The existing literature on gene–environment interactions for 5-HTTLPR genotype and life stress has produced conflicting results, possibly due to methodological heterogeneity across studies with respect to subject recruitment and life stress assessment (Uher and McGuffin, 2008
; Uher and McGuffin, 2010
). We have now conducted the first study to assess such a gene–environment interaction across multiple age cohorts, using a consistent and highly sensitive life stress assessment methodology, and a quantitatively measured biologically based endophenotype. On the basis of three age cohorts of children, younger adults, and older adults, and with each cohort similarly powered to detect a gene–environment interaction, we now report evidence for a significant interaction in the cohort of younger adults only.
Our analyses suggest that, indeed, subject age and SLE assessment both influence whether a significant interaction between 5-HTTLPR genotype and SLE will be observed. When the measure of SLE was the total number of such events (SLE-T), accumulated over a lifetime, none of the three age cohorts exhibited a significant gene–environment interaction. Similarly, when the measure of SLE was the number of such events during the first 15 years of life (SLE-15), neither younger nor older adults exhibited a significant gene–environment interaction (no SLE-15 data were available for the cohort of children studied here). However, when the focus was on SLE that had occurred during the first 5 years of life (SLE-5), there was a significant interaction between 5-HTTLPR genotype and SLE, but only in the younger adults, not in children (no SLE-5 data were available for older adults). Thus, this study is consistent with the hypothesis (Lewis et al, 2010
; Uher and McGuffin, 2008
) that age but also SLE assessment may be critical variables in gene–environment interactions involving 5-HTTLPR genotype (Uher and McGuffin, 2008
; Uher and McGuffin, 2010
). Future work will also need to address why significant G × E interactions were absent in adolescents but present in younger adults. We speculate that the emergence of significant G × E interactions over this period may reflect developmental processes in brain maturation that may be epigenetically regulated (Bennett-Baker et al, 2003
; Munoz-Najar and Sedivy, 2010
Among younger adults, the interaction between 5-HTTLPR genotype and life stress was significant only when the analysis focused on SLEs that had occurred in the first 5 years of life: individuals homozygous for the LA
allele exhibited a negative correlation between the number of SLEs during the first 5 years of life and peak cortisol response to social stress, whereas carriers of the S allele exhibited a positive correlation. Strikingly, a very similar pattern was reported by Canli et al (2006)
, who in an imaging study observed a negative correlation between SLEs and amygdala activation in homozygous L allele carriers and a positive correlation in S allele carriers (Canli et al, 2006
), although these analyses did not separate early SLEs from other SLEs.
The interesting ‘crossover' found in the cortisol response when analyzing is similar G × E and seems to be similar to what has been observed for depression and suicide studies showing that the S allele may be associated with enhanced cortisol release after early exposure to stress (Caspi et al, 2003
; Roy et al, 2007
These data suggest that the short variant of the 5-HTTLPR does not constitute a ‘vulnerability' allele, but rather that the endophenotype associated with the short allele may be optimal or not, depending on the environment: if one operationalizes optimal socioemotional functioning in terms of low cortisol reactivity to a social stressor, then individuals in the S group outperform individuals in the L group, if both groups had experienced low numbers of early SLEs. Indeed, Taylor et al (2006)
made a similar observation with respect to homozygous S allele carriers, noting that in benign environments, their genotype assumes properties that protect from depression, relative to L allele carriers (Taylor et al, 2006
). Contradictory, recent results showed that newborns with two S alleles have been found to show the highest response after a physical stressor (Mueller et al, 2010
The data from this study, as in Canli et al (2006)
, also show that responsiveness to environmental influences is not a special attribute of the short variant of the 5-HTTLPR, as some have argued in favor of the short allele as a ‘plasticity gene' (Belsky et al, 2009
; Belsky and Pluess, 2009
) providing a genetic basis to improved performance in an array of cognitive tasks and increased social conformity (Homberg and Lesch, 2010
). Instead, both studies have shown that homozygous long allele carriers exhibit markedly different behavioral and neural responses to social stressors, visual stimuli, and in the scanner at rest, depending on their early life stress history. This may explain why some studies associated the L allele with less favorable phenotypes, such as increased cardiovascular reactivity, greater risk of myocardial infarction (Coto et al, 2003
; Fumeron et al, 2002
; Williams et al, 2001
), increased risk of psychosis (Goldberg et al, 2009
), and increased risk of chronic PTSD (Grabe et al, 2009
; Lee et al, 2005
; Thakur et al, 2009
). In light of these results, it is possible that, in the absence of early SLEs, homozygous L allele carriers are susceptible to multiple physical and psychological adverse outcomes.
The interaction of 5-HTTLPR genotype and SLEs may explain, at least in part, why cortisol studies that have not taken life stress history into account have produced inconsistent findings. For example, Jabbi et al (2007)
and Gotlib et al (2008)
found larger cortisol responses to a stressor in homozygous carriers of the S allele (Gotlib et al, 2008
; Jabbi et al, 2007
). We recently reported that newborns with two S alleles exhibit the highest cortisol response after a physical stressor (Mueller et al, 2010
). In contrast, Alexander et al (2009)
and Wüst et al (2009)
did not find significant differences in cortisol reactivity between 5-HTTLPR groups after a stressful task. Contrary to these previous studies, we now report that homozygous carriers of the L allele exhibited the largest cortisol response to a stressor. However, this vigorous reactivity has to be viewed in the context of early SLEs: it was only present in those individuals who had not experienced any early SLEs (see ). Indeed, the cortisol response of the L group with no early adversity was almost thrice as high
as the cortisol response of the S group. However, in individuals who had experienced three or more early SLEs, the pattern was reversed: the cortisol response of the L group with early SLEs was almost one-third below
the cortisol response of the S group. Thus, SLEs likely represent a major confounding variable that needs to be included in future analyses of cortisol reactivity as a function of 5-HTTLPR genotype. Indeed, the fact that among younger adults in our study the L group exhibited greater cortisol reactivity than the S group can therefore be largely attributed to the fact that the majority of L group members (N
=12 (out of 28)) had not experienced any early SLEs.
Only one previous study had measured cortisol reactivity as a function of SLEs: Alexander et al (2009)
reported higher cortisol levels in males homozygous for the S allele with a history of SLEs (Alexander et al, 2009
). Our sample was too small to formally conduct the same analysis, but inspection of the subsample of men in the younger sample revealed higher cortisol reactivity in the L group, in contrast to Alexander et al's (2009)
findings. These different outcomes might at least be partly due to the use of different stressors across the two studies. Future work with much larger samples will need to address these and other factors that may have driven these conflicting gender effects.
Our study is limited by its reliance on self-report: although the LHC and LEQ have been shown to reliably elicit memories, biases cannot be entirely ruled out. Furthermore, our participants may have differed in their ability to remember events accurately despite the use of memory cues and the calendar format of the LHC method. Also, because the number of SLE during the first 5 years of life was not available for older adults because of the nature of the assessment method, analyses between early life stress and genotype could not been carried out for older adults. Another limitation is that recruiting strategies differed between the three groups: children were recruited at schools in Dresden, and younger adults were recruited via flyers from campus, but older adults were recruited via flyers in a fitness center for older people and an advertisement in a local newspaper. Thus, our sample of older adults may represent a cohort of healthier and more active individuals, compared with the other cohorts tested in this study. Finally, our current sample is too small to conduct analyses in regard to possible quadruple interactions between 5-HTTLPR genotype, SLEs, age, and gender. We plan to continue increasing our sample to be able to conduct these analyses in the future.
In sum, our findings support the notion that the effects of functional genetic variations are further modulated by environmental factors, particularly by those occurring during the early 5 years of life. Instead of a simple additive effect, by which early life stress may exacerbate a genetically predisposed individual to respond to social stressors in later life, we observed an interaction of 5-HTTLPR genotype and early life SLEs. On the basis of this interaction, characterizing an individual's stress response on genotype alone leads to potentially erroneous conclusions. For example, in our younger adult data set, the cortisol response of homozygous L allele carriers may be substantially greater or smaller than that of S allele carriers, depending on the number of SLEs they experienced. Fascinatingly, a high number of early SLEs in homozygous L carriers are associated with reduced cortisol stress reactivity, suggesting positive adaptation in the face of early environmental stressors. This is the first study, to our knowledge, to investigate the influence of 5-HTTLPR and SLEs in a cross-sectional design. We find that 5-HTTLPR is associated with individual differences in stress reactivity that may endow individuals with variable stress reactivity, with early SLEs modulating the direction of this response.
All results for tri-allelic genotypes have been discussed as well (see Supplementary Information).