Differential effects of 5-HTTLPR and DRD2/ANKK1 polymorphisms on electrocortical measures of error and feedback processing in children

1. Introduction 

Psychiatric disorders have a well-established genetic background (Sanders et al., 2004).Yet, only relatively few specific genetic polymorphisms have been identified as being associated with mental disorders. Moreover, associations of these polymorphisms with the disorders are typically weak. This may be explained by the large heterogeneity and complexity of clinical phenotypes that are based on rather global diagnostic criteria (Faraone et al., 2005).

Recent advances in the field of imaging genetics suggest that the effects of genes on brain morphology and function are larger than those on disease phenotypes. Hariri and Weinberger (2003) created the concept of “imaging genomics”, in which the phenotype has been proposed to be the physiological response of the brain during specific information processing (Brown and Hariri, 2006). Given the much stronger link between genomic variation and brain activity, samples to be investigated may be much smaller than those that have been used in patient-control comparisons based on clinical diagnoses (Fallgatter et al., 2004).

The present study employed the measurement of brain function by means of electroencephalogram (EEG) event-related potentials (ERPs) obtained during error and feedback processing in relation to common polymorphisms of two genes, the serotonin transporter (5-HTTLPR) gene and the D2 dopamine receptor (DRD2) gene. These two genes may differentially contribute to learning from feedback on errors as they have been suggested to mediate different personality traits that, however, have both been found to predispose for alcohol dependence (Wu et al., 2008). Based on a neurobiological learning model (Gray, 1985, Gray, 1987), Cloninger, 1987a, Cloninger, 1987b suggested that alcohol dependency might develop from either an overactive behavioral inhibition system (BIS) which is associated with harm avoidance and proposed to be mediated by serotonergic processes or an overactive behavior activation system (BAS), which is associated with novelty seeking and should be mediated by the dopaminergic system.

The 5-HTTLPR gene plays an important role in serotonergic neurotransmission by facilitating serotonin (5HT) reuptake from the synaptic cleft (Heils et al., 1995). It is known to have two alleles, which differ in the number of variable repeat sequences in the promoter region: a low activity short (S) variant and a long (L) variant. Compared to carriers of the L variant, individuals carrying the S variant have repeatedly been suggested to be prone to anxiety-related personality traits (Brown and Hariri, 2006, Jacob et al., 2004, Sen et al., 2004), to show augmented neural processing of aversive stimuli (Canli et al., 2005) and greater sensitivity to stimuli associated with punishment (Finger et al., 2007) as well as to be liable to alcohol dependence (e.g., Feinn et al., 2005, Lin et al., 2007). Although recently a tri-allelic variation has been identified suggesting functionally different polymorphisms within the long variant (e.g., Hu et al., 2006), we followed the vast amount of the literature by grouping according to the assumption that the S variant functionally dominates upon the L variant (Hariri et al., 2005, Otte et al., 2007) and therefore compared a group of homozygous carriers of the L variant (LL) with carriers of at least one S allele (SL and SS).

The DRD2 gene has multiple allelic forms, one of which, the Taq1 A1 polymorphism has been related to a reduced D2 dopamine receptor binding affinity (Noble, 2003) and lower dopamine receptor density in the striatum (Jonsson et al., 1999). Its presence has been suggested to play a central role in the neuromodulation of appetitive behaviors, and to be associated with smoking and alcoholism (Bowirrat and Oscar-Berman, 2005, Munafo et al., 2007, Preuss et al., 2007), gambling (Comings et al., 1996), and sensitivity to stress (Bau et al., 2000, Pani et al., 2000). The DRD2 Taq1 A1 allele has therefore been related to what is conceptualized as the Reward Deficiency syndrome, pointing to an inefficiency in the acquired reward system (Bowirrat and Oscar-Berman, 2005). Different from natural rewards that include the satisfaction of only physiological drives, acquired rewards are defined as positive reinforcers, i.e. events that increase the probability of a subsequent response. Note that although the Taq1 A1 variant has recently been described to alter an amino acid in a protein kinase gene (ankyrin repeat and kinase domain containing 1; ANKK1) identified in a less than 10kb downstream region of the DRD2 locus (Neville et al., 2004) implying the possibility that changes in ANKK1 activity may explain the described associations between the DRD2 variant and neuropsychiatric disorders, we decided to keep referring to the variant as the DRD2 Taq1 A1 polymorphism, because this agrees with the nomenclature used in the majority of published studies to date.

Augmented neural processing of aversive stimuli, greater sensitivity to stimuli associated with punishment, and a less efficient processing of positive reinforcement can be studied by measuring electrocortical correlates of error and feedback processing. This type of processing is generally referred to as performance monitoring (Stuss et al., 1995, Ullsperger, 2006). Performance monitoring is described as a process of adapting behavior by making use of negative and positive feedback from the environment or comparing the action at hand to an internal representation of the intended action. These abilities are conceptualized as external and internal performance monitoring, respectively (Müller et al., 2005). Since the early nineties they have been thoroughly studied by means of EEG ERPs (e.g., Falkenstein et al., 1991, Gehring et al., 1990, Miltner et al., 1997). A paradigm allowing for measuring both aspects of internal and external performance monitoring originates from the probabilistic learning task developed by Holroyd and Coles (2002). In this task, subjects are required to learn particular stimulus–response combinations by making use of performance feedback that is contingent to their responses. It allows for investigating the transition from external to internal monitoring as learning by feedback proceeds throughout the course of the task. To this end ERPs that are time-locked to the response as well as time-locked to the feedback stimuli are examined.

Our study was conducted on a group of primary school-aged children who had participated in an experiment that was aimed at investigating whether different psychopathological conditions differentially affect error and feedback processing (Groen et al., 2008). From the majority of these children DNA samples were obtained.

The task applied was a modified version of the probabilistic learning paradigm adapted for completion by children (Crone et al., 2004, Groen et al., 2007). This task has been shown to evoke several distinguishable ERP components that were highly sensitive to the task manipulations (Groen et al., 2007) as well as to differences between children with different types of psychopathology (Van Meel et al., 2005, Groen et al., 2008). In the present study we investigated four of these components, two of them related to the children’s response and two related to the feedback upon their responses.

Response-locked components were an early error-related negativity (ERN; Falkenstein et al., 1991, Gehring et al., 1990, Gehring et al., 1993, Groen et al., 2007, Van Meel et al., 2007) with a fronto-central scalp distribution and an onset at or shortly before the commission of an incorrect response until about 100ms thereafter, as well as a later occurring error-related positivity (Pe) peaking approximately 200–400ms post-response with a maximum at parietal electrode sites (Davies et al., 2001, Falkenstein et al., 1991, Groen et al., 2007). While the ERN has been associated with a rather unconscious process of error detection, the Pe has been suggested to reflect conscious error processing that facilitates adaptive behavior (Davies et al., 2001, Leuthold and Sommer, 1999, O’Connell et al., 2007, Overbeek et al., 2005). Both components have been found to be increased in response to the commission of errors.

Concerning the feedback-locked ERPs, a feedback-related ERN as described in previous studies (Gehring and Willoughby, 2002, Holroyd and Coles, 2002, Müller et al., 2005, Van Meel et al., 2005) appeared not to be sensitive to the manipulations of the task paradigm applied in the present study (see, Groen et al., 2007). Yet, two other components were shown to vary with the task conditions. These were a feedback-related P3, maximal at centro-parietal sites in the range of 200–450ms after feedback onset and another, later (up from 450ms after feedback onset) occurring and longer lasting centro-parietal positivity, which is referred to as the Late Positive Potential. Both were found to be larger in response to negative feedback as compared to positive feedback. While the feedback-P3 has been suggested to reflect the updating of task rules from long term memory in response to error feedback (Donchin and Coles, 1988, Groen et al., 2007), the LPP, has been thought to reflect increased attention to affective-motivational stimuli because it has repeatedly been found in response to highly arousing pleasant and unpleasant pictures (Cuthbert et al., 2000, Hajcak et al., 2006, Schupp et al., 2000).

In a group of healthy children, the response-locked error potentials could be shown to increase when learning proceeded while the feedback-locked potentials appeared to decrease with task progression, thus reflecting an increase in internal monitoring that is accompanied by decreasing dependency on external feedback (Groen et al., 2007).

The present study investigated whether the above-described error- and feedback-related ERP components are differentially affected by the serotonergic 5-HTTLPR and dopaminergic DRD2 polymorphisms as well as whether the combined occurrence of the short 5-HTTLPR variant with the DRD2 Taq1 A1 allele might amplify possible effects. Given that enhanced neural processing of aversive stimuli and greater sensitivity to stimuli associated with punishment has been reported for carriers of the 5-HTTLPR S-allele we expected the group of children with one or both S-alleles to show greater sensitivity to internally monitored errors (ERN and Pe) as well as to externally monitored negative feedback compared to the children with the L alleles. More specifically, this would imply the occurrence of a greater early frontal negativity and a greater somewhat later observable parietal positivity related to incorrect responses as well as a greater parietal response to negative feedback in S allele carriers as compared to homozygous L allele carriers. As the Taq1 A1 allele of the DRD2 gene has predominantly been related to deficient reward processing we expected the children carrying this allele to be different from the non-carriers specifically in their feedback-related ERP responses, in particular their LPP related to positive feedback. Group differences emerging from these pairwise group comparisons might become even more evident when comparing the children carrying both the S-variant of the 5-HTTLPR and the Taq1 A1 variant of the DRD2 gene to the children possessing neither of these variants.

2. Methods 

2.1. Subjects 

The sample consisted of 65 normally intelligent children (51 boys and 14 girls; mean age=11.41, SD=0.91, range=10–12 years), either with a Pervasive Developmental Disorder (PDD; N=18), or Attention Deficit Hyperactivity Disorder (ADHD, N=27), or being healthy controls (N=20) who had all participated in an experiment investigating performance monitoring, and of whom DNA samples had been taken for genotyping. Clinical diagnoses, as established by independent child psychiatrists, were based on DSM-IV-TR criteria (American Psychiatric Association, 2000) and several standardized PDD and ADHD behavior questionnaires (for more details we refer to Groen et al. (2008)).

In the original study by Groen et al. (2008) there were 35 children with ADHD, 31 of whom were methylphenidate (MPH) responders, all taking this drug during the main part of the year preceding the experiment. These MPH responders were randomly assigned to an MPH-treated or medication-free condition. Given an MPH half-life of about 2h, those assigned to the medication-free condition were asked to discontinue MPH-intake for at least 17h before they entered the experiment. The remaining four of the 35 children with ADHD did not yet use medication for their ADHD-symptoms and were directly assigned to the medication-free group.

For the present study, DNA was obtained from 27 children with ADHD; 13 were taking MPH, while 14 had been medication-free for at least 17h at the time of the experiment, two of them not having used MPH before. All children in the PDD group were medication-free at the time of the experiment; children taking any other psychotropic drug were excluded from the study. How the heterogeneity of the sample has been dealt with when grouping according to the polymorphisms is described below.

Aim and study procedures were fully explained to the patients and their parents before written consent was obtained from the parents as well as the12-year-olds. The study had been approved by the medical ethics committee of the University Medical Center Groningen.

2.1.1. Genotyping 

Buccal smears were collected using cervical brushes. For the sake of obtaining reliable DNA three samples were taken from each of the participating children, one in the morning, one in the noon and one in the evening. The samples were stored in buffer containing proteinase K and sodium dodecylsulfate. DNA was isolated using salt extraction followed by iso-propanol precipitation. Based on validation experiments in the laboratory, we expect an error rate below 1% for the 5-HTTLPR genotyping because errors were minimized by cross-checks during the crucial steps by the technicians and the use of automated systems for samples and PCR buffers. For the DRD2, all samples were duplicated with a verification rate of 100%.

2.1.2. DRD-2 genotyping 

The DRD-2 Taq1 polymorphism was determined using real-time polymerase chain reaction (PCR). We used primers DRD2-GAF (5′-GCAACACAGCCATCCTCAAAG-3′) and DRD2-GAR (5′-GTGCAGCTCACTCCATCCT-3′) for DNA amplification and probes DRD2-GAV2 (VIC-CTGCCTCGACCAGC) and DRD2-GAM (FAM-CTGCCTTGACCAGC) to detect the Taq I A2 allele (G) and Taq I A1 (A) allele, respectively (Assay by design, Applied Biosystems, Nieuwerkerk a/d IJssel, The Netherlands). PCR was performed using Taqman Universal master mix (Applied Biosystems) and 5‰ bovine serum albumin. After initial denaturation (95°C, 10min) amplification took place using 40 cycles of denaturation (92°C, 15s) and annealing/extension (60°C, 60s). PCR and detection were carried out using an Applied Biosystems 7500 Real-Time PCR system. Primer and probe sequences were based on the NCBI sequence AF050737.

2.1.3. Genotyping 5-HTTLPR 

5-HTTLPR genotypes were determined using the HTTp2a and HTTp2B primer set to amplify 406 (S) and 450 (L) bp fragments using PCR (Cook et al., 1997). The LA, LG and S alleles13 were determined by incubation of the PCR product with the restriction enzyme Msp I (New England Biolabs, Westburg, Leusden, The Netherlands) for at least 3h at 37°C. Msp I cuts the GGCC sequence, resulting in fragments of 329, 62, and 59 (LA), 174, 155, 62 and 59bp (LG), and 285, 62 and 59bp (S), respectively. The resulting restriction fragments were separated using a 2% agarose gel and visualized using GelStar (SYBR-green; Cambrex Bio Science, Rockland, ME).

2.1.4. Grouping by variants of the 5-HTTLPR serotonin transporter gene 

Fifteen children were 5-HTTPLR S homozygotes (SS), 20 L homozygotes (LL), and 30 heterozygotes (SL). Assuming functional dominance of the S allele (Brown and Hariri, 2006), two groups were formed consisting of 45 S carriers (SS/SL) and the 20 LL carriers, respectively. Since these groups were not equal with respect to the presence of the DRD2 Taq1 A1 allele, gender, clinical, and medication status, we matched 20 of the 45 S carriers to those with only L variants on these variables. This matching was considered necessary as effects of gender, clinical diagnosis, and the use of MPH on the investigated ERPs have previously been reported (Davies et al., 2001, Van Meel et al., 2005, Jonkman et al., 2007, Groen et al., 2008). It resulted in two groups of 20 children perfectly matched on the DRD2 gene variants, clinical status, and medication, with, however, three more girls (n=6) in the LL group. The distribution of the children across the matching variables before and after matching is presented in Table 1. The two 5-HTTLPR groups did not differ in age [M1=M2=11.4] or intelligence [M1 (SD)=103.8 (13.2); M2 (SD)=102.4 (10.6)].

Table 1. Distribution of the children in the two 5-HTTLPR groups across gender, clinical status, and medication (ADHD+: taking MPH; ADHD−: being MPH-free during the experiment). Numbers between brackets present the original numbers from which the matched groups were drawn.

		5-HTTLPR
		LL					SS/SL
		GG		Taq1A		Total	GG		Taq1A		Total
		Boys	Girls	Boys	Girls		Boys	Girls	Boys	Girls
	Controls	3	1	1	2	7	3	1	1	2	7 (13)
	ADHD+	2	0	2	0	4	2	0	2	0	4 (9)
	ADHD−	1	2	0	0	3	3	0	0	0	3 (11)
	PDD	5	0	0	1	6	5	0	1	0	6 (12)
	Total	11	3	3	3	20	13 (23)	1 (5)	4 (14)	2 (3)	20 (45)

2.1.5. Grouping by the DRD2 gene variants 

Twenty-three children possessed at least one Taq1 A1 allele (three of them had both copies), the remaining 42 children were non-carriers (GG). As here again neither the presence of the S and L variants of the 5-HTTPLR gene nor the children’s clinical status and gender were equally distributed across the groups we matched the groups by these three variables taking the distributions of the Taq1 A1 group as point of departure. Dropping two girls from the Taq1 A1 group resulted in two groups of 21 children who were perfectly matched on the 5-HTTPLR variants, gender, and clinical status, and medication (Table 2). The groups did not differ in age [M1=M2=11.5] or intelligence [M1 (SD)=102.1 (13); M2 (SD)=101.1 (11.1)].

Table 2. Distribution of the children in the two DRD2 groups across gender, clinical status, and medication (ADHD+: taking MPH; ADHD−: being MPH-free during the experiment). Numbers between brackets present the original numbers from which the matched groups were drawn.

		DRD2
		Taq1A					GG
		SS/SL		LL		Total	SS/SL		LL		Total
		Boys	Girls	Boys	Girls		Boys	Girls	Boys	Girls
	Controls	2	2	3	1	8	2	2	3	1	8 (10)
	ADHD+	3	0	1	0	4	3	0	1	0	4 (9)
	ADHD−	3	0	0	0	3	3	0	0	0	3 (11)
	PDD	4	1	1	0	6	4	1	1	0	6 (12)
	Total	12	3	5	1	21	12 (23)	3 (5)	5 (11)	1 (3)	21 (42)

2.1.6. Grouping by combined variants of the 5-HTTPLR and DRD2 gene 

Another two groups were formed for the comparison of those children carrying both one or two 5-HTTPLR S alleles and the DRD2 Taq1 A1 allele (SA: n=17) to those children possessing neither of them (LL/GG: n=14). Due to small sample sizes, these groups could not be matched according to gender and clinical or medication status. However, the groups turned out rather similar on these variables (see Table 3) as well as on age [M1 (SD)=11.5 (0.9); M2 (SD)=11.3 (0.9)] and intelligence [M1 (SD)=100.4 (10.9); M2 (SD)=101.6 (11)].

Table 3. Distribution of the two groups carrying either both the DRD2 Taq1 A1 and 5-HTTLPR S allele or none of these, respectively.

		5-HTTLPR/DRD2
		SA		Total	LL/GG		Total
		Boys	Girls		Boys	Girls
	Controls	4	3	7	3	1	4
	ADHD+	2	0	2	2	0	2
	ADHD−	3	0	3	1	2	3
	PDD	5	0	5	5	0	5
	Total	14	3	17	11	3	14

2.2. Task and experimental procedure 

The children performed a probabilistic learning task in which they had to learn stimulus–response (S–R) combinations by making use of performance feedback. As the task has been thoroughly described in a previous report of our group (Groen et al., 2007), only the essentials of the task are described here. The whole experiment consisted of nine different task blocks. Within each block, which consisted of 96 stimulus presentations (trials), four colored pictures belonging to the categories animals, fruits, music, and sports, were randomly presented on a PC screen. For each of the four pictures, the children had to discover which of two keys to press by attending to feedback stimuli. In the beginning of each block, they were ignorant of the two feedback conditions that were assigned to the stimuli. Two of the four pictures (A and B) were always followed by informative feedback. Pressing the left key to picture A always resulted in positive feedback (indicated by a green square appearing at the PC screen), while pressing the right key resulted in negative feedback (indicated by a red square). For picture B this coupling was opposite: pressing the right key resulted in positive feedback and pressing the left key was followed by negative feedback. The other two pictures (C and D) were followed by uninformative feedback. The feedback valence for picture C was always positive and that for picture D always negative, i.e. the feedback stimuli were unrelated to the children’s response. The uninformative feedback condition had been included to control for the validity of the feedback manipulations. An example of a single trial is presented in Fig. 1.

View Large Image Download to PowerPoint

Fig. 1. Time course of a single trial. Within one task block each trial started with the presentation of one out of four stimuli. The feedback stimulus appeared 1000ms after stimulus off-set and stayed on the screen for 1500ms. The next trial started after a variable Inter Trial Interval (ITI) of 500, 750, or 1000ms. Originally published in Groen et al. (2007).

Each of the nine blocks, which were randomly presented, initiated a new learning process with four new pictures. Within the informative condition the number of trials for each feedback valence was variable as it depended on the individual error rate of the child. In the uninformative condition the number of trials for both positive and negative feedback was 24. Instructions were to win as many points as possible. Positive feedback and negative feedback indicated the win or loss of one point, respectively. Feedback indicating loss of two points appeared on the screen when the child responded too late, i.e. after a previously determined individual deadline. This individual response deadline (mean reaction time+10%) was introduced to elicit enough error trials for computing error-related potentials and to take into account individual differences in response speed. It was determined in a deadline determination block before the start of the actual experiment, in which a black square appeared on the screen when children responded too late. The children started with 52 points in the beginning of each block, which could add up to a maximum of 100 points. Standardized instructions and a practice block of 24 trials proceeded the deadline determination block containing 96 trials. When prepared for physiological recording the children performed the nine task blocks with a 20min break after the fifth block. At the end of the experiment all children received a present independent of the number of points they won.

3. Results 

3.2. Event-related potentials 

Task manipulation effects (tested on the whole group of 65 children) as well as group effects are summarized in Table 4, Table 5. Fig. 2, Fig. 3, Fig. 4 show the ERPs for the three pair-wise group comparisons conducted locked ERN and Pe (Fig. 2, Fig. 3) and feedback-locked P3 and LPP (Fig. 4). Group interaction effects are depicted in Fig. 5(a)–(i). While the tables present (minimum and maximum) F-ratios (of consecutive intervals) with corresponding p-values and effect sizes the figures show mean amplitudes of the (successive) intervals that contained group effects with at least medium effect sizes. Note that all interactions shown in the figures are significant at p⩽.05.

Table 4. ANOVA results of the response-locked ERN and Pe measured at Fz and Pz, respectively. F-ratios and corresponding significance levels as well as effect sizes for interaction effects are presented below the (successive) intervals for which they were found. Note that minimal F-ratios (Fmin) with a p⩾.5<.1 form part of a series of successive effects with at least medium effect size (η2⩾.06).

		ERN at Fz (−100 to 100 ms)		Pe at Pz (100 to 600 ms)
	Task manipulations	Fmin(1, 64)	Fmax(1, 64)	Fmin(1, 64)	Fmax(1, 64)
	Response type	−100 to 60 ms		100 to 550 ms
		F=14.9; p<.001; η2=.19	F=59.2; p<.001; η2=.49	F=13.5; p<.001; η2=.18	F=306; p<.001; η2=.83
	Response type by task half	−80 to 40 ms		100 to 450 ms
		F=4.2; p=.04; η2=.07	F=14.9; p<.001; η2=.19	F=5.4; p=.02; η2=.08	F=57.1; p<.001 ; η2=.48
	Group comparisons
	1) 5-HTTLPR: SS/SL vs. LL	Fmin(1, 38)	Fmax(1, 38)	Fmin(1, 38)	Fmax(1, 38)
	Group	− 80 to 80 ms		n.s.
		F=3.7; p=.06; η2=.09	F=5.3; p=.03; η2=.12
	Group by response type	n.s.		200 to 400 ms
				F=4.1; p=.05; η2=.10	F=6.1; p=.02; η2=.14
	Group by response type by task half	− 60 to 0 ms		n.s.
		F=3.7; p=.06; η2=.09	F=6.6; p=.01; η2=.15
	2) DRD2: Taq 1 A vs. GG	Fmin(1, 40)	Fmax(1, 40)	Fmin(1, 40)	Fmax(1, 40)
	Group	n.s.		n.s.
	Group by response type	n.s.		n.s.
	Group by response type by task half	n.s.		n.s.
	3) 5-HTTLPR / DRD2: SS/SL + Taq1 A vs. LL + GG	Fmin(1, 29)	Fmax(1, 29)	Fmin(1, 29)	Fmax(1, 29)
	Group	n.s.		n.s.
	Group by response type	n.s.		n.s.
	group by response type by task half	− 20 to 0 ms		n.s.
		F=5.3; p=.03; η2=.15

Table 5. ANOVA results of the feedback-locked P3 and LPP both measured at Pz. F-ratios and corresponding significance levels as well as effect sizes for interaction effects are presented below the (successive) intervals for which they were found. Note that minimal F-ratios (Fmin) with a p⩾.5<.1 form part of a series of successive effects with at least medium effect size (η2⩾.06).

		P3 at Pz (200 to 450 ms)		LPP at Pz (450 to 900 ms)
	Task manipulations	Fmin(1, 64)	Fmax(1, 64)	Fmin(1, 64)	Fmax(1, 64)
	Feedback valence	200 to 450 ms		450 to 800 ms
		F=12.9; p=.001; η2=.17	F=29.8; p<.001; η2=.31	F=4.1; p=.05; η2=.06	F=9.5; p=.003; η2=.13
	Valence by task half	n.s.		500 to 700 ms
				4.0; .05; .06	6.5; .01; .09
	Group comparisons
	1) 5-HTTLPR: SS/SL vs. LL	Fmin(1, 38)	Fmax(1, 38)	Fmin(1, 38)	Fmax(1, 38)
	Group	n.s.		n.s.
	Group by valence	n.s.		n.s.
	Group by valence type by task half	350 to 450 ms		n.s.
		F=2.82; p=.1; η2=.07;	F=4.8; p=.03; η2=.11
	2) DRD2: Taq 1 A vs. GG	Fmin(1, 40)	Fmax(1, 40)	Fmin(1, 40)	Fmax(1, 40)
	Group	n.s.		n.s.
	Group by valence	350 to 450 ms		450 to 850 ms
		F=3.66; p=.06; η2=.08	F=4.06; p=.05; η2=.09	F=2.89; p=.1; η2=.07;	F=9.3; p=.004; η2=.19
	Group by valence by task half	n.s.		450 to 550 ms
				F=4.4; p=.04; η2=.10	F=5.8; p=.02; η2=.13
	3) 5-HTTLPR/DRD2: SS/SL + Taq1 A vs LL + GG	Fmin(1, 27*)	Fmax(1, 27)	Fmin(1, 27)	Fmax(1, 27)
	Group	n.s.		n.s.
	Group by valence	300 to 450 ms		450 to 650 ms
		F=2.9; p=.1; η2=.10;	F=4.8; p=.04; η2=.15	F=3.08; p=.1; η2=.10	F=3.66; p=.07; η2=.12
	Group by valence by task half	400 to 450 ms		450 to 550 ms
		F=5.4; p=.03; η2=.17		F=5.4; p=.03; η2=.19	F=6.7; p=.02; η2=.20

View Large Image Download to PowerPoint

Fig. 2. ERN time-locked to the response (0ms) at Fz. ERPs are depicted for both the first and second half of the task and correct and incorrect responses: (a) 5-HTTLPR S carriers; (b) homozygous 5-HTTLPR L carriers; (c) DRD2 Taq1 A1 carriers; (d) DRD2 Taq1 A1 non-carriers; (e) carriers of both the 5-HTTLPR S variant and the Taq1 A1 allele of the DRD2 gene; and (f) carriers of neither the 5-HTTLPR S variant nor the Taq1 A1 allele of the DRD2 gene.

View Large Image Download to PowerPoint

Fig. 3. Pe time-locked to the response (0ms) at Pz. ERPs are depicted for both the first and second half of the task and correct and incorrect responses: (a) 5-HTTLPR S carriers; (b) homozygous 5-HTTLPR L carriers; (c) DRD2 Taq1 A1 carriers; (d) DRD2 Taq1 A1 non-carriers; (e) carriers of both the 5-HTTLPR S variant and the Taq1 A1 allele of the DRD2 gene; and (f) carriers of neither the 5-HTTLPR S variant nor the Taq1 A1 allele of the DRD2 gene.

View Large Image Download to PowerPoint

Fig. 4. P3 and LPP time-locked to the feedback stimulus (0ms) at Pz. ERPs are depicted for both the first and second half of the task and positive and negative feedback: (a) 5-HTTLPR S carriers; (b) homozygous 5-HTTLPR L carriers; (c) DRD2 Taq1 A1 carriers; (d) DRD2 Taq1 A1 non-carriers; (e) carriers of both the 5-HTTLPR S variant and the Taq1 A1 allele of the DRD2 gene; (f) carriers of neither the 5-HTTLPR S variant nor the Taq1 A1 allele of the DRD2 gene.

View Large Image Download to PowerPoint

Fig. 5. Mean amplitudes (with standard errors) of the intervals that turned out to contain significant group by task variable effects. (a)–(c) reflect the group interactions found for the response-locked ERN and Pe. (d)–(i) reflect the interactions found for the feedback-locked P3 and LPP.

Before group comparisons on ERP amplitudes were carried out we checked whether the corresponding groups differed in the number of trials included in the (condition-dependent) ERP averages. Across groups the mean number of trials varied from 25 to 35 in the incorrect response condition and from 122 to 144 in the correct response condition of the second task half. For none of the eight conditions (i.e. four response-locked and four feedback-locked conditions) differed the compared groups significantly from each other in their number of trials included.

3.2.1. Response-locked: ERN on Fz 

The existence of an ERN on error trials is reflected by main effects of “response type”, while significant two-way interactions “response type by task half” reflect the expected greater ERN in the second half of the task (see Table 4 and Fig. 2).

Comparison of the 5-HTTLPR-based groups revealed main effects of “group” for 10 successive intervals and three-way interactions “response type by task half by group” for the three successive intervals running from −60 to 0ms (Table 4). Fig. 2(a) and (b) shows that 5-HTTLPR S allele carriers exhibit a greater ERN in especially the second task half. Post hoc comparison of the two groups on the incorrect responses of the second task half revealed significant group differences [t(38)=2.89; p=.006]. Mean amplitude values of these intervals are depicted in Fig. 5(a) demonstrating this three-way interaction.

Comparison of the DRD2-based groups showed that there were neither main effects of group or significant interactions of group with any of the task variables for any of the investigated intervals (Fig. 2(c) and (d)).

Comparison of the 5-HTTLPR/DRD2 combination groups resulted in a significant three-way interaction “response type by task half by group” with a high effect size for the interval running from −20 to 0ms (Table 4). Fig. 2(e) and (f) shows that children carrying both the 5-HTTLPR S allele and the DRD2 Taq1 A1 variant exhibit a greater ERN in especially the second task half [t(29)=2.8; p=.01]. Mean values are presented in Fig. 5(a).

3.2.2. Response-locked potentials: Pe on Pz 

The ANOVA on the task variables revealed significant effects of “response type” as well as significant interactions “response type by task half” (see Table 4). Amplitudes were greater for incorrect than for correct responses with this effect being significantly greater for the second task half (see Fig. 3). This corroborates the presence of a response-dependent Pe.

The only significant group effect was found for the 5-HTTLPR polymorphism. Here we found significant “response type by group” interactions for the four successive intervals running from 200 to 400ms after feedback occurrence (Table 4). Fig. 3(a) and (b) shows that the amplitude difference between correct and incorrect responses is greater for the 5-HTTLPR S allele carriers than for the L allele carriers. This two-way interaction is depicted in Fig. 5(c) presenting the mean amplitudes of the four intervals.

3.2.3. Feedback-locked potentials: P3 and LPP on Pz 

Test of the task manipulations resulted in significant main effects of feedback valence for the five successive P3 and for seven successive LPP intervals. As expected, these effects reflected larger amplitudes for negative feedback stimuli. In general, these feedback effects were smaller during the second task half as is reflected by significant two-way interactions “valence by task half”, for the four LPP intervals running from 500 to 700ms after feedback (Table 4 and Fig. 4).

Comparison of the 5-HTTLPR-based groups showed that these groups differed significantly with respect to task half dependent differences in only their P3 amplitude. This is reflected by three-way interactions “valence by task half by group” for the two P3 intervals running from 350 to 450ms after feedback (Table 5). Fig. 4(a) and (b) shows that, within these P3 intervals, only 5-HTTLPR L carriers demonstrate a decreased response to negative feedback during the second task half. Fig. 5(d) illustrates this three-way interaction on the mean amplitudes of the two intervals. Post hoc comparison on the difference between negative feedback responses during the first and second task half showed nearly significant group differences [t(38)=1.86; p=.07] with medium effect size (η2=.08).

Comparison of the DRD2-based groups resulted in “valence by group” interactions for 10 successive intervals comprising both the P3 (350–450ms) and LPP (450–850ms). Moreover, significant three-way interactions “valence by task half by group” were found for the two intervals within the LPP period running from 450 to 550ms (Table 5). Fig. 4(c) and (d) shows that Taq1 A1 allele carriers demonstrate greater amplitudes in response to negative compared to positive feedback during both task halves while no such prominent difference is seen for the Taq1 A1 non-carriers. In this latter group no feedback valence effect was present for the first task half while for the second task half it was even reversed, with greater amplitudes for positive than for negative feedback. This interaction is reflected by Fig. 5(e). Fig. 4(c) and (d) moreover shows that in contrast to non-carriers, the Taq1 A1 carriers demonstrate a decreased response to positive feedback during the second half of the task. When testing group differences on the children’s responses to only positive feedback, we indeed found (nearly) significant “task half by group” interactions with medium effect sizes for the six successive intervals running from 250ms until 550ms after the feedback stimulus [Fmin(1,40)=3.04; p=.1; η2=.07; Fmax(1,40)=4.27; p=.04; η2=.10]. Mean amplitude values of these intervals are depicted in Fig. 5(f) reflecting this two-way interaction, which was statistically significant [F(1,40)=4.25; p=.046; η2=.10].

Comparison of the 5-HTTLPR/DRD2 combination groups revealed (nearly) significant “valence by group” interactions (with medium and high effect size) for three successive P3 and four successive LPP intervals as well as significant three-way interactions “valence by task half by group” (with high effect sizes) for two successive LPP intervals (see Table 5). In Fig. 4, Fig. 5(g) and (h) we see that the group possessing both the 5-HTTLPR S variant and the DRD2 Taq1 A1 variant exhibits greater feedback valence differences for the P3 and LPP intervals than does the other group. The three-way interactions for the LPP moreover reflect that, different from non-carriers, the children with both the 5-HTTLPR S allele and DRD2 Taq1 A1 variant respond with a decreased potential to positive feedback during the second task half (Fig. 5(h)). Analysis on the mean amplitudes of the same six intervals as computed for the DRD2 groups revealed again a significant interaction between group and task half for only the ERP responses to positive feedback [F(1,29)=4.63; p=.04; η2=.14]. This interaction is depicted in Fig. 5(i).

4. Discussion 

The present study demonstrates that the serotonin transporter gene 5-HTTLPR and the dopamine D2 receptor gene DRD2 differentially affect distinct aspects of error and feedback processing. Whereas children with the short variant of the 5-HTTLPR gene appeared to show greater sensitivity to error processing, children possessing the DRD2 Taq1 A1 allele differed in their sensitivity to both negative and positive feedback, as compared to children who did not possess the respective gene variants. As the groups did not differ in their task performances, the group differences that were seen in the ERPs could not be explained by differences in the number of committed errors (Ullsperger, 2006).

Both the ERN and Pe were found to be significantly more enhanced in the 5-HTTLPR S group than in the LL group, the ERN especially during the second task session. These findings are in line with those by Fallgatter et al. (2004) who also reported on both an enhanced ERN and Pe amplitude in a (smaller) sample of adult S allele carriers compared to a sample of age- and gender-matched homozygous L allele carriers. The ERN has been considered to reflect anterior cingulate cortex (ACC) activity in response to a negative reinforcement signal from the dopaminergic mesencephalon (Holroyd and Coles, 2002), and indeed there are several studies that indicated the ACC as the main source of the ERN (see for a review, Taylor et al., 2007). The serotonergic part in the generation of the ERN might therefore in the first place be explained by the role of the ACC, which has previously been described as a structure that is rich in 5-HT receptors (Haznedar et al., 1997), while moreover ACC metabolic activity has been reported to be normalized in depressive patients through using the serotonin reuptake inhibitor sertraline (Mann et al., 1996).

Variations in the ERN, however, might also be explained by the involvement of cortico-limbic circuits, in which both the ACC and the amygdala play an important role. Similar to the ACC, the amygdala is innervated by serotonergic neurons, and 5-HT receptors are present throughout its sub-nuclei (Azmitia and Gannon, 1986, Smith et al., 1999). A series of independently conducted neuro-imaging studies (functional magnetic resonance imaging [fMRI] and positron emission tomography) on both phobic patients and healthy adults revealed that subjects carrying the 5-HTTLPR S allele exhibited significantly increased amygdala activity when processing aversive stimuli or engaging in anxiety provoking activity such as public speaking (Bertolino et al., 2005, Brown and Hariri, 2006, Heinz et al., 2005). Here it is important to note that Brown and Hariri (2006) were able to show that the S-allele-driven enlarged amygdala responsiveness appeared to be equally pronounced in both sexes and in carriers of one or two S-alleles. Another extensive (f)MRI study conducted by Pezawas et al. (2005) revealed that 5-HTTLPR S-allele carriers showed significantly reduced grey matter volume of both the perigenual ACC and the amygdala, with moreover less structural covariation and a weaker functional connectivity between the amygdala and the ACC, suggesting amygdala hyper-responsivity to be due to weaker inhibitory control by the ACC.

Given the above-mentioned findings of greater amygdala responsiveness to (social) performance-anxiety evoking situations and aversive, especially fear- and anger-expressing stimuli as well as the repeatedly reported increased sensitivity to stress (Caspi et al., 2003, Covault et al., 2007) exhibited by carriers of the 5-HTTLPR S allele, we suggest that an increased ERN as a measure of the individual’s sensitivity to error commission may reflect a predisposition to serotonergically driven (social) performance-anxiety. This suggestion may be further supported by the nearly significant positive correlation (r(52)=.26, p=.06) between the magnitude of the ERN and children’s scores on a scale reflecting internalizing (i.e. anxiety- and depression-related) behavior as measured by the Child Behavior Checklist (Achenbach and Rescorla, 2001), which has previously been found for the group of patients included in the present study (Groen et al., 2008).

Concerning the second response-related somewhat later occurring error positivity, the Pe, we also found an effect of only the 5-HTTLPR polymorphisms, the S allele carriers showing significantly larger amplitudes in response to errors. The Pe has been proposed to be a P3-like response (O’Connell et al., 2007, Overbeek et al., 2005) reflecting phasic changes in locus ceruleus norepinephrine (LC-NE) activity (Nieuwenhuis et al., 2005). Moreover, as, in contrast to the ERN, the Pe was shown to be related to the post-error slowing of response times, it has been suggested to reflect error awareness and subsequent adaptive behavior (Overbeek et al., 2005). Yet, how could LC activity be affected by the serotonin transporter gene? As the association between 5-HTTLPR polymorphisms and amygdala activation has been rather well-established (see for a review, Munafo et al., 2008), here again the amygdala may play a mediating role. Phasic changes in LC activity have been repeatedly observed to be triggered by signals that the LC receives from especially the central nucleus (CeN) of the amygdala, which has been proposed to not only participate in emotional learning but also in attentional, i.e. conscious processing (Bouret et al., 2003, LeDoux, 2007). These CeN-related LC responses are associated with the predictive value or meaning of a stimulus rather than with its physical properties (see for a review, Bouret et al., 2003). The somewhat longer pathway along the LC may therefore explain not only the later occurrence of the Pe as compared to that of the ERN but also its proposed functional meaning, i.e. reflecting error awareness. Possessing the short variant of the 5-HTTLPR gene may hence play an important role in both unconscious error detection and the conscious processing of errors needed for behavior adjustment.

Turning to the feedback-related ERPs, and, first of all, to the feedback P3, we found that the 5-HTTLPR S allele carriers showed no decrease in (negative) feedback dependency, a decrease that has previously been found to accompany increased internal monitoring as is expressed by an increasing ERN with task progression. As, however, these children did show an increase in internal monitoring, we propose that the absence of a decreased feedback P3 with task progression reflects a remaining state of alertness to negative feedback stimuli. We speculate that this may be due to an enhanced sensitivity to negative information or criticism. In combination with greater error sensitivity and error awareness this would agree with the notion that carriers of the 5-HTTLPR S allele have a predisposition to developing (social) performance anxiety.

DRD2 polymorphism dependent variations were especially found for the later occurring and longer lasting LPP complex. Our findings suggest that carrying the Taq1 A1 allele is associated with a generally greater sensitivity to negative feedback, yet at the same time diminishing sensitivity to positive feedback with task progression. Although our findings do not agree with recently published results from a fMRI study (Klein et al., 2007) where a weaker response to negative feedback in DRD2 Taq1 A carriers was found, they do agree with another finding from that study reporting on a reduced reward-related increase in nucleus accumbens (NAc) activity in Taq1 A1 allele carriers, as well as with results from previous studies suggesting the DRD2 Taq1 A1 polymorphism to be involved in the Reward Deficiency syndrome (Balleine et al., 2007, Bowirrat and Oscar-Berman, 2005). The Reward Deficiency syndrome has been referred to as a reduced sensitivity to reward associated with abnormalities in dopaminergically driven cortico-striatal brain regions including the ventral striatum and the NAc. Striatal D2 signaling has been shown to regulate motivational processes in mice (Drew et al., 2007), and the NAc in particular has been proposed to be the dopaminergic structure that is most reliably linked to reward-related processes and alcohol dependence (Bowirrat and Oscar-Berman, 2005). Our findings on the LPP may be in support of the hypothesis that gradually getting insensitive to a regularly offered positive reinforcement, as found here for only the DRD2 Taq1 A1 allele carrying children, may lead to seeking other types of reward in order to keep the neuronal release of dopamine at a level that counteracts the rise of negative feelings (Bowirrat and Oscar-Berman, 2005).

Returning to the common association found for both genes with alcohol dependence (e.g., Feinn et al., 2005, Bowirrat and Oscar-Berman, 2005, Preuss et al., 2007) there hence may be indeed different neurophysiological systems and mechanisms leading to the same behavior: it may arise from the need to reduce anxiety-related feelings (i.e. an overactive BIS) mediated by the S allele of the 5-HTTLPR gene or have a reward and sensation-seeking origin (related to an overactive BAS) that is mediated by the Taq1 A1 allele of the DRD2 gene. The latter may explain the reported liability to other types of drug addiction and gambling as well. Our findings of differential 5-HTTLPR and DRD2 effects on ERPs that are related to the distinct aspects of error and feedback processing as outlined above are quite supportive of this hypothesis.

Still, the mechanisms underlying reward processing are probably more complex than resulting from dopamine release alone. Bowirrat and Oscar-Berman (2005) describe a “reward cascade” involving the release of serotonin that finally leads to a fine tuning of dopamine release by stimulating enkephalin which in turn inhibits the release of γ-aminobutyric acid. The authors therefore pointed to the combined effects of various genes for different neurotransmitters resulting in a final inefficiency of the reward system.

Our findings on the feedback-related ERP differences between the two groups formed on the basis of both the 5-HTTLPR and the DRD2 gene suggest the involvement of both serotonin and dopamine in feedback processing as, different from the response-related ERPs, group by task variable interaction effects appeared to be greater in the comparison of the 5-HTTLPR and DRD2 combination groups than in the comparison of the groups that were formed on the basis of the DRD2 gene alone. The ERP plots presented in Fig. 2, Fig. 4 are suggestive of such combined effects. Yet, as direct comparison of these groups with any of the 5-HTTLPR or DRD2 groups is complicated by overlap in participants a straightforward conclusion about additive or interactive effects of the two polymorphisms cannot be drawn.

Other study limitations need to be acknowledged. First, although many differences reached statistical significance, our sample size was relatively small, not allowing for the study of the recently suggested tri-allelic 5-HTTLPR genotypes (Hu et al., 2006) or to form groups large enough in order to statistically test interactive effects between the 5-HTTLPR and DRD2 variants. A second weakness may have been the use of a somewhat heterogeneous study sample consisting of children with ADHD or PDD along with healthy controls. To control for this heterogeneity, however, we succeeded in optimal matching of groups with regard to both genotype and clinical status. With respect to gender, however, there were three more girls in the 5-HTTLPR LL group than in the S group. As previously an interaction effect of age and gender on the ERN has been found, with girls showing a smaller amplitude at the age of 10 but amplitudes similar to boys at the age of 11 and 12 (Davies et al., 2004), we also tested for interaction effects of gender by response type on amplitudes in the ERN (Fz) intervals. We did so on the whole group of 65 children (51 boys and 14 girls) participating in this study and indeed found significant interactions (p<.05) for the intervals running from −100ms before to 20ms after the response. These, however, showed greater negativities to incorrect responses for the girls. The smaller ERN found for the 5-HTTLPR LL group is therefore unlikely to be caused by the three more girls who, moreover, had a mean age of 11.8 (SD=0.68) years, which was slightly higher than that of the three girls in the 5-HTTLPR S group (M=11.1; SD=.41).

One finally might question whether the same results were obtained in a sample of adults, as the structures involved in the generation of the ERP components investigated may not yet have been fully matured. Next to the fact that genetic profiles are invariant there are two arguments for assuming comparability. (1) Our findings on the 5-HTTLPR polymorphisms agree with those on the ERN and Pe of an adult study conducted by Fallgatter et al. (2004), and (2) all components investigated have previously been found in adult studies and shown to be sensitive to the same type of task manipulations in healthy children (Groen et al., 2007) of the same ages as investigated in the present study.

In conclusion, the present study points to differential effects of common polymorphisms of the 5-HTTLPR and DRD2 genes on reinforcement-related learning, with 5-HTTLPR S carriers having increased sensitivity to error processing, and DRD2 Taq1 A1 carriers exhibiting greater sensitivity to negative feedback and task progression dependent decreasing sensitivity to positive feedback. Our findings are in line with what has repeatedly been suggested in the literature, i.e. the 5-HTTLPR S allele contributing to a predisposition for anxiety-related behavior and the DRD2 Taq1 A1 allele to a predisposition for the reward deficiency syndrome.

Disclosure/conflict of interest 

Dr. Hoekstra has been a consultant to and/or speaker for Eli Lilly and has received research support from Eli Lilly. Dr. Minderaa is a paid consultant to Eli Lilly. These financial ties are not related to the present paper.