A gene expression survey by microarrays in three brain regions (DLPFC, NAcc and ACC) revealed alterations in the serotonergic and second messenger systems in the DLPFC and a consistent down-regulation of several members of the metallothionein subfamilies 1 and 2 genes in the NAcc and ACC in mood disorder suicides. Our strategy consisted on using high quality brains (short PMI, high pH, high RIN) with no history of drug dependence, as well as the development of a general linear model in which factors known to influence RNA quality and expression were taken into account when exploring the suicide specific expression changes. This is to our knowledge the first study investigating suicide specific changes in mood disorders in three highly relevant areas as they are involved in the control of impulses and in mood disorders (DLPFC, ACC) and in motivation-reward mechanisms (NAcc).
We studied gene expression changes in the DLPFC, an area previously associated with both mood disorders, control of impulses and suicide. Interestingly, the most statistically significant gene differentially expressed in suicide victims was the 5-HT2A serotonergic receptor, with a1.28 fold-change reduction in mood disorder subjects that committed suicide versus non-suicide mood disorder subjects. Genetic association studies have often shown that mutations in the 5-HT2A receptor are associated with mood disorders and suicide 
, as well as with the response to antidepressants 
. Additionally, gene expression post-mortem studies have revealed alterations 5-HT2Areceptor levels in mood disorder subject suicides when compared to controls. However the direction of the alterations in 5-HT2A is controversial, with some reports showing elevations in the 5-HT2A binding and expression 
while others showed reduced levels or no differences 
. Recently, a well designed study including matched suicides and controls also confirmed the absence of changes in 5-HT2 binding but reported a decrease in neuronal densities in the prefrontal cortex in a sample of predominantly major depressed suicides 
. In another recent study designed to investigate 5-HT2A levels in major depression, Shelton et al. reported an increase in 5-HT2A expression in major depression subjects, but reported no differences when comparing suicides versus non-suicide subjects 
. In this study, we aimed to investigate suicide specific changes in mood disorder patients and observed a significant decrease in 5-HT2A levels in suicide which can at least partially explain previous discordant findings when comparing mood disorder to controls or suicides to controls. The reduction in 5-HT2A levels is also in concordance with another study comparing suicide versus non-suicide schizophrenia subjects using also Affymetrix microarrays 
, suggesting that this reduction in 5-HT2A levels might be involved in suicide irrespective of the diagnosis. We also observed decreased expression in suicide group of the brain specific ADCY1, a brain specific form of adenylate cyclase, suggesting altered neurotransmission and cyclic AMP signaling in suicide and confirms previous observations pointing to cAMP signaling alterations in suicide and mood disorders 
and in animal models of depression 
. Both serotonergic alterations and second messenger systems underline global neurotransmission alterations in the DLPFC that seem to play a role in the predisposition to commit suicide.
In both the ACC and NAcc decreased expression was observed in gene members of two subfamilies (1 and 2) of the metallothionein super family in the suicides versus the non-suicides mood disorder subjects. These changes were confirmed successfully Metallothioneins are small single chain proteins with a high content of cysteine residues organized in specific sequences enabling the formation of thiolate cluster and the binding of certain metals react to oxidative stress, glucocorticoids and inflammatory mediators 
. There are four subfamilies of metallothioneins proteins in mammals corresponding to types 1 to 4, all metallothionein genes are closely located in chromosome 16q13. MT1 and MT2 are expressed ubiquitously and highly co-expressed, whereas MT4 is not expressed in the brain and MT3 is brain specific and expressed mainly in hippocampus amygdala and cortex in zinc enriched neurons and astrocytes 
. Consistent with this, MT3 was also observed to be expressed in the three brain areas investigated but was not differentially expressed like MT1 and 2 subfamilies. MT3, unlike MT 1 and 2 is not regulated by glucocorticoids and stress, which can explain this difference 
MT1 and MT2 are not only known to be responsive to several forms of stress and to glucocorticoids 
, they also play an important role in neuroprotection 
. In animal models of stress, both MT1 and MT2 mRNA were elevated as much as 30-fold following just 12 h (one cycle) of restraint stress whereas MT3 gene expression was not altered by stress 
. The induction occurred at the transcriptional level and was mediated by the activation of glucocorticoid receptor 
. In a stress mouse model, treatment with a glucocorticoid receptor antagonist (RU 486) prior to restraint stress inhibited MT induction by at least 50% indication that at least in part MT induction is under direct control of the glucocorticoid receptor at the transcriptional levels 
. Glucocorticoids induction of metallothioneins expression is done through the direct binding of the ligand activated glucocorticoid receptor to the glucocorticoid response elements (GRE) in the promoter of MTs 
. Dexamethasone has been shown for instance to increase significantly the levels of MT and immobilization stress has also been shown to increase selectively MT1 and MT2 of rat brains 
. MT1 and 2 play a neuroprotective role against stress and glucocorticoids elevation by a mechanism of induced expression of metallothioneins in astrocytes followed by an extracellular release and internalization by neurons after binding to the receptor megalin that mediates the transport of MTs into neurons 
. Thus, MT1 and 2 sub-families, that play an important role in neuroprotection from the effects of behavioral stress and in response to cortisol, are significantly reduced in mood disorder suicides compared to non-suicides. Interestingly, in another recent microarray study using Affymetrix Exon 1.0 ST arrays, Shelton et al. observed and confirmed by qPCR a decreased expression of MT1M in the BA10 of depressed subjects compared to controls 
Alterations in the hypothalamic-pituitary-adrenal (HPA) axis in depressive states and suicide have been consistently observed for more than 40 years 
. It has also been extensively shown that depressed states are specifically linked to both chronic stress and elevated cortisol levels relevant of HPA axis hyperactivity in depression 
. Many studies have observed higher baseline cortisol values in young depressed patients 
and cortisol hypersecretion in depression 
. We propose in this study that decreased levels of MT1 and 2 can result in a higher sensitivity in suicide completers to the effects of depression-associated cortisol levels in the brain. If these results are confirmed in a larger independent cohort, pharmacological agents that increase the expression of MT 1 and 2 metallothionein sub-families or their transport to neurons by the megalin receptor without necessarily altering HPA axis activity can be envisioned for the treatment and prevention of suicide and suicidal behaviors in mood disorder patients.
The effect of known confounding variables in post-mortem brain gene expression studies (pH, PMI, age, gender, diagnosis and RNA quality) was controlled by conducting an ANCOVA using significant variable as covariates. One of the weaknesses of this study is the relatively small size of the microarray sample. However, we were able to confirm the observed results in a larger sample and to explore MT expression in control subjects also by qPCR in a larger sample in both the ACC and NAcc ( and
). Another problem observed was the difference in age between the non-suicides and the suicides in the NAcc and ACC. This was expected as suicide is particularly prevalent in younger populations 
with about 200,000 young people that die every year in the world by suicide 
. For that reason, we included age into the statistical model in order to exclude genes altered by age only. Ultimately however, MT subfamilies 1 and 2 down-regulation in suicide mood disorder subjects should be tested in peripheral tissue such as blood lymphocytes, where they have been shown to be expressed 
, in order to confirm these results in a clinical population and to investigate the validity of these changes as a biomarker for suicide.
In conclusion, while expression profiling confirmed and further extended the implication of serotonergic genes in suicide in the DLPFC, the present study also reveals a significant decrease in expression of metallothioneins subfamilies1 and 2, a family of genes involved in neuroprotection and in cortisol/stress response, in suicide victims compared to non-suicide mood disorder patients. Because cortisol levels are known to be elevated in depressive states, the observed decreased in MT expression in the suicide group might point to a failed neuroprotective mechanism against cortisol in this group and point to a new marker and possible pharmacological target for suicide and suicidal behavior.