Quality control (QC) parameters for RNA. The relationships between RNA QC (determined by the average of the 5′/3′ signal ratios of β-actin and glyceraldehyde-3-phosphate dehydrogenase across U133A and U133B chips) and the postmortem interval (PMI) in Brodmann area (BA) 4, BA11, and BA8/9 show no effect of the PMI on the QC measurements.
Results of principal component analysis. Analysis was based on the approximately 22 000 genes on the HG-U133A chip from control samples in 3 cortical regions (Brodmann area [BA] 11, BA4, and BA8/9) and the cerebellum. The first 3 components accounted for 31.7% of the total variance. Components 1, 2, and 3 accounted for 14.1%, 10.7%, and 7.0% of the variances, respectively.
Venn diagrams show the number of genes identified as differentially expressed and the overlap of genes between the different comparisons in the motor cortex (Brodmann area [BA] 4), the dorsolateral prefrontal cortex (BA8/9), and the orbital cortex (BA11). Each circle represents a single contrast. Upward and downward arrows indicate the number of up- and down-regulated genes. The numbers of genes that are differentially expressed in a single contrast are shown in parentheses. The intersections of the circles indicate the number of genes common between contrasts. The between-region Venn diagram shows the differentially expressed genes in BA4, BA8/9, and BA11. The numbers of genes that are differentially expressed in a single brain region are shown in parentheses. The intersections of the circles indicate the number of genes differentially expressed between brain regions. C indicates control subjects; DSC, depressed suicide completers; and SC, suicide completers.
Hierarchical cluster analysis of the 375 genes identified as being differentially expressed in Brodmann area (BA) 11. Both expression patterns in individuals and genes were clustered. A, Clustered image map. The color and intensity indicate direction and level of change. Blue spectrum colors indicate down-regulated expression; red spectrum colors, up-regulated expression. C indicates controls; DSC, depressed suicide completers; and SC, suicide completers. B, Cluster set plots display the average expression profile (light green line) of all genes in each of the clusters, along with the minimum and maximum deviation around the mean (black vertical lines). C, Principal component analysis based on the differentially expressed genes. The first 3 components accounted for 63.3% of the total variance (components 1, 2, and 3 accounted for 41.9%, 15.9%, and 5.5% of the variance, respectively). The DSC, SC, and C groups show separation on the first component. D, Graphical representation of the percentage of the most common gene ontology terms within each of the clusters in BA11. ATP indicates adenosine triphosphate.
Motor cortex (Brodmann area [BA] 4) analysis. A, Clustered image map of the hierarchical cluster analysis of the 163 genes identified as being differentially expressed genes in BA4. Both expression patterns in individuals and genes were clustered. The color and intensity indicate direction and level of change. Blue spectrum colors indicate down-regulated expression; red spectrum colors, up-regulated expression. Most of the genes (99 or approximately 60%) were misregulated in the depressed suicide completers (DSC) in relation to controls (C). Of these, 63 were up-regulated and 36 were down-regulated in relation to the C group. SC indicates suicide completers. B, Cluster set plots display the average expression profile (light green line) of all genes in each of the clusters, along with the minimum and maximum deviation around the mean (black vertical lines). C, Principal component analysis based on the differentially expressed genes. The first 3 components accounted for 61.2% of the total variance. Components 1, 2, and 3 accounted for 36.4%, 20.0%, and 4.9% of the variance, respectively. The DSC, SC, and C groups show separation on the first component. D, Graphical representation of the percentage of the most common gene ontology terms within each of the clusters. ATP indicates adenosine triphosphate.
Dorsolateral prefrontal cortex (Brodmann area [BA] 8/9) analysis. A, Clustered image map of the hierarchical cluster analysis of the 282 genes identified as being differentially expressed genes in the BA8/9. Both expression patterns in individuals and genes were clustered. The color and intensity indicate direction and level of change. Blue spectrum colors indicate down-regulated expression; red spectrum colors, up-regulated expression. Most of the genes were misregulated in suicide completers (SC) in relation to controls (C) (140 or approximately 50%) and depressed suicide completers (DSC) (75 or approximately 25%). B, Cluster set plots display the average expression profile (light green line) of all genes in each of the clusters, along with the minimum and maximum deviation around the mean (black vertical lines). C, Principal component analysis based on the differentially expressed genes. The first 3 components accounted for 65.6% of the total variance. Components 1, 2, and 3 accounted for 43.6%, 17.2%, and 4.7% of the variance, respectively. The DSC, SC, and C groups show separation on the first component. D, Graphical representation of the percentage of the most common gene ontology terms within each of the clusters. ATP indicates adenosine triphosphate; GTP, guanosine triphosphate.
Graphic representations of the observed changes in spermidine/spermine N1-acetyltransferase gene (SSAT) expression. Raw Affymetrix data (Microarray Analysis Suite version 5.0; available at: http://www.affymetrix.com) illustrate the differential expression of SSAT in Brodmann area (BA) 4, BA8/9, and BA11. C indicates controls; DSC, depressed suicide completers; and SC, suicide completers.
Spermine/spermidine N1-acetyltransferase gene (SSAT) and β-actin polymerase chain reaction (PCR) products after reverse transcription and PCR amplification. A, Agarose gel stained by ethidium bromide. Bands illustrated as an example are from Brodmann area (BA) 4. C indicates controls; DSC, depressed suicide completers; and SC, suicide completers. B, Distribution of the SSAT342A/C locus in a sample of SC (n = 181) and matched controls (n = 80). CI indicates confidence interval; OR, odds ratio. C, Graphical representation of the relative SSAT messenger RNA (mRNA) levels in C, SC, and DSC groups (percentage of β-actin). The semiquantitative analysis by reverse transcription–PCR of SSAT mRNA levels was performed in the motor cortex (BA4), dorsolateral prefrontal cortex (BA8/9), and orbital cortex (BA11). Asterisks indicate significant differences compared with the C group.
Confirmation of the spermine/spermidine N1-acetyltransferase gene (SSAT) expression differences at the protein level. A, Immunohistochemical staining photomicrographs of adjacent brain sections using an anti-SSAT polyclonal antibody (1:75) in the motor cortex (Brodmann area [BA] 4), the dorsolateral prefrontal cortex (BA8/9), and the orbital cortex (BA11) of a control subject (C), a suicide completer (SC), and a suicide completer with major depression (DSC). Arrows indicate examples of positive immunohistochemical staining. B, Western blot analysis using an anti-SSAT polyclonal antibody (1:1000) in the motor cortex (BA4), dorsolateral prefrontal cortex (BA8/9), and orbital cortex (BA11). Asterisks indicate significant differences in comparisons with the control group.
Sequeira A, Gwadry FG, ffrench-Mullen JMH, Canetti L, Gingras Y, Casero RA, Rouleau G, Benkelfat C, Turecki G. Implication of SSAT by Gene Expression and Genetic Variation in Suicide and Major Depression. Arch Gen Psychiatry. 2006;63(1):35-48. doi:10.1001/archpsyc.63.1.35
Copyright 2006 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2006
A large body of evidence suggests that predisposition to suicide, an important public health problem, is mediated to a certain extent by neurobiological factors.
To investigate patterns of expression in suicide with and without major depression and to identify new molecular targets that may play a role in the neurobiology of these conditions.
Brain gene expression analysis was performed using the Affymetrix HG-U133 chipset in the orbital cortex (Brodmann area [BA] 11), the dorsolateral prefrontal cortex (BA8/9), and motor cortex (BA4). Subsequent studies were carried out in independent samples from adjacent areas to validate positive findings, confirm their relevance at the protein level, and investigate possible effects of genetic variation.
We investigated 12 psychiatrically normal control subjects and 24 suicide victims, including 16 with and 8 without major depression, in the brain gene expression analysis, validation, and protein studies. The genetic studies included 181 suicide completers and 80 psychiatrically normal controls. All subjects investigated were male and of French Canadian origin.
Main Outcome Measures
Gene expression measures from microarray, semiquantitative reverse transcription–polymerase chain reaction, immunohistochemistry, and Western blot analyses.
Twenty-six genes were selected because of the consistency of their expression pattern (fold change, >1.3 in either direction [P≤.01] in at least 2 regions). The spermine/spermidine N1-acetyltransferase gene (SSAT) was successfully validated by reverse transcription–polymerase chain reaction, immunohistochemistry, and Western blot analyses. A variant located in the SSAT polyamine-responsive element regulatory region (SSAT342A/C) demonstrated a significant effect of genotype on SSAT brain expression levels (F1 = 5.34; P = .02). Further investigation of this variant in an independent sample of 181 male suicide completers and 80 male controls showed a higher frequency of the SSAT342C allele among suicide cases (odds ratio, 2.7; 95% confidence interval, 1.4-5.3; P = .005), suggesting that this allele may increase predisposition to suicide.
These data suggest a role for SSAT, the rate-limiting enzyme in the catabolism of polyamines, in suicide and depression and a role for the SSAT342 locus in the regulation of SSAT gene expression.
Suicide is a major public health problem, and in many countries it is the leading cause of death for men younger than 35 years.1 During the past few decades, it has become increasingly clear that individuals who commit suicide have a certain biological predisposition, part of which is given by genes.2 Psychopathology, particularly major depressive disorder, is commonly associated with suicide, but the genetic liability to suicide is likely independent from the liability to psychiatric disorders.3- 6 A growing effort has been in place to identify biological markers for suicide and depression,7 but most of the studies have focused on components of the serotonergic2 and noradrenergic systems.2 However, it is clear that additional systems play a role in the neurobiology of these conditions.
This study aimed to identify new molecular targets that may play a role in the neurobiology of suicide with and without major depression. To identify potential risk factors, a gene expression study was initially conducted as a screening strategy. This was conducted in the following 3 brain cortical regions: Brodmann area (BA) 4, BA8/9, and BA11. Postmortem and neuroimaging studies during the past decades have implicated BA8/9 (dorsolateral prefrontal cortex) and BA11 (orbital cortex) in suicide and depression.8,9 Because of the common motor function alterations in depressed patients,10,11 BA4 (motor cortex), primarily a motor region linked to motor deficits,12 was also investigated. The brain gene expression screening study identified the gene that codes for spermine/spermidine N1-acetyltransferase (SSAT) as an interesting target, which was further investigated using a series of complementary strategies. Herein we present evidence suggesting that polyamines, and particularly SSAT, may play a role in the predisposition to suicide and major depression.
All subjects were male and of French Canadian origin, a homogeneous population with a founder effect.13 Cases and control subjects were matched on the basis of age and postmortem interval (PMI). All subjects died suddenly without a prolonged agonal state or protracted medical illness. Brain samples were obtained from the Quebec Suicide Brain Bank, Montreal, and were collected with PMIs of less than 36 hours at autopsy. We sampled BA4, BA8/9, and BA11 at 4°C and snap-froze the samples in liquid nitrogen before storage at −80°C. This study was approved by our local institutional review board, and informed consent was obtained from next of kin.
All subjects were psychiatrically characterized by psychological autopsies, which are validated methods to reconstruct psychiatric history by means of extensive proxy-based interviews, as outlined elsewhere.3 Briefly, psychological autopsies were performed for Axis I of the DSM-IV as assessed by the Structured Clinical Interview for DSM-IV, which was administered by trained clinicians with an average of 2 informants per family. After the interview, we reviewed the coroner's notes and all relevant medical records and wrote a case report for the purpose of a best-estimate diagnosis. Best-consensus DSM-IV Axis I diagnoses were made by a panel of psychiatrists after analysis of the case reports. The sample consisted of 16 depressed suicide completers (depressed suicide group) who died during an episode of major depression, 18 suicide completers (suicide group) with no lifetime history of major depression, and 12 matched controls with no history of suicidal behavior or a major psychiatric diagnosis.
Samples from all subjects included in the microarray screening were used for the semiquantitative reverse transcription–polymerase chain reaction (RT-PCR) and Western blot experiments. Three subjects per group and per region (n = 27) were included in the immunohistochemistry studies. Subjects were selected with the researchers blinded to the microarray results, and selections were based on the quality of the tissue for immunohistochemistry experiments.
The population-based gene association study was conducted in a larger sample consisting of all subjects included in the microarray studies and an independent sample totaling 181 male suicide completers and 80 male controls, all of whom were of French Canadian origin. Suicide completers were consecutively collected from the Montreal Central Morgue, Montreal, Quebec, and controls were psychiatrically normal subjects according to Diagnostic Interview Schedule assessments that were drawn from the Quebec general population and were matched by age, sex, and ethnic origin.
Extractions of RNA used in the present study had a minimum A260/A280 ratio of more than 1.9. The samples were further checked for evidence of degradation and integrity. Samples had a minimum 28S/18S ratio of more than 1.6 (2100 Bioanalyzer; Agilent Technologies, Palo Alto, Calif). We used the Human Genome U133 set, which consists of 2 GeneChip arrays with 45 000 probe sets representing more than 39 000 transcripts derived from approximately 33 000 well-substantiated human genes (available at: http://www.affymetrix.com).
GeneChip analysis was performed with Microarray Analysis Suite version 5.0, Data Mining Tool 2.0, and Microarray database software (available at: http://www.affymetrix.com). All of the genes represented on the GeneChip were globally normalized and scaled to a signal intensity of 100.
Various microarray RNA integrity indicators were used in this study (Table 1 and Table 2) to filter samples for quality for final analysis. Principal component analysis (PCA) was used to identify outlier arrays quickly. Microarray quality control parameters included the following: noise (RawQ), consistent number of genes detected as present across arrays, consistent scale factors, and consistent β-actin and glyceraldehyde-3-phosphate dehydrogenase 5′/3′ signal ratios. Outlier subjects were excluded on a region basis without any subject being excluded from all the regions. Similar numbers of subjects were included in the final analysis across the 3 regions (for BA4: 7 controls, 5 suicide completers, and 10 depressed suicide completers; for BA8/9: 6 controls, 6 suicide completers, and 7 depressed suicide completers; and for BA11: 6 controls, 5 suicide completers, and 8 depressed suicide completers).
We selected genes for analysis on the basis of “present calls” by Microarray Analysis Suite 5.0. In the present study, for a gene to be included, it had to be present (detectable) in at least 75% of the subjects in at least 1 of the 3 groups to reduce the chances of false-positive findings. Expression data were analyzed using Genesis (GeneLogic, Gaithersburg, Md) and AVADIS software (Strand Genomics, Redwood City, Calif). Gene expression values were floored to 1 and then log2-transformed.
One-way analysis of variance was performed for each gene to identify statistically significant gene expression changes. To identify differences between depressed and nondepressed suicide completers, statistically significant genes were subjected to a post hoc test for the contrasts of depressed suicide completers vs controls, suicide completers vs controls, and depressed suicide completers vs suicide completers. In all, 2 criteria were used to determine whether a gene was differentially expressed. A gene had to have a 1-way analysis of variance P value of less than or equal to .01. Second, for a given contrast, a gene had to have a fold change (FC)–P value combination of 1.3 FC in either direction and P≤.01.
Cluster analysis was performed using average-linkage hierarchical cluster analysis with a correlation metric. Both expression patterns in individuals and genes were clustered. We performed PCA on the basis of the initial gene sets and on the selected genes (according to our significance criteria). The PCA based on the initial gene set did not discriminate the 3 groups; the PCA based on the selected genes did.
We performed RT in a total volume of 40 μL with 2 μg of total messenger RNA using M-MLV RT (Gibco BRL Life Technologies, Burlington, Ontario) and oligo(deoxythymidine)16 primers. We performed PCR amplification using AmpliTaq Gold (Applied Biosystems, Foster City, Calif). Messenger RNA–specific primers were designed using Primer3 (available at: http://www-genome.wi.mit.edu/cgi-bin/primer/primer3_www.cgi) to avoid amplification of contaminating genomic DNA. The PCR products were visualized using ethidium bromide staining after electrophoresis in a 3% agarose gel. Images were digitalized and analyzed using Gene Tools software (Syngene, Cambridge, England). New samples collected from adjacent tissue from all of the subjects included in the microarray expression studies were used in this analysis (16 depressed suicide completers, 8 suicide completers, and 12 controls).
Sections from tissue adjacent to that used for the microarray experiment were used for immunohistochemistry. Three subjects per group and per region (n = 27) were included in the analysis, and on average 3 slides per subject were examined. Immunohistochemical labeling was performed using standard protocols. In brief, frozen samples were sectioned at 10 μm, air dried at room temperature, fixed in acetone, and conserved at −80°C. Before the incubation with the primary antibody, slides were acclimated at room temperature for 15 minutes and incubated with Tris-buffered saline solution for 10 minutes and with normal rabbit antiserum (Santa Cruz Biotechnology Inc, Santa Cruz, Calif) to avoid nonspecific binding. The SSAT antibody was diluted at 1:75. For the staining, the LSAB2 system peroxidase (Dako Corp, Carpinteria, Calif) was used according to the manufacturer's indications. The sections were evaluated by 2 of us (A.S. and L.C.) who were blinded to phenotype and brain region. The immunopositive cells were counted with ImageJ software (version 1.29x; National Institutes of Health, Bethesda, Md).
Western blot analyses were carried out on additional samples adjacent to the previous dissections from all the subjects used in the microarray experiments (16 depressed suicide completers, 8 suicide completers, and 12 controls). Briefly, 50 μg of total protein was loaded in 4% to 20% precast gels (Tris-Glycine; Invitrogen Corp, Carlsbad, Calif) and transferred onto nitrocellulose membranes. Membranes were blocked with 6% milk in Tris-buffered saline with Tween and hybridized to the anti-SSAT polyclonal primary antibody (1:1000) overnight at 4°C and then to a peroxidase-conjugated secondary antibody (1:5000; Santa Cruz Biotechnology Inc). The proteins were visualized by means of chemiluminescence (Bio-Rad Laboratories, Hercules, Calif). For standardization and comparisons, the membranes were also hybridized to a primary anti–β-actin antibody (1:5000; Sigma-Aldrich Corp, St Louis, Mo). Films were digitalized, and the bands were counted with Gene Tools software.
Genomic DNA was extracted from blood or from frozen brain tissue samples using standard procedures.14 A description of the PCR method used for amplification can be found elsewhere.15 Genotyping was performed using the SNaPshot16 procedure and the ABI 3100 genetic analyzer (Applied Biosystems) following the manufacturer's instructions. Genotypes were automatically generated using GeneScan 1.0 and Genotyper 1.0 (Applied Biosystems). Four single nucleotide polymorphisms were genotyped, 2 in the coding sequence (SSAT460 and SSAT495) and 2 located a few nucleotides away from the polyamine responsive element motif and within the regulatory region (SSAT342 and SSAT624). The SSAT342A/C genotypes were confirmed by MspI (New England Biolabs, Mississauga, Ontario) digestion followed by visualization by ethidium bromide staining after electrophoresis in a 2.5% agarose gel.
Analysis of demographic parameters showed no significant difference in terms of age and PMI between the groups (Table 3). Consistent with previous reports,17- 19 analysis of PMI on RNA quality control parameters showed no significant effect in our sample (Figure 1).
As an initial evaluation of the discriminatory ability of the microarray gene expression assay, we compared cortical samples (BA4, BA8/9, and BA11) from normal controls with samples obtained from the cerebellar tissue of the same individuals (n = 9). We performed PCA on the basis of approximately 22 000 genes on the HG-U133A chip (Figure 2). Spatial separation on the first component demonstrates the sensitivity and power of the microarray experiment design to detect the subtle gene expression pattern changes between the cortical and cerebellar tissue due to their respective functional and neuroanatomical particularities.
To reduce the number of comparisons and the chances of false-positive findings, statistical comparisons were performed on selected genes instead of the total number of genes present in the chipset. The Venn diagrams in Figure 3 summarize the number of differentially expressed genes observed per comparison using as criteria P≤.01 and FC≥1.3 and the extent of overlap of genes differentially expressed between the different brain regions. The common genes differentially expressed in BA4, BA8/9, and BA11 between the 2 suicide groups vs the controls are represented in the intersection of the Venn diagrams. A total of 26 genes were common between 2 or 3 regions, and a total of 200 common genes were differentially expressed between the groups across brain regions (Figure 3). As shown in Table 4, the FCs for the 2 comparisons, depressed suicide completers vs controls and suicide completers vs controls, are highly consistent and going in the same direction for all the genes in all the regions.
The PCA performed with differentially expressed genes in each region showed good separation on the first component between the 3 groups (Figure 4C for BA11, Figure 5C for BA4, and Figure 6C for BA8/9). The spatial discrimination observed demonstrates that the difference between the groups of subjects is based on genes that are differentially coregulated between the groups.
In BA11, 375 genes from an initial set of 14 864 selected genes were identified as being differentially expressed (Figure 3). Most of these genes (275/375) were misregulated in the depressed suicide completers compared with the controls as shown by the Venn diagrams (Figure 3). Cluster analysis showed mainly 4 clusters of genes with similar expression as seen in the cluster image map in Figure 4A and B. Genes in cluster 1 (Figure 4B) were on average overexpressed in both suicide groups when compared with the controls but showed no differences between the 2 suicide groups, suggesting that genes in this cluster are coregulated in the same manner in suicide completers indistinct of the diagnosis. Cluster 2 genes were on average down-regulated in the depressed suicide completers vs the suicide completers and the controls, with no major differences between the suicide completers and the controls, suggesting that these genes may be associated with depression. The SSAT gene implicated in the stress response was present in this cluster. A graphical representation of SSAT levels in BA11 is shown in Figure 7. Clusters 3 and 4 were much smaller and showed an opposite pattern from each other. Gene ontology analysis based on these clusters showed clear differences in terms of the most common gene ontology terms between the clusters (Figure 4D).
In BA4, 162 genes were identified as being differentially expressed from the initially selected 14 034 genes (Figure 3). Cluster analysis of differentially expressed genes in BA4 showed 4 main clusters of genes (Figure 5A). The largest cluster was composed of 61 genes on average overexpressed in the 2 suicide groups compared with controls (Figure 5A and B, cluster 1). The second cluster in BA4 consisted of 45 genes generally down-regulated among the depressed suicide completers compared with the suicide completers and the controls and mostly implicated in protein metabolism and signal transduction. The SSAT gene was also present in this cluster, and a graphic representation of its differential expression is shown in Figure 7. Cluster 3 was composed of 37 genes with average lower expression among suicide completers compared with depressed suicide completers and the controls. Cluster 4 was the smallest in BA4, with 21 genes mainly having higher expression in both suicide groups vs the controls. Gene ontology analysis was also performed in the BA4 based on the clusters observed and showed clear differences in terms of the most frequent terms between the clusters (Figure 5D).
In BA8/9, 282 of an initial set of 14 519 selected genes were identified as being differentially expressed (Figure 3). The distribution of differentially expressed genes shows that most genes were misregulated in the suicide completers in relation to the controls (140 or approximately 50%) and to the depressed suicide completers (75 or approximately 25%). Nine genes, of which 5 were up-regulated and 4 were down-regulated, were common in the suicide groups in relation to the controls (Figure 3 [Venn diagram]). Cluster analysis of BA8/9 showed a particular expression pattern, with the 2 major clusters of genes showing, on average, primarily differences between the suicide completers vs the depressed suicide completers and controls, suggesting that this region may be more suicide specific (Figure 6A). Cluster 1 was composed of 151 genes that were on average less expressed among the suicide completers compared with the controls. Cluster 2 was the second largest cluster in BA8/9 with 114 genes. Genes in this cluster were also differentially expressed in the suicide completers compared with the depressed suicide completers and controls, with the difference that they were overexpressed in the suicide completers in this cluster. Cluster 3 was composed of only 17 genes mainly up-regulated in the 2 suicide groups compared with the controls. Gene ontology analysis based on the 3 clusters observed in BA8/9 also showed clear differences, confirming the specificity of these clusters (Figure 6D).
A total of 26 genes were found to be differentially expressed in at least 2 of the 3 regions investigated as shown by the intersections in the Venn diagram (Figure 3 and Table 5). According to 1-way analysis of variance, one of these genes, SSAT, was differentially expressed in BA4 and BA11 at the P<.001 level and in BA8/9 at the P<.05 level (Figure 7). In BA4, SSAT was significantly down-regulated in the depressed suicide completers and suicide completers in relation to the controls, with FCs of −1.6 (P = .005) and −1.4 (P = .02), respectively. In BA11, SSAT was significantly down-regulated in suicide groups in relation to the controls, with FCs of −1.8 (P = .002) and −1.4 (P = .005), respectively. Finally, in BA8/9, SSAT was down-regulated in the depressed suicide completers in relation to the controls, with an FC of −1.4 (P = .02).
As indicated in Table 3, some of the subjects included in the study had psychopathology other than major depressive disorder. In addition, several subjects had a history of substance dependence/abuse, which may act as an important confounder of the gene expression analyses. An analysis of covariance controlling for a history of substance dependence/abuse, presence of psychopathology other than major depressive disorder, and result of the postmortem toxicological screening examination indicated that the group effect on SSAT expression in BA4 (F2 = 3.89; P = .04) and BA11 (F2 = 12.2; P = .001) was independent of the effect of these possible confounders.
Differential expression of SSAT in BA4, BA8/9, and BA11 was confirmed by semiquantitative RT-PCR analyses (Figure 8A and C) on independent samples from the same individuals. Lower expression of SSAT was confirmed in BA4 for the suicide completers (FC = −1.2) and depressed suicide completers (FC = −1.3) compared with the controls (F2 = 3.75; P = .04). In BA8/9, SSAT expression was lower among the suicide completers and depressed suicide completers compared with the controls, with FCs of −1.5 and −1.4, respectively (F2 = 8.08; P = .002). Similarly, in BA11, SSAT expression was 1.5-fold lower in the suicide completers and 1.4-fold lower in depressed suicide completers than in the controls (F2 = 9.20; P = .01).
Altered expression at the transcriptional level does not necessarily lead to altered protein expression, and SSAT is known to undergo extensive posttranscriptional regulation.20,21 Therefore, confirmation of the observed changes at the protein level was carried out by immunohistochemistry analysis in tissue sections prepared from the same brain regions using an SSAT polyclonal antibody.22Figure 9A illustrates the observed changes in SSAT immunoreactivity in BA4, BA8/9, and BA11 of a control, a suicide completer, and a depressed suicide completer. Quantification of immunopositive cells in a subgroup of subjects (3 subjects per group per region [n = 27]) showed a lower SSAT protein expression in both depressed suicide completers (FC2 = 1.36; P = .04) and suicide completers (FC2 = 1.35; P = .005) compared with controls. The changes at the protein level were also confirmed by quantification of SSAT by Western blot analysis in adjacent samples from the same regions from all subjects initially selected for the microarray analysis (16 depressed suicide completers, 8 suicide completers, and 12 controls). Significant alteration of SSAT protein levels was observed between the groups in BA4 (F2 = 4.47; P = .02), BA8/9 (F2 = 4.80; P = .02), and BA11 (F2 = 4.08; P = .03), with lower expression among the suicide groups, particularly the depressed suicide completers, compared with the controls (Figure 9B). Thus, the microarray evidence, confirmed by the semiquantitative RT-PCR results, reflect relevant changes at the protein level and suggests a possible role of SSAT in the pathophysiology of suicide and depression.
It was recently shown that SSAT expression is closely regulated by a cis-acting polyamine-responsive element located in the promoter region of this TATA-less gene.23 We studied the genetic variation at 4 loci in the SSAT gene and investigated their influence on SSAT expression. Because SSAT is an X-linked locus not located in the pseudoautosomal region, male subjects are hemizygous, and because all brains were from male donors, we could investigate the direct relationship between SSAT allelic variants and the altered expression of SSAT. The only polymorphic locus, SSAT342A/C, which is located in the polyamine-responsive element–regulatory region, showed a significant effect on SSAT expression levels in BA4, BA8/9, and BA11 (F1 = 5.34; P = .02), with subjects having the SSAT342A variant showing more expression. Because we observed lower SSAT expression levels in the brains of suicide completers, we hypothesized that suicide completers from the general population would have less frequency of SSAT342A. To test this hypothesis, a sample of 181 French Canadian male suicide completers and 80 French Canadian male controls from the general population underwent genotyping. This analysis (Figure 8B) showed a protective role of the SSAT342A variant because not having this variant significantly increased the risk of committing suicide (odds ratio, 2.7; 95% confidence interval, 1.4-5.3; P = .005). Thus, the SSAT342A variant, which is associated with a higher level of expression of SSAT in BA4 and BA11, is found significantly less frequently among suicide completers compared with controls, suggesting that this locus may play a role in suicide predisposition through a regulatory influence on SSAT expression in the brain.
Using microarray brain expression analysis as a screening tool in a group of suicide completers with and without major depression and a group of controls, we have identified SSAT as a gene that is differentially expressed in BA4, BA8/9, and BA11. Differential expression of SSAT was confirmed by semiquantitative RT-PCR, immunohistochemistry analysis, and Western blot findings. Analysis of the genetic variation at the SSAT342A/C locus in the vicinity of the polyamine-responsive element located in the promoter of the SSAT gene demonstrated an effect of the genotype on gene expression, with the A allele associated with higher levels of SSAT expression. Conversely, the evaluation of the SSAT342 polymorphism in an independent sample of 183 suicide completers and 80 controls showed a lower frequency of the A allele among the suicide completers, suggesting a protective role against suicide and depression. The main conclusions of our study concerning SSAT are based on the consistency of the significance in different brain regions (BA4, BA8/9, and BA11), the validation of these differences using alternative (RT-PCR) and complementary methods (immunohistochemistry and Western blot), and the observation that variation at the promoter region influences levels of expression and may play a role in predisposition to suicide and depression.
Spermine/spermidine N1-acetyltransferase is the rate-limiting enzyme in the catabolism of polyamines24 (spermidine and spermine) and is implicated in the polyamine stress response.25,26 Polyamines, especially spermine, are stored in synaptic vesicles and released by depolarizationlike neurotransmitters.27 Evidence suggests the implication of polyamines in mood disorders, such as the observation that lithium prevents the stress-induced polyamine response in rats.25,28,29 In addition, spermidine and spermine block the serotonin transporter transient current in a manner similar to fluoxetine hydrochloride and cocaine.30 Finally, glutamatergic neurotransmission is closely controlled by intracellular levels of polyamines, with spermine and spermidine being specific modulators of N-methyl-D-aspartate and AMPA (α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid) receptor activity.31 Consequently, significant down-regulation of SSAT would be expected to disrupt polyamine homeostasis, resulting in regional increases in levels of spermine, spermidine, or both. Considering the multiple processes in which polyamines of the central nervous system have been implicated, changes in polyamine levels may produce profound effects.
Some of the most important limitations of this study include a possible confounding effect of comorbidity with substance dependence/abuse and limited power of the microarray expression study. In our sample, as expected, several subjects had a history of substance dependence/abuse (alcohol and cocaine). This was the case in both suicide groups, and to a lesser degree, in the control group (alcohol). However, after controlling for the presence of these factors and other relevant potential confounders, the group effect on SSAT expression remained significant in BA4 and BA11, suggesting that the results observed in the gene expression screening, at least with SSAT, are not a consequence of these other factors.
Although the sample used in the microarray study is of limited power, and multiple testing may lead to a high rate of false discoveries, the following should be taken into account when interpreting our major findings. (1) We used a number of procedures to avoid false-positive results. For instance, we filtered out genes not present in at least 75% of the subjects per group. This procedure reduces significantly the number of comparisons by decreasing probe sets being tested (approximately 15 000 probe sets instead of the approximately 44 000). In addition, the criteria we used to determine whether a gene was differentially expressed combined the FC and P value criteria. (2) Our brain expression studies were used only as a first step of a screening procedure to identify potential targets of interest. Several levels of internal consistency were used to select the target of further study. (3) The findings implicating SSAT are based on different levels of observation, including (a) consistency between different gene probes signals, (b) consistency between brain regions, (c) validation using an alternative method (RT-PCR), (d) confirmation at the protein level using 2 complementary methods (immunohistochemistry and Western blot), and (e) genetic evidence suggesting that variation at the promoter region may influence levels of expression.
In a recent study, Sibille et al32 performed a microarray analysis comparing expression patterns in BA9 and BA47 of depressed suicide completers vs psychiatrically normal controls who were matched on the basis of sex, age, PMI, and race. They observed no evidence of differences in gene expression that correlated with major depression and suicide. Many differences between the 2 studies could explain the discrepant results obtained. First, in our study, we included male subjects only, and as demonstrated by the same group,33 prefrontal cortex gene expression has a strong sex-related component, probably increasing the gene expression variability if male and female subjects are combined in the study. Second, all subjects included in our study were of French Canadian origin, a population with a well-known and well-characterized founder effect.13 It is possible that by investigating subjects from this young (approximately 12 generations) and isolated population, we reduced the total variability in gene expression patterns in our study. Finally, another significant difference is that our analysis was performed using the Human Genome U133 set, which consists of 2 GeneChip arrays with approximately 45 000 probe sets, whereas the analysis that was performed in the study by Sibille et al32 used only the U133A GeneChip, which contains approximately half the probe sets (22 000 probe sets) of the U133 set.
In this study, we simultaneously screened the expression levels of genes using microarray analysis in postmortem cortical regions from suicide completers with and without major depression vs a group of controls. We identified SSAT as a candidate mediating risk for suicide. This effect appears to be moderated, to a certain extent, by the presence of major depressive disorder. However, our study design does not allow us to completely separate the effect of suicide from the underlying psychopathology. Such resolution could be obtained by an investigation of controls with depression who were not suicide completers. However, collecting such a sample is operationally challenging given the mean age of the suicide completers. Confirmation of our results and further investigation of the role of SSAT and other polyamine-metabolizing enzymes in the neurobiology of suicide and major depressive disorder is warranted.
Correspondence: Gustavo Turecki, MD, PhD, McGill Group for Suicide Studies, Douglas Hospital, McGill University, 6875 LaSalle Blvd, Montreal, Quebec, Canada H4H 1R3 (email@example.com).
Submitted for Publication: October 13, 2004; final revision received February 17, 2005; accepted June 15, 2005.
Author Contributions: Mr Sequeira and Dr Gwadry contributed equally to this study.
Funding/Support: This study was supported in part by grants MOP-38078 and MOP-53321 from the Canadian Institutes of Health Research, Ottawa, Ontario.
Acknowledgment: We thank the Bureau du Coroner du Québec, Montreal, for their support; and W. H. Zheng, PhD, and Amanda Li for their technical support.