Relative and Absolute Quantification of

Dataset

Relative and Absolute Quantification of Postsynaptic Density Proteome Isolated from Rat Forebrain and Cerebellum*□ S

Dongmei Cheng,a,b Casper C. Hoogenraad,b,c,d,e John Rush,f Elizabeth Ramm,c Max A. Schlager,d,g Duc M. Duong,a Ping Xu,a Sameera R. Wijayawardana,h John Hanfelt,h Terunaga Nakagawa,c Morgan Sheng,c,i and Junmin Peng a,j The postsynaptic density (PSD) of central excitatory synapses is essential for postsynaptic signaling, and its components are heterogeneous among different neuronal subtypes and brain structures. Here we report large scale relative and absolute quantification of proteins in PSDs purified from adult rat forebrain and cerebellum. PSD protein profiles were determined using the cleavable ICAT strategy and LC-MS/MS. A total of 296 proteins were identified and quantified with 43 proteins exhibiting statistically significant abundance change between forebrain and cerebellum, indicating marked molecular heterogeneity of PSDs between different brain regions. Moreover we utilized absolute quantification strategy, in which synthetic isotope-labeled peptides were used as internal standards, to measure the molar abundance of 32 key PSD proteins in forebrain and cerebellum. These data confirm the abundance of calcium/calmodulin-dependent protein kinase II and PSD-95 and reveal unexpected stoichiometric ratios between glutamate receptors, scaffold proteins, and signaling molecules in the PSD. Our data also demonstrate that the absolute quantification method is well suited for targeted quantitative proteomic analysis. Overall this study delineates a crucial molecular difference between forebrain and cerebellar PSDs and provides a quantitative framework for measuring the molecular stoichiometry of the PSD. Molecular & Cellular Proteomics 5:1158 –1170, 2006.

In excitatory synapses of the mammalian brain, the postsynaptic density (PSD)1 is a specialized membrane-assoFrom the aDepartment of Human Genetics, Center for Neurodegenerative Disease, and hDepartment of Biostatistics, Emory University, Atlanta, Georgia 30322, cThe Picower Center for Learning and Memory, RIKEN-MIT Neuroscience Research Center, Howard Hughes Medical Institute, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, dDepartment of Neuroscience, Erasmus Medical Center, P. O. Box 1738, 3000 DR Rotterdam, The Netherlands, and fCell Signaling Technology, Inc., Beverly, Massachusetts 01915 Received, September 6, 2005, and in revised form, January 17, 2006 Published, MCP Papers in Press, February 28, 2006, DOI 10.1074/ mcp.D500009-MCP200 1 The abbreviations used are: PSD, postsynaptic density; AQUA, absolute quantification; SRM, selective reaction monitoring; CaMK,

1158

Molecular & Cellular Proteomics 5.6

ciated structure containing a high concentration of glutamate receptors, cell adhesion molecules, and associated scaffold proteins and signaling enzymes (1–3). Glutamate receptors in the PSD are assembled into large protein complexes by binding to PDZ domain-containing scaffold proteins. In a well characterized example, the cytoplasmic C termini of NR2 subunits of the NMDA-type glutamate receptor interact with the PDZ domains of the PSD-95 family of scaffold proteins, which are highly enriched in the PSD (2). PSD-95 in turn binds to cytoplasmic signaling proteins such as SynGAP, the synaptic GTPase-activating protein (GAP) for Ras/Rap small GTPases (4, 5). Assembly of such protein complexes facilitates specific coupling of postsynaptic receptors to trafficking mechanisms and downstream signaling pathways that control synaptic strength, cytoskeletal rearrangements, and nuclear responses (1). Many if not most of the protein constituents of the PSD are dynamically influenced by synaptic activity via mechanisms such as protein phosphorylation, local translation, ubiquitination, degradation (6, 7), and protein translocation into and out of synapses (8). Altered composition and structural remodeling of the PSD are believed to play critical roles in the formation/elimination and plasticity of synapses. Because it is amenable to biochemical purification and because of its compact size (a few hundred nanometers in diameter and 20 – 40 nm thick), the PSD is a highly suitable “organelle” for proteomic analysis by MS. Recently LC-MS/MS has become a high throughput platform for proteome analysis with the sensitivity in low or even subfemtomole (⬍10⫺15 mol) range (9, 10). Using this platform, we and others have identified several hundred proteins in purified PSD fractions (11–14), revealing the diverse protein composition and unexpected activities in the PSD. calcium/calmodulin-dependent protein kinase; PSD-95, postsynaptic density protein of 95 kDa; NMDA, N-methyl-D-aspartate; AMPA, ␣-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid; GAP, GTPaseactivating protein; SynGAP, synaptic GTPase-activating protein; GluR, glutamate receptor; GKAP, guanylate kinase-associated protein; geLC/LC-MS/MS, one-dimensional SDS gel with two-dimensional LC-MS/MS; SCX, strong cation exchange; IP3R, inositol trisphosphate receptor; mGluR, metabotropic glutamate receptor; SAPAP, SAP90/PSD-95-associated protein.

© 2006 by The American Society for Biochemistry and Molecular Biology, Inc. This paper is available on line at http://www.mcponline.org

Quantification of the PSD

With numerous PSD proteins identified, one critical question is whether the diversity of proteins found represents similar complexity in all PSDs or whether the diversity arises from the heterogeneity of PSDs from different types of neurons and/or different brain regions. In addition, our view of the organization of the PSD remains unsatisfyingly “topological” and schematic (1–3) even for the well studied PSDs from pyramidal neurons of the hippocampus/cortex. A more useful map of the PSD requires much more quantitative information on the stoichiometry of individual proteins and on the threedimensional structure of protein complexes within the PSD. Ultrastructural insights into the architecture of postsynaptic protein complexes are likely to come from electron microscopy, which has been used successfully to image the native AMPA receptor complex (15). The measurement of protein stoichiometry (the number of molecules of specific proteins) in the PSD can be addressed utilizing quantitative proteomic strategies based on stable isotope dilution in MS. In the quantitative MS strategies, protein samples are analyzed with added internal standards that are chemically identical forms of the proteins/peptides of interest containing stable heavy isotopes (e.g. 2H, 13C, 15N, etc.). The labeled internal standards and their native counterparts have the same chemical properties and behave similarly under any conceivable chemical isolation step. Finally they are separated by mass spectrometry, and their intensity ratio provides an accurate measurement of the amount of native peptides/ proteins. Many methods have been independently reported, and they differ in the details of isotope incorporation (e.g. metabolically labeling in vivo or chemically labeling in vitro) and in the analytical procedures used. One of the popular methods is the ICAT strategy (16) utilizing a Cys-based labeling reaction. The strategy has been successfully adopted into many proteomic studies in yeast and mammals (17, 18). The sensitivity and reliability of the method have been further improved by the generation of cleavable ICAT reagents (17, 19). Although the ICAT strategy allows a comparison of the relative quantities of expressed proteins, the absolute quantity of a targeted protein could not be readily measured from a complex mixture until the development of a strategy termed absolute quantification (AQUA) of proteins and modifications (20). In the AQUA method, labeled synthetic peptides are used as internal standards and detected in the mode of selective reaction monitoring (SRM) to improve sensitivity. The AQUA method has been used in several studies to probe phosphorylated proteins (20), membrane proteins (21), and a few key proteins in the PSD (12). A more comprehensive “census” of the PSD protein population would be exceedingly valuable. Here we report a quantitative comparison between the protein makeups of PSDs purified from the forebrain versus the cerebellum using the ICAT strategy: this analysis revealed profound differences in molecular composition in different brain regions. Moreover we extended our AQUA analysis to

measure the absolute molar quantities of 32 core proteins/20 ␮g of total PSD proteins, again comparing these values between forebrain and cerebellar PSD fractions. EXPERIMENTAL PROCEDURES

Preparation of Postsynaptic Density from Rat Brain—PSD fractions were prepared from rat forebrains and cerebellums at 4 °C (22). The brain structures were collected from adult rat and homogenized in ice-cold Buffer A (5 mM HEPES (pH 7.4), 1 mM MgCl2, 0.5 mM CaCl2, phosphatase inhibitors (1 mM NaF and 1 mM ␤-glycerophosphate), and protease inhibitors (0.1 mM PMSF, 1 ␮g/ml aprotinin, 1 ␮g/ml leupeptin, 1 mM benzamidine, and 0.1 mM pepstatin)) with a Teflon homogenizer (12 strokes). The resulting extract was centrifuged at low speed (1,400 ⫻ g for 10 min) to collect the first supernatant (S1). The pellet (P1) was re-extracted with the homogenizer (five strokes) and centrifuged at 700 ⫻ g for 10 min. The supernatant (S1⬘) was pooled with S1 followed by high speed centrifugation at 13,800 ⫻ g for 10 min. The supernatant (S2) was removed, and the pellet (P2) was resuspended in Buffer B (0.32 M sucrose and 6 mM Tris (pH 8.0) with the same inhibitors of phosphatase and proteases) by a Teflon homogenizer (five strokes), loaded onto a discontinuous sucrose gradient (0.85/1/1.15 M in 6 mM Tris (pH 8.0)), and centrifuged at 82,500 ⫻ g for 2 h. The synaptosome fraction (Syn) between 1 M and 1.15 M sucrose was collected and adjusted to 4 ml with Buffer B. An equal volume of Buffer C (6 mM Tris (pH 8.1) and 1% Triton X-100) was added, mixed for 15 min, and centrifuged at 32,800 ⫻ g for 20 min. The resulting pellet (PSDI) was extracted again with Buffer D (6 mM Tris (pH 8.1) and 0.5% Triton X-100) for 15 min and centrifuged again at 201,800 ⫻ g for 1 h to obtain a pellet (PSDII) for proteomic analysis. PSD proteins were dissolved in 50 mM Tris (pH 8.5) with 1.0% SDS at 95 °C for 5 min. The protein concentration was determined by Pierce BCA protein assay using BSA as standard and was further confirmed by staining PSD samples loaded side by side on an SDS gel (Fig. 1A). Antibodies—NR1 (BD Biosciences), GluR␦ (Santa Cruz Biotechnology, Inc.), GluR2/3 (Chemicon), GluR1 (Oncogene Research), calcium/calmodulin-dependent protein kinase II␣ (CaMKII␣) (Sigma), ␣-tubulin (Sigma), PSD-95 (Upstate Biotechnology), Nir2 (23, 24), Septin4 (25), and homemade antibodies against Homer, GKAP, and Shank. Immunocytochemistry—Sprague-Dawley rats were anesthetized with sodium pentobarbital intraperitoneally (75 mg/kg) and perfused with 4% paraformaldehyde in 0.12 M phosphate buffer. The brains were cryoprotected in 0.1 M phosphate buffer containing 30% sucrose and cut into 40-␮m-thick transverse sections on a freezing microtome. All sections were collected in 50 mM TBS (pH 7.6) with 0.5% Triton X-100, blocked in 10% normal horse serum in the same buffer for 4 h, and then incubated for 48 h at 4 °C with primary antibodies in 2% normal horse serum in 50 mM TBS (pH 7.6) with 0.5% Triton X-100. After incubation, the sections were rinsed and processed with the use of the avidin-biotin complex method (ABC Elite kit, Vector Laboratories). Finally the sections were incubated in 0.05% 3,3-diaminobenzidrine and 0.01% H2O2 in 50 mM Tris buffer (pH 7.6) and mounted. ICAT Profiling by Combining One-dimensional SDS Gel with Twodimensional LC-MS/MS (geLC/LC-MS/MS)—The experiment was performed according to the manufacturer’s protocol (Applied Biosystems, Framingham, MA) with modifications. PSD proteins were dissolved in 50 mM Tris (pH 8.5) with 1.0% SDS at 95 °C for 5 min and diluted to adjust SDS to 0.2% and protein concentration to 1.0 ␮g/␮l. Then the proteins (⬃100 ␮g) were reduced by the addition of 2 ␮l of Tris(2-carboxyethyl)phosphine hydrochloride and incubation at 95 °C for 10 min. The reduced samples (the PSDs from cerebellum or forebrain) were labeled at 37 °C for 2 h with one vial of cleavable ICAT light or heavy reagent, respectively, which contains four moieties: a thiol-reactive group, a linker of light or heavy isotopes, another acid-


1159


cleavable linker, and a biotin tag. The labeled protein were combined, separated in one lane on a 12% SDS gel, and excised into five pieces (⬍20, 20 –50, 50 – 80, 80 –160, and ⬎160 kDa) followed by in-gel tryptic digestion (26). Cys-labeled peptides were affinity-purified by an avidin cartridge and completely dried in a SpeedVac. The biotin moiety of labeled peptides was then removed using cleaving reagents (Applied Biosystems). The samples were dried again before the analysis by mass spectrometry. The analysis was carried out based upon multidimensional protein identification technology (MudPIT), a methodology of LC/LC-MS/MS (27). The cleaved ICAT-labeled peptides were dissolved in Buffer A (0.4% acetic acid, 0.005% heptafluorobutyric acid, and 5% acetonitrile), loaded onto a three-phase 100-␮m-inner diameter self-packed column (7 cm of 5-␮m C18 reverse phase resin, 2.5 cm of strong cation exchange resin, and 2.5 cm of C18 resin) (28), and analyzed in six cycles. In each cycle, the column was first eluted using a saltcontaining buffer (5% acetonitrile and 1.0% formic acid with the addition of 0, 25, 50, 100, 150, or 500 mM ammonium acetate) for 5 min, washed with Buffer A for 5 min, and then eluted with Buffer B (0.4% acetic acid, 0.005% heptafluorobutyric acid, and 95% acetonitrile) in a three-step gradient (5–10% in 15 min, 10 –25% in 150 min, and 25– 40% in 15 min). The column was further cleaned with 100% Buffer B for 5 min and re-equilibrated with Buffer A for 10 min before the next cycle. Two more cycles with 37.5 and 75 mM ammonium acetate were performed in the analysis of the gel piece (20 –50 kDa). Eluted peptides were ionized and detected in a MS survey scan from 400 –1700 m/z with three microscans followed by one data-dependent MS/MS scan (three microscans each; isolation width, 3 m/z; 35% normalized collision energy; dynamic range of 1 min) on an LCQ-Deca XP-Plus ion trap mass spectrometer (Thermo Finnigan). ICAT Data Processing—Using the SEQUEST algorithm (29), all MS/MS spectra were searched against a rat reference database (ftp.ncbi.nih.gov/refseq; released on May 11, 2005; 23,983 proteins) tagged with a randomized rat database of the same size that enabled the estimation of false-positive matches (30). The parameters were set to allow precursor ion mass tolerance to be 3.0, to consider only b and y ion series, and to use monoisotopic mass type for precursor and product ions. The search was run without enzymatic restriction, and the maximum number of missed cleavages was set to three. The fixed mass of cysteine was added with 227.0 because of ICAT labeling. Dynamic mass changes were permitted to allow for the detection of modified methionine (⫹16 Da, oxidation) and cysteine (⫹9 Da, ICAT differential labeling). We used more stringent SEQUEST criteria than described previously (30, 31) including the following. (i) Non-tryptic peptides or singly charged peptide matches were discarded. (ii) ⌬Cn score was at least 0.08. (iii) The minimum length of peptides was seven. (iv) XCorr cutoff was adjusted to keep the estimated false-positive rate below 1% before removing the peptide redundancy (30), leading to the following thresholds: 2.23 for the fully tryptic, doubly charged peptides; 2.93 for the fully tryptic, triply charged peptides; 3.26 for the partially tryptic, doubly charged peptides; and 4.09 for the partially tryptic, triply charged peptides. After filtering the initial 48,467 assignments and deleting redundant peptides and common contaminants (e.g. trypsin and keratin), 758 peptides were passed including nine peptides (⬃1.2%) matched to randomized sequences, suggesting that about 1.2% of assignments in the original database were false-positives. To further increase the certainty of identification, we manually verified proteins matched by less than three peptides and discarded 30 peptides (⬃4.0%). Therefore the final list of 728 peptides was accepted with very high confidence. If we counted multiple modified peptide forms (oxidation of Met and ICAT labeling of Cys) as a single peptide, the list was reduced to 545 peptides (Supplemental Table S1).

1160


Protein assignment and quantification were based on the identified peptides; this was confounded by the presence of different names and accession numbers for the same protein and by peptides shared by several proteins (e.g. paralogs) (32). To alleviate this problem, we separated peptides into two categories, unique and shared peptides, and grouped together the proteins that had shared peptides. This led to the identification of 443 proteins in 276 groups. Thereafter all proteins were verified manually to remove the redundancy. In general, only one protein was selected to represent one group except that other group members were identified by at least one unique peptide, namely, in each group, we first selected the protein with the highest number of peptide matches, also kept other proteins matched by additional unique peptides, and then removed proteins identified only by shared peptides. Finally we accepted a total of 296 proteins in 276 groups (Supplemental Table S2). Quantification of Cys-containing peptides was carried out with the XPRESS software (18). For seven proteins (2%), only one peptide of the labeled pair was detected, and the other was too weak to be quantified accurately; we treated the peptide abundance ratio as 10-fold. Another nine proteins (3%) were identified only by non-Cyscontaining peptides and could not be quantified. In addition, a significant portion of peptides (37%) were identified multiple times because they split in chromatographic fractions (e.g. in consecutive strong cation exchange (SCX) factions), and/or their different charge states were analyzed by MS/MS independently, allowing the acquisition of multiple data points for a single peptide. Statistical Analysis of ICAT Data—Statistical evaluation was performed in several steps according to the methods proposed previously by Li et al. (33) with modifications on the calculation of variance. (i) All abundance ratios for peptides obtained in the ICAT experiment were transformed into log2(ratio). (ii) The mean and associated variance of log2(ratio) was calculated for every peptide quantified multiple times (Supplemental Table S1). (iii) The mean and variance of log2(ratio) was derived for each protein based on either all assigned peptide(s) or only on unique peptide(s) that are not shared among proteins. Regarding proteins identified by a single peptide, variance of the peptide if available was directly transferred to the corresponding protein. (iv) We assigned variance for all proteins using a shrinkage estimator that is based on an empirical Bayes approach with inverse Gamma prior (34). A shrinkage estimator is often used for analyzing variance of individual measurement (i.e. log2(ratio) for each protein) according to common variance of the whole dataset. In our study this was desirable because data points for most of the proteins were relatively small.

␴˜ 2i ⫽ 共1 ⫺ ␻i兲␴i2 ⫹ ␻i␸

(Eq. 1)

n n ␸ ⫽ 关兺i⫽1 共ni ⫺ 1兲␴2i 兴/关兺i⫽1 共ni ⫺ 1兲兴

(Eq. 2)

␻i ⫽ ␯/共␯ ⫹ ni ⫺ 1兲

(Eq. 3)

n n ␯ ⫽ 4 ⫹ 关2␸2兺i⫽1 共ni ⫺ 1兲兴/关兺i⫽1 共ni ⫺ 1兲共␴2i ⫺ ␸兲2兴

(Eq. 4)

where ␴˜ i is derived variance for each protein weighted by its number of measurements (i.e. peptides) (ni), ␴i2 is original variance for the protein, ␸ is an estimator of the common variance given by a weighted average of original variances, ␻i is a weighting parameter varying from 0 to 1, and ␯ ranges from 4 to infinity. (v) The experimental log2(ratio) was normalized to reduce systematic errors in sample handling. Because the majority of proteins appeared to display highly similar abundance in forebrain and cerebellar PSDs (Fig. 1A), the log2(ratio) data of proteins were fitted with a normal distribution on the basis of central limit theorem, 2

np共x, A, ␮, ␴0兲 ⫽ A ⫻ exp关⫺共x ⫺ ␮兲2/2␴02兴

(Eq. 5)


where exp is the exponential function, np is the number of proteins with certain log2(ratio) assigned by x, A is a constant, and ␮ and ␴02 are the population mean and its variance, respectively (Fig. 3). The log2(ratio) for each protein was then normalized by subtracting the fitted mean (␮) as the mean should have been 0 if sample preparation was ideal. (vi) The probability (p) value that the a protein remained unchanged was calculated by

p ⫽ erfc兵兩y兩/sqrt关2共␴˜ 2i ⫹ ␴02兲兴其

(Eq. 6)

where erfc is the complementary error function, y is the normalized log2(ratio) of each protein, and sqrt is the function of square root. Finally the proteins were organized according to their molecular or functional categories as described previously (Supplemental Table S2) (12). Quantification of Selected Proteins by the AQUA Strategy—AQUA peptides for 32 proteins of interest were selected based on the following rules: (i) to ensure their uniqueness, namely not being shared in other proteins in one species; (ii) to be conserved in as many species as possible to allow their cross-species application; (iii) to examine whether they are present in splicing variants; (iv) to select peptides compatible with the LC-MS/MS setting (i.e. with reasonable hydrophobicity and ionization efficiency); (v) to include amino acid residues such as Leu, Val, Phe, Pro, Gly, or Ala because these heavy isotope-labeled residues are commercially available and affordable (Cambridge Isotope Laboratories, Inc.); (vi) to avoid peptides with labile amino acids if possible, for example, Met, Cys, and Trp are likely to be oxidized; Asp-Pro and Asp-Gly peptide bonds are unstable; Asn and Gln are subjected to deamidation; and N-terminal Gln can undergo cyclization; and (vii) to eliminate peptides with flanking sequences that inhibit proteolysis. All AQUA peptides were synthesized, purified (generally ⬎90% pure), and quantified by amino acid analysis (twice, by Cell Signaling Technology, Inc., Beverly, MA). An equal amount of forebrain and cerebellar PSD fractions (⬃20 ␮g) was resolved on a 9% SDS gel. Each lane was excised into two pieces (less than 80 kDa and at least 80 kDa) followed by in-gel digestion (25 ng/␮l trypsin with ⬃1:2 trypsin/substrate ratio). Isotopically labeled AQUA peptides (0.2 pmol each except 2 pmol of CaMKII␣ and CaMKII␤) were implemented and soaked into the dried gel pieces with trypsin. The peptide mixtures were then analyzed by capillary reverse phase chromatography combined with tandem mass spectrometry using an LCQ-Deca XP-Plus ion trap mass spectrometer or an LTQ linear ion trap instrument (Thermo Finnigan). Every sample was analyzed at least three times, and ⬃10% of the sample (equivalent to ⬃2 ␮g of total PSD proteins) was loaded onto the column each time. The instrument was operated in the SRM mode; see Supplemental Table S3 for detailed conditions. The quantification analysis was carried out using Xcalibur software (Thermo Finnigan). The analysis of immunopurified AMPA receptor complex (15) was performed in a similar manner. RESULTS AND DISCUSSION

Comparison of Cortical and Cerebellar PSDs Using Western Blotting and the ICAT Strategy The protein composition of PSDs (PSDII fractions) purified from cerebellum and forebrain was first compared by SDSPAGE and silver staining (Fig. 1A). Although there were some differences in band intensities, the overall pattern of major bands looked highly similar between forebrain and cerebellum. Western blotting of glutamate receptors showed that NMDA receptors (NR1, NR2A, and NR2B) were enriched in PSDs compared with synaptosomal fractions and much more

abundant in forebrain than in cerebellar PSDs (Fig. 1B). By contrast, GluR␦ subunit was found and highly enriched only in cerebellar PSDs as reported previously (35). Different from NMDA receptors, AMPA receptors (GluR1 and GluR2/3) were more evenly distributed between forebrain and cerebellum PSDs. Several other major PSD proteins were also detected by Western blotting. CaMKII␣, a serine/threonine kinase, and PSD-95, an NMDA receptor scaffolding protein, were highly abundant in the PSDs, as expected, but much more abundant in forebrain than cerebellum. The PSD scaffolds GKAP and Homer were concentrated in both forebrain and cerebellum PSD fractions. The scaffold protein calmodulin-associated serine/threonine kinase showed only a weak enrichment in the PSDs and like ␣-tubulin was approximately equally distributed between forebrain and cerebellum (Fig. 1B). These results indicate that our PSD preparations are valid and display the expected differential expression of known forebrain versus cerebellar proteins. We also assessed by Western blotting the level of enrichment of PSD-95, CaMKII␣, and GluR␦ in the preparation of PSD fractions from forebrain and cerebellum (Fig. 1, C and D). In linear titration experiments, PSD-95 and CaMKII␣ immunoblot signals showed a similar degree of enrichment (PSDII/S1 ratio) in forebrain versus cerebellum PSDs even though both proteins were much more highly expressed in forebrain (Fig. 1). Interestingly GluR␦ accumulated in cerebellar PSDs to a degree similar to that of CaMKII␣ (mean PSDII/S1 ratio, ⬃18) in three independent experiments (Fig. 1E). Although there are some variations in the degree of enrichment of specific proteins among different experiments, the general consistency within each experiment (Fig. 1E) indicates that we can quantitatively compare PSD proteins isolated from various brain regions in the same experiment. For subsequent quantitative analysis, we used the PSD fractions purified in experiment 3. ICAT-labeled PSD fractions were analyzed by LC-MS/MS to quantify the relative amounts of proteins in cerebellar versus forebrain PSDs (Fig. 2). Thousands of MS/MS spectra were acquired and analyzed by database searching and stringent data filtering. Finally we accepted 545 peptides (Supplemental Table S1) of which the vast majority (88%) were labeled with ICAT reagents, indicating that labeled peptides were isolated at high efficiency. Because some peptides are shared among proteins (32), we clustered the proteins with shared peptide(s) into one group. Thus all identified peptides were assigned to 296 proteins in 276 groups, and the proteins were further classified under functional categories as described previously (12) (Supplemental Table S2). Almost all PSD proteins were identified in both forebrain (281 of 296) and cerebellum (282 of 296) preparations. As protein quantification was based upon the ratio of light and heavy labeled Cys-containing peptides (Fig. 2D), we quantified 287 proteins and failed to analyze the other nine proteins that were only identified by non-Cys-containing peptides.


1161


FIG. 1. Analysis of PSDII fractions from cerebellum and forebrain. A, comparison of PSDII fractions purified from forebrain (F) and cerebellum (C) by SDS-PAGE and silver staining at various protein concentrations. Molecular mass markers are shown on the left (kDa). B, enrichment of PSD proteins by subcellular fractionation. Forebrain and cerebellum from adult rats were homogenized, fractionated by differential centrifugation, and analyzed by Western blotting. S1, the first supernatant after low speed spin to remove nuclei and cell debris; S2, cytosol plus light membranes; P2, crude synaptosomal fraction; Syn, synaptosomal fraction after sucrose gradient; and PSDII, the PSD fractions prepared by double Triton X-100 extraction. C, comparison of S1 and PSDII fractions purified from forebrain and cerebellum by Western blotting. D, quantification of CaMKII␣ immunoblot signals in S1 and PSDII fractions from forebrain that allowed the calculation of the -fold enrichment (PSDII/S1 ratio). E, comparison of the -fold enrichment of three proteins in three independent experiments.

To evaluate the significance of quantitative data acquired for every protein in the ICAT analysis, we used the statistical approaches recently developed by Li et al. (33) to derive probability values. Logarithm transformation of abundance ratios was first carried out to convert multiplicative error into additive error (36). As most of the proteins did not change in the two PSD samples (Fig. 1A), we applied null hypothesis and fitted the experimental data to a normal distribution (Fig. 3). Most of the protein data were fitted very well, but a portion of proteins that exhibited larger changes obviously were not fitted to the curve, indicative of the biological difference between cerebellar and forebrain PSDs. Interestingly the fitted population mean of normal distribution can be utilized to correct systematic errors introduced in sample handling (e.g. starting with a different amount of PSD proteins in ICAT labeling). In this case, the mean value of log2(c/f) (cerebellum/ forebrain ratio) is 0.362, equivalent to 1.29-fold, suggesting

1162


that the initial labeled sample of cerebellar PSD may be 29% more than expected and this error could result from the measurement of total protein concentration and/or the variation during isotopic labeling. Moreover p values were calculated based on how well the data of each protein was fitted to the normal distribution, considering both the protein abundance ratio and its associated variance. Despite the effects of some factors such as limited sample size and fitting errors, the p values represented a reasonable index to assess whether protein changes identified in the ICAT assay is biologically relevant or is purely caused by experimental variability. Of the 287 proteins that were quantified with this approach, 43 proteins (15%) exhibited p values less than 0.05 (Table I) suggesting significant molecular heterogeneity of PSDs in these brain regions. Because the quantification of proteins was complicated by common peptides shared among related family members, we acquired two sets of data according to their


FIG. 2. Relative quantification of the PSD proteome isolated from forebrain and cerebellum by the ICAT strategy. A, flow chart for the ICAT analysis using one-dimensional SDS gel and two-dimensional LC/LC-MS/MS. B, a representative run of LC/LC-MS/MS. Labeled PSD peptides from one of the gel pieces (80 –160 kDa) were loaded on an SCX column, eluted in six steps using different salt concentrations of ammonium acetate, and analyzed by reverse phase chromatography in a 120-min gradient. The signal of the strongest base peak is also shown in every run. C, total peptides identified in all gel pieces and SCX elution steps. D, quantification of NMDA receptor 2B (NR2B) based on one of its identification peptides (DSVSGGGPCTNR with monoisotopic mass of 1385.5 Da). The ion current intensities of doubly charged heavy and light labeled peptides (m/z 693.8 and 689.3) are shown in solid and dashed lines, respectively.

FIG. 3. The distribution of abundance ratio of identified proteins in forebrain and cerebellar PSDs. The PSD protein ratios between cerebellum and forebrain (c/f) were converted into logarithm values (base 2), ranging from ⫺3.5 to 3.5. The range was split into windows of 0.5 units in which proteins were counted and plotted in the solid line to fit to a normal distribution using the equation under “Experimental Procedures” and parameters shown.

total assigned peptides and their unique peptides (Supplemental Table S2). In some cases (e.g. plakoglobin), the two data sets were not completely consistent with each other, reflecting the fact that shared peptides represent the total

amount of all assigned proteins, whereas the unique peptides are generated from single protein, and their derived data should be more accurate. In addition to numerous proteins previously documented to be present in the PSD fraction (11–14), we identified a set of novel PSD proteins, the majority of which are selectively present in the cerebellar PSDs, e.g. CaMKIV (37), inositol trisphosphate receptor (IP3R) (38), phosphatidyl inositol transfer protein Nir2 (39), and G-protein signaling regulator RGS8 (40). Focusing on the 43 proteins displaying p values less than 0.05, we found that these proteins are implicated in a broad range of cellular activities and that forebrain and cerebellar PSDs often contain different members within a protein family (Table I). For instance, glutamate receptor NMDAR2B (NR2B) was found to accrue in forebrain PSDs, whereas GluR␦ was specifically enriched in cerebellar PSDs. However, most of the PSD proteins with significant difference have no specific functional homologue in the other brain region (Table I). Several GTPases and regulators were found selectively in cerebellar PSDs, whereas only one was preferentially present in forebrain PSDs. We also noted that a considerable number of proteins selectively present in cerebellar PSDs attribute to the IP3R and protein kinase C signaling pathway (Table I), such as IP3R (38) and diacylglycerol kinase ␨ (41). To verify the differential expression of some PSD proteins, we performed immunostaining on sagittal sections of adult rat


1163


TABLE I Protein difference in forebrain and cerebellar PSDs measured by ICAT Proteins are sorted by categories and then by protein names. log2(c/f) is the log ratio between cerebellar and forebrain PSDs based on total peptides. p values are the probability values based on all matched peptides. log2(c/f)

p values

Densin-180

⫺3.0

0.004

Actin-binding LIM protein 1 Adducin ␣ Adducin ␤ Ankyrin erythroid Ankyrin repeat-containing protein Band 4.1 erythrocyte

2.8 2.6 2.4 2.9 ⫺2.5 2.9

0.006 0.011 0.020 0.004 0.016 0.005

3.0

0.004

GEFII, cAMP-regulated Rab6-interacting protein 1 Rab6-interacting protein 2 RGS8 SynGAP

2.2 2.3 2.9 2.8 ⫺2.5

0.033 0.026 0.005 0.006 0.043

CaMKIV Casein kinase 1␦ Diacylglycerol kinase ␨ Inositol-polyphosphate 5-phosphatase A Phosphatase 2 regulatory subunit A

2.4 ⫺3.7 2.9 3.0 ⫺2.2

0.020 0.000 0.006 0.004 0.032

2.5

0.015

Dihydropyrimidinase-like 3 Glutamate-oxaloacetate transaminase 2

⫺2.2 ⫺2.4

0.032 0.017

Mitochondrial F0 complex, subunit d

⫺2.3

0.027

2.9

0.005

ATPase family, AAA domain-containing Cajalin 2, E2a-Pbx1-associated protein Calmin ␦ Familial cylindromatosis CYLD F-box and leucine-rich repeat protein 16 FXYD domain-containing ion transport regulator 5 G2 protein Hypothetical protein LOC305021 KIAA1045 protein NDRG family member 2 Nir2, phosphatidylinositol transfer protein Signal peptidase complex subunit 2 Visinin-like Ca2⫹-binding protein type 3

3.0 ⫺2.8 2.5 ⫺2.5 ⫺2.7 3.0 ⫺2.2 3.0 ⫺3.0 2.7 3.0 2.7 2.4

0.004 0.010 0.015 0.017 0.009 0.004 0.036 0.004 0.004 0.008 0.004 0.009 0.021

GluR␦, glutamate receptor ␦ Inositol 1,4,5-trisphosphate receptor mGluR1 NMDAR2B

2.2 2.2 2.2 ⫺2.5

0.032 0.040 0.034 0.020

Shank3

⫺3.3

0.001

2.2 2.9

0.038 0.005

Accession no. Cell adhesion NP_476483.1 Cytoskeleton Actin XP_217645.3 NP_058686.1 NP_036623.1 XP_240464.3 XP_221028.3 XP_232771.3 Others NP_001011893.1 GTPases and regulators XP_215985.3 XP_578521.1 NP_740769.2 NP_062217.1 NP_851606.1 Kinases/phosphatases and regulators NP_036859.1 NP_620691.1 NP_112405.1 XP_344972.2 NP_476481.1 Membrane trafficking NP_671479.1 Metabolism NP_037066.1 NP_037309.1 Mitochondria NP_062256.1 Motor proteins XP_236444.3 Others XP_220076.3 XP_235052.3 XP_234500.3 XP_341653.2 NP_001009504.1 NP_068709.1 XP_230346.3 NP_001014007.1 XP_575808.1 NP_598267.1 NP_001008370.1 XP_214994.2 NP_059052.1 Receptors and channels NP_077355.1 NP_001007236.1 NP_058707.1 NP_036706.1 Scaffolds NP_067708.1 Translation XP_212832.2 XP_574243.1

1164

Protein names

Septin 4

SNAP25-interacting protein 30

Myosin VI

Ribosomal protein L10 Ribosomal protein S16



(cerebral cortex, striatum, and hippocampus), and the enrichment level of CaMKII␣ in forebrain was not as striking as that for the other proteins tested in cerebellum (Fig. 4), consistent with the ICAT quantitative data. Furthermore we used another quantitative mass spectrometric approach to confirm much of the ICAT data in the following sections.

Absolute Quantification of 32 Core Components in Forebrain PSDs

FIG. 4. Immunohistochemistry staining of forebrain and cerebellum enriched PSD proteins in adult rat brain. Rat sagittal brain sections were immunostained for CaMKII␣, Nir2, Septin 4, and GluR␦.

brain (Fig. 4). As shown in Table I, Nir2, Septin 4, and GluR␦ were much more abundant in cerebellar PSDs with log2(c/f) larger than 2.5, equivalent to ⬃6-fold. As expected, Nir2, Septin 4, and GluR␦ were selectively expressed in the molecular layer of the cerebellum, thus corroborating our quantitative proteomic data. Previously uncharacterized in the PSD, Nir2 is a phosphatidylinositol transfer protein involved in maintenance of diacylglycerol levels and regulation of Rho GTPase, actin cytoskeleton, and cell morphogenesis (23, 24). Thus Nir2 could be involved in PSD signaling or spine remodeling in the cerebellum. Despite frequent utilization for statistical tests, the 95% confidence level (p ⫽ 0.05) may exclude some interesting proteins with biological relevance, especially for current large scale proteomic analyses that require significant input of time and resources and are normally not repeated, resulting in limited data points for statistical analysis for most proteins. However, many proteins with a slightly lower confidence level may be verified by independent approaches. We selected CaMKII␣ that was more abundant in forebrain PSDs with log2(c/f) of ⫺2.1, equivalent to an ⬃4-fold decrease in cerebellum (p ⫽ 0.056). Immunostaining experiments indicated that CaMKII␣ showed preferential staining in the forebrain

Because the absolute quantity of proteins cannot be measured by ICAT, a complementary strategy termed AQUA has been developed to quantify in absolute molar terms any designated protein in a sample mixture (20). This targeted proteomic approach involves the addition of internal standards of synthetic peptides or recombinant proteins, which are isotopically labeled with heavier masses. To examine the linearity of the AQUA method using our instrumentation setting, a pair of synthetic light and heavy peptides was mixed at various concentration ratios over a range of 3 orders of magnitude (0.013–13). The result demonstrated remarkable agreement between the known levels and the measured ratios (Fig. 5, A and B). We then quantified 32 major PSD proteins using the AQUA method (Table II). In the case of CaMKII␣, the native and isotope-labeled peptides were eluted at the same retention time during reverse phase chromatography (Fig. 5C). In the scan of SRM (Fig. 5D), the native product ion was shown as two strong peaks and one weak peak around m/z 273 due to naturally occurring isotopes, whereas the labeled internal standard was indicated by several peaks around m/z 279. Therefore two m/z ranges (271–275 and 276 –280) permitted the quantification of both native and labeled ions, respectively (Fig. 5C). CaMKII—Of the 32 proteins measured by AQUA in the rat forebrain PSDs, CaMKII␣ (27.8 pmol/20 ␮g of PSD protein) and CaMKII␤ (4.7 pmol/20 ␮g) were by far the most abundant in molar terms. By mass percentage, CaMKII␣ and CaMKII␤ constituted ⬃7.4 and 1.3% of the PSD, respectively. Our numbers are largely consistent with the literature, which has long recognized CaMKII␣ as the most abundant protein of the PSD (42). The amount of CaMKII␣/␤ associated with PSD preparations is known to increase substantially within minutes following decapitation of the rat, probably by translocation of the enzyme from the cytosolic fraction (43). Much of the CaMKII may exist as contaminating “clusters” (⬃100-nm diameter) that copurify with actual PSDs (44), thereby giving rise to overestimations of CaMKII content in the PSDs, especially following hypoxic conditions. The function and regulation of CaMKII are critical for synapse development and plasticity (45, 46). It remains a mystery why a catalytic enzyme would be present at such high molar amounts in the PSD. SynGAP—SynGAP, the synaptic RasGAP, is well known to be abundant in the forebrain PSDs (4, 5). However, we were surprised that SynGAP (⬃2.1 pmol/20 ␮g) was about as abun-


1165


FIG. 5. Quantitative analysis of 32 PSD proteins by the AQUA strategy. A and B, linearity of the quantitative method. A pair of synthetic light (L) and heavy (H) peptides (TLSDYNIQK phosphorylated at Ser and labeled on Leu) was mixed at various ratios, 0.013, 0.043, 0.13, 1.3 and 13, and then measured by LC-SRM. A linear equation was used to calculate the least squares fit for the line. The left lower corner in A was enlarged and is shown in B. C, LC-SRM of native and labeled CaMKII␣ peptides (ITQYLDAGGIPR) in the mixture of digested PSD proteins. Both peptides were eluted at the exact same time and peaked at 15.5 min. The areas and signal-to-noise ratios (S/N) of the peaks are indicated. D, the average of SRM scans during 15.3–15.6 min in the LC. Although only one product ion was monitored for native and AQUA peptides, several peaks were shown for each product ion due to naturally occurring isotopes.

dant as the scaffold proteins that it binds to, the PSD-95 family (which total ⬃2.3 pmol/20 ␮g). We note that SynGAP is a large protein (⬃135 kDa) with several motifs for proteinprotein interaction; therefore it could have a scaffolding function in addition to its activity as a Ras/Rap GAP. Glutamate Receptors—Although a substantial portion of AMPA receptors are extracted from the PSD by Triton X-100, AMPA receptors remain fairly abundantly represented in our MS quantitation. GluR2 was the most abundant subunit in the forebrain PSD fraction (0.16 pmol/20 ␮g) followed by GluR1 (0.15 pmol/20 ␮g); in comparison, GluR3 was rather meager (0.03 pmol/20 ␮g). GluR4 is reported to be rare in hippocampus (47) and was not quantified by AQUA in this study. In addition, we also quantified by AQUA method the relative stoichiometry of GluR subunits in AMPA receptors immunopurified from rat forebrain (15) (Table II). The ratio among GluR1/GluR2/GluR3 from the PSD (94:100:19) is somewhat similar to that measured from the purified AMPA receptor complex (79:100:13). The slightly higher proportion of GluR2 in the latter preparation might be attributable to the purification of AMPA receptors by anti-GluR2 antibody (15). In both cases, it is notable that GluR2 molar abundance was approximately equal to the sum of GluR1 and GluR3. Our quantita-

1166


tion suggests that the majority of AMPA receptors in forebrain PSDs are GluR1/GluR2 heteromers and only a minority are GluR2/GluR3 heteromers; this is roughly consistent with coimmunoprecipitation studies of AMPA receptors solubilized from CA1/CA2 regions of the hippocampus (47). NMDA receptor subunits were substantially more abundant than AMPA receptors in the PSDII fraction from forebrain (Table II). NMDA receptors are believed to be tetramers comprised of two NR1 and two NR2 subunits. NR1 (0.24 pmol/20 ␮g) was more abundant than NR2A (0.15 pmol/20 ␮g); this is predictable because NR2B-containing receptors are also present in the PSD. Unfortunately we were not able to measure NR2B molar quantity because its synthesized AQUA peptide appeared to be unstable due to deamidation. Metabotropic glutamate receptors (mGluR1 and mGluR5) were relatively rare in the PSD fraction, consistent with their localization in a perisynaptic “ring” outside of the PSD (48). PSD-95—The PDZ domain-containing protein PSD-95 is perhaps the best characterized scaffold protein of the PSD. Befitting its fame, PSD-95 is the most abundant protein (1.73 pmol/20 ␮g) that we measured in the forebrain PSD fraction other than CaMKII␣/␤ and SynGAP. In molar terms, PSD-95 was ⬃6-fold more abundant than PSD-93 (also known as


TABLE II Absolute and relative quantification of the 32 core proteins in PSD fractions Measurements were made in ⬃20 ␮g of forebrain or cerebellar PSD sample. ⬘AMPAR⬘ shows the measurement using immunoisolated AMPA receptors. “Previous” shows our published measurements based on labeled proteins (12). log2(c/f) is the log abundance ratio between cerebellar and forebrain PSDs measured by AQUA. “ICAT” shows the log ratio of cerebellar and forebrain PSDs measured by all assigned peptides in ICAT. ND, not determined. Protein name

Afadin AKAP79/150 CaMKII␣ CaMKII␤ N-cadherin ␤-Catenin GluR1 GluR2 GluR3 Homer1 Homer2 Homer3 IRSp53 mGluR1 mGluR5 NR1 NR2A NR2B PKC ␥ PSD-93 PSD-95 SAP102 SAP97 SAPAP1 SAPAP2 SAPAP3 SAPAP4 Septin7 Shank1 Shank2 Shank3 SynGAP

Forebrain

AMPAR

Previous

pmol

pmol

pmol

0.07 ⫾ 0.02 0.12 ⫾ 0.02 27.8 ⫾ 5.3 4.72 ⫾ 0.21 0.19 ⫾ 0.01 0.26 ⫾ 0.03 0.15 ⫾ 0.01 0.16 ⫾ 0.01 0.03 ⫾ 0.001 0.29 ⫾ 0.04 0.02 0.07 ⫾ 0.01 0.47 ⫾ 0.1 0.05 ⫾ 0.02 0.05 ⫾ 0.003 0.24 ⫾ 0.01 0.15 ⫾ 0.01 ND 0.10 ⫾ 0.01 0.31 ⫾ 0.06 1.73 ⫾ 0.09 0.21 ⫾ 0.04 0.04 ⫾ 0.01 0.35 ⫾ 0.02 0.16 ⫾ 0.01 0.24 ⫾ 0.07 0.07 ⫾ 0.02 0.60 ⫾ 0.06 0.20 ⫾ 0.03 0.35 ⫾ 0.01 0.30 ⫾ 0.05 2.05 ⫾ 0.18

0.048 ⫾ 0.006 0.061 ⫾ 0.015 0.008 ⫾ 0.001

chapsyn-110) and ⬃8-fold more abundant than SAP102 in the PSD fraction. These three members of the PSD-95 family are known to be core components of the PSD, and they interact via their first two PDZ domains with the C termini of NR2 subunits of NMDA receptors (2). It is significant, however, that even by itself PSD-95 outnumbers by ⬎14-fold the abundance of NMDA receptors in the PSD (assuming two NR1 molecules per NMDA receptor). If one also includes SAP102 and PSD-93, then the PSD-95 family is ⬃20-fold more abundant in the PSD fraction than NMDA receptors. These numbers imply that binding to NMDA receptors is only a small portion of the scaffolding function of PSD-95. Other membrane proteins, such as neuroligins, stargazin/transmembrane AMPA receptor-regulatory protein, and ErbB receptors could contribute interacting partners for the PDZ domains of PSD-95 family proteins (2). Moreover the cytoplasmic protein SynGAP was also extremely abundant in the PSD (2.05

0.08 0.19

0.31

1.60

0.35

0.21

log2(c/f)

1.1 ⫺2.3 ⫺1.8 ⫺0.3 0.1 ⫺0.7 ⫺0.5 0.1 ⫺0.1 ⫺0.6 ND 2.8 ⫺1.5 2.0 ⫺2.3 ⫺2.5 ⫺2.7 ND 1.1 ⫺1.3 ⫺1.3 ND ⫺0.5 ⫺0.6 ⫺0.1 ⫺0.7 0.1 0.2 0.6 0.4 ⫺1.6 ⫺1.3

ICAT

⫺2.1 ⫺0.6 ⫺0.8 ⫺1.0 ⫺0.5 ⫺0.7

2.2 ⫺1.9 ⫺2.5 0.7 ⫺1.0 ⫺1.8 ⫺1.0 ⫺1.4

0.0 1.4 ⫺1.4 ⫺3.3 ⫺2.5

pmol/20 ␮g), roughly similar to the molar quantity of PSD-95. Because SynGAP also interacts with the PDZ domains of PSD-95, it is likely to be the major binding partner of PSD-95 in stoichiometric terms in the PSD. The fourth member in this family, SAP97 (0.04 pm/20 ␮g), was an order of magnitude less represented than PSD-93 or SAP102 in keeping with the easy extractability of SAP97 in Triton X-100 (49). Other Scaffolding Proteins—The PSD-95 family scaffolds interact by their guanylate kinase-like domains directly with GKAP/SAPAP proteins, which in turn bind to the Shank family of proteins (2). Shank proteins interact with Homer family proteins (1–3). We used AQUA to systematically measure the amount of these scaffold proteins in the forebrain PSD. In the GKAP/SAPAP family, the order of abundance was GKAP/ SAPAP1 (0.35 pm/20 ␮g) ⬎ SAPAP3 (0.24 pm/20 ␮g) ⬎ SAPAP2 (0.16 pm/20 ␮g) ⬎ SAPAP4 (0.07 pm/20 ␮g). The order of abundance in the Shank family was Shank2 (0.35


1167


pm/20 ␮g) ⬎ Shank3 (0.30 pm/20 ␮g) ⬎ Shank1 (0.20 pm/20 ␮g). Homer1 (0.29 pmol/20 ␮g) was much more abundant than Homer2 (0.02 pm/20 ␮g) and Homer3 (0.07 pm/20 ␮g). Assuming that members of the PSD-95, GKAP/SAPAP, and Shank families have the same biochemical functions within each family, then the relative stoichiometry of these interacting scaffolds in forebrain PSD is roughly as follows: 6 PSD95:2 GKAP/SAPAP:2 Shank:1 Homer. These measurements reveal a diminishing molar abundance of scaffold proteins along the chain of interactions from PSD-95 to Shank and Homer that will constrain models of how these proteins are interlinked in the PSD.

total PSD proteins. One difficulty is that PSDs are highly variable in size, and the second difficulty is that they are likely to be heterogeneous in composition even within a given brain region. Nevertheless estimates of number of molecules per “average” or “typical” PSD would be extremely valuable to the field. Recently scanning electron microscopy has been used to estimate the mass of an average PSD (⬃1.1 GDa for a PSD of 360-nm diameter), and from this value the number of copies of PSD-95 was deduced to be ⬃300 (52). With this number in hand, the copies of other proteins in the PSD can be computed simply by knowledge of the stoichiometric ratio relative to PSD-95. Thus, the information provided here offers a powerful quantitative insight into the PSD proteome.

Comparison with Previous Results Taken together, the numbers obtained in this study (Table II) are in rough agreement with previous AQUA measurements that were performed on a small subset of these proteins in another PSD preparation (12). For instance, for GluR2 we found 0.16 pmol/20 ␮g in this study versus 0.19 pmol/20 ␮g previously; NR1, 0.24 versus 0.31 pmol/20 ␮g; GKAP/SAPAP1, 0.35 versus 0.35 pmol/20 ␮g; PSD-95, 1.73 versus 1.6 pmol/20 ␮g; and Shank1, 0.20 versus 0.21 pmol/20 ␮g. On the other hand, there were larger differences in GluR1 (0.15 versus 0.08 pmol/20 ␮g previously). The difference could be due to inherent impreciseness of the AQUA method, or it could reflect variation in the composition of the PSD as a result of differing degrees of hypoxia during euthanasia of the animal and brain dissection. As the AQUA strategy requires complete digestion of target proteins that is often affected by the selection of digestion conditions (50, 51), we used the extremely stable protein ubiquitin to optimize digestion conditions and found that all tryptic sites except one flanked by two acidic residues in ubiquitin could be reliably cleaved under denaturing in-solution or in-gel conditions with high trypsin/substrate ratio (data not shown). Our selected AQUA peptides are surrounded by no acidic residues and are anticipated to be digested well under the optimal in-gel digestion condition because the results are largely consistent with our previous measurements (12) in which recombinant proteins (rather than synthetic peptides) were used as internal standards for controlling variations in digestion efficiency.

Stoichiometry of Proteins per PSD Rather than per Microgram of Protein The AQUA measurements in this study reveal only the molar quantities of specific proteins per mass of total PSD protein (in this case, per 20 ␮g). From these numbers, valuable information can be deduced about the relative molar amounts of specific proteins in the PSD (e.g. relative abundance of members of protein families or of mutually interacting proteins). However, it is not possible to calculate the absolute numbers of a particular protein within an individual PSD because we could not count the number of PSDs represented by 20 ␮g of

1168


Relative Quantification of 32 Major PSD Components by the AQUA Strategy The AQUA strategy also allows the relative quantification of proteins because the internal standards can be pooled to serve as a reference in any sample. We measured the absolute amount of 32 proteins within the PSDs from forebrain and cerebellum and then calculated their ratios in abundance (Table II). The results are largely consistent with the results based on the ICAT method. Although hundreds of proteins were quantified using the ICAT strategy, a few known PSD proteins were missing in the analysis. For example, 13 (40%) of 32 proteins measured by AQUA were not quantified in the ICAT analysis despite the utilization of geLC/LC-MS/MS, which provided four-dimensional separation power and generated more than 48,000 MS/MS spectra. The incomplete coverage might be caused by (i) absence of Cys residue in some protein sequences, (ii) Cys-labeled peptides in low abundance and/or low recovery in geLC/LC-MS/MS, (iii) low ionization efficiency causing the signal to be too weak to compete for “sequencing” by MS/ MS, (iv) the nature of “shotgun” proteomics because only a subset of ions can be selected for sequencing when many peptides are co-eluted together, and (v) irregular MS/MS spectra with low database matching scores that were discarded during data filtering. By contrast, in the AQUA method only three (9%) of the selected proteins were not quantified in this study (Table II). The three proteins might be quantified after further improvement of sensitivity using other AQUA peptides. Currently AQUA remains a relatively small scale approach requiring the synthesis of labeled peptides. However, only a small amount of each peptide (subpicomole) is consumed in one assay. With hundreds of peptides covering most or all the identified PSD proteins, AQUA could offer a highly sensitive tool for large scale quantitative analysis of the PSD composition. Acknowledgments—We thank Dr. F. Strobel for the usage of the LTQ-FT mass spectrometer and Drs. M. Kinoshita and S. Lev for the Septin 4 (H5) and Nir2 antibodies, respectively. We also thank Dr. M. P. Epstein for discussion on statistical approaches.


* This work was supported in part by National Institutes of Health Grants DK069580 (to J. P.), DA019937 (to J. P.), and AG025688 (to J. P. and J. H.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www. mcponline.org) contains supplemental material. b Both authors contributed equally to this work. e Recipient of long term fellowship from the Human Frontier Science Program Organization. g An investigator of the Howard Hughes Medical Institute. i To whom correspondence may be addressed. Tel.: 404-712-8510; Fax: 404-727-3728; E-mail: [email protected]. j To whom correspondence may be addressed. Tel.: 617-452-3716; Fax: 617-452-3692; E-mail: [email protected]. REFERENCES 1. Sheng, M., and Kim, M. J. (2002) Postsynaptic signaling and plasticity mechanisms. Science 298, 776 –780 2. Kim, E., and Sheng, M. (2004) PDZ domain proteins of synapses. Nat. Rev. Neurosci. 5, 771–781 3. Kennedy, M. B., Beale, H. C., Carlisle, H. J., and Washburn, L. R. (2005) Integration of biochemical signalling in spines. Nat. Rev. Neurosci. 6, 423– 434 4. Kim, J. H., Liao, D., Lau, L. F., and Huganir, R. L. (1998) SynGAP: a synaptic RasGAP that associates with the PSD-95/SAP90 protein family. Neuron 20, 683– 691 5. Chen, H. J., Rojas-Soto, M., Oguni, A., and Kennedy, M. B. (1998) A synaptic Ras-GTPase activating protein (p135 SynGAP) inhibited by CaM kinase II. Neuron 20, 895–904 6. Steward, O., and Schuman, E. M. (2003) Compartmentalized synthesis and degradation of proteins in neurons. Neuron 40, 347–359 7. Ehlers, M. D. (2003) Activity level controls postsynaptic composition and signaling via the ubiquitin-proteasome system. Nat. Neurosci. 6, 231–242 8. Inoue, A., and Okabe, S. (2003) The dynamic organization of postsynaptic proteins: translocating molecules regulate synaptic function. Curr. Opin. Neurobiol. 13, 332–340 9. Aebersold, R., and Mann, M. (2003) Mass spectrometry-based proteomics. Nature 422, 198 –207 10. Yates, J. R., III (2004) Mass spectral analysis in proteomics. Annu. Rev. Biophys. Biomol. Struct. 33, 297–316 11. Li, K. W., Hornshaw, M. P., Van Der Schors, R. C., Watson, R., Tate, S., Casetta, B., Jimenez, C. R., Gouwenberg, Y., Gundelfinger, E. D., Smalla, K. H., and Smit, A. B. (2004) Proteomics analysis of rat brain postsynaptic density. Implications of the diverse protein functional groups for the integration of synaptic physiology. J. Biol. Chem. 279, 987–1002 12. Peng, J., Kim, M. J., Cheng, D., Duong, D. M., Gygi, S. P., and Sheng, M. (2004) Semiquantitative proteomic analysis of rat forebrain postsynaptic density fractions by mass spectrometry. J. Biol. Chem. 279, 21003–21011 13. Yoshimura, Y., Yamauchi, Y., Shinkawa, T., Taoka, M., Donai, H., Takahashi, N., Isobe, T., and Yamauchi, T. (2004) Molecular constituents of the postsynaptic density fraction revealed by proteomic analysis using multidimensional liquid chromatography-tandem mass spectrometry. J. Neurochem. 88, 759 –768 14. Jordan, B. A., Fernholz, B. D., Boussac, M., Xu, C., Grigorean, G., Ziff, E. B., and Neubert, T. A. (2004) Identification and verification of novel rodent postsynaptic density proteins. Mol. Cell. Proteomics 3, 857– 871 15. Nakagawa, T., Cheng, Y., Ramm, E., Sheng, M., and Walz, T. (2005) Structure and different conformational states of native AMPA receptor complexes. Nature 433, 545–549 16. Gygi, S. P., Rist, B., Gerber, S. A., Turecek, F., Gelb, M. H., and Aebersold, R. (1999) Quantitative analysis of complex protein mixtures using isotope-coded affinity tags. Nat. Biotechnol. 17, 994 –999 17. Li, J., Steen, H., and Gygi, S. P. (2003) Protein profiling with cleavable isotope-coded affinity tag (cICAT) reagents: the yeast salinity stress response. Mol. Cell. Proteomics 2, 1198 –1204

18. Han, D. K., Eng, J., Zhou, H., and Aebersold, R. (2001) Quantitative profiling of differentiation-induced microsomal proteins using isotope-coded affinity tags and mass spectrometry. Nat. Biotechnol. 19, 946 –951 19. Hansen, K. C., Schmitt-Ulms, G., Chalkley, R. J., Hirsch, J., Baldwin, M. A., and Burlingame, A. L. (2003) Mass spectrometric analysis of protein mixtures at low levels using cleavable 13C-isotope-coded affinity tag and multidimensional chromatography. Mol. Cell. Proteomics 2, 299 –314 20. Gerber, S. A., Rush, J., Stemman, O., Kirschner, M. W., and Gygi, S. P. (2003) Absolute quantification of proteins and phosphoproteins from cell lysates by tandem MS. Proc. Natl. Acad. Sci. U. S. A. 100, 6940 – 6945 21. Barnidge, D. R., Dratz, E. A., Martin, T., Bonilla, L. E., Moran, L. B., and Lindall, A. (2003) Absolute quantification of the G protein-coupled receptor rhodopsin by LC/MS/MS using proteolysis product peptides and synthetic peptide standards. Anal. Chem. 75, 445– 451 22. Cho, K. O., Hunt, C. A., and Kennedy, M. B. (1992) The rat brain postsynaptic density fraction contains a homolog of the Drosophila discs-large tumor suppressor protein. Neuron 9, 929 –942 23. Litvak, V., Dahan, N., Ramachandran, S., Sabanay, H., and Lev, S. (2005) Maintenance of the diacylglycerol level in the Golgi apparatus by the Nir2 protein is critical for Golgi secretory function. Nat. Cell Biol. 7, 225–234 24. Litvak, V., Argov, R., Dahan, N., Ramachandran, S., Amarilio, R., Shainskaya, A., and Lev, S. (2004) Mitotic phosphorylation of the peripheral Golgi protein Nir2 by Cdk1 provides a docking mechanism for Plk1 and affects cytokinesis completion. Mol. Cell 14, 319 –330 25. Kinoshita, A., Noda, M., and Kinoshita, M. (2000) Differential localization of septins in the mouse brain. J. Comp. Neurol. 428, 223–239 26. Shevchenko, A., Wilm, M., Vorm, O., and Mann, M. (1996) Mass spectrometric sequencing of proteins silver-stained polyacrylamide gels. Anal. Chem. 68, 850 – 858 27. Washburn, M. P., Wolters, D., and Yates, J. R., III (2001) Large-scale analysis of the yeast proteome by multidimensional protein identification technology. Nat. Biotechnol. 19, 242–247 28. McDonald, W. H., Ohi, R., Miyamoto, D. T., Mitchison, T. J., and Yates, J. R., III (2002) Comparison of three directly coupled HPLC MS/MS strategies for identification of proteins from complex mixtures: singledimension LC-MS/MS, 2-phase MudPIT, and 3-phase MudPIT. Int. J. Mass Spectrom. 219, 245–251 29. Eng, J., McCormack, A. L., and Yates, J. R., III (1994) An approach to correlate tandem mass spectral data of peptides with amino acid sequences in a protein database. J. Am. Soc. Mass Spectrom. 5, 976 –989 30. Peng, J., Elias, J. E., Thoreen, C. C., Licklider, L. J., and Gygi, S. P. (2003) Evaluation of multidimensional chromatography coupled with tandem mass spectrometry (LC/LC-MS/MS) for large-scale protein analysis: the yeast proteome. J. Proteome Res. 2, 43–50 31. Link, A. J., Eng, J., Schieltz, D. M., Carmack, E., Mize, G. J., Morris, D. R., Garvik, B. M., and Yates, J. R., III (1999) Direct analysis of protein complexes using mass spectrometry. Nat. Biotechnol. 17, 676 – 682 32. Baldwin, M. A. (2004) Protein identification by mass spectrometry: issues to be considered. Mol. Cell. Proteomics 3, 1–9 33. Li, X. J., Zhang, H., Ranish, J. A., and Aebersold, R. (2003) Automated statistical analysis of protein abundance ratios from data generated by stable-isotope dilution and tandem mass spectrometry. Anal. Chem. 75, 6648 – 6657 34. Gelman, A., Carlin, J., Stern, H., and Rubin, D. (1995) Bayesian Data Analysis, Chapman & Hall, London 35. Kashiwabuchi, N., Ikeda, K., Araki, K., Hirano, T., Shibuki, K., Takayama, C., Inoue, Y., Kutsuwada, T., Yagi, T., Kang, Y., Aizawa, S., and Mishina, M. (1995) Impairment of motor coordination, Purkinje cell synapse formation, and cerebellar long-term depression in GluR␦2 mutant mice. Cell 81, 245–252 36. Listgarten, J., and Emili, A. (2005) Statistical and computational methods for comparative proteomic profiling using liquid chromatography-tandem mass spectrometry. Mol. Cell. Proteomics 4, 419 – 434 37. Ribar, T. J., Rodriguiz, R. M., Khiroug, L., Wetsel, W. C., Augustine, G. J., and Means, A. R. (2000) Cerebellar defects in Ca2⫹/calmodulin kinase IV-deficient mice. J. Neurosci. 20, RC107:1–5 38. Matsumoto, M., Nakagawa, T., Inoue, T., Nagata, E., Tanaka, K., Takano, H., Minowa, O., Kuno, J., Sakakibara, S., Yamada, M., Yoneshima, H., Miyawaki, A., Fukuuchi, Y., Furuichi, T., Okano, H., Mikoshiba, K., and Noda, T. (1996) Ataxia and epileptic seizures in mice lacking type 1 inositol 1,4,5-trisphosphate receptor. Nature 379, 168 –171


1169


39. Lev, S. (2004) The role of the Nir/rdgB protein family in membrane trafficking and cytoskeleton remodeling. Exp. Cell Res. 297, 1–10 40. Saitoh, O., and Odagiri, M. (2003) RGS8 expression in developing cerebellar Purkinje cells. Biochem. Biophys. Res. Commun. 309, 836 – 842 41. Goto, K., and Kondo, H. (1999) Diacylglycerol kinase in the central nervous system—molecular heterogeneity and gene expression. Chem. Phys. Lipids 98, 109 –117 42. Kennedy, M. B., Bennett, M. K., and Erondu, N. E. (1983) Biochemical and immunochemical evidence that the “major postsynaptic density protein” is a subunit of a calmodulin-dependent protein kinase. Proc. Natl. Acad. Sci. U. S. A. 80, 7357–7361 43. Suzuki, T., Okumura-Noji, K., Tanaka, R., and Tada, T. (1994) Rapid translocation of cytosolic Ca2⫹/calmodulin-dependent protein kinase II into postsynaptic density after decapitation. J. Neurochem. 63, 1529 –1537 44. Dosemeci, A., Reese, T. S., Petersen, J., and Tao-Cheng, J. H. (2000) A novel particulate form of Ca2⫹/calmodulin-dependent [correction of Ca2⫹/CaMKII-dependent] protein kinase II in neurons. J. Neurosci. 20, 3076 –3084 45. Colbran, R. J., and Brown, A. M. (2004) Calcium/calmodulin-dependent protein kinase II and synaptic plasticity. Curr. Opin. Neurobiol. 14, 318 –327 46. Lisman, J., Schulman, H., and Cline, H. (2002) The molecular basis of CaMKII function in synaptic and behavioural memory. Nat. Rev. Neuro-

1170


sci. 3, 175–190 47. Wenthold, R. J., Petralia, R. S., Blahos, J., II, and Niedzielski, A. S. (1996) Evidence for multiple AMPA receptor complexes in hippocampal CA1/ CA2 neurons. J. Neurosci. 16, 1982–1989 48. Lujan, R., Nusser, Z., Roberts, J. D., Shigemoto, R., and Somogyi, P. (1996) Perisynaptic location of metabotropic glutamate receptors mGluR1 and mGluR5 on dendrites and dendritic spines in the rat hippocampus. Eur. J. Neurosci. 8, 1488 –1500 49. Kim, E., Naisbitt, S., Hsueh, Y. P., Rao, A., Rothschild, A., Craig, A. M., and Sheng, M. (1997) GKAP, a novel synaptic protein that interacts with the guanylate kinase-like domain of the PSD-95/SAP90 family of channel clustering molecules. J. Cell Biol. 136, 669 – 678 50. Havlis, J., and Shevchenko, A. (2004) Absolute quantification of proteins in solutions and in polyacrylamide gels by mass spectrometry. Anal. Chem. 76, 3029 –3036 51. Barnidge, D. R., Hall, G. D., Stocker, J. L., and Muddiman, D. C. (2004) Evaluation of a cleavable stable isotope labeled synthetic peptide for absolute protein quantification using LC-MS/MS. J. Proteome Res. 3, 658 – 661 52. Chen, X., Vinade, L., Leapman, R. D., Petersen, J. D., Nakagawa, T., Phillips, T. M., Sheng, M., and Reese, T. S. (2005) Mass of the postsynaptic density and enumeration of three key molecules. Proc. Natl. Acad. Sci. U. S. A.