The neurobiological basis of suicidal behaviour is multifactorial and complex. Several lines of evidence indicate that environmental factors as well as multiple genes and interactions of both are implicated in its aetiology. The aim of this study was to establish the transcriptomic expression profile of post-mortem brain tissue of suicide victims in order to identify new candidate genes and biological patterns for suicidal behaviour. Post-mortem orbitofrontal cortex tissue was derived from 11 suicide victims and 10 non-psychiatric controls carefully selected from a brain bank of over 150 brains, and the expression of more than 23000 messenger RNAs was assessed in this case-control study. Validation experiments were carried out using quantitative RT–PCR as an independent method. A classification of the differentially expressed genes according to their biological function and statistical analyses of the data were performed in order to identify biological pathways that are over-represented in the suicide group. In total, 124 transcripts demonstrated significant changes (fold changes ⩾1.3, p value ⩽0.01), with 59 showing under-, and 65 over-expression in the suicide group. The results could be validated for nine particularly interesting transcripts (CDCA7L, CDH12, EFEMP1, MLC1, PCDHB5, PTPRR, S100A13, SCN2B, and ZFP36). The pathway analysis showed that the Gene Ontology categories ‘central nervous system development’, ‘homophilic cell adhesion’, ‘regulation of cell proliferation’ and ‘transmission of nerve impulse’ were significantly enriched. The differentially expressed genes and significant biological processes might be involved in the pathophysiology of suicide and warrant further attention.
The predisposition to suicidal behaviour is influenced by both genetic and environmental factors and interactions between these two (Balazic and Marusic, 2005). It is probable that many genetic variants with small effects contribute to the aetiology of suicide, but most of the research efforts to date focused on the serotonergic system and serotonergic genes (Rujescu et al., 2007). Recently, additional genes like neutrophins (Dwivedi et al., 2005; Kunugi et al., 2004) or SSAT (Sequeira et al., 2006) have been implicated. However, the number of susceptibility genes, the attributable risk conferred by each, and the degree of interaction between them remain unknown. Thus, finding genes that are implicated in suicide has a high priority, given that the biological mechanisms as well as the pathobiology are far from being understood. Microarray technology provides the possibility to simultaneously compare the levels of thousands of gene transcripts in one experiment and thus to screen for new candidate genes and mechanisms without an a-priori hypothesis (Konradi, 2005), and have the potential to reveal interesting gene expression alterations in psychiatric disorders. This was well demonstrated by several studies investigating gene expression profiles in schizophrenic subjects that yielded clearly intersecting results (Mirnics et al., 2006). This paper describes the expression analysis of ∼23000 transcripts in the orbitofrontal cortex of 11 suicide victims and 10 matched controls carefully selected from a brain bank of over 150 brains. The orbitofrontal cortex is uniquely positioned to use associative information to project into the future, and to use the value of perceived or expected outcomes to guide decisions (Schoenbaum et al., 2006). Impaired decision making is a neuropsychological intermediate phenotype for suicidal behaviour (Jollant et al., 2005). We performed a differential expression analysis and classified the identified genes according to their biological function in order to reveal candidate mechanisms for suicidal behaviour.
Autopsy brain samples
The human brain tissue was obtained from the Institute of Forensic Medicine of the Johann Wolfgang Goethe University in Frankfurt am Main, Germany. The research was approved by the University's Internal Review Board. The basic characteristics of the subjects are shown in Table 1. Regarding psychiatric diagnoses seven out of the 11 suicide victims had depression, three adjustment disorder, and one schizophrenia. Two suicide victims additionally had alcohol abuse. Furthermore, the toxicological analyses of the brains showed that five suicide victims had received antidepressants (doxepine, amitriptyline and clomipramine), two bromazepam, three analgetics (tramadol) and two sleep medication (7-aminoflunitrazepam, doxylamin). Information on demographic data, agonal factors and psychopathology of the decedents were derived from medical records, coroner's investigations, the medical examiner's results and interviews of next-of-kin. Post-mortem tissue was derived from the orbitofrontal cortex [Brodmann Area (BA) 11] of 11 suicide victims and 10 controls. All brains that showed signs of infarcts, subarachnoidal haemorrhages or tumours were excluded prior to this study. Eight specimens of the suicide group as well as of the control group were obtained from the left hemisphere, respectively. Three specimens of the suicide group and one of the control group originate form the right hemisphere, and no information was available for one control subject. However, the two groups showed no significant differences regarding the side of the orbitofrontal cortex. Tissue pH is a robust indicator for RNA quality (Kingsbury et al., 1995), and therefore the two groups were matched very closely by this factor. The pH value was measured by means of a pH meter after homogenizing ∼500 mg of brain tissue in 2 ml distilled water. Furthermore, both groups were also matched according to age, post-mortem interval (PMI) and gender. The mean pH of the controls and the suicide victims (6.71±0.20 vs. 6.74±0.10, p=0.663; mean value±s.d.) as well as age (yr) (64.1±11.4 vs. 55.4±14.3, p=0.140), PMI (hours) (69.5±18.9 vs. 59.7±26.1, p=0.340) or male:female ratio (3:7 vs. 3:8) showed no significant differences between the two groups.
PMI, Post-mortem interval; n.a., not available.
Li and colleagues conducted microarray experiments with post-mortem brain tissue of 40 subjects and showed that the terminal medical conditions and the duration of the agonal state had the strongest influence on the RNA quality of the samples (Li et al., 2004). Subjects with longer agonal states generally tended to have a lower brain tissue pH, more degraded RNA and also showed systematic expression changes of genes coding for transcription factors, and proteins involved in energy metabolism and stress response. Another study on microarray experiments yielded similar results, showing that prolonged agonal states associated with coma and hypoxia affect RNA integrity and gene expression profiles more than age, gender and post-mortem factors (Tomita et al., 2004). Therefore, tissue pH is the most important covariate, followed by age, PMI and gender. The extraction of intact and biologically active mRNA after more than 4 d was reported by several groups (Bahn et al., 2001; Johnston et al., 1997; Schramm et al., 1999; Yasojima et al., 2001). Therefore, we also included subjects with longer PMI and focused on appropriate pH values and stringent RNA quality control, which was carried out with an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA, USA). Almost all subjects in our sample experienced brief deaths due to violent suicide methods or cardiac events with short agonal states.
RNA extraction and preparation
RNA was isolated from ∼75 mg of frozen brain tissue using the RNeasy Lipid Tissue Mini Kit (Qiagen, Hilden, Germany). The tissue samples were homogenized in 1 ml Qiazol Reagent using an Ultraturrax at 4°C for 20 s. The subsequent steps were carried out according to the manufacturer's protocol with an additional DNAse digestion in order to remove potential genomic DNA. The quality of the isolated RNA was first assessed by agarose gel electophoresis, and subsequently analysed with the Agilent 2100 Bioanalyzer (Agilent Technologies). All RNA samples showed clearly defined 28S and 18S ribosomal bands, indicating good RNA quality as proposed for microarray and validation experiments (Mimmack et al., 2004). The mean ribosomal RNA ratio (28S/18S), measured with the Agilent 2100 Bioanalyzer, was 1.57. The isolated RNA was stored at −80°C until further processing and utilized for both the microarray and real-time quantitative RT–PCR experiments.
Microarray experiments were performed using Illumina Sentrix HumanRef-8 Expression BeadChips (Illumina, San Diego, CA, USA) which contain eight single arrays per chip. An experimental comparison of this platform with the Affymetrix HG-U133 plus 2.0 array yielded comparable data, especially for genes predicted to be differentially expressed (Barnes et al., 2005). Each Illumina bead-type carries several thousand 50-mer transcript-specific probe sequence oligonucleotides, and each bead-type is represented on average by 30 copies in each array. This internal technical replication of the Illumina HumanRef-8 Expression BeadChip provides the statistical accuracy for multiple measurements and thus increases precision and reliability (Kuhn et al., 2004). The array set contains 24000 different bead types representing ∼23000 NCBI Reference Sequence gene transcripts, as some transcripts are measured in duplicate by two different probes. Each of the 21 RNA samples was hybridized to a single array, and including five technical replicates, a total of 26 experiments using four BeadChips were performed. One hybridization replicate was hybridized on two different BeadChips in order to evaluate the influence of different slides and the hybridization step itself on the results. Concerning the labelled replicates, four RNA samples were split in two aliquots in each case, labelled in different reactions and hybridized to separate arrays. Target cDNA for the microarrays was prepared following the manufacturer's protocol. Briefly, 100 ng of total RNA of each brain sample was reverse transcribed into cDNA and subsequently in-vitro transcribed and biotinylated. The Ambion Illumina® RNA amplification kit (Ambion, Austin, TX, USA) is based on the classic Eberwine T7-RNA amplification protocol (Van Gelder et al., 1990). After hybridization and stringent washing, the bound cRNA was stained using Streptavidin-Cy3. The gene signals on the chips were scanned using a submicron-resolution Bead Array Reader and the Sentrix Scan Software (Illumina, San Diego, CA, USA).
Microarray data analysis
Analysis was carried out with Illumina AnEx Software version 184.108.40.206. The arrays were normalized with the AnEx Average Normalization Algorithm, so that the means of all gene expression signals for the different arrays become equal. This method has been chosen as the HumanRef-8 BeadChips compare large numbers of genes, and the expression levels can be assumed to be roughly similarly distributed. Prior to this normalization, the software automatically applied unspecific background subtraction, which especially improves the assessment of expression changes for genes with dim signals. The arrays contain several negative control beads with random sequence probes which provide a measurement of non-specific hybridization, non-specific dye signal, and scanner background. After normalization, each bead-type signal was calculated by averaging corresponding bead signals after removing the outliers using median absolute deviation. The BeadChip technology provides sufficient power to detect fold changes of 1.3 with 95% confidence (Dickinson and Craumer, 2003), and therefore all genes showing a differential expression of ⩾1.3 were considered in the subsequent analyses. The detection p value for each probe was computed from the background model, indicating the chance that the target signal can be distinguished from the negative controls. Only genes that could be detected with p⩽0.01 in both groups were included in further analyses. Finally, the differential expression between the control group and the suicide group of each gene was calculated using the unpaired t test.
One microgram of total RNA was used for cDNA synthesis with random hexamer primers and Superscript II reverse transcriptase (Invitrogen, Karlsruhe, Germany) according to the manufacturer's recommendations. The oligonucleotide primers were designed using the web-based primer design program Primer3 from the Whitehead Institute for Biomedical Research (http://frodo.wi.mit.edu/cgi-bin/primer3/primer3_www.cgi; Rozen and Skaletsky, 2000). The resulting amplicons ideally overlapped with the oligonucleotide probe sequences of the Illumina microarrays, or were located as near as possible. All primer pairs were further optimized to ensure the specific amplification of the PCR product under investigation and the absence of any byproducts (Mimmack et al., 2004). Amplicons were typically 50–150 bp in length, and we favoured primers spanning exon–exon junctions in order to avoid amplification of genomic DNA. The primer sequences used are listed in Table 2.
Quantitative RT–PCR using the SYBR-Green dye (ABgene, Hamburg, Germany) was performed in a Rotor-Gene 2000 Cycler (Corbett Research, Sydney, Australia) under the following typical conditions: 95°C for 15 min, 40 cycles at 95°C for 15 s and 60°C for 60 s. The annealing temperature was adjusted in some cases from 58°C up to 63°C, depending on the primer pair. Amplification of the designated product was confirmed by subsequent melting curve analysis (60–99°C).
Each sample was determined in duplicate, whereas the standard curve was generated in triplicates from 1 ng up to 100 ng cDNA. The experiments were analysed using Rotor-Gene 4.6 software (Corbett Research). A normalization factor was calculated, derived from the geometric means of four internal control genes. First, the expression stability of six frequently used internal control genes was determined by geNorm (available at http://medgen.ugent.be/∼jvdesomp/genorm/; Vandesompele et al., 2002). This robust strategy identified the most stable internal control genes (ACTB, GAPDH, POLR2A and UBC) and provided accurate normalization factors.
More than 15680 probes delivered clearly detectable signals with p⩽0.01. The Illumina RS-8 BeadChip features sufficient precision to detect 1.3-fold differences with 95% confidence within the dynamic range (Kuhn et al., 2004). Therefore, we included all genes with fold changes ⩾1.3 and detection p values ⩽0.01 for further analyses. A total of 1585 transcripts fulfilled these criteria and were included in the subsequent analyses. An initial unpaired t test was performed in order to identify differentially expressed genes between the two groups. A total of 124 transcripts demonstrated significant changes (p⩽0.01), with 59 showing under-expression, and 65 over-expression in the suicide group. The results of the initial unpaired t test, depending on different p values and fold-change combinations, are listed in Table 3.
We conducted a second analysis using the web-based T-Rex tool of the Gene Expression Profile Analysis Suite GEPAS (http://gepas.bioinfo.cipf.es/; Vaquerizas et al., 2005). This method identifies differentially expressed genes and simultaneously controls for the false discovery rate as described by Benjamini and Hochberg (2000). The GEPAS software therefore utilizes the p.adjust function of the Bioconductor stats R package (http://www.bioconductor.org). This analysis yielded nearly identical results, classifying 117 genes as being differentially expressed with p⩽0.01 (55 transcripts under-expressed and 62 over-expressed in the suicide group). A total of 115 genes reached p values ⩽0.01 in both analyses and showed complete accordance. The transcripts exclusively identified by AnEx marginally failed to reach significance in this second analysis reaching p values between 0.0101 and 0.0118. The T-Rex software also provides image maps that visualize the data (Figure 1). The false discovery rate was estimated at 0.12, e.g. 14 out of the 117 transcripts are expected to be false-positive results.
Confirmation of differential gene expression
The validation of the results with a second method is required by all means, as multiple testing may lead to false positives (Chuaqui et al., 2002). For an initial confirmation of the microarray results, we decided to validate changes of three transcripts with p⩽0.05 and fold changes greater than ±2. We chose the genes EFEMP1, PENK and ZFP36 and carried out SYBR Green quantitative RT–PCR experiments. The direction of the expression change could be reproduced for all three genes; however, only the data for EFEMP1 and ZFP36 reached statistical significance (Table 4).
Subsequently, we focused on nine particularly interesting genes (AMPH, CDCA7L, CDH12, CDH22, CHGB, MYR8, PCDHB5, S100A13 and SCN2B) out of the 124 transcripts with p⩽0.01. The direction of expression change could be confirmed for all transcripts except for CHGB, and statistical significance was reached for CDCA7L, CDH12, PCDHB5, S100A13 and SCN2B. Additionally, we assessed the gene expression levels of two transcripts with higher p values, namely MLC1 and PTPRR, which are significantly differentially expressed in the suicide group (Table 4). The results of the quantitative RT–PCR experiments are depicted in Figure 2.
Biological function of identified genes
In order to identify putative biological patterns in the gene expression changes, the identified genes were classified according to their biological function. The Gene Ontology (GO) database (http://www.geneontology.org) organizes gene products according to their biological process, molecular function and cellular component (Harris et al., 2004). The 124 genes with a p value ⩽0.01 and fold change ⩾1.3 were analysed with the GO Tree module of the Web-based Gene Set Analysis Toolkit (WebGestalt http://bioinfo.vanderbilt.edu/webgestalt; Zhang et al., 2005) in order to identify GO categories with significant enrichments. This analysis is based on a hypergeometric test without correction for multiple testing, and we chose the WebGestalt human genome set as the reference set, as the Illumina arrays measure the expression of more than 23000 transcripts. Four GO terms at level 5 with a minimum of four genes are represented more often among the 124 differentially expressed genes than would be expected by chance (Figure 3). These genes belong to the GO categories ‘Central nervous system development’, ‘Homophilic cell adhesion’, ‘Regulation of cell proliferation’ and ‘Transmission of nerve impulse’ (Table 5). This indicates that these biological processes might be involved in suicidal behaviour.
Ratio of enrichment for the category.
Cytogenetic location of the differentially expressed genes
The cytogenetic location of the identified genes was visualized using the WebGestalt Toolkit. The Chromosome Distribution Chart (Figure 4) shows the location of all 124 genes with p⩽0.01; every gene is symbolized by a cross on the appropriate chromosome.
This microarray analysis identified several transcripts to be differentially expressed in the orbitofrontal cortex of suicide victims. The 124 transcripts demonstrated significant changes (fold changes ⩾1.3 p value ⩽0.01), with 59 showing under-, and 65 over-expression in the suicide group. Validation experiments using quantitative RT–PCR showed similar expression changes for almost all investigated transcripts, and nine particularly interesting genes (CDCA7L, CDH12, EFEMP1, MLC1, PCDHB5, PTPRR, S100A13, SCN2B and ZFP36) were found to be statistically differentially expressed when using this independent method.
The S100A13 protein is a homodimeric protein that belongs to the S100 subfamily of EF-hand Ca2+-binding proteins. It is released into the extracellular compartment as a multiprotein complex consisting of S100A13, synaptotagmin (Syt1), and a hFGF-1 homodimer (Carreira et al., 1998) and seems to be involved in the development of the central nervous system (CNS), as it is specifically expressed in the developing human brain (Chan et al., 2003). Interestingly, it is located within a S100 protein family cluster on chromosome 1q21, in a region which has been strongly linked to schizophrenia (Craddock et al., 2005).
The CDCA7L gene encodes a transcription factor which inhibits the monoamine oxidase A (MAOA) promoter and enzymatic activity by binding at the MAOA SP1 core promoter binding site (Chen et al., 2005). MAOA preferentially degrades monoamines including serotonin in the CNS and is thus involved in the central serotonergic responsivity. Interestingly, Du and colleagues found a significant association of a high activity-related MAOA allele with suicidal behaviour in males (Du et al., 2002). A more recent investigation examined a functional polymorphism in the promoter region of the MAOA gene (Courtet et al., 2005). Although there was no overall association with suicidal behaviour, the high-activity variant was significantly over-represented in men who had attempted violent suicide, compared to non-violent attempters. Overall these findings emphasize the potential importance of the MAOA interacting protein CDCA7L for the aetiology of suicidal behaviour.
Additionally, we could confirm the differential expression of PTPRR. The PTPRR protein is a receptor-type protein tyrosine phosphatase that participates in phosphorylation-dependent signalling pathways. Vital cellular functions such as neuronal survival, synaptic plasticity and signal transduction are regulated in part by the balance between the activities of protein-tyrosine phosphatases and protein-tyrosine kinases (Zhang, 2005). PTPRR belongs to a family of protein tyrosine phosphatases that specifically inactivate mitogen-activated protein kinases (MAPKs; Eswaran et al., 2006). These enzymes are highly interesting in the context of suicidal behaviour, as Dwivedi and colleagues (2001) reported significantly changed transcript levels of several genes of the MAPK pathway. This study showed decreased mRNA and protein levels of the MAP kinase 1 (MAPK1) and the MAP kinase 3 (MAPK3) and increased protein levels of the MAP kinase phosphatase 2 (MKP2) in the prefrontal cortex and hippocampus of depressed suicide subjects. The MKP2 protein is a ‘dual function’ protein phosphatase that specifically dephosphorylates MAPK1 and MAPK2 (Gopalbhai and Meloche, 1998).
Another gene which has been found to be differentially expressed is MLC1. Mutations in the MLC1 gene cause the progressive neurological disease megaloencephalic leukoencephalopathy with subcortical cysts, although its exact function is still unknown (Boor et al., 2005). Interestingly, two studies demonstrated the association of mutations in MLC1 with both, schizophrenia and bipolar disorder (Meyer et al., 2001; Verma et al., 2005).
The genes CDH12, EFEMP1 and PCDHB5 belong to the GO category ‘cell adhesion’. These cell adhesion proteins are involved in the neural development and the establishment of neural cell–cell connections. Cell adhesion proteins participate both in the development of the CNS and in signal transduction via synaptogenesis and synaptic vesicle recycling (Togashi et al., 2002). Cadherins and associated proteins seem to contribute physiologically to synaptic plasticity, and a disturbed cadherin function results in abnormal synapse formations (Takeichi and Abe, 2005).
PTPRR, S100A13 and SCN2B play a role in the signal transduction in the CNS. The classification of the 124 genes according to their biological function showed that complex mechanisms like cell–cell adhesion, signal transduction, cell proliferation and the development of the CNS might be involved in the pathophysiology of suicide.
In contrast to this investigation, the recent transcriptional profiling of 19 suicide victims with major depression and 19 non-psychiatric control subjects revealed no significant expression changes on the level of single genes or biological pathways (Sibille et al., 2004). This study utilized Affymetrix U133A microarrays with RNA extracted from the prefrontal cortex (BA 9 and BA 47), and the subjects had been matched according to age, PMI and sex, but not on the base of tissue pH. This difference may to some extent explain the discrepant findings, as tissue pH seems to be crucial for RNA stability and the mRNAs differ in terms of sensitivity to pH (Hynd et al., 2003; Li et al., 2004). Furthermore, the study of Sibille and colleagues (2004) used Affymetrix microarrays whereas our investigation was performed using Illumina Sentrix HumanRef-8 BeadChips. This represents another important difference between the two studies, although these platforms seem to yield comparable data (Barnes et al., 2005).
A Canadian investigation, however, identified several expression changes and demonstrated evidence for the implication of SSAT in suicide and depression (Sequeira et al., 2006). The SSAT gene was differentially expressed in BA 11 of both non-depressed suicide completers, as well as in depressed suicide victims, compared with controls. This transcript exhibited a fold change of −1.4 in the suicide victims. Most interestingly, this result could be confirmed in our sample, where the SSAT transcript showed a 1.3-fold decrease in suicide victims (p=0.052). We subsequently compared the data for all transcripts at a significance level of p⩽0.05 with the results of Sequeira and colleagues (2006). This group published the expression changes for transcripts that were differentially expressed in the orbitofrontal cortex of both depressed suicide victims and non-depressed suicide completers, compared to controls. This investigation was performed using the Affymetrix HG-U133 chipset, and thus cannot be directly compared, as for example, some probes are not specific for a single transcript of one gene. However, as mentioned above, the two platforms seem to generate similar data (Barnes et al., 2005). The comparison of these two experiments showed a great overlap, as seven genes (ATP6V0E, CDC42EP4, CKLFSF6, CTSH, RBP1, SEMA4F and SYT13) showed similar expression changes in the investigation of Sequeira and colleagues (2006) and in our sample.
Two genome-wide linkage studies investigated suicidal behaviour in alcohol-dependent subjects and patients with mood disorders. The study by Hesselbrock and colleagues found genome-wide significant evidence for linkage near the marker D2S1790 at cytogenetic location 2p11 for the qualitative phenotype ‘ever tried suicide’ and modest evidence for the regions near the markers D1S1602 and D3S2398 on chromosomes 1 and 3, respectively (Hesselbrock et al., 2004). The second analysis demonstrated genome-wide adjusted significant ΔLOD scores for the regions 2p12, 5q31-q33, 6q12, 8p22-p21, 11q25 and Xq25-26.1 (Zubenko et al., 2004). The differentially expressed genes PCDHB5 and PDGFRB are located in the region 5q31-33 identified by Zubenko et al., but some megabases (Mb) distant from the marker D5S1480 (Figure 4). Furthermore, HS6ST2 is located near the marker DXS1047 on chromosome Xq26, but also here ∼3 Mb away.
The study is not without limitations. The rate of false-positive findings according to the quantitative RT–PCR data was found to be higher than predicted by the data analysis described above. The divergent findings could be due to secondary RNA structures, splice variants or complex polyA tails that influence both methods (Mimmack et al., 2004). This might, in part, also be caused by the different normalization methods, as the global normalization of the microarray data uses all expression signal intensities and is expected to be more robust. Additionally, the validation of the transcript expression changes on the protein level would be desirable, as there is even less correlation in this case (Chuaqui et al., 2002). This discrepancy may partly be caused by the methodologies used to determine protein expression and the absolute abundance of the proteins. We did not carry out validation experiments on the protein level, as the availability of post-mortem brain tissue is limited and thus restricts the sample size and challenges appropriate matching. Instead, we focused on the validation of particularly interesting results with quantitative RT–PCR, as this method is suitable for high-throughput validation and requires only minimal amounts of material.
Suicidal behaviour is a complex and heterogeneous phenotype, ranging from suicide ideation and suicide attempts to completed suicide. It has been suggested that suicidal behaviour is, at least partially, inherited independently from psychiatric disorders (Brent and Mann, 2005), and therefore we decided to focus more on a certain phenotype of this behaviour rather than on diagnostic categories.
We tried to generate a homogenous group of suicide and selected only subjects for the study that died by means of violent suicide. Six suicide victims of the sample had a history of affective disorders, and there is evidence for alcohol abuse in some individuals of the suicide group. Thus, the suicide group represents a ‘typical’ sample, as affective disorders are found in ∼40% of all cases (Arsenault-Lapierre et al., 2004; Bertolote et al., 2004). Given the relatively small number of subjects, it was not possible to calculate the data for subsamples, e.g. for depressed in comparison to non-depressed suicide victims or controls. Substance-related problems are also common, concerning approximately one out of four suicide victims (Arsenault-Lapierre et al., 2004). This finding is reflected in our sample, where alcohol abuse is reported for two subjects.
In conclusion, we have conducted a microarray study and assessed transcriptional changes in the orbitofrontal cortex of suicide victims. The initial analysis of the arrays identified 124 genes to be differentially expressed with p⩽0.01. The observed changes have been validated for nine genes.
A functional analysis revealed significantly enriched transcriptional alterations of genes involved in cell–cell communication, signal transduction, cell proliferation and the development of the CNS. These alterations might contribute to the pathophysiology of suicidal behaviour.
Statement of Interest