (@'-I6) of the AD protein.
A mass of 1752.5 was observed for fraction B (Fig. 4B).
This value did not correspond to any of the masses predicted for lysylendopeptidase-generated fragments of the AB protein.
A computer search suggested that the peptide Glu-3 to Lys-16 might constitute fraction B, because its theoretical molecular weight, 1770, was within approximately 18 atomic mass units of the experimentally observed mass of 1752.5 We speculated that this mass difference might be the result of dehydration of glutamic acid to form pyroglutamate. Such a modification could also explain the failure of the NHz-terminal sequencing of this fragment. In an attempt to confirm the presence of pyroglutamate as the first amino acid of the  peptide in fraction B, the fraction was treated with pyroglutamylpeptidase followed by protein sequencing. As expected, Phe-Arg-His-Asp-Ser were identified as the second through sixth amino acids of this peptide. Hence, we conclude that fraction B was another amino-terminal fragment (P3-l6) of AB protein that was blocked at position 3, and that this fragment represented a minor species (15-20%) of the total AB NHZ-terminal fragments, as judged from the HPLC chromatogram (Fig. 3).

(E-F-R-H-D-S-G-Y-E-V-H-H-Q-K) is 1770; and of pyroglutamate @-16 (pE-F-R-H-D-S-G-Y-E-V-H-H-Q-K)
Fraction C produced the spectrum shown in Fig. 4C. The mass of 1326.3 corresponds well with that predicted for the peptide Leu-17 to Lys-28, 1326.5. The predominant peak at 1347.5 is consistent with the sodium-cationized molecular ion of this peptide. Taken together with the protein sequence data, fraction C was concluded to contain the middle portion (@'7"8) of A@ protein without any detectable modification. The mass spectrometric analysis of fraction D showed two families of peaks separated by -285 atomic mass units (Fig.  4 0 ) ; both had the same amino-terminal sequence (see above). The ions at masses 1085.9 and 1370.1 correspond closely to the theoretical masses for @z9"0 (1086.4) and @29-43 (1371.71, respectively. The theoretical mass difference between these, 285.3, agrees well with the difference observed experimentally (284.2). The signals of mass 1104.8 and 1123.7 were thought to represent essentially the same peptide as that of mass 1085.9, with the former likely to represent the sodium-cationized molecular ion of @29-40. The signal of mass 1390.7 is also consistent with the sodium-cationized form of the peptide of mass 1370.1. Fig. 4 0 shows the spectrum of the one case among the five Alzheimer cases we examined that had the highest peak for the Gly-29 to Thr-43 peptide. In all five brains, the latter peptide was consistently present, although the content of this peak varied from 5 to 25% of the total COOH-terminal fragments. Taken together with the protein sequence data for this fraction (above), the two molecules were concluded to be the A@ carboxyl-terminal fragments, @2g-40 and @29-43, respectively.
We failed to detect peptides with Ile-41 or Ala-42 as the last amino acids of A@ protein, although care was taken to search for ions of 1199.5 or 1270.6 that would correspond to peptides with the sequence Gly-29 to Ile-41 or Gly-29 to Ala-42, respectively.
Taking all the data into account, A@ protein was principally composed of the 40-residue sequence Asp-1 to Val-40, associated with a minor species of the peptide, pyroglutamate-3 (pGlu-3) to Val-40. The longest A@ protein species detected extended to Thr-43 at its carboxyl terminus.

DISCUSSION
The new method for purifying A@ protein from human brain reported here provides a faster, more efficient means of isolating intact A@ protein because it does not require an incubation time of 1-2 days for digestion with proteases like pepsin (3) or collagenase (2, 7, lo), nor the time for gel filtration prior to HPLC purification (2,lO). Homogenization of frozen tissue directly into a high concentration of SDS in the presence of multiple protease inhibitors was designed to minimize possible in vitro proteolysis by quickly inhibiting endogenous proteases. When 1 or 2% SDS was used in the homogenization buffer instead of 10% SDS, considerable lipid together with some residual protein and nucleic acid were present as contaminants in the amyloid fraction (data not shown). In order to decrease the likelihood of artifactual generation of ragged NH2 and COOH termini during purification, we avoided the use of proteases such as pepsin and collagenase, commercial preparations of which often contain additional proteolytic activities as trace contaminants. An additional advantage of our method is that it is applicable to tissues with a low content of A@ protein and can be readily modified for small to large scale preparations.
We applied this purification method to the cerebral cortex of five different brains that had neuropathologically typical AD. Immunocytochemical examination of briefly formalin-fixed slices taken directly from the cortical regions examined biochemically demonstrated the presence of many A@-immunoreactive plaques having both diffuse (28-30) and compacted morphologies. The number of intracortical microvessels bearing A@ deposits varied from a few to a moderate number per 50-pm vibrotome section among the five cases. Meninges were purposefully removed prior to homogenization to eliminate meningovascular A@ deposits that might be present.
The central finding emerging from our analyses is that the peptide Asp-1 to Val-40 is the major A@ species in the cerebral cortex of the AD cases we examined. A minor species beginning at residue Glu-3 but blocked by conversion to pyroglutamate and thus not amenable to sequencing was obtained from each of the 5 brains. Its relative amount varied within a narrow range of -15 to 20% of the total A@ NHz-terminal fragments. It is possible that this species arose artifactually during our extraction, but the immediate homogenization of frozen tissue in 10% SDS with multiple protease inhibitors and the fact that the major portion of cortical A@ could be readily sequenced from Asp-1 (as can meningovascular A@; Refs. 2, 7-10 and 12) make this unlikely. Assuming that it occurs in uiuo, the type of cortical deposit that contains this NH2-truncated, blocked species is not known. We speculate that this species may be present in the compacted, fibrillar cores of so-called "mature" senile plaques. Support for this speculation includes: ( a ) the finding by several laboratories (4,6,9) of an apparent blockage of the AB NH2 terminus in compacted amyloid cores isolated by different methods that all avoided the use of protease digestion; ( b ) reports by several investigators (3,4,6) of the much greater resistance of compacted cores than meningovascular amyloid to solubilization in reagents such as guanidine-HC1, suggesting greater chemical modification of the A@ in the cores; ( c ) the observation emerging from studies of the temporal progression of A@ deposits in Down's syndrome patients (see e.g. Ref. 31), normal aged humans, and lower primates' that diffuse, nonfilamentous deposits of AB precede the appearance of compacted, dense-core plaques, suggesting that the latter are more mature and may thus be modified chemically. However, these lines of evidence are indirect, and we plan to purify compacted plaque cores by a modification of the present method to determine whether the blocked AB species we have detected is enriched in such deposits. It should be emphasized that no difficulty in obtaining an NHz-terminal sequence beginning with residue Asp-1 has been found in the studies that have analyzed meningovascular AB (2,(7)(8)(9)(10)12).
Of the five AD brains examined here, we found a minor A@ species beginning with Ala-2 in only one case (data not shown). Thus, we did not obtain evidence of major NHzterminal heterogeneity of cortical AB, as reported in an earlier study (3). It is possible that the Ala-2 species derives from postmortem proteolysis by an exopeptidase. On the other hand, it is possible that some degree of NHz-terminal heterogeneity may occur in uiuo in amyloid deposits of varying age and location among AD subjects. There is evidence of allelic heterogeneity among the autosomal dominant AD families genetically linked to chromosome 21 that have been examined to date (19,22,23,32). In addition, many cases have no identifiable genetic basis, making it probable that environmental factors (for example head trauma; Ref. 33) could play a role in initiating some cases of widespread cerebral @amyloidosis. Although deposition of AB in plaque-like deposits is an invariant feature of AD, biochemical microheterogeneity in the composition of these deposits, including the presence of longer @APP fragments that contain the AB region, cannot * M. Podlisny and D. Selkoe, unpublished observations.

Mass Spectrometry of
Purified Amyloid p Protein in AD be ruled out and might be expected.
Regarding the carboxyl terminus of A@, we found the major species of cortical AB to end at residue Val-40 in all five cases examined. However, a minor species ending at Thr-43 was detected in each case, varying in amount from -5 to 25% of the total COOH-terminal fragments. We failed to detect peptides ending with Val-39, Ile-41, or Ala-42 in these five cases, except for one case that had a small amount of fragment after lysylendopeptidase digestion (data not shown).
Our current data are consistent with at least two possibilities: 1) that is the initial amyloid fragment excised from PAPP, and that many of these molecules are digested by carboxypeptidase(s) in the tissue to yield a major species of or 2) that and arise independently from precursor molecules of different cellular origin (e.g. vascular uersus neuropil) or in different anatomical sites (e.g. neurons in one cortical layer uersus another). Available data indicate that vascular deposits can contain (7,9), but these are unlikely to be the sole source of the abundant molecules in our extracts, since microvascular deposits were less abundant than diffuse and compacted plaques in the cases we studied.
The mechanisms by which the known mutations in human PAPP (PAPPss3 Glu + Gln; PAPPv17 Val + Ile, Val + Gly or Val + Phe) lead to enhanced cerebral and vascular deposition of AP are currently unclear. One report has suggested that synthetic A@ peptides containing the PAPPsg3 Gln substitution form fibrils more rapidly i n uitro than the wild-type peptide (34). The PAPP717 mutations occur carboxyl-terminal to the A@ peptides that we have detected to date in AD cerebral cortex. However, it may be that future studies will identify minor species of extracellular A@ peptides that extend beyond Thr-43, toward the carboxyl terminus of PAPP. This possibility raises the question of whether several distinct proteases may be involved in the progressive processing of amyloidogenic @APP fragments intracellularly and extracellularly prior to the occurrence of the apparently stable A@'-40 or peptides. The peptide composition of AP deposits in Alzheimer brain tissue must be understood in detail if drugs designed to prevent the formation of AP peptides are to be truly effective.