INTRODUCTION

Currently used classification systems in psychiatry are not able to accurately display the inter-individual differences in etiology and outcome of major depressive episodes. Phenotype heterogeneity contributes to the observed differences in response and remission rates to antidepressant drug treatment between patients. Despite the availability of numerous drugs, only half of the patients afflicted with major depression respond to a first antidepressant medication, and about 30% do not reach remission even after several treatment trials (Rush et al, 2006; Hennings et al, 2009). In recent years, studies on biological differences in depressed patients have primarily focused on genetic analyses and certain brain circuits (eg, monoamine systems). In some cases, these studies revealed associations between genes or proteins and the occurrence of a major depression (Carpenter et al, 2004; Geracioti et al, 1997; Jokinen et al, 2008; Sullivan et al, 2006). Due to the complexity of the disease and the involvement of a multitude of brain systems, single markers are unlikely to have the ability to improve current diagnostic and therapeutic strategies. Although genome-wide association studies have shown that we can categorize patients as high or low number of response allele carriers to predict antidepressant treatment outcome (Ising et al, 2009), we still do not know which antidepressant drug a patient will benefit from. At the same time, an overall sub-classification based on individual biological features of depressed patients is still missing. To examine the complexity of pathological brain mechanisms accounting for a depressive episode, the most relevant molecular compartment to analyze is the proteome. The proteome is the entire collection of proteins encoded by the genome of an organism, including protein isoforms and posttranslational modifications (Taurines et al, 2011). Due to its close proximity, cerebrospinal fluid (CSF) reflects the metabolic status and biochemical alterations of the brain. Hence, for the in vivo investigation of a brain disease, CSF represents the most relevant specimen (Zougman et al, 2008). The analysis of the CSF proteome can reveal protein differences between diseased and healthy individuals, allowing the discovery of metabolic pathways linked to the pathogenesis of psychiatric disorders (Yuan et al, 2002). In the present report, we are employing two complementary methods for the interrogation of the CSF proteome. Two-dimensional polyacrylamide gel electrophoresis (2D-PAGE) is able to analyze intact proteins, and is a straightforward method for the detection of posttranslational modifications. Compared with 2D-PAGE, mass spectroscopy-based peptide analysis methods have greater sensitivity and reproducibilty with less material consumption (Lottspeich et al, 2010). This is an important aspect in the case of clinical specimens that are often of limited availability. However, the mass spectroscopy-based peptide analysis approach requires digestion of the proteins before differential analysis, resulting in an increased sample complexity. The combined application of the two complementary proteomic methods provides a comprehensive approach for biomarker discovery. In an earlier study, we have detected a 33-kDa protein with an isoelectric point of 5.2, whose expression level in CSF differed significantly between depressed patients with and without suicide attempt (Brunner et al, 2005). Due to a lack of material, we were unable to identify the protein. In the present paper, we report on a comprehensive hypothesis-free proteomic analysis of CSF from depressed patients. We have successfully identified several differentially expressed proteins that are discussed with regard to their potential role in the pathogenesis of unipolar depression.

MATERIALS AND METHODS

CSF Samples

The study included CSF from 12 individuals fulfilling ICD-10 criteria for a depressive episode, either as a single episode or within recurrent episodes (mean age 54±16.01 years; 10 females, 2 males). A total of 12 CSF control samples were matched according to sex and age (mean age 53±18.08 years; 8 females, 4 males). Participants had provided written informed consent to take part in the study (German Federal Research Ministry, Competence Nets in Medicine, subproject 1.5). Samples from individuals with additional severe systemic or central nervous system disease were excluded from the study. CSF samples with abnormal levels of glucose, lactate, total protein, cells, positive oligoclonal bands, or disturbance of the blood–brain barrier, as well as blood-contaminated samples were not considered.

Sample Preparation for 2D-PAGE

A total of 24 CSF samples were prepared for 2D-PAGE. Each sample (5 ml) was desalted and concentrated to 80 μl on a Vivaspin spin column with a molecular weight cutoff of 5 kDa (Vivascience AG, Hannover, Germany). Afterwards, immunodepletion of albumin, transferrin, IgG, IgA, antitrypsin, and haptoglobin in CSF was performed using a Multiple Affinity Removal System column (Agilent Technologies, Böblingen, Germany) as described previously (Maccarrone et al, 2004). To apply the same amount of protein on each gel, protein concentration of HPLC flowthrough fractions was determined using the Bradford method. Equal amounts of protein (136 μg) were then diluted in isoelectric focusing (IEF) buffer (8M urea, 2% CHAPS, 0.2% Biolyte, 100 mM DTT, 0.001% bromophenol blue) to a final volume of 300 μl.

Two-dimensional Polyacrylamide Gel Electrophoresis

Depleted CSF samples in IEF sample buffer were applied to a 17 cm, pH 3–10 Immobiline strip (BioRad, Hercules, CA). After active rehydration at 50 V for 12 h IEF, SDS polyacrylamide gel electrophoresis and gel staining were carried out as described previously (Jacob and Turck, 2008). A total of 24 gels were run, 12 each from depressed individuals and controls.

Gels were first stained with ProQ Diamond (Molecular Probes, Eugene, OR) to visualize phosphorylated proteins with a Molecular Imager FX Pro fluorescence scanner (BioRad). Subsequent total protein staining was performed with Coomassie Brilliant Blue G (Sigma Aldrich, St Louis, MO) in colloidal solution (17% (w/v) ammonium sulfate, 2% (v/v) phosphoric acid, and 34% (v/v) methanol). Coomassie-stained gels were scanned with a GS-800 densitometer (BioRad).

Data Analysis

Scanned images were analyzed with the help of PDQUEST software (BioRad). After normalization according to total density, spots of relevant protein intensity were detected using the automated spot detection and matching function. The automated analysis was verified and completed by manual matching. In order to find the most interesting spots, ie., those with relevant protein intensity differences between the two groups, an explorative comparison analysis based on t-statistics calculated by PDQUEST was performed. Differences in the most interesting protein spots were finally statistically evaluated using multivariate and univariate tests of significance (F-tests for simple effects in MANOVA). To avoid singularities, the all protein spots were first partitioned in subsets of three to five items, and then subjected to analyses of variance. A nominal level of significance of α=0.05 was accepted and corrected according to the Bonferroni procedure for all à-posteriori tests.

In-gel Protein Digest and Mass Spectrometry Analysis

Protein spots of interest were excised from the gel using the Proteineer Sp automated spot cutter (Bruker Daltonics GmbH, Bremen, Germany). Destaining, drying, and tryptic digestion of gel spots was carried out as previously described (Jastorff et al, 2009).

For matrix-assisted laser desorption ionization TOF mass spectrometry (MALDI-TOF-MS) analysis, samples were prepared by a dried droplet method on a 600-μm AnchorChip MALDI Target (Bruker Daltonics). α-cyano-4-hydroxycinnamic was dissolved in 30% (v/v) acetonitrile and 70% (v/v) 0.1% trifluoroacetic acid to saturation. The matrix preparation was diluted 10-fold in a 2 : 1 ethanol : acetone solution. Each sample (1 μl) was spotted onto the AnchorChip sample target, and 1.5 μl of the prepared matrix solution was added to the spot. For instrument calibration, the peptide calibration standard I (Bruker Daltonics) was used. Peptide mass fingerprints were acquired using the Ultraflex I TOF/TOF mass spectrometer (Bruker Daltonics).

Protein Identification

MALDI-TOF-MS data analysis was carried out with Biotools software (Bruker Daltonics). For protein identification, the database search program Mascot (Matrix Science, London, UK) was applied to identify proteins according to their tryptic mass fingerprints. The search was carried out against NCBI non-redundant and human protein database. Mascot search was performed with a 25-ppm parent mass tolerance, and one possible tryptic miscleavage. Chemical modifications were fixed carbamidomethylation modification for cysteine and oxidation for methionine residues. Protein hits that were significant according to the Mascot score (P<0.05) were accepted. All protein hits were further verified by reviewing the position of the spot with regard to molecular weight and isoelectric point (pI) on the 2D-gel image.

Western Blot

Protein concentration of CSF samples was determined with the Bradford method (BioRad) and bovine serum albumin as protein standard. Proteins (2 μg) were separated on 12% SDS gels, and electrotransferred to a PVDF membrane. For IEF-immunoassays, proteins were applied to pH 3–10 Immobiline strips (BioRad) and separated by IEF as described above. Membranes were incubated with primary antibody (anti-cystatin C, 1 : 7500 (Abcam, Cambridge, UK), anti-pigment epithelium derived factor (PEDF), 1 : 5000 (Chemicon International, Hofheim, Germany), anti-prostaglandin D synthase (lipocalin-type), 1 : 5000 (Cayman Chemical, Ann Arbor, MI), anti-apolipoprotein E, 1 : 5000 (Abcam)) in 5% milk-PBS for 90 min at RT, followed by 60 min incubation at RT with a secondary antibody (IgG, 1 : 25.000 (Sigma Aldrich)). Immunocomplexes were visualized by chemoluminescense with the help of ECL Plus reagent (GE Healthcare, Piscataway, NJ) and exposed to autoradiography films. Films were scanned and processed with QuantityOne (BioRad) to quantify band and isoform signal intensities. The statistical analysis was carried out with the t-test. The difference was considered to be significant when P<0.05.

Liquid Chromatography Mass Spectrometry (LCMS) Profiling

Instrumentation

Agilent HPLC-Chip/MS system interfaced to an Agilent LC/MSD TOF mass spectrometer. The HPLC-Chip/MS system is a microfluidic chip-based device that integrates sample preparation (enrichment column), analytical separation (nanocolumn), and nanoelectrospray formation (emitter tip).

Sample Preparation

Human CSF fractions were immunodepleted using an Agilent Multiple Affinity Removal system. A 20 μg portion of each sample was then desalted using spin tubes that incorporate 5-kDa molecular weight cutoff filters (Vivaspin). The desalted samples were reduced and denatured using 50% TFE in a 50 mM ammonium bicarbonate buffer with 10 mM DTT at 95 °C for 20 min. The reduced samples were subsequently alkylated with 30 mM iodoacetamide at room temperature for 1 h, and the unreacted iodoacetamide was quenched by the addition of DTT. The samples were diluted 1 : 10 with 25 mM ammonium bicarbonate buffer, then treated with trypsin at a 1 : 25 enzyme : substrate ratio overnight at 37 °C.

Chromatographic conditions

Chromatographic separation of the digested samples was accomplished using a protein ID chip, which consists of a 40 nl enrichment column and 75 μm × 43 mm analytical column packed with ZORBAX 300SB-C18, 5-μm material. The mobile phase A was 0.1% formic acid in water and B was 90% acetonitrile +0.1% formic acid in water. Sample was loaded on the enrichment column at 4 μl/min by the loading pump. After loading, the microvalve was switched to analysis mode, to forward-flush the sample onto the analysis column. The analytical flow rate was 300 nl/min, and the solvent gradient went from 3 to 45%B in 75 min, then to 80% B at 80 min for 2 min. Each sample was run in triplicate.

Mass spectrometer conditions

Electrospray data for the CSF samples was obtained using orthogonal nanospray in positive mode. MS conditions were as follows: drying gas temperature, 300 oC; drying gas flow, 4 l/min; capillary voltage, 1900 V, and fragmenter voltage, 150 V. The mass spectrometer was operated with a mass range of m/z 300 to 2400 U and two cycles/second. To ensure low ppm mass accuracy, the internal reference mass correction was utilized to correct for scan-to-scan variations.

Ms data analysis

Data analysis was done using Agilent MassHunter Qualitative Analysis software and Mass Profiler Professional software, designed to facilitate rapid differential analysis of samples for profiling applications. The MassHunter algorithm extracts chemically qualified molecular features from complex LC/MS TOF data sets by first finding the mass peaks in all mass spectra, and then removing the non-chromatographic chemical background. Next, peaks are clustered in RT (in seconds) and m/z to form 3-D peaks. The 3-D peaks are centroided and a peak volume determined for each peak. Related 3-D peaks (isotopes, adducts, dimers, trimers, multiple charge states) are combined and assigned a neutral mass and total volume. Mass Profiler Professional software combines capabilities for cross-sample alignment of molecular features in both the retention time and mass dimensions, including several normalization options, with statistical methods and visualization tools to aid in identification of differentially expressed features. The Profiler software compensates for scan-to-scan shifts in retention time (seconds) and measured masses (ppm). Common features are identified and cross-sample response RSD values are calculated. Results filters can be used to reduce the number of differentially expressed biomarkers to be investigated.

RESULTS

We analyzed CSF proteomes of depressed patients and compared them to controls. For this purpose, we used a proteomic workflow that incorporates two complementary methodologies, 2D-PAGE and LCMS-based differential expression analysis.

Image analysis of colloidal Coomassie-stained 2D-PAGE gels resulted in 213 spots that revealed quantitative differences between the two groups, with 11 being significantly different (univariate F-tests in MANOVA, P<0.05). Mass spectrometry analysis of the excised gel spots identified PEDF (two isoforms), apolipoprotein E precursor (ApoE), prostaglandin D2 synthase (PGDS; 21 kDa), transthyretin precursor, α-1B-glycoprotein, vitamin D-binding protein (DBP, two isoforms), cystatin C, β-2-glycoprotein, and hemopexin. In case of PEDF, ApoE, PGDS, cystatin C, and DBP, the excised spot represented a single isoform of the protein. Figure 1 shows a representative Coomassie-stained gel of CSF from a depressed patient with differentially expressed proteins indicated. Table 1 gives an overview of the identified differentially regulated proteins; Table 2 shows their molecular function and role in biological processes. Comparative analysis of the phosphoprotein-specific ProQ Diamond-stained gels indicated a difference in phosphorylation pattern between the CSF proteome of depressed and control individuals. A total of 16 differences of phosphorylated proteins were detected between the replicate groups. As ProQ Diamond is a fluorescence stain not detectable with visible light, we had difficulties cutting out the relevant gel pieces. We were only able to identify five of the differentially phosphorylated proteins by mass spectrometry. Different phosphorylation patterns were detected for heat-shock cognate 71-kDa protein, heterogeneous nuclear ribonucleoprotein H, ApoE precursor, fructose-bisphosphate aldolase C, and cAMP-dependent protein kinase catalytic subunit alpha. Figure 2 shows a representative ProQ Diamond-stained gel of CSF from a control person with differentially phosphorylated proteins indicated (see also Tables 1 and 2). To further verify our 2D-PAGE results, we additionally analyzed the depleted CSF samples by LCMS-based profiling. Successful LCMS-based differential expression analysis requires good reproducibility for both chromatography (retention time) and mass spectrometry (accurate mass), so that feature lists can be compared between different samples. The reproducibility of the replicate runs for this study was monitored by clustering the data in GeneSpring (Figure 3). Statistical analysis of the data revealed 144 target features that were significantly different between the control and depressed samples. A log2 plot of the features (Figure 4) shows the distribution of those features. Targeted MS/MS results identified all nine proteins found to be differentially expressed by 2D-PAGE. With the exception of α-1B-glycoprotein, cystatin C, and PEDF, the same direction of protein regulation was found by LCMS-based profiling and 2D-PAGE analysis. The discrepancy in regulation found for the above-named proteins is probably due to the quantitation of peptides derived from different protein isoforms during LCMS profiling and 2D-PAGE. The list of 32 corresponding proteins derived from significant different mass features found by mass spectrometry is shown in Table 1 (indicated as mass spectrometry profiling results). A summary of identified mass features is presented in Supplementary Table 1; peptide sequences and the number of peptides used for protein identification are shown in Supplementary Table 2.

Figure 1
figure 1

(a) Representative Coomassie-stained two-dimensional polyacrylamide gel electrophoresis (2D-PAGE) gel image of cerebrospinal fluid (CSF) from a depressed patient. Differentially regulated proteins and isoforms between disease and control groups are indicated, and were identified as 1. transthyretin precursor; 2. cystatin C; 3. prostaglandin D2 synthase; 4. apolipoprotein E; 5 and 6. pigment epithelium derived factor; 7. vitamin D-binding protein; 8. β-2-glycoprotein; 9. hemopexin; and 10. alpha-1B-glycoprotein. (b) 2D-PAGE comparison of proteins that were further validated by immunoblots.

PowerPoint slide

Table 1 Regulated Proteins in Cerebrospinal Fluid of Depressed Patients
Table 2 Molecular Function of Regulated Proteins in Depression
Figure 2
figure 2

ProQ Diamond-stained two-dimensional polyacrylamide gel electrophoresis (2D-PAGE) gel image of cerebrospinal fluid (CSF) from a control person. Differentially phosphorylated proteins and isoforms between disease and control groups are circled and identified proteins are numbered as 1. apolipoprotein E; 2. heterogeneous nuclear ribonucleoprotein H; 3. heat-shock cognate 71 kDa protein; 4. cAMP-dependent protein kinase catalytic subunit alpha; 5. fructose-bisphosphate aldolase C.

PowerPoint slide

Figure 3
figure 3

Hierarchical clustering tree for molecular features extracted from the liquid chromatography mass spectrometry (LCMS) analysis of depressed and control cerebrospinal fluid (CSF) samples. The clustering analysis can be used to monitor the reproducibility of the analysis (technical replicates), as well as the similarity of the biological replicates. The clustering analysis was performed for the feature list from t-test of features found in 100% of at least one group.

PowerPoint slide

Figure 4
figure 4

Log 2 plot of molecular feature abundance from cerebrospinal fluid (CSF) samples of depressed and control individuals. Features that are significantly upregulated or downregulated are shown.

PowerPoint slide

To validate the quantitative differences detected by 2D-PAGE and LCMS-based differential expression analysis, we performed immunoblotting for four representative proteins, using CSF samples from the same and independent individuals. PGDS, ApoE, PEDF, and cystatin C were analyzed. SDS gel electrophoresis does not allow to distinguish protein isoforms of similar molecular weights since they tend to run as one band. Except for cystatin C and PEDF, no differences in expression were found in the SDS gel-based western blot between depressed and control CSF samples. We therefore next performed IEF-based immunoblots that allow the separation of proteins with different isoelectric points. Using this method, we were able to quantitate individual protein isoforms and confirm the 2D-PAGE results. Figures 5, 6 and 7 show the results of the SDS- and IEF-immunoblots for cystatin C, PEDF, and PGDS, respectively.

Figure 5
figure 5

Cystatin C SDS gel western blot (a) and isoelectric focusing (IEF) western blot (b). Total cystatin C protein, as well as the most acidic isoform of the protein show higher expression levels in cerebrospinal fluid (CSF) from control persons ((a)+(b): *p<0.01).

PowerPoint slide

Figure 6
figure 6

(a) Pigment epithelium-derived factor western blot. Two bands are detectable due to different molecular weights of the isoforms. The upper band isoform showed a significant increase in cerebrospinal fluid (CSF) of depressed individuals compared with controls (b), *p<0.05.

PowerPoint slide

Figure 7
figure 7

Prostaglandin D2 synthase SDS gel western blot (a) and isoelectric focusing (IEF) western blot. (b) Statistical analysis of western blot band intensities did not reveal a significant difference between depressed patients and controls ((a), p=0.134), whereas analysis of individual isoforms showed a significant decrease of isoform 2 in depressed patients ((b), *p<0.05).

PowerPoint slide

DISCUSSION

The precise pathophysiological mechanisms leading to the clinical phenotype of a major depression still remain unclear (Krishnan and Nestler, 2010; aan het Rot et al, 2009; Lee et al, 2010). The availability of biological markers would ultimately allow to account for inter-individual variance in etiology and phenotype of depressive episodes in different individuals, and thereby improve personalized therapeutic strategies and outcome (Figure 8; Luo et al, 2010; Turck et al, 2005). The comparison of the CSF proteome of depressed patients and controls with the help of 2D-PAGE and LCMS-based differential expression analyses revealed a number of proteins that showed altered expression levels between the two groups. Interestingly, two of the identified proteins, ApoE and cystatin C, have already been implicated in other CNS diseases (Levy et al, 2006). Proteins that were found to be differentially regulated in depression also included PEDF, PGDS, transthyretin precursor, alpha-1B-glycoprotein, DBP, β-2-glycoprotein, and hemopexin. Among the protein markers that were only identified by LCMS-based differential expression, the neurofilament heavy polypeptide and neuronal cell adhesion molecule (NCAM) are of particular interest. Neurofilament heavy polypeptide is a protein involved in nervous system development and cell death. Neurofilaments have been discussed as biomarker candidates for neurodegenerative diseases (Petzold et al, 2009; Pijnenburg et al, 2007). NCAM is a protein involved in neuronal growth and repair (Gnanapavan et al, 2010). The protein's expression was found to be altered in schizophrenic patients, where it was detected as its variable alternative-spliced exon isoform in CSF (Vawter, 2000; Vawter et al, 2000). A selective increased expression of primarily the 120 kDa NCAM isoform has been reported in CSF from bipolar mood disorder type I and recurrent unipolar major depression patients (Poltorak et al, 1996). In our study, peptides of the 140-kDa NCAM isoform were found to be elevated in CSF of depressed patients. These are probably derived from the variable alternative-spliced exon isoform, which is in good agreement with previous observations.

Figure 8
figure 8

3D principal component analysis (PCA) for feature list from t-test of features found in 100% of at least one group allows patient categorization.

PowerPoint slide

A set of four proteins (PEDF, ApoE, PGDS, cystatin C) known to have important functions in brain metabolism were chosen for the validation phase of our biomarker study.

PEDF is a neurotrophic glycoprotein, which is a member of the serine protease inhibitor gene family. In neuronal cells, PEDF exhibits neuroprotective effects that seem to involve the activation of cyclic AMP-responsive element binding protein (CREB) and nuclear factor-κB (Tombran-Tink and Barnstable, 2003; Yabe et al, 2005). CREB has been implicated as a molecular state marker for depression and for the response to antidepressant treatment (Yamada et al, 2003). In this study, an increase of PEDF expression was detectable in 2D-PAGE of CSF from depressed patients. PEDF is known to activate CREB in primary neuronal cultures (Yabe et al, 2005). In depression, this could represent a compensation mechanism to counteract endogenous pro-apoptotic stress.

Cystatin C is a known cysteine proteinase inhibitor, which is produced by most cells including neurons (Levy et al, 2006). In AD, cystatin C co-localizes with amyloid-β deposits and inhibits its fibril formation. It was shown that overexpression of cystatin C in brains of amyloid precursor protein transgenic mice reduces cerebral amyloid-β-deposition (Kaeser et al, 2007). The CST3 Thr25 allele of CST3, which encodes cystatin C, leads to reduced cystatin C levels and promotes susceptibility to AD. In our study, we found a decrease of cystatin C, especially the pI 7.8 isoform, in 2D-PAGE of CSF from depressed patients. As decreased cystatin C levels in CSF could reflect a higher susceptibility to a neurodegenerative disease like AD, reduced levels in depressed patients could possibly indicate the presence of cognitive symptoms in depression. However, we do not have any information about amyloid plaque load in the brains of our study patients to allow correlation with cystatin C levels. PGDS, a glycoprotein with a molecular weight of approximately 26 kDa, synthesizes prostaglandin D2 (PGD2), which is one of the most potent endogenous sleep-inducing substances (Yamashima et al, 1997). Beside sleep induction, PGD2 is involved in sedation, nociception, and release of hormones (Eguchi et al, 1999; Sri Kantha et al, 1994; Urade and Hayaishi, 1999). Sleep disturbances belong to the key symptoms of major depression. Especially, slow-wave sleep seems to be reduced in depression. PGD2 infusion into the subarachnoidal space in rat brains preferentially induced slow-wave sleep (Matsumura et al, 1994). We detected a decrease of PGDS isoforms in CSF of depressed patients, which could contribute to bemoaned sleep disturbances.

Apo E has an important role in lipoprotein metabolism. A homozygote constellation of the E2 allele is linked to the hyperlipoproteinemia type III, a disease that leads to an increased risk for cardiovascular diseases like stroke or heart attack (Brummer et al, 1998). We detected an ApoE isoform (pI6.3) that is more abundant in CSF of controls compared with depressed individuals, and probably represents an E2 or E3 isoform. The observed expression pattern could be due to interindividual differences of isoform expression according to different individual allele constellations. On the other hand, major depression has been repeatedly linked to the presence of a metabolic syndrome in afflicted persons (Jakovljevic et al, 2007). Patients with psychiatric disorders are known to have an increased risk of developing the metabolic syndrome, and patients with a metabolic syndrome are at a higher risk of developing a depressive episode (Capuron et al, 2008; Richter et al, 2010; Vogelzangs et al, 2007). The underlying mechanism linking both syndromes are still not identified, but differences in Apo E isoform expression could be one connecting link.

Next to CSF protein level differences, we also investigated CSF protein phosphorylation by using a specific gel stain. In the central nervous system, protein phosphorylation and second messenger mechanisms are linked to neurotransmission (Popoli et al, 2000), and altered CSF protein phosphorylation patterns have been discussed as a marker for neurodegenerative diseases (Henneman et al, 2009; Shaw et al, 2009). With the help of 2D-PAGE, we found several CSF protein phosphorylation differences between depressed patients and controls, of which five were identified (Figure 2). If and how these differences contribute to the phenotype of a major depression needs to be analyzed in future studies.

In conclusion, we detected several protein expression and phosphorylation differences in the CSF proteome of depressed patients compared with controls. With the help of mass spectrometry, we were able to identify the majority of them. A subset of the identified proteins has important roles in brain metabolism, neuronal protection, and differentiation or neurotransmission. Others, eg, PGDS, could be related to certain clinical aspects of major depression such as sleep disturbances.

The herein reported results support the notion that markers can be generated to identify subgroups of depressed patients who could share a causal mechanism. Such biomarkers are needed to match the right patient with the right antidepressant drug. Proteomics provides a promising tool for the detection of disease-specific markers for depression, and thereby has the ability to account for inter-individual molecular differences between patients and an adjusted and improved pharmacological intervention.