The Florida Hurricanes Study
Three of the six published GxE investigations have come from our research team using data from the 2004 Florida Hurricanes Study (Amstadter et al., 2009
; Kilpatrick et al., 2007
; Koenen et al., in press
), an NIMH funded project designed to assess hurricane related outcomes in a representative sample of older adults (n
= 1130, age 60 and over) and a comparison sample of younger adults (n
= 413, age 21–59) residing in Florida disaster sites. This household probability sample of adults was interviewed 6–9 months after the hurricanes about demographic variables, exposure to the hurricane(s), PTSD status, prior exposure to potentially traumatic events (PTEs), and social support (Acierno, Ruggiero, Kilpatrick, Resnick, & Galea, 2006
). The study design included collecting saliva samples via mail from adults who had been interviewed about their exposure to the 2004 hurricanes, and over 600 participants returned saliva samples. Comparisons of participants who did and did not return saliva samples found that there were no significant differences on any major variables (Galea, Acierno, Ruggiero, Resnick, & Kilpatrick, 2006
), underscoring the viability of biologic data collection via US mail for the purposes of genetic analyses. The saliva collection kits used in the Florida Hurricanes Study obtained approximately 10–30 µg of DNA from each sample, with a failure rate of about 3% (no higher than found in studies in which samples are not mailed to the laboratory prior to DNA extraction), which is an adequate yeild to support a plethora of genetic assays.
The first paper published from this project, and notably, the first published GxE PTSD study, tested the hypothesis that the polymorphism in the serotonin transporter gene (also known as 5-HTTLPR or SLC6A4
) would modify risk of post-hurricane PTSD under the high environmental risk conditions (Kilpatrick et al., 2007
). This commonly studied polymorphism is a variable number tandem repeat (VNTR) located in the promoter region of the gene. The 5-HTTLPR is a functional polymorphism, with the short, “s,” allele being less transcriptionally efficient than the long, “l,” allele (Lesch et al., 1996
). This polymorphism has since been found to be triallelic, in that a third functional allele, ‘LG
’, has been identified (Nakamura, Ueno, Sano, & Tanabe, 2000
is characterized by an A>G substitution at nucleotide 6 of the first of two extra 22-bp repeats in the ‘l’ allele, resulting in transcriptional capacity comparable to the ‘s’ allele. In this study, the LG
and s alleles were coded as s’ alleles; LA
were coded as l’ alleles. Environmental stress conditions were dichotomized into high versus low hurricane exposure, and high versus low social support. As illustrated by , the low expression (s) variant of the 5-HTTLPR increased risk of post-hurricane PTSD, but only under the high stress conditions of high hurricane exposure and low social support. At Highest risk (n
= 27) were those with the s’/s’ genotype and high stress exposure. Medium risk (n
= 54) were those with the l/s genotype and high stress exposure. All others (n
=498) all others were considered low risk. There was a strong association between risk group and prevalence of PTSD, χ2
= 579) = 19.94, p
<.001. High-risk individuals (high hurricane exposure, the low expression 5-HTTLPR variant, low social support) were at 4.5 times (95% CI = 1.2, 17.9) the risk of developing PTSD as compared to low-risk individuals.
Level of stress exposure modifies the association between the 5-HTTLPR genotype and post-hurricane PTSD.
A second paper was published by our team that examined the 5-HTTLPR gene and PTSD (Koenen et al., 2009). Whereas previous GxE investigations of psychiatric phenotypes have only examined individual-level environmental variables (e.g., (Caspi et al., 2003
; Kilpatrick et al., 2007
), this study investigated the role of group-level social environment. Specifically, crime rate (taken from the Serious Crimes Known to Police U.S. Federal Bureau of Investigation, Uniform Crime Reporting Program unpublished data from 1999; high crime counties were defined as counties in which the crime rate was above the mean crime rate of the counties in our study) and percent unemployment (taken from the 2000 U.S. Census Summary File 3; high unemployment was defined as counties in which the unemployment rate was above the mean of the studied counties) were the examined environmental variables. Of particular note, county-level crime rate data provide an example of an environmental variable that may be objectively observed. Results indicated that county-level crime rate and percent unemployment modified the association between the 5-HTTLPR genotype and PTSD; low expression allele carries (‘s’) were at increased risk for PTSD in high environmental stress conditions (high unemployment; high crime), and were at decreased risk for PTSD in low-risk environments (low unemployment, low crime). This paper underscores the importance of investigation of not just individual level variables, but also group level factors of the overall social environment.
The third PTSD GxE paper using Florida Hurricane Study data examined a polymorphism that had not previously been studied in the context of PTSD (Amstadter et al., 2009
). Polymorphisms in the RGS2
(regulator of G-protein signaling 2) gene were found to be associated with anxious behavior in mice (Leygraf et al., 2006
) and anxiety in humans (Smoller et al., 2008
). We examined whether rs4606, a single nucleotide polymorphism (SNP) of RGS2
, and social support moderated risk for post-hurricane and lifetime PTSD. Rs4606 (‘C’ allele was the risk allele) was associated with increased symptoms of post-hurricane PTSD symptoms under conditions of high hurricane exposure and low social support (p
< .05). Further, this polymorphism was associated with lifetime PTSD symptoms under conditions of lifetime exposure to a PTE (other than the current hurricane), and low social support (p<0.001). These GxE interactions remained significant after adjustment for sex, ancestry, and age, indicating that RGS2
rs4606 modifies risk of post-disaster and lifetime PTSD under conditions of high stressor exposure.
GxE Under Conditions of High Trauma Exposure
The 5-HTTLPR polymorphism of SLC6A4 was recently examined in 408 Rwandan refugees (Kolassa et al., in press
). Participants endorsed extremely high levels of trauma exposure, reporting an average of 12.6 different traumas; not surprising, given the rates of traumatic experiences, 81% of participants met full criteria for lifetime PTSD and the probability of PTSD was nonlinearly associated with number of traumas, asymptotically approaching a probability of 1 for participants reporting 15 or more traumatic events. Findings supported a main effect for genotype, with the resiliency associated with the presence of an l allele diminishing across exposure to increasing numbers of traumatic events. More specifically, whereas the fitted probability of lifetime PTSD for participants homozygous for the low expression allele (s’/s’) was 100% regardless of the number of traumatic experiences, carriers of the l’ alleles evidenced lower levels of PTSD following exposure to few trauma experiences but showed significantly increased risk associated with increasing numbers of traumatic experiences. The extremely high levels of trauma exposure, and concomitant high rates of PTSD, found in this sample demonstrate the overwhelming effects of trauma under conditions of extreme exposure and highlight the importance of careful characterization of trauma exposure in GxE research.
Child Abuse as a GxE Moderator
A recent paper by Ressler’s research team at Emory University also tested GxE models for PTSD (Binder et al., 2008
). The sample (n
= 762) consisted predominately of low income African American patients who presented to general medical clinics at a large urban hospital. Current PTSD was assessed, as well as a retrospective assessment of exposure to child abuse. Fine mapping of the FKBP5
locus was undertaken, as FKBP5
is a glucocorticoid regulating gene, and has been previously found to be related to peritraumatic dissociation (a potent predictor of PTSD) (Koenen et al., 2005
), it was a fitting candidate gene. No main effects of polymorphisms in FKBP5
were identified, but a significant GxE interaction between severity of child abuse (but not adult traumatic events) and FKBP5
polymorphisms was found for PTSD. illustrates the interaction findings for one of the FKBP5
polymorphisms that remained significant after correction for multiple testing. The results show that the ‘C’ allele was associated with higher levels of PTSD symptoms, but only among individuals exposed to two or more types of child abuse. There was no effect of genotype for individuals who experienced only one type of child abuse or no child abuse. Dexamethasone suppression test data also demonstrated that these polymorphisms have functional consequences for glucocorticoid system response sensitivity. They concluded that variation in FKBP5
, in the presence of child abuse, may alter sensitivity of this stress-response pathway, “placing those individuals who have had significant child abuse at significant risk for PTSD in the face of other traumatic experiences” (p. 1304).
Severity of child abuse modifies the association between FKBP5 polymorphisms and adult PTSD symptoms.
Finally, Nelson et al (2009)
examined the interaction of gamma-aminobutyric acid receptor, alpha-2 (GABRA2
) polymorphisms and childhood trauma in risk of lifetime PTSD diagnosis in adults (n=259). The authors found significant (p<.05) interactions between three SNP and childhood abuse; only homozygous individuals were at significantly increased risk of PTSD. Findings remained similar after controlling for nicotine and alcohol dependence. No evidence for significant GxE interactions were found for major depression.
Novel Environmental Variables
An important direction for PTSD GxE research to advance in is not only the expansion of studied ‘G’ variables, but also the broadening of ‘E’ variables that are investigated. Environmental variables previously found to moderate the relationship between genetic variants and PTSD include individual-level variables (e.g., social support, level of trauma exposure), and group-level environmental variables (i.e., county level crime rate, county level unemployment rate). Although the studied ‘E’ variables have been promising, they likely only represent a small fraction of the possible ‘E’ characteristics that may play a role in the etiology and maintenance of PTSD, leaving a wide-open field of research in need of being conducted. Expansion of both individual-level and group-level environmental variables is warranted.
Extant psychosocial studies of PTSD may help to inform the selection of ‘E’ variables for inclusion in GxE studies. A number of psychosocial risk factors for PTSD have been identified (Brewin, Andrews, & Valentine, 2000
) that may be fitting ‘E’ variables for use in future research (e.g., education, general childhood adversity, current life stressors, female gender). Further, aspects of an individuals’ trauma history (e.g., PTE severity, duration, interpersonal victimization, age of onset, multiple victimization history) have been related to increased risk for the disorder (Brewin et al., 2000
), and could also be studied in the context of a GxE study of PTSD. Additionally, findings from the non-human primate literature may help inform novel ‘E’ variable selection, particularly for pediatric PTSD, as early rearing environment has also been shown to moderate the relationship between serotonin transporter variation and central nervous system function (Bennett et al., 2002
) as well as hypothalamic adrenal axis response to stress (Barr et al., 2004
). PTSD GxE studies may also benefit from findings in other areas of psychiatric genetic study. For example, a recent GxE study of depression found that parenting style, specifically perceived maternal rejection, moderated the relationship between genetic variants and depression in juvenile detainees (Haeffel et al., 2008
), suggesting that family environment variables (e.g., family cohesion, communication, conflict) may be fitting ‘E’ variables in the study of child and adolescent PTSD.