In this study, gene expression was investigated using genome-wide microarrays in 17 brain areas thought to be involved in the neurobiology of suicide and major depression, comparing suicides with and without major depression to psychiatrically normal controls. This is, to our knowledge, the first large-scale brain expression study aiming at identifying global brain alterations associated with suicide and major depression. The extent of the expression changes varied considerably between the diverse brain areas investigated, with certain areas, such as those that comprise the prefrontal cortex and hippocampus, accounting for the majority of expression changes. This is consistent with what one would expect according to neuroanatomical studies of depression and suicide and is also consistent with previous studies looking at discrete brain regions 
. The functional analysis using gene ontologies also revealed that an over-representation of genes involved in cell communication processes were globally altered. More specifically, among genes involved in synaptic transmission, a striking number of GABAergic receptor subunit genes were generally up-regulated among the suicides with major depression, but showed lower expression levels among the 2 other groups. We also observed for the suicide with major depression group a general up-regulation of AMPA receptors subunit genes and a global down-regulation of GRM3 receptors and glutamine synthase (GLUL) gene expression. Our study suggests the presence of consistent alterations of several genes coding for components of the same pathways across different brain regions.
The HG-133AB chipset contains around 44,000 probe sets many of which may not be expressed at biologically significant or detectable levels. Accordingly, Jongeneel et al. estimated that between 10 to 15 thousand transcripts are actually expressed in several types of human cell lines 
. For that reason and in order to reduce the multiplicity problem, we used a combination of filtering methods in order to include in our analysis only transcripts that were actually expressed and reliably detectable. This approach efficiently allows to significantly reduce the total number of analyzed probe sets without notably decreasing the number of truly positive genes 
. This resulted in an average of around 15,000 probe sets analyzed per region Second, in order to control for type I errors, we also used a combination of stringent P
-value thresholds (≤0.01 both at the ANOVA and post-hoc test), as well as a fold change of at least 1.3 in either direction. Most importantly, by focusing on results that replicate across several different brain regions, which constitute partially independent experiments, we are likely to have significantly reduced the occurrence of type I errors in our study.
The current approach led to the identification of 4,472 differentially expressed probe sets over the 17 brain regions (). As expected, and in accordance with the neuroanatomical and post-mortem biomarkers literature, three prefrontal cortex areas, BA8,9, BA10, BA46, and the hippocampus, had the highest number of differentially expressed probe sets, thus confirming the implication of these regions in the pathophysiology of suicide and major depression ( and ). These four areas have been well characterized and have been previously shown to be implicated in suicidal behaviors and depression in numerous post-mortem studies 
. In vivo, neuroimaging studies have also pointed to alterations in the prefrontal cortex and in the hippocampus in patients suffering from major depression 
. Our study provides on a genomic scale, potential molecular targets that may account for those alterations in the brains of suicide victims with and without major depression.
Functional analysis using gene ontologies 
was performed across the 17 regions using a new tool (ErmineJ) that efficiently addresses many of the limitations and problems of the initial gene ontology tools 
, by implementing more comprehensive algorithms and the possibility of performing analyses in parallel. This global functional ontological profiling revealed specific ontological categories commonly overrepresented in all the regions investigated in this study, and further investigation showed that an important proportion of genes belonged to cell-communication processes. Among these, a remarkable number of probe sets corresponded to genes coding for various molecular units of the GABAergic and glutamatergic neurotransmitter systems.
L-glutamic acid (glutamate) and GABA are respectively the main excitatory and inhibitory neurotransmitters in the central nervous system 
. Growing evidence has supported alterations in both of these neurotransmitter systems in major depression 
and suicide 
. Sanacora et al. 
using a magnetic resonance spectroscopy protocol observed elevated levels of glutamate and lower levels of GABA in the occipital cortex of subjects diagnosed with major depression. Furthermore, Hasler et al. 
demonstrated that abnormal reductions in glutamate/glutamine and GABA concentrations are present in the prefrontal cortex of unmedicated depressed patients. Our results are also in concordance with those of Choudary et al. 
, who performed a gene expression study in the cingulate and prefrontal cortex brain areas of suicides and depressed suicides using one of the chips (HG-U133A) of the microarray set used in our study. Interestingly, their results point to similar alterations in glutamate recycling (glutamine synthase, GLUL), glutamate receptors (GRIA1, GRIA3, GRIK1, GRM3) and GABA receptors (GABARB3, GABRD, GABARG2) in depressed suicides versus controls. Also, recently, Merali et al. 
observed altered levels of GABA(A) receptor subunits (α1, α3, α4 and δ) in the BA10 of depressed suicide victims versus non-depressed controls.
GRIA3, which was also confirmed to be differentially expressed by SemiQ RT-PCR (), is of particular interest in suicide as it was significantly down-regulated in the prefrontal cortex in both suicide groups (BA46, ), with and without major depression, suggesting an implication in suicide irrespective of the presence of major depression. This result is particularly important in the light of the recent observation by Laje et al. 
that genetic variation at the GRIA3 gene seems to be associated with suicidal ideation during citalopram therapy and suggests that expression changes in this gene may also confer susceptibility to suicide and suicidal ideation in antidepressant treated patients.
Glia and astroglia in particular are responsible for the uptake,via the glial glutamate transporter (EAAT2) and metabolism and recycling, via glutamine synthase (GLUL) of glutamate 
. Glutamine synthase is responsible for the recycling of glutamate by its conversion into glutamine, which is then released by the astrocytes and taken up at the synaptic terminals where it can be reconverted into glutamate or GABA 
. Glutamine synthase was down-regulated in several prefrontal and parietal areas of brains of suicides with major depression, but not in suicides without major depression suggesting a depression specific dysregulation of glutamate recycling probably leading to altered glutamatergic and/or GABAergic neurotransmission. At the same time the majority of ionotropic glutamatergic receptors differentially expressed were up-regulated in these brain regions in depressed suicides, reinforcing the idea of a substantial alteration of glutamatergic neurotransmission in this group. Given the importance of some of these molecules in glial metabolism, and the growing evidence pointing to astroglial alterations in major depression 
, future studies should investigate cell specific changes in gene expression by means of laser capture microdissection in the brains of depressed suicides.
This hypothesis, if true, is in agreement with the observation that a single dose of Ketamine, an NMDA antagonist, is sufficient to produce a rapid and long lasting antidepressant effect 
. Glutamate seems to mediate stress-induced neuronal atrophy in the hippocampus 
. In addition, although not always consistent, there are different lines of evidence, comprising peripheral studies 
, postmortem brain studies 
, and in-vivo imaging studies 
reporting glutamatergic dysfunction in major depression. Interestingly, glutamatergic neurotransmission is closely controlled by intracellular levels of polyamines, spermine and spermidine being specific modulators of NMDA and AMPA receptors activity 
. Polyamines, and more specifically SSAT, the rate limiting enzyme in the catabolism of polyamines, were associated with suicide and depression in a previous study by our group 
. Polyamines modulate GABAergic and glutamatergic neurotransmission, genes of those systems as well as SSAT were also found to be altered in the present study. In light of these observations, it is important to consider the polyamine-glutamatergic systems as a possible target for future strategies for the treatment of major depression.
Serotonergic and adrenergic dysfunction has been implicated in suicide 
yet we did not detect a significant representation of genes coding for components of these neurotransmitter systems in our differential expression analyses. This result is in accordance with all other microarryay experiments performed to date using suicide brains, where no serotonergic genes have been detected as differentially expressed 
While our microarray experiment sampled multiple brain regions making our analysis global in nature, although enriched for frontal cortical regions due to the implication of these regions in depression and suicide, not all differentially expressed GABAergic and glutamatergic genes were differentially expressed across all regions. In general, we observed a particular probe set as differentially expressed across 2–3 regions, many of which did not overlap with other probe sets identified as differentially expressed. Still, the consistency of the dysregulation in the GABA-glutamate gene systems was striking.
Even though possible limitations regarding pre- and post-mortem factors, such as agonal period, alcohol abuse/dependence and post mortem interval were experimentally controlled for in this study, the conclusions presented here are to be taken with caution and need to be confirmed in an independent and larger sample. Our study design does not allow to clearly differentiating the alterations solely related to suicide from those specific to major depression. This would be possible to resolve only by including a group of matching patients with major depression who did not die by suicide, but such a group would be too difficult to obtain due to the demographic characteristics of suicide victims and depressed patients. Nevertheless, our results are interesting as they shed light into the molecular alterations simultaneously taking place in several important brain regions of individuals with and without major depressive disorder at the moment of their suicide. Another limitation of this study is the choice of 1.3 as a fold-change cut-off. While this allowed us to focus on more robust effects, it prevented us from detecting more subtle changes in gene expression that may be at play.
In conclusion, this is, to our knowledge, the first study attempting to determine global brain expression changes taking place in the brain of suicide victims with and without major depression. We observed global changes in genes implicated in synaptic transmission, and more specifically, in genes involved in GABAergic (inhibitory) and glutamatergic (excitatory) neurotransmission. Further studies are warranted in order to examine in detail the cellular origin of the alterations observed in our analyses, to validate the observed changes using complementary approaches and to investigate possible genetic factors related to the observed alterations.