Abstract
Psychiatric disorders such as Alzheimer’s disease, schizophrenia and mood disorders are severe and disabling conditions of largely unknown origin and poorly understood pathophysiology. An accurate diagnosis and treatment of these disorders is often complicated by their aetiological and clinical heterogeneity. In recent years proteomic technologies based on mass spectrometry have been increasingly used, especially in the search for diagnostic and prognostic biomarkers in neuropsychiatric disorders. Proteomics enable an automated high-throughput protein determination revealing expression levels, post-translational modifications and complex protein-interaction networks. In contrast to other methods such as molecular genetics, proteomics provide the opportunity to determine modifications at the protein level thereby possibly being more closely related to pathophysiological processes underlying the clinical phenomenology of specific psychiatric conditions. In this article we review the theoretical background of proteomics and its most commonly utilized techniques. Furthermore the current impact of proteomic research on diverse psychiatric diseases, such as Alzheimer’s disease, schizophrenia, mood and anxiety disorders, drug abuse and autism, is discussed. Proteomic methods are expected to gain crucial significance in psychiatric research and neuropharmacology over the coming decade.
Keywords
Abbreviations
The following abbreviations are used in this review:
two-dimensional polyacrylamide gel electrophoresis;
Abeta binding alcohol dehydrogenase;
amyloid beta;
Alzheimer’s disease;
alcohol dehydrogenase;
apolipoprotein;
amyloid precursor protein;
autistic spectrum disorder;
brain-derived neurotrophic factor;
bipolar disorder;
creatine kinase;
central nervous system;
cAMP response element-binding protein;
cerebrospinal fluid;
difference gel electrophoresis;
dystrophin related protein 2;
enzyme-linked immunosorbent assay;
electrospray ionization;
glyceraldehyde-3-phosphate dehydrogenase;
glutamate dehydrogenase;
glial fibrillary acidic protein;
gamma-glutamyltransferase;
high-anxiety-related behaviour;
hippocampal cholinergic neurostimulating peptide precursor protein;
haptoglobin;
high-performance liquid chromatography;
heat shock protein;
Human Proteome Organisation;
isotope-coded affinity tags;
immunoglobin;
isobaric tag for relative and absolute protein quantification;
intravenous;
low-anxiety-related behaviour;
liquid chromatography electrospray ionization tandem mass spectrometry;
lactate dehydrogenase;
matrix-assisted laser desorption/ionization time of flight mass spectrometry;
methamphetamine;
mitogen-activated protein kinase 1;
mild cognitive impairment;
major depressive disorder;
malate dehydrogenase;
[+]-5-methyl-10, 11-dihydro-5H- dibenzo-[a,d]-cyclohepten-5,10-imine hydrogen maleate;
mass spectrometry;
N-methyl-D-aspartate;
obsessive compulsive disorder;
polymerase chain reaction;
Parkinson’s disease;
pyruvate dehydrogenase;
phosphatidylethanol-binding protein;
phosphatidylethanolamine-binding protein 1;
pyruvate kinase;
peptidyl prolyl cis-trans isomerise;
peroxiredoxin;
precursor;
presenilin 1;
quantitative reverse transcription polymerase chain reaction;
schizophrenia/schizophrenic;
surface-enhanced laser desorption/ionization;
soluble N-ethylmaleimide-sensitive factor attachment protein;
superoxide dismutase;
synaptotagmin;
targeted replacement;
ubiquitin carboxyl-terminal hydrolase;
ubiquitin carboxy-terminal hydrolase L1;
VGF nerve growth factor inducible.
Introduction
Psychiatric disorders such as AD, schizophrenia, mood and anxiety disorders, substance abuse and autism are severe and disabling diseases, and some of them are major causes of morbidity even in childhood and adolescence. In contrast to diseases adhering to strict Mendelian inheritance, mental disorders are complex, polygenetic and often poorly understood with regards to pathomechanisms and biological pathways. In most cases, it is likely that the interplay of multiple gene products with environmental factors produces a given psychiatric phenotype. Over the last decade psychiatric research has primarily focused on genomic approaches, with the human genome project syndicate producing essential progress in the understanding of mental diseases. There had been expectations by researchers and clinical practitioners of significant improvements in diagnostic and therapeutic opportunities based on the new molecular genetic data. However, to date few revolutionary tests have been developed to differentiate between similar phenotypes and states of disease, to monitor therapeutic progress or to assess the prognosis of individual patients.
The new research approach of proteomics, based on MS, has recently been applied to psychiatric research. Proteomics essentially refers to the systematic analysis of all expressed proteins. The ‘proteome’ describes primarily the pool of proteins encoded by the genome of an organism at a specific point in time, incorporating the set of isoforms, post-translational modifications, covalent structures and complex protein–protein interactions present therein.
Often proteomic studies are applied in tandem with transcriptomic or metabolomic approaches. Transcriptomics refers to the assessment of gene transcripts/mRNA abundance in a tissue (Hegde et al., 2003), whilst metabolomics is the quantitative and qualitative analysis of metabolites and small molecules acting in biochemical networks (Dinge et al., 2006, Godfrey et al., 2009, Oldiges et al., 2007).
In the past and at present, molecular methods such as in situ hybridization, qRT-PCR and blotting techniques have been performed to reveal abnormally expressed gene products in individuals affected by mental diseases. However, many of these methods are restricted to the detection of a few gene products at a time. Newer approaches, such as microarray technologies, allow for large-scale transcriptome analysis, but still have limitations. For example, post-transcriptional events, such as alternative mRNA splicing, and post-translational protein modifications increase the diversity of products that can be synthesized from a fixed number of genes. Post-transcriptional events may not be detected by mRNA arrays hampering efforts to reveal the full variety of gene expression (Carter et al., 2005). Post-translational modifications, which produce altered findings on the protein level (e.g. with altered position of protein spots in 2D-PAGE), may occur without detectable alterations of corresponding mRNA levels. These differential effects are a possible reason why MS and microarray results sometimes vary within the same study. In contrast to microarray approaches, proteomics has the great advantage of facilitating analysis at the protein level, thereby reflecting more closely the pathophysiological processes underlying the clinical phenomenology of specific psychiatric conditions. Proteomic tools allow for an automated, technology-driven large-scale mode of examination that provides the chance to determine the whole proteome in a given tissue without a priori assumptions about candidate molecules. The rapid development of proteomics witnessed in recent times is due to the increasing sophistication of biological MS, improvements in bioinformatics and the magnitude of data resulting from genomic sequencing of different organisms. Indeed, proteomics is complemented by functional genomics approaches that provide complete genomic sequences, allowing for protein identification by correlation of MS measurements with sequence databases.
Proteomic approaches can be divided into the main fields ‘expression proteomics’, ‘functional proteomics’, ‘structural proteomics’ and ‘interaction proteomics’. Basic research primarily deals with functional and structural aspects of proteins. However, the most common approach in psychiatry and other applied settings is the assessment of protein expression in human tissue and animal models under different conditions. As one might imagine, proteomic research in psychiatry may identify trait and state biomarkers of psychiatric diseases by comparing protein expression in patients and control subjects. The same approach is applicable to the comparison of the proteome from patients receiving different therapeutic interventions (e.g. drug versus placebo); see Figure 1. A trait marker reflects the properties of the behavioural and biological processes that play an antecedent, potentially causal role in the aetiopathophysiology of a disorder, whereas a state marker represents the state of the clinical manifestation in patients (Chen et al., 2006b). The assessment of biomarkers in easily accessible tissues such as saliva, blood, urine or CSF provides promising opportunities in diagnostic and prognostic processes. Similar phenotypes with distinct nosological entities could be compared by assessing their differential protein patterns. It might be possible to evaluate the individual prognosis of a psychiatric disease, to predict the susceptibility for a specific mental disorder and perhaps to prevent its appearance (in case of a positive family history). The analysis of proteomic profiles may further allow the prediction of drug responses/efficacy and further be used to monitor therapy. Therefore, proteomics may contribute to the development of a more personalized and specific therapy on the basis of the individual’s proteome.
Using proteomic tools, the expression levels, amino acid structure, post-translational modifications (e.g. glycosylation, phosphorylation, oxidation), interactions and functions of proteins can be detected and compared between patient-derived and control post-mortem tissue samples, and also between animals models for a psychiatric disease and the respective controls. A detailed analysis of specific protein modifications in different disease stages will further contribute to a better understanding, for example by uncovering pathomechanisms causing amyloid deposits in AD or deranged cell migration, proliferation and connectivity in other severe psychiatric disorders.
Proteomics also provides the opportunity to observe the effect of different psychotropic drugs on protein expression in post-mortem tissue, in vivo in patients and animal models, and in vitro in cell lines. Findings from proteomic research can hopefully point to new drug targets for psychiatric diseases and optimize current treatment strategies (see Figure 1).
Previous proteomic studies have often found changes in a large number of proteins, with varying results across studies and the inherent problems of lack of reproducibility. One challenge of proteomic research is therefore to deal reasonably with this multitude of observed expression changes and suggested biomarkers. There are many possibly confounding factors during a proteomics project that might cause false-positive or false-negative results. From the point of view of subject selection, there is the complication of clinical heterogeneity of biologically distinct phenotypes that cannot easily be distinguished by current diagnostic criteria. From a procedural point of view, sample collection, handling and storage are sensitive and crucial steps for a successful mass spectrometric analysis of proteins. An insufficient optimization and standardization of procedures can potentially produce false-positive results (Luque-Garcia and Neubert, 2007). Characteristic of many mass spectrometric methods, digested proteins/peptides are analysed and assigned to proteins. These methodological characteristics demand a high sensitivity and specificity of mass analysers and present complex computational and statistical challenges to cut down false protein identifications (Nesvizhskii et al., 2007). The use of search algorithms (each with their own statistical ‘scoring’ methodology) has been suggested to obtain protein identifications with confidence. Whilst each package usually applies its own threshold for the identification of proteins from peptide fragmentation profiles, it is generally accepted that ideally more than one peptide should identify a protein from a protein digest. Furthermore the Human Proteome Organisation (HUPO) has defined guidelines for accurate protein identification and is studying mechanisms for the comparison of protein identifications from different software packages. Furthermore the inclusion of raw data (fragment spectra) of LC-ESI MS/MS is recommended as supplementary material accompanying academic publications.
In proteomic biomarker research it is certainly necessary to verify preliminary mass spectrometric findings in independent, sufficiently large sample sets and to validate results with different methods such as Western blot or ELISA to confirm relevant proteomic findings. Such validation is common in microarray studies, with Northern blots and Q-PCR being used to confirm mRNA changes. Furthermore, pilot studies may be used to calculate false-positive rates, allowing for better confidence in protein identification.
Bearing all of the above in mind, proteomics research can undoubtedly provide new insight into the study of neuropsychiatric diseases, especially in the identification of protein modifications and expression changes related to these disorders.
Proteomic methods
A proteomic analysis comprises the following main work steps (see Figure 2):
(1) protein isolation from different tissues; (2) separation and fractionation of complex protein mixtures into fractions containing fewer components; (3) the analysis of the sample components by MS; and (4) the use of specific databases for data processing.
Sample preparation
As a first work step, proteins have to be isolated from the tissue (serum/plasma, CSF, specific cells or cell lines, post-mortem tissue etc.). An optimized and standardized sample collection, handling and storage procedure is essential to minimize the risk of false-positive and false-negative results. For a review on different strategies for sample preparation in the field of biomarker research, see Luque-Garcia and Neubert (2007).
Every type of tissue is dominated by a few proteins in high concentrations. Highly abundant proteins, such as structural proteins and metabolic enzymes in the brain, or albumin, transferrin, lipoproteins and immunoglobulins in the serum, hamper mass spectrometric detection of low abundance proteins. To address these difficulties, different fractionation and enrichment procedures are required prior to mass spectrometric analysis. To reduce the sample complexity of tissue, so-called subproteomes may be obtained, focusing on membrane proteins, nuclear or cellular fractions (Vercauteren et al., 2004a). On an anatomical level, methods such as laser capture microdissection allow for the precise dissection of specific nuclei and neuronal populations in brain tissue samples (Moulédous et al., 2003). Highly abundant serum proteins may be depleted using affinity chromatography-based methods, prior to further separation. The most commonly used separation techniques for proteins are 2D-PAGE and HPLC, as well as the usage of pre-coated protein chips, centrifugal filters or magnetic beads (see Figure 2). The details and relative merits of sample preparation and protein fractionation techniques are discussed elsewhere (Freeman and Hemby, 2004; Issaq and Veenstra, 2007; Issaq et al., 2002; Luque-Garcia and Neubert, 2007; Newton et al., 2004; Tannu and Hemby, 2006).
Mass spectrometry
MS has emerged as a central method to most proteomic strategies and thus in this review we are focusing on studies in which MS tools were used. Mass spectrometers for protein and peptide analysis can be configured for either ESI or MALDI, both of which are ‘soft’ ionization techniques that enable the transfer of intact proteins and nucleic acids into the gas phase without fragmentation (Aebersold and Mann, 2003, Domon and Aebersold, 2006); see Figure 3. Simple peptide samples may be examined with MALDI-ToF MS, while more complex mixtures may require separation using an analytical HPLC column coupled to the ESI MS systems (LC-ESI MS). The analysis of tryptic peptides of a protein extract using this method is often called shotgun proteomics. A variation of the MALDI technique, SELDI MS, has been developed. This approach is based on the retention of proteins of a sample via protein chip systems with pre-activated surfaces (e.g. with antibodies, receptors, ionic or hydrophobic material). The retained proteins are then ionized and detected by MS similar to the process using MALDI-ToF MS (Issaq et al., 2002; Merchant and Weinberger, 2000).
Quantitation
Quantitation using 2D-PAGE is performed using sophisticated image analysis to compare gel images for differential spot intensities of the same proteins. To reduce any gel-based variation, two-dimensional DIGE has been developed, which allows multiple protein samples to be separated on a single gel through the use of fluorescent dyes (Monribot-Espagne and Boucherie, 2002; Unlu et al., 1997). Briefly, as a first step a pooled reference standard is created from all samples, which is then labelled with one of the three fluorescent dyes (cy2, cy3 or cy5). Each individual sample is then labelled with one of the remaining dyes, and mixed for equal protein loading prior to the first and second dimensional separation. The advantage of this technique is that the pooled reference standard image can be used for alignment and normalization of all of the gels for multiple comparisons, allowing for any number of gels to be compared directly. For identification of differentially expressed proteins visualized on the gel, protein spots are excised and enzymatically digested. Identification of the enzymatic peptides may be performed by LC-ESI MS/MS or by MALDI-ToF MS. Using a relatively large amount of replicates, or utilizing the DIGE technology, the quantitation of proteins is reproducible and reliable, although labour intensive (Monribot-Espagne and Boucherie, 2002; Unlu et al., 1997).
Gel-free semi-quantitative comparison of different sample groups, such as patients and controls, has been enabled using stable isotope labelling and MS (e.g. ICAT or iTRAQ) (von Haller et al., 2003; Wiese et al., 2007). Isotopic-labelling can minimize variations during the analysis as proteins of different samples are differentially labelled prior to mixing and analysis. The isotopic labels differ solely in molecular weight and allow for the relative protein quantification. However, using isotopic labelling, only a limited number of samples can be compared and replicates are therefore often pooled. In studies of AD, the comparison of CSF protein profiles with these techniques has already proven successful (Abdi et al., 2006; Choe et al., 2007; D’ascenzo et al., 2008).
Using label-free quantitation methods an unlimited number of samples can be compared (Finney et al., 2008; Ru et al., 2006). However, with the application of these ‘gel-free/label-free’ proteomics methods, quantitation of individual proteins is more complicated. Here several proteins are digested and the tryptic peptides are mixed; thus, quantitation of an individual peptide may not necessarily reflect levels of the parent protein. It has been reported (Tang et al., 2004) that multiple sources of variation affecting the peak intensity may be introduced during the analysis, such as differences in ESI efficiencies among different peptides and samples, as well as differences in separation for replicate runs and trypsin digestion efficiency. These issues are often peptide dependent, resulting in differences in relative abundance in peptides from the same protein, making an automated approach using all peptides identified from one protein for quantitation more difficult. Nevertheless, recently it was described that applying the correction of the peptides from the same protein across all samples can indicate whether the peptide is a good candidate for quantitation of the entire protein (Schwarz et al., 2007). This has been shown to be particularly accurate for low abundance proteins.
Proteomic studies in psychiatric disorders
Alzheimer’s disease
Cell cultures
Alzheimer’s disease: cell cultures
Another goal of in vitro-proteomic studies is to test the effect of drugs on protein profiles of cell lines relevant for neuropsychiatric disorders. Sultana et al. (2006c) sought to identify individual proteins that are protected by the potential AD therapeutic D609 (an inhibitor of phosphatidylcholine-specific phospholipase and a glutathione mimetic) against Abeta(1–42) induced oxidation. Pretreatment of rat neuronal cultures with this antioxidant protected GAPDH, 14-3-3 protein zeta, PK, MDH from oxidation. Post-translational modifications and expression changes of these four proteins were also reported in animal models of AD and human post-mortem studies (Cottrell et al., 2005; David et al., 2005; Korolainen et al., 2006; Schonberger et al., 2001; Shin et al., 2004; Sizova et al., 2006; Sultana et al., 2006e, 2007; Wang et al., 2005; Woltjer et al., 2005). Thus, the antioxidant effect of D609 may indicate a possible benefit of this drug in AD treatment.
Animal models
Alzheimer’s disease: animal models
Although the aetiopathology of AD dementia is currently incompletely ascertained, there is growing evidence that Abeta deposits and the reaction of the plaque surrounding tissue play major roles in disease development and progression (Haass and Selkoe, 2007; Montalto et al., 2007). With the ambition of developing a possible vaccination against Abeta deposits, transgenic mice with mutant APP and PS1 (for more details on all animal models, see Table 2), that develop Abeta deposits in brain, were immunized with Abeta attenuates (Vehmas et al., 2001). In agreement with former reports (Schenk et al., 1999) these animals developed high serum antibody titres against Abeta(1–42) and showed reduced Abeta deposits in the brain, illustrating the potential validity of this approach.
In a 2004 study on the hippocampal proteome of APP/PS1 transgenic rats, more than 70 protein spots on 2D-PAGE showed a differential expression in the pre-plaque stage of the disease prior to the appearance of cognitive impairments (Vercauteren et al., 2004b); 44 of the identified proteins and their isoforms had been shown previously to play a role in learning and memory formation. The remaining proteins displayed functions in cytoskeletal and membrane stability, glucose metabolism, electron transport, signal transduction, protein/mRNA synthesis and modification as well as protein transport. At the same time Shin et al. (2004) published the cortical proteome of the Tg2578 mouse line with a mutant APP. In comparison to respective control brain they identified changes in concentrations of several proteins with functions in cellular energy metabolism (amongst others enolase 1, 2, ATP synthase alpha-chain, GAPDH, MDH, PK, aconitase, GFAP, DRP2). Furthermore in this study, increased oxidation and nitration of many proteins was observed. In a following similar study (Carrette et al., 2006), the up-regulation of enolase 1 was confirmed. Soreghan and colleagues used a non-gel-based quantitation method to assess oxidized proteins in an APP/PS1 mutant mouse model in comparison with aged wild-type mice (Soreghan et al., 2005). Fifty-nine of a total of 117 oxidatively modified proteins were specifically associated with the transgenic mice. These proteins mainly had roles described in signalling pathways, transcriptional regulation and vesicular trafficking. Gillardon et al. (2007) analysed the mitochondrial proteome of Tg2576 mice at the onset of cognitive impairment but before plaque deposition. Examination of synaptosomal and non-synaptic mitochondria revealed numerous changes in the subunit composition of the respiratory chain complexes I and III; however, the corresponding mRNA levels were not altered as measured by microarray analysis. Within this study brain mitochondria of young Tg2576 mice showed an impaired state 3 respiration and uncoupled respiration. Abeta oligomers were detected in synaptosomal fractions as was impaired glucose metabolism in the anterior cingulate cortex. So abnormalities in mitochondrial protein profiles and function could be detected in the pre-plaque state indicating that mitochondria are early targets of Abeta aggregates. Lewis et al. (2004) aimed to characterize Abeta(40) and Abeta(42) accumulation in mice, derived from the Tg2576 line, and in a double transgenic line. Abeta(1–40) and Abeta(1–42) were validated as the two main amyloid peptide species. The double mutation line showed in general higher SELDI-ToF MS peak intensities for Abeta(1–40) and Abeta(1–42), but no major qualitative alterations of the brain peptide profile: in the spectra of both mice lines truncated peptides Abeta(1–39), Abeta(1–38) and Abeta(1–37) appeared. In a further study the chemical features of another transgenic APP mouse model were examined (Esh et al., 2005). In these animals, a variety of C-terminally elongated Abeta peptides were detected in addition to Abeta(40) and Abeta(42), N-terminally truncated peptides and altered APP degradation. Sizova et al. (2006) studied the brain proteome associated with amyloid plaque deposition and identified 15 proteins which were significantly altered in a slightly different transgenic mouse to that used by Esh et al. (2005). These identified proteins play a role in glial response/inflammatory and oxidative processes, in cholesterol metabolism and neuronal signal transduction.
Aside from elevated Abeta(42) peptides, there are also increased levels of Abeta binding alcohol dehydrogenase (ABAD), which contains an intracellular binding site for Abeta. Over-expression of Abeta and ABAD in transgenic mice results in an enhanced neuronal cytotoxicity with subsequent changes in spatial learning memory (Lustbader et al., 2004). Yao et al. (2000) used this animal model to identify the antioxidant protein Prdx 2 as being consistently up-regulated in the transgenic animals as well as in human AD post-mortem brain. It was also observed that neuronal cell lines transfected with Prdx 2 were better protected against toxic Abeta levels than mock transfected cells. Therefore, it can be suggested that the up-regulated Prdx 2 expression in AD reflects part of the cellular response to protect the brain from further oxidative damage.
The other major histopathological AD hallmark, apart from amyloid deposits, is tau-containing neurofibrillary tangles. David and colleagues analysed a transgenic mouse line that expresses mutant tau protein (David et al., 2005). Proteomic techniques and functional analyses revealed that tau accumulation induces mitochondrial dysfunction, with a diminished capacity in electron transport and a significant reduction in ATP levels. Further results imply a metabolic dysregulation, modifications in the oxidative state of the brain and a synaptic pathology in the presence of the mutant tau. In a second study of this group, Abeta(42)-induced effects on the proteome were studied in the mutant tau animal model, and also in an in vitro cell model (David et al., 2006). Many of the differentially expressed proteins in these two models are linked to energy metabolism, stress response and protein folding, proteasome degradation, and transcriptional processes. Analogous findings were also found in human AD post-mortem brain (David et al., 2006).
Apolipoprotein (apo) E knock-out mice have also been used as animal models in AD research (Choi et al., 2004a), as studies have shown an influence of APOE gene variants on the oxidation status in the brain (Ramassamy et al., 1999). In ApoE knock-out mice an approximately two-fold increase in hippocampal protein oxidation was reported along with an elevated sensitivity for oxidation of specific proteins (GFAP, CK B, disulfide isomerase, chaperonin subunit 5, DRP2, 75 kDA glucose-regulated protein).
In another study the impact of APO gene variants on the hippocampus proteome was assessed in APOE targeted replacement (TR) mice (Osorio et al., 2007). The level of the chaperone protein 75 kDA glucose-regulated protein (mortalin) was found to be differentially expressed in APOE3 and APOE4 mice, and in human AD post-mortem tissue APOE levels were correlated with APOE genotypes APOE 3/3 and APOE 4/4. These two studies support previous findings that the apoE gene product regulates the cellular response mechanisms on oxidative stress associated with AD pathology.
In AD patients, increased levels of acrolein has been detected (Lovell et al., 2001). Acrolein (2-propen-1-al), the most reactive of the alpha, beta-unsaturated aldehydes (Esterbauer et al., 1991) is present in automobile exhaust gases, but is also endogenously produced during lipid metabolism (Adams and Klaidman, 1993). Mello et al. (2007) sought to identify oxidized proteins in gerbils after application of acrolein. Acrolein increased protein oxidation in a dose-dependent manner and the identified oxidized proteins were linked to energy metabolism, protein synthesis, neurotransmission and the cytoskeleton.
Many findings of AD animal models are in accordance with abnormalities seen in post-mortem tissue of AD patients. Therefore, animal models seem to be valid tools in AD research seeking to characterize Abeta isoforms implicated in AD and their functions, to test the effect of potential vaccinations and drug treatments and to identify oxidized proteins that play a role in the disease pathophysiology.
Post-mortem studies
Alzheimer’s disease: post-mortem studies
Amongst the proteins that exhibited changes in their oxidation status are also components of glutamate metabolism, such as glutamine synthetase and GDH (Butterfield et al., 2006b; Korolainen et al., 2006). Glutamine synthetase regulates extra neuronal glutamate levels and thereby neurotransmission. An impaired activity of this enzyme due to the observed increase in oxidation would result in an augmented neurotoxicity and neurodegeneration as glutamate synthetase converts the potential neurotoxin glutamate into glutamine. Interestingly, there is an increase in total amount of soluble GDH in AD, but a decrease in its oxidation (Korolainen et al., 2006). It is thought that the GDH pathway is mostly active when the glucose levels are low. Thus, these reports complement the findings of an altered energy metabolism in AD.
UCHL1 is part of the ubiquitin-proteasome system and plays an important role in the degradation of misfolded, damaged and short-lived proteins. An overload of this system by undegradable molecules together with a functional impairment due to oxidation might eventually result in the accumulation of abnormal proteins and in neuronal degeneration (Butterfield, 2004; Butterfield et al., 2006b). These suggestions are in agreement with recent proteomic findings of dysregulation of the proteasome, as well as the established finding of abnormal deubiquitination and consecutive aggregation of damaged proteins in AD dementia (Chen et al., 2006a).
Proteins involved in axonal growth and cytoskeletal regulation have also been reported to be oxidized in AD: DRP2 has been proposed to modulate neuronal plasticity in the aging brain, and altered DRP function in AD results in cytoarchitectural changes, such as reduced dendritic length (Butterfield et al., 2006b; Coleman and Flood, 1987; Lubec et al., 1999). A dysregulation of DRP2 has been repeatedly observed in human AD brain tissue (Pamplona et al., 2005; Schonberger et al., 2001; Tsuji et al., 2002). Another target of elevated oxidation, actin beta, is a key player in cytoskeletal pathways and has been shown to rapidly influence the shape of dendritic spines. Impaired function of this protein would be consistent with the observed abnormalities in synaptic plasticity and consequential reduced memory function in AD (Butterfield et al., 2006b). Expression changes or increased oxidation could be detected for several other cytoskeleton- and membrane transport-associated proteins such as actin binding protein, tau, vimentin, dynamin 1, dynein, tubulin isoforms, neurofilament triplet L or stathmin (Cheon et al., 2001; Liao et al., 2004, Pamplona et al., 2005). Cytoskeletal proteins are also potentially relevant targets for processes that cause decreased protein solubility in early AD (Wang et al., 2005; Woltjer et al., 2005).
Another prime candidate for proteomic analysis in AD is carbonic anhydrase II, an enzyme which is preferentially oxidized and catalyses the reversible hydration of CO2, a reaction fundamental to many cellular and systemic processes including glycolysis and acid and fluid secretion. Thus, carbonic anhydrase II represents an essential enzyme for general brain function and a potentially important target for post-translational modification in AD.
The supposition that post-translational modifications, such as oxidation and nitration, significantly alter protein function has already been proven correct for several of the above-mentioned proteins (Butterfield et al., 2006b). Apart from protein oxidation and nitration, changes in other post-translational modifications such as S-glutathionylation, phosphorylation and glycosylation have been found in AD brain tissue (Kanninen et al., 2004; Korolainen et al., 2005, Newman et al., 2007). Therefore, proteomic approaches, combined with other study designs have been deployed successfully to test the hypothesis that increased oxidation and nitration together with other post-translational modifications are involved in the complex mechanism of brain pathology in AD disease.
Normal aging is associated with increased oxidative stress in the body and an increased probability of developing neurodegenerative disorders (Squier, 2001). Antioxidative proteins, such as peroxiredoxins and Cu/Zn SOD protects cells from oxidative damage. Peroxiredoxins are ubiquitously in all living organisms and remove cellular hydrogen peroxide and SOD catalyses the dismutation of superoxide radical to hydrogen peroxide (Kim et al., 2001). In AD post-mortem brain, elevated concentrations of Prdx 1 and 2, as well as of SOD were reported (Kim et al., 2001; Krapfenbauer et al., 2003; Melanson et al., 2006; Schonberger et al., 2001) providing evidence for compensatory response mechanisms to increased oxidative stress and cell loss in AD.
In AD post-mortem tissue altered concentrations of proteins linked to amyloidosis and cell death were found. Proteomic characterization of amyloid plaques confirmed the presence of Abeta peptides, alpha1-antichymotrypsin, apoE, collagen type XXV, cystatin C, alpha-synuclein and proteoglycans in plaques (Liao et al., 2004). Liao et al (2004) also reported an enrichment of 26 proteins of the following categories in amyloid plaques: cytoskeletal proteins, chaperones, and proteins associated with membrane trafficking, inflammation, proteolysis and phosphorylation/dephosphorylation. Sergeant et al. (2003) performed a qualitative study on aggregated Abeta peptides in human AD brain. Their results show that in the earliest stage of AD it is the amino-truncated Abeta(42) peptides that seed amyloid deposition and therefore this Abeta species seems to be instrumental in basic mechanisms of plaque formation. A dysregulation of Abeta peptides as a major hallmark of AD pathophysiology has been confirmed in human AD brain by other proteomics studies (Lewczuk et al., 2004; Yang et al., 2007). Proteomics tools, coupled with the use of laser capture micro-dissection has allowed for the characterization of proteins in neurofibrillary tangles (Wang et al., 2005). Seventy-two of 155 detected proteins could be identified leading to the discovery of new proteins that had not yet been associated with this central feature of AD neuropathology (see Table 3).
Chaperone proteins have been strongly implicated in AD as these play important roles in proper protein folding and maturation, in protein degradation and transport across membranes (Mayer and Bukau, 2005). Yoo et al. (2001b) reported aberrant expression patterns of six out of nine chaperone proteins analysed in AD: mitochondrial HSP 60 kDA, HSP 70 kDA 4, heat shock cognate 71 kDA, alpha-crystalline B chain, glucose regulated protein 75 and 94 kDA. Several further studies support the relevance of altered chaperone or chaperone-associated protein expression to AD pathogenesis: altered levels of mitochondrial HSP 60 kDA (Tsuji et al., 2002), GFAP (Liao et al., 2004; Tsuji and Shimohama, 2001; Tsuji et al., 2002); increased GFAP glycosylation (Kanninen et al., 2004), increased S-glutathionylation of alpha-cristallin B (Newman et al., 2007), increased oxidation of heat shock cognate 71 kDA (Castegna et al., 2002b), peptidyl prolyl cis-trans isomerase (PPIase) (Sultana et al., 2006a,b) and GFAP (Pamplona et al., 2005) and increased nitration of HSP 70 kDA (Sultana et al., 2007) have all been reported in AD. Furthermore HSP 90 kDA alpha, heat shock cognate 71 kDA, alpha crystallin B2, GFAP, alpha spectrin were observed in complexes with APP in human brain lysates (Cottrell et al., 2005). Given that modifications in the function of chaperones can cause major effects on the targeted proteins, these findings support the hypothesis that dysfunction of protein folding/degradation contributes to AD pathology (Cottrell et al., 2005; Sultana et al., 2006a,b).
Proteomic analysis of synaptic components have been conducted in AD post-mortem brains (Schonberger et al., 2001; Yoo et al., 2001a). Yoo et al. (2001a) found decreased beta-SNAP expression in temporal cortex of AD patients, whilst synaptotagmin I (p65) was reduced in cerebellum, temporal and parietal cortex and synaptotagmin I (pI 7.0) was down-regulated in temporal, parietal cortex and thalamus. Schonberger et al. (2001) reported that amongst 76 differentially expressed proteins were the synaptic proteins vesicular fusion protein NSF, synaptotagmin, transformation sensitive protein and phosphatidylethanolamine-binding protein. Dysregulated synaptosomal proteins and an increased oxidation of synaptic proteins, such as gamma-SNAP (Sultana et al., 2006b) support previous reports of impaired synaptogenesis and/or synaptosomal loss as a consequence of neuronal loss in the AD brain.
CSF and peripheral biomarkers
Alzheimer’s disease: CSF biomarkers
Apart from the often-used MALDI-ToF MS, SELDI-ToF MS was also used in several AD CSF studies. Carrette et al. (2003) proposed a combination of five polypeptides as disease markers: cystatin C, two beta 2-microglobulin isoforms, an unknown 7.7 kDA polypeptide, and a 4.8 kDA VGF polypeptide. CSF changes for beta 2-microglobulin had already been described by Davidsson et al. (2002) and Puchades et al. (2003), using different MS methods. These results were confirmed by two further studies (Hu et al., 2005; Simonsen et al., 2007b,c). As a component of the class I major histocompatibility complex, beta 2-microglobulin plays a role in the balance between membrane protein turnover and elimination (Hoekman et al., 1985). In vitro, partially folded beta 2-microglobulin represents a key intermediate in the generation of amyloid fibrils (Hong et al., 2002). Changes in cystatin C CSF expression were replicated by Choe et al. (2007), Hu et al. (2005) and Simonsen et al. (2007c,d). Cystatin C seems to be the main cysteine proteinase inhibitor in most human biological fluids (Grubb, 1992) and influences Abeta oligomer formation, as reported in a recent in vitro study (Selenica et al., 2007). It might be suggested that cystatin regulates the transformation of monomeric Abeta to larger and perhaps more toxic molecular species in vivo.
Abeta peptides in CSF have also been tested for their potential as disease markers. Lewczuk et al. (2003, 2004) described changes in the pattern of CSF Abeta peptides and, concordantly to some previous data, report lower CSF Abeta(42) concentration in AD. According to further data hitherto unreported Abeta peptides seem to play a role in AD pathogenesis. Maddalena et al. (2004) compared the mass profiles of CSF Abeta peptides of AD patients and healthy controls and detected the following peptides in both groups: Abeta(2–14), Abeta(1–17), Abeta(1–18), Abeta(1–33), Abeta(1–34), Abeta(1–37), Abeta(1–38), Abeta(1–39), Abeta(1–40) and Abeta(1–42). A decrease of the Abeta(1-38) level was seen in the patient group. The extent to which quantitation of CSF Abeta varies, as influenced by different analysis methods, was recently discussed by Simonsen et al. (2007a). The group compared different SELDI-ToF MS methods and ELISA for the Abeta determination. Elevated Abeta levels in test samples could only be observed by one of the SELDI-ToF arrays. These findings might explain varying results in AD research and therefore caution is demanded for the comparison of data derived from different study designs. In another SELDI-ToF MS study of this group the CSF proteome of patients with a stable MCI and patients with MCI progressing to AD were compared (Simonsen et al., 2007b). The authors identified a panel of 17 putative biomarkers for the prediction of progression from MCI to AD. The same group determined the CSF protein profiles in 95 AD patients and 72 healthy controls (Simonsen et al., 2007c). Amongst the potential marker proteins with increased concentrations were Abeta(1–38) and Abeta(1–40), the integral membrane protein 2B and the complement component C3a anaphylatoxin; amongst the down-regulated proteins were the above-mentioned beta 2-microglobulin and apoC1 and, furthermore, neurosecretory protein VGF and cystatin C. In another recent study this group verified a panel of protein biomarkers for the differentiation between AD and patients with frontotemporal dementia (Simonsen et al., 2007d). Zhang et al. (2005) were able to quantify relative levels of 163 proteins in AD patients and controls. In half of the proteins the AD/control ratio differed more than 20% (see Table 4), with the most consistent finding being up-regulation of APP and cathepsin B in AD. The abnormally expressed proteins play a role in inflammation/immune reaction, extracellular matrix/adhesion processes, neurotransmission, signalling cascades, apoptotic and metabolic processes.
In 2007 Finehout and co-workers published a panel of CSF proteins that differentiated among AD, healthy controls and control individuals with neurological diseases. As observed in several studies, the identified proteins are linked to inflammatory processes, amyloid transport and proteolytic inhibition. This study has been criticised because it neither controlled for blood contamination in CSF nor for age-related changes in control samples (Zhang and Montine, 2007).
The occurrence of increased protein oxidation in CSF of probable AD patients with MCI was studied by Korolainen and colleagues in order to complement findings from post-mortem tissue (Korolainen et al., 2007). In general the carbonylation levels did not vary between the probable AD patients and controls subjects and only two proteins showed regulation: lambda chain was up-regulated in AD, and an unidentified protein down-regulated. However prominent difference between the carbonylation status between women and men was seen, with a higher carbonylation of Vitamin D-binding protein, apoA1 and alpha 1-antitrypsin in men.
In some proteomics CSF biomarker studies, modern labelling techniques, such as iTRAQ, have been used. Abdi, Pan and colleagues identified down-regulation of neurexin 1 (involved in pre-synaptic neurotransmitter release) and an increase of the acute-phase protein alpha 1-acid glycoprotein (Abdi et al., 2006; Pan et al., 2008). Choe and co-workers (2007) determined the change of protein profiles of two AD patients undergoing a 6-month treatment with IV Ig and 3 months of drug washout by iTRAQ labelling. In addition to an initial clinical improvement, consistent protein alteration during treatment (with opposite changes during washout) could be seen for several proteins that are supposedly linked to AD. Even though only two individuals were included, such a study design seems in general useful to reduce the risk of false-positive results, as each individual can serve as his/her own control.
Taken together, proteomic approaches have consistently identified putative AD CSF markers that play roles in amyloid homeostasis, protein transport and turnover (such as Abeta variants, transthyretin, beta 2-microglobulin, cystatin C, hemopexin), apolipoprotein metabolism (apoA1, apoC1, apoE, apoJ), inflammation reaction (alpha-1 antichymotrypsin, complement factors) and synaptic processes (VGF fragments). However, in synopsis of all results from proteomic CSF studies the variability of results is striking. Differences in sample preparation, recruited patients, applied separation and MS techniques as well as different types of used databases for protein identification might cause these controversial findings. To address this topic of inconsistency in CSF biomarkers results, Hu and colleagues studied the variability of CSF protein profiles within one person and between different individuals (Hu et al., 2005). In an undoubtedly small sample of six people, the authors discovered proteins with a significant fluctuation of abundance within the same individual. In general they observed a greater profile similarity between CSF samples from the same individual compared with samples from different individuals. Amongst the proteins with higher intra-individual variability several had been suggested as potential AD markers (transthyretin, ubiquitin and apoE apo H precursor, alpha 2 macroglobulin, transferrin and albumin). Therefore, results from CSF biomarker studies have to be considered with caution bearing these limitations in mind.
Alzheimer’s disease: peripheral biomarkers
In 2005 a study utilizing a proteomic approach sought to identify a mass peak pattern, not a single protein or a pattern of a few identified proteins, that could differentiate between groups with AD patients, patients with MCI and individuals without cognitive impairments (Lopez et al., 2005). In a blinded trial these authors were able to classify the named groups with the help of a protein pattern using a specific algorithm. German and colleagues used a very similar analytical platform to determine a marker pattern in the sub-set of proteins that bind to carrier molecules such as albumin (German et al., 2007). A differentiation between AD and PD patients, as well as control subjects was possible with a marker pattern of four peaks. Three of these peaks were identical to three of those observed in Lopez et al.’s (2005) study.
Liu et al. (2006) have confirmed an apolipoprotein dysregulation in AD reporting lower serum apoA1 levels in AD. Hye et al. (2006) described that image analysis of the plasma protein gel pattern alone allowed identification of AD patients with 56% sensitivity and 80% specificity. Of 15 altered gel spots, 13 contained some component of Ig or serum albumin precursor. In a very recent approach Ray et al. (2007) measured the abundance of 120 known signalling proteins by ELISA and presented a panel of 18 proteins that demarcated AD patients and controls. Furthermore these authors were able to differentiate between MCI patients that later developed AD and MCI patients without progression. The observed alterations point to a dysregulation of hematopoiesis, inflammation, neuroprotection, neurotrophic activity, phagocytosis and energy homeostasis in AD.
As the existing data shows considerable variability, further efforts are necessary to define universal standards in biomarker research so that studies become more comparable. Current results certainly have to be verified in different cohorts, addressing reproducibility, effectiveness and disease specificity. In the future the design of reliable diagnostic tests will probably embed a combination of several biomarkers.
Drug addiction
Cell cultures
Drug addiction: cell cultures
Animal models
Drug addiction: animal models
As the mechanisms of nicotine addiction have not yet been completely understood, Yeom and co-workers analysed the proteome of the striatum in nicotine-addicted rat brain, as this region, together with the nucleus accumbens, is postulated to play a pathophysiological role in addiction (Yeom et al., 2005). Increases in the levels of zinc-finger binding protein-89 after repeated nicotine administration were found, and as a protein that induces growth arrest, these changes might reflect apoptosis induction by nicotine. Further proteins that are involved in cell survival and apoptotic processes were differentially regulated, such as death effector domain-containing DNA binding protein and BDNF. These findings contribute to a better understanding of the effects of nicotine administration at the molecular level. In which ways these findings can finally be translated and used for progress in nicotine addiction therapy remains open.
Kim et al. (2004, 2005), in two studies, determined the protein pattern in rat frontal cortex under chronic opioid administration (for either butorphanol tartrate or morphine) and focused on one specific post-translational modification, the phosphorylation of tyrosine residues. They reported multiple cases of altered, in most cases increased, expression of proteins with phosphorylated tyrosine in opioid-dependent rats. The abnormally expressed proteins had functions in intermediary metabolism (e.g. gluthamine synthetase), cell cytoskeleton/differentiation (e.g. tubulin, actin beta) and cell signalling (e.g. GTP-binding proteins). In another study on the effect of chronic morphine administration on rat brain, the subproteome of the synaptic plasma membrane was analysed (Prokai et al., 2005). Amongst several differentially expressed synaptic membrane proteins, the alpha 3 subunit of the Na+/K+ ATPase pump was identified; this enzyme had already been reported to be involved in the effects of chronic morphine dependence (Biser et al., 2002). Another group has investigated the effect of chronic morphine administration in different brain regions. In a first study they assessed the whole rat brain proteome and identified several potential dependence markers (Bierczynska-Krzysik et al., 2006a). The reported proteins are mainly cytoplasmic and mitochondrial enzymes, others belong to GTPase and gluthation S-transferase superfamilies, ATPase, asparaginase or proteasome subunit p27 families. In a further approach, this group analysed the cerebral cortex, hippocampus and striatum protein profiles (Bierczynska-Krzysik et al., 2006b). In total 26 altered proteins were identified differentially in the three brain regions. In agreement with previous studies, regulation in bioenergetics pathways (reflected by a differential expression of ATP synthase), cell metabolism, protein handling, signalling and oxidative stress were seen. Li and co-workers assessed the nucleus accumbens protein profile in rats after repeated treatment with morphine (Li et al., 2006). Under repeated administration the authors reported an augmentation of neurofilaments and changes in actin beta post-translational modifications. Furthermore proteins involved in neurotransmission, energy metabolism and protein degradation were regulated. In a further study after chronic, but not acute morphine administration, the hippocampal proteome of mice showed a down-regulation of three metabolic enzymes (Fe-S protein 1 of NADH dehydrogenase 2, pyruvate dehydrogenase complex E2 subunit, LDH 2) (Chen et al., 2007). The simultaneous intake of morphine and the opioid antagonist naltrexone inhibited this down-regulation. Moron and coworkers examined mouse hippocampus after chronic morphine administration, but analysed only proteins of the post-synaptic density (Moron et al., 2007). A total of 102 proteins could be identified in this hippocampal subproteome, amongst them 10 with differential expression. Clathrin, which plays a role in endocytosis, showed increased levels to the largest extent. Further study results additionally point to a redistribution of endocytic proteins at the synapse and, as a consequence, the modulation of synaptic plasticity at excitatory synapses in the hippocampus by morphine. Yang et al. (2007) studied chronic morphine dependence on the prefrontal rat cortex proteins. The 58 identified modulated molecules belonged to the following classes: bioenergetic pathways, signal transduction, synaptic transmission, cytoskeleton, chaperones, and local synaptic protein synthetic machinery. Although most changes were described for the first time, several of the identified proteins had already been reported in morphine-dependence research (ATPase H+ transporting V1 subunit B isoform 2, guanine nucleotide-binding protein beta polypeptide 2-like 1, guanine nucleotide-binding protein beta-2 subunit, cyclic nucleotide phosphodiesterase 1, GDH 1, calcium/calmodulin-dependent protein kinase II alpha, CK B, lamin-A) (Bierczynska-Krzysik et al., 2006b; Kim et al., 2005; Li et al., 2006; Neasta et al., 2006; Prokai et al., 2005).
In addition to the effects of cocaine, nicotine and opiates, the influence of psychostimulants on protein profiles was also examined by means of proteomics. Iwazaki et al. (2006) aimed at identifying protein expression profiles in the striatum after an acute low dose of MAP, as this brain region is involved in the locomotor response to this drug. Amongst the differentially expressed proteins, the authors found most proteins linked to mitochondrial function (phosphoglycerate kinase 1, ATP synthase F(1) beta chain and enolase 1). These findings are in line with previously reported MAP effects on mitochondria (Davidson et al., 2001). Further modified proteins are related to apoptotic (dynamin-like protein 1) and cytoskeletal processes (actin beta), oxidative stress (Prdx 2, 5) and protein degeneration (UCH L1) or are markers of cell damage (CK B, LDH). It seems that these systems are regulated in the striatum by a single, small MAP dose. In a second study this group examined the increased behavioural MAP response and the striatal protein pattern during subsequent dose exposure. A repeated MAP intake caused, as described for morphine before, behavioural sensitization (MAP-induced sensitization has also been used as an animal model for MAP-induced psychosis and schizophrenia). Only two proteins overlapped between the acute dose and repeated dose study, but a congruent regulation of mitochondrial, oxidative stress and apoptotic proteins was seen, implicating neuronal stress and/or neurotoxicity. The observed expression changes of synaptic and cytoskeletal proteins could reflect modulations of pre-synaptic function/plasticity, microfilament turnover and axonal growth during MAP-induced sensitization. Assessing the impact of amphetamine on the hippocampal proteome, Freeman and colleagues compared the profile in rats naive to amphetamine, during a self-administration session, during voluntarily abstinence and after reinstatement of self-administration (Freeman et al., 2005). The authors revealed a crucial role of actin and other cytoskeletal proteins during abstinence and suggested that alterations in the neuronal tone could affect neurotransmission and finally behaviour.
It is hoped that the identification of drug-induced modifications in neuronal plasticity and signal transduction will lead to a clearer understanding of mechanisms of drug dependency, which in turn will assist in the design of specific drugs for withdrawal therapy.
Post-mortem studies
Drug addiction: post-mortem studies
The findings of these first proteomic post-mortem studies in drug addiction research make valuable contributions to a better understanding of the multitude of protein pathways and interactions that are modulated by cocaine, alcohol and other abused drugs.
Peripheral markers
Drug addiction: peripheral biomarkers
Schizophrenia
Animal models
Schizophrenia: animal models
In a 2006 study on protein expression under different antipsychotic treatments La et al. (2006) used chlorpromazine and clozapine in healthy Sprague–Dawley rats. After antipsychotic treatment, the hippocampal proteome differed in levels of MDH, Prdx 3, vacuolar ATP synthase subunit beta and mitogen-activated protein kinase kinase 1. The pattern of protein regulation differed between the typical and atypical antipsychotic, only the reduction of MDH was a consistent finding in both groups. The mitochondrial matrix protein MDH is a key enzyme in the malate shuttle system and facilitates the conversion of malate to oxaloacetate and the replenishing of oxaloacetate levels by reductive carboxylation of pyruvate (La et al., 2006). A role of this enzyme in the pathophysiology of schizophrenic psychosis is supported by the above-mentioned findings in MK-801 treated rats (Paulson et al., 2003, 2004a), by human post-mortem microarray studies (e.g. Middleton et al., 2002; Vawter et al., 2004) and also by proteomics post-mortem studies (Prabarakan et al., 2004). One could possibly conclude that neuroleptics exert their antipsychotic effect partially by targeting MDH. Such observations could in the long term be helpful in the design of more effective and specific pharmacological agents.
Recently a different animal model of schizophrenia-like behaviours, neonatal ventral hippocampal lesions in rats, has been investigated (Vercauteren et al., 2007). These authors have established a protocol to enrich plasma membrane- and vesicle-associated proteins in the sample. Eighteen of 392 2D-PAGE spots of prefrontal cortex proteins showed intensity variations in lesioned rats compared with controls. The majority of the dysregulated proteins were associated with various neurotransmitter systems, with major roles in plasma membrane receptor and synaptic vesicle turnover (such as syntaxin-binding protein 1b). Subproteome-based proteomic methods as utilized by Vercauteren et al. (2007), when compared with whole cell proteomic approaches, help to reduce interference from highly abundant proteins in the sample. Thus, these subproteomic approaches allow for a better analysis of low abundance proteins, thereby increasing the chance to unravel the complex protein dysregulation and interaction in schizophrenic psychosis.
Post-mortem studies
Schizophrenia: post-mortem studies
Presently, proteomics studies in schizophrenia have revealed a number of protein systems that may be involved in the condition, with some of the proteins being identified in more than one study (such as DRP2, fructose-biphosphate aldolase C, CK B, GFAP, HSP 70 kDA protein 1). Differences in published results might be due to heterogeneous patient samples, diverse analysed brain areas and different MS methods.
CSF and peripheral biomarkers
Schizophrenia: CSF and peripheral biomarkers
In several studies, abnormalities in the expression of apolipoproteins in schizophrenia were reported as already addressed. The above-mentioned down-regulation of apoA1 in plasma of schizophrenic patients (Yang et al., 2006) was confirmed by La et al. (2007) in a small patient group under treatment with clozapine or chlorpromazine. The authors had focused on the expression of different apolipoproteins after initially finding increased levels of the apoA1, apoA4 and apoE in plasma of Sprague–Dawley rats treated with chlorpromazine but not with clozpine. La and co-workers speculate that the down-regulation of apoA1 might be associated with the pathology of schizophrenia and that chlorpromazine increases apoA1 expression as part of its therapeutic action. A regulation of apoA1 in schizophrenia was also observed in another study (Huang et al., 2007b). Increased apoA1 levels in CSF, serum, post-mortem liver and brain tissue and red blood cells could be seen consistently through different methodological approaches. Although all these studies imply an association between the pathophysiology of schizophrenia and apolipoproteins, there are also hints that these alterations emerge just as side effect from antipsychotics or are influenced by confounding factors such as nicotine usage (Huang et al., 2007b); hence, there is still need for continuative research to clarify this topic.
In a very recent study Prabakaran et al. (2007) published corrobative data on their previous post-mortem findings (Prabakaran et al., 2004) that metabolic alterations followed by oxidative stress are linked to the disease process of schizophrenic psychosis. The group studied the proteome of liver and red blood cells and discovered six of 14 discriminating proteins in the liver and four of eight altered red blood cell proteins associated with oxidative stress. The observed changes in peripheral protein patterns, in CSF, serum or other tissues have certainly to be tested for reproducibility, but show potential to find surrogate disease markers in easily accessible tissues.
Mood disorders
Cell cultures
Mood disorders: cell cultures
Animal models
Mood disorders: cell cultures/animal models
Investigating the effects of antidepressant medication on protein profiling in the rat hippocampus, Khawaja et al. (2004) reported 33 differentially regulated proteins after a 2-week treatment with either venlafaxin or fluoxetine. Amongst these modified proteins were those involved in neurogenesis, in outgrowth/maintenance of neuronal processes and in neural regeneration/axonal guidance collapsin response mediator protein systems. Other modulated proteins were associated with neuronal vesicular cell trafficking and synaptic plasticity, with neurosteroidogenic responses, and possible anti-apoptotic pathway-mediated regulatory events. In a following similar study design, the rat hippocampus and frontal cortex proteome were analysed after chronic treatment with fluoxetine and two putative novel antidepressants: an NK1 receptor antagonist (GR205171) and a corticotropine-releasing factor receptor 1 antagonist (DMP696) (Carboni et al., 2006b). All treatments resulted in modified levels of actin isoforms, whereas both fluoxetine and GR205171 caused decreased synapsin II levels. Fluoxetine intake augmented extracellular signal-regulated kinase 2 and transgelin 3 and decreased vacuolar ATP synthase. After GR205171 treatment, protein disulphide isomerase A was down-regulated, whereas dynamin 1 and aldose reductase were increased. Application of DMP696 influenced PK, LDH, DRP2 and ATP synthase concentrations. Although this report revealed a specific pattern of protein modulation for each pharmacologically active compound, the authors suggested that all antidepressants share the ability of modulating neural plasticity.
Post-mortem studies
Mood disorders: post-mortem studies
Depression is often associated with suicidal behaviour. In a recent post-mortem comparative study the prefrontal cortex proteome of suicide victims and controls was assessed (Schlicht et al., 2007). Five 2D-PAGE proteins spots differed significantly in intensities between both groups, the following three appearing only in suicide victims: manganese superoxide dismutase, alpha-crystallin B chain and GFAP. Manganese superoxide dismutase, as a major antioxidant enzyme, protects cells against oxidative stress, alpha-crystallin B chain belongs to the low molecular HSP proteins and GFAP is reported to be involved in astrocytic activation in gliosis. Whether the observed expression changes in proteins connected with glial function, neurodegeneration and neuronal injury have a real impact upon suicidal behaviour, or whether they just occurred as an effect of previous medication of the suicide victims, has to be ascertained.
Peripheral biomarkers
Mood disorders: CSF biomarkers
Anxiety disorders
Animal models
Anxiety disorders: animal models
Autistic spectrum disorder
Post-mortem studies
Autistic spectrum disorder: post-mortem studies and peripheral biomarkers
Peripheral biomarkers
To date no disease markers for the diagnosis of autism have been validated and diagnostic procedures are mainly based on the observation of behavioural abnormalities. However, a reliable protein biomarker would facilitate an early and exact ASD diagnosis, a crucial precondition for an early behaviour-modifying therapeutical approach. Recently Corbett et al. (2007) analysed the serum proteome in a group of autistic children aged four to six years for differentially expressed proteins (see Table 18). The authors of this LC-MS/MS study reported altered levels of the following apolipoprotein and complement factors: apoB100, complement factor H related protein, complement C1q and fibronectin 1. The findings of this first MS-based study on peripheral markers are in line with former studies, that revealed alterations in immune modulating and inflammation systems (e.g. complement) in ASD patients (Molloy et al., 2006; Nelson et al., 2006; Zimmerman et al., 2005).
Discussion
Our overview on widely used proteomic methods and currently available data reflects new developments and opportunities in modern neuropsychiatric research. Proteomic analysis of brain compartments and peripheral tissue, of cell and animal models, is now a promising tool to better understand the complexity of brain disorders and drug effects as well as to search for corresponding early disease markers.
Despite the encouraging progress in proteomic technologies this approach is still in a developmental stage with pitfalls and hurdles to overcome. In general, in a given tissue, a proteome of millions of proteins could be expected, but so far only a small fraction has been detected with current proteomic methods. One reason for this is that certain characteristics of brain molecules complicate their analysis. Proteins of interest in CNS are often transmembrane- and membrane-associated proteins, including ion channels, G proteins and receptors. These molecules unfortunately are rather insoluble. To give consideration to these limitations, proteomic methods in neuropsychiatric research have to be specially adapted. The widely used 2D-PAGE imposes clear limitations in representation and detection of these insoluble proteins, as well as for the detection of low abundance gene products, proteins of small size, hydrophobic, acidic and basic proteins, all of which often escape 2D-PAGE analyses. Furthermore, neuropeptides and proteins only exist in very low concentrations, in addition, the amount of available brain and CSF tissue is limited.
Every type of tissue is dominated by few proteins in high concentrations that hamper mass spectrometric detection of low abundance proteins. To address these difficulties, different fractionation and enrichment procedures are inevitable prior to mass spectrometric analysis, as explained in more detail in the methods section. A sophisticated and efficient sample prefractionation and the recovery of more homogenous sample fractions/subproteomes prior to MS will eventually lead to the reward of higher information yield. Apart from adapting 2D-PAGE methods, non-gel-based techniques such as LC-ESI MS will gain in importance as they seem more suitable for the analysis of interesting protein populations in brain research. As a progressive method, nanoscale capillary LC-ESI MS systems actually allow the detection of very small protein or peptide amounts.
In addition to advances in prefractionation methods, improvements in the reliability of mass spectrometers are desirable to allow proteomics to enter in a clinical setting where only small tissue samples are available. Strong efforts are also needed in the field of proteomics data processing. A systematic study in 2005 tested various algorithms currently in use for MS/MS data analysis (Kapp et al., 2005). The authors of this study compared five search algorithms with respect to their sensitivity and specificity, and evaluated them based on specified false-positive rates. As the algorithms still displayed weaknesses in sensitivity or specificity, the authors suggested the use of multiple (at least two) search algorithms to reduce the number of false-positive identifications. According to these results there is still room for improvement regarding proteomic software tools.
With the rapid growth in the research area of proteomics vast quantities of biological data (especially in association with other ‘omics’ approaches such as metabolomics or lipidomics) are generated. It is crucial to think about efficient data-mining technologies and the establishment of international, public databases. As a general goal, multicentre collaborations are desirable to obtain large-scale data from multiple levels of analysis and to integrate these data. This requires a joint effort and collaborations between clinical researchers of the medical sector and academia from chemistry, biochemistry, cell biology and molecular genetics alongside specialists in bioinformatics and statistics. ‘SwissProt’ represents an already available database that provides reliable protein sequences associated with a high level of annotation (such as protein function, domains structure, post-translational modifications, variants, etc.), a minimal level of redundancy, the usage of standardized nomenclature and a high level of integration with other databases. In this context also HUPO has to be mentioned. This project (see http://www.HUPO.org) aims to identify all protein isoforms in cells (in health and disease state) of organisms such as humans or mice. In the sub-project the Human Brain Proteomics Project (HBPP), the human and mouse brain proteome shall be characterized and the data will be compared with mouse models of human disease and to relevant autopsy materials from human neurodegenerative diseases. A further important aim of the HUPO is to define standards concerning the handling, exchange and dissemination of proteomics data as no gold standard exists up to date to guide investigators and enhance comparability of proteomics and biomarker studies. For AD, at a 2003 consensus panel, researchers gave the following guidelines for biomarker research in this dementia: the ideal biomarker for AD should aim to detect a fundamental feature of neuropathology and should be validated in neuropathologically confirmed cases; the diagnostic sensitivity should be >80% for detecting AD and the specificity of >80% for distinguishing other dementias; it should be reliable, reproducible, non-invasive, simple to perform and inexpensive. For biomarker validation it is recommended to confirm the findings by at least two independent studies conducted by qualified investigators and the results published in peer-reviewed journals. Furthermore, it would be especially useful if the biomarker could also capture the beneficial effect of disease-modifying therapy (Ho et al., 2005).
Despite all of the above-mentioned difficulties and challenges, proteomic technologies provide rapid progress and immense benefits and already now create important biological data in neuropsychiatric research. Proteomic approaches allow a large-scale high-throughput qualitative and quantitative protein analysis complementing other traditional methods used in molecular genetics. Proteomics has the advantage of being relatively unbiased without a priori assumptions about differences between sample groups. It is a powerful research field that can reveal the function of so far uncharacterized proteins and generate new hypotheses to improve the understanding of the basic physiology of CNS under normal and disease conditions. In contrast to molecular genetic studies, proteomics has the great advantage of analysing processes at the protein level, thereby possibly being closer to the pathophysiological processes underlying the clinical phenomenology of specific psychiatric conditions. This is a crucial point as there exist more proteins than genes due to modifications of gene products by alternative splicing and post-translational modifications of the proteins expressed. Analyses on the protein level become even more important because of the already-mentioned reports that mRNA levels and protein concentrations correlate insufficiently and it seems not possible to predict protein expression levels from quantitative mRNA data (Gygi et al., 1999). For some genes in this study the protein levels varied by more than 20-fold with constant mRNA levels. As discussed in the introduction, on the one hand technical issues can contribute to an imperfect correlation, whilst on the other hand these differences can be attributed to the increasing complexity of gene products from the gene to the protein level by alternative RNA splicing and post-translational modifications. Current mRNA analysis methods are not set up to systematically capture different mRNA splice variants, whereas proteomic approaches are able to identify these variants (Hegde et al., 2003). As diverse isoforms of proteins can differ in their molecular characteristics and finally their function, it is crucial to distinguish between these variants by proteomic methods. Through amino acid sequencing, detection of post-translational modifications and protein–protein interactions of molecules involved in neuronal transmission and signal cascades, a better context-based functional understanding of cellular protein networks in neuropsychiatric disorders will emerge. The detailed knowledge, concerning molecular pathways in vivo together with proteomics data of drug effects in vitro, could facilitate the discovery of new potential drug targets and the design of more specific medications with fewer side effects.
As already mentioned for several psychiatric disorders, proteomic studies present the immense opportunity to identify surrogate biomarkers in easily accessible tissue for early disease detection, perhaps disease prevention, and the differentiation of stages and similar phenotypes of distinct nosological entities. Furthermore, biomarkers are suitable for a personal drug-monitoring scheme. The assessment of the patients’ individual protein profile before and in the course of medication might in the future allow a prediction of drug response and an adequate treatment modification. Ideally the marker recovery should be simple and non-invasive enough to perform in easily accessible tissue whose gene expression profile is similar to more inaccessible CNS tissues. In a recent study the gene expression patterns in blood and brain were compared (Sullivan et al., 2006). The authors found significant expression similarities in whole blood and multiple CNS tissues with a median correlation of about 0.5. About half of a set of schizophrenia candidate genes were found in both blood and prefrontal cortex. The authors concluded that surrogate marker search in blood is a useful tool when it has been determined that the relevant genes are expressed in both tissues.
After a successful identification of disease biomarkers proteomic technologies will hopefully overcome the obstacle of designing clinically useful and easily applicable laboratory or bedside tests.
Conclusion
Proteomic high-throughput technologies based on MS are increasingly used as a valuable tool in psychiatric research. This approach provides the opportunity to analyse and identify the complexity and dynamics of pathophysiological processes in neuropsychiatric disorders at the protein level. Understanding the molecular mechanisms of synaptic transmission, protein–protein interactions and signalling cascades will provide crucial insights into brain diseases, allowing the search for diagnostic and prognostic biomarkers as well as new therapeutic targets.
Footnotes
Acknowledgment
This work was supported by the German Research Council (HU 1536/1-1, HU 1536/1-2).