Analysis of heterogeneous A4 peptides in human cerebrospinal fluid and blood by a newly developed sensitive Western blot assay.

The betaA4 peptide, a major component of senile plaques in Alzheimer's disease (AD) brain, has been found in cerebrospinal fluid (CSF) and blood of both AD patients and normal subjects. Although betaA4 1-40 is the major form produced by cell metabolism and found in CSF, recent observations suggest that the long-tailed betaA4 1-42 plays a more crucial role in AD pathogenesis. Here, we established new monoclonal antibodies against the C-terminal end of betaA4 1-40 and 1-42, and used them for the specific Western blot detection. After optimizing the assay conditions, these antibodies detected low picogram amount of betaA4, and both betaA4 1-40 and 1-42 levels in CSF could be determined by direct loading of the samples. Blood levels of betaA4 1-40 and 1-42 were also determined by specific immunoprecipitation followed by Western blot detection. We found that CSF betaA4 1-42 level is lower in AD patients compared with non-demented controls, although there was a significant overlap between the groups. The level of betaA4 1-40 in CSF, and of betaA4 1-40 as well as betaA4 1-42 in plasma, were not different between AD patients and controls. Besides the 4-kDa full-length betaA4 band, we could also detect several N-terminal variants of betaA4 in CSF and plasma of both AD patients and controls. Two N-terminally truncated betaA4 species migrating at the position of 3.3 and 3.7 kDa were found in CSF, while 3.7- and 5-kDa forms were found in plasma. The relative abundance of these various species were considerably different in the CSF and plasma, suggesting that the cellular source and/or clearance of betaA4 is different in these two compartments.

A major neuropathological feature of Alzheimer's disease (AD) 1 is the presence of senile plaques in the brain. These extracellular deposits of fibrillar aggregates are mostly composed of a 4-kDa peptide called ␤A4 or ␤-amyloid (1,2). ␤A4 is a proteolytic fragment derived from larger protein precursor called amyloid precursor protein (APP) (3). Although ␤A4 exists as an aggregated, poorly soluble form in brain deposits, it is secreted from cells by normal metabolism as a soluble molecule and this soluble ␤A4 is also detected in cerebrospinal fluid (CSF) and blood of both AD patients and healthy controls (4 -6).
On the other hand, several recent biochemical and immunohistochemical analyses indicated that plaques and brain homogenates of AD and Down's syndrome patients contain considerable amounts of N-terminally truncated ␤A4 (12, 13, 19 -23). These truncated forms of ␤A4 are especially abundant in diffuse plaques, which are assumed to represent a premature form of plaques, suggesting an important role of these species in the initial phase of AD pathogenesis. N-terminal heterogeneity was also detected in ␤A4 isolated from cell culture conditioned medium (24) and pooled CSF (5,9), although their levels in individual CSF samples has not been reported. So far, there is no information about N-terminal variants of ␤A4 in blood, and it is not known whether the N-terminal raggedness of soluble ␤A4 is altered during preclinical or clinical AD.
Total ␤A4 levels in CSF have been examined by several groups. One study showed increased level of ␤A4 in early-onset AD (25), while others showed no significant difference between AD and controls (26 -28). Recently, an ELISA specific for ␤A4 1-42 was established and applied to CSF measurement, showing the somewhat surprising results that the CSF ␤A4 1-42 level is lower in AD cases (29). Taken together, the contribution of soluble ␤A4 occurring in body fluids to AD pathogenesis remains obscure.
So far, ␤A4 levels in body fluid samples were mostly measured by sandwich ELISA method. Although ELISA is a sensitive and simple method suitable for routine clinical use, one drawback in ␤A4 measurement is that it can be affected by the presence of other proteins which may cross-react or mask the epitope of antibodies used in the assay. In this report, we established a sensitive Western blot assay to specifically measure various ␤A4 variants. By using newly developed monoclonal antibodies that recognize the C-terminal end of ␤A4 1-40 (C40) and 1-42 (C42) for detection, we could determine ␤A4 1-40 and 1-42 levels in CSF by direct loading of samples. This assay was also applied to analyze ␤A4 1-40 and 1-42 levels in plasma, as well as N-terminally truncated variants of ␤A4 in CSF and plasma of both AD patients and control individuals.

EXPERIMENTAL PROCEDURES
Establishment of Monoclonal Antibodies-␤A4 C-terminal specific monoclonal antibodies (mAbs) were generated as follows. Synthetic peptides corresponding to ␤A4 33-40 (GLMVGGVV) or ␤A4 35-42 (MVGGVVIA) were conjugated with keyhole limpet hemocyanin or bovine serum albumin (BSA) through a cysteine residue added at the N terminus of the peptides. BALB/c mice were immunized with 50 -100 g of the peptides three to four times with about 2-week intervals. Three days after the final boost, spleen cells were isolated and fused with SP2/0 myeloma cells using polyethylene glycol 1500. Fused cells were cultured in HAT (hypoxanthine/aminopterin/thymidine) selection medium supplemented with 15% fetal calf serum, growth promoting reagent (HFCS, Boehringer Mannheim) and 4 ng/ml of human interleukin-6. Antibody-producing hybridoma cells were screened by solid phase ELISA, and positive cells were cloned twice by limited dilution. Two clones designated G2-10 and G2-11 were selected as C40-and C42-specific mAbs, respectively. mAbs against ␤A4 N-terminal region were obtained from mice immunized with full-length ␤A4 1-40 or 1-42 with the same procedure, and one clone designated W0-2 was selected for further use. Isotypes of mAbs were determined by using an ELISA kit (Bio-Rad).
Western Blot Assay for the Determination of ␤A4 1-40 and ␤A4 1-42 Levels in CSF-For the determination of ␤A4 1-40 levels, 8 l of CSF sample was mixed with 4 l of 3 ϫ sample loading buffer (6% SDS, 15% 2-mercaptoethanol, 30% glycerol, and 0.3 mg/ml bromphenol blue in 188 mM Tris-HCl, pH 6.8), heated at 90°C for 10 min, and separated by 16% Tris-Tricine SDS-PAGE. Separated proteins in the gels were electrophoretically transferred onto nitrocellulose membrane at 380 mA for 45 min (30). The blotted membrane was heated in boiling phosphate buffered saline (PBS; 8.1 mM disodium hydrogen phosphate, 1.5 mM potassium dihydrogen phosphate, 137 mM sodium chloride, and 2.7 mM potassium chloride, pH 7.4) for 5 min to enhance the signal (31), and blocked with 5% skim milk in PBS containing 0.05% Tween 20 (PBS-T buffer) for 15-30 min. After washing the membrane with PBS-T, G2-10 antibody (3 g/ml), diluted in PBS-T containing 0.25% BSA, was added and incubated for overnight at 4°C. The bound antibodies were detected by horseradish peroxidase-conjugated anti-mouse Ig secondary antibody (Amersham Corp.) followed by ECL detection system (Amersham) according to the manufacturer's instruction. For the detection of ␤A4 1-42, 100 l of CSF sample was vacuum-dried (Speed Vac), dissolved in 13 l of 3ϫ sample loading buffer, heated at 90°C for 10 min, and loaded onto a 16% Tris-Tricine SDS-PAGE gel. Western blot detection was performed by the same procedure employed for the ␤A4 1-40 detection, except that the membrane was blocked with 0.25% BSA instead of skim milk and that G2-11 antibody (6 g/ml in PBS containing 0.25% BSA and 0.25% Tween 20) instead of G2-10 was used as primary antibody with 2 h incubation at room temperature. For the detection of total ␤A4, W0-2 antibody (1 g/ml) was used as primary antibody for staining. In some experiments, 10 -20% gradient Tris-Tricine gel (Novex) was also used. Band density was quantitated by densitometric analysis using Mac Bas (Fuji) program. In each case, known amount of synthetic ␤A4 1-40 or 1-42 was loaded onto the same gel and measured in parallel to draw calibration curves for quantitation. The amount of the standard ␤A4 was determined by amino acid analysis.
Determination of ␤A4 1-40 and 1-42 Levels in Plasma-Plasma ␤A4 was analyzed by two-step immunoprecipitation: 1) purification of total ␤A4 from plasma by W0-2-coupled gel beads and 2) specific precipitation with G2-10 (for ␤A4 1-40 detection) or G2-11 (for ␤A4 1-42 detection), followed by Western blot detection. First, 700 l of plasma sample was mixed with 70 l of 10ϫ immunoprecipitation buffer (250 mM Tris-HCl, pH 8.0, containing 5% Triton X-100 and 5% Nonidet P-40) and centrifuged at 13,000 rpm for 3 min. W0-2 antibody covalently coupled to Affi-Gel 10 (Bio-Rad) (11 g of antibody coupled to 7.5 l of gel) was added and incubated at 4°C for 20 h with continuous rocking of the tubes. After centrifugation and washing the pellet three times with PBS-T, bound proteins were eluted by the addition of 300 l of 50 mM glycine HCl buffer, pH 2.4, containing 0.1% BSA. The sample was centrifuged, and the supernatant was transferred to fresh tube containing 30 l of 1 M Tris-HCl, pH 8.0, and 0.5% Tween 20. G2-10 antibody (4 g) and 20 l of protein G-agarose beads (Boehringer Mannheim) was added and incubated for 4 h at room temperature. Beads were collected by centrifugation and washed twice with PBS-T. Immunoprecipitated protein was solubilized from the beads by adding 60 l of 3ϫ sample loading buffer and heating at 90°C for 10 min. A 30-l sample was loaded onto 10% Tris-Tricine SDS-PAGE, and ␤A4 band was detected by Western blot using W0-2 as primary antibody. For ␤A4 1-42 detection, the supernatant after G2-10 immunoprecipitation was used for the assay. First, 20 l of protein G-agarose (without antibody) was added and incubated for 2 h at room temperature to remove any remaining G2-10 antibody, centrifuged, and then G2-11 antibody (8 g) and 20 l of protein G-agarose were added to the supernatant. Tubes were incubated at 4°C overnight, centrifuged, and precipitated proteins were eluted by 30 l of 3ϫ sample loading buffer and analyzed as described above. Known amount of synthetic ␤A4 1-40 and 1-42 peptides, diluted in 1 ml of 0.1 M Tris-HCl, pH 7.4, containing 0.25% BSA and 0.05% Tween 20, were analyzed in parallel and used as standards to calculate ␤A4 concentrations.
Detection of N-terminally Truncated Forms of ␤A4 from CSF-CSF samples (1 ml) mixed with 200 l of Tris-HCl, pH 7.4, containing 0.25% BSA and 0.05% Tween 20 were incubated with W0-2 (4 g) and protein G-agarose (20 l) at 4°C overnight. After centrifugation, supernatant was transferred to fresh tube and again incubated with W0-2 (4 g) and protein G-agarose (20 l) at room temperature for 4 h, followed by supernatant transfer and incubation with protein G-agarose (20 l) alone at room temperature for 2 h. During these three cycles of immunoprecipitation, most of the full-length ␤A4 was removed. Then, G2-10 antibody (4 g) and protein G-agarose (10 l) were added to the supernatant and incubated at 4°C overnight. Beads were collected by centrifugation, washed twice with PBS-T, and bound protein was solubilized by incubation with 15 l of 3 ϫ sample loading buffer at 90°C for 10 min. Samples were separated by 16% Tris-Tricine SDS-PAGE and detected by Western blotting using G2-10 (3 g/ml) for detection.
Detection of N-terminally Truncated Forms of ␤A4 from Plasma-Plasma samples (1 ml) were mixed with 100 l of 10 ϫ immunoprecipitation buffer and centrifuged at 13,000 rpm for 3 min. G2-10 coupled Affi-Gel 10 (7.5 g of antibody coupled to 5 l of gel) was added to the supernatant and incubated at 4°C overnight. Gels were collected by centrifugation, washed three times with PBS-T, and bound protein was eluted by incubating with 15 ml of 3ϫ sample loading buffer at 90°C for 10 min. Samples were separated by 16% Tris-Tricine SDS-PAGE and detected by Western blotting using G2-10 (3 g/ml) for detection.
Subjects-CSF samples were collected from 59 subjects, and they were separated into three groups as follows; AD patients (n ϭ 39), demented non-AD patients (n ϭ 14), and non-demented patients with other neurological diseases (n ϭ 11). Clinical diagnosis of AD was based on NINCDS/ADRDA criteria (32). Among AD cases, 11 were also affected with cerebrovascular diseases and therefore should be classified as possible AD. Demented non-AD cases include vascular dementia (n ϭ 10) and Pick's disease (n ϭ 4). Non-demented cases include major depression (n ϭ 6), schizophrenia (n ϭ 2), mania (n ϭ 1), personality disorder (n ϭ 1), and Parkinson's disease (n ϭ 1). Samples were obtained by lumber puncture and stored frozen until use.
Plasma samples (heparinized plasma) were collected from a different group of patients. Total number of samples measured was 63, consisting of early-onset AD (onset age Ͻ 65) (n ϭ 20), late-onset AD (onset-age Ն 65) (n ϭ 12) and healthy controls (n ϭ 31). Summarized profiles of patients are listed in Table I (CSF samples) and Table II (plasma  samples).

Establishment and Characterization of Monoclonal Antibod-
ies against ␤A4 -In order to obtain mAbs which are specific for the different C termini of short-tailed and long-tailed ␤A4, we immunized two mice with synthetic ␤A4 33-40 peptide and four mice with synthetic ␤A4 35-42 peptide, both conjugated to carrier protein. After ELISA screening and cloning of positive wells, we established 10 different clones (seven IgG class and three IgM class) specific for the C terminus of ␤A4 1-42 (C42) and five clones (four IgG class and one IgM class) specific for the C terminus of ␤A4 1-40 (C40). Unexpectedly, three out of five clones specific for C40 were obtained from mice immunized with 35-42 peptide. We focused on IgG class mAbs and evaluated their affinity and specificity. One of each MAbs from C40-specific clones and C42-specific clones was selected for further use and designated G2-10 (class IgG2b,) and G2-11 (class IgG1,), respectively. We also obtained eight mAbs that recognize both ␤A4 1-40 and 1-42 (all IgG class) from mice immunized with ␤A4 1-40 or 1-42 (two mice each). Epitopes of these eight clones were all mapped to the residues 1-16 of ␤A4. All of these mAbs also reacted with secreted APP cleaved at the ␣-secretase site (APP␣sec) (34). The clone designated W0-2 (class IgG2a,), which showed the highest affinity, was selected for further analysis. Epitope of W0-2 was further examined using shorter peptides and the major recognition site was mapped to residues 5-8 of ␤A4. The specificities of the selected three antibodies were evaluated by solid phase ELISA and are demonstrated in Fig. 1. W0-2 reacted with ␤A4 1-40, 1-42, and 1-43 with nearly same affinity, as expected from its epitope. G2-10 was found to be completely specific for 1-40 in this assay. G2-11 did not cross-react with 1-40, but showed 1-2% of cross-reactivity with 1-43 peptide.
The three mAbs were next evaluated for their use in Western blot detection of ␤A4 (Fig. 2). After optimizing the assay conditions such as blotting time, membrane heat treatment, blocking reagent, antibody concentrations, and incubation times, G2-10 could detect 3 pg (0.7 fmol) of ␤A4 1-40 without any detectable cross-reactivity with 1-42. Using G2-11, we could visualize 12 pg (3 fmol) of ␤A4 1-42 without cross-reactivity with 1-40. W0-2, which recognizes both 1-40 and 1-42, turned out to be superior over the other mAbs in terms of the sensitivity and detected 0.4 pg (0.1 fmol) of ␤A4 1-40 and 1.6 pg (0.4 fmol) of ␤A4 1-42. The staining for 1-40 peptide was always about four times higher compared with 1-42, when the same amounts of peptide (determined by amino acid analysis) were loaded on the gels.
Analysis of ␤A4 in CSF-Direct loading of 8 l of CSF onto Tris-Tricine SDS-PAGE gel followed by Western blot detection with G2-10 or W0-2 mAb gave one clearly detectable ␤A4 band (Fig. 3). APP␣sec was also detected by W0-2. When G2-11 was used, we could not see staining of a 4-kDa band of ␤A4 from direct loading of 8-l samples (data not shown). However, when 100 l of CSF was concentrated by vacuum-drying (Speed Vac) and loaded onto the gel, a ␤A4 1-42 band became visible (Fig.  3). We analyzed 59 CSF samples (34 AD patients, 14 demented non-AD patients, and 11 non-demented controls) and measured the levels of both ␤A4 1-40 and 1-42. ␤A4 concentrations were determined by densitometric analysis of the ␤A4 bands in comparison with the known amount of synthetic ␤A4 peptides. As shown in Fig. 4 and Table I, there were no significant differences in ␤A4 1-40 levels between AD patients, demented non-AD patients, and non-demented control groups. For ␤A4 1-42 levels, control (non-demented patients) showed slightly higher levels, and the differences between the control group and the AD group, and between the control group and the demented non-AD group, were both statistically significant (p Ͻ 0.05, t test). However, there was no difference between the AD and the demented non-AD group, and considerable overlap was seen between control and demented groups. There was no obvious correlation between ␤A4 levels and disease severity (mini-mental state examination score) or age (data not shown).
Analysis of ␤A4 in Plasma-The Western blot assay was next applied to blood specimens. We found that plasma ␤A4 was not detectable by direct loading of samples, because of the low content of ␤A4 and of high total protein concentration that limit the possible loading volume onto the gels. Therefore, we used immunoprecipitation treatment before Western blot analysis. First, total ␤A4 was purified by immunoprecipitation with W0-2-coupled gel followed by acid elution. Then ␤A4 1-40 and 1-42 were precipitated with G2-10 and G2-11 antibody, respectively, and stained with W0-2 in Western blotting. Representative pictures of the exposed films are shown in Fig. 5. We applied the assay to 63 plasma samples (20 early-onset AD, 12 late-onset AD, and 31 normal controls) and determined the levels of ␤A4 1-40 and 1-42. Results are shown in Fig. 6 and Table II. In this study, we did not find significant differences between each groups for both ␤A4 1-40 and 1-42 levels. When analyzing ␤A4 1-40, besides full-length ␤A4 1-40, one additional minor band at 5 kDa was detected that was not observed in CSF samples. From longer exposure of the films, this 5-kDa band was detected from all of the samples, and the amount of the 5-kDa band seemed to be proportional to the 4-kDa ␤A4 band.
Detection of N-terminally Truncated Forms of ␤A4 -We next tried to detect N-terminally truncated forms of ␤A4 in CSF and plasma samples in order to find out whether there is any difference in their levels between AD patient and controls. From direct loading of CSF samples, we could not detect any band which migrate below full-length ␤A4 (Fig. 3). Immunoprecipitation with G2-10 followed by the detection with the same antibody in Western blotting was also not successful, because the strong signal of full-length ␤A4 1-40 on the exposed film interfered with the detection of the closely migrating N-terminally truncated ␤A4 species. Therefore, we used W0-2 for immunodepletion of the full-length ␤A4 prior to the immunoprecipitation and Western blot detection with G2-10. From 1 ml of CSF, two bands below the full-length ␤A4 were visualized in this assay (Fig. 7). From longer exposure of the film, these    ␤A4 in CSF and Blood two bands were detected in all of the nine samples analyzed. The molecular weight of the upper and the lower bands were estimated to be 3.7 and 3.3 kDa, respectively. The 3.3-kDa band migrated slower than "p3" band derived from conditioned medium of SY5Y cells transfected with SPA4CT fragment (100 amino acid peptide of APP C-terminal plus signal peptide) (35). Because the "p3" peptide comprises residues 17-40 of ␤A4, it suggests that both the 3.7-and 3.3-kDa species are N-terminally extended from Leu-17. Although the amount of these truncated ␤A4 differed considerably between each samples, these peptides were detected both in AD and control CSF samples, and their amount did not correlate with the AD diagnosis. When G2-11 instead of G2-10 was used to precipitate N-terminally truncated ␤A4 species ending at Ala-42 (C42), no band below 4 kDa was detected (data not shown). In contrast to the results of CSF, some plasma samples were shown to contain N-terminally truncated ␤A4 with comparable amounts to full-length ␤A4. When 700 l of plasma (12 samples, 6 AD, and 6 controls) were analyzed by G2-10-coupled gel immunoprecipitation (without immunodepletion of full-length ␤A4) and Western blot detection, the 3.7-kDa band was found in five samples (three AD and two controls) (Fig. 8). The amount of this band varied considerably between each sample and did not correlate with the amount of full-length ␤A4 1-40, which is in contrast to the 5-kDa band in Fig. 5, the intensity of which was proportional to that of the 4-kDa ␤A4 band. There was no obvious correlation between the amount of the 3.7-kDa band and the diagnosis. DISCUSSION Although amyloid deposition in the form of senile plaques is a central feature of AD pathology, the source of ␤A4 in the deposits is not known. Since the amyloid exists in extracellular spaces which are in direct contact with CSF, one possible model for plaque formation is that increased production (or decreased clearance) of ␤A4 in AD causes accumulation of soluble ␤A4 in CSF, which then is converted to aggregated forms or directly deposits onto a nucleating aggregate of ␤A4. There are also observations that suggest the possibility of a hematogenic origin of ␤A4 in brain deposits (15,36). In order to address these issues, it is essential to analyze ␤A4 levels in body fluids of individual patients. Especially, when considering the recent observations that suggest a critical role of long-tailed ␤A4 1-42 in plaque formation, it is becoming more desirable to measure the levels of ␤A4 1-40 and 1-42 separately. For this purpose, we, first, developed several mAbs which are specific for the C termini of ␤A4 1-40 (C40) and ␤A4 1-42(C42). Unexpectedly, three of the five clones specific for C40, including G2-10, were obtained from the mice immunized with 35-42 peptides. These clones were obtained from two mice, both of which also produced C42 specific clones. Although we do not have an explanation for this unexpected outcome, we assume that there is a possibility that the 35-42 peptide was processed in vivo in those mice, and the newly generated peptides with C40 end may have served as immunogen.
mAbs with the highest affinities were selected and applied for Western blot detection of ␤A4. In order to increase the sensitivity of our detection system, we optimized the assay conditions such as membrane blotting time, blocking reagents, antibody concentrations, and staining times. Blocking reagent was one of the critical factors for high sensitivity. For example, we found that for the staining with G2-11, BSA was better compared with the commonly used skim milk, which caused much higher backgrounds in this assay (data not shown). We also found that the heat treatment of the nitrocellulose membrane after the blotting is essential for high sensitivity. For heating the membranes, we used boiling PBS instead of water, which is used in the initial report of this technique (31), since the signal enhancement was much more pronounced in PBS compared to water (data not shown). Although the mechanism of this signal enhancement is not clear, we found that this heat treatment is effective for several other antigen-antibody combinations (data not shown), indicating the general usefulness of this method for sensitive Western blot analysis. After these assay conditions were optimized, both C40-specific mAb (G2-10) and C42 specific mAb (G2-11) could detect ␤A4 peptides in the low picogram range. The sensitivity obtained with W0-2 (0.4 pg for ␤A4 1-40, 1.6 pg for 1-42) was especially remarkable and is more than 1000 times higher compared with the previous reports on ␤A4 detection by Western blot (6,25). The reason why the ␤A4 1-40 synthetic peptide always gives a stronger signal compared with the same amount of ␤A4 1-42 is not clear. One possible explanation is that 1-40 and 1-42 peptides have different conformations and the binding affini- FIG. 7. Detection of N-terminally truncated forms of ␤A4 from CSF samples. CSF samples from eight patients (four AD patients and four non-demented subjects) were analyzed according to the procedure described under "Experimental Procedures" (lanes 2-9). 4-kDa band represents full-length ␤A4 1-40, which remained after immunodepletion step. Conditioned medium from SPA4CT (signal peptide plus Cterminal 100 amino acids of APP) transfected SY5Y cells was analyzed in parallel (lane 10). In this sample, the full-length ␤A4 band migrated at a slightly higher position due to the additional 2 amino acids at the N-terminal end of ␤A4 derived from the signal peptide (35). Lane 1 is the synthetic ␤A4 1-40 peptide (19 pg) directly loaded onto the gel. ties to the antibody is not the same. However, we think that it is more plausible that 1-40 and 1-42 peptides bind to nitrocellulose membranes with different efficiency, since this difference in signal intensity between 1-40 and 1-42 was also seen by other antibodies directed against N-terminal part of ␤A4 (data not shown) and since W0-2 reacted with 1-40 and 1-42 peptides with the same efficiency by the ELISA evaluation (Fig. 1).
By using this Western blot assay, we could directly quantitate both ␤A4 1-40 and 1-42 in CSF samples. We found that there was no need for an immunoprecipitation step. We argue that this direct detection has the advantage to measure total ␤A4 levels, since ␤A4 in CSF is reported to bind to carrier proteins such as apolipoprotein J (apoJ) (37,38), apoE (39), and transthyretin (40), and if antibody epitopes of ␤A4 are masked by such binding proteins, it cannot be detected by sandwich ELISA or Western blot analysis which depends on a prior immunoprecipitation step.
Our results showed that there is no difference in the ␤A4 1-40 levels among AD patients, demented non-AD patients, and non-demented cases. In contrast, we found that some of the non-demented CSF samples contain higher levels of ␤A4 1-42 compared with demented cases. The mean values of CSF ␤A4 1-42 in non-demented cases (0.501 ng/ml) was 1.8 times higher compared with AD cases (mean ϭ 0.277 ng/ml) or non-AD demented cases (0.282 ng/ml). This result is consistent with the recent ELISA measurement by Motter et al. (29), which showed similar 1.7 times increase when non-demented controls (mean ϭ 0.632 ng/ml) were compared with AD patients (mean ϭ 0.383 ng/ml). In our assay, however, more considerable overlapping of the values between control and AD samples was observed, and we found no difference between AD and non-AD demented groups, suggesting that the measurement of ␤A4 1-42 levels in CSF may not be useful for diagnostic purpose. A correlation with disease severity (mini-mental state examination score) or age of patients was also not clear, although we think that this must be further examined with larger number of samples.
As to the blood levels of ␤A4, there are not many studies in the literature, probably because of the difficulty to detect the low levels of ␤A4 in the blood. Very recently, one group measured ␤A4 1-40 and 1-42 levels in the plasma by an ELISA method and reported that familial Alzheimer's disease patients of a Swedish family having a clinical APP mutation (APP Lys-670 to Asn, Met-671 to Leu double mutation) have two to three times higher plasma concentration of both ␤A4 1-40 and 1-42 compared with non-carrier of the mutant gene (41). Furthermore, they also reported that a clinical presenilin-1 (S182) gene mutation in another familial Alzheimer's disease case also had about two times higher ␤A4 1-42 levels in plasma (42). In sporadic AD cases, this increase of blood ␤A4 levels were less prominent, but still about 10% of patients are reported to have elevated levels of ␤A4 1-42 (41). In our study, which includes mainly sporadic AD patients, no such increase was observed both for 1-40 and 1-42 levels. It must be clarified in future studies whether this discrepancy is simply because our assay for blood samples was not accurate enough to detect this subtle increase of 1-42 levels or because of some other reason such as different severity of the disease in the two patient groups or due to the different assay methods used in these two studies.
With our Western blot assay, we were also able to detect N-terminally truncated forms of ␤A4 from CSF and blood. The measurement of N-terminally truncated ␤A4 in individual samples was of considerable interest, since recent studies indicated that brain deposits contain large amounts of these shortened ␤A4 species along with full-length ␤A4, and espe-cially the diffuse plaques, which are assumed to be a premature form of senile plaques, are reported to contain N-terminally ragged forms as the major component, suggesting its contribution in early stages of AD pathogenesis (21). In CSF, we detected two different forms of N-terminally truncated ␤A4 species, which are ending at C40 and migrating at a positions corresponding to 3.3 and 3.7 kDa. In comparison with the results of former biochemical analyses using pooled CSF (9), we assume these two band are most likely to be ␤A4 11-40 and ␤A4 6 -40. Although ␤A4 ending at Leu-34 was also identified as one of the major components (9), we could not detect this C-terminally truncated form in our assay using W0-2 for staining (Fig. 3). We were also unable to detect the p3 peptide (␤A4 17-40), which is abundant in cell culture conditioned medium, suggesting either the ratio of various ␤A4 species produced in vivo and in the cultured cells are different or the in vivo clearance of p3 peptide is faster than of the longer ␤A4 species. The amount of truncated ␤A4 species was very low and visible only after the immunodepletion of full-length ␤A4. There was no obvious correlation between their amount and the diagnosis of AD, in this measurement of a relatively limited number of samples. However, since we could not detect N-terminally truncated ␤A4 species ending at C42, there is still the possibility that the levels of this more hydrophobic peptide is different between AD patients and controls and that this long-tailed ␤A4 species could contribute to AD pathogenesis. Further studies will be needed to examine this possibility.
There were several differences between the N-terminally truncated variants of ␤A4 found in CSF and in plasma. In plasma, we detected only a 3.7-kDa form and relative amount of this molecule (p3.7/full-length ␤A4 ratio) was much higher compared with CSF. We also detected a 5-kDa band only in plasma. This band presumably represents an APP fragment generated by cleavage at the C40 end of ␤A4 and at N-terminal upstream of ␤A4 residue Asp-1 (43). These differences between plasma and CSF suggest that the cellular sources and/or metabolism of ␤A4 are different in these two compartments.
The results of this study do not support the hypothesis that the ␤A4 deposition in the brain of AD patients is caused by extracellular accumulation of soluble ␤A4. There still remains a possibility that minor structural variants of ␤A4, such as Glu 3 -pyroglutamated ␤A4 (23), ␤A4 isomerized at Asp or Ser residues (44,45), or conformational variant of ␤A4 (46) are increased in CSF of AD patients and contribute to the pathogenesis, since our assay may not distinguish these species from unmodified forms. However, we rather assume that soluble ␤A4 in body fluids is neither a useful marker of the disease without additional information, nor does it play a direct role in the pathogenesis, at least in cases of sporadic AD. Whether the mechanism of ␤A4 deposition between familial AD and sporadic AD is the same or not is an important issue to be addressed in future studies.