The association between neuroticism and the serotonin transporter polymorphism depends on structural differences between personality measures

Abstract

Over the last decade the relationship between the serotonin transporter polymorphism (5-HTTLPR) and anxiety related personality traits has been intensely investigated. Although there are some meta-analytic approaches, the question remains open why the results of these studies are inconsistent. One suggestion is, that structural differences between personality measures lead to this heterogeneity. Based on this assumption we investigated the relationship between the 5-HTTLPR and neuroticism (N) as measured by the EPQ-R and NEO-FFI as well as harm avoidance (HA) as measured by the TCI in a sample of N = 410 healthy Caucasians. There were significant main effects of the s-allele for the EPQ-R-N (F = 6.98; df = 1; p = 0.009) as well as for the NEO-FFI-N (F = 4.08; df = 1; p = 0.044), but not for HA. In order to detect which aspects of negative emotionality are essential for the association with the 5-HTTLPR, all items of the three scales were tested with respect to their relation to the 5-HTTLPR. The 12 items which were significantly associated with the 5-HTTLPR form a new scale with a high internal consistency (α = .84). With respect to the item contents it can be summarized that especially the components of depression and stress sensitivity relate to the transporter polymorphism.

Introduction

There is strong evidence that a great part of the variability in personality is explained by genes (up to 60%, for example Bouchard, Lykken, McGue, Segal, & Tellegen, 1990), yet the molecular genetic basis of personality remains widely unknown. The best starting point to investigate, which genetic factors influence personality traits are personality theories that propose a biological basis. Some personality theories relate the activity of distinct neurotransmitter systems to distinct personality traits (e.g. Cloninger, 1987) thereby suggesting potential candidate genes to be investigated. Especially gene loci on candidate genes with known functional effects on neurotransmitter pathways can be investigated with respect to their associations to personality traits.

In this field of research the serotonergic system is of special interest. Both personality and psychopathological diseases are supposed to be associated with differences in the activity of serotonergic pathways. Serotonin (5-HT) plays a major role in depression and anxiety disorders (e.g. Bell and Nutt, 1998, Deakin and Graeff, 1991) and the modulation of the 5-HT system is crucial for their therapy. For example, selective serotonin re-uptake inhibitors (SSRIs) are highly effective antidepressants (Steffens, Krishnan, & Helms, 1997). Furthermore, the 5-HT system was also linked to anxiety-related personality traits. Cloninger (1987) for example proposed 5-HT to play a decisive role in harm avoidance (HA). In the same line Gray (1970) related the behavioral inhibition system (BIS) to serotonergic pathways. Thus this neurotransmitter system is especially suitable to be investigated with respect to the relationship between personality, psychiatric disorders and their underlying neurotransmitter functioning.

In their pioneer work Lesch et al. (1996) were able to demonstrate a significant association between the 5-HTTLPR polymorphism and anxiety-related personality traits, namely neuroticism, harm avoidance and the anxiety factor of Cattel’s 16PF questionaire. The 5-HTTLPR is a 44-bp insertion/deletion polymorphism in the promotor region of the serotonin transporter (5-HTT) gene. The shorter (s) allele of 5-HTTLPR was associated with reduced 5-HTT gene transcription rate leading to reduced 5-HTT binding sites and reduced 5-HT reuptake (Heils et al., 1996, Lesch et al., 1996). In the study of Lesch et al. the s-allele was identified as the dominant allele, implying that the presence of at least one short allele (genotypes ss or sl) is associated with significantly higher scores in personality traits of negative emotionality.

Since the first stimulating findings of Lesch et al. (1996) many attempts to replicate these results followed. While some studies support the association between the s-allele and neuroticism (e.g. Greenberg et al., 2000), or HA (e.g. Ricketts et al., 1998), other studies did not (e.g. Lang et al., 2004).

The first systematic approach to summarize the outcomes of these heterogeneous results in a meta-analysis found little evidence for an association between the 5-HTTLPR polymorphism and any anxiety-related personality trait (Munafo et al., 2003). However, the following attempts by Sen, Burmeister, and Ghosh (2004) and by Schinka, Musch, and Robichaux-Keene (2004) identified a significant main effect of the s-allele of the 5-HTTLPR on neuroticism but not on HA. Recently, in a fourth meta-analysis, Munafo, Clark, and Flint (2005a) included all studies with nonclinical samples that were previously taken into consideration plus two new available studies on this issue. Surprisingly, this most thorough investigation led to the opposite result: only the homozygous short ss genotype of the 5-HTTLPR polymorphism was associated with higher scores in HA but not in neuroticism. Hereupon both, Sen et al., 2005, Schinka, 2005 reanalyzed the data used by Munafo et al. (2005a) and were able to confirm their initial results, namely that the s-allele of the 5-HTTLPR is associated with higher scores in neuroticism but not in harm avoidance. Munafo, Clark, and Flint (2005b) concluded that different weighting technique in meta-analyses may bias results. Both neuroticism and harm avoidance seem to be related to the 5-HTTLPR, but the effect for neuroticism may be greater.

It was also suggested, that gender differences could be a source of the inconsistent findings, because men and women differ in their mean neuroticism scores, with women generally having higher scores (e.g. Jorm, 1987). But neither a study of Greenberg et al. (2000), nor a meta-analysis on this topic by Munafo, Clark, and Flint (2004) were able to find gender specific effects on the association between the 5-HTT gene and neuroticism.

Based on the data of the meta-analyses it can be concluded that the strength of the association between 5-HTTLPR and traits related to anxiety depends on the questionnaire used. Most of the studies used the NEO-FFI, NEO-PI or NEO-PI-R to measure neuroticism as conceptualized by the five factor model of personality by Costa and McCrae (1992). Although harm avoidance of the tridimensional personality questionnaire (TPQ)/temperament and character inventory (TCI) (Cloninger, 1987, Cloninger et al., 1994) and neuroticism of the big five share some variance (De Fruyt, Van De Wiele, & Van Heeringen, 2000), it can be assumed that there are some conceptual differences essential for the strength of the association between neuroticism and the 5-HTTLPR. For example the correlation between the N-score obtained by the NEO and HA as measured by use of the TCI in the study of De Fruyt et al., 2000 is .54. This indicates that taking a reliability <1 into account for both measures more than 70% of the variance remains unexplained. It is possible that special aspects of neuroticism could be crucial for the association with the s-allele of the 5-HTTLPR and that the broad concept of neuroticism may more or less overshadow the core dimension associated with serotonin. One way to identify these facets is to investigate similar traits by means of various measurement instruments with a slightly different focus.

There have been some attempts to analyze the relationship between the 5-HTTLPR and negative emotionality by using the NEO-PI and the TCI simultaneously in one sample. Three of them did not obtain any association to the 5-HTTLPR, neither with HA nor with N (Kumakiri et al., 1999, Nakamura et al., 1997, Samochowiec et al., 2004). But given the fact that the studies of Nakamura et al., 1997, Kumakiri et al., 1999 both obtained significant different genotype distributions in their Japanese samples than usually found in Caucasian samples, these negative findings indicate that the association is dependent on ethnicity. Furthermore, the sample of Samochowiec et al. (2004) only consists of 100 subjects, which might be too small in order to discover effects of one polymorphism. By contrast, Osher, Hamer, and Benjamin (2000) were able to detect associations between the s-allele of the 5-HTTLPR and both, HA as measured by the TPQ and two subscales of N (anxiety and depression) as measured by the NEO-PI-R in an Israeli sample. In this direct comparison HA seems to be more strongly related to the 5-HTTLPR than N.

It would be very interesting to use the EPQ-R, which covers a wide spectrum of facets among the construct of neuroticism (e.g. anxiety, somatic complaints, irritability, depression, tension), as an additional instrument to examine the association with the 5-HTTLPR. Unfortunately, there are only a few genetic association studies, that measured neuroticism in a nonclinical sample by means of the EPQ-R, and none of them reported a significant association (Jorm et al., 1998, Cruz et al., 1995, Willis-Owen et al., 2005). But there are some plausible reasons, why these studies could have failed to identify any significant association with the 5-HTTLPR. First, Cruz et al. (1995) had a very small sample size (N = 38) which renders this per se not very representative sample vulnerable for a stratification bias. Second, the sample of Jorm et al. (1998) was almost not in Hardy–Weinberg equilibrium questioning the quality of the data (i.e. possible genotype errors and stratification biases can not be excluded). Furthermore, Willis-Owen et al. (2005) favoured extreme-groups instead of a population-based design. There are hints, that the association between neuroticism and the 5-HTTLPR is caused by the normal variation in the middle range of neuroticism and does not occur in extreme groups (Sirota, Greenberg, Murphy, & Hamer, 1999).

Therefore, the aim of the present study is threefold. First, we try to replicate the association between the 5-HTTLPR polymorphism and anxiety related personality traits, i.e. neuroticism of the NEO-FFI and HA. Second, we want to investigate the association between the 5-HTTLPR and EPQ-R-N for the first time in a sample of appropriate size with a better likelihood to meet the Hardy–Weinberg equilibrium. Third, we try to clarify, which components of neuroticism are essential for the association between negative emotionality and 5-HTTLPR.