Proteomic Characterization of Postmortem Amyloid Plaques Isolated by Laser Capture Microdissection

The presence of amyloid plaques in the brain is one of the pathological hallmarks of Alzheimer’s disease (AD). We report here a comprehensive proteomic analysis of senile plaques from postmortem AD brain tissues. Senile plaques labeled with thioflavin-S were procured by laser capture microdissection (LCM), and their protein components were analyzed by liquid chromatography coupled with tandem mass spectrometry (LC-MS/MS). We identified a total of 488 proteins co-isolated with the plaques and found multiple phosphorylation sites on neurofilament intermediate chain, implicating the complexity and diversity of cellular processes involved in the plaque formation. More significantly, we identified 26 proteins enriched in the plaques of two AD cases by quantitative comparison with surrounding non-plaque tissues in each case. The localization of several proteins in the plaques was further confirmed by the approach of immunohistochemistry. In addition to previously identified plaque constituents, we discovered novel association of dynein heavy chain with the plaques in human postmortem brain and in a double transgenic AD mouse model, suggesting that neuronal transport may play a role in neuritic degeneration. Overall, our results revealed for the first time the sub-proteome of amyloid plaques that is important for further studies on disease biomarker identification and molecular mechanisms of AD pathogenesis.

Introduction sodium dodecylsulfate (SDS), 10% glycerol, 10 mM dithiothreitol (DTT), 1 mM di-Na ethylenediamine tetra-acetate (EDTA), and protease inhibitor cocktail (Roche Applied Science, Indianapolis, IN). After the extraction, the samples were alkylated with 50 mM iodoacetamide in dark at room temperature for 30 min. The total amount of proteins in the samples was estimated on a silver-stained SDS gel according to a standard protein marker with known concentration.
For mass spectrometry analysis, Proteins in each sample were separated on a 6-12% SDS gel (0.75 mm thick) and stained with Coomassie blue G-250. The entire lane was cut into 15 pieces followed by in-gel trypsin digestion (28). The resulting peptides from each gel piece were dissolved in buffer A (0.4% acetic acid, 0.005% heptafluorobutyric acid (HFBA), 5% acetonitrile). A pressure cell was used to load each sample onto a 50 µm i.d. x 12 cm self-packed, fused-silica C18 capillary column as described (29). Peptides were eluted during a 2-hr gradient from 10% to 30% buffer B (0.4% acetic acid, 0.005% HFBA, 95% acetonitrile; flow rate: ~300 nl/min). Eluted peptides were ionized under high voltage (1.8-2 kV), detected in a MS survey scan from 400-1700 atomic mass unit (amu) with 2 µscans followed by three data-dependent MS/MS scans (3 µscans each, isolation width 3 amu, 35% normalized collision energy, dynamic range 3 min) in a completely automated fashion on an LCQ-DECA XP-Plus ion trap mass spectrometer (Thermo Finnigan, San Jose, CA).

Database searching for protein identification
The Sequest algorithm (30) was utilized for searching all MS/MS spectra against the human reference database (ftp://ftp.ncbi.nih.gov/genbank, July, 2003). The parameters were set to allow parent ion mass tolerance to be 3 and to consider only b and y ion series. Modifications were permitted to allow the detection of the following (mass shift shown in Daltons): oxidized methionine (+16), carboxymethylated cysteine (+57), and phosphorylated serine, threonine, and tyrosine (+80). We used more stringent Sequest criteria than previously described (31,32) including: (1) only fully-tryptic peptides were considered; (2) Cn score is at least 0.08; (3) Xcorr should be larger than 2.0, 1.7 or 3.3 for charge states of +1, +2, +3, respectively. To further reduce false-positives, we manually verified proteins matched by less than three peptides, since no false-positives were found among proteins identified by at least three distinct peptides (32). Therefore all peptides were accepted with high confidence. The conversion from the identified peptides to proteins was complicated by the presence of different names for the same protein and/or by the sharing of peptides within several proteins (e.g. protein paralogs) (33). Thus we manually verified all proteins and removed the redundancy. Typically we accepted proteins identified by at least one "unique peptide". Obvious contaminants such as trypsin and keratins were removed. Finally we merged the datasets of the plaque samples from two independent experiments.

Protein quantification by mass spectrometry
Quantitative protein comparison between the plaques and the non-plaque control was carried out in two steps. The first step was based on the number of peptides identified for an assigned protein, indicative of protein abundance. We discarded proteins that were identified by more peptides in the control than in the plaques from either AD case. The second step was based on extracted ion current (XIC) of corresponding peptides in MS survey scans (15,34,35). The ratio of peak intensities of selected peptides was measured in the 15 pairs of peptide mixtures that were generated by in-gel digestion of the 15 pairs of gel pieces as shown ( Figure 1). We analyzed each pair of samples in two consecutive LC-MS/MS runs on the same column and found that the quantitative variation was within two-fold in general by using a trypsin auto-cleavage peptide (VATVSLPR, m/z 842.5 for singly charged ion) as internal control. Therefore, we used two-fold as the threshold for protein enrichment in the plaques. The trypsin autocleavage peptide was also used to normalize the measured peptide ratios, and to normalize the elution time of selected peptides between the pair of LC-MS/MS analysis.
It should be mentioned that, in contrast to the peptide identification that was primarily derived from its MS/MS spectrum ( Figure 3B), the peptide quantification was resulted from the MS survey scan ( Figure 3A). An MS survey scan allowed the detection of many peptide ion peaks, of which only the most three predominant were selected for sequencing by MS/MS analysis. However, the other non-sequenced peaks on MS survey scans could be useful for quantification. For example, when a peptide was sequenced by MS/MS analysis in the plaque sample but not sequenced in the control, it is still possible to find and quantify the peptide ion in MS survey scans of the control sample according to its predicted m/z value and adjusted elution time. Otherwise, if the peptide could not be reliably located in the control sample, we estimated that the plaque/control (P/C) ratio was more than two-fold. Finally, we accepted proteins that were found to be enriched at least two-fold in both AD cases.

Identification of proteins enriched in AD amyloid plaques by quantitative proteomics
Senile plaques in AD brain tissues were stained with thioflavin-S and isolated using laser capture microdissection (LCM) ( Figure 1A). Despite some background staining on the postmortem tissue sections, the plaques were easily distinguishable under the microscope.
Approximately 2,000 plaques from one AD brain were captured to yield ~4 µg of total protein after extraction with SDS-containing lysis buffer. The non-plaque regions from the same brain sections were also procured as a control. The total protein samples extracted from the plaques and the non-plaque control were resolved in parallel on a SDS gel ( Figure 1B), which indicates a similar protein composition in the two samples. The proteins in the entire gel lanes were analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS) as described in Methods. Approximately 30,000 MS/MS spectra were acquired for each sample and searched against a human protein database using the Sequest program (30). The matched peptides were further filtered rigorously and led to identification of 331 proteins in the plaques and 327 proteins in the adjacent non-plaque regions of cortex in this AD case (figure 2).
As the plaques are complex and heterogeneous neuropathological structures and are expected to vary among human cases, the protein components may differ in AD patients. We repeated the entire proteomic analysis using samples isolated from the second AD case and identified 413 proteins in the plaques and 384 proteins in its own non-plaque control (figure 2). The datasets from the two AD cases were combined to result in a list of 488 proteins detected in the plaque samples (Table S1). The current literature identifies about 53 proteins present in the plaques, of which 44 proteins were found in our large-scale analyses (Table S1). As expected, many of these proteins were also found in the cortical areas without the plaques, since some normal cellular elements are components of the plaques, including glia and neurons. Alternatively, these proteins may reflect the capture of normal cellular elements along with the plaques.
To determine which of the above proteins were enriched in the plaque regions, we utilized a strategy outlined in figure 2. First, only the 256 plaque proteins detected in both AD cases were considered to increase the dataset reliability. Although it is difficult to estimate protein abundance directly from its peptide ion current, the number of peptides are identified for each protein is roughly correlated with the abundance of the protein after the protein size is normalized (38,39). Based on this principle, we removed the proteins that were identified more frequently in the controls than in the plaques according to the corresponding peptide numbers (figure 2 and Table S1), and kept the remaining 168 proteins for further quantitative analysis. More recently, several groups proposed a simple relative quantification method via extracted ion current of peptides in successive analyses (15,34,35). We used this method to quantify the 168 proteins by manual inspection of the raw files. For example, during the LC-MS/MS analysis, an A tryptic peptide was detected in an MS survey scan ( Figure 3A) and sequenced by MS/MS ( Figure 3B) in the plaque sample and the control. The extracted ion current signal for the peptide is shown in figure 3C, which allows the peptide quantification in both samples. It is worth noting that a trypsin auto-cleavage peptide was used as an internal standard to normalize the experimental variation and to fit the elution time of selected peptides. The plaque/control ratio of A abundance was determined to be around 80, indicative of the enormous enrichment of A in the plaques. The relative abundance of all other proteins was quantified in the similar manner, resulting in the final acceptance of 26 proteins enriched at least two-fold in the plaques of both AD cases (Table 1). However, the majority of previously identified plaque proteins are not in this list because some abundant proteins (e.g. actin and tubulin, Table S1) are present but not necessarily concentrated in the plaque, and some low abundance proteins (e.g. collagen XXV and antichymotrysin, Table S1) were detected only in one AD case maybe due to sensitivity limitation of the LC-MS/MS approach. Nevertheless, the list of 26 proteins that were identified and concentrated in the plaque regions in both AD cases are more likely to be genuine plaque proteins.

Classification of identified plaque proteins
To evaluate the proteins identified in the plaques from AD postmortem samples, we group them under functional categories in alphabetical order (Table 1 and Table S1) and discuss the implication of some enriched plaque proteins in AD pathogenesis.

Cell adhesion, cytoskeleton and membrane trafficking
A number of intercellular adhesion molecules have been shown to be localized in amyloid plaques in AD patients (40). We found that collagen I and fibrinogen were concentrated in the plaques. In addition to extracellular structures, integrity of the intracellular cytoskeleton is important for neuronal physiological functions such as axoplasmic flow of essential synaptic components (41). In our study, all major isoforms of cytoskeletal components, actin, tubulin and neurofilament, were identified with high numbers of peptides, which indicates that they are abundant species in the plaques as well as in non-plaque regions. Interestingly, we identified one actin binding proteins, coronin, is concentrated in the plaque, although its potential role in pathogenesis remains unclear.
Also enriched in plaques is a microtubule-associated protein tau that is more commonly known to be associated with neurofibrillary tangles and neurites (42), and likely represents dystrophic neurite component in the plaques. The identification of a variety of cytoskeletal protein elements, some of which are known to be relevant to AD, suggests that these elements may be involved in the plaque formation and cytoskeletal impairments may lead to deficit in axoplasmic flow and eventually to neuritic dystrophy.
Indeed, numerous proteins involved in membrane trafficking and protein sorting were revealed to be concentrated in the plaques by the mass spectrometry analysis, such as clathrin heavy chain, dynamin and dynein heavy chain (Table 1). This observation implicates that the AD plaque might sequester some key proteins to perturb the protein sorting system that is crucial for maintaining normal synaptic plasticity.

Chaperones and inflammation
It is well known that the senile plaque core is surrounded by activated astrocytes, microglia and dystrophic neurites (43,44), and the heat shock proteins exhibit high expression levels in reactive astrocytes in areas rich in senile plaques (45). Consistently, we identified many heat shock proteins (Table S1) and found that HSP90 was enriched in the plaques (Table 1). We also identified glial fibrillary acidic protein (GFAP) and vimentin with high numbers of peptides; both are the major components of intermediate filaments in activated glial cells.

Kinase/phosphatase and regulators
The imbalance of phosphorylation/dephosphorylation activity is believed to contribute to AD pathology, as evidenced by tau hyperphosphorylation in the neurofibrillary tangles.
We identified multiple kinases (Table S1) but none of them were specifically enriched in the plaque regions. Instead, three 14-3-3 isoforms showed a significant degree of enrichment in the isolated plaques. Previous studies have demonstrated that 14-3-3 proteins are present in neurofibrillary tangles (46), and at least one 14-3-3 protein has further been shown to be an effector of tau phosphorylation (47). Moreover, we attempted to detect protein phosphorylation sites in the plaque samples by tandem mass spectrometry and located two phosphorylated amino acid residues in neurofilament 3 (SPVPKSPVEEK and KAESPVKEEAVAEVVTITK with modified sites shown bolded and italicized). The first phosphorylation site was documented in tandem repeats in the neurofilament sequence and was excessively modified in the AD brain (48). The second peptide indicated a phosphorylation site on serine 736 that we show for the first time.
These phosphorylation events may play a role in the formation of dystrophic neurites surrounding the plaque core.

Proteolysis
The ubiquitin-proteasome system plays a crucial role in the degradation of misfolded proteins and turnover of cell signaling molecules (49). This study identified ubiquitinactivating enzyme E1 enriched in the plaques. More strikingly, numerous subunits of lysosomal ATPase and cathepsin D were found to be concentrated in the plaques, suggesting the high proteolytic activity in the plaques versus non-plaque regions.
Moreover, we found in the plaques antitrypsin, cystatin B and cystatin C. Cystain C is a cysteine proteinase inhibitor and has been shown to be upregulated in AD and coaggregated with A (50,51). Cystain C is also proposed as a potential risk factor for lateonset AD (52). Overall, the relative abundance of proteolytic enzymes and inhibitors in multiple cellular proteolytic pathways strongly suggests the activation of protein degradation mechanisms and the interplay between proteolysis and inhibition activities during the plaque formation.

Validation of selected plaque components
To further verify the localization of some of the proteins identified, we undertook immunostaining analysis on postmortem brain tissues for their colocalization with A . In figure 4A, monoclonal antibody against vimentin specifically labeled activated astrocytes surrounding the fibrillary plaque core that was recognized by A antibody; and some astrocyte processes were extended deeply into the plaque core, indicating a local inflammatory response in the plaque region (53,54). Polyclonal antibodies against the Hsp70 and Hsp90 complex show strong immunoreactive signal in dot-like structure located in the plaque regions ( Figure 4B), consistent with our proteomic analysis. More strikingly, dynein heavy chain was detected as punctate and thread-like structures in many plaques labeled by A antibody ( Figure 4C), suggesting that dynein heavy chain is a constituent of plaque neurites. Confocal microscopic images clearly indicated that dynein heavy chain was enriched in the plaque, compared with the surrounding neuropil ( Figure 4D).
The colocalization of dynein with amyloid plaque was corroborated in an AD model, PS1/APP double transgenic mice (36,37). As expected, no plaque was visualized in 9-month-old control animals and dynein heavy chain antibodies labeled the soma and neurites in the cortex and hippocampus ( Figure 4E). In great contrast, the similar brain regions in 9-month-old AD mice manifested high density of amyloid plaques, the majority of which are clearly surrounded by dynein immunoreactive structures that may represent enlarged neurites ( Figure 4F). Recently, missense mutations in dynein heavy chain have been genetically linked to progressive motor neuron degeneration and the formation of Lewy body-like inclusions (55). It has also been observed previously that dynein immunoreactivity in AD brain tissue is significantly increased compared with normal brain (56). Given that dynein is responsible for retrograde transport of vesicles along microtubules in the axon (57), the enrichment of dynein encircling the plaque core might trigger trafficking malfunction to cause neuritic dystrophy.

Discussion
Our studies revealed a total of 488 proteins in amyloid plaques, representing three histopathological components: the plaque core, activated glial cells, and dystrophic neurites. Identified intracellular proteins are likely derived from neurites and/or glial cells surrounding the plaque core, while extracellular proteins may be components of the plaque core. It is also possible that some intracellular proteins can leak out after plasma membrane damage during neuronal degeneration. By quantitative mass spectrometry, we further identified 26 proteins enriched in the plaques when comparing with the nonplaque control sample from the same AD case. This comparison is particularly valuable because both samples were derived from the same postmortem tissue specimen, which essentially eliminates the effects of many confounding factors (e.g. genetic variation, postmortem interval, etc.) that are often encountered in human tissue studies.
Our results confirmed the presence of most proteins that were previously detected in the plaques mainly by immunohistochemistry (40). The classic plaque components identified in this study include A , 1-antichymotrypsin (58), apolipoprotein E (59), collagen type XXV (60), cystatin C (61), -synuclein (62), proteoglycans (63), and clusterin (64). On the other hand, only a few proteins known to be amyloid plaque components were missed in this study. Those included complement inhibitors (40), myeloperoxidase (5), 2-macroglobin (65), SOD (66), HO-1 (67), catalase (68), and cholineseterase (69). The apparent absence of these proteins in the plaques in our study could be due to the low abundance of the proteins, and/or incompatibility of tryptic peptides with the LC-MS/MS system (12). Our results from two independent studies of AD cases revealed that about two-thirds of the proteins were identified in both cases. The difference between the two datasets may be contributed by the sample variation in the patients and the nature of shotgun proteomics strategy, as only a fraction of peptides was selected and sequenced by mass spectrometry when a complex peptide mixture was analyzed by the LC-MS/MS approach (29).
Immunohistochemistry was employed as an independent approach to confirm the proteomic findings. Eight proteins (in Table 1 and Table S1)  A is widely accepted as the major component of amyloid plaque core, consistently, our mass spectrometry analysis identified A in both plaque and non-plaque Among the two phosphorylated peptides identified in neurofilament 3 (medium chain), one peptide displays the repeated KSPV motif that is homologous to neurofilament heavy chain and tau (70). Hyperphosphorylation of neurofilament as well as tau may lead to abnormal microtubule network assembly and disruption of vesicle transport (71,72), potentially resulting in neuritic degeneration around the pathological plaque structure. We failed to find with confidence phospho-peptides derived from other proteins such as tau, because in LC-MS/MS analysis, only peptides with strong signals were selected for sequencing; thus, the majority of peptides generated from a complex mixture were missed by the mass spectrometer. To gain a more complete view of the phosphorylation events in the plaques, specific enrichment of the modified forms of proteins/peptides will be required.
The combination of laser capture microdissection with LC-MS/MS provides a general method integrating a cellular staining approach with biochemical protein analysis, which permits the direct sequencing of proteins present in a specific microscopic region with high sensitivity, as evidenced by the identification of several hundreds of proteins from less than five micrograms of total plaque proteins. By applying different staining methods that specifically label other types of plaques such as diffuse and primitive plaques, this methodology can be further extended to determine proteins involved in early stages of aggregation, and possibly illustrate the molecular events that initiate the plaque formation. This approach can also be applied to study plaque evolution in transgenic mouse model. Furthermore, it is possible to use this technology for the analysis of other pathological structures such as Lewy bodies in Parkinson disease, protein inclusions in Huntington's disease, or ubiquitin-positive inclusions in frontotemporal dementia. On the other hand, the LC-MS/MS approach itself has been used as a primary tool to allow highly sensitive protein identification, and to provide protein quantification information by integrating the extracted ion current of eluted peptides. More accurate quantification of proteins could be achieved by applying stable isotope labeling-based techniques such as isotope-coded affinity tags (ICAT) strategy (73).
To our knowledge, this is the first large scale analysis of proteins from AD amyloid plaques. The results of this study demonstrate that the protein molecules in amyloid plaques are highly complex and diverse, implicating the involvement of many cellular pathways in disease development. The plaque subproteome identified in our study will be instructive for subsequent hypothesis-driven experiments on disease biomarker identification and molecular genesis of Alzheimer's disease.       The proteins and functional categories were sorted in the alphabetic order. Ctl: the number of different peptides that identify a protein in the control samples of non-plaque regions Plaque: the number of peptides that identify a protein in the plaque samples P/C: the abundance ratio of proteins from the plaque samples versus non-plaque regions When a peptide in the non-plaque control was not detected, we assumed the ratio was more than 2. Sometimes, even if a peptide ion was not sequenced by MS/MS, it was still possible to identify it in MS survey scan to enable quantification (see Methods).