Background The subthalamic nucleus (STN) is an efficient target for treating patients with Parkinson's disease as well as patients with obsessive–compulsive disorder (OCD) using high frequency stimulation (HFS). In both Parkinson's disease and OCD patients, STN-HFS can trigger abnormal behaviours, such as hypomania and impulsivity.
Methods To investigate if this structure processes emotional information, and whether it depends on motor demands, we recorded subthalamic local field potentials in 16 patients with Parkinson's disease using deep brain stimulation electrodes. Recordings were made with and without dopaminergic treatment while patients performed an emotional categorisation paradigm in which the response varied according to stimulus valence (pleasant, unpleasant and neutral) and to the instruction given (motor, non-motor and passive).
Results Pleasant, unpleasant and neutral stimuli evoked an event related potential (ERP). Without dopamine medication, ERP amplitudes were significantly larger for unpleasant compared with neutral pictures, whatever the response triggered by the stimuli; and the magnitude of this effect was maximal in the ventral part of the STN. No significant difference in ERP amplitude was observed for pleasant pictures. With dopamine medication, ERP amplitudes were significantly increased for pleasant compared with neutral pictures whatever the response triggered by the stimuli, while ERP amplitudes to unpleasant pictures were not modified.
Conclusions These results demonstrate that the ventral part of the STN processes the emotional valence of stimuli independently of the motor context and that dopamine enhances processing of pleasant information. These findings confirm the specific involvement of the STN in emotional processes in human, which may underlie the behavioural changes observed in patients with deep brain stimulation.
- Parkinson's Disease
- Behavioural Disorder
- Cognitive Electrophysiology
- Evoked Potentials
- Psychology, Experimental
Statistics from Altmetric.com
- Parkinson's Disease
- Behavioural Disorder
- Cognitive Electrophysiology
- Evoked Potentials
- Psychology, Experimental
High frequency stimulation (HFS) of the subthalamic nucleus (STN) alleviates motor disabilities in Parkinson's disease (PD).1 Apart from its motor effects, STN-HFS has been shown to provoke behavioural disorders such as hypomania and impulsivity, as well as deficits in facial emotion recognition2 ,3 and modification of subjective emotional experience,4 suggesting a role in emotional information processing. Interestingly, these side effects of STN-HFS have been reported not only in PD5 ,6 but also in patients with obsessive–compulsive disorder (OCD) treated successfully with STN-HFS.7 Importantly, similar behavioural modifications have been observed following STN lesion/inactivation in normal rats and monkeys.8 ,9 Thus these side effects seem to result from modifications of STN functioning induced by STN-HFS, independently of the pathology considered.
In the basal ganglia system, the STN has a crucial position as it receives input from other basal ganglia nuclei but also direct projections from cortical areas.10 ,11 Anatomical studies in primates have demonstrated that the STN can be segregated into sensorimotor, associative and limbic territories according to the topography of external pallidal inputs.12 In monkeys and also in rodents, the STN limbic territory is located in the anterior and ventral part of the nucleus,12 which has reciprocal connections with the limbic ventral pallidum.13–15
Electrophysiological studies in animals have shown that STN neuronal activity is modified in relation to motor but also emotional information, such as reward prediction and obtainment.16–18 In humans, electrophysiological studies have shown modification of STN activity in relation to motor preparation and execution,19–21 but also in relation to behaviourally relevant22 ,23 and emotional stimuli, with larger event related desynchronisation for emotional compared with neutral stimuli.24
These observations suggest that the STN processes motor, cognitive and limbic information. However, the interaction between the coding of emotional and motor information has not been specifically assessed. To study how the STN encodes emotional information in relation to the motor programmes they may trigger, we recorded local field potential (LFP) of 16 PD patients implanted in the STN while they were performing an emotional categorisation task in which motor responses depended on the emotional valence of each stimulus. STN-LFPs were analysed in relation to the precise anatomical location of the recording contacts, computed using a three-dimensional atlas of the basal ganglia.25 ,26 As dopaminergic replacement therapy (DRT) has been shown to modify STN-LFP in PD patients,27 ,28 the same paradigm was performed with and without treatment.
Materials and methods
Sixteen patients with idiopathic PD implanted in the STN participated to the study with informed consent. The protocol was approved by the local ethics committee for biomedical research and supported by INSERM (RBM C07-42). Inclusion criteria were those used for deep brain stimulation29: severe form of the disease (mean Unified Parkinson's Disease Rating Scale III off treatment±SD=27±14, mean disease duration±SD=11±4 years), excellent response to dopaminergic medication (mean improvement in Unified Parkinson's Disease Rating Scale III on treatment=79±17%, mean levodopa equivalent dosage=992±346 mg/day), absence of cognitive decline (mean Mattis Dementia Rating Scale=141±3/144, mean Frontal score=54±3/60), absence of psychiatric problems at the time of surgery, normal brain MRI and absence of surgical contraindications. Patients were also assessed using the Hospital Anxiety Depression Scale (mean±SD item anxiety=6±4/21; item depression=5±3/2130) and the State–Trait Anxiety Inventory (STAI) (mean±SD STAI-state=34±11/80; STAI-trait=40±12/8031).
Stimuli were complex pictures selected from the International Affective Picture System32 and via the Internet. The valence and arousal of stimuli were assessed in 80 control subjects (20–70 years old; mean age 40±15 years) using the Self-Assessment Manikin Scale.33 The pictures were separated into pleasant (mean±SD: valence=6.6±1.3; arousal=4.1±2.2), unpleasant (valence=2.5±1.2; arousal=5.6±2.0) and neutral (valence=50±0.9; arousal=1.7±1.2). The luminance and spatial frequency34 of each stimulus were calculated in Matlab.
The experiment was composed of unpleasant and pleasant tasks. In each task, emotional or neutral pictures were presented for 1 s, followed by a question mark initiating the response phase (figure 1). Each task comprised two active (A and B) and one passive (C) condition. In condition A, the patient had to evaluate the unpleasantness/pleasantness of the pictures according to their own feelings and indicate a unpleasant/pleasant picture by pressing a button during the response phase. In condition B, the rules were inverted and patients had to indicate a neutral picture. In condition C, patients had to look at the pictures passively. This design generated ‘emotional motor’ and ‘neutral non-motor’ trials (A), ‘emotional non-motor’ and ‘neutral motor’ trials (B), and ‘emotional passive’ and ‘neutral passive’ trials (C). Non-motor and passive trials differed as the valence of the stimuli had to be evaluated in non-motor trials but not in passive ones. Blocks containing 50% neutral and 50% emotional stimuli (either all pleasant or unpleasant) were assigned to each condition, for a total of 50 trials per valence and condition, with equal mean luminance and mean spatial frequency content between blocks. Each picture was presented only once during the whole experiment. The sequence of pictures within blocks, the association between block and condition, and the order of conditions were randomised between patients.
Due to fatigue and the postoperative state, all patients could not perform all of the tasks in all conditions. All patients included in the study (n=16) performed the unpleasant task off treatment; 13 patients also performed the pleasant task off treatment. Seven of the patients included then performed the unpleasant task on treatment and five of them also performed the pleasant task on treatment. Off treatment evaluations were made 4 days after surgery in the morning, after overnight withdrawal of DRT. On treatment evaluations were performed the day before, 1 h after the intake of the patient's usual morning dose. For the evaluations, patients were seated in front of a 15 inch screen displaying the stimuli and responded with their right hand.
LFP were recorded bilaterally using chronic stimulating electrodes (model 3389, Medtronic Neurological Division, Minneapolis, Minnesota, USA), which have four platinum–iridium cylindrical contacts (1.27 mm in diameter and 1.5 mm in length) separated by 0.5 mm, contact 0 being the most ventral and contact 3 the most dorsal. Bipolar recordings were made between adjacent contacts of the electrode with three channels per electrode (0-1, 1-2, 2-3). Signals were amplified, low pass filtered at 250 Hz and digitised at 512 Hz using an EEG amplifier (Digital EEG-EP Multifunction System, EBNeuro, Italy) and monitored online using Galileo software (EBNeuro).
A trial was considered to be correct if the patient attributed to the picture the emotional valence previously defined and responded at the appropriate time (during the question mark). Reaction times (RTs) and performances were compared across conditions using repeated measures one way ANOVA.
Signal analysis and statistics
Signal processing, visualisation and statistical analyses were performed using Matlab (MathWorks) and software developed locally (http://cogimage.dsi.cnrs.fr/logiciels/). Continuous bipolar recordings were divided into trials starting 1 s before to 3 s after stimulus onset. Signals were corrected for their mean baseline amplitude (−700 ms to 0) and trials with artefacts or incorrect behaviour were rejected. Correct artefact free trials were averaged time locked on stimulus onset, low pass filtered at 20 Hz and high pass filtered at 0.2 Hz to remove DC offset. Changes in LFP activity evoked by the onset of the stimuli event related potential (ERP) were considered significant if the amplitude deviated by more than 3 SD from the prestimulus period over more than 100 ms. We defined ERP latency as the time at which this deviation started. We selected channels with ERP and evaluated the main effects of stimulus valence (emotional vs neutral trials) and motor context (motor vs non-motor vs passive trials) using two different analyses. (1) A group analysis was performed to evaluate stimulus valence, motor context and interaction effects on the normalised area under the curve (NAUCs) of the ERP across subjects. The AUC was computed from the ERP on a (0.1 s) time window and normalised (NAUC) per each channel by dividing AUC for each condition by the mean AUC across all conditions of the same task. NAUCs were averaged across all channels for each subject, and mean values were compared by repeated measures two way ANOVAs. Paired tests were corrected using Bonferroni's method. (2) An individual analysis was performed to evaluate the co-occurrence of emotional valence and motor context effects on each channel and to assess the effects of the localisation on STN activity. Repeated measures two way ANOVAs were performed for each channel and time sample and the statistical significance of the observed effects was assessed using non-parametric randomisation tests (see supplementary material, available online only). Effects were considered significant when we observed a p value<0.05 on a time window of at least 25 ms.
To perform the correlation between the emotional effect and the dorsoventral coordinates, emotional effects were quantified as the difference between the NAUCs of emotional and neutral ERPs.
Localisation of recordings contacts
The location of the electrodes and contacts within the STN was obtained using an adjustable three-dimensional histological atlas of the basal ganglia in which coordinates are measured relative to the posterior commissural point25 ,26 (see supplementary material, available online only). As we performed bipolar recordings between two adjacent contacts on the electrode, the coordinates of the halfway point between the two recording contacts were used for anatomical localisation.
Accuracy and RT in the different tasks and conditions are reported in table 1. The patients responded with the same degree of accuracy in unpleasant, pleasant and neutral trials (ANOVA, F(3,36)=1.80, p=0.16), and at the same RT (ANOVA, F(3,36)=0.33, p=0.80). In patients tested both on and off treatment, treatment had no effect on patient accuracy in the pleasant or unpleasant task (unpleasant task, ANOVA, stimulus valence: F(1,6)=0.01, p=0.93, treatment: F(1,6)=0.06, p=0.81; pleasant task, ANOVA, stimulus valence: F(1,4)=0.24, p=0.65, treatment: F(1,4)=0.67, p=0.46); and had no effect on RT (unpleasant task, ANOVA, stimulus valence: F(1,6)=1.82, p=0.22, treatment: F(1,6)=0.73, p=0.43; pleasant task, ANOVA, stimulus valence: F(1,4)=0.39, p=0.56, treatment: F(1,4)=2.84, p=0.17).
In the 16 patients studied in the three experiments, 90 of the 96 channels were analysed. Due to technical problems (artefact or cable failure), six channels could not be included. The onset of the picture evoked a significant ERP in 15 patients, on 65 channels, in the right (n=35) and left (n=30) hemispheres (table 2). This ERP was observed at a mean latency of 168±74 ms and reached a maximum of 363±125 ms after stimulus onset (figure 2A). A polarity reversal of the ERP was observed in nine patients, on 11 channels, in the right (n=7 electrodes) and left (n=4 electrodes) hemispheres (figure 2A).
Effect of unpleasant and pleasant emotion on STN-LFP off treatment
Thirteen patients performed the unpleasant and pleasant tasks off treatment, with 72 channels recorded. ERPs were observed in 92% of patients (n=12), on 71% of the channels (n=51).
In the unpleasant task, NAUCs were significantly larger in unpleasant than in neutral trials in all motor contexts (ANOVA, stimulus valence: F(1,11)=50.96, p<10−3, stimulus valence×motor context interaction: F(2,22)=2.50, p=0.15; paired test in motor: p<10−2, non-motor: p<10−3, passive trials: p<10−3; figure 3C), but did not vary according to the subsequent motor behaviour (motor context: F(2,22)=2.96, p=0.08). In the pleasant task, NAUCs did not vary significantly, according to stimulus valence or subsequent motor behaviour (ANOVA, stimulus valence: F(1,11)=2.18, p=0.17; motor context: F(2,22)=0.78, p=0.47; stimulus valence×motor context interaction: F(2,22)=4.02, p=0.07) (figure 3D). Similar results were found in the individual analysis (table 2) with significant modification of activity in unpleasant compared with neutral trials in 67% of the channels in the unpleasant task (figure 3A) versus 24% in the pleasant task (figure 3B). Conversely, the change in ERP in relation to subsequent motor behaviour was not significantly different between the two tasks and did not exceed 18%. Simultaneous effects of the emotional valence and the subsequent motor behaviour did not exceed 10% in any task. In summary, ERP amplitude was modified by stimulus valence in the unpleasant task whereas no effect was observed in the pleasant task off treatment.
Effect of unpleasant and pleasant emotion on STN-LFP on treatment
Seven patients performed the unpleasant task and five patients the pleasant task while on and off treatment. On treatment, ERP were observed in 100% of patients and 86% of the channels in the unpleasant task; and in 100% of patients and 80% of the channels in the pleasant task. On treatment, mean NAUCs were significantly larger in emotional than in neutral trials in both unpleasant and pleasant tasks (unpleasant task: ANOVA, stimulus valence: F(1,6)=64, p<10−3; stimulus valence×motor context interaction: F(1,6)=1.12, p=0.32; pleasant task: ANOVA, stimulus valence: F(1,4)=38.71, p<10−2; stimulus valence×motor context interaction: F(1,4)=1.29, p=0.30). The difference between unpleasant and neutral trials was significant whatever the subsequent motor behaviour (motor: p<10−2, non-motor: p<10−3 and passive trials: p<10−2) (figure 4C), whereas the difference between pleasant and neutral trials was significant in motor trials and tended to be significant in the non-motor and passive trials (motor: p<10−2, non-motor p=0.04, passive trials: p=0.03) (figure 4D). These results were confirmed at the individual level (table 2) with significant changes in activity in relation to the emotional valence in 69% and 71% of the channels in the unpleasant and pleasant tasks, respectively (figure 4). In the same patients off treatment, mean NAUCs were significantly larger in unpleasant than in neutral trials (ANOVA, stimulus valence: F(1,6)=58.41, p<10−3, stimulus valence×motor context interaction: F(1,6)=1.18, p=0.29) but did not differ in pleasant compared with neutral trials (ANOVA, stimulus valence: F(1,4)=1.33, p=0.31, stimulus valence×motor context interaction: F(2,8)=0.76, p=0.40). The individual analysis showed similar results with significant changes in ERP in 67% and 33% of channels in the unpleasant and pleasant tasks, respectively (table 2).
In summary, in the unpleasant task, ERP amplitude was similarly modified by stimulus valence on and off treatment. In the pleasant task, ERP amplitude was modified by stimulus valence only on treatment.
Location of the ERPs
Localisation of the recording contacts showed that 92% of the channels located within the STN showed an ERP, and 66% of the channels located outside of the STN showed no ERP (figure 2B). Contacts outside of the STN were located in the Forel field (n=5), zona incerta (n=15), substantia nigra (n=2) and mesencephalic reticular formation (n=10). Among these, contacts showing an ERP were located in the zona incerta (n=5), substantia nigra (n=2), Forel field (n=2) and mesencephalic reticular formation (n=2).
Location of the emotional effect
A significant positive correlation was found between the emotional effects and the dorsoventral coordinates of the channels. In the unpleasant task, the emotional effect was maximum on ventral channels and diminished along the dorsoventral axis (R=0.55, p<0.001) (figure 5). The small number of channels showing a significant effect of valence in the pleasant task did not allow assessment of the effect of electrode location in this task.
Using a visual paradigm, we demonstrated that the STN processes emotional information in PD patients and that DRT modifies the STN response to pleasant information.
Modality independent cognitive STN evoked activity
We have first shown that visual stimuli evoked modification of activity in the STN at similar latencies in all motor contexts. This evoked response was therefore independent of any motor related process such as movement preparation, initiation or execution. This is true for limb movement but also saccadic ones, as similar ERP were observed in our study when preventing saccadic activity (see supplementary material, available online only), and similar ERP have been reported in the STN using auditory stimuli.35 Thus the cognitive process can activate the STN independently of the modality of stimulus presentation. It should be noted that the timing of the ERP observed in this study was similar to the P300, a scalp ERP that is sensitive to the amount of attentional resources allocated to the evaluation/categorisation of task relevant stimuli.36
Emotion modulates stimulus evoked activity in the STN
In our task, we manipulated the emotional valence of the stimuli but also their relevance to the task at hand. In other words, the task relevant stimuli triggering a motor response (target) were either emotional or neutral according to the instruction. LFP studies in humans have previously shown that the STN is implicated in the evaluation of task relevant stimuli,23 and in the preparation and initiation of motor responses.19–21 Consequently, we could have expected greater STN activity in response to task relevant stimuli regardless of their emotional valence. Interestingly, the opposite results were observed, with larger ERPs for emotional compared with neutral stimuli whatever their relevance for the task. This effect was even observed in passive trials, in which no stimuli were task relevant and no motor response was triggered. This result demonstrates that at the time of stimulus presentation, the STN processes the emotional valence of stimuli, highlighting its role in limbic processes. Given the anatomy of the STN afferent projections, emotional information might be transmitted to the STN by subcortical area such as the ventral striatum and ventral pallidum but it could also originate from cortical areas projecting directly to the STN.
Different effect of dopamine according to emotional valence
Strikingly, in PD patient off treatment, the amplitude of STN-ERP was modulated by unpleasant but not pleasant emotional valences. As no patient had depression, this effect cannot be related to mood disorder. Another explanation could be the differences in arousal capacities of unpleasant and pleasant pictures, pleasant pictures being less arousing. However, consistent with the result of a previous study,37 the level of arousal in the unpleasant task did not modify the ERP (see supplementary material, available online only). Alternatively, the STN could be implicated in the processing of unpleasant but not pleasant stimuli. This is supported by the selective deficit of negative facial emotion discrimination reported in PD following STN-HFS,2 but is in contradiction with a previous study reporting changes in STN oscillatory activity induced by pleasant pictures.24 As patients in the latter study were on treatment, we tested the effect of DRT on the STN response in a subset of patients. We observed that encoding of pleasant information was increased by treatment. While these results were obtained in a smaller number of patients, they strongly suggest that the encoding of pleasant information in the STN depends on the integrity of the nigrostriatal dopaminergic system. This is in accordance with a previous report of altered positive reinforcement learning in off treatment PD patients, which was restored after DRT intake.38 However, in the same study, negative reinforcement learning was altered off treatment whereas in our study, the encoding of unpleasant information was not modified by treatment. This discrepancy raises the question of the relation between the encoding of aversive events and the dopaminergic system.
The observation of a decrease in ERP amplitude between adjacent channels on the same electrode suggests that the extent of bipolar LFP recordings is quite restricted. Moreover, the correspondence between the location of the contacts inside the STN and the presence of an ERP, together with the observation of polarity reversal in nine patients on adjacent channels of the same electrode, confirms that the ERP was generated by a local source. Polarity reversals were preferentially localised in the ventral part of the nucleus with a gradient of the emotional effect along the dorsoventral axis. This observation strongly suggests that cell assembly, which receives and encodes the emotional value of a stimulus, is located ventrally in the nucleus, in accordance with the existence of a limbic territory in the anterior and ventromedial part of the STN.12 ,39
There are some limitations to the results reported here. First, in the area of emotional responses, one man's food can be another's poison. However, the categorisation task allowed us to control for subjects’ own feelings as the trials considered as emotional or neutral were selected according to the answer given by the subject. Second, it must be stated that these results have been obtained in the context of PD. In the behavioural analysis, when evaluating patients’ responses based on control subjects’ ratings, we observed 80% accuracy. This shows that the process of valence categorisation was not altered in our patients. Most importantly, to test whether recorded activity reflected the normal physiological STN response, we performed an appropriate control experiment by comparing the STN response on and off treatment. Third, the validity of the anatomofunctional correlation depends on the accuracy of channel localisation. An assessment of potential localisation errors was previously published,26 stating that ‘a 1 mm accuracy for target localisation was achievable’.
Our results show that the emotional value of stimuli is transmitted to the STN, regardless of the presence of a specific motor demand and of stimulus relevance to the task at hand. The ventral STN thus appears to be part of a brain limbic network. Even if delineating its specific function remains a challenging question, the STN has been hypothesised to act as a regulator of thoughts and behaviour, computing the threshold for decision making in high conflict situations.40 From our data, we can hypothesise that representing the affective value of the environment is necessary for the STN to achieve this computation. Accordingly, when disrupting information transfer through the ventral STN, neuropsychiatric side effects are observed in PD and OCD patients. Interestingly, both hypomania and depression have been reported in PD following STN HFS, but only hypomania has been reported in OCD. Increasing evidence suggests that hypomania is related to STN-HFS whereas depression would be triggered in PD patients by the decrease in DRT.6
We thank the patients for their participation. We also thank C Tallon-Baudry, D Schwartz, A Ducorps and K N'Diaye for technical assistance on data analysis, as well as WIA Haynes and B Lau for their comments on the manuscript.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
Contributors AB was involved in study design and conception, study set-up, recordings, data analysis and interpretation, and was responsible for drafting, revising and ﬁnalising the manuscript. She is the guarantor. M-LW was involved in obtaining funding, study design, patient care, and revising and ﬁnalising the manuscript. CK was involved in patient surgery as well as revising and ﬁnalising the manuscript. J-BP was involved in study design and conception. EB was involved in the atlas development. JY was involved in the atlas development, as well as revising and ﬁnalising the manuscript. LM was involved in obtaining funding, study design and data interpretation, as well as revising and ﬁnalising the manuscript.
Funding This work was supported by grants ANR 05-JCJC-0235-01, 06-NEURO-006-01 and by France Parkinson. The protocol was sponsored by INSERM (RBM C07-42).
Competing interests None.
Ethics approval The study was approved by the local ethics committee.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.