Peter McGuffin (pictured) is Director, Shaza Alsabban is a PhD student and Rudolf Uher is a clinical lecturer at the MRC Social Genetic and Developmental Psychiatry (SGDP) Centre at the Institute of Psychiatry, King's College London.
The idea that individuals differ in their response to stress is commonplace and every clinician who has ever prescribed antidepressants knows that there is considerable variation in the extent to which patients with depression show a therapeutic improvement or develop unwanted effects. Similarly, the idea that genes might contribute to these types of individual differences is hardly contentious. It is somewhat surprising therefore that attempts to identify such sources of variation at a molecular level have provoked a good deal of debate and controversy, much of which, as exemplified by a paper in this issue of the Journal by Lewis et al, Reference Lewis, Mulligan, Wiles, Cowen, Craddock and Ikeda1 has been focused on the gene that encodes the serotonin transporter.
A polymorphism (common variant) in the promoter region of the serotonin transporter gene affecting expression was first reported by Lesch and colleagues in 1994. Reference Lesch, Balling, Gross, Strauss, Wolozin and Murphy2 This is a so-called length polymorphism, 5-HTTLPR, that exists in two forms, a long allele (‘l’) containing a sequence of 43 base pairs and a short allele (‘s’) in which this sequence is deleted. The long form was shown to be associated with higher and the short form with lower expression of the gene product. Early reports suggested association between carrying the short form and depressive disorder, Reference Collier, Stober, Li, Heils, Catalano and Di Bella3 but there were subsequent failures of replication and the authors of the most recent meta-analysis Reference Clarke, Flint, Attwood and Munafo4 conclude that, although there may indeed be a very small effect, the picture is clouded by publication bias. Indeed, much more interest has been concentrated, not on whether 5-HTTLPR has a main effect, but on whether it moderates the effects of adversity in provoking depressive symptoms and whether there is an interaction with response to antidepressants.
Interaction with adversity
In 2003, Caspi and colleagues Reference Caspi, Sugden, Moffitt, Taylor, Craig and Harrington5 published a report in the journal Science that has since become one of the most talked about papers in psychiatry and has been cited over 2000 times. They reported that, in a cohort of young men and women aged 26 from Dunedin, New Zealand, individuals with one or two copies of the short allele of the 5-HTTLPR exhibited more depressive symptoms, diagnosable depression and suicidality following stressful life events than individuals homozygous for the long allele. Subsequently, Eley et al Reference Eley, Sugden, Corsico, Gregory, Sham and McGuffin6 published a partial replication in which there was a significant genotype–environmental risk interaction for 5-HTTLPR in female (but not male) adolescents, with the effect being in the same direction as in the Caspi et al study. Many other studies, some supportive, and some negative, have followed and in 2009 two meta-analyses were published Reference Munafo, Durrant, Lewis and Flint7,Reference Risch, Herrell, Lehner, Liang, Eaves and Hoh8 where both sets of authors concluded that the overall evidence did not in fact favour the existence of an interaction between 5-HTTLPR and stressful life events in depression. However, neither meta-analysis took into account the substantial heterogeneity in methodology of the studies reviewed which, as we had earlier pointed out, Reference Uher and McGuffin9 had systematic effects on the results.
Specifically, negative studies tended to have used brief self-report measures of adversity rather than semi-structured interviews or more objective evidence, such as third-party reports. Negative studies also appeared to be associated with the age of individuals, with studies on older adults or adolescents, particularly boys, tending not to show an interaction effect, consistent with the early study by Eley and colleagues. Reference Eley, Sugden, Corsico, Gregory, Sham and McGuffin6
We subsequently performed an updated systematic review and found 34 published studies on the interaction between 5-HTTLPR and adversity (as opposed to 5 and 14 studies scrutinised in the two ‘negative’ 2009 meta-analyses). Reference Uher and McGuffin10 We found that 17 of these replicated the original Caspi et al Reference Caspi, Sugden, Moffitt, Taylor, Craig and Harrington5 finding of environment interaction – that is, those carrying the short 5-HTTLPR allele have higher rates of depression or depressive symptoms following life events than those who only carry the long allele. We classified eight studies as partial replications (including that of Eley et al Reference Eley, Sugden, Corsico, Gregory, Sham and McGuffin6 ) where there was an interaction only in females or only with a subset of types of adversity and nine were non-replications. We found a statistically significant relationship between the method used to measure environmental adversity, and the outcome of the study, in that all studies that used objective indicators or semi-structured interviews replicated the environment interaction and all non-replications used brief self-reports. There was also a statistically significant preponderance of non-replications or partial replications in adolescent samples. For reasons that were not apparent, we also found that the negative meta-analyses preferentially included studies that used self-report measures of adversity.
There has now been a third meta-analysis which took a more inclusive approach, combining findings at the level of significance tests rather than raw data. Reference Karg, Burmeister, Shedden and Sen11 This was positive overall and was most highly significant for a moderating effect of 5-HTTLPR on depression in relation to childhood maltreatment and medical illness rather than recent life events. This meta-analysis has confirmed a hypothesis previously proposed by Brown & Harris Reference Brown and Harris12 that the moderating effect of 5-HTTLPR is relatively specific to childhood maltreatment. It has also confirmed our previous finding that the negative results of some studies can be explained by the use of brief self-report instruments to assess stress and that the positive genotype–environment is very unlikely to be due to publication bias.
An additional source of variability might be a single-base substitution (rs25531, A/G) in the long form of 5-HTTLPR that appears to have functional significance. Hu et al Reference Hu, Oroszi, Chun, Smith, Goldman and Schuckit13 found that lG behaves equivalent to the low-expressing s allele and therefore studies that include many lG alleles within s/l and l/l genotypes may underestimate the effect of 5-HTTLPR. This was confirmed by Zalsman et al Reference Zalsman, Huang, Oquendo, Burke, Hu and Brent14 who investigated the association of 5-HTTPLR with stressful life events and severity of depression in Caucasian participants, genotyped for lG,lA and s. They found that lower-expressing alleles (lG and s) independently predicted greater depression severity compared with the higher-expressing lA allele, and that lower-expressing alleles increased the impact of life events on severity of depression. The authors found that 10.5% of the l alleles were the lower-expressing lG allele that otherwise would have been treated as the higher-expressing l allele.
Response to antidepressants
Since the serotonin transporter is the presumed site of action of selective serotonin reuptake inhibitor (SSRI) antidepressants, it is unsurprising that polymorphisms in the serotonin transporter gene have been seen as prime candidates for pharmacogenetic studies of antidepressants. A paper by Lewis et al Reference Lewis, Mulligan, Wiles, Cowen, Craddock and Ikeda1 in this issue of the Journal, on the GENPOD study, is the latest such investigation and fails to provide any support for the hypothesis that the 5-HTTLPR genotype is associated with response to citalopram or indeed to the selective noradrenergic antidepressant, reboxetine. Previously, our own investigation on the moderating effect of 5-HTTLPR based on the multisite European GENDEP study also compared an SSRI, escitalopram, with a mainly noradrenergic drug, nortriptyline. Reference Huezo-Diaz, Uher, Smith, Rietschel, Henigsberg and Marušič15 In contrast with Lewis et al, we found that the polymorphism moderated the response to escitalopram (but not nortriptyline) with long-allele carriers improving more than short-allele homozygotes. We also found a significant threeway interaction between 5-HTTLPR, drug and gender, indicating that the effect was concentrated in men treated with escitalopram. The GENPOD study did not have sufficient power to rule out a small effect but the authors argued that their results make it unlikely that 5-HTTLPR genotype has a sufficiently large effect to form the basis of a clinically useful test.
It is worth considering these differing results against the background of other studies. GENDEP and GENPOD are two of the largest pharmacogenetic studies of antidepressants to date but are exceeded in size by the multisite US STAR*D study. This was not designed primarily as a pharmacogenetic study but two pharmacogenetic analyses have been performed. The first pharmacogenetic analysis of the STAR*D study, like the GENPOD analysis, found no effect of 5-HTTLPR on response to citalopram. However, this was carried out on a mixed-ethnicity sample. When a subsequent analysis was performed, dividing up the participants according to ethnic origin, a moderating effect of 5-HTTLPR on improvement of depressive symptoms in response to citalopram was indeed found, but only among White non-Hispanics. Ethnicity has also been a confounding factor in other analyses. A meta-analysis of 15 mainly fairly small studies, but with a combined total of 1435 individuals, showed that long-allele carriers were more likely to respond to SSRI antidepressants and/or enjoy a remission than short-allele homozygotes. Reference Serretti, Kato, De Ronchi and Kinoshita16 Again the effect was only in Europeans and was not present in Korean and Japanese populations included in the meta-analysis. Reference Kim, Lim, Kim, Kim, Chang and Carroll17,Reference Yoshida, Ito, Sato, Takahashi, Kamata and Higuchi18 One possible explanation of these ethnic differences could lie in the other common variants in the serotonin transporter gene that are thought to have functional effects.
One of these, a single nucleotide polymorphism (SNP) called rs2020933 in the first intron of the gene, had a small but significant effect on response to both escitalopram and nortriptyline in the GENDEP study Reference Huezo-Diaz, Uher, Smith, Rietschel, Henigsberg and Marušič15 and is known to have a large variation in allele frequency between populations. Variation in this SNP, which was not tested in the GENPOD study, is unlikely to explain GENDEP/GENPOD differences as both studies were exclusively of White Europeans, but it might explain some of the differences in other studies of Europeans versus non-Europeans.
An additional complexity is that an analysis of the GENDEP study taking self-rated life events into account showed that adversity in the 6 months before treatment interacted with 5-HTTLPR genotype in predicting response to antidepressants. Reference Keers, Uher, Huezo-Diaz, Smith, Jaffee and Rietschel19 This perhaps suggests that the 5-HTTLPR genotype has in fact a rather broad role as a marker of emotional reactivity and we can now turn to consider the experimental evidence relating to this more general hypothesis.
Experimental evidence of 5-HTTLPR effects on stress response
Interestingly, humans are not the only mammals that have a common functional variant in the promoter region of their serotonin transporter gene. Rhesus macaque monkeys also have a serotonin promoter polymorphism with long (more active) and short (less active) alleles. Macaques can be stressed early in life by being separated from their mothers and reared instead with a peer group. Their resultant ‘anxiety’ can be assessed both behaviourally and by measuring adrenocorticotropic hormone (ACTH) and cortisol levels at baseline and during separation stress. When such a study was performed on monkeys of known 5-HTTLPR genotypes the findings were somewhat analogous with the findings on adversity in humans. There was a separation×rearing×5-HTTLPR interaction, such that animals reared with peers who carried a short 5-HTTLPR allele had higher ACTH levels during separation than did the other animals studied. Reference Barr, Newman, Shannon, Parker, Dvoskin and Becker20 Rodents do not have a similar polymorphism but the effect of having a short allele early in development can be mimicked by giving them SSRIs as pups or by complete or partial ‘knockout’ of the serotonin transporter gene. Such disruptions also lead to exaggerated stress responses in animals with low or absent serotonin transporter activity compared with normal mice of the same strain. Reference Li, Wichems, Heils, Van De Kar, Lesch and Murphy21–Reference Caspi, Hariri, Holmes, Uher and Moffitt23
Two types of relevant experiment have been performed in human volunteers and in volunteers who have recovered from depression. In the first, healthy volunteers are exposed to stimuli such as fearful faces during functional brain magnetic resonance imaging (fMRI) when they typically show activation of the amygdala and associated regions. Hariri et al Reference Hariri, Mattay, Tessitore, Kolachana, Fera and Goldman24 showed that those individuals with one or two copies of the short allele of 5-HTTLPR showed greater amygdala neuronal activity, as assessed by blood oxygen level dependent fMRI, in response to fearful stimuli compared with individuals homozygous for the long allele. It has also been shown that short-allele homozygotes, more than other genotype subgroups, showed significantly greater positive functional connectivity between right amygdala and right fusiform gyrus and between right fusiform gyrus and right ventrolateral prefrontal cortex in response to prototypically fearful faces. Reference Surguladze, Elkin, Ecker, Kalidindi, Corsico and Giampietro25 In the second type of study, individuals have a low mood induced by giving them an amino acid mixture without tryptophan, which depletes them of this amino acid. One such study looked at women with or without a family history of depression who were genotyped for 5-HTTLPR. All women showed a reduction in plasma tryptophan levels but the most pronounced reduction in mood was in women who had both a family history of depression and who were homozygous for the short allele. Women who were long-allele homozygotes did not develop depressive symptoms, irrespective of family history. Reference Neumeister, Konstantinidis, Stastny, Schwarz, Vitouch and Willeit26
Taken together, these experimental studies reveal mechanisms of greater sensitivity to the environment that likely mediate the relationship between adversity and psychopathology, observed in the epidemiological studies. Reference Caspi, Hariri, Holmes, Uher and Moffitt23
Conclusions
Both the moderating effect of the serotonin transporter gene on antidepressant response and the alleged relationship between 5-HTTLPR genotype and depressive symptoms following life events have provoked controversy, most recently exemplified by Lewis et al's paper on the GENPOD study in the current issue of the Journal. Reference Lewis, Mulligan, Wiles, Cowen, Craddock and Ikeda1 We have argued that there are systematic effects that characterise the negative studies in the life events area and that when the problems resulting from low power and/or mixed ethnicity are taken into account there probably is a real, if small, effect of 5-HTTLPR on response to SSRIs. We agree with Lewis et al that testing for 5-HTTLPR genotype on its own is not likely to be clinically useful but it might, however, form part of a battery of tests that does turn out to have predictive utility. Finally, we are not just dealing with a body of evidence concerning 5-HTTLPR that is restricted to antidepressant response and response to life events. There is also a whole set of animal and human experiments that consistently point in the same direction in which possession of one or more short alleles is associated with greater reactivity to stress. The truth is that there are just too many straws travelling in the same direction for us not to know which way the wind is blowing.
eLetters
No eLetters have been published for this article.