This study investigated whether somatic markers mediate the effect of serotonin transporter genotype on Iowa Gambling Task (IGT) performance. Participants (N = 135) were genotyped for the insertion/deletion and single-nucleotide (rs25531) polymorphisms in the promoter region of the serotonin transporter gene (5-HTTLPR). The results of mediation analyses indicated that skin conductance responses that anticipated IGT card selections partially (i.e. 42% of the total effect) mediated the effect of genotype on IGT performance. In comparison with high-functioning 5-HTTLPR genotypes, the low-functioning genotypes were associated with higher total IGT scores. This suggests that the higher synaptic availability of serotonin, associated with the low-functioning 5-HTTLPR genotypes, may confer differential susceptibility to decision making under risk, and that almost half of this effect is explained by facilitated somatic markers during IGT.
The Iowa Gambling Task (IGT) is extensively used in basic and clinical research on decision making. The original success of IGT was due to its sensitivity to ‘myopia for the future’, a type of functional impairment displayed by patients with ventromedial prefrontal cortex lesions (Bechara et al. 1994). These patients failed to develop a preference for the decks of cards yielding lower immediate gains but also smaller future losses during IGT, and persevered in choosing from the decks that yielded higher immediate gains but also higher future losses. In healthy controls, the preference for long-term advantageous decks in IGT was paralleled by the development of skin conductance responses (SCR) prior to the selection of cards (Bechara et al. 1997). Anticipatory SCR were higher before picking a card from the disadvantageous decks in healthy controls, but not in patients with ventromedial prefrontal cortex lesions (Bechara et al. 1999). These discoveries put IGT in the center of the somatic marker hypothesis (Bechara et al. 2000; Damasio 1996), which has argued that emotions, as reflected by anticipatory SCR during IGT, guide decision making. Although this theory has been debated (Bechara et al. 2005; Dunn et al. 2006; Guillaume et al. 2009; Maia & McClelland 2004), it remains highly influential in cognitive psychology, cognitive neuroscience and neuroeconomics.
Recent interest in IGT has concerned the neurochemical mechanisms that underlie optimal performance, with implications for understanding the decision-making dysfunctions in various neurological and psychiatric conditions (Bechara 2004; Verdejo-Garcia & Bechara 2009). It has been hypothesized that the biasing action of somatic markers (e.g. anticipatory SCR) on IGT performance may involve changes in neurotransmitter release in cortical and subcortical neural structures associated with executive functioning, working memory, emotions and the implementation of behavioral decisions (Bechara & Damasio 2005; Li et al. 2010). One of the candidate neurotransmitters involved in IGT is serotonin. It was shown that the administration of fluvoxamine, a selective serotonin reuptake inhibitor (SSRI), improved the selection of advantageous choices in IGT, and this effect was stronger in the second part of the task (Bechara et al. 2001). This part of IGT is thought to reflect decision making under risk because, by this stage, participants have probably acquired some subjective sense of the outcome probabilities associated with the different decks of cards (Bechara et al. 1997). Therefore, serotonin stimulation by SSRI may specifically facilitate decision making under risk in IGT, but this hypothesis should be further investigated considering that the preliminary findings of Bechara et al. (2001) have not been replicated. However, it was also observed that enhancing serotonin activity in humans by administering its precursor, 5-hydroxytryptophan, reduced decision-making biases (i.e. the reflection effect) in a task similar to IGT (Murphy et al. 2009). In contrast, tryptophan depletion in rats, which effectively reduced brain serotonin synthesis, impaired decision-making performance in a murine analog of IGT (Koot et al. 2012). Psychopharmacological studies in animals and humans seem to agree on the involvement of serotonin in decision making, but it is yet unclear whether increased serotonin activity has enhancing or impairing effects on IGT (Homberg 2012; Rogers 2011).
Selective serotonin reuptake inhibitor acts on the serotonin transporter (5-HTT), which is the key molecule responsible for the reuptake of serotonin. The human 5-HTT is encoded by a gene located on the chromosome 17 (17q11.1-q12). The promoter of this gene (5-HTTLPR) hosts two polymorphic loci, which result in variations in 5-HTT gene expression and 5-HTT function in the brain. An insertion/deletion polymorphism in the human 5-HTTLPR has been first reported (Lesch et al. 1996). This polymorphism has a short (S) allele comprising 14 copies of a 20–23 base pair unit, and a long (L) allele comprising 16 copies of the same repeat unit. The short allele is associated with reduced expression of 5-HTT gene and consequently reduced availability of 5-HTT in the brain (Little et al. 1998; Reimold et al. 2007). A more recently described polymorphism (rs25531) involves the substitution of an adenine to a guanine in the L allele of 5-HTTLPR (Hu et al. 2006). The LG allele functionally resembles the S allele (Ehli et al. 2012; Hu et al. 2006), so it is necessary to genotype both 5-HTTLPR polymorphisms (i.e. triallelic 5-HTTLPR) in order to correctly categorize the L alleles into low-functioning (LG) and high-functioning (LA). Therefore, an alternative to pharmacological manipulations of 5-HTT by SSRI is to study the influence of low- (S and LG) and high-functioning (LA) alleles of 5-HTT gene on IGT performance.
In light of the potentially enhancing effects of SSRI on IGT, one would expect that the low-functioning alleles of 5-HTT facilitated decision making in this task. Genetic knockout of 5-HTT in rats was indeed associated with a higher level of performance in a murine analog of IGT (Homberg et al. 2008). However, the lack of 5-HTT by gene knockout in rats may have different effects on IGT, in comparison with low 5-HTT functioning in humans. Indeed, the findings of recent genetic association studies on 5-HTT and IGT in healthy humans have been heterogenous. Two studies were able to confirm that S carriers made more advantageous choices in the first block of IGT (Stoltenberg et al. 2011; Stoltenberg & Vandever 2010), whereas another study found that only S carriers who also had a certain dopamine receptor type 4 genotype displayed IGT advantage (Ha et al. 2009). In contrast, three studies reported that S carriers showed reduced IGT performance (van den Bos et al. 2009; He et al. 2010; Homberg et al. 2008), and another study indicated no effect of 5-HTTLPR on IGT (Lage et al. 2011). Similar studies in clinical samples found that the S and LG alleles were associated with reduced IGT performance in obsessive-compulsive disorder (da Rocha et al. 2008), or had no influence on IGT in suicide attempters (Jollant et al. 2007).
Clearly, additional studies are crucial in order to elucidate the effects of 5-HTT on IGT, and set the stage for meta-analyses. Different groups focused on various IGT outcomes (e.g. total score, scores from a single 20-trial block and scores from the first block vs. the other four blocks), and these variables need to be considered. The effects of 5-HTT may also be different in clinical samples compared with healthy volunteers because the effects of the genotype may interact with the duration of the disease or previous medication in patients with emotional disorders. Moreover, only two of the previous studies in healthy participants (Ha et al. 2009; Stoltenberg et al. 2011) genotyped for rs25531 alleles, and this may have contributed to their success in finding positive effects on IGT. Another unexplored aspect, which would greatly increase the theoretical significance of this line of research, is related to the effects of 5-HTT genotype on somatic markers during IGT. Recent evidence confirmed that the difference between anticipatory SCR before advantageous and disadvantageous choices predicted IGT performance, and suggested that somatic markers and declarative knowledge independently contribute to IGT (Guillaume et al. 2009).
This study was designed to investigate the effects of 5-HTT genotype, including 5-HTTLPR and rs25531, on IGT performance and somatic markers during IGT. On the basis of the previous literature related to somatic marker hypothesis, we specifically expected that anticipatory SCR would mediate the effect of genotype on IGT performance.
Material and methods
N = 135 participants (118 women) volunteered for this study, following in-class presentations. Prior to study participation, written informed consent was obtained from all the volunteers. They were all Caucasians of Romanian descent, and came from the same well-circumscribed geographical area. Age ranged from 16 to 42 (M = 21.6 years). All the participants were compensated for their time. The study followed the recommendations of AMA's Declaration of Helsinki, and it was approved by the Babeş-Bolyai University Research Council.
DNA was extracted from leukocytes [ethylenediaminetetraacetic acid (EDTA)-anticoagulated blood] using Genomic DNA Extraction Kit (Fermentas, Vilnius, Lithuania) and kept at −20°C. Both biallelic and triallelic (i.e. including rs25531) 5-HTTLPR genotyping were performed using a protocol adapted after Lonsdorf et al. (2009) and Kosek et al. (2009). The polymerase chain reaction (PCR) assay conditions were optimized as follows: each reaction was carried out in a 25 µl volume [50 ng of genomic template, 12.5 µl PCR mastermix (2x)]; the forward primer (5′-GGCGTTGCCGCTCTGAATGC-3′) and reverse primer (5′-GAGGGACTGAGCTGGACAACCAC-3′), from Generi-Biotech (Hradec Kralove, Czech Republic), were used to amplify a region encompassing 5-HTTLPR. These primers yield amplicons of 529 (for L allele) or 486 bp (for S allele). Thermal cycling consisted of 3 min of initial denaturation at 94°C followed by 31 cycles of 94°C (40 seconds), 57°C (40 seconds) and 72°C (40 seconds), each with a final extension step of 4 min at 72°C. The LG and LA alleles were subsequently studied by enzymatic digestion of 10 µl of PCR products that were digested by HpaII (an isoschizomer of MspI) type FastDigest (Fermentas) in a 30 µl reaction assay at 37°C for 5 min. The restriction enzyme MspI recognizes and cuts a 5′-C/CGG-3′ sequence resulting in the following fragments: 340, 127 and 62 bp for the LA allele; 174, 166, 127 and 62 bp for the LG allele; 297, 127 and 62 bp for the SA allele; and 166, 131, 127 and 62 bp for the SG allele. Finally, 10 µl of remaining PCR product and 15 µl of restriction enzyme assay solution were loaded onto a 2.5% agarose gel, run for 2 h at 160 V in 0.5 × TBE running buffer and visualized by ethidium bromide for size estimation. The 5-HTTLPR allele frequencies were 0.45 for SA allele, 0.48 for LA allele and 0.05 for LG allele, similar to the ones reported by Hu et al. (2006) for Caucasians. One of the participants carried an ultralong (>16 repeats) allele (XL) (Ehli et al. 2012). The genotypes were categorized into low-functioning (i.e. carriers of two low-expressing alleles: SS, LGLG, SLG: N = 32), intermediate-functioning (i.e carriers of one low-expressing allele: SLA, LGLA: N = 66), and high-functioning (i.e. carriers of two high-expressing alleles: LALA, LAXL: N = 30). These genotypes were in Hardy–Weinberg equilibrium (χ2 = 0.13, not significant).
Iowa Gambling Task
The computerized version of IGT was used (Bechara et al. 1994). Briefly, IGT involves four decks of cards (A–D), from which participants are allowed to choose 100 cards. Each card from decks A and B is associated with a high gain ($100), whereas each card from C and D comes with a lower gain ($50). However, choosing a card is followed by unpredictable losses in all decks: for every 10 cards from A and B, there is a $1000 gain and $1250 loss; for every 10 cards from C and D, there is a $500 gain and $250 loss. Therefore, decks A and B are disadvantageous because of the net $250/10 cards loss, and decks C and D are advantageous due to the net $250/10 cards gain. The IGT score is obtained by subtracting the total number of disadvantageous choices from the total number of advantageous choices: [(C + D) – (A + B)]. Participants are allowed to choose 100 cards. Learning in IGT is operationalized as the improvement of the [(C + D) – (A + B)] scores from one block of 20 successive selections to another.
SCR was recorded during IGT, using a Biopac MP150 system (Biopac Systems, Goleta, CA, USA). Area under the curve (µS/s) was estimated from reward or punishment intervals (i.e. 5 seconds after the result of each selection was displayed), and anticipatory intervals (i.e. 5-second intervals before the result of each selection was displayed). The difference between anticipatory SCR before disadvantageous and advantageous cards was computed.
Analysis of covariance (ancova) was used to test for differences in IGT performance and SCR by 5-HTTLPR group. Pearson correlations were computed to determine associations between IGT performance and SCR. Mediation was tested using multiple regression analyses (Baron & Kenny 1986; Frazier et al. 2004; Hoyt et al. 2008). According to Baron and Kenny (1986), the model had to satisfy three conditions in order to be confirmed (Fig. 1): (1) the initial variable (5-HTTLPR group) was significantly related to the outcome variable (CD–AB score) (Path c); (2) the initial variable was related to the mediator (anticipatory SCR difference between CD and AB cards) (Path a) and (3) the mediator was significantly associated with the outcome variable when regressed on both the mediator and the initial variable (Path b), and the effect of the initial variable (Path c′) was reduced compared with that in the first regression (Path c). The Aroian version of the Sobel test was used to test whether the indirect effect was significant (Aroian 1947; Baron & Kenny 1986). All the analyses were run in spss V.13.0.
Iowa Gambling Task performance and somatic markers by 5-HTTLPR group
A 3 (genotype: low- vs. intermediate- vs. high-functioning) × 5 (block of trials) ancova, with sex as covariate, indicated significant main effects of genotype (F2,132 = 4.61, P = 0.01, Cohen's d = 0.96) and block (F4,130 = 5.79, P = 0.003, Cohen's d = 1.56) on CD–AB scores, as well as a significant interaction of genotype and block (F7,127 = 4.1, P = 0.01, ohen's d = 0.87). In comparisons with the high-functioning group, the low- and intermediate-functioning groups had significantly higher CD–AB scores in blocks 2–4 (i.e. trials 21–80) (Fig. 2a). A similar ancova analysis found a significant effect of genotype on CD–AB anticipatory SCR (F2,132 = 3.09, P = 0.04, Cohen's d = 0.45), with higher SCR in low- and intermediate-functioning genotypes compared with high-functioning genotype (Fig. 2b). There was a statistically significant correlation between CD–AB anticipatory SCR and CD–AB scores (r = 0.29, P = 0.04). There were no statistically significant effects on SCR during reward or punishment intervals.
Somatic markers mediating the association between 5-HTTLPR and Iowa Gambling Task performance
Table 1 describes the results of the analyses that examined the mediation hypothesis (see also Fig. 1). As predicted, genotype was significantly related to IGT performance (Path c) and to anticipatory SCR (Path a). When regressing IGT performance on both genotype and anticipatory SCR, the coefficient associated with the relation between anticipatory SCR and IGT performance, while controlling for genotype (Path b), was significant. The coefficient associated with the relation between genotype and IGT performance, while controlling for anticipatory SCR (Path c′), remained significant. However, Path's c′ coefficient was smaller than Path's c and anticipatory SCR was a significant mediator (Aroian's z = 5.33, P = 0.00). The ratio a×b/c indicated that about 42% of the total effect of genotype on IGT performance was mediated by anticipatory SCR. A reverse-causality analysis that tested a path from IGT performance to anticipatory SCR indicated a consistently smaller indirect effect (18% of the total effect). The mediation model was also tested after excluding the fewer men from the sample, and the results were closely similar (i.e. 41% mediation by anticipatory SCR).
Table 1. Path coefficients for the mediation by anticipatory SCR of the relation between genotype and IGT performance
Testing steps in mediation model
B, unstandardized regression coefficient; β, standardized regression coefficient; CI, confidence interval; SE, standard error.
*0 = high-functioning genotype (i.e. LA/LA and LA/XL genotypes); 1 = intermediate-functioning genotype (i.e. S/LA and LG/LA genotypes); 2 = low-functioning genotype (i.e. S/S, LG/LG and S/LG genotypes).
*P < 0.05;
**P < 0.01.
Testing step 1 (Path c)
Outcome: [(C + D) – (A + B)] score
Predictor: genotype (low- vs. intermediate- vs. high-functioning)*
As hypothesized, we found that anticipatory SCR during IGT partially mediated the relation between 5-HTTLPR genotype and IGT performance. In terms of causation, genotype preceded both SCR development (the mediator) and learning IGT (the outcome). In addition, we tested an alternative model in which IGT performance mediated anticipatory SCR and found that the indirect effect dropped significantly. Therefore, these results support the view that low-functioning 5-HTTLPR genotypes are associated with increased IGT performance (partially) because they lead to higher anticipatory SCR during IGT.
The present finding that low-functioning alleles of 5-HTTLPR facilitate IGT performance is in line with previous genetic association studies in humans (Stoltenberg et al. 2011; Stoltenberg & Vandever 2010). This effect is similar to that of SSRI, which may facilitate IGT performance and particularly decision making under risk (Bechara et al. 2001), as well as to observations from psychopharmacological manipulations of tryptophan (Koot et al. 2012; Murphy et al. 2009). Our results are also supported by the previous observation that 5-HTT knockout in rats increased performance in the second half of a murine IGT analog (Homberg et al. 2008). Overall, these studies indicate that the reduced function of 5-HTT, which probably leads to an increased synaptic availability of serotonin, is beneficial to decision making in IGT. 5-HTTLPR may thus illustrate differential susceptibility (Belsky & Pluess 2009; Homberg & Lesch 2010) to IGT learning environment, particularly in the second part of the task. From this perspective, carriers of low-functioning alleles of 5-HTTLPR may be particularly susceptible to learning from environments that include risk, in which the decision maker has complete information regarding the stochastic relationship between actions and outcomes (Rangel et al. 2008).
We found that anticipatory SCR mediated 42% of the total effect of genotype on IGT performance. This partial, yet consistent mediation provides a new type of support for the somatic marker hypothesis (Bechara et al. 2000). To our knowledge, this is the first study showing that a genetic-driven difference in somatic markers has impact on IGT performance. These results contribute to the recent literature on how genes contribute to individual differences in decision making (Ebstein et al. 2010), and may open a new and important line of research related to somatic marker hypothesis and the contribution of emotions to IGT. In light of debates surrounding the somatic marker hypothesis, future studies might test a complementary model in which declarative knowledge acquired during IGT might mediate another portion of the effect of 5-HTTLPR on IGT performance. Somatic markers and declarative knowledge independently contribute to IGT performance (Guillaume et al. 2009), and it is thus possible that they are influenced by different genetic polymorphisms or perhaps gene × environment interactions. To date, twin studies have not approached the genetic and environmental influences on IGT and its underlying mechanisms. An intriguing possibility, although entirely speculative, is that heritability contributes to IGT through somatic markers, and individual environment influences IGT through declarative knowledge. Therefore, subsequent research may investigate whether 5-HTTLPR interacts with other genetic polymorphisms (Ha et al. 2009; da Rocha et al. 2011) and environmental factors (e.g. socioeconomic status) in influencing the susceptibility to learning IGT. Somatic markers and IGT performance may qualify for endophenotypes of neuropsychiatric conditions (e.g. obsessive-compulsive disorder, bipolar disorder, pathological gambling, addiction), and shed light on the pathogenesis of these conditions (Verdejo-Garcia & Bechara 2009).
This design was based on a priori predictions, which were derived from an empirically supported theory. Our hypotheses were also in line with previous experimental results from psychopharmacological studies in humans and genetic engineering in animals. As expected, the results supported our model. However, it is noteworthy that the association of low-functioning 5-HTTLPR alleles with higher IGT performance is not in line with several previous studies (He et al. 2010; Homberg et al. 2008; van den Bos et al. 2009). Most of these studies did not genotype rs25531, and the failure in categorizing L alleles into low-functioning LG and high-functioning LA may explain the divergence of their results from the theoretically and empirically supported view. Future studies might thus try to replicate these results by genotyping both 5-HTTLPR polymorphisms (Ehli et al. 2012; Hu et al. 2006; Lesch et al. 1996), and measuring IGT, somatic markers and declarative knowledge.
In conclusion, this study showed that anticipatory SCR during IGT mediate part of the effect of 5-HTTLPR on IGT performance, with low-functioning alleles conferring an advantage in learning IGT by their association with increased somatic markers.
We are grateful to Simona Pană, Silviu Matu, Ioana Cocia, Bianca Blaj and Julia Avram for help with data collection. This research was supported by grant 411/2010 from the National Council of Scientific Research in Higher Education (CNCSIS) to A.C.M.