The Profile of Soluble Amyloid β Protein in Cultured Cell Media

To study the metabolism of amyloid β protein (Aβ) in Alzheimer's disease, we have developed a new approach for analyzing the profile of soluble Aβ and its variants. In the present method, Aβ and its variants are immuno-isolated with Aβ-specific monoclonal antibodies. The identities of the Aβ variants are determined by measuring their molecular masses using matrix-assisted laser desorption ionization time-of-flight mass spectrometry. The levels of Aβ variants are determined by their relative peak intensities in mass spectrometric measurements by comparison with internal standards of known identities and concentrations. We used this method to examine the Aβ species in conditioned media of mouse neuroblastoma cells transfected with cDNAs encoding wild type or mutant human amyloid precursor protein. In addition to human Aβ-(1-40) and Aβ-(1-42), more than 40 different human Aβ variants were identified. Endogenous murine Aβ and its variants were also identified by this approach. The present approach is a new and sensitive method to characterize the profile of soluble Aβ in conditioned media and biological fluids. Furthermore, it allows direct measurement of each individual peptide in a peptide mixture and provides comprehensive information on the identity and concentration of Aβ and Aβ variants.

The discovery that soluble A␤ (sA␤) is a constituent of cerebrospinal fluid (CSF) (19 -21) and cultured cell media (20,22) indicates that A␤ is a normal product of cellular metabolism of APP. The major form of A␤ in biological fluids is A␤- . Both NH 2 -terminal and COOH-terminal truncated sA␤s have been identified in pooled CSF specimens (21). A␤-(17-X) has also been detected in cultured cell media (22). The hypothesis that APP metabolism and A␤ production play central roles in the pathogenesis of AD is supported by the observation that certain familial AD mutations cluster within or immediately adjacent to the A␤ domain and these mutations influence APP processing and A␤ production. For example, the APP 670/671 double mutations produce elevated A␤-  and A␤-  in vitro and in vivo (23,24). Further, APP 717 mutations increase the relative amount of highly amyloidogenic A␤-(1-42/1-43) (27,28), which may, in turn, recruit A␤-  to be deposited into amyloid plaques (29).
Numerous A␤ species, in various concentrations, have been described in amyloid deposits from AD and control brain tissues (30). However, no significant difference that distinguishes AD patients and controls has been described in CSF studies of either sA␤ peptide species (30) or concentration (20,31,32). We consider it likely that earlier methodologies used for A␤ analysis may have failed to detect subtle variations in A␤ levels between samples.
To investigate the production and metabolism of A␤ in tissue, biological fluids, and cultured mammalian cell conditioned medium, we have developed a sensitive method for measuring sA␤ and its variants using the strategy of microscale immuno-affinity capture (immunoprecipitation) and mass-specific identification (33,34). The method relied on the combined approaches of immunoprecipitation and mass spectrometric analysis (IP/MS). The sA␤ and its variants were selectively isolated by immunoprecipitation with anti-A␤ mAbs. The identities of these isolated A␤ peptides were determined by measuring their molecular masses using matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF-MS) (35). The relative signal intensities were used to estimate the concentrations of A␤. Using this approach, we have detected several novel A␤ variants and have successfully quantified sA␤s in the conditioned media of cultured mammalian cells.

EXPERIMENTAL PROCEDURES
Cell Culture-Mouse neuroblastoma cells (N2a), and cells stably transfected with Myc-epitope tagged wild type human APP 695 cDNA (N2a/APP 695 ) or APP 695 cDNA encoding Swedish mutant (N2a/APP 695:595/596NL ) (36) were cultured in Dulbecco's modified Eagle's medium (high glucose)/Opti-MEM (1:1) containing 5% fetal bovine serum and 1% penicillin/streptomycin. All cell culture media were obtained from Life Technologies, Inc.. Geneticin (G418) was added to the growth media for the stably transfected cells at a concentration of 0.2 mg/ml. Cells were plated in 100-mm tissue culture dishes and grown in 5% CO 2 in a 37°C incubator.
A␤ Degradation in Cell Culture Media Conditioned by Non-transfected N2a Cells-Cell conditioned media of non-transfected N2a cells conditioned for 18 h were freshly prepared, as above but without addition of protease inhibitors. Synthetic human A␤-(1-40) or A␤-(1-42) (50 nM) was added to freshly harvested conditioned media and incubated at 37°C for 22 h. At the end of the incubation period, 1 ml of medium was collected and protease inhibitor mixture (see above) was immediately added to the sample media. The sample was split, and 0.5 ml of aliquots were immunoprecipitated with either mAbs 4G8 or 6E10 (see above). A␤-related peptides were analyzed by MALDI-TOF-MS using trifluoroacetic acid/water/acetonitrile (1:20:20, v/v/v) as elution buffer (see above). . Spectrum A shows the mass range from 1800 to 4800 Da, and spectrum B is an expansion of the range from 1800 to 3100 Da. The identities of the observed peaks are indicated using murine A␤ sequence numbers (see Fig. 2). The measured masses for these peaks are listed in Table I. Peaks corresponding to doubly protonated A␤ peptides were labeled 2ϩ. The peak labeled u corresponds to the unidentified background peaks. No murine A␤ peptides were detected from IP/MS spectrum using mAb 6E10 (C). dium from non-transfected N2a cells was used as a control to examine the specificity of IP/MS for human A␤. 1 ml of conditioned medium was immunoprecipitated by mAbs 4G8 and 6E10 (separately) with Protein G/A-agarose beads and the immunoprecipitated peptides were analyzed by MS. Using mAb 4G8, we observed a series of peaks in the spectrum (Fig. 1, A  and B). The molecular masses of the observed peaks indicated that they correspond to the amino acid sequences of murine A␤ peptides (42). A summary of molecular masses of the observed peaks and the calculated molecular masses and amino acid sequences of their corresponding murine A␤ peptides are provided in Table I. Although mAb 4G8 was raised against the amino acid sequence of human A␤, we have successfully detected murine A␤. The mAb 4G8 detects epitope between amino acids 17-24 of A␤, a sequence conserved between murine and human A␤s (Fig. 2). In contrast, mAb 6E10 is human-specific and fails to recognize murine A␤s (Fig. 1C). The amino acid sequences of murine A␤ and human A␤ differ at amino acids 5, 10, and 13 ( Fig. 2). The differences in amino acid sequence also contribute to the differences in their molecular masses and made it possible to differentiate human and mouse A␤ species by mass spectrometry.

Human A␤ and Its Variants Can Be Detected and Distinguished in Conditioned Medium of Cells Expressing Human
APP-Cultured mammalian cells constitutively produce and secrete A␤. The profile of total soluble A␤ expressed by cultured cells was characterized using mouse neuroblastoma (N2a) cells, which were stably transfected with human APP cDNAs encoding either wild type (N2a/APP 695 ) or Swedish familial AD mutant APP (N2a/APP 695:595/596NL ). These mouse neuroblastoma cells were incubated in serum-reduced media (0.2% fetal bovine serum) for 24 h. The sA␤ peptides secreted into the cultured media (0.5 ml) were analyzed by immunoprecipitation with anti-A␤ mAbs and MALDI-TOF-MS. The mass spectrum resulting from IP/MS analysis of A␤ peptides from N2a/APP 695: 595/596NL cell conditioned medium using mAb 4G8 is shown in Fig. 3. The identities of peaks were assigned according to the measured molecular masses and several criteria (see below) and are identified in the figure according to human A␤ amino acid sequence numbers. In addition to the species containing A␤ amino acid sequences 1-40 (A␤-(1-40)) and 1-42 (A␤-(1-42)) (the major components found in senile plaque of AD), we observed several short forms of A␤-related peptides. A sum-mary of these A␤-related peptides is given in Table II. A similar A␤ profile was observed for cells expressing wild type human APP 695 (N2a/APP 695 ) (data not shown).
The molecular masses of individual peptides (observed as peaks in the mass spectrum) were measured and used to determine the identities of the these peptides. Several criteria were used for determining the identities of A␤ peptide variants. (i) The peak observed in the spectrum of cell culture medium conditioned with N2a/APP 695:595/596NL cells (or N2a/APP 695 cells) should not exist in the control spectrum, which resulted from assay of medium conditioned by non-transfected N2a cells. (ii) The measured molecular mass of that peak should match the calculated molecular mass of A␤ peptide(s) based on the human A␤ amino acid sequence. (iii) The tentative matched A␤ peptide should contain the epitope site for the mAb used in the immunoprecipitation. (iv) If the predicted A␤ peptide contains a methionine residue, then its mass should increase by 16 Da when formic acid/water/isopropanol is used as elution buffer, since the methionine residue in A␤ peptide at position 35 is readily oxidized under these conditions (42).
The (430 ng/ml to 43 pg/ml), and A␤-(1-28) (320 ng/ml to 32 pg/ml) using serum-reduced media conditioned by non-transfected N2a cells for 12 h and immediately treated with protease inhibitors upon harvest. Conditioned medium of N2a cells was used to mimic the conditions used in a time-course study (see below). Only a small amount of endogenous murine A␤ was produced within 12 h and did not interfere with the measurement of the human A␤ peptides. A␤-(12-28), as internal standard, was added to each of the diluted solutions with a constant concentration of 20 nM prior to immunoprecipitation. We used A␤-(12-28) peptide as the internal standard because it appears not to be produced naturally (we did not detect it in our analysis of secreted peptides from N2a cells). The effects of experimental variations (e.g. signal variations due to the absolute amount of Protein G/A-agarose beads or to the laser intensity used during the MS measurement) on the resulting MS peak height was corrected by normalization to the internal standard, i.e. the peak height of A␤-(12-28). The peak heights of A␤ peptides were measured, and the relative peak heights of A␤-(1-42), A␤-(1-40), and A␤-(1-28) to A␤-(12-28) were calculated. These relative peak heights were used to evaluate the reproducibility and sensitivity of IP/MS in measuring A␤ peptides. Fig. 5 depicts a series of mass spectra resulting from immunoprecipitation of five different concentrations of A␤ peptides using mAb 4G8. The peaks corresponding to A␤-(1-42), A␤-(1-40), A␤-(1-28), and A␤- (12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28) are labeled and also indicated by dotted lines. Other peaks appear in the spectra, especially in spectra of low A␤ concentrations. These peaks correspond to endogenous murine A␤ peptides. The spectra in Fig. 5 clearly show that A␤s can be detected at levels as low as 0.1 nM (0.01 nM for A␤1-28), with a signal-to-noise level greater than 2. This . Spectrum A shows the mass range from 1800 to 4800 Da, in which A␤ peptides were observed, and spectrum B is an expansion of the range from 1800 to 3100 Da. The identities of the observed peaks are indicated using human A␤ sequence numbers (see Fig. 2) and denoted as described in Fig. 1 legend. The measured masses for these peaks and the peaks observed in Fig. 4 are summarized together in Table II. FIG. 4. MALDI-TOF-MS spectrum of human soluble A␤ and its variants detected from 0.5 ml of N2a/APP 695:595/596NL cell conditioned medium by IP/MS using mAb 6E10 (see "Experimental Procedures" and Fig. 1 legend). The peak labeled insulin, 2ϩ corresponds to doubly protonated bovine insulin, which is added for mass calibration. The ϫ 5.0 label in A indicates 5-fold amplification in y scale. corresponds to about 0.4 ng/ml for A␤-(1-40). The average relative peak height for each A␤ peptide obtained by three independent measurements are plotted versus concentration in Fig. 6. This plot indicates that IP/MS has a very wide dynamic range (linearity between the relative peak height and the A␤ concentration) for A␤ measurement. The correlation coefficients between the relative peak heights and peptide concen- N2a/APP 695:595/596NL cells were propagated in 1.5 ml of growth medium in 12-well tissue culture plates. To prepare condi-tioned media, cell monolayers were washed three times with phosphate-buffered saline and then placed in 1 ml of serumreduced medium and incubated at 37°C, 5% CO 2 for times between 1 h and 30 h. The harvested media were treated immediately with a protease inhibitor mixture (see above) to prevent proteolytic degradation. A␤s were immunoprecipitated together with an internal standard, A␤-(12-28), by mAb 4G8 and analyzed by MS. The mass spectra from four time points are shown in Fig. 7. In order to reduce the fluctuation in the A␤ measurements that is caused by the variation in cell number in each measurement, two separate measurements were performed for each time point. The average peak heights of A␤ peptides relative to standard (A␤-(12-28)) and measured A␤ peptide concentrations using standard curve (Fig. 6) were plotted versus time to view the relationship between the truncated A␤ and A␤-(1-40/1-42) (Fig. 8). We observed that the total level of A␤s increased during the 20-h incubation time. The data shown in Fig. 8 represent the integrated values of sA␤ peptides in cell culture media. In other words, these results represent net A␤ levels, which are presumably determined by the balance of production and degradation of A␤ in the cell culture media. The rate of A␤ turnover cannot be determined from our current data due to the continuous production of A␤ by cells. Proteases in Conditioned Medium Degrade Exogenous A␤ to an Insignificant Degree-To examine the origin of the short A␤ peptides observed in cell culture media, we studied A␤ proteolysis in conditioned medium of non-transfected N2a cells. Conditioned medium was freshly prepared by conditioning serumreduced medium with non-transfected N2a cells. Synthetic human A␤-(1-40) (50 nM) was incubated at 37°C for 22 h in the medium of cell conditioned for 18 h. The degraded A␤ peptides were analyzed by IP/MS with mAbs 4G8 and 6E10. The representative mass spectra resulting from 0-h and 22-h samples of A␤-  are shown in Fig. 9. We observed very low levels of truncated A␤ peptides after 22 h of incubation (Fig. 9, B and D). These degraded A␤ peptides, primarily generated by proteolytic cleavages after lysine 28, histidine 13/14, and hydrophobic amino acids (phenylalanine 19/20, glycine 33, and leucine 34), are present at low levels in comparison to full-length A␤-(1-40).

Amyloid ␤ Protein and Its Variants Are Readily Detected by
Immunoprecipitation/Mass Spectrometric Analysis-The present study reports a comprehensive profile of sA␤ secreted by cultured cell lines using a novel IP/MS method. Sixty-four A␤-related peptides (44 from human and 20 from murine A␤ sequences) have been identified from immunoprecipitated conditioned cell culture media without further purification. Compared to other methodologies used to measure sA␤, the IP/MS has several significant advantages. High affinity antibodies were utilized to specifically purify and enrich the peptides of interest. This enrichment enabled us to analyze the peptides of interest directly from a small volume of biological fluid with low A␤ concentrations. This technique should be very useful in the analysis of sA␤ from human specimens (e.g. CSF). Accurate molecular masses of A␤ peptides were determined in the present measurement. Because highly specific A␤ monoclonal antibodies were utilized, these accurate molecular masses provide essential information regarding the identities, amino acid composition, and sequence length of A␤ and its related variant peptides. Another important feature of this mass-specific detection assay is the ability to simultaneously analyze complex mixtures of A␤s in biological samples in a single assay. Hence, both the molecular masses and the concentrations of sA␤s can be determined in a single measurement using appropriate A␤peptide analogs as internal standards with appropriate standard curves (see below).
In contrast to the A␤-related peptides observed from cell culture media conditioned with N2a/APP 695:595/596NL and N2a/ APP 695 cells, a different proteolytic digestion pattern was observed from cell culture media conditioned by non-transfected N2a cells. The predominant murine A␤ peptide fragments observed in this study apparently resulted from cleavages around amino acid residue Glu 11 . These peptides start at either Glu 11 or Val 12 at their NH 2 termini. The different termini in A␤ species of human versus mouse A␤s may be due to alternative proteolytic events that are directed by the differences in human and mouse A␤ peptide sequences (Fig. 2).
We have observed a complex set of A␤ peptides from cultured cell media. We are quite confident that these peptides are not artificially produced by proteolysis of full-length 1-40/42 peptides in the medium after harvest, or during our antibody incubation steps. In support of this view, we have demonstrated that synthetic A␤-(1-40) and A␤-(1-42) peptides are not degraded upon addition to the medium of non-transfected N2a cells conditioned for 12 h and subsequently subject to IP/MS (Fig. 5), in the existing of protease inhibitors. Furthermore, these A␤ peptide fragments are also not generated extrocellularly by proteolytic degradation as consequence of proteolytic activity in fetal bovine serum that was used in the conditioning medium (data not shown). The mechanisms by which these A␤-related peptides are generated is presently not clear. Nevertheless, clarification of the relationship of these short A␤ forms and longer A␤-(1-40/1-42) forms may provide some insight into the production, metabolism and deposition of A␤ in AD.
Broad Concentration Ranges of A␤ Peptide Can Be Measured by IP/MS-In conventional immunoprecipitation, quantitative recovery of a specific protein (antigen) is the primary goal. This is achieved by using excess antibody and excess precipitation reagents. Normally, 10 l of mAb and 15 l of Protein G/Aagarose beads are used for immunoprecipitation from 1 ml of cellular extraction or cell culture media (according to the manufacturer's protocol, Oncogene Science, Inc). In comparison, our present approach uses only 1 l of mAb and 3 l of Protein G/A-agarose beads for immunoprecipitation. The small amount of immunoprecipitation reagent used here is intended to reduce the final elution volume and allow direct mass spectrometric measurement of antigens. We have attempted to achieve quantitative sampling of the antigens rather than 100% immunoprecipitation of them. For quantitative sampling (using small amount of precipitation reagent), a proper ratio of mAb and precipitation reagent is essential to provide high binding capacity and high detection sensitivity for the peptide (antigen) of interest. If the ratio is too low, the peptide binding capacity will be reduced and only the most abundant species and some minor species (possibly certain polymeric A␤ peptides) with substantially higher avidity to the mAbs will be detected. If the ratio is too high, the sensitivity will be reduced by the increased occupation of free antibodies on the binding site of the precipitating reagent. Therefore, the amount of mAb and Protein G/A-agarose bead and the ratio of these two reagents was carefully optimized. Using a mAb/precipitation reagent ratio of 1:3 (v/v), we have been able to measure a very broad concentration range of A␤ peptides (from 0.01 nM to 100 nM) (Fig. 6). The concen- tration of A␤ in CSF, plasma, and cultured cells has been reported to be in the range of 0.6 -8 nM, 0.1-0.3 nM, and 0.7-1 nM, respectively (as the black bar indicates in Fig. 6) (18,31). The results from our experiments suggest that IP/MS will be useful in measuring A␤ and its variants in human body fluids. The sensitivity of IP/MS is comparable with enzyme-linked immunosorbent assay, which is Յ0.1 ng/ml (0.02 nM) for A␤s (18).
Our data also indicate that different A␤ peptides have very different signal intensities when assayed by mass spectrometry. For example, at the same concentration, A␤-(1-28) results in a peak intensity that is almost 5-fold higher than that of A␤-(1-40). These different peak intensities are likely due to the different ionization efficiencies of the different peptides in the mass spectrometric measurement. These different responses may also arise from differences in solubility of each individual peptide and/or differences in affinity for the antibodies. However, these experimental effects on the quantitative measurement of A␤ can be corrected by using appropriate A␤-peptide analogs as internal standards and standard curves for each peptide as present in the mixture. Our data strongly suggest that it is inappropriate to compare the concentrations of any two peptides from their peak heights in the mass spectrum. The concentration of a given peptide can only be measured using the relative peak height of this peptide to an internal standard and compared with a standard curve of the corresponding peptide. In the absence of a standard curve, the relative peak height can only be used to semi-quantitatively compare the levels of the same peptide.

The Concentrations of A␤-(1-40) Peptide and A␤-Related Peptides Increase in Parallel over 20 h in Cell Culture Media-
The results of our time-course experiments indicate that the concentrations of both A␤-(1-40) and A␤-(1-42) increase over a 20-h period (Fig. 8). More interestingly, the major fragment peptides of A␤ (for example, A␤-(1-34) and A␤-(1-28)) observed at late time points were also detected at very early time points and their amounts increased in proportion to the increasing amounts of A␤- . This result suggests two possibilities regarding the metabolism of these A␤ fragment peptides. The first possibility is that these A␤ fragment peptides are generated directly from APP by bona fide proteolytic processing events and are secreted simultaneously with A␤-(1-40) into the cell culture medium. The second possibility is that A␤ is constantly degraded by proteases immediately after its secretion into the cell culture medium.
A␤ Degradation in Conditioned Medium-Our analysis of degradation products of synthetic human A␤ peptides revealed four primary cleavage sites (His 13 /His 14 , Phe 19 /Phe 20 , Lys 28 , and Gly 33 /Leu 34 ) with three different endoprotease substrate specificities. These A␤ peptides contribute to the low levels of certain A␤ peptide variants normally observed in cell culture media of transfected cells. However, only a small portion of synthetic A␤-(1-40) peptide was degraded in conditioned medium over 22 h. A␤ degradation by proteases released from cultured Chinese hamster ovary (CHO) cells and monkey kidney COS cells have been reported recently (44,45). In the published studies (44), pulse-chase analysis revealed that the concentration of A␤ in the CHO conditioned medium reaches a peak level about 6 h into the time course, and diminishes over the next 18 h. The inhibition of A␤ degradation in CHO conditioned medium required the presence of all four major classes of protease inhibitors. Rapid degradation of A␤ in COS conditioned medium has also been reported (26). A serine protease-␣ 2 -macroglobulin complex was described in the latter report as being responsible for A␤ degradation in COS conditioned medium. This novel serine protease cleaves A␤ primarily after hydrophobic amino acid residues (A␤ residues 17-19 and 32-34) and degrades synthetic A␤-(1-40) almost completely in 16 h. This protease activity can be inhibited by certain types of serine protease inhibitors (26). In the present report, we demonstrated A␤ degradation in N2a conditioned medium. However, the cleavage sites and degradation rates were different from those shown in CHO and COS cells. These differences most likely indicate that cell line-dependent processes are operative. However, in view of our demonstration that the shortened A␤-related species appear at the earliest time points and accumulate in proportion to A␤-(1-40) over the time course (Figs. 7 and 8) and our observation that only a small portion of exogenously added A␤ peptide was degraded in conditioned medium (Fig. 9), we conclude that the vast majority of shortened A␤-related species are generated during intracellular processing of APP.
Here, we have shown that the combination of immunoprecipitation and mass spectrometry provides a specific and sensitive approach for studying amyloid ␤ protein and its variants in conditioned media. This method should greatly facilitate studies of the metabolism of A␤ in vitro and in vivo.
Acknowledgments-We thank Mrs. Florence Martin for her generous contributions to support research in Alzheimer's disease. We thank Dr. Gopal Thinakaran for supplying N2a cells and N2a cells stably expressing human APP and human APP with the "Swedish" mutation. We thank Dr. Brian T. Chait of The Rockefeller University for advice and helpful discussions. are the resulting spectra of IP/MS analysis using mAb 4G8, and C and D are the resulting spectra from mAb 6E10. The identities of peaks observed in these spectra were labeled with human A␤ sequence numbers (see Fig. 1 legend).