Background
Suicide represents a worldwide public health problem and is a leading cause of death among those aged 15-44 years in many developed countries, representing a significant social and economic burden
http://www.who.int/mental_health/prevention/suicide. According to the stress-diathesis model of suicidal behaviour, risk of suicide is not exclusively determined by a psychiatric illness (or an alternative stressor), but rather that individuals may have a predisposition or tendency towards suicidal ideation, which is aggravated further by one or more stressors [
1]. A greater understanding of clinical and biological risk factors will help us understand more fully this model of suicidal behaviour and inform new therapies and tailored interventions.
The molecular pathology of suicidal behaviour remains complex; epidemiological studies have demonstrated familial clustering of suicide and suicidal behaviour [
2]. Meta-analysis of published twin case studies for suicide found a higher rate of concordance for suicide and for suicidal behaviour in monozygotic (MZ) versus dizygotic (DZ) twins, establishing a solid genetic link [
3]. Currently, risk assessment of suicidal behaviour relies heavily on clinical variables [
4]. Indeed, patients with psychiatric disorders are at higher risk for suicide attempt (SA) [
5‐
8], however protective and risk genetic variants for suicide appear to be independent of underlying psychiatric disorders [
9‐
11]. Identifying relevant genetic variants or single nucleotide polymorphisms (SNPs) in genes intimately linked to various neurobiological pathways, whose alteration may contribute to suicidal behaviour, may provide insight into the etiological processes that underlie suicidal behaviour and identify novel biological therapeutic targets.
Numerous neurobiological pathways have been implicated in the etiology of suicide and suicidal behaviour. Deregulation of the monoaminergic transmitter systems (norepinephrine, dopamine and serotonin) [
12,
13], gamma aminobutyric acid (GABA) inhibitory neurotransmitter system [
14,
15], glutamatergic excitatory neurotransmitter system [
14,
15], lipid metabolism [
16,
17], cannabinoid system [
18] and neurotrophic signalling factors [
19] are linked to suicidal behaviour. Interestingly, the hypothalamic-pituitary-adrenal (HPA) axis may be involved in increased suicidal behaviour in response to adversity during development [
20], hence highlighting the contribution of early environmental factors to suicide risk.
Suicidal behaviour is a complex trait; therefore it is likely to be the result of a combination of numerous environmental and genetic factors with an added layer of complexity via potential gene-gene (GxG) and gene-environment (GxE) interaction effects [
21]. Consequently, investigating the relationship between genetic variants and environmental factors (e.g. childhood abuse or alcoholism) is an important step towards elucidating the mechanisms behind suicidal behaviour.
Here we test the hypothesis that genetic variants in candidate genes important for neurobiological pathways linked to suicidal behaviour and/or associated endophenotypes, may confer a protective [
22] or risk effect on SA among patients with co-existing psychiatric illness (suicide attempters versus non-attempter controls). In addition to main genetic effects, we also tested selected GxG and GxE interactions for association with SA.
Methods
Clinical sample collection
The study consisted of 159 psychiatric patients recruited as part of the Ireland North/South Urban/Rural Epidemiologic (INSURE) collaborative project [
23]. Briefly, consecutive newly referred patients to six Community Psychiatric Clinics on the island of Ireland (Donegal, Belfast, Omagh, Dublin, Portlaoise and Ballinasloe) were invited to take part in a clinical and molecular genetics study of suicidal behaviour in major psychiatric disorders over a year long recruitment period. In total, 159 patients gave written informed consent for the study. Patients were interviewed for Axis I Psychiatric Disorders using the Structured Clinical Interview for DSM-IV [
24]. History of attempted suicide was recorded using the Columbia University Suicide History Questionnaire [
25], which incorporates the Scale for Suicide Ideation [
26], and the Suicide Intent Scale [
27]. A suicide attempt was defined as a completed act of self-harm with at least some expressed intent to die (scoring at least 1 on item 10 of Suicide Intent Scale [
27]). Seventy-six (48% Male, 52% Female; Mean age: 34.7 yrs) had a history of SA and 83 (61.4% Male, 38.6% Female; Mean age: 40.7 yrs) had not. Childhood abuse was defined as the presence or absence of a history of sexual and/or physical abuse under the age of 16. Further clinical variables from clinical and demographic assessments are outlined in Table
1. All clinical research interviews were performed by a trained research psychiatrist or research psychiatric nurse. Ethics approval was obtained for the study from the St Vincent's Healthcare Group Ethics and Medical Research Committee.
Table 1
DSM-IV psychiatric diagnosis and clinical features of patient samples
Sex
| | | 0.124 |
Male | 36(48) | 51 (61.4) | |
Female | 39 (52) | 32 (38.6) | |
Mean age (years)
| 34.7 | 40.7 | 0.017 |
Positive 1st degree family history of suicide attempt
| 26 (35.6) | 14 (17.1) | 0.014 |
Child abuse (Physical and/or sexual)
| 37 (49.3) | 27 (32.5) | 0.047 |
Axis I Diagnosis
| 42 (55.3) | 36 (43.4) | 0.18 |
Substance-related disorders
| 24 (31.6) | 21 (25.3) | 0.334 |
Substance abuse/dependence | 38 (50) | 29 (34.9) | 0.078 |
Alcohol abuse/dependence | 6 (8.8) | 7 (8.6) | 1 |
Schizophrenia and other psychotic disorders
| | | |
Bipolar disorders
| 5 (7.4) | 4 (4.8) | 0.786 |
Major depressive disorder
| 48 (63.2) | 37 (44.6) | 0.022 |
Anxiety disorders
| 32 (42.7) | 25 (30.1) | 0.14 |
Eating disorders
| 1 (1.3) | 0 (0) | 0.965 |
Gene and SNP selection
In this study, genes intimately linked to numerous neurobiological pathways considered important for the pathogenesis of suicide and/or associated disorders or endophenotypes, were selected. Genes from the following pathways were considered, monoaminergic transmitter systems (norepinephrine, dopamine and serotonin), GABA inhibitory neurotransmitter system, glutamatergic excitatory neurotransmitter system, HPA axis and cannabinoid system. Genes important for cholesterol transport and neural growth and differentiation were also selected.
Twenty-eight SNPs from the 18 selected genes (
COMT, 5-HT2A, 5-HT1A, 5-HTR1B, TPH1, MAO-A, TPH2, DBH, CNR1, BDNF, ABCG1, GABRA5, GABRG2, GABRB2, SLC1A2, SLC1A3, NTRK2, CRHR1) were included in our analysis. SNPs, within these 18 genes, were screened for known genetic associations with suicide, suicidal behaviour and/or associated endophenotypes using biomedical online resources PubMed
http://www.ncbi.nlm.nih.gov/pubmed and OMIM
http://www.ncbi.nlm.nih.gov/omim. Priority was given to functional SNPs or SNPs within regulatory regions utilising online genetic databases ensembl
http://www.ensembl.org/index.html and DbSNP
http://www.ncbi.nlm.nih.gov/projects/SNP/.
Genotyping
DNA was extracted from patient blood samples using standard techniques. Flanking SNP sequences were obtained from DbSNP and all genotyping assays were designed and validated by KBioscience (Hertfordshire, UK). Genotyping for SNP analysis was performed by KBioscience using the competitive allele specific PCR (KASP) chemistry coupled with a FRET-based genotyping system
http://www.kbioscience.co.uk/reagents/KASP/KASP.html. Genotyping data was viewed using SNPviewer (KBioscience, Hertfordshire, UK), allowing graphical viewing of the clusters that group the allele calls.
Data analysis
Statistical tests of association examining the relationship between patient's alleles or genotype, at each of the 28 genetic loci, and suicide attempt were performed using Pearson's Chi-square test or Fisher's exact test (when counts are low). This analysis was performed both in SPSS and SNP and Variation Suite for verification purposes.
SNPs which showed a modest level of association in the original allelic and genotype tests of association (P < 0.1) and one SNP which showed a gender specific association with SA, were further analysed by applying four genotype models (Additive (XX = 0, XY = 1, YY = 2), Dominant (XX = 0, XY & YY = 1), Over-dominant (XX & YY = 0, XY = 1) and Recessive (XX & XY = 0, YY = 1); X = Major allele, Y = Minor allele). For all models, except the additive model, Pearson's Chi-square test of association was used. For the additive model an exact Armitage trend test was preformed. Binary logistical regression analyses were performed to evaluate the contribution of the genotype model associated with SA at each locus in the prediction of attempted suicide, correcting for potential confounders such as age, gender and certain psychiatric disorders and to interrogate associated SNPs for potential GxG and GxE interactions. Analysis of potential GxE interactions were applied to SNPs with evidence of association in at least one genetic model (n = 3) and to one SNP which showed evidence of association in a gender specific manner (n = 1) and not examined for all candidate SNPs listed in this study. In addition, GXG interactions analysis was performed on these 4 SNPs, which showed evidence of association with SA and not examined for all candidate SNP listed in this study.
A non-parametric Mann-Whitney test was used to calculate the difference in age between cases and controls. For all other comparisons of clinical and demographic variables a Fisher's exact test or Chi-squared test of association was used. All statistical analysis was performed on SPSS (PASW statistics 18, Chicago, Illinois, USA) and genotypic associations verified using SNP and Variation Suite (SVS) 7 software. For all tests, significance was ascribed at P < 0.05.
Discussion
We have identified 4 SNPs (rs4755404, rs2269272, rs6296 and rs1659400) in genes SLC1A2, SLC1A3, 5-HTR1B and NTRK2, respectively, which showed evidence of association with SA compared to a non-attempter psychiatric control group. At present, there is no published data on 3/4 of the genetic variants (rs4755404, rs2269272 and rs1659400) reported here and their association with suicide or SA. In addition, we identified a 3-locus (rs6296 (C/G) × rs4755404 (C/G or G/G) × rs1659400 (C/C)) G×G interaction, which was a significant predictor of SA, suggesting that variation in SLC1A2, 5-HTR1B and NTRK2 genes may contribute to the risk of SA independently, and in an interactive manner. This paper provides evidence for the first time of a putative G×E interaction, where genetic variation at the rs1659400 locus may moderate the risk associated with history of childhood abuse and subsequent suicidal acts.
SLC1A2 and
SLC1A3 are glial high-affinity transporter molecules that regulate glutamate concentrations at synapses [
15]. Aberrant expression of these key transporters could impair reuptake of glutamate, hence prolong synaptic activation and potentially lead to cytotoxic damage to neurons and glia [
15].
SLC1A2 and
SLC1A3 are downregulated in the brains of MDD patients compared to controls [
28]. Moreover, decreased
SLC1A3 expression has been observed in suicide brains [
29].
In this study, the
SLC1A2 gene variant, rs4755404, was associated with SA and schizophrenia and other psychotic disorders. Consistent with these findings, rs4755404 was previously associated with schizophrenia in a Japanese patient cohort [
30]. In contrast, the
SLC1A3 gene variant, rs2269272, was significantly associated with male non-attempters, where a gene-sex interaction was observed. The T/T genotype is absent from the attempter group and over-represented in the male non-attempter group, suggesting a T/T genotype may confer a protective effect on risk of SA, particularly in males with underlying psychiatric disorders. HWE analysis revealed that the control group was not in equilibrium for the rs2269272 locus, which is likely due to the fact that our control group is not "disease-free" but rather consists of individuals with underlying psychiatric conditions. However, the possibility that some other factor may be influencing the over-representation of T allele homozygotes in the non-attempter group cannot be ruled out. Therefore, further investigation in a larger cohort of patients is warranted. Taken together, our findings support the glial hypothesis of mood abnormalities [
28] and concur with the literature on a putative role of glutamate dysregulation in suicidal behaviour.
The serotonin (5-HT) system is the most widely studied area of neurobiological suicide research [
21]. Aberrant 5-HT receptor binding has been implicated in suicide and SA [
12].
5-HTR1B binding is decreased in the prefrontal cortex of suicide brains [
31]. Previously a
5-HTR1B promoter CpG island genetic variant, rs6296, was associated with SA [
32]. However, findings are inconsistent and a number of studies have found no evidence of association [
33,
34]. Rs6296 has also been associated with substance use disorders, major depression and inconsistently with alcohol abuse [
33,
35]. Here we report a significant association between the rs6296 locus and SA. In addition, persons with a C/G genotype and a history of abuse or dependence on alcohol were significantly associated with SA, but a gene × alcohol interaction was not evident, suggesting that they represent additive risk factors. These findings provide the impetus for future studies in a cohort of alcohol abuse/dependence patients to further understand the relevance of rs6296 genetic variants and SA in patients with a history of alcohol abuse/dependence.
NTRK2 (
TRKB) encodes the receptor for BDNF. Aberrant neurotrophic signalling has been implicated in suicide risk by various studies [
36,
37].
BDNF and
NTRK2 mRNA and protein expression are reported downregulated in the prefrontal cortex and hippocampus of suicide victims compared to controls [
36,
37]. The majority of studies have focused on functional
BDNF variant, rs6265, which has been inconsistently implicated in suicide and MDD [
38,
39]. In this study, we found no evidence of association between rs6265 and SA, consistent with a previous report [
40]. Previously, a number of genetic variants within the
NTRK2 gene have been associated with SA among depressed patients [
41]. We observed a significant association with
NTRK2 intronic genetic variant, rs1659404, and SA in females. Rs1659400 is in strong LD with several other SNPs within the
NTRK1 gene, some of which have been implicated in alcohol dependence and depression [
42,
43].
To date a number of gene × childhood adversity interactions have been reported in psychiatric patients [
44‐
46]. Childhood trauma (including abuse) has been reported to interact with low expressing
5-HTTLPR genotypes and moderate the risk of suicidal behaviour [
47]. Recently, an interaction between child maltreatment and
5-HTT polymorphisms and suicidal ideation among children was described [
48]. Here we report, for the first time, a possible moderation effect of a
NTRK2 polymorphism on childhood abuse and risk for future suicidal acts. It is important to note that a history of childhood abuse was assessed in this study by a trained clinician as opposed to self-report questionnaires. Recent research has demonstrated that clinician ratings of developmental histories, such as childhood sexual and physical abuse, are in agreement with patient's self-reports and thus supports the validity of clinician reports for numerous clinical variables, including childhood abuse assessment [
49]. Future studies could investigate the putative interaction of this genetic variant with childhood trauma score, which would include sexual, physical and emotional abuse and neglect. Such a study would provide a more comprehensive assessment of gene × childhood adversity interaction and risk for SA at this locus.
A number of limitations are apparent in the current study. Firstly, case/control samples were not age-matched, with the SA group having a significantly younger mean age. The SA group also contained a greater number of individuals with a family history of SA and a history of childhood abuse. In addition, rates of MDD diagnoses were different in SA and non-attempter psychiatric controls. However, the genetic variants identified in this study were not associated with either MDD, a family history of SA or abuse. Moreover, our sample size may have reduced power to detect small genetic effects, or alternative interactions between genetic and environmental risk factors. In this study, multiple testing correction, such as Bonferroni correction, was not applied. Bonferroni correction can be regarded as ultraconservative [
50] and ignores the functional candidate gene study design utilised, which is likely to increase the prior probability of detecting an association. In addition, given our modest sample size, Bonferroni correction is likely to result in large type II error rates. Arguably the best solution to type I error is replication [
50].
Competing interests
The authors declare that they have no competing interests.