J. Hoenicka, Servicio de Psiquiatría, Hospital Universitario 12 de Octubre, AV de Córdoba SN, Madrid 28041, Spain. E-mail:firstname.lastname@example.org
Polymorphisms of DRD2 and ANKK1 have been associated with psychiatric syndromes where there is believed to be an underlying learning process deficit such as addiction, post-traumatic stress disorder and psychopathy. We investigated the effects of the DRD2 C957T and ANKK1 TaqIA single nucleotide polymorphism (SNP), which have been associated with psychopathic traits in alcoholic patients, on fear conditioning and aversive priming in healthy volunteers. We found that the DRD2 C957T SNP, but not the ANKK1 TaqIA SNP, was associated with both differential conditioning of the skin conductance response and the aversive priming effect. There were no differences between the genotype groups with respect to the extinction of the skin-conductance conditioned response. These results suggest that the C957T SNP could be related to learning differences associated with the risk of developing psychiatric disorders in individuals that are carriers of the C homozygous genotype. Our genetic data raise the possibility that the dopaminergic system functional variations determined by this SNP could affect fear learning.
The study of biological and environmental influences on psychopathic personality has indicated that individual differences in antisocial behaviour are highly heritable (Fontaine et al. 2008; Viding et al. 2005). In clinical samples, the clearest link between genetic variation and psychopathic traits has been established with genes of the locus 11q22-q23 (Ponce et al. 2008). The most recent data show that the single nucleotide polymorphisms (SNPs) of the D2 dopamine receptor gene (DRD2) and the ankyrin repeat and kinase domain 1 (ANKK1), C957T (rs6277) and TaqIA (rs1800497), respectively, are epistatically associated with psychopathic traits in alcoholic patients (Ponce et al. 2008). The DRD2 gene (OMIM # 126450) encodes a G protein-coupled receptor which plays a central role in reward and learning mechanisms (Hoenicka et al. 2007), and its C957T SNP has also been associated with schizophrenia (Hoenicka et al. 2006; Lawford et al. 2005) and post-traumatic stress disorder (PTSD) (Voisey et al. 2009). The ANKK1 gene (OMIM # 608774), which is adjacent to DRD2, codes for a novel predicted kinase (Neville et al. 2004). The ANKK1 TaqIA SNP, which is also a DRD2 gene marker, has been widely associated with addiction, especially with severe and antisocial alcoholism (Noble 2000; Ponce et al. 2003), thus suggesting a role for this SNP in reward and learning systems. These DRD2 and ANKK1 gene variants have also been studied in different learning paradigms. In healthy volunteers, C957T has been linked to reinforcement learning (Frank et al. 2007). TaqIA has been associated with high reward sensitivity (Lee et al. 2007), the capacity of learning from errors (Klein et al. 2007), a greater sensitivity to negative feedback in neuropsychological tasks (Althaus et al. 2009) and reversal learning (Jocham et al. 2009).
Psychopathy has been associated with significant impairment in several types of learning. Individuals with psychopathy show impaired stimulus-reinforcement learning in the context of aversive conditioning (Birbaumer et al. 2005; Flor et al. 2002). However, Ishikawa et al. (2001) reported increased cardiovascular reactivity in stressed ‘successful psychopaths' and they suggest that this higher response could help to prevent risk behaviours. Thus, psychopathy could essentially entail a specific deficit in emotional response to the harm suffered by others and not a general hyporeactivity to threat.
These data, together with the molecular findings at locus 11q22-q23 in clinical and control populations, suggest that some aversive learning paradigms could be endophenotypes linked to DRD2 and ANKK1 gene variants. For example, fear conditioning and aversive priming are potential endophenotypes that could be involved in the pathophysiology of certain psychiatric disorders associated with learning abnormalities, such as psychopathy and PTSD. Fear conditioning is a basic form of associative learning involved in the emotional response to threats, and aversive priming is a bias favouring identification of threats after an aggression. Because genetic, neuroimaging and clinical evidence indicates that ANKK1 and DRD2 polymorphisms may play an important role in affective information processing, reactivity to stress and fear learning, we investigated the relationships between both DRD2 C957T and ANKK1 TaqIA, and psychophysiological and behavioural indices of fear to examine whether these SNPs are related to fear conditioning and aversive priming.
Sixty-three healthy undergraduate students (32 females) at the Universidad Complutense de Madrid, Madrid, Spain, were recruited for the study. Their age range was 19–27 years and they received €20 as a monetary incentive. Data from one participant were excluded because the genetic analysis could not be completed, and data from a further two participants were excluded because of extreme values on the recognition time task. The study was approved by the Ethics Committee of Hospital 12 de Octubre, Madrid. Participants signed two informed consent forms, one for the experimental procedure and one for the genetic study.
Aversive learning experiment
The experimental study was carried out in the Faculty of Psychology of the Universidad Complutense de Madrid. The participant was seated in a sound-attenuated and tenuously illuminated experimental room that was connected to an adjacent control room.
During the conditioning phase, four pictures of neutral black-and-white faces (2 male, 2 female) from the Eckman and Friesen set (numbers 21, 41, 65 and 99) were presented to each participant. Half of the participants were presented the two male faces as conditioned stimuli (CS + and CS −) and the two female faces served as additional faces to familiarize the participant with them. For the remaining participants, the faces were presented in the opposite manner. The face that was coupled to the CS + was also balanced among participants. During the priming phase, these four faces were used as those to be recognized (old faces). In addition, 10 neutral faces from the Eckman and Friesen set (numbers 6, 13, 28, 33, 47, 56, 72, 83, 92 and 110) were used as distractors (distractor-faces). The faces were projected from the control room on a screen located about 110 cm in front of the participant. The image size was 22 × 30 cm.
Electric shocks were generated by a Mark 100 stimulator (Farrall Instruments, Inc., Grand Island, NE, USA) fed by batteries and optically isolated from the computer. The electric shocks were applied to the internal surface of the forearm through a Tursky concentric electrode. The intensity of the shock was fixed by the participant following the instruction ‘it must be clearly uncomfortable but not painful’. During the conditioning phase, shocks were applied to the dominant forearm for 200 milliseconds, and during the priming phase they were applied to the non-dominant forearm for 180 milliseconds.
Neutral tones were 1000 Hz and the intensity was set by the participant following the instruction ‘you must hear it clearly, but it must not be unpleasant at all’. The tones were presented for 200 milliseconds during the conditioning phase and for 180 milliseconds during the priming phase.
Psychophysiological and behavioural recordings
The skin conductance response (SCR) was recorded using two Ag–AgCl Beckman electrodes (8-mm diameter), filled with a 0.05 m NaCl gel. Electrodes were fastened by adhesive collars to the palmar surface of the second phalanges of the first and second fingers of the non-dominant hand. The signal was amplified by an S71–22 bioamplifier (Coulbourn Instruments, Whitehall, PA, USA) and sampled at 50 Hz. The SCR was scored as the maximum increase in conductance between the first and the fourth second after the onset of the stimulus. The SCR value was considered zero if changes were lower than 0.01 μS or no increase was observed. The reaction time (RT) was registered by a locally built console with two buttons, marked ‘yes’ and ‘no’.
An IBM-compatible personal computer was used to monitor the presentation of stimuli, and to acquire and store the SCR and RT data. It was equipped with a PCL-812 PG Multi-Lab card and a PCLD-785 Relay Output Board card (both from Advantech Co., Cincinnati, OH, USA). The software was developed by the Technical Service of the Faculty of Psychology of the Universidad Complutense de Madrid.
The experiment had four phases: presentation of distractors, habituation, conditioning and aversive priming.
Upon arrival, the participants were reminded that they could leave the experiment at any time. They were fitted with skin conductance electrodes, and then they viewed each of the 10 faces that would operate as distractors in the priming phase two times. Each picture was presented for 8 seconds and the inter-trial interval (ITI) was 4 seconds. Next, the shock electrodes were fitted, and the intensity of the shock and the tone were set. During the breaks in the experiment, participants were asked again about the unpleasantness of the stimuli and the intensity was modified when necessary. The participant could also ask to change the intensity at any time.
Next, the participants were informed that they would view four new faces several times each, and that one of the faces would sometimes be paired with the shock (CS +) and the other would sometimes be paired with the tone (CS −). During the habituation phase, the procedure included a non-reinforced presentation of each of the CS +, CS − and the two other old faces. During the acquisition phase, the procedure included eight presentations of the CS + paired with the aversive shock and eight presentations of the CS − paired with the neutral tone (paired trials), three presentations of each of the CS + and the CS− alone (test trials), and four presentations of each of the other two old faces.
The set habituation-conditioning was organized in six blocks with a total of 34 trials and a single break after the 20th trial. The first block was the habituation phase. The second block consisted of two paired trials of the CS + and two paired trials of the CS −. Blocks 3, 4 and 5 consisted of two paired trials of each of the CS + and the CS −, one test trial of each of the CS + and the CS−, and one presentation of each of the other two faces. The sixth block consisted of one presentation of each of these last two faces. Within the blocks, the trials of each type were presented in semi-random order, with the restriction that none of the faces appeared in more than three consecutive trials. The presentation time for the faces was 8 seconds, the interval between the appearance of the face and the onset of shock or tone [stimulus onset asynchrony (SOA)] was 3 seconds and the duration of the shock and the tone was 200 milliseconds. The ITI varied randomly between 16 and 20 seconds.
Next, the aversive priming (Huertas-Rodríguez 1991) phase began. Participants were informed that they would see several faces and should indicate ‘as quickly as possible, but trying not to make mistakes', if the face was one of the four faces presented during the conditioning phase or one of the faces seen at the beginning of the experiment. It was therefore a recognition task, and not only a familiarity-judgement task. Participants were warned that any face could be preceded by the neutral tone, the aversive shock or neither of them. The recognition decision was made by pressing the corresponding buttons on the console. By pressing the button, the face disappeared. When the participant did not press any button in a period of 3 seconds after the face onset, it disappeared automatically. The SOA was 190 milliseconds. The duration of the prime (shock or tone) was 180 milliseconds. The ITI varied randomly between 19 and 21 seconds.
Participants underwent two presentations of the CS + and two of the CS −, seven presentations of the other two old faces and 10 presentations of the distractors. The CSs were preceded once by the aversive prime, and once by the neutral one. The rest of the faces were preceded four times by the aversive prime, seven times by the neutral prime and six times by neither of them. There was one presentation of each prime–CS sequence, because the aversive-prime/CS − sequence and neutral-prime/CS + sequence could operate as counterconditioning trials, thus reducing or eliminating the expected effect in subsequent presentations. After trial 14, a break of 4 min was taken. The four critical prime/CS sequences were presented in trials 9, 13, 17 and 20. There were four presentation orders for these sequences, according to the eight possible permutations that fulfill the requirement for an alternate conditioned stimulus (CS + and CS −) and prime (aversive and neutral) presentations. The presentation order for each participant was random.
The priming phase finished with a presentation of each CS + and CS − without being preceded by shock or tone. The order of presentation was counterbalanced between the participants.
Samples from epithelial cheek cells were collected with buccal swabs. Genomic DNA extraction and amplification were performed using the illustra GenomiPhi V2 DNA Amplification Kit (GE Healthcare Europe, Barcelona, Spain) following the manufacturer's protocol. Genotyping was performed by Taqman assays, which were designed to run on an ABI 7900HT machine with Sequence Detection System software (Applied Biosystems, Foster City, CA, USA). In a previous work, we evaluated the non-random association of the two SNPs and found a low degree of linkage disequilibrium between TaqIA and C957T in the control population (D′ = 0.58, r2 = 0.14) (Ponce et al. 2008).
The sequence-specific primers for ANKK1 TaqIA (5′CTGCCTCGAC CAGC 3′ and 5′CTGCCTTGACCAGC 3′) were designed for the C (A2) and T (A1) alleles, respectively. A common reverse primer (5′-GCAACACAGCCATCCTCAAAG-3′) was used. Five samples for each genotype were confirmed by direct sequencing analysis. The resulting genotypes for ANKK1 TaqIA were clustered according to the presence of at least one A1 allele (A 1+ genotype, A1 allele homozygous and heterozygous; A1− genotype, homozygous for the A2 allele) as previously described (Blum et al. 1990). DRD2 C957T was analysed as previously described by Hoenicka et al. (2006). Briefly, the sequence-specific primers for the Taqman assays (5′-CTGTCGGGAGTGCTG-3′ and 5′-CTGTCAGGAGTGCTG-3) were used for the C and T alleles, respectively, as was the common reverse primer 5′-GCCCATTCTTCTCTGGTTTGG-3′. The genotypes obtained were grouped assuming a recessive model for the C957 allele: homozygous individuals for the C allele vs. heterozygous and homozygous individuals for the T allele.
We performed the Hardy–Weinberg equilibrium test using the Haploview software (version 3.2; Whitehead Institute for Biomedical Research; http://www.broad.mit.edu/mpg/haploview/index.php) and we found no deviation from either in ANKK1 TaqIA (P = 0.44) or in DRD2 C957T (P = 0.83). In addition, we observed in this sample the same ANKK1 and DRD2 genotype distributions that were previously found for both Spanish (Ponce et al. 2008) and European (NCBI: http://www.ncbi.nlm.nih.gov/) healthy populations.
All SCR magnitudes were range corrected, by dividing each value by the mean of the participant's three maximal responses and multiplying the result by 100 (PC SCR). A square-root transformation was also applied to help to normalize the distribution [SQRT (1 + PC SCR)]. In the case of the RTs, data were log transformed. However, means are presented as milliseconds to allow a better interpretation. The incorrect recognition responses (4.6% of total data) and RT values over the grand mean ± 2.5 SD (1.7% of total data) were replaced by the mean of the participant.
A repeated-measures general linear model (SPSS 15.0 for Windows, SPSS Inc., Chicago, IL, USA) was used for each SNP, with SCR and RT as consecutive dependent variables. We used a two-tailed t-test for post hoc comparisons between genotypes, and a one-tailed t-test for the comparisons for which we had a specific prediction a priori. Given the unequal size of the genetically defined groups (C957T SNP: CC group = 9, CT /TT group = 51; TaqIA SNP : A 1− = 18, A 1+ = 42), additional Mann–Whitney tests were used for post hoc comparisons between genotypes. This analysis gave rise to an identical pattern of results to that of the t-test.
Relationship between the C957T and TaqIA SNPs, and fear conditioning
The SCR analysis showed a significant CS×test-trial× C957T interaction (F 2.116 = 3.154; P = 0.046). This interaction effect was subsequently parsed into orthogonal linear and quadratic contrasts to evaluate the effect of the test-trial on the differential response to the CS + and the CS − depending on the C957T genotype. Only the CS × test-trial × genotype linear interaction achieved significance (F 1.58 = 5.687; P = 0.020) (Fig. 1b). Next, we analysed the test-trial × genotype linear interaction for the CS + and CS − separately. Only the interaction for the CS + was significant (F 1.58 = 8.401; P = 0.005; partial η2 = 0.127). The CT/TT group exhibited a lower SCR in the third trial when compared with the first trial (t = 2.975; d f = 50; P = 0.003), whereas the SCR in the CC group not only failed to show a decrement but rather tended to increase (P = 0.073). Thus, differential conditioning of SCR was higher in the CC carriers during the acquisition phase. In these individuals, the response to CS + did not decrease during the conditioning process. Moreover, this divergence between the genotype groups was not because of a difference in the differential SCR during the habituation phase (P = 0.490), or a difference in the SCR to the unconditioned stimulus (P = 0.794), or a difference in electrodermal responsivity (P = 0.453).
With respect to the TaqIA polymorphism, the repeated measures ANOVA did not show a CS × test-trial ×genotype interaction (P = 0.688) (Fig. 1c).
The CS × genotype interaction analysis showed no significant differences between groups in the extinction of the SCR for the C957T SNP (P = 0.782) (Fig. 1b) or for the TaqI-A (P = 0.147) (Fig. 1c).
Relationship between the C957T and TaqI-A SNPs and aversive priming
We first confirmed an overall aversive priming effect. When all the data were taken into account, there was a significant prime × CS interaction (F 1.59 = 7.867; P = 0.007). The RT in the aversive-prime/CS + condition was shorter when compared with each of the other three conditions (P < 0.001).
When the C957T genotype was included in the model, a significant prime × CS × C957T interaction pattern was identified (F 1.58 = 4.337; P = 0.042) (Fig. 2b). In the aversive-prime/CS + condition, but not in the other three conditions, the CC carriers showed a shorter RT than the CT/TT group (t = 2.392, df = 58, P = 0.02; Cohen ′s d = 0.879). Thus, after the differential conditioning process, the presence of an aversive primer facilitates CS + recognition, and this effect is more prominent in the CC carriers.
When these analyses were performed by grouping according to the TaqIA SNP genotype, we did not find a significant prime × CS ×TaqIA interaction (P = 0.538) (Fig. 2c).
We found that DRD2 C957T, but not ANKK1 TaqIA, was associated with both SCR differential conditioning and aversive priming in healthy participants. Specifically, during acquisition, the CT/TT group exhibited a continued decrease in the SCR to the CS + with repeated presentation of the stimulus, whereas the response of the C homozygous carriers was maintained or tended to increase. On the contrary, when we examined the extinction of the CS + response, we found no differences between the genotype groups. Thus, the higher physiological response did not seem to be associated with a learning impairment; instead, it seems that the C homozygous carriers did not adapt to the CS + exposure, as long as this stimulus represented a threat, thus producing a sustained emotional response.
With respect to aversive priming, that has a clear similarity to cognitive biases associated with anxiety, our paradigm assesses the speed of selective attention towards the threatening stimulus. After the differential conditioning process, the C homozygous group had a shorter recognition time in the aversive-prime/CS + condition, thus reflecting that either the CS + had become more threatening, or that priming is facilitated.
Fear conditioning and aversive priming usually have an adaptive function. Normal fear to a threat warns the person of danger and the aversive priming facilitates a fast and accurate perception of real threats. However, excessively intense fears could have maladaptive consequences and enhanced cognitive-affective biases can increase the likelihood of processing an unreal threat that in turn would increase anxiety. Therefore, our study may also have clinical implications. Indeed, our results are in line with data and theories pertaining to psychiatric disorders associated to the C957 homozygous genotype such as the PTSD (Voisey et al. 2009), schizophrenia (Hoenicka et al. 2006; Lawford et al. 2005) and substance abuse (Perkins et al. 2008). For instance, individuals with PTSD seem to be more conditionable than trauma-exposed individuals without PTSD (Orr et al. 2000). Further, several studies suggest that there is an increased likelihood of flashbacks and intrusive phenomena related to the traumatic experience in these patients as a consequence of a subsequent increase in arousal. This retroactive effect occurs even if the arousal is artificially triggered [e.g. (Bremner et al. 1997; Nixon and Bryant 2005)]. A similar phenomenon has been found in schizophrenia where stressful situations raise the likelihood of hallucinations (Freeman and Garety 2003). In the case of drug abuse, situations leading to the increment of the arousal also increase the likelihood of images and thoughts related to the substance of abuse (Sinha 2008). Indeed, in the case of nicotine addiction, the smokers homozygous for the C957 allele took more cigarette puffs during negative vs. positive mood (Perkins et al. 2008).
On the contrary, our results in the C957 homozygous, a risk genotype for psychopathic traits (Ponce et al. 2008), seem not to be consistent with one of the most replicated findings in criminal psychopaths, i.e. electrodermal hyporeactivity in the anticipation of aversive stimuli (Fowles 2000). However, we must remember that our study was performed in healthy volunteers. Heightened autonomic stress reactivity has also been observed in successful psychopaths when compared wiht both unsuccessful psychopaths and controls (Ishikawa et al. 2001). On the other hand, in our previous work on alcoholic patients, the C homozygous genotype was associated with psychopathy exclusively when the ANKK1 TaqIA A1 allele was present. Thus, the effect of DRD2 C957T effect on learning processes may vary according to the biological background. Poor modulation of the electrodermal response has been related to a deficit in inhibitory control, a clinical trait that may be linked to a variety of psychiatric disorders marked by impulsivity or disinhibition (Taylor et al. 1999). Thus, the C homozygous genotype, which was linked to an autonomic over-reaction in this study, could be related to lesser control and greater reactive aggressiveness, which are characteristic of addictive and antisocial personality disorders.
An increased conditioned behavioural response to threats during conditioning acquisition has also been observed in mice when amphetamine was withdrawn (Pezze et al. 2002). This heightened response may correlate with differences in the functioning of the nucleus accumbens (NAc), which is involved not only in reward learning but also in aversive conditioning (Pezze et al. 2001). Although aversive conditioning has mostly been related to functioning of the amygdala and prefrontal cortex, it has been showed that the NAc also has a relevant role. Elevated dopamine levels have been observed in the NAc in response to aversive outcomes (Kalivas and Duffy 1995; Robinson et al. 1987; Young 2004), CSs that are predictive of such outcomes or exposure to the conditioning context (Pezze et al. 2002; Young 2004; Young et al. 1993, 1998). Dopamine signalling in the amygdala and prefrontal cortex is mediated mainly by the D1 dopamine receptor, whereas D2 is highly expressed in the striatum. The C957T SNP has been associated with mRNA stability and protein synthesis in vitro (Duan et al. 2003), as well as with an increased D2 receptor density in several brain areas (Hirvonen et al. 2009). Furthermore, in mice, overexpression of the D2 receptor in the striatum has been found to affect dopamine functioning in the prefrontal cortex, thus leading to working memory deficits (Kellendonk et al. 2006). Therefore, learning differences related to cortical function may be a reflection of a primary striatal variation.
The limitations of this study are related to the nature of association studies; therefore, our results must be interpreted with caution. The strength of our results is tempered by a relatively small sample size, especially in some genotype subgroups. Thus, to validate our results, it would be desirable to replicate our genetic studies in independent samples. On the other hand, our approach to measure the evidence of extinction is somewhat atypical, because it involved the last two trials of the priming phase and the need for the participant to press the button on the recognition test. Extinction tests generally involve presentation of the CS to the participant with no additional tasks that would increase the SCR to both the CS + and the CS−. Despite these limitations, we believe that our study has identified a genetic variant of the dopaminergic system that is related to aversive learning, and that this variant deserves greater attention. Future research should clarify to what extent this increased differential SCR was because of a lower inhibitory control of emotional response to danger rather than to fear-enhanced learning.
We cannot disregard the potential role of ANKK1 TaqIA in other learning processes related to psychopathic traits. This polymorphism has been associated with reward-related personality traits (Lee et al. 2007) and different learning paradigms (Althaus et al. 2009; Klein et al. 2007). In any case, the study of other aversive learning paradigms would shed light on the role, if any, of the epistatic association between C957T and TaqIA in learning impairment in psychopathy.
In summary, our results suggest that functional variations in D2 signalling determined by the C957T SNP could be related to learning differences associated with the risk of developing psychiatric disorders in carriers of the C homozygous genotype.
This work has been supported by the ‘Fondo para Investigaciones Sanitarias' (FIS), Madrid, Spain, grants no. 05/0731 and 08/0529. CIBERSAM is an initiative of the Instituto de Salud Carlos III.