J. Neurochem. (2011) 119, 617–629.
Bipolar disorder is a devastating illness that is marked by recurrent episodes of mania and depression. There is growing evidence that the disease is correlated with disruptions in synaptic plasticity cascades involved in cognition and mood regulation. Alleviating the symptoms of bipolar disorder involves chronic treatment with mood stabilizers like lithium or valproate. These two structurally dissimilar drugs are known to alter prominent signaling cascades in the hippocampus, but their effects on the post-synaptic density complex remain undefined. In this work, we utilized mass spectrometry for quantitative profiling of the rat hippocampal post-synaptic proteome to investigate the effects of chronic mood stabilizer treatment. Our data show that in response to chronic treatment of mood stabilizers there were not gross qualitative changes but rather subtle quantitative perturbations in post-synaptic density proteome linked to several key signaling pathways. Our data specifically support the changes in actin dynamics on valproate treatment. Using label-free quantification methods, we report that lithium and valproate significantly altered the abundance of 21 and 43 proteins, respectively. Seven proteins were affected similarly by both lithium and valproate: Ank3, glutamate receptor 3, dynein heavy chain 1, and four isoforms of the 14-3-3 family. Immunoblotting the same samples confirmed the changes in Ank3 and glutamate receptor 3 abundance. Our findings support the hypotheses that BPD is a synaptic disorder and that mood stabilizers modulate the protein signaling complex in the hippocampal post-synaptic density.
false discovery rate
glutamate receptor 3
The hippocampus and related limbic structures are critical to regulating affect, stress responses, and cognition related to memory formation. Dysfunction in these brain regions is hypothesized to contribute to affective illnesses like bipolar disorder. Observational and functional studies have investigated the neuroanatomical differences in the hippocampus of bipolar patients, although few differences in overall structure and cell type distributions have been noted (Knable et al. 2004; Videbech and Ravnkilde 2004; Frazier et al. 2005). However, the two most prominent drugs used to treat the manic and depressive elements of bipolar disorder, lithium and valproate, have been shown to have a significant impact on the hippocampus in both pre-clinical and clinical studies. These mood stabilizers enhance the rate of hippocampal neurogenesis in rodents (Chen et al. 2000), patients (Yucel et al. 2007, 2008), and can drastically alter prominent signaling cascades as well. Previous studies have shown that lithium and valproate alter prominent intracellular signaling pathways, such as the glycogen synthase kinase-3, extracellular-regulated kinase, B-cell chronic lymphocytic leukemia/lymphoma 2 (Bcl-2) signaling cascade and the wnt/β-catenin pathway (Hunsberger et al. 2009). Although it is unclear whether altering these signaling cascades can affect the overall hippocampal structure, there is evidence that significant changes emerge at the subcellular level, particularly at the neuronal synapse.
Located at neuronal terminals, the post-synaptic density (PSD) comprises a complex network of cytoskeletal scaffolding and signaling proteins that facilitate the movement of receptor and signaling proteins within the synaptic active zone (Sheng and Hoogenraad 2007). It is thought to facilitate many functions that are critical to interneuronal signaling and to reflect cellular response to external environmental changes (Emes et al. 2008). Metabotropic glutamate receptors (mGluRs) in the PSD control long-term potentiation or depression, providing the mechanisms thought to underlie cognitive and mood processes (Luscher and Huber 2010). Dysfunction in plasticity cascades has been implicated in patients suffering from bipolar disorder, as well as many other psychiatric disorders, and the therapeutic basis of conventional and novel mood stabilizers may be mediated through improvements in synaptic plasticity regulation (Coyle and Duman 2003; Zarate et al. 2006). Observational studies have examined the composition of the post-synaptic complex in the rodent brain (Collins et al. 2006; Trinidad et al. 2006), but methods for quantitatively evaluating protein expression are still emerging (Dosemeci et al. 2007). The over-arching remodeling of the post-synaptic protein complex following chronic treatment with mood stabilizers is unknown. In this work, we used mass spectrometry to quantitatively investigate the composition of the rat hippocampal post-synaptic complex and examined how this proteomic network is affected by chronic treatment with lithium and valproate.
Materials and methods
Animals and treatment with mood stabilizers
Eight-week-old male Wistar Kyoto rats (Taconic, Germantown, NY, USA) were housed in pairs and treated for 5 weeks with either unmedicated chow (control group, n = 12), or medicated chow containing either lithium carbonate (2.4 gm/kg, n = 12) or sodium valproate (20 gm/kg, n = 10) (BioServ, Frenchtown, NJ, USA). All animals were housed in a facility with constant temperature (22 ± 1°C) and 12 h light/dark cycle with ad libitum access to food and water. An additional water bottle containing 0.9% saline solution was provided to ensure electrolyte balance was maintained throughout the experiment. All experimental procedures were approved by the Animal Use Committee of the National Institute of Mental Health and were conducted according to the NIH guidelines.
Isolation of hippocampal post-synaptic densities
The PSD fraction was isolated from hippocampal tissue of rats that were housed in the same cage using procedures previously published (Carlin et al. 1980; Dosemeci et al. 2000). Hippocampi were rapidly collected, pooled from two animals so that there would be sufficient final PSD yield (∼60–100 μg) for proteolysis and peptide fractionation. Tissue was immediately homogenized using a chilled glass Teflon homogenizer in ice-cold buffer (0.3 M mannitol, 1 mM EDTA, protease inhibitor cocktail; Gamm et al. 2008), omitting aprotinin because of observed interference with subsequent analyses (Aminoethyl-benzene sulfonyl fluoride hydrochloride 104 mM, bestatin HCl 4 mM, E-64 1.4 mM, leupeptin 2 mM, pepstatin 1.5 mM and phosphatase inhibitors I and II diluted 1 : 100, all from Sigma Aldrich, St Louis, MO, USA). The homogenate was centrifuged at 3000 g for 10 min at 4°C [SS-34 rotor, Sorvall (Thermo Fisher Scientific, Waltham, MA, USA)] to remove nuclei and cell debris. The resulting supernatant was saved and pellet was resuspended in another 3 mL ice-cold mannitol buffer and centrifuged as above. Pooled supernatants were centrifuged at 26 900 g for 30 min at 4°C. The pellet was resuspended again in 0.7 mL ice-cold mannitol buffer and fractionated on a 6-tier Ficoll® gradient (20, 16, 12, 8, 2, 0%) and centrifuged for 90 min at 59 600 g at 4°C [SW-60ti rotor; Beckman XL-90 Ultra (Beckman Coulter, Inc; Miami, FL, USA)].
The fractions between the 8–12% and 12–16% interfaces were first collected, pooled together, and then diluted 1 : 4 in 4 mL mannitol buffer. These combined fractions were centrifuged at 26 900 g (SS-34 rotor) for 20 min at 4°C. The synaptosomal pellet was resuspended in 0.9 mL of a 20 mM HEPES solution (pH 7.2) containing 0.5% Triton X-100 detergent (supplemented with protease and phosphatase I and II inhibitors; diluted 1 : 100 as recommended). The suspension was centrifuged at 20 800 g (Eppendorf 5417R) for 30 min at 4°C. The pellet was resuspended in 0.5 mL of HEPES buffer (containing 0.5% Triton X-100 and 75 mM KCl) and centrifuged again at 20 800 g for 30 min at 4°C. Finally, the enriched PSD pellet was resuspended in 0.2 mL detergent-free 20 mM HEPES buffer. The protein concentration was estimated by BCA protein assay (Thermo Scientific, Rockford, IL, USA) and aliquots were frozen at −80°C.
Immunoblotting was performed using a standard protocol. Briefly, membranes were probed with the primary antibodies to Ankyrin G (1 : 100, NB20; EMD Biosciences, Gibbstown, NJ, USA), glutamate receptor 3 (Grm3; 1 : 300, sc-47137; Santa Cruz Biotechnology, Santa Cruz, CA, USA) and PSD-95 (1 : 5000, No. 4970; Cell Signaling, Beverley, MA, USA). Protein bands were visualized using ECL Plus enhanced chemiluminescent signal detection kit (GE Healthcare, GE Healthcare Bio-Sciences Corp, Piscataway, NJ, USA) and exposed to Kodak Biolight film (Rochester, NY, USA). When necessary, membranes were stripped using Re-Blot Plus (Millipore, Waltham, MA, USA). Protein levels were normalized to a loading control (PSD-95) and subsequently analyzed by densitometric film analysis using AlphaImager software (Alpha Innotech, San Leandro, CA, USA). GraphPad Prism software was used for statistics, performing one-way anova with a post hoc Tukey (compares all columns) and individual t-tests (Control vs. Lithium, Control vs. Valproate).
Strong cation exchange chromatography
From each treatment group, three different PSD preparations of 50 μg proteins were denatured by 8 M urea at 60°C for 45 min. To reduce denatured PSD proteins, 1 M dithiothreitol (final concentration 45 mM) was added and incubated at 60°C for 15 min. The proteins were alklyated by adding 1 M iodoacetamide (final concentration 100 mM) and incubating in the dark for 15 min. Subsequently, the PSD sample was digested by diluting from 8 to 1 M urea concentration by adding 100 mM NH4HCO3 and sequencing grade trypsin (Promega, Madison, WI, USA) (substrate to enzyme ratio of 10 : 1; w/w). The sample was digested at 37°C overnight. The digested samples were desalted using an UltraMicroSpinTM reverse phase column (The Nest Group Inc., Southboro, MA, USA) according to the manufacturer’s instructions.
The desalted sample was concentrated to dryness in vacuo (SpeedVac; Thermo Scientific, USA). The dried peptides were resuspended in 100 μL of 10 mM ammonium formate, 10 mM formic acid, and 25% acetonitrile by vortexing and sonicating for 15 min respectively. The resuspended sample was then loaded onto the PolySULFOETHYL A column (1.0 mm × 50 mm, PolyLC Inc., Columbia, MD, USA) to separate peptides by strong cation exchange (SCX) chromatography. The composition of solution A was 10 mM ammonium formate, 10 mM formic acid, 25% acetonitrile; solution B contained 500 mM ammonium formate, 500 mm formic acid, 25% acetonitrile. The peptides were separated by 12 non-linear gradient steps of 3 min each at 100 μL/min flow rate. The fractions were collected every minute for a total of 36 fractions. The concentration of solution B in each step of the gradient was 0%, 1%, 3%, 5%, 7%, 10%, 15%, 20%, 30%, 50%, 70% and 100%, respectively. Each fraction was dried in vacuo and then re-suspended in nanopure water to remove volatile ammonium formate salts in vacuo. After three cycles of aqueous suspension and in vacuo drying, the peptides were re-suspended in 30 μL of 5% acetonitrile, 0.1% formic acid before analyzing by mass spectrometry.
Samples were separated on reverse phase nanocolumn (PicoFritBioBasicC18 column with 75-m inner diameter and 15-m tip, New Objective, Woburn, MA, USA) at 500 nL/min, running a 60-min linear gradient from 5% to 80% acetonitrile on an Eksigent nano LC 2D HPLC system. The LC unit was coupled to Triversa Nanomate (Advion, Ithaca, NY, USA) spray source attached to LTQ-Orbitrap (Thermo Electron, San Jose, CA, USA). To minimize the variation in liquid chromatography between fractions, individual fractions from each treatment were run in a set. That is, replicates of fraction 1 from the SCX of the three treatment groups (nine samples) were analyzed on LC/MS as a single set. In each set, between every sample analyzed, one blank with a short LC cycle of 30 min was run to minimize peptide carry-over. All of the fractions from all replicates and conditions were analyzed, a total of 324 LC/MS runs. The peptides were analyzed in positive ion mode; for each MS scan, the top five most intense ions were selected for collision-induced dissociation and MS/MS recording. The collision energy was set at 35%. The resolution for MS was set at 60 000 and data were collected in a centroid mode.
Data were searched using MASCOT version 2.1 (Matrix Science, Boston, MA, USA) (Perkins et al. 1999). The data were searched against the Swiss-Prot database (version 57. 15.0, release date 24 March 2009) with the species filter ‘mammals’. The other parameters were as follows: (i) enzyme specificity: trypsin; (ii) one allowed missed cleavage; (iii) fixed modification: cysteine carbamidomethylation; (iv) variable modification: methionine oxidation; (v) precursor mass tolerance was ±50 ppm; and (vi) fragment ion mass tolerance was ±0.8 Da. The false discovery rate (FDR) for each sample set was determined with Mascot using a concatenated reversed sequence decoy database. Control 1: 0.63% (11 decoy hits/1751 peptides), Control 2: 0.27% (6 decoys/2184 peptides), Control 3: 0.35% (8 decoys/2306 peptides), Lithium 1: 0.43% (10 decoys/2314 peptides), Lithium 2: 0.32% (9 decoys/2175 peptides), Lithium 3: 0.47% (11 decoys/2343 peptides), Valproate 1: 0.46% (10 decoys/2193 peptides), Valproate 2: 0.23% (6 decoys/2598 peptides) Valproate 3: 0.25% (6 decoys/2418 peptides). It should be noted that most reversed sequence decoy database hits were single spectra observations. As a result, the FDR calculated for spectral hits is lower than that for peptides (average FDR for all nine sample sets based on spectral hits is 0.07 ± 0.021%). The protein inference from the peptide search data was carried out by parsimony analysis (Yang et al. 2004) using the NCBI software MassSieve (Slotta et al. 2010). In MassSieve, all single-peptide hits are removed and only peptide identifications with MASCOT ion scores greater than or equal to their identity scores and with probability scores less than 0.05 accepted. The net effect of this filtering is that the calculable peptide FDR drops to zero. We used a broad species filter because the rat database is incomplete, and then reviewed all mammalian proteins designated as equivalent (i.e. containing identical sets of peptides) (Liska and Shevchenko 2003; Junqueira et al. 2008). Using this strategy, protein identifications were assigned to the species rat by default, but wherever a rat homolog was not found, a secondary rodent species (mouse) was preferentially selected, followed by human and bovine, respectively. We included proteins only if there was at least one unique peptide (observed multiple times) identified in each of the groups (control, lithium, valproate) based on the criteria described above. Mass spectrometric data files (raw data files, data files from the MASCOT search engine, and the protein inference list from MassSieve) associated with this paper have been deposited in the NCBI Peptidome Repository for public access (accession PSE 150) (http://www.ncbi.nlm.nih.gov/projects/peptidome/).
Protein quantification and statistical analyses
The software program suite DBparser 3.0 was used for the protein quantification as described previously (Dosemeci et al. 2007; McFarland et al. 2008). The program extracts retention time and peak intensities from MS1 raw data based on the precursor mass assigned by MASCOT. Proteins were quantified by summing the ion current intensities of all constituent peptides, and have not been normalized. The log2-fold change between treated and control was calculated by comparing average intensities from three replicates. Proteins that were not identified in all three replicates were not considered for quantification. A two-tailed t-test and Cohen’s d effect size correlation was performed based on log2-transformed intensities. The p-value cutoff of > 0.9, effect size of > ±0.8, and log2-fold change of ±0.8 was considered significant. The effect size (Cohen’s d) is the number of standard deviations difference between conditions. So, a Cohen’s d of 1.0 means there is a one standard deviation difference between conditions. Thus, the effect sizes take into account the size of the standard deviation for the variable in question.
The networks, functional analyses, and canonical pathways were generated using Ingenuity Pathway Analysis (Ingenuity® Systems, http://www.ingenuity.com). The composite data of post-synaptic density proteome used is the dataset reported in Table S1.
Composition of hippocampal post-synaptic density proteome
We identified 605 proteins in the PSD of the rat hippocampus, based on concatenated data sets derived from nine sets of mass spectrometric analyses [3 conditions × 1/2 (6 paired hippocampi) × 36 SCX fractions, or 324 runs] (Table S1). Of the total, 584 proteins (96.5%) were found in both treatment groups and control (but not necessarily in all three replicates), whereas 332 (55%) of the proteins were found in all of the replicates of all three groups. The functional classification of the proteins, shown in Table S1, follows the format used previously in a comprehensive literature survey of PSD proteins (Collins et al. 2006). Based on this classification scheme, our PSD preparation was predominantly composed of signaling proteins (23%), cytoskeletal and cell adhesion (20%) and synaptic vesicle proteins (13%) (Fig. 1) similar to those reported in the survey of PSD proteomics literature (Collins et al. 2006). In the Collins study, a consensus list of 466 proteins was compiled from seven different studies profiling the rodent post-synaptic density proteome. From this consensus list, we found 424 proteins in our PSD preparation from rat hippocampus (data not shown). The overlap of our dataset list with the published PSD consensus list indicates that our preparations are relatively specific, and typical of sucrose density gradient purified post-synaptic fractions (Dosemeci et al. 2007). Trinidad et al. reported 2159 proteins in PSD isolated from several regions of mouse brain. This higher number of PSD proteins may be attributed to improved instrumentation, synaptic heterogeneity, and/or sample quantity. The Trinidad study used tenfold higher amounts of purified PSD preparation (500 μg, tissues pooled from several animals) in their analyses compared to 50 μg in the current study (Trinidad et al. 2006).
Effects of lithium and valproate on hippocampal post-synaptic proteome
We compared the changes in protein abundance at the post-synaptic density following chronic treatment with lithium or valproate. Based on our statistical criteria (see Materials and methods section), we found that lithium treatment significantly altered the level of 20 proteins (Table 1) (for complete list, see Table S2) and valproate significantly affected abundance of 41 proteins (Table 2) (for complete list, see Table S3). Seven proteins were significantly altered by both lithium and valproate treatment: metabotropic Grm3, ankyrin 3 (Ank3), dynein heavy chain 1, and 14-3-3 protein isoforms T, F, E and Z. All seven proteins identified responded similarly, in direction and magnitude, to both mood stabilizers (Tables 1 and 2 shown in bold letters). To validate these results, we performed immunoblotting experiments on three selected proteins Ank3 and Grm3 with the same nine PSD fraction sample sets (n = 3/treatment group) that underwent mass spectrometry analysis (Fig. 2). After testing for antibody specificity and linearity (data not shown), we confirmed that Grm3 and Ank3 abundance were increased, as expected in the PSD fraction by mood stabilizers (Fig. 2).
|Protein||Log2-fold change||p-value||Disease function||Description|
|IGSF8_MOUSE||2.4||0.97||Glioblastoma/AD (Zhang et al. 2005)||Implicated in neurite outgrowth and maintenance of neural networks. Over-expression is shown to inhibit glioblastoma in vivo and in vitro (Murdoch et al. 2003; Kolesnikova et al. 2009)|
|E41L3_MOUSE||1.9||0.95||Interacts with actin and spectrin and plays a role in recruitment of NMDARs to synaptic junctions (Hoy et al. 2009)|
|PDE2A_RAT||1.8||0.96||Depression||Implicated in depression and elevated levels are shown in hippocampus and cerebellum on treatment of antidepressants (Esposito et al. 2009)|
|VA0D1_MOUSE||1.4||0.98||BD (Wellcome Trust Clinical Consortium 2007)||ATP synthetase|
|RIMB2_RAT||1.2||0.90||Regulation and coordination of synaptic vesicle trafficking and exocytosis (Hibino et al. 2002)|
|GRM3_RAT||1.0||0.93||Schz||Metabotropic glutamate receptor 3|
|QCR1_RAT||0.8||0.99||HD (Trushina and McMurray 2007)/Depression||Level shown to be reduced in patients suffering with depression (Johnston-Wilson et al. 2000)|
|ANK3_HUMAN||0.8||0.94||BD||Regulation of the assembly of voltage-gated sodium channels|
|CSKI1_RAT||−0.8||0.92||Forms tripartite complex of unknown function with CASK and neurexin-1 (Tabuchi et al. 2002)|
|CTNA2_MOUSE||−1.2||0.92||BD (Wellcome Trust Clinical Consortium 2007)/PMN||Regulator of synaptic remodeling and plays role in stability of dendritic spines and contacts (Abe et al. 2004)|
|RGRF1_RAT||−1.3||0.93||Activation of Erk and cAMP signaling by coupling of NMDARs to Ras (Sepulveda et al. 2010)|
|TPIS_RAT||−1.5||0.98||HD (Strand et al. 2005)||Glycolytic enzyme linked to inositol depletion during lithium treatment (Shi et al. 2005)|
|CCG8_RAT||−1.5||0.97||Regulation and translocation of AMPA receptors on synaptic membranes (Rouach et al. 2005)|
|Protein||Log2-fold change||p-value||Disease function||Description|
|MAP2_RAT||2.6||1.00||HD (Hodges et al. 2006)/BD||Reduced expression in hippocampus in individuals suffering with bipolar disorder (MacDonald et al. 2006)|
|AT2B1_RAT||2.4||0.90||HD (Luthi-Carter et al. 2000)||Calcium transport and homeostasis|
|PRDX1_RAT||2.4||0.92||AD (Cumming et al. 2007)/BD (Wellcome Trust Clinical Consortium 2007)||Redox regulation of cell|
|HS12A_HUMAN||2.0||1.00||Schz||Expression is reduced in prefrontal cortex in patients suffering with schizophrenia (Pongrac et al. 2004)|
|NCDN_RAT||1.9||0.97||Epilepsy (Dateki et al. 2005)||Deletion results into epileptic seizures and may be negative regulator of CamKII (Dateki et al. 2005)|
|ADA23_MOUSE||1.4||0.98||Interacts with LGI1 and required for regulation of neurite growth (Owuor et al. 2009)|
|PSD3_MOUSE||1.4||0.92||Modulator/guanine exchange factor for Arf6|
|CAPZB_RAT||1.4||0.90||Regulation of actin polymerization (Cooper and Schafer 2000)|
|NSF_RAT||1.3||0.96||Schz (Hurko and Ryan 2005)||Regulated targeting of synaptic vesicles|
|CAZA1_RAT||1.2||0.90||Regulation of actin polymerization (Cooper and Schafer 2000)|
|DPYL2_RAT||1.2||1.00||BD/Schz/AD (Takata et al. 2009)||Implicated as common genetic marker between schizophrenia and bipolar disorder (Fallin et al. 2005)|
|RUFY3_RAT||1.2||0.98||Parkinson’s disease (Maraganore et al. 2005)||Neuronal polarization (Mori et al. 2007)|
|PROF2_RAT||1.2||0.94||Schz (Hakak et al. 2001)||Actin polymerization (Pilo Boyl et al. 2007)|
|NAC2_RAT||1.1||0.99||Calcium transport and homeostasis|
|GRM2_RAT||1.0||0.97||Metabotropic glutamate receptor 2|
|COF1_RAT||1.0||0.97||Cytoskeleton protein/disassembly of actin filaments (Bellenchi et al. 2007)|
|VATB2_RAT||1.0||1.00||HD (Strand et al. 2005)/Schz (Hakak et al. 2001)||ATP synthesis and up-regulated in dorsolateral prefrontal cortex in patients suffering with schizophrenia (Hakak et al. 2001)|
|MAP4_RAT||0.9||0.97||Parkinson’s disease (Fung et al. 2006)||Microtubule-associated protein|
|ANK3_HUMAN||0.9||0.95||BD||Regulation of the assembly of voltage-gated sodium channels|
|PHAR1_RAT||0.9||0.93||BD (Wellcome Trust Clinical Consortium 2007)||Protein phosphatase I and actin regulatory protein|
|SYT1_RAT||0.8||0.99||Schz (Hemby et al. 2002)||Calcium sensor for coordinated synaptic vesicle exocytosis. Expression is reduced in patients suffering with mesial temporal lobe epilepsy (Yang et al. 2006)|
|SNP25_RAT||0.8||0.98||BD/Schz||Plays role in synaptic vesicle exocytosis and expression is found to be reduced in hippocampus of patients suffering with schizophrenia and bipolar disorder (Fatemi et al. 2001)|
|GBB1_RAT||0.8||0.92||HD (Abou-Sleymane et al. 2006)/AD (Li et al. 2008)||Beta subunit of G-protein|
|GRM3_RAT||0.8||0.96||Schz||Metabotropic glutamate receptor 3|
|GEPH_HUMAN||0.8||0.94||Hypereplexia||GABAergic and glycinergic inhibitory synapse. Form tripartite complex with neurolign-2 and collybistin (Poulopoulos et al. 2009)|
|IQEC3_RAT||−1.2||0.91||Modulator/guanine exchange factor for Arf1|
|IP3KA_RAT||−1.4||0.94||Activity of the enzyme is regulated by Ca2+ and calmodulin and plays role in calcium homeostasis (Xia and Yang 2005)|
|UN13A_RAT||−1.7||0.97||ALS (Schymick et al. 2007)||Synaptic vesicle priming|
|NLGN2_RAT||−2.0||0.91||GABAergic and glycinergic inhibitory synapse. Form tripartite complex with neurolign-2 and collybistin (Poulopoulos et al. 2009)|
|NOE1_RAT||−2.3||1.00||Schz||Uncharacterized/novel Physically interacts with DISC-1 (Camargo et al. 2007)|
|MYO6_MOUSE||−5.3||0.91||Regulate the clathrin mediate endocytosis of AMPA (Osterweil et al. 2005) receptors (Camargo et al. 2007)|
Lithium- and valproate-specific quantitative changes in hippocampal PSD proteome
There are 13 additional protein concentrations that changed specifically in response to lithium as described in Table 1. These proteins are known from the literature to be implicated in different neurological and psychiatric disorders and are associated with different functions as indicated in the table. The quantitative changes in the protein phosphodiesterase Pde2a and the scaffolding protein Dlg1 are consistent with the observations made by others in the literature (Sato et al. 2008; Esposito et al. 2009) Valproate treatment has more robust effect on altering PSD protein levels in comparison with lithium. There are an additional 35 proteins whose abundance has changed on valproate treatment (Table 2): especially those in different functional categories like cytoskeletal (especially actin dynamics as discussed later), signaling proteins, enzymes, synaptic vesicles and transport. As shown in Table 2, many of these proteins have been implicated in various psychiatric and neurological disorders.
Functional assignment by ingenuity pathway analysis of the 605 identified proteins showed the top five functions as cell-to-cell signaling, cellular organization, morphology, cellular transport, and cell signaling. The top five canonical pathways mapped were oxidative phosphorylation (38 proteins), CREB signaling in neurons (36 proteins), synaptic long-term potentiation (28 proteins), glutamate receptor signaling (20 proteins), and calcium signaling (32 proteins). The enrichment of mitochondrial proteins in PSD preparation supports the assignment of oxidative phosphorylation as a top canonical pathway. Network analysis of non-canonical pathways is summarized in Table S4, which includes the proteins in each network, their score, and the number of associated focus proteins identified. Ingenuity pathway analysis calculates a score for each network that indicates the likelihood that this set of focus genes in a network could be explained by random chance alone. The score is generated by taking into account the number of network eligible molecules, the size of the network, and the total number of molecules in the ingenuity knowledge base that can be included in networks. The score is calculated using a right-tailed Fisher’s exact test and is displayed as the negative log of that p-value. A score of 6 indicates that there is a one in million chance of deriving this network due to random chance.
We set out to evaluate the composition of the hippocampal post-synaptic region and how two very different chronic mood stabilizer treatments affect its composition using an unbiased mass spectrometry-based proteomics approach. We found 584 proteins of 605 proteins in the control and both treatment conditions. These results suggest that chronic lithium and valproate treatments do not promote synthesis or pruning of new protein families, but instead modulate significant increases or decreases in the abundance of proteins present in the post-synaptic proteome as defined by isolation methods commonly used in this field. Although this preparation is enriched with respect to histologically confirmed post-synaptic density proteins, there are some proteins co-isolated that are known to be pre-synaptic or mitochondrial in origin. The abundance of scaffolding proteins, with the exception of Dlg1 (Disks large homolog 1, or synapse-associated protein 97, SAP97), did not change significantly following drug treatment (see Table S5). The graphical comparison of the log2-fold changes of lithium versus valproate treatment (Fig. 3) reveals that most proteins are centered towards zero in all four quadrants. This suggests that mood stabilizer treatments lead to few, specific quantitative changes in the PSD proteome. Remarkably, there is not a single protein in this scatter plot analysis whose abundance is changed in opposite directions by the two drug treatments (see quadrants 1 and 3 of Fig. 3) and there are few that show similar changes in abundance (see quadrants 2 and 4). There are a significant number of proteins that show changes in abundance levels that are specific to either lithium or valproate treatment (points close to the X- and Y-axes in Fig. 3), suggesting differences in mechanism and pharmacology of these two drugs. The effectiveness of valproate on alteration of protein level, could also be due to its inhibitory action on histone deacetylase, which is thought to epigenetically repress transcription (Gottlicher et al. 2001; Phiel et al. 2001).
To get further insight, we analyzed our data set with ingenuity pathway analysis to determine significant non-canonical networks of proteins and corresponding changes distinct to the valproate and lithium treatment. The top five networks identified with this analysis are detailed in Table S4. Because it is known that valproate induces growth cone spreading and affects the actin polymerization leading to changes in growth cone morphology, the network illustrating that the abundance of 50 actin-interacting proteins change on valproate treatment (Fig. 4) is shown. The change in actin cytoskeleton is important for neuronal guidance and may be responsible for establishing new neuronal contacts resulting in changes in the neuronal circuit. In this network of proteins, Cof1(CFL1), Dpysl2, Phactr1, IP3KA (ITPKA) and E41L3 (EPB41L3) are involved in regulation of actin polymerization and dynamics (Fig. 4). Although not mapped in Fig. 4, Prof2 (1.2 log2-fold change, Table 2) is suggested to interact with the Wave1 complex at the synapse to control the actin polymerization (Pilo Boyl et al. 2007) whereas Cof1 is implicated in the disassembly of actin filaments (Bellenchi et al. 2007). The role of Dpysl2 is shown to be critical in axon formation and dendrite specification. It has been suggested Dpysl2 plays a role in transport of Sra-1/Wave1 complex to growth cones, augmenting the actin reorganization, and thereby inducing the axon outgrowth and formation (Kawano et al. 2005). The role of IP3KA in calcium transport is well known, but it also contains an F-actin binding domain. Kim et al. (2009) showed that this actin binding domain (independent of IP3KA catalytic activity), regulates the remodeling of dendritic spine actin by the scaffolding Rac protein. They also observed the accumulation of IP3KA in the synaptic area after induction of long term potentiation (LTP). Phactr1 and E41L3 also physically interact with actin but their role in actin dynamics is still not very clear (Parra et al. 2000; Allen et al. 2004). The expression of E41L3 is significantly increased also on lithium treatment (Table 1). The actin-capping protein Caza2 (CAPZA2), regulates the actin polymerization by binding to barbed ends of actin filament (Cooper and Schafer 2000). The loss of capping proteins reduces the cellular motility and their role may be ‘funneling’, that is, to regulate actin polymerization at specific time and place (Cooper and Schafer 2000). Our quantitative and network analyses support the pathophysiology of neuronal remodeling and neurogenesis on valproate treatment.
In lithium-treated rats, the expression of the transmembrane protein Igfs8, belonging to immunoglobulin superfamily, increases prominently and significantly (2.5 log2-fold change). Igfs8 is expressed in adult brain and implicated in neurite outgrowth and maintenance of neural networks (Murdoch et al. 2003). Its expression is reduced in glioblastoma and its over-expression inhibits glioblastoma in vivo and in vitro (Kolesnikova et al. 2009). Our data are consistent with the observation that lithium has neuroprotective and neurotrophic effects. In Table 1, the reduced abundance of triose phosphate isomerase after lithium treatment is particularly notable. The therapeutic value of lithium has been attributed to the depletion of inositol in the human brain, although the molecular basis for this depletion is not completely clear. Shi et al. have shown that yeast with a mutation in the gene encoding triphosphate isomerase leads to hypersensitivity to lithium and valproate, inositol auxotrophy and accumulation of dihydroxyacetone phosphate, an intermediate in the glycolytic pathway. The authors reasoned that accumulation of dihydroxyacetone phosphate leads to competitive inhibition of myo-inositol phosphate 3 synthetase, which is a likely cause of inositol auxotrophy (Shi et al. 2005). Our data suggest that it is possible that a reduction in inositol during lithium treatment could be due to inhibition of the de novo biosynthetic pathway.
The seven proteins that changed in common after lithium and valproate treatment do not map to any single set of canonical or non-canonical pathways upon ingenuity pathway analysis. However, the role of the 14-3-3 protein group has been implicated in psychiatric disorders previously. Middleton et al. (2005) have shown that expression of 14-3-3 gene group is decreased in patients suffering with schizophrenia. In support of the transcriptomics data, Martins-de-Souza et al. (2009) showed the reduction in 14-3-3 zeta/delta, 14-3-3 gamma and 14-3-3 eta in patients suffering with schizophrenia using differential labeled quantitative proteomic technique. Two different studies have found evidence of association of two members of this group, 14-3-3F and 14-3-3E with bipolar disorder and schizophrenia (Ikeda et al. 2008; Grover et al. 2009). It was reported that lower levels of 14-3-3E showed developmental defects in hippocampal neurons (Ikeda et al. 2008).
Ankyrin 3 is an adapter protein that regulates the assembly of voltage-gated sodium channels. Genome-wide association studies have shown that the genetic variant of ANK3 (which encodes ankyrin 3) is significantly associated with risk of bipolar disorders (Phiel et al. 2001; Schulze et al. 2009). Some genome-wide association studies imply GRM3 (also known as mGluR3) genetic variants as risk factors for bipolar and major depressive disorder (Fallin et al. 2005; Kato 2007). More studies are needed to explore causality between BPD and ANK3 or GRM3 genetic variants. The mGluR2/3 antagonist, LY-341495, increases mobility time in the forced swim test. The mGluR2/3 agonist, LY-379268, treatment mimics nicotine withdraw in induction of reward deficit monitored with intracranial self-stimulation reward thresholds. mGluR2/3 agonists, LY-379268 and LY-354740 attenuate amphetamine-induced locomotion (Cartmell et al. 1999, 2000), an experimental model for mania and behavioral action of mood stabilizers. Our data suggest both lithium and valproate increase concentrations of Ank3 and mGluR3 in the PSD proteome, supporting the roles of these proteins in mood regulation.
In conclusion, we used a broad, discovery approach to profile the chronic effects of lithium and valproate on the post-synaptic density-enriched proteome isolated from rat hippocampus. The results support the multiple known protein networks of this synaptosomal preparation with regard to neuronal plasticity, and the notion that dysfunction of this synaptic proteome contributes to BPD and other major psychiatric illnesses. Furthermore, we found that chronic treatment with mood stabilizers regulated levels of PSD proteome proteins linked to several key signaling pathways. Our data specifically support the changes in actin dynamics on valproate treatment. Lithium and valproate treatments increased levels of Ank3 and mGluR3 in the PSD proteome, supporting the roles of these two proteins in mood regulation. Future studies are required to test whether targeting Ank3 or mGluR3 produce lithium-like mood-stabilizing effects in animal models and in BPD patients.
This work was supported by the Intramural Research Program of the National Institute of Mental Health, NIH (MH000274 and MH000279). The authors received significant advice and assistance from staff in LNT, especially Dr Jeffrey Kowalak, Anthony J. Makusky, Jason Harrington, and Ronald Finnegan.
This work was supported by the Intramural Research Program of the National Institute of Mental Health, National Institutes of Health (NIMH-NIH), MH000274 and MH000279. The authors report no biomedical financial interests or potential conflicts of interest. At the date of manuscript submission, Dr Manji and Dr Chen are employees of Johnson and Johnson Pharmaceutical Research and Development, Titusville, NJ; Dr Catapano is an employee of George Washington University, Washington, DC, and Dr Nanavati is an employee of Northwestern University, Evanston, IL. This work was initiated while they were employees of the NIMH.