[1] 16w Proteomic analysis of brain tissue from an Alzheimer's disease mouse model by two-dimensional difference gel electrophoresis

[1] 160w We used a ␤-amyloid precursor protein (APP) transgenic (Tg) mouse model that displays some of the typical Alzheimer-associated pathological features to study the brain proteoma associated with amyloid plaque deposition. Two groups (male and female) of 14-month-old Tg mice were compared with their wild type littermates. We used differential 2D electrophoresis coupled with mass spectrometry to generate one of the first complete image of changes in brain protein expression occurring in this well-recognized model of Alzheimer's disease (AD). We identified 15 different proteins, which are significantly regulated in this pathology (p < 0.05, ≥1.5-fold variation in expression comparing with the wild type samples). These comprise a number of proteins that were already known to be implicated in AD and neurodegeneration, as well as several proteins which relationship with AD had not been shown before. Identified proteins were grouped according to their biological key pathways. Results obtained are discussed in view of existing bibliographic data on human AD transcriptoma and proteoma.

[1] 111w Alzheimer's disease (AD) is the most common senile dementia with ultimately fatal progression for which there is no early diagnostics, as well as no effective treatment at present. A number of pathological processes characterize this neurological disorder. Formation of cerebral plaques containing the ␤-amyloid peptide as the primary component is considered today to be a "prime mover" of the pathogenesis [58]. Dysfunction and death of neurons and loss of synapses in different brain regions lead to reduction in neuronal signal transmission. In addition, many neurons develop neurofibrillary tangles (filamentous inclusions in neuronal cell bodies and proximal dendrites) that are predominantly composed of a filamentous, hyperphosphorylated form of the microtubuleassociated protein Tau.

[2] 125w A␤ peptide, the major constituent of neuritic plaques and vascular deposits in AD patients, is derived by proteolysis from the larger ␤-amyloid precursor protein (APP). Point mutations of the APP gene have been identified in some familial early onset AD cases. There were some unsuccessful attempts to develop transgenic mouse model of Alzheimer's disease [42] before creating at the end of 1990s of animal lines based on expression of different mutant human APPs with brain specified promoters and optimized translation 0197-4580/$ -see front matter © 2006 Elsevier Inc. All rights reserved. doi:10.1016/j.neurobiolaging.2006.01.011 initiation site. The Thy1-APP 751 SL transgenic mice that were utilized in this study reproduce several characteristic features of Alzheimer disease like A␤ peptide deposition, dystrophic neurites formation, and progressive neuronal death [6].

[3] 103w Animal models of Alzheimer's disease and, in particular, APP transgenic mice, provide the means of defining relationships between age-related alterations in behavior and neuropathological and biochemical abnormalities in the brain. Such models are also invaluable for developing new diagnostics and new therapies for detecting and treating AD in humans. The need for new targets as well as validated biological markers for AD justifies intensive research of the model mouse proteoma. In the present study, we applied the powerful technique of differential 2D electrophoresis (DIGE) coupled with mass spectrometry to the analysis of protein modifications occurring in the cortex of 14-month-old APP transgenic mice.

[1] 131w The transgenic mice (Tg) used in this study overexpress a mutated form of human APP in the brain [6]. The transgenic construct contains the human APP gene with two mutations found in familial forms of AD-the so-called London (L: V642I) and Swedish (S: KM595/596NL) mutations. The brain specific Thy1 promoter controls the transcription of this sequence. As a consequence, these mice present a high level of human amyloid peptide A␤ in the brain. From 6 months of age, they develop fibrillar amyloid deposits, associated with dystrophic neurites and neuronal stress markers [6], therefore representing an animal model for Alzheimer's disease. The proteomic assays were performed using 14-month-old animals, which exhibit many amyloid deposits in various brain regions, in particular cortex and hippocampus, as well as memory deficits in behavioral tests [40].

[2] 127w The proteomic analysis was performed using 12 Tg animals and 12 wt littermates, with six females and six males in each group. Mice were perfused transaortically before removing the brain to eliminate serum protein contamination. Cortices were used because plaque deposits are most abundant in this brain region. For the analysis, pooled samples of protein extracts were prepared for each group of six animals (male or female, Tg or wt, see Fig. 1A). Separate proteomic assays were performed for males and females, since it has been observed that the kinetics of intracellular A␤ peptide accumulation as well as plaque deposition were different with mouse gender: on average females start to develop histological signs of amyloid deposition several weeks earlier than males [T. Canton, V. Blanchard, pers. commun.].

[3] 82w Cortex protein extracts from Tg and wt mice were compared using the DIGE technique that was developed for identification of protein variations between samples by 2D electrophoresis. To improve protein spot resolution, each comparison (Tg male versus wt male, and Tg female versus wt female) was carried out using three overlapping linear pH gradients for the first dimension electrophoresis: pH 4.5-6.0, 5.5-6.7, and 6.0-9.0 (see Fig. 1B). Thus, six complete DIGE experiments (three pH gradients for two mouse gender groups) were performed.

[4] 137w Experimental design of a single DIGE Assay is shown in Fig. 1C. The whole experimental procedure can be represented as a sequence of three consecutive steps: (1) experimental determination of a threshold value which allows to determine the level of experimental noise (calibrating assay, Fig. 1C, left panel), ( 2) determination of regulated spots (analytical assay, Fig. 1C, center panel), and (3) physical matching of spots on preparative gels for extraction of the regulated proteins and identification by MS (preparative assay, Fig. 1C, right panel). To make the result of calibrating and analytical assays statistically significant, all gels were run in triplicates. The preparative gels were run in duplicates to ensure enough material for identification of regulated spots. Altogether ten gels (three calibrating, three analytical and four preparative) were run in parallel in one complete DIGE assay.

[5] 537w In the calibrating assay (Fig. 1C, left panel) the same sample -mixture of equivalent amounts of Tg and wt poolswas labeled with three different dyes (Cy2, Cy3, or Cy5). The resulting labeled protein mixtures were combined and separated by 2D-E. Three digital pictures were obtained by scanning the gel at appropriate wavelengths. Simultaneous analysis of the three images was achieved using the DeCyder software, which performs the following operations: delineation of spot borders, calculation of integrated intensity for each spot and each channel, then calculation for each spot of the abundance ratio between the Cy5 and Cy2 channels after normalization using the Cy3 channel (Fig. 2). This methodology ensures that spots are delineated in the same way for the three channels, and that expression ratio between the Cy5 and Cy2 channels is calculated in a reliable manner through normalization by a common reference channel Fig. 1. Experimental design of the proteomic study. (A) Samples used for the proteomic study: 12 APP-transgenic mice (six males and six females) as well as 12 wild-type littermates (six males and six females) were used. Protein extracts were generated from the cortex of each animal and pooled in four samples: Tg male, wt male, Tg female, wt female. Separate analysis comparing Tg and wt was conducted for each gender. (B) Summary of the different DIGE experiments performed: three complete DIGE analyses corresponding to three overlapping pH gradients were conducted for each gender: a total of six complete DIGE analyses were performed. (C) Experimental design of a single DIGE assay-(Left panel) Calibration assay. A mixture of three samples labeled with Cy2, Cy3 or Cy5 (each of them representing the same mixture of equivalent amounts of Tg and wt samples) was separated by 2D-E. The resulting gel was scanned at appropriate wavelengths for each dye and three pictures of the same gel were generated. Variations in intensity of each protein spot were assessed to determine a threshold value for further determination of regulated spots in the analytical assay. Three calibrating gels were run in parallel to make the obtained threshold value statistically significant. (Center panel) Analytical assay. Three differently labeled samples (Tg sample labeled with Cy5, wt sample-with Cy2, and mixture of equivalent amounts of both-with Cy3) were separated by 2D-E on the same gel. Three pictures of the same gel obtained at appropriate wavelengths were analyzed. Three analytical gels were run in parallel for statistical significance. (Right panel) Preparative assay. Preparative amounts of unlabeled Tg or wt sample were separated by 2D-E. Each preparative gel was run in duplicate, stained by SYPRO Ruby and scanned. Appropriate spots were excised and proteins identified by MS. (Cy3). The resulting distribution of ratios was approximated by a Gaussian curve, and the ratio corresponding to 2S.D. (95% confidence interval) was determined. Since in the calibration assay all three samples were the same Tg/wt mixture, differences in spot intensities detected after comparison of the corresponding gel pictures result from experimental noise only. These experimental noise variations were calculated for three independent calibrating gels run in parallel and the ratio corresponding to the 95% confidence interval was found to be 1.48 ± 0.02. Thus, the value of 1.5 was taken as a threshold level for the subsequent analytical assay.

[6] 134w For the analytical assay, a set of triplicate analytical gels was run to identify spots differing in protein amounts between Tg and wt samples. They were run exactly as calibrating gels, except that three different samples were labeled with the three dyes: 50 g of proteins from wt extract pool were labeled with Cy2, 50 g of proteins from Tg extract pool were labeled with Cy5, and a reference sample corresponding to a mixture of both (25 g wt + 25 g Tg) was labeled with Cy3 (Fig. 1C, center panel). For each spot, an expression ratio was calculated between the Tg and wt channels after normalization using the Cy3-labeled reference sample (Fig. 2). The final expression ratio reported is the average of the expression ratio obtained for the triplicate analytical gels (Fig. 2C).

[7] 23w According to the threshold level detected in the calibration assay, only spots with expression ratios higher than 1.5 were considered as significantly regulated.

[8] 198w An illustration of the whole analysis process is shown in Fig. 2. Fig. 2A represents the image of one analytical gel run with male mouse cortex samples (pH gradient 4.5-6.0) and taken in the Tg-Cy5 fluorescence channel. The regulated spots (identified after image analysis) are surrounded by white borders. An area containing several regulated spots (outlined by the white rectangle) is shown with magnification in Fig. 2B, with the wt-Cy2 image on the left, and the corresponding Tg-Cy5 image on the right. Finally, the table in Fig. 2C shows the calculation of the abundance ratio (Tg versus wt) for one selected spot (marked by an arrowhead in Fig. 1B). The Tg/wt ratio was calculated for the three analytical gels, and the final value taken into account is an average of these three independent measurements. By convention, when a calculated ratio was smaller than 1, the inverse ratio is indicated with a negative sign. A Student's t-test was performed and the corresponding p-value was calculated for each spot ratio. These values reflect the probability of obtaining the observed data if the two groups (Tg and wt) had the same protein intensity and show the statistical significance of the results.

[9] 80w Preparative gels were loaded with non-labeled protein samples, using 10-times more protein than analytical gels (Fig. 1C). Two gels were run with the wt samples, and two gels with the Tg samples, to ensure recovery of enough material for spot identification. After gel staining with the SYPRO Ruby fluorescent dye, spots selected as regulated in the analytical gels were matched on the preparative gels, excised using a robotic system, and analyzed by mass spectrometry for identification of the corresponding protein(s).

[10] 109w For MS identification, the excised spots were submitted to in-gel trypsin proteolysis. The resulting peptide mixtures were then analyzed by MALDI-TOF-MS. In the cases where MALDI-TOF-MS gave ambiguous results, tandem mass spectrometry was used after chromatographic separation using a capillary LC-MS/MS system. This hierarchical approach was effective in reducing the number of samples for analysis by the time-consuming LC-MS/MS procedure. Furthermore, this methodology was also essential for identification of low expression and/or low molecular weight proteins for which MALDI-TOF-MS analysis generated poor signals. The two approaches (MALDI-TOF-MS and nanoLC-MS/MS) are complementary, and the use of two different ionization processes contributed to the detection of a wider range of peptides.

[11] 161w Thirty-five regulated spots were identified in this way and are listed in Table 1. In this table, average expression ratios corresponding to each regulated spot are shown together with the corresponding t-test p-values. Twenty-two spots were successfully analyzed by MALDI-TOF-MS, and the corresponding proteins were identified by peptide mass fingerprinting using the Mascot search engine against SWISS-PROT and TrEMBL databases. The average sequence coverage of the identified proteins was around 45% (Table 1). Eighteen spots were successfully analyzed by LC-MS/MS, and the corresponding proteins were identified by database searching using generated collision-induced dissociation MS/MS spectra. The sequence coverage was lower (26% in average), but this was compensated by protein sequence information. When the sequence coverage was lower than 20%, a manual sequencing was performed on several peptides to confirm the presence of the protein in the mixture. Five spots were analyzed by both techniques, which led to identify the same proteins, but in several cases additional proteins were identified by LC-MS/MS.

[12] 164w The 35 identified spots are distributed in all six DIGE experiments. More spots were found in the pH gradient 4.5-6.0 (14 spots for males, 5 spots for females) than in the two others (pH 5.5-6.7: four spots for males, six spots for females; pH 6.0-9.0: three spots for males, two spots for females) (Table 1). Although no precise calibration of isoelectrical point or molecular weight was made on the 2D gels, we ensured that the theoretical pI and MW of identified proteins (Table 1) were consistent with the pH range where the spot had been detected and its height on the gel. Additional browsing of sequence databases was also performed to retrieve only entries corresponding to mouse proteins, and to replace distinct database entries corresponding in fact to the same protein (with the exception of spot 13, for which MALDI-TOF-MS data matched better with mouse GFAP sequence reported in the TrEMBL entry Q295K3 than with mouse GFAP sequence corresponding to the Swissprot entry P03995).

[13] 172w Table 1 presents the complete list of regulated spots and the corresponding proteins. For 28 spots out of 35, a unique protein per spot was identified, and was then considered to be responsible for the observed variation in spot intensity. The remaining spots (7) were found to contain several proteins, and the regulated protein could thus not be unequivocally assigned. However, for six of them, we observed that one protein had already been unequivocally identified in another regulated spot, and that it was often identified with a better sequence coverage than the other proteins found in the same spot. Therefore, we considered that this protein was likely to account for the variation observed. For the sake of exhaustiveness, we still mention the other proteins in italics in Table 1. The last spot (#27) contains two proteins unique to this spot and identified with similar MS sequence coverage. In this case, no experimental element allowed favoring one of the two proteins, and we further considered both as potentially regulated in the transgenic animals.

[14] 65w As shown in Table 1, several proteins were identified in more than one spot. Amongst the 35 spots identified, 11 spots contained the glial fibrillary acidic protein (GFAP), 4 spots contained the apolipoprotein E (ApoE), 3 spots contained the dihydropyrimidinase-related protein 2 (DRP-2, also known as CRMP2), and 3 spots contained the peroxiredoxin 6 (Prxd-6). Thus, these 21 spots account for only four distinct proteins.

[15] 81w Multiple spots for one single protein result from three different situations. They can be found in gels run with different pH gradients, because of some overlap between the three pH ranges (e.g. DRP-2, Prxd-6) or they can be identified in gels run with similar pH gradients but with different samples (male and female extracts). In these cases, they correspond to the identification of the same protein in independent experiments and they are regulated in the same way and with comparable ratios.

[16] 150w Alternatively, different spots corresponding to the same gene can be found in the same gel (same gradient, same gender) suggesting the regulation of several isoforms from the same protein. This is the case for GFAP, which is found in 8 spots in one experiment (male samples, pH gradient 4.5-6.0) and in three spots in a separate experiment (female samples, pH gradient 4.5-6.0). The existence of several transcripts and numerous protein isoforms is well documented for this protein [18,34,33]. Isoforms and post-translational events are not necessarily regulated in the same way and to the same extent under conditions where protein expression is modified. Interestingly, in our study, the various isoforms identified for the GFAP protein are all regulated in the same way (upregulated) but with some variability in the intensity of the regulation depending on the spot/isoform considered. This is also observed for Serum albumin, ApoE and CaMK-II (see Table 1).

[17] 311w As a result of this analysis, 15 proteins were identified as regulated in the cortex of Tg mice compared to control mice. They are listed in Table 2. Eight of these proteins were identified in more than one spot, amongst which five proteins were regulated in both male and female samples. The remaining proteins were more often identified in the male group only. A possible explanation for this discrepancy is that females, known to have a more complex hormonal regulation, may have a greater inter-individual variability which would mask less significant variations after pooling individual samples. Most proteins were found to be upregulated, with average ratios ranging from 1.51 (arbitrary threshold) to 4.3 for GFAP. Only two proteins (CaMK-II and OMP) were found to be downregulated, with average ratios of -1.6 and -2.0, respectively. Additional biological information concerning the 15 identified proteins, gathered using various databases such as Swissprot, LocusLink, or Pubmed is also reported in Table 2. Proteins were clustered in functionally relevant groups, according to the biological pathways in which they are mainly involved. The use of mouse tissue profiling data from Affymetrix mRNA microarrays [F. Bassilana, et al., unpublished results] and bibliographic data from PubMed revealed that 7 proteins out of 15 are enriched in the brain. Moreover, eight proteins are located intracellularly, four proteins are secreted, but only two are transmembrane proteins, in line with the known under-representation of this class of proteins in 2D electrophoresis. A literature survey revealed that eight of the identified proteins have been reported to be regulated in previous proteomic studies, either in AD patient samples [14,19,29,34,45,51], or in transgenic animals [47,54]. We also confirmed the regulation of five proteins which had only been identified as regulated at the transcriptome level [20,24,60]. Finally, five proteins have been identified as regulated in an AD-related model for the first time in the present study.

[1] 140w The present study allowed to identify a set of 15 proteins regulated in the cortex of 14-month-old transgenic mice compared to wild-type littermates. These proteins can be grouped into three main pathways: inflammation and oxidative stress, cholesterol metabolism and neuronal and synaptic signaling. Regulation of a small subset of neuronal proteins in this APP transgenic model is in agreement with the lack of major neuronal loss described in these transgenic mice. The present study validates several earlier findings about AD transcriptoma and proteoma, and brings up new data about genes regulated during AD development. It will be very informative to extend the description of this animal model to younger animals in order to follow the changes in protein expression as a function of aging and kinetics of A␤ deposition, as well as the distribution of changes in distinct brain regions.

[2] 136w A conjugation of efforts will be needed to get an exhaustive view of the AD-related proteoma. New techniques are currently developing, which will enlarge the proteoma subset analyzed and shorten the time of analysis, such as MudPIT (multidimensional protein identification technology), based on multidimensional high-pressure liquid chromatography coupled with tandem mass spectrometry [37] or protein microarray technologies. The nature of the samples analyzed will also evolve with new techniques allowing microdissection of plaques or small brain areas such as laser capture microdissection [32]. The combination of multiple proteomic approaches on various biological samples in both rodent model and human samples will progressively generate a detailed description of the proteomic and cellular changes associated with AD. This will provide essential knowledge for the identification of disease pathogenesis as well as for the discovery of potential biological markers.

[1] 128w The Tg mouse line was generated and established as described in [6]. These mice bear a transgenic construct containing the human APP 751 cDNA with the London (V642I) and Swedish (KM595/596NL) mutations under control of the murine Thy1 promoter (Thy1-APP751). To improve the translation initiation site of APP, an optimized Kozack consensus sequence was introduced. Mice used in this study were 14-month-old Thy1-APP 751 transgenic animals (Tg) and wild-type littermates (wt) from a pure C57Bl6 background. Two groups of 6 males and 6 females were anesthetized (60 mg/kg of pentobarbital and 40 mg/kg of ketamine, i.p.) and perfused transaortically with phosphate-buffered saline for 5 min. Brains were then removed, dissected as described in [25], and cortices were rapidly frozen on dry ice and kept at -80 • C.

[2] 143w Whole individual cortices were lyophilized for 48 h, then 10 mg were crushed in aluminium foil and resuspended in 80 l of a boiling buffer solution containing 32 mM Tris-HCl/Tris Base and 1.2% (v/v) Triton X-100. After 5 min of incubation at 100 • C then 5 min of incubation on ice, 7.5 l of a mixture containing 100 mM Pefabloc and 100 mM EDTA was added, followed by incubation with 18.6 l of a DNase/RNase solution (containing 53.2 u of DNase I (Roche) and 6.4 u of RNase A (Sigma-Aldrich)). The reaction was allowed to proceed at 4 • C for 10 min. Proteins were then solubilized by adding 105 mg of urea, 38 mg of thiourea and 47 l of a solution containing 21.4% (v/v) CHAPS in water. Protein samples were gently mixed at room temperature until complete solubilization of urea.

[3] 58w Insoluble material was quickly removed by centrifugation at room temperature at 14,000 × g for 5 min, and the supernatant was then centrifuged at 100,000 × g for 45 min. The final supernatants were aliquoted and stored at -80 • C. The protein concentration of the extracts was determined by the Bradford method (Protein Assay Dye Reagent, Bio-Rad).

[4] 205w Four pools of protein extracts were generated according to mouse type and gender (Tg male, wt male, Tg female, wt female) by mixing equal protein amount of cortex extract for each of the six corresponding animals. Protein extract pools were conjugated with fluorescent dyes through Nhydroxysuccinimidyl linkages to result in labeling of around 10% of protein molecules. For analytical gels, either 50 g of proteins from wt extract pool, 50 g of proteins from Tg extract pool, or a mixture of both (25 g wt + 25 g Tg), were incubated with 400 pmoles of cyanine dyes Cy2, Cy5 or Cy3 (Amersham Biosciences, Inc.), respectively. For calibrating gels, the same mixture of 25 g of proteins from wt extract pool plus 25 g of proteins from Tg extract pool was incubated with 400 pmoles of each of the three dyes Cy2, Cy5 and Cy3. The reaction was performed for 30 min at room temperature in the dark and stopped by adding 10 mM lysine for 10 min at room temperature. Labeled protein samples were diluted with an equal volume of a solution containing 7 M urea, 2 M thiourea, 4% CHAPS, 20mg/ml DTT, 2% Pharmalyte 3-10, 0.2% (v/v) Triton X-100 before loading on the gel.

[5] 35w Cy-dye images as well as SYPRO Ruby-dye images were collected using a ProXpress fluorescent gel scanner (Perkin-Elmer, Norwalk, CT). Gel images were normalized by adjusting the exposure times according to the average pixel values observed.

[6] 124w The quantification of spot abundance ratios for analytical and calibration gels was carried out using the DeCyder software (Amersham Bioscience, Inc.), which was developed specifically for 2D-DIGE using the experimental design described above. First, for each gel, the Cy2, Cy5, and Cy3 images were merged, allowing the co-detection of spot boundaries on the three images. For each spot, the spot volume (sum of pixel intensities) was calculated in the Cy2 or Cy5 channels then normalized according to the corresponding Cy3 spot volume. Comparison of normalized Cy2 and Cy5 protein expression intensities within each gel gave a standardized expression ratio. This value was compared across all gels for each matched spot, and an average value was calculated using the triplicate values from each experimental condition.

[7] 116w The threshold of significance for differentially regulated spots was defined from the analysis of three calibration gels, loaded with a mixture of wt plus Tg proteins labeled with each of the three dyes (Cy2, or Cy3, or Cy5). After image analysis as described above, the distribution of normalized expression ratios was approximated by a Gaussian distribution and the ratio value corresponding to 2S.D. (95% confidence limit) was determined. On analytical gels, only the spots with an average expression ratio greater or equal to the determined threshold ratio were considered as significantly regulated. Spots were defined as upregulated when their expression was increased in Tg compared with wt, and downregulated when their expression was decreased in Tg.

[8] 60w Finally, spots found out as significantly regulated after analysis of the Cy-dye-labeled images were matched on the SYPRO Ruby protein patterns, and spots were excised from preparative gels by means of an automated spot picker (Amersham Biosciences, Inc.). Upregulated spots were excised from preparative gels containing Tg protein samples, and downregulated spots were extracted from preparative gels containing wt samples.

[9] 116w In-gel digestion was performed with an automated protein digestion system, MassPREP Station (Micromass, UK). The gel slices were washed twice with 50 l of 25 mM NH 4 HCO 3 and 50 l of acetonitrile. The cysteine residues were reduced by 50 l of 10 mM dithiothreitol at 57 • C and alkylated by 50 l of 55 mM iodoacetamide. After dehydration with acetonitrile, the proteins were cleaved in the gel with 8 l of 12.5 ng/l of modified porcine trypsin (Promega, USA) in 25 mM NH 4 HCO 3 , overnight at room temperature. The peptides generated were extracted with 60% acetonitrile in 5% formic acid. The peptide extracts were used for MALDI-TOF-MS and/or nanoLC-MS-MS.

[10] 116w MALDI mass measurements were carried out on an Ultraflex TM TOF/TOF (Bruker Daltonics, USA). 0.5 l of peptide extract was used for the MALDI-TOF-MS analysis and co-crystallized in the matrix ␣-cyano-4hydroxycinnamic acid. The spectra were internally calibrated using two trypsin autolysis peaks at m/z 842.510 and 2211.105. Monoisotopic peptide mass were assigned and used for databases research. Proteins were identified by peptide mass fingerprinting using the program MASCOT (Matrix Science, UK) against SWISS-PROT and TrEMBL databases. The research was carried out in all species. One missed cleavage per peptide was allowed, a mass tolerance of 35 ppm was used and some variable modifications were taken into account, such as carbamidomethylation for cysteine and oxidation for methionine.

[11] 153w Nanoscale capillary liquid chromatography-tandem mass spectrometric (LC-MS-MS) analysis of the digested proteins was performed using a CapLC capillary LC system (Micromass, UK) coupled to a hybrid quadrupole orthogonal acceleration time-of-flight tandem mass spectrometer (Q-TOF II, Micromass). Chromatographic separations were conducted on a reversed-phase (RP) capillary column (Pepmap C18, 75 m i.d., 15 cm length, LC Packings) with a 200 nl/min flow. The gradient profile used consisted of a linear gradient from 95% A (H 2 O/0.05% HCOOH) to 45% B (acetonitrile/0.05% HCOOH) in 35 min, followed by a linear gradient to 95% B in 1 min. Mass data acquisitions were piloted by MassLynx software (Micromass, UK). Mass data collected during a LC-MS/MS analysis were processed and converted into a PKL file to be submitted to the search software MAS-COT (Matrix Science, UK). Searches were conducted with a tolerance on mass measurement of 150 ppm in MS mode and 0.25 Da in MS/MS mode.

[12] 73w Western blot analysis of ApoE, GFAP protein expression was performed according to standard procedures using NuPAGE Western blot equipment from Invitrogen. Samples of protein extracts from individual animals containing 6 or 12 g of proteins were migrated on 4-12% gradient polyacrylamide gels and transferred onto Invitrogen 0.45 m nitrocellulose membranes. The same amount of protein extract was loaded in each well, as confirmed by Ponceau Red staining of the membrane after protein transfer.

[13] 96w Antibodies anti-GFAP (rabbit polyclonal serum, Dako), anti-ApoE (goat polyclonal serum, Calbiochem) and anti-beta-actin (mouse monoclonal antibody AC74, Sigma-Aldrich) were used as primary antibodies; anti-rabbit IgG antibody, anti-mouse IgG antibody (Amersham) or anti-goat IgG antibody (Sigma-Aldrich) conjugated to HRP were used as secondary antibody conjugates. Membranes were treated with ECL mixture (Amersham Biosciences), and the chemiluminescence signal was recorded using a CCD system (LAS-3000, Fujifilm). The quantitative analysis of the resulting images was performed using the Multigauge software provided with the LAS-3000 system, and the results were normalized using the beta-actin signal detected on the same membrane.

[14] 176w Protein samples from mouse cortex corresponding to individual wt or Tg animals were analyzed by Western blot to estimate their content of GFAP or ApoE. As shown in Fig. 3, the intensity of the beta-actin signal was similar for all samples (Tg or wt), and was therefore used as an internal reference for cross-sample normalization. Fig. 3 shows a representative experiment with six wt and six Tg samples analyzed for GFAP expression (Fig. 3A), or ApoE expression (Fig. 3B). The intensity of the GFAP band is roughly similar in all wt samples, whereas it is consistently stronger in all Tg samples compared to the wt ones (Fig. 3A). In Fig. 3B, the band corresponding to ApoE displays a strong intensity in Tg samples, whereas it is hardly detectable in wt samples. A similar experiment was also performed for Peroxiredoxin 6. Despite a faint immunochemical signal, we were able to observe a stronger signal in the Tg samples compared to the WT ones (signal was 1.8 times higher in Tg animals than in wt ones, not shown).

[15] 85w Upon quantification and normalization to the beta-actin signal, the signal obtained for GFAP in Tg cortex samples was 2.1 ± 0.4 higher than in wt samples, a value very similar to the one obtained by DIGE (Table 2). For ApoE, the signal was 6.2 ± 1.5 higher in Tg samples than in wt ones, whereas the ratio determined by DIGE was 2.2. Altogether, these data confirm the increase of GFAP and ApoE in Tg cortices compared to wt ones, as observed by the DIGE technique.

[1] 245w The first dimension gel electrophoresis was carried out using narrow immobilized pH gradient gels (Immobiline Dry Strip, 0.5 mm × 3 mm × 180 mm, containing Immobilines NL 3-10, pH 4.5-6.0, 5.5-6.7, 6.0-9.0, Amersham Pharmacia Biotechnology) and a horizontal electrophoresis apparatus (Multiphor II, APB). Strips were rehydrated in a solution containing 6 M urea, 2 M thiourea, 1% (v/v) CHAPS, 0.5% Pharmalytes 3-10, and 0.4% DTT. Fifty micrograms of labeled proteins (for analytical and calibrating gels) or 500 g of unlabeled proteins (for preparative gels) were loaded. Isoelectric focusing (IEF) for pH 4.5-6.0 and 5.5-6.7 was performed as follows: 50 V for 2 h, 100 V for 2 h, 300 V for 2 h, 2000 V for 16 h and 3500 V for 18 h. IEF conditions for pH 6.0-9.0 were: 1 h at 600 V and then 9 h at 3500 V. After equilibrating in a solution containing 0.1 M Tris-HCl, 36% urea, 30% glycerol, and either 0.5% DTT or 4.5% iodoacetamide (10 min of incubation in both cases), strips were ready to be applied to the second dimension gels (12.7% SDS-PAGE). Gels were run at 10 • C as follows: 50 mA for 1.5 h, 100 mA for 1.5 h, and then 185 mA overnight. For each condition (mouse gender and pH), three analytical gels, three calibration gels, and four preparative gels were run in parallel. Preparative gels containing unlabeled protein samples were finally stained with the SYPRO Ruby dye (Genomic Solutions, Inc.).

[2] 110w We describe here one of the first proteomic analysis of a transgenic mouse model of Alzheimer disease. These mice carry a human form of the APP gene with the London (V642I) and Swedish (KM595/596NL) mutations under the control of the Thy1 promoter, resulting in strong expression of the transgene in the brain and develop amyloid plaques from 6 months of age [6]. Such mice represent a partial model for Alzheimer's disease, as they produce amyloid plaques but no neurofibrillary tangles, the other important hallmark of AD. As such, their proteomic analysis provides a description of the regulatory events taking place in response to the presence of amyloid plaques in brain.

[3] 177w Although the model does not reproduce all the features of AD, transgenic mouse samples present some advantages compared to human samples for studies of cellular, biochemical, and molecular events in AD. Indeed, several AD-related proteomic assays have been carried out earlier using postmortem human brain tissues [13,14,45,51]. Although these samples directly reflect AD pathology, their comparison at the proteomic level is hampered by the possible heterogeneity in post-mortem delays between samples (AD and controls), and protein degradation during this delay. Another source of unwanted heterogeneity when working with human samples is the diversity in terms of disease severity, possible coexisting pathologies and age of the patients. On the contrary, using Tg mice allows to pool only highly homogeneous samples, in terms of pathology development (which takes place at a very similar rate in each mouse), genetic background (use of littermate animals as controls), and sample preparation (use of freshly removed and perfused brain cortices). Consequently, the variations in protein expression that were detected are more likely to be related to the amyloid pathology than to inter-individual differences.

[4] 175w In general, proteomic analyses on AD patient brain samples have led to identify a smaller number of regulated genes than transcriptomic studies [9]. This fact can be partially explained by the nature of separation techniques generally used in proteomics that are based on the protein physicochemical properties and can resolve efficiently only a selected subset of the whole proteoma. For example, it can be expected that membrane and basic proteins, as well as very small polypeptides, will be considerably underrepresented when using 2D electrophoresis. In consequence, it is likely that a subset of regulated proteins will not be found out. In our case, a good example of that is the APP transgene, which is known to be strongly expressed only in transgenic animals but was not identified in the present study. On the other hand, proteomic studies have the strong advantage of directly measuring the protein expression, a parameter more physiologically relevant than the RNA level. Moreover, they enable to resolve distinct isoforms of a protein and identify post-translational modifications which can be functionally important.

[5] 156w In the present study we used a powerful and relatively new method for protein investigation, i.e. the 2D electrophoresis DIGE technique, which has been developed especially for identifying qualitative and quantitative differences in protein expression between two biological samples. First, this method allows a high level of resolution and sensitivity by the use of large 2D-gels and fluorescent dyes. In addition, the possibility to run the transgenic and control samples on the same gel eliminates the usual pitfall of matching of 2D patterns from different gels, making the comparison more straightforward and reliable. Finally, the efficient and automated computer image analysis allowed a thorough and robust quantification including statistical analysis, in order to select only significant changes. Altogether, by using this technique coupled to mass spectrometry for protein identification, we were able to identify 15 proteins which expression is altered in the plaque-bearing cortex of transgenic mice and give a reliable quantitative assessment for each one.

[6] 41w The proteins identified in the present study were grouped according to their key biological pathway. Despite the low number of proteins identified, they can be grouped into three main pathways: inflammation and oxidative stress, cholesterol metabolism and neuronal and synaptic signaling.

[7] 301w Two glial proteins related to the inflammatory response show increased levels in Tg mice in our study: GFAP, which overexpression was confirmed by western blotting, and the complement protein C1q. Chronic inflammation has been consistently observed in the brains of AD patients or transgenic mice developing amyloid plaques [12,36], and it is believed to participate in the pathological process of AD, although its role may be both protective and harmful [46]. Characteristic inflammatory features are the presence of activated microglial cells and reactive astrocytes surrounding plaques, as well as the expression of inflammatory mediators such as cytokines or complement factors [12]. GFAP, an intermediate filament protein specifically expressed in astrocytes, is dramatically upregulated during reactive astrogliosis [38]. C1q is the initial component of the classical complement pathway, and it can be secreted by both microglia and astrocytes. Several findings suggest a direct contribution of C1q in the pathology induced by amyloid plaque formation. Indeed, C1q has been shown to bind A␤ fibrillar aggregates, resulting in the activation of the complement cascade [53], and also to favor A␤ fibrillogenesis, both in vitro and in transgenic mice [7,56]. Moreover, the absence of C1q in Tg2576 mice decreases the level of activated glia without changing the number of amyloid plaques compared to regular Tg2576 animals [22]. Consistent with the observed upregulation of GFAP and C1q in our study, the level of both proteins was reported to be increased in AD patient brain samples [44,61], and amyloid transgenic models [28,36]. The glycosylation of GFAP was also shown to be increased in AD brains [29]. Moreover, immunochemistry studies showed that GFAP and C1q are concentrated in the surrounding of plaques: GFAP was found in reactive astrocytes, and C1q was observed in microglia and amyloid plaques, both in AD patients brain samples and transgenic mouse brains [1,36].

[8] 138w The upregulation of peroxiredoxin 6 (Prxd-6) observed in the transgenic animals may reflect changes in the oxidative status of the Tg mouse brain. This is in agreement with the elevated oxidative stress reported in AD patients brain samples [10], as well as with observations of oxidative stress markers associated with plaque formation in other amyloid transgenic models [6,45]. Prxd-6 (also named 1-Cys peroxiredoxin and antioxidant protein-2) is a bifunctional enzyme with a glutathione peroxidase activity and a phospholipase A2 activity [17], which may be able to both reduce and cleave oxidized lipids in order to restore membrane integrity. This is supported by overexpression and inactivation experiments showing a role for this protein in oxidative defense [35,55]. Prxd-6 was also found to be upregulated in a proteomic study of AD patient brains [45], strengthening its potential relationship with AD.

[9] 328w Cholesterol metabolism is another process known to be implicated in AD pathology [12], as indicated by the association of several cholesterol-related gene polymorphisms with the disease. In particular, the 4 form of the apolipoprotein E gene is associated with an increased risk of sporadic AD at least eightfold, making it the strongest identified risk factor for late-onset forms of the disease [11]. The key role of the cholesterol metabolism pathway in AD pathology was also suggested by retrospective studies showing a 70% lower prevalence of AD for individuals chronically treated with inhibitors of cholesterol synthesis [59]. In this respect, it seems important that two of the proteins that we found to be regulated in the present mouse AD model are related to the cholesterol pathway. First, apolipoprotein E was shown to be highly upregulated in both male and female animals (expression ratios 2.04 and 2.37, respectively). This is consistent with the increased level of ApoE in AD patient brains and its co-localization with senile plaques [3]. As the major apolipoprotein in the brain, ApoE is generally thought to play a role in cholesterol transport, and it is mainly synthetized by astrocytes [49]. It favors the formation of amyloid plaques, since APP transgenic mice develop less plaques when they are deficient for ApoE [4], but it is not clear whether this is due to the modulation of cholesterol metabolism (known to affect A␤ processing), or to a direct effect on A␤ aggregation or clearance [3]. Another protein involved in the cholesterol metabolism pathway was found to be upregulated in our model: the protein ACAT-2 (acyl-coA:cholesterol acyltransferase 2) is one of two enzymes that catalyze the formation of cholesterol esters stored in cytosolic lipid droplets [15]. A possible functional relationship of ACAT-2 to amyloid processing is suggested by the fact that the modulation of ACAT activity (either ACAT-1 or -2) and consequently of cholesterol ester production in a cellular model, is correlated with a parallel modulation of A␤ generation [43].

[10] 27w Altogether, the increased level of expression of ApoE and ACAT-2 in our model could be an indication of altered cholesterol metabolism, possibly in relation with neuronal stress.

[11] 284w A set of neuronal proteins was also found to be regulated in the present model of AD pathology. The regulation of the CaMK-II ␣ subunit, a very abundant kinase essential for synaptic signaling, or of synaptotagmin I and NSF, two proteins involved in vesicle trafficking, may reflect changes in synaptic properties. It should be noted that both synaptotagmin I and NSF were previously shown to be overexpressed in AD patient brain [45], and that NSF was identified in a functional proteomic study as a partner for APP [19]. DRP-2, a protein known to be involved in axon path finding during development [16], was found to be increased in the cortices of transgenic mice. Several proteomic studies reported the altered expression of DRP-2 in AD brains [14,34,45]. However, depending on the study, the protein was either dysregulated [34], downregulated [45], less glycosylated [29] or oxidatively modified [14]. DRP-2 is also known to be hyperphosphorylated in neurofibrillary tangles, another pathological mark of AD [26]. The decreased level of DRP-2 reported in AD brains could be related to neuronal loss, a feature which is not observed in our animal model devoid of neurofibrillary tangles. Rather, the increased level of DRP-2 observed in our model could be related to neuritic reorganization and formation of dystrophic neurites around amyloid plaques. Neuritic reorganization may also be the cause of altered expression of moesin, a protein involved in modulating actin cytoskeleton in relation with cell polarity and motility. However, the changes in neuronal proteins are overall of lower intensity compared to changes described above, and this is consistent with the lack of major neuronal loss described in these mice and in similar APP transgenic models, even at this late age [6,27].

[12] 235w Two serum proteins, albumin and transferrin, were also found to be strongly upregulated. Serum albumin displays the largest change (ratio 5.4). Interestingly, increased expression of this protein had already been demonstrated in another proteomic study analyzing AD patient brain samples [45]. Such an increase of albumin level in AD brain is usually explained by the leakage of the blood-brain-barrier (BBB) [57], since albumin is highly present in serum but almost absent in adult brain [48]. A similar BBB breakdown has been described in some areas of the cerebral cortex of 4-month-old APP transgenic mice [52]. In addition, Poduslo et al. [41] reported adsorption of albumin on vessel walls in APP/PS1 double transgenic mice and Kumar-Singh et al. [31] showed a high association of dense-core amyloid plaques with blood vessels and infiltration of albumin around plaques in two transgenic mice models similar to the one used here. These two reports suggest that an increased level of albumin in the brain of APP transgenic mice is possible even in the absence of BBB leakage. BBB breakdown could also explain the upregulation of transferrin (ratio 2.2), but the latter is also expressed in normal adult brain. The relationship of transferrin with AD includes the genetic association of the C2 allele with the disease [39], as well as its iron transporter function, since metal cations have been proposed to play a role in the development of the pathology [21,50].

[13] 154w A final set of proteins found to be regulated in our transgenic model was not previously described in relationship with this pathology. The increased abundance of M-pyruvate kinase, a ubiquitous enzyme involved in glycolysis, could reflect the increased metabolic activity of activated glial cells, as suggested before [5]. Dnm1l (dynamin-1 like) is a recently characterized protein, with a possible role in mitochondria and peroxisome fission [23,30]. Olfactory marker protein (OMP) is almost exclusively expressed in mature olfactive neurons, but also in small groups of neurons in several areas of the mouse CNS including the cortex [2,8]. Being one of the rare downregulated proteins in our model, the decrease of its expression in APP transgenic mouse brain may be an indication of alterations in neuronal signaling. Regulation of these proteins could suggest new pathways involved in the pathogenesis of AD. However, further investigations will be needed to determine their relationship to amyloidosis in the brain.