|Home | About | Journals | Submit | Contact Us | Français|
Impulsive behavior is thought to reflect a trait-like characteristic that can have broad consequences for an individual’s success and well-being, but its neurobiological basis remains elusive. Although striatal dopamine D2-like receptors have been linked with impulsive behavior and behavioral inhibition in rodents, a role for D2-like receptor function in frontostriatal circuits mediating inhibitory control in humans has not been shown. We investigated this role in a study of healthy research participants who underwent positron emission tomography with the D2/D3 dopamine-receptor ligand [18F]fallypride, and blood oxygen level-dependent functional magnetic resonance imaging (BOLD fMRI) while they performed the Stop-signal Task, a test of response inhibition. Striatal dopamine D2/D3-receptor availability was negatively correlated with speed of response inhibition (stop-signal reaction time), and positively correlated with inhibition-related fMRI activation in frontostriatal neural circuitry. Correlations involving D2/D3 receptor availability were strongest in the dorsal regions (caudate and putamen) of the striatum, consistent with findings of animal studies relating dopamine receptors and response inhibition. The results suggest that striatal D2-like receptor function in humans plays a major role in the neural circuitry that mediates behavioral control, an ability that is essential for adaptive responding and is compromised in a variety of common neuropsychiatric disorders.
Poor ability to inhibit an ongoing response (response inhibition) is manifested in several psychiatric disorders involving impulsive behavior (Moeller et al., 2001; Aron et al., 2007b; Chamberlain and Sahakian, 2007; de Wit, 2009). Such disorders have been associated with dopaminergic dysfunction (Swanson et al., 2007; Koob and Volkow, 2010), suggesting that dopamine contributes to response inhibition. This view is supported by pharmacological findings (de Wit et al., 2002; Eagle and Robbins, 2003b; Eagle et al., 2011), evidence that links dopamine-related genetic polymorphisms with response inhibition (Congdon et al., 2008; Hamidovic et al., 2009), and demonstrations of impaired response inhibition in stimulant abusers (Monterosso et al., 2005; Colzato et al., 2007), who have low striatal dopamine D2-like receptor availability (Volkow et al., 2001; Lee et al., 2009).
Rodent studies using the Stop-signal Task (SST) (Verbruggen and Logan, 2008) to assess response inhibition have suggested a specific role of dopamine D2/D3 receptors in the dorsal striatum (Eagle et al., 2011). Although a relationship between response inhibition and dopaminergic neurochemical markers has not been established in humans, dopamine receptors have been linked to self-report measures of impulsivity. Impulsivity has been correlated negatively with striatal dopamine D2/D3-receptor availability in a combined sample of methamphetamine-dependent and healthy-control participants (Lee et al., 2009), and positively with amphetamine-induced striatal dopamine release in healthy participants (Buckholtz et al., 2010). These findings highlight the relevance of striatal dopamine in trait-like impulsivity without elucidating its relationship to behavioral measures of response inhibition.
Studies of SST performance in patients and fMRI responses in healthy controls have highlighted the importance of fronto-basal ganglia circuitry in response inhibition (e.g., Rieger et al., 2003; Aron et al., 2007b; Chambers et al., 2009), emphasizing the role of right inferior frontal cortex (IFC), anterior insula, presupplementary motor area (preSMA)/anterior cingulate, and the subthalamic nucleus (STN) (Mirabella et al., 2011; Swann et al., 2011). It has been proposed that the right IFC and/or preSMA project to the STN via a “hyperdirect” pathway to stop a response quickly (Aron et al., 2007a). However, striatal activation also accompanies stopping (Vink et al., 2005; Aron and Poldrack, 2006; Li et al., 2008a; Boehler et al., 2010; Zandbelt and Vink, 2010), possibly reflecting inhibitory control in preparation for stopping or occurring via an “indirect” basal ganglia pathway. Further insights may come from exploring the relationship between frontostriatal activation during stopping and dopaminergic markers.
To determine how D2/D3 receptors relate to response inhibition and associated frontostriatal activation, we assessed striatal D2/D3-receptor availability using positron emission tomography (PET) in healthy participants who performed the SST during fMRI. As suggested by the aforementioned literature, we anticipated a negative correlation of striatal D2/D3-receptor availability with stopping speed. We also expected that striatal D2/D3-receptor availability would be related to the fMRI response during stopping. Although prior literature directly linking dopaminergic neurochemical markers with human fMRI responses is sparse, striatal dopamine release has been positively correlated with striatal fMRI activation, suggesting that D2/D3-receptor availability and inhibition-related striatal fMRI activation would be positive (Schott et al., 2008).
All procedures were approved by the UCLA Office for the Protection of Research Subjects. Participants were recruited using radio, Internet, and local newspaper advertisements. After receiving a detailed explanation of the study, participants gave written informed consent and underwent eligibility screening using questionnaires, the Structured Clinical Interview for DSM-IV (SCID, First et al., 1996), and medical examination. Participants were excluded from the study for the following criteria: current Axis I diagnosis other than nicotine dependence; use of psychotropic medications or substances, except some marijuana or alcohol (not meeting DSM-IV criteria for dependence); central nervous system, cardiovascular, pulmonary, or systemic disease; HIV seropositivity, hepatic disease, or hematocrit < 32; pregnancy; lack of fluency in English; and MRI ineligibility due to metal implants or claustrophobia. Corrective lenses were used to ensure adequate visual acuity when participants viewed stimuli in the scanner. Participants were required to provide negative urine samples for illicit drugs of abuse on days of testing. They received compensation in the form of cash, gift certificates, and vouchers.
Twenty-one healthy adults participated in the study. Data from three participants were excluded from analyses because the participants did not meet performance criteria on the Stop-signal Task. To be included, participants were required to reach the following criteria: successful inhibition on 40% to 60% of stop trials, at least 80% responding on Go trials, less than 10% errors in discriminating the arrow direction on Go trials, and an estimated SSRT greater than 50 ms. All three excluded participants responded on less than 80% of Go trials and had SSRT estimates of < 50 ms (unrealistically fast, suggesting erroneous task performance); one had more than 10% errors on direction discrimination. In total, data from 18 participants were included in the reported analyses (age: M=32.5 years, SD=8.2; 8 female).
During scanning, participants performed the Stop-signal Task (Logan and Cowan, 1984), which required them to execute a motor response (Go trials) on most trials and to withhold their response on some trials (Stop trials). The task implemented was similar to the one used by Aron & Poldrack (2006) except that two adaptive staircases were used instead of four (see below). On each trial, a left- or right-pointing arrow stimulus was displayed. On Go trials (75% of trials), participants were to respond as fast as possible with a left or right key press (using index and middle fingers of the right hand). For the Stop trials (25% of trials), participants were instructed to stop their response when they heard a Stop signal (an auditory tone) that was presented at a particular delay (stop-signal delay, SSD) subsequent to the arrow stimulus. The SSD value for the Stop trial was adaptively determined to ensure that participants inhibited their response successfully on approximately 50% of trials.
Performance on the Stop-signal Task has been characterized as a race between independently running Go and Stop processes (Logan and Cowan, 1984). Although Go response times are directly measurable (Go RT), Stop response times are not and must be estimated using the stop-signal reaction time (SSRT). The SSRT is computed by subtracting the SSD from the Go RT, and relies on the assumption that the response-time distribution on Stop trials is the same as on Go trials. Using the staircase procedure to adjust the SSD dynamically for producing inhibition on 50% of Stop trials ensures consistency with this assumption and accounts for differences in Go RTs across participants (Band et al., 2003).
A Go trial consisted of the following sequence. First, a white circular fixation ring (subtending a 4.3 × 6.2° viewing angle) appeared in the center of the screen against a black background. After 500 ms, a white arrow appeared within the fixation circle. The arrow pointed to the left on half of the trials, and to the right on the other half. The direction of the arrow was randomized across trials. The fixation ring and arrow remained on the screen for up to 1 s, after which they disappeared and the background was shown. The fixation ring and arrow disappeared as soon as the participant made a choice, and the background was shown for the remainder of the 1-s period plus the rest period. If the participant pressed the incorrect key in response to the arrow, the computer recorded the error, but no feedback was provided to the participant. A Stop trial was identical to a Go trial in all respects, except that a tone (900 Hz; duration, 500 ms) was played at the SSD after onset of the arrow stimulus. If a participant inhibited his or her response on Stop trials, the arrow and fixation ring remained on the screen for the duration of the 1-s period. If the participant responded, the arrow and fixation ring disappeared, and the background screen was shown for the remaining period. The SSD fluctuated dynamically throughout the experiment, depending on the participant’s performance. If a participant successfully inhibited a response on a Stop trial, inhibition was made more difficult on a subsequent Stop trial by increasing the SSD by 50 ms; if a participant did not successfully inhibit, inhibition was made easier by decreasing the SSD by 50 ms. Two independent staircases were used in this way. Starting SSD values were determined on the basis of each participant’s performance of the task outside the scanner in a behavioral testing session on a prior day (SSD: M=256.1, SD=106.5).
On the scanning day, both staircases began at the same SSD value. This value was determined for each participant in the earlier behavioral test session. Two blocks (fMRI runs) of trials were presented, with 96 Go trials and 32 Stop trials in each block. Each SSD staircase was incremented or decremented 16 times within each block. The staircases were independent but randomly interleaved (i.e., each particular Stop trial was randomly selected from one of the two staircases). To maintain continuity of performance between the two blocks, the starting SSD values for the two ladders in the second block were set to the last SSD values from the ladders in the previous block.
We used custom Matlab (MathWorks, Natick, MA) code to select sequences of Go and Stop events and to select the distribution of rest time in a way that optimized the detection of hemodynamic responses for the critical contrast of Stop and Go events. Approximately four Go trials appeared between Stop trials (M=4.08, SD=2.97), and the number of leftward and rightward pointing arrows was equal. Rest periods were imposed following each trial. The duration of rest time ranged between 0.5 and 4 s (mean, 1 s; sampled from an exponential distribution truncated at 4 s). A Monte-Carlo approach was used to generate a large number of sequences within these constraints in order to select those with the highest efficiency for detecting differences between Go and Stop events (c.f., Liu et al., 2001).
The presentation of all stimuli and response collection were programmed using Matlab (Mathworks, Natick, MA) and the Psychtoolbox (www.psychtoolbox.org) on an Apple MacBook running Mac OSX (Apple Computers, Cupertino, CA). Visual and auditory stimuli were presented using MRI compatible goggles and headphones, respectively (Resonance Technologies, Van Nuys, CA).
Task directions were explained to participants on their pre-scan behavioral testing day. They were informed that stopping and going were equally important and that it would not always be possible to stop. Within the scanner, participants responded with their right hands on an MR-compatible button box. Stop tones were played through headphones at a level both sufficient to exceed scanner noise and comfortable to the subject. Each scan was preceded by an instruction screen with a reminder to the subject: “Remember, respond as FAST as you can once you see the arrow. However, if you hear a beep, your task is to STOP yourself from pressing. Stopping and going are equally important.”
Imaging was performed using a 3T Siemens AG (Erlangen, Germany) Allegra MRI scanner. We acquired 364 functional T2*-weighted echoplanar images (EPI) across two runs [slice thickness, 4 mm; 34 slices; repetition time (TR), 2 s; echo time (TE), 30 ms; flip angle, 90°; matrix, 64 × 64; field of view (FOV), 200 mm]. Two additional volumes at the beginning of each run were discarded to allow for T1 equilibrium effects. For registration purposes, a T2-weighted, matched-bandwidth, high-resolution anatomical scan (same slice prescription as EPI) and a magnetization-prepared, rapid-acquisition, gradient-echo (MPRAGE), high-resolution scan were acquired for each participant. The MPRAGE scan was conducted in a separate session on a Siemens Sonata 1.5T scanner (slice thickness: 1 mm; 160 slices, TR, 1900 ms; TE, 4.38 ms; flip angle, 15; FOV, 256 mm). The orientation for matched bandwidth and EPI scans was oblique axial so as to maximize full brain coverage and to optimize signal from brain regions prone to signal dropout.
Dopamine D2/D3-receptor availability was assayed using [18F]fallypride, a radioligand with high affinity for D2/D3 receptors (Mukherjee et al., 1995). Images were acquired using a Siemens ECAT EXACT HR+ scanner [in-plane resolution full-width at half-maximum (FWHM) = 4.6 mm, axial FWHM = 3.5 mm, axial field of view = 15.52 cm] in three-dimensional (3D) mode. Subjects were placed in the supine position with the brain centered in the transaxial field of view. A 7-min transmission scan was acquired using a rotating 68Ge/68Ga rod source for attenuation correction. PET dynamic data acquisition was initiated with a bolus injection of [18F]fallypride (~5 mCi in 30 s). To reduce radiation exposure to the bladder wall and to minimize discomfort, participants were given a 20-min break after the first 80 min of emission data acquisition, and then returned to the scanner bed after the break. After a second 7-min transmission scan, emission data were collected for another 80 min. The total dynamic scanning sequence consisted of six 30-s frames, seven 1-min frames, five 2-min frames, four 5-min frames, and 12 10-min frames. Data were reconstructed using ECAT v7.3 OSEM (Ordered Subset Expectation Maximization; 3 iterations, 16 subsets) after corrections for decay, attenuation, and scatter.
The median and standard deviation of Go response times (Go RT) were calculated for each participant as well as percent Go responses and percent inhibition on Stop trials. The average SSD was computed for each subject from the values of the two staircases, and since the 50% inhibition criteria were met, SSRT was estimable by subtracting each participant’s average SSD from his/her median Go RT. All behavioral data were analyzed using Matlab (Mathworks, Natick, MA) and the R statistical package (www.r-project.org).
Analysis of fMRI data was performed using the FSL (4.1) toolbox from the Oxford Centre for fMRI of the Brain (www.fmrib.ox.ac.uk/fsl). The image time-course for each participant was first realigned to compensate for small head movements (Jenkinson et al., 2002), and all non-brain matter was removed using FSL’s brain extraction tool. Images were denoised for motion-related artifacts using independent components analysis (FSL’s MELODIC). Motion-related components were identified manually using a set of heuristics (Tohka et al., 2008), and the data were then reconstructed after removing the motion-related components. Data were spatially smoothed using a 6-mm full-width-half maximum Gaussian kernel. Registration was conducted through a three-step procedure, whereby EPI images were first registered to the matched-bandwidth high-resolution structural image, then to the MPRAGE structural image, and finally into standard [Montreal Neurological Institute (MNI)] space (MNI avg152 template), using 12-parameter affine transformations (Jenkinson and Smith, 2001). Registration from MPRAGE structural images to standard space was further refined using FNIRT nonlinear registration (Andersson et al., 2007a, b). Statistical analyses at the single-subject level were performed in native space, with the statistical maps normalized to standard space prior to higher-level analysis.
Whole-brain statistical analysis was performed using a multi-stage approach to implement a mixed-effects model treating participants as a random-effects variable. Statistical modeling was first performed separately for each imaging run. Regressors of interest were created by convolving a delta function, representing trial onset times with a canonical (double-gamma) hemodynamic response function. Regressors of interest included Go trials, Successful Stop trials, and Unsuccessful Stop trials. Go trials in which participants omitted a response or made discrimination errors were modeled using a single nuisance covariate, and motion parameters were included as covariates of no interest to account for variance associated with residual motion not captured by the MELODIC denoising procedure. No participant exceeded translational motion above 2.5 mm (M=0.49, SD=0.39).
For all analyses, time-series statistical analysis was carried out using FILM (FMRIB’s Improved Linear Model) with local autocorrelation correction (Woolrich et al., 2001) after highpass temporal filtering (Gaussian-weighted LSF straight line fitting, with sigma = 33 s). Contrast images for the two runs within each session were combined using fixed-effects analyses. The results of these contrasts were submitted to whole-brain correlations with D2/D3-receptor binding-potentials (see Statistical Analyses below).
The caudate nucleus, putamen, and nucleus accumbens were defined as volumes-of-interest (VOIs) using an automated procedure (FSL FIRST; FMRIB Integrated Registration and Segmentation Tool, Oxford University, Oxford, UK, www.fmrib.ox.ac.uk/fsl/first/index.html) that provides a 3D binary mask for these regions in native space. The caudate nucleus VOI included the head and body of caudate, and the putamen VOI included the entire putamen.
Reconstructed PET images were combined into 16 images, each containing an average of 10-minute dynamic frames of data. These 16 images were motion-corrected using FSL’s McFLIRT (Jenkinson et al., 2002) (using least squares as the cost function and the first image of the second scanning block as the reference volume). After motion correction, PET images were co-registered to the participant’s structural MRI using a six-parameter, rigid-body transformation computed with the ART software package (Ardekani et al., 1995). Specifically, the transformation matrix for registration of the first PET image (containing maximal intensities across the brain) to the MRI was computed and applied to all 16 PET images in the series. Time-activity data within anatomically defined VOIs (see above) were extracted from motion corrected, co-registered PET data and imported into the PMOD 3.2 (PMOD Technologies Ltd., Zurich, Switzerland) kinetic analysis program (PKIN) for kinetic modeling. Time-activity curves were fit using the simplified reference-tissue model (SRTM; Lammertsma and Hume, 1996) to provide an estimation of k2′, the rate constant for transfer of the tracer from the reference region to plasma. The cerebellum was used as a reference region, a common practice that assumes the cerebellum is devoid of specific binding sites for the radiotracer (Lammertsma and Hume, 1996). Although non-negligible D2-like receptor binding in cerebellum has been measured using [11C]FLB457, specific binding of [18F]fallypride (which has a higher equilibrium dissociation constant) was not detected in the cerebellum (Vandehey et al., 2010). A volume-weighted average of k2′ estimates from high-activity regions (caudate and putamen) was computed. The time-activity curves were refit using the SRTM2 model (Wu and Carson, 2002) with the computed k2′ values applied to all VOIs. Binding potential (BPND) was then calculated by subtracting 1.0 from the product of the tracer delivery (R1) and the tracer washout (k2′/k2a). Whole-brain BPND maps were computed by carrying out the above computations on a voxel-by-voxel basis.
Since BPND in the left and right regions of each striatal VOI were highly correlated (caudate: r=0.99, p<0.001; putamen: r=0.89, p<0.001; nucleus accumbens: 0.71, p<0.001), a volume-weighted average of the left and right regions for each VOI was used for correlation analyses. Volume-weighted averages of left and right striatal VOI Sucessful Stop vs. Go fMRI parameter estimates were also computed since they were also highly correlated (caudate: r=0.82, p<0.0001; putamen: r=0.81, p=<0.0001; nucleus accumbens: r=0.66, p<0.005).
We used both VOI-based and voxel-wise approaches to examine correlations between BPND in striatal regions, SSRT, and the fMRI contrast of Successful Stop vs. Go. VOI-based correlations were computed using the non-parametric Spearman’s Rho test for consistency with the non-parametric whole-brain correlations (VOI results were comparable with those obtained using Pearson correlations). In order to obtain sensitivity in substructures within VOIs, BPND images (thresholded to exclude voxels with negligible D2/D3 binding, i.e., BPND < 1) were subjected to linear regression to test associations with SSRT. Control for familywise error was implemented using nonparametric randomization tests with the FSL RANDOMISE v2.8 tool (Permutation-based nonparametric inference, Oxford University, Oxford UK), with variance smoothing of 10 mm (FWHM Gaussian) (www.fmrib.ox.ac.uk/fsl/randomise). This method repeatedly reorders the rows of the design matrix and computes the statistical map for each sample, recording the maximum statistic across the brain for each sample in order to obtain an empirical null distribution (Nichols and Holmes, 2002). Threshold-free cluster enhancement (TFCE) (Smith and Nichols, 2009) was used to detect significant clusters of activation. For each analysis, 5000 randomization runs were performed. Statistical maps were thresholded at p < 0.05 (one-tailed) corrected for the entire search volume, though some results are shown at more lenient thresholds for visualization purposes, as noted in the Results section.
In addition to VOI-based correlations between striatal binding potentials and successful Stop vs. Go fMRI parameter estimates, we conducted whole-brain analyses examining regions throughout the brain that showed a correlation with BPND. Whole-brain between-subject correlations were conducted using the FMRIB Local Analysis of Mixed Effects (FLAME1) module in FSL (Beckmann et al., 2003; Woolrich et al., 2004) using binding-potential values for each VOI as covariates (zero-meaned). Z (Gaussianised T) statistic images were thresholded using cluster-corrected statistics with a height threshold of Z > 1.96 and a cluster probability threshold of p < 0.05, whole-brain corrected using the theory of Gaussian Random Fields (Worsley et al., 1992). All analyses were subjected to robust outlier deweighting (Woolrich, 2008). Conjunction analyses to indicate overlap between statistical maps was conducted using the revised minimum statistic approach proposed by Nichols et al. (2005) and cluster-corrected statistics. Locations of cortical activations were identified using the Harvard-Oxford Probabilistic Atlas, which is included in the FSL software package, and the sectional brain atlas of Duvernoy and Bourgouin (1999).
Behavioral results from the Stop-signal Task are shown in Table 1. Participants performed at 99% correct on Go trials and inhibited their responses on approximately half of the Stop trials (51%) indicating that the adaptive staircase procedure for equating Stop trials across participants was successful. SSRT and median Go RT values were similar to those reported in the literature from similar samples (e.g., Logan et al., 1997; Boehler et al., 2010).
SSRT showed a significant negative correlation with BPND values in the anatomically-defined caudate (r=-0.57) and putamen (r=-0.50) VOIs but not in the nucleus accumbens (r=-0.18) (Table 2, Figure 1) indicating that better response inhibition performance was associated with greater receptor availability. Voxel-wise analyses revealed a significant negative correlation bilaterally in the putamen (p<0.05, TFCE-corrected for multiple comparisons) and one that approached significance in the caudate (p<0.06, TFCE-corrected) (Figure 1), but correlations in the vicinity of the nucleus accumbens only appeared at lower thresholds (p<0.15, TFCE-corrected). No correlations were observed between striatal BPND and median Go RT (caudate: r=0.04, p=0.88; putamen: r=0.08, p=0.74; nucleus accumbens: r=0.10, p=0.70), suggesting that the observed caudate and putamen correlations were specific to variation in inhibiting a response, not to motor engagement in general.
Whole-brain analyses of the Successful Stop vs. Go fMRI contrast revealed regions of activation typically reported for this task, including the right inferior frontal gyrus, supplementary motor area/pre-supplementary motor area, anterior cingulate, bilateral anterior insula, bilateral middle frontal gyrus, bilateral superior temporal gyrus, bilateral posterior parietal cortex (Figure 2). Among striatal regions, bilateral caudate activation did not survive cluster-corrected thresholds, but showed significant activation at an uncorrected threshold of P<0.001. A whole-brain correlation analysis using participants’ SSRT values as a covariate revealed a negative correlation in clusters of activation within the superior frontal gyrus, anterior cingulate, left inferior frontal gyrus/lateral orbitofrontal cortex, and on the right side, caudate, putamen, thalamus, and amygdala.
VOI-based analyses revealed a significant positive correlation between BPND and Successful Stop vs. Go fMRI activation in the caudate but not in the putamen or nucleus accumbens (Table 2, Figure 3A & B). When removing an outlier in the caudate fMRI data (one time point greater than two standard deviations, see scatterplot in Figure 3B), the correlation remained significant (r=0.79, p=0.0002, N=17).
To determine the extent to which frontostriatal functional activation related to caudate BPND, we performed a whole-brain voxel-wise correlation of Successful Stop vs. Go fMRI activation and caudate BPND (corrected for multiple comparisons). In addition to the caudate, several regions within the frontal lobes showed significant clusters (Table 4, Figure 3C &D); these included the right lateral orbital/fronto-polar cortex, left anterior insula, and rostral anterior cingulate/superior frontal gyrus (medial portion), regions previously reported to be related to response-inhibition behavior. A conjunction analysis used to statistically identify the extent to which these regions overlapped with those activated during the Successful Stop vs. Go contrast revealed overlap in the rostral anterior cingulate and left anterior insula but at a liberal voxel-wise height threshold of Z>1 (cluster-probability threshold of P<0.05). A conjunction with the correlation of Successful Stop vs. Go and SSRT also showed overlap in rostral anterior cingulate as well as right superior frontal gyrus (medial portion) and left inferior frontal gyrus at the same liberal threshold. No regions showed a negative correlation with BPND.
Striatal dopamine-receptor availability was associated both with response-inhibition performance and stopping-related frontostriatal activation. Specifically, D2/D3-receptor availability in the dorsal but not the ventral striatum was negatively correlated with SSRT (i.e., those with greater receptor availability stopped more quickly). These correlations held for SSRT but not for Go RT, suggesting that they were specific to stopping. Caudate D2/D3-receptor availability was positively correlated with fMRI Stop vs. Go activation in the caudate and prefrontal regions, including right lateral orbitofrontal/fronto-polar cortex, left anterior insula, and anterior cingulate cortex.
This study is the first to relate SSRT to a dopaminergic neurochemical marker in humans. The results support prior human studies showing a relationship between dopamine function and SSRT using pharmacological manipulations (Tannock et al., 1989; de Wit et al., 2002), indirect indicators of dopamine function measured via spontaneous eye-blink rate (Colzato et al., 2009), and observations of longer SSRTs in stimulant abusers (Monterosso et al., 2005; Colzato et al., 2007), who, as a group, exhibit low striatal D2-like receptor availability. They also extend our observation of a negative correlation between self-reported impulsivity and striatal D2/D3-receptor availability in humans, particularly those who abuse methamphetamine (Lee et al., 2009), by showing a positive relationship between response inhibition and striatal D2/D3-receptor availability in healthy participants. It is also consistent with studies in which baseline striatal D2-like receptor availability predicted impulsive behavior and rates of cocaine administration in drug-naive rodents (Dalley et al., 2007) and non-human primates (Nader et al., 2006).
The anatomical specificity (dorsal/ventral striatum) of the SSRT-receptor availability correlations corroborate findings from rodent lesion and pharmacological studies suggesting that the nucleus accumbens does not participate in response inhibition (as measured by SSRT) (Eagle and Robbins, 2003a; Eagle et al., 2011). These results cohere with a distinction between the ability to ‘stop’ an initiated response and ‘wait’ for an impending reward as separate phenotypes of impulsivity served by dorsal and ventral striatum, respectively (Robinson et al., 2009). Further support for this distinction comes from observations of lower D2/D3-receptor availability in the ventral but not dorsal striatum of highly impulsive rats that exhibit greater premature responding (deficient ‘waiting’) (Dalley et al., 2007), and studies showing that nucleus accumbens lesions induce more impulsive selections for sooner, smaller versus larger, later rewards (Cardinal et al., 2001). Nonetheless, while SSRT and receptor availability in the nucleus accumbens were not correlated in this study, measurement error may have contributed as the region is a small structure susceptible to partial volume effects in PET.
Our results are partially consistent with a model of striatal control of behavior that makes functional distinctions between dopamine receptor subtypes (Frank, 2005). This distinction posits that striatonigral neurons primarily expressing D1 receptors facilitate “Go” responding while striatopallidal neurons expressing D2 receptors support “NoGo” responses (Frank and O’Reilly, 2006; Frank et al., 2007). Although we did not measure D1 receptor availability, the fact that D2/D3-receptor availability was related to SSRT and not Go RT generally supports this model. However, caution should be taken when considering these results with respect to the model. Assessments comparing studies of stop-signal and go/no-go tasks indicate clear dissociations in the pharmacology and neural circuitry underlying ‘action cancelation’ (i.e., stopping) and ‘action restraint’ (i.e., NoGo) (Eagle et al., 2008; Swick et al., 2011), suggesting that the two represent different processes. Thus, models drawing from Go/No-Go results may not generalize to SST findings. Moreover, pharmacological studies do not support a clear double-dissociation between receptor subtype and the two basic components of these tasks (i.e., Go and Stop/NoGo). Eagle et al. (2011) found that a D2-receptor antagonist infused in the dorsal striatum slowed SSRTs but also slowed Go RTs, whereas a D1 antagonist decreased SSRTs with no effect on Go RT. These results are mostly parallel with our findings but only partially consistent with the model, suggesting that, beyond separating receptor function by action inhibition and go processes, models of striatal behavioral control may benefit from taking into account a balance of both D1 and D2 receptor activity in support of inhibitory control.
Further evidence for the role of D2 receptors in stopping comes from a human genetic study linking a polymorphism that controls D2-receptor expression (Zhang et al., 2007) and SSRT (Hamidovic et al., 2009). Although we could not directly distinguish between D2- and D3-receptor availability using [18F]fallypride, our results in the dorsal striatum likely reflect involvement of D2 receptors as dopamine receptors in this region in humans are primarily D2 whereas the ventral striatum exhibits substantial D3 receptor binding (Seeman et al., 2006; Searle et al., 2010).
We also found that dorsal striatal D2/D3-receptor availability is related to fMRI activation associated with stopping. Previous fMRI and patient studies have identified key brain regions important for mediating SSRT. These include ventrolateral (right inferior frontal gyrus (rIFG)/anterior insula) and medial prefrontal [pre-supplementary motor area (preSMA)/anterior cingulate (ACC) cortex], subthalamic nucleus (e.g., Aron et al., 2003; Aron et al., 2007a; Chambers et al., 2009), and the striatum (Vink et al., 2005; Li et al., 2008b; Jahfari et al., 2009; Boehler et al., 2010; Zandbelt and Vink, 2010). Within the striatum, the caudate nucleus shows greater activation during stopping with fast versus slow SSRTs (Li et al., 2008b) and when participants are more likely to anticipate stopping (Vink et al., 2005; Jahfari et al., 2009; Zandbelt and Vink, 2010), highlighting its role in controlled execution of motor behavior (whether by preparing to stop, representing stopping rules, or implementing stopping itself). Whether this activation is related to dopaminergic function per se had not been previously established. The positive correlation between caudate D2/D3-receptor availability and Successful Stop versus Go activation reported here provides potential evidence for D2- like neurotransmission underlying stopping-related caudate activation. However, since these correlations do not demonstrate a causal relationship, caudate D2/D3-receptor availability may only be an index of stopping ability and not necessarily directly influencing activation. Further studies are required for a definitive answer.
We also observed a positive relationship between activation in several prefrontal regions and caudate D2/D3-receptor availability. These included the right superior frontal cortex, right lateral orbital/fronto-polar cortex, left anterior insula, ACC and paracingulate cortices. Insula and ACC activations are commonly reported in fMRI studies of response inhibition (see Swick et al., 2011 for a meta-analysis) and have been identified in a “mega”-analysis of several Stop-signal fMRI datasets to negatively correlate with SSRT (Congdon et al., 2010). Activation in these regions overlapped across the Stop vs. Go contrast and its correlation with BPND and SSRT, adding support for their role in response inhibition with potential mediation by dopamine receptor function.
We observed a correlation between striatal receptor availability and activation within two other prefrontal regions important for stopping, preSMA and rIFG, only with uncorrected, voxel-wise thresholds (preSMA: p<0.005, x,y,z=6, 10, 62; rIFG, pars opercularis, p<0.001, x,y,z=54, 24, 0), providing only preliminary evidence for a relationship to dopamine receptor availability.
The correlation between ACC activation and caudate D2/D3-receptor availability further supports a dopaminergic influence on response inhibition given the direct connections between dorsal striatum and ACC (Haber and Knutson, 2010). It is possible that dopaminergic function influences the strength of frontostriatal functional connectivity (e.g., ACC-caudate strength) mediating SSRT (Jahfari et al., 2011), but this remains to be established.
Together with a PET study showing a positive relationship between striatal fMRI activation and dopamine release (Schott et al., 2008), this study provides evidence for an influence of striatal dopamine neurotransmission on BOLD responses. However, the coupling of dopamine receptor availability and the neurovasculature on which the BOLD signal depends is not fully understood. Rodent studies indicate that indirect dopamine agonists (e.g., amphetamine) increase regional cerebral blood volume (rCBV), a strong correlate of BOLD fMRI, by a primarily non-vascular action; however, direct D2/D3 agonists reduce rCBV and antagonize rCBV enhancement by indirect agonists (Chen et al., 2005; Choi et al., 2006). These observations suggest that striatal D2/D3 receptor availability would be negatively related to fMRI responses associated with dopamine release. The discrepancy between the positive relationship observed in humans and these findings using systemic dopaminergic manipulations may reflect a lack of correspondence between the animal and human models or differences between actions of activity-related dopamine release in a discreet network as compared with globally induced pharmacological effects.
In summary, we show that dopamine-receptor availability is related to response inhibition and activation of related frontostriatal circuitry in humans. Although SSRT is primarily considered a measure of motor-response inhibition, some evidence suggests that it is related to other forms of self-control, such as drug craving in stimulant abusers and regulation of emotional states (Cohen and Lieberman, 2010; Tabibnia et al., 2011) and is predictive of substance abuse vulnerability (Nigg et al., 2006; Ersche et al., 2012). Several psychiatric populations (e.g., individuals with ADHD, schizophrenia, drug dependence) show irregular dopamine function combined with impulsive behavior as common phenotypes. This study provides evidence for a direct link between the two, affording a basis for pharmacotherapeutic treatments that target specific dopamine-receptor function for disorders marked by poor response inhibition.
Supported by NIH grants P20 DA022539, R01 DA020726 (EDL), the Consortium for Neuropsychiatric Phenomics (NIH Roadmap for Medical Research grant UL1 DE019580, RL1 DA024853 (EDL)) and M01 RR00865 (UCLA GCRC). Additional funding provided by endowments from the Thomas P. and Katherine K. Pike Chair in Addiction Studies, and the Marjorie M. Greene Trust.
Disclosure/Conflict of Interest
The investigators have no conflicts of interest or financial disclosures to report.