Abstract
Impulsivity is a behavioral trait that is elevated in many neuropsychiatric disorders. Parkinson's disease (PD) patients can exhibit a specific pattern of reward-seeking impulsive-compulsive behaviors (ICBs), as well as more subtle changes to generalized trait impulsivity. Prior studies in healthy controls (HCs) suggest that trait impulsivity is regulated by D2/3 autoreceptors in mesocorticolimbic circuits. While altered D2/3 binding is noted in ICB+ PD patients, there is limited prior assessment of the trait impulsivity–D2/3 relationship in PD, and no prior direct comparison with patterns in HCs. We examined 54 PD (36 M; 18 F) and 31 sex- and age-matched HC (21 M; 10 F) subjects using [18F]fallypride, a high-affinity D2/3 receptor ligand, to measure striatal and extrastriatal D2/3 nondisplaceable binding potential (BPND). Subcortical and cortical assessment exclusively used ROI or exploratory-voxelwise methods, respectively. All completed the Barratt Impulsiveness Scale, a measure of trait impulsivity. Subcortical ROI analyses indicated a negative relationship between trait impulsivity and D2/3 BPND in the ventral striatum and amygdala of HCs but not in PD. By contrast, voxelwise methods demonstrated a positive trait impulsivity–D2/3 BPND correlation in ventral frontal olfactocentric-paralimbic cortex of subjects with PD but not HCs. Subscale analysis also highlighted different aspects of impulsivity, with significant interactions between group and motor impulsivity in the ventral striatum, and attentional impulsivity in the amygdala and frontal paralimbic cortex. These results suggest that dopamine functioning in distinct regions of the mesocorticolimbic circuit influence aspects of impulsivity, with the relative importance of regional dopamine functions shifting in the neuropharmacological context of PD.
SIGNIFICANCE STATEMENT The biological determinants of impulsivity have broad clinical relevance, from addiction to neurodegenerative disorders. Here, we address biomolecular distinctions in Parkinson's disease. This is the first study to evaluate a large cohort of Parkinson's disease patients and age-matched healthy controls with a measure of trait impulsivity and concurrent [18F]fallypride PET, a method that allows quantification of D2/3 receptors throughout the mesocorticolimbic network. We demonstrate widespread differences in the trait impulsivity–dopamine relationship, including (1) loss of subcortical relationships present in the healthy brain and (2) emergence of a new relationship in a limbic cortical area. This illustrates the loss of mechanisms of behavioral regulation present in the healthy brain while suggesting a potential compensatory response and target for future investigation.
Introduction
Impulsivity reflects a deficiency in withholding actions, executing long-term plans, and making adjustments according to evaluation of future consequences (Patton et al., 1995). Aberrant levels of impulsivity are widely implicated across the spectrum of neuropsychiatric disease, evident in substance abuse, attention-deficit hyperactivity disorder, and antisocial personality disorder (Patton et al., 1995; Barratt et al., 1997; Avila et al., 2004). Abnormal mesolimbic dopamine transmission may be one common cause, where impulsiveness is modulated by disruptions to dopamine release, especially midbrain projections to the ventral striatum (Pettit et al., 1984; Piazza et al., 1991; Dalley and Roiser, 2012).
Dopamine D2 and D3 receptors located at mesolimbic somatodendritic (midbrain), presynaptic axonal, and postsynaptic somatodendritic (basal ganglia) sites are molecular structures of particular interest. D2 inhibitory autoreceptors regulate the synthesis and release of dopamine; and inhibitory postsynaptic receptors are expressed on GABAergic neurons of the basal ganglia (Mercuri et al., 1992; Khan et al., 1998). Preclinical studies reveal an association between heightened impulsivity (indexed by premature responses) and decreased ventral striatal D2/3 binding (Dalley et al., 2007). This work is complemented by data in healthy humans, where impulsivity, as measured by the Barratt Impulsiveness Scale-11 (BIS-11), is negatively correlated with midbrain D2/3 binding and positively correlated with amphetamine-evoked ventral striatal dopamine release (Barratt et al., 1997; Buckholtz et al., 2010). However, regulation of this trait is not solely attributable to changes in mesolimbic or accumbal dopamine function. Other dopaminergic and nondopaminergic systems, such as those of the orbitofrontal cortex (OFC) (Volkow et al., 2001, 2008; Barker et al., 2013), or amygdala (Kim et al., 2018) likely play significant roles in impulsive behavior (Dalley and Roiser, 2012).
The involvement of aberrant dopamine neurotransmission in impulsive behavior is particularly evident in Parkinson's disease (PD). Medication-naive patients often display decreased novelty-seeking tendencies and a preference for risk-averse, conservative choices (Menza et al., 1993), while chronic dopamine replacement therapies (especially D2/3 agonist medications) are known to induce the compulsive pursuit of naturally rewarding activities, including sex, gambling, shopping, eating, and hobby participation (Weintraub et al., 2015). These impulsive-compulsive behaviors (ICBs) bear a notable impact on patient disability and caregiver burden (Leroi et al., 2012; McKeown et al., 2020), providing a clinically relevant impetus for further study. ICBs are associated with mesolimbic dopaminergic abnormalities, including increased ventral striatal dopamine release in response to reward (Steeves et al., 2009), associations between ventral striatal dopamine release and self-reported ICBs (Song et al., 2022), and decreased ventral striatal D2/3 binding (Steeves et al., 2009; Payer et al., 2015; Stark et al., 2018b). Recent data also suggest that ICBs may be related to a derangement of the typical relationship between mesolimbic D2 autoreceptor expression and ventral striatal dopamine release (Song et al., 2022), where ICB+ subjects exhibited the absence of a negative relationship between midbrain D2/3 and ventral striatal DA release which was present in ICB– subjects.
Few studies have evaluated the relationship between D2/3 binding and a general measure of impulsivity (as opposed to mixed impulsivity–compulsivity as measured by ICB severity scales) in a PD population. A key early, albeit small [11C] FLB-457 PET study reported a positive correlation between impulsivity assessed with the BIS-11 and D2/3 binding in a ROI encompassing parts of the left subgenual cingulate and neighboring OFC (Ray et al., 2012). No studies have evaluated whether these associations differ between healthy control (HC) and PD groups. In addition, the inability of [11C]FLB-457 to derive stable estimates of D2/3 receptor availability in D2/3 receptor-rich regions hinders the ability to address associations in the striatum (Olsson et al., 1999). Thus, it remains unclear how the circuit localization of the D2/3 impulsivity relationship shifts in PD.
As PD develops, changes occur among multiple DA parameters because of the regionally specific patterns of DA neuron loss, as well as compensatory processes arising in response to either DA losses or dopamine treatments. As a result, the mechanisms of impulsivity regulation may shift to different nodes within this network. [18F]fallypride PET effectively measures both striatal and extrastriatal D2/3 binding (Kessler et al., 2000), allowing for the interrogation of the D2/3–impulsivity relationship across the mesocorticolimbic network. We hypothesized a change in the D2/3–impulsivity relationship in PD; specifically, the loss of negative associations in the midbrain and ventral striatum that characterize younger HCs. Based on the limited prior [11C]FLB-457 data, we also tentatively hypothesized the emergence of a positive ventromedial frontal cortical D2/3–impuslivity relationship in PD. We evaluated these hypotheses by determining [18F]fallypride PET D2/3 nondisplaceable binding potential (BPND) in a cohort of PD patients and age-matched HC subjects.
Materials and Methods
Participants
Participant selection methods followed those reported in Stark et al. (2018b). Participants with PD were recruited from the Movement Disorders Clinic at Vanderbilt University Medical Center. All met U.K. Brain Bank PD diagnosis criteria (Gibb and Lees, 1988). All were receiving dopamine replacement therapy, including levodopa and DA agonist medications, with daily doses recorded in levodopa equivalents (Tomlinson et al., 2010). Patients were excluded if they had an implanted deep brain stimulator, received antipsychotic medications or experienced symptoms of psychosis, suffered from comorbid psychiatric, cerebrovascular, or cardiovascular disease, or could not tolerate an MRI/PET and dopaminergic medication withdrawal. HC subjects, described by Dang et al. (2016, 2017) had no history of psychiatric illness, head trauma, substance abuse, diabetes, or medical condition that precluded MRI. No participants took psychostimulant or psychotropic medications (with an exception for occasional benzodiazepine use as sleep medication) over the preceding 6 months, consumed excessive alcohol, or used tobacco. All participants completed urine drug tests to ensure the absence of amphetamine, barbiturates, cocaine, marijuana, or opiates.
A neurologic examination was performed on all participants. HC and PD participants completed the BIS-11, a questionnaire designed to assess impulsiveness (Patton et al., 1995). PD patients completed the BIS-11 with the help of a caregiver or partner to provide additional behavioral insight, and before withholding of DA replacement therapy. PD patients completed the Movement Disorders Society-United Parkinson's Disease Rating Scale Part III (a clinician assessment of acute motor function). Before PET and MRI imaging, DA agonists and levodopa were withheld for >40 h and >16 h, respectively (based on considerations of the half-lives of carbidopa-levodopa [1.5 h; 2.5 h when combined with entacapone] ropinirole [6 h], and pramipexole [8-12 h]) (Fabbrini et al., 1987; Wright et al., 1997; Bennett and Piercey, 1999; Hauser, 2004; Tompson and Oliver-Willwong, 2009). Nineteen PD subjects participated in an additional intervention described by Song et al. (2022); these subjects received either placebo or study drug, but only the initial placebo scans are included in the current study. The other 35 PD subjects did not receive a placebo intervention. Dementia was screened by administration of the Montreal Cognitive Assessment (Nasreddine et al., 2005), with a cutoff score of 22. In PD patients, depression was screened using the Center for Epidemiologic Studies Depression Scale Revised (Radloff, 1977). The presence of medication-induced ICBs was assessed using a semistructured interview with patient and partner. Written informed consent was obtained from all subjects, and the study was performed in accordance with the Institutional Review Board at Vanderbilt University Medical Center, adhering to the ethical standards stipulated by the Declaration of Helsinki and its amendments.
Demographic and clinical features for PD patients (n = 54; M = 36; F = 18) and HC subjects (n = 31; M = 21; F = 10) are presented in Table 1. Both groups had similar average age and sex distributions. In PD subjects, symptom progression was moderate (average disease duration of 5.6 ± 3.6 years; Hoehn & Yahr Stages II-III).
Demographic and clinical characteristics of the two participant groupsa
Table 1-1
PET acquisition time schema. Download Table 1-1, DOCX file.
Table 1-2
PET acquisition time parameters. Acquisition parameters related to DY period start times and break lengths for both groups. Download Table 1-2, DOCX file.
Fallypride PET data acquisition
[18F]fallypride PET imaging and analysis procedures followed those reported by Stark et al. (2018b). [18F]fallypride was synthesized in the radiochemistry laboratory following procedures outlined by U.S. Food and Drug Administration INDs 47245 and 120035. Data for 66 subjects (35 PD and 31 HC) were collected on a GE Discovery STE PET/CT scanner; data for 19 PD subjects were acquired on a Philips Vereos PET/CT scanner with a 3D emission acquisition and a transmission attenuation correction. Serial scan acquisition began simultaneously with a 5.0 mCi slow bolus injection of [18F]fallypride (specific activity > 3000 Ci/mmol). CT scans were collected before each of the three emissions scans for attenuation correction. Scans lasted ∼3.5 h. Two 15-20 min breaks, beginning ∼70 min and 135 min after the beginning of the scan, were included for patient comfort. Data for PD and HC subjects were acquired using identical PET technical parameters, with the single exception of a slightly different PET acquisition time protocol for the second and third dynamic runs. Although total scan and dynamic run duration were the same, at 3000 and 3200 s for the second and third dynamic runs, respectively, the protocol in the PD group used less numerous frames of longer length (with two 1500 s vs four 750 s frames in the PD and HC groups, respectively, in the second run, and two 1800 s vs three 1200 s frames in the PD and HC groups, respectively, for the third dynamic run; for further description of dynamic structure, see Extended Data Tables 1-1 and 1-2; for description of a comparable HC acquisition protocol, see Dang et al., 2016, 2017). In past [18F]fallypride studies, these subtle acquisition time protocol differences did not significantly impact estimates of BPND or introduce a confound, as kinetic modeling accounts for decay along the time-activity curve (Smith et al., 2016). PET acquisition protocols using these validated procedural outlines have proven sufficient to reliably quantify D2/3 BPND in striatal and extrastriatal regions (Kessler et al., 2000; Mukherjee et al., 2002).
Fallypride PET data processing
Following attenuation correction and decay correction, serial PET scans were coregistered using Statistical Parametric Mapping software (SPM8, Wellcome Trust Center for Neuroimaging; http://www.fil.ion.ucl.ac.uk/spm/software/) to correct for motion with the last dynamic image of the first series serving as the reference image. The mean PET image produced by realignment was then coregistered to the corresponding high-resolution T1 MRI using FSL's FLIRT with 6 degrees of freedom (FSL version 5.0.2.1, FMRIB).
D2/3 receptor levels were estimated using the simplified reference tissue model (Lammertsma et al., 1996), with PMOD software (PMOD Technologies) to measure [18F]fallypride binding potential (BPND; the ratio of specifically bound [18F]fallypride to its nondisplaceable concentration as defined under equilibrium conditions). Voxel-wise estimates were generated using a published basis function fitting approach (Gunn et al., 1997) conducted in the PXMOD module. Rate constants were specified as k2a minimum = 0.006 min−1 and k2a maximum = 0.6 min−1. Because of limited D2/3 expression in the cerebellum (Camps et al., 1989), it was selected as the reference region (Kessler et al., 2000). Subject space analyses were conducted by registering baseline BPND images to T1 space with the saved FSL FLIRT transform matrices. For voxel-wise analyses, subject space BPND images were registered to MNI space (2 mm3) using FSL's FNIRT (FSL version 5.0.2.1, FMRIB).
ROI delineation
Subcortical ROIs, including the bilateral caudate (head), putamen (whole body; precommissural and postcommissural), ventral striatum, amygdala, midbrain (ventral portion/substantia nigra), and cerebellum, were manually segmented on the T1-weighted MRI scans by a neurologist (D.O.C.) experienced in PET and MRI data analysis and transferred to the coregistered PET studies through the FLIRT FSL transformation matrix. Manual segmentation methods followed established anatomic criteria, capturing the central portion of the selected region to gather the most representative sample of voxels while limiting potential partial volume effects. This method was selected to limit partial volume contamination in densely arranged subcortical nuclei, and was applied so as to avoid the potential confound of intersubject structural variability (Kessler et al., 2009). The caudate, putamen, and globus pallidus were manually drawn on axial slices ∼2-12 mm above the anterior commissure-posterior commissure (ACPC) line. The ventral striatum was segmented on coronal slices with the criteria of Mawlawi et al. (2001). The amygdala can be identified on axial slices 6-20 mm below the ACPC line, 12-28 mm lateral to the midline, and 2-12 mm behind the plane of the anterior commissure (Schaltenbrand and Wahren, 1998). To minimize partial volume contamination of the amygdala from the overlying striatum, amygdala ROIs were defined 10-16 mm beneath the ACPC plane. The midbrain was drawn on axial slices in 9-14 mm below the ACPC line capturing the ventral portion of the structure, and the thalamus was segmented 2-12 mm above the ACPC line (Schaltenbrand and Wahren, 1998). The cerebellar ROI was drawn centrally to include the vermis (a region with low levels of D2/3) and to avoid partial volume contamination from the midbrain or temporal cortex. It contained an approximately equal distribution of gray and white matter.
MRI
MRI scanning was performed with a Philips Ingenia system at 3.0T (Philips Healthcare) in the off-medication state (during the same visit in which [18F]fallypride data were captured). Subjects underwent a high-resolution anatomic T1-weighted scan (3D MPRAGE; spatial resolution = 1 × 1 × 1 mm3; TR/TE = 8.9/4.6 ms) and T2-weighted FLAIR (spatial resolution = 1 × 1 × 1 mm3; TR/TE = 4000/120 ms). These were assessed for incidental findings by a neurologist (D.O.C.).
Experimental design and statistical analysis
Group differences in demographic parameters were evaluated using Mann–Whitney U tests. Sex distribution was evaluated with a Fisher's exact test.
To test the hypothesis that PD patients express a distinct relationship between impulsivity and dopamine D2/3 receptor expression in subcortical areas, mean group regional subcortical [18F]fallypride BPND was analyzed via a general linear regression model (GLM), where BIS-11 Total score was the dependent variable, within-ROI BPND, and PD status were the independent variables, and age, UPDRS Part III (off-medication), and PET scanner (GE Discovery vs Philips Vereos) were covariates because of previous evidence that these factors influence D2/3 receptor status (Antonini et al., 1997; Mukherjee et al., 2002) and/or BPND. The interaction between BPND and PD status was the effect of interest. HC subjects were assigned “0” for UPDRS Part III score. Separate interaction analyses were conducted for the BIS-11 attentional, motor, and Nonplanning Impulsiveness subscales. Predefined subcortical ROIs for this analysis included the caudate, putamen, ventral striatum, amygdala, and midbrain. No variable centering was completed. For ROIs exhibiting a significant interaction term, separate descriptive Pearson's correlations were then conducted, covarying for age, PD motor symptom severity off-medication (UPDRS Part III), and PET scanner, to quantify distinct within-group relationships between BIS-11 Total score/subscore and subcortical ROI BPND. False discovery rate (FDR) was controlled at 0.05 to correct for multiple comparisons in all subcortical analyses. Subcortical ROIs included in correction were n = 5 for the GLM interaction analysis, where correction was separate for each BIS-11 subscale; n was variable for post hoc correlations, as two analyses (one per patient group) were completed with each significant interaction term. All analyses were performed using SPSS Statistics 24 (IBM) and R.
A voxel-wise correlation analysis was used to investigate the relationship between BIS-11 Total score and D2/3 BPND across cortical regions. This method was preferred in cortical areas because of the lack of strong a priori cortical hypotheses, to account for subregional heterogeneity, and to limit the number of ROIs assessed, thereby preserving statistical power. Age, PD motor symptom severity off-medication (UPDRS Part III), and scanner were included as covariates. Significance criteria consisted of an uncorrected p < 0.001 for magnitude and an extent threshold of 50 voxels, with multiple comparisons correction using cluster-level FDR at 0.05. This analysis was repeated for BIS-11 subscales, including Attentional Impulsiveness, Motor Impulsiveness, and Nonplanning Impulsiveness, using identical statistical parameters. Any clusters produced by the cortical voxel-wise analysis (for either group) were used to define ROIs in standard space as well as subject space, to provide scatterplot visualization and to confirm that results were present in both standard and subject space, and not misattributed or because of partial voluming of nearby areas. Since the cortical ROIs were defined based on clusters that were significant, these analyses were not intended to provide an estimate of the true unbiased effect size. These derived cortical ROIs were then analyzed with GLM and partial correlation analyses in parallel to subcortical ROIs, and were included in multiple comparisons correction where appropriate.
Although all dopaminergic medication was held acutely for scan procedures, a supplementary analysis was completed to assess whether any observed regional relationships between BPND and impulsivity in the PD group were mediated in part by chronic dopaminergic medication dosage. To complete this assessment, a GLM was specified, where BIS-11 score was the dependent variable, within-ROI BPND, medication dose (either dopamine agonist single dose equivalent [SDE] or levodopa equivalent daily dose [LEDD]) were the independent variables, and age, UPDRS Part III (off-medication), and PET scanner were covariates. The interaction between dose and within-ROI BPND on BIS-11 score was the effect of interest for these analyses, as it reflects the influence of medication effects on the D2/3-trait impulsivity relationship. Agonist SDE and LEDD were assessed in separate models. BIS-11 Total score and all subscales were assessed in the same manner. This process was completed for both a priori subcortical regions and any cortical regions defined via the exploratory voxelwise analysis.
Results
Impulsivity differences in PD
The PD group was characterized by significantly elevated BIS-11 Total score (p = 0.01; PD average = 62.3 ± 10.3; HC average = 56.4 ± 10.4), BIS-11 Attentional Impulsiveness score (p < 0.01; PD average = 16.8 ± 3.8; HC average = 14.6 ± 3.5), and BIS-11 Nonplanning Impulsiveness score (p = 0.01; PD average = 24.1 ± 5.5; HC average = 21.0 ± 4.9). This is consistent with prior results from a larger PD cohort, which this sample partially composes (Aumann et al., 2020). While there is no strictly defined BIS-11 score threshold to indicate clinically significant impulsivity, scores observed here constitute a pattern consistent with other behavioral phenotypes characterized by mild to moderate impulsive behavior (Stanford et al., 2009).
Subcortical [18F]fallypride BPND and impulsivity relationship differences in PD
Analysis of total BIS-11 scores revealed significant interaction between group and ventral striatal BPND (unstandardized B = −3.40, p = 0.004): a significant negative correlation (controlling for age, UPDRS Part III, and scanner where appropriate) was present in the HC but not the PD group (HC: partial r = −0.48, p = 0.008; PD: partial r = 0.22, p = 0.13). A similar finding emerged for the amygdala (unstandardized B = −17.04, p = 0.007), with a similar correlation pattern (PD: partial r = 0.19, p = 0.17; HC: partial r = −0.41, p = 0.02). No significant interaction terms were observed for the caudate, putamen, or midbrain. Figure 1 presents scatterplots of BPND versus BIS-11 Total score for the amygdala and ventral striatum, respectively, after controlling for age, UPDRS Part III, and scanner. Visualization of the segmentation protocol for all subcortical regions is displayed in Extended Data Figure 1-1, while a full report of analysis results (including nonsignificant findings) is presented in Extended Data Figure 1-2.
Mean [18F]fallypride binding potential versus BIS-11 Total score. Scatterplots of the partial correlation between BPND (x axis) and BIS-11 Total score (y axis; unstandardized residuals covarying for age, UPDRS Part III, and PET scanner). Significant group × BPND interaction terms were observed in the amygdala (A,B) and ventral striatum (C,D). Scanner type is depicted, where data for 19 PD subjects were acquired using a Philips Vereos PET/CT scanner; and data for 35 PD and 31 HC subjects were acquired using a GE Discovery STE PET/CT scanner. Visualization of the segmentation protocol for all subcortical regions is displayed in Extended Data Figure 1-1, while a full report of analysis results (including nonsignificant findings) is presented in Extended Data Figure 1-2. Analysis of the effect of chronic dopaminergic medication dose on the relationship between regional BPND and BIS-11 score is reported in Extended Data Figure 1-3. * indicates uncorrected p<0.05; ** indicates uncorrected p<0.01.
Figure 1-1
Manual segmentation protocol for subcortical [18F]fallypride binding potential analysis. (A-E) Representative axial, sagittal, and coronal slices for a single subject exhibiting the manual segmentation routine for the five subcortical structures included analyses: (A) caudate, (B) putamen, (C) ventral striatum, (D) midbrain, and (E) amygdala. Download Figure 1-1, TIF file.
Figure 1-2
Results of group × BPND interaction on BIS-11 Score. Unstandardized B values are listed alongside p values for the interaction term of the general linear model, which included BIS-11 score as the dependent variable, and group, age, UPDRS Part III, scanner, and ROI BPND as independent variables. The interaction between group and ROI BPND was the effect of interest. For ROIs expressing a significant interaction term, post-hoc partial Pearson's correlation analyses were conducted for each group in order to better define BIS-11 score-BPND relationships, covarying for age, PD severity (UPDRS Part III), and scanner. The correlation coefficients and p values for these analyses are also reported. This process was completed for BIS-11 Total score, as well as the Attentional Impulsiveness, Motor Impulsiveness, and Nonplanning Impulsiveness subscales. Download Figure 1-2, DOCX file.
Figure 1-3
Medication dose × BPND interaction on BIS-11 score (subcortical analysis). Unstandardized B values are listed alongside p values for the interaction term of a supplementary general linear model, which included BIS-11 score as the dependent variable, and age, UPDRS Part III, scanner, ROI BPND, and medication dose (either dopamine agonist single dose equivalent [SDE] or levodopa equivalent daily dose [LEDD]) as independent variables. The interaction between medication dose and ROI BPND was the effect of interest. Results for all subcortical ROIs are included. Download Figure 1-3, DOCX file.
Subsequent analysis of BIS-11 subscales uncovered significant interactions for several ROIs. A significant interaction involving Motor Impulsiveness was present in the ventral striatum (unstandardized B = −1.088, p = 0.003; PD: partial r = −0.03, p = 0.84; HC: partial r = −0.59, p = 0.001). For Attentional Impulsiveness, this included the amygdala (unstandardized B = −6.48, p = 0.003; PD: partial r = 0.16, p = 0.27; HC: partial r = −0.47, p = 0.009); the ventral striatum exhibited a marginally significant interaction term that did not pass multiple comparisons correction (unstandardized B = −0.817, p = 0.037; PD: partial r = 0.10, p = 0.48; HC: partial r = −0.33, p = 0.08). Finally, for Nonplanning Impulsiveness, the ventral striatum (unstandardized B = −1.44, p = 0.018; PD: partial r = 0.32, p = 0.02; HC: partial r = −0.34, p = 0.06) and amygdala (unstandardized B = −9.01, p = 0.005; PD: partial r = 0.16, p = 0.26; HC: partial r = −0.46, p = 0.01) exhibited significant interaction terms. No significant interaction terms were observed for the caudate, putamen, or midbrain. A full report of BIS-11 subscale analysis results (including nonsignificant findings) is presented in Extended Data Figure 1-2. Broadly, a pattern of negative correlation between BPND and BIS-11 subscale score was present in control, but not PD, subjects.
To ensure that the PET scanner system did not introduce an additional confounding effect despite its inclusion as a GLM covariate, the subcortical analysis was repeated exclusively with subjects scanned using the GE Discovery STE system (PD: n = 35, HC: n = 31); results were broadly consistent with full group findings. These included a significant interaction between group and ventral striatal BPND on BIS-11 Total (unstandardized B = −3.369, p = 0.004; PD: partial r = 0.22, p = 0.23; HC: partial r = −0.48, p = 0.008) and Nonplanning Impulsiveness (unstandardized B = −1.615, p = 0.006; PD: partial r = 0.37, p = 0.03; HC: partial r = −0.34, p = 0.06) scores, significant interactions between group and amygdala BPND on BIS-11 Total (unstandardized B = −16.018, p = 0.02; PD: partial r = 0.11, p = 0.60; HC: partial r = −0.41, p = 0.02) and Nonplanning Impulsiveness (unstandardized B = −9.827, p = 0.005; PD: partial r = 0.17, p = 0.35; HC: partial r = −0.46, p = 0.01) scores, and marginally significant interactions between group and amygdala/ventral striatal BPND on Attentional Impulsiveness scores which did not pass multiple comparisons correction.
Analysis of the effect of chronic dopaminergic medication dose on the relationship between regional BPND and BIS-11 score was unrevealing; for both dopamine agonist SDE and LEDD, the dose × BPND interaction on BIS-11 score was not significant for any region or any BIS-11 subscale. Full subcortical results are reported in Extended Data Figure 1-3.
Cortical [18F]fallypride BPND and impulsivity relationship differences in PD
Using the voxel-wise method covarying for age, UPDRS Part III, and PET scanner, a significant positive relationship between [18F]fallypride BPND and BIS-11 Total score was present in an area encompassing the caudal aspect of the gyrus rectus, the medial orbital gyrus, and the posterior orbital gyrus, that is, in the caudal OFC (cluster MNI coordinates [mm]: −18, 8, −22; 10, 12, −18; qFDR: 0.14; 0.14; cluster threshold: 77) (Ongur and Price, 2000; Ongur et al., 2003) of PD patients but not control subjects (Fig. 2). A similar positive correlation cluster pattern in the bilateral caudal OFC was evident for the Attentional Impulsiveness subscale in PD subjects (cluster MNI coordinates [mm]: −18, 6, −22; 12, 10, −18; qFDR: 0.006; 0.006; cluster threshold: 100) (Fig. 3). No significant clusters were observed for motor or Nonplanning Impulsiveness. The significant clusters observed in the BIS-11 Total and Attentional Impulsiveness analyses were used to directly define new standard space caudal OFC ROIs, where ROI boundaries exactly reflected those of the significant clusters for each analysis (for cluster/ROI visualization, see Figs. 2, 3). In parallel to the subcortical analyses, a GLM was specified wherein BIS-11 Total score was the dependent variable, mean ROI BPND, and PD status were the independent variables, and age, UPDRS Part III, and scanner were covariates. BPND was defined separately for BIS-11 Total score and the Attentional Impulsiveness subscale, as the unique voxelwise cluster-derived ROI was used for each scale, respectively. Significant interaction terms were present for the BIS-11 Total cluster-derived ROI (unstandardized B = −4.406, p = 0.007; PD: partial r = 0.52, p < 0.001; HC: partial r = −0.08, p = 0.69) as well as the BIS-11 Attentional Impulsiveness cluster-derived ROI (unstandardized B = −1.620, p = 0.003; PD: partial r = 0.53, p < 0.001; HC: partial r = −0.10, p = 0.609). These results passed multiple comparisons correction when included with relevant subcortical ROIs for BIS-11 Total and Attentional Impulsiveness. Because no significant voxelwise result and subsequent ROI were defined for the Motor Impulsiveness or Nonplanning Impulsiveness subscales, they did not undergo further analysis.
Voxel-wise [18F]fallypride correlation with BIS-11 Total. A, Map of a significant cluster in the caudal OFC where [18F]-fallypride BPND expressed a positive relationship with BIS-11 Total score, overlaid on an MNI template brain. This cluster survived FDR correction at 0.05, and localized to the caudal OFC. This cluster was used to define an ROI in standard space. Subsequently, mean BPND values were extracted. Scatterplots of the partial correlation between BPND (x axis) and BIS-11 Total score (y axis; unstandardized residuals covarying for age, UPDRS Part III, and scanner) express a positive correlation in the PD group (B) that is not expressed in the HC group (C). Scanner type is depicted, where data for 19 PD subjects were acquired using a Philips Vereos PET/CT scanner; and data for 35 PD and 31 HC subjects were acquired using a GE Discovery STE PET/CT scanner. Visualization of the hand-drawn subject space caudal OFC ROI is provided in Extended Data Figure 2-1.
Figure 2-1
Supplementary Figure 8. Subject space caudal OFC [18F]fallypride binding potential versus BIS-11 Total/Attentional Impulsiveness score. (A) Representative axial, sagittal, and coronal slices for a single subject exhibiting the manual segmentation routine for the subject space caudal OFC ROI. This ROI was defined based on the results of the exploratory voxelwise analysis for BIS-11 Total and Attentional Impulsiveness scores. Scatterplots of the partial correlation between subject space caudal OFC ROI BPND (x axis) and BIS-11 Total score (y axis; unstandardized residuals covarying for age, UPDRS Part III, and scanner) express a positive correlation in the PD group (B) that is not expressed in the HC group (C). A similar pattern was exhibited for BIS-11 Attentional Impulsiveness for the PD (D) and HC (E) groups. Scanner type is depicted, where data for 19 PD subjects were acquired using a Philips Vereos PET/CT scanner; and data for 35 PD and 31 HC subjects were acquired using a GE Discovery STE PET/CT scanner. Download Figure 2-1, TIF file.
Voxel-wise [18F]fallypride correlation with BIS-11 Attentional Impulsiveness. A, Map of significant clusters where [18F]-fallypride BPND expressed a positive relationship with BIS-11 Attentional Impulsiveness score, overlaid on an MNI template brain. Clusters survived FDR correction at 0.05, and localized to the caudal OFC. These clusters were used to define an ROI in standard space. Subsequently, mean BPND values were extracted. Scatterplots of the partial correlation between BPND (x axis) and BIS-11 Total score (y axis; unstandardized residuals covarying for age, UPDRS Part III, and scanner) express a positive correlation in the PD group (B) that is not expressed in the HC group (C). Scanner type is depicted, where data for 19 PD subjects were acquired using a Philips Vereos PET/CT scanner; and data for 35 PD and 31 HC subjects were acquired using a GE Discovery STE PET/CT scanner.
To ensure that the PET scanner system did not introduce an additional confounding effect in the cortical analysis despite its inclusion as a voxelwise covariate, the analysis was repeated exclusively with subjects scanned using the GE Discovery STE system (PD: n = 35, HC: n = 31); results were broadly consistent with full group findings. The BIS-11 Total and Attentional Impulsiveness analyses produced a significant positive correlation cluster in the PD group that localized to the same caudal OFC area as the main analysis, although the BIS-11 Total cluster was unilateral rather than bilateral in this subcohort (cluster MNI coordinates [mm]: −16, 8, −22; qFDR: 0.005; cluster threshold: 131). The BIS-11 Attentional Impulsiveness cluster was bilateral, paralleling the results of the main analysis (cluster MNI coordinates [mm]: −20, 6, −20; 12, 10, −18; qFDR: 0.002; 0.018; cluster threshold: 69). Following standard space ROI extraction, significant interaction terms were present for the GE-only BIS-11 Total cluster-derived ROI (unstandardized B = −6.618, p = 0.001; PD: partial r = 0.64, p < 0.001; HC: partial r = −0.10, p = 0.59) as well as the GE-only BIS-11 Attentional Impulsiveness cluster-derived ROI (unstandardized B = −1.835, p = 0.002; PD: partial r = 0.66, p < 0.001; HC: partial r = −0.08, p = 0.68).
Subsequently, to confirm that the effect observed in the main full-cohort voxel-wise analyses was consistent in subject space as well as standard space, a new hand-drawn ROI was defined that captured the caudal OFC in subject space. This ROI was larger than the cluster-derived standard space ROI, capturing the posterior half of the gyrus rectus and posterior potions of the medial and posterior orbital gyri. Partial Pearson's correlations, covarying for age and PD severity, confirmed that, in PD but not HC subjects, a positive correlation was present between subject space caudal OFC BPND and BIS-11 Total (PD: r = 0.39, p = 0.004; HC: r = 0.09, p = 0.65) and Attentional Impulsiveness scores (PD: r = 0.33, p = 0.02; HC: r = 0.20, p = 0.29). Visualization of the hand-drawn subject space caudal OFC ROI is provided in Extended Data Figure 2-1. Because of the potential confounds of partial volume contamination between the ventral striatum and caudal OFC cluster, partial correlations were repeated in the PD group using ventral striatal BPND as an additional covariate, again revealing significant positive correlations between BIS-11 and BPND derived from both standard and subject space caudal OFC ROIs (standard space PD: r = 0.48, p < 0.001; subject space PD: r = 0.34, p = 0.018). Finally, visual inspection of voxel-wise t value maps verified that the correlation was restricted to the observed caudal OFC cluster, and subthreshold separate clusters were not present in other regions, such as the ACC or the dorsolateral PFC. No significant clusters were identified for other subscales, or for any subscale within the HC group.
Analysis of the effect of chronic dopaminergic medication dose on the relationship between cortical BPND and BIS-11 score was also unrevealing; for both dopamine agonist SDE and LEDD, the dose × BPND interaction on BIS-11 score was not significant for any cortical region (either in standard space or subject space) or any BIS-11 subscale. All p values were >0.20.
Discussion
We observed divergent impulsivity–D2/3 associations in HC and PD cohorts, where HCs exhibited negative BIS-11/BPND relationships along mesolimbic regions (ventral striatum and amygdala) absent in PD. Conversely, a positive BIS-11/BPND association localized to the caudal OFC in PD but not HCs, suggesting a role of heightened importance for ventromedial mesocortical dopamine. This could constitute a compensatory response to PD-induced neurodegeneration or dopaminergic treatments.
Mesoaccumbal dopamine and impulsivity
Ventral striatal D2/3 is a key target of the mesoaccumbal circuit. PET studies indicate an association between impulsivity-related traits and reduced ventral striatal D2/3 BPND (Dalley et al., 2007; Trifilieff and Martinez, 2014). Additionally, negative correlations are reported between ventral striatal D2/3 BPND/BIS-11 in a healthy cohort (Reeves et al., 2012), methamphetamine users (Lee et al., 2009), and pathologic gamblers (Clark et al., 2012). We observed a significant group × BPND interaction in the ventral striatum, where D2/3/BIS-11 correlations were absent in PD but present in HCs. This negative impulsivity–BPND relationship is similar to that seen in the midbrain (Buckholtz et al., 2010); and consistent with prior ventral striatal receptor-behavior relationships (Reeves et al., 2012; London, 2020). The lack of PD findings likely relates to dopaminergic denervation. While mean ventral striatal [18F]fallypride BPND is comparable between HC/PD, and ventral striatal projections are less vulnerable than dorsal projections (Gibb and Lees, 1991; Stark et al., 2018a), this region experiences losses masked by postsynaptic D2/3 receptor upregulation (Falardeau et al., 1988). These changes may be sufficient for PD patients to lose an important D2/3-mediated mechanism of trait impulsivity regulation.
Motor Impulsiveness (reflecting acting without thinking) and Nonplanning Impulsiveness (difficulty deferring gratification) emerged as significant subscale interactions for the ventral striatum. Motor Impulsiveness exhibited a negative association with HC ventral striatal DA transporter binding previously (Smith et al., 2018), where low DA transporter could indicate decreased DA regulatory capacity, and high Motor Impulsiveness could reflect difficulty modulating behavior at the limbic-motor interface (motivation-action translation). Here, a negative relationship in HCs may similarly reflect DA release properties influencing the limbic-motor interface, which become notable in older populations. In PD, derangements of the mesoaccumbal tract could disrupt this coupling.
A prior PET study reported a negative relationship between BIS-11 score and midbrain D2/3 BPND in young adult HCs (Buckholtz et al., 2010). We observed only a nonsignificant trend for this association in our older HC population, and the group × BPND interaction in the midbrain was not statistically significant. As such, the present data emphasize the ventral striatum over the midbrain as a critical location where PD alters D2/3 impulsivity relations.
Amygdala impulsivity regulation
In the BLA, D2/3 receptors regulate reward learning processes (Di Ciano and Everitt, 2004; Berglind et al., 2006; Thiel et al., 2010) as well as associative learning and fear extinction (de Oliveira et al., 2011; Shi et al., 2017). The central nucleus is another amygdalar structure of interest, where D2 KOs exhibit elevated impulsivity extinguished by restoring local D2 (Kim et al., 2018). Impulsivity–D2/3 correlations have not been observed in humans and were notably absent in prior work (Buckholtz et al., 2010); however, differences in both cohort age and ROI protocols are notable.
Paralleling the ventral striatum, the negative HC relationship could be influenced by presynaptic nerve terminals regulating local DA release. Postsynaptic receptors on BLA interneurons (Kroner et al., 2005) or in the central nucleus (Scibilia et al., 1992) are also relevant. The results here suggest that amygdalar D2/3 may relate to a notable behavioral parameter in the healthy brain. The lack of a relationship in the PD cohort is likely because of PD-amygdala pathology, including Lewy body inclusion susceptibility (Braak et al., 1994) and dopaminergic projection loss. Past studies have tied this degeneration to the development of PD depression, where symptom severity was associated with amygdala volume (van Mierlo et al., 2015) and function (Sheng et al., 2014).
Attentional Impulsiveness (associated with difficulty concentrating and racing or extraneous thoughts) and Nonplanning Impulsiveness (difficulty deferring gratification) emerged as amygdalar subtraits of interest. While the amygdala is not central to either cognitive control or sustained attention networks, it is necessary for attending to stimuli with significant emotional or reward valence (Jacobs et al., 2012; Zhang et al., 2021), encoded via dopaminergic signaling (Lutas et al., 2022). This function could be captured in the HC relationship. Increased Attentional Impulsiveness appears to be characteristic of PD, where high scores on the subscale may drive elevated BIS-11 Total scores relative to HCs (Aumann et al., 2020). This increase could be because of the loss of mechanisms of behavioral regulation normally administered by D2 in amygdala (Kim et al., 2018).
Cortical impulsivity regulation
Cortical areas also receive midbrain dopaminergic input (Thierry et al., 1973), with considerable D2/3 expression in ventromedial frontal cortical areas (Lidow et al., 1989). While we observed no relationships in HCs, PD subjects exhibited a positive relationship between trait impulsivity and D2/3 BPND in the caudal-medial OFC, which we consider analogous to the olfactocentric subdivision of the paralimbic cortex (olfactocentric-paralimbic cortex). This cluster broadly included the caudal straight gyrus (gyrus rectus), portions of the posterior orbital gyrus, and the inferior segment of the subcallosal area. This corresponds to portions of Brodmann areas 13, 14c, and 25, and is a histologic transition zone from agranular to granular cortex (Mesulam and Mufson, 1982; Kringelbach and Rolls, 2004; Mackey and Petrides, 2009). Some human studies have produced similar results but do not perfectly parallel the region exhibited here. One prior [11C]FLB-457 PET study of ICB+ PD patients reported a positive BIS-11/frontal cortical D2/3 BPND correlation (Ray et al., 2012); however, the association observed by Ray et al. (2012) was based on an ROI encompassing a more anterior and superior region within the OFC and subgenual ACC, and thus did not test for association in the precise cluster observed here.
Dopaminergic signaling in this region may be important in regulating switches between broader networks and associated modes of cognitive processing, via targeting segregated D1- or D2-expressing cortical cell populations with different projection targets. While D1/D2 are widely coexpressed in cortical areas, there are also considerable neuron populations with segregated expression (Gaspar et al., 1995; Jenni et al., 2017; Wei et al., 2018). In this manner, DA acting on infralimbic/caudal OFC D2 may facilitate cortex-amygdala projections (rather than cortex-accumbens projections) encoding reward contingencies and regulating cognitive flexibility (Jenni et al., 2017), and preferentially engages connectivity with specific cortical areas (Kahnt and Tobler, 2017). Through these networks, infralimbic/caudal OFC D2 is broadly associated with updating salience value through mechanisms, including behavioral flexibility in reward-seeking (Barker et al., 2013), which may affect the association between medial OFC activation and reward-based decision-making (Engelmann and Tamir, 2009; Blankenstein et al., 2017).
In PD, neurodegeneration and DA derangements could increase the importance of caudal-medial OFC D2/3 to trait impulsivity. One potential mechanism is ventral striatal input regulation, where tonic DA maintains limbic-cortical balance. Increases in ventral striatal tonic DA attenuate cortical inputs, while reductions facilitate them (Charara and Grace, 2003; Goto and Grace, 2005). In PD, dopaminergic projection loss, changes to DA release and regulation, and DA replacement could all alter tonic DA disturbing this balance and favoring PFC input. D2/3 agonist medications could further affect this balance by favoring D2-expressing cortical projections. Combined, the relevance of caudal OFC D2/3 to behavior may increase. The likelihood of this mechanism is unclear given the multifactorial complexity and lack of distinct ventral striatal decrements (despite likely regional DA changes). Dopaminergic medication dose did not mediate impulsivity–D2/3 relationships in our analyses, but chronic effects might not be captured in continuous dose vectors. The reason for a positive relationship between caudal OFC D2/3 and impulsivity (compared with negative subcortical relationships in HCs) is unclear; however, this may relate to differences in cortical versus subcortical DA release and receptor populations, or the involvement of the caudal-medial OFC activity in facilitating impulsivity-related behavior (e.g., risk-taking) (Blankenstein et al., 2017).
The amygdala-OFC axis is of particular interest given relevant subscale findings. PD exhibits a loss of the negative Attentional Impulsiveness relationship in the amygdala but the appearance of a new positive Attentional Impulsiveness relationship in the caudal OFC. This could indicate a pattern of compensation along related tract structures similarly involved in salience assignment and attentional regulation (Jacobs et al., 2012; Li et al., 2017), resulting in the difficulty concentrating that appears to drive elevations in BIS-11 Total score in PD patients (Aumann et al., 2020). However, this transition could also represent broader cortical dysfunction or early dysexecutive symptoms, which could impair attentional regulation in PD.
Future directions and conclusions
While a positive aspect of our investigation was PD cohort size, this group included a larger proportion (∼50%) of individuals with ICBs than would be anticipated (Garcia-Ruiz et al., 2014). However, as a study goal was assessment across the impulsivity spectrum, we considered this larger proportion valuable. Because of basal ganglia structure proximity, one possible concern is partial volume error. This was especially relevant for the caudal OFC, which was proximal to D2/3-dense regions. However, effects remained significant in supplemental analyses that including ventral striatal BPND as a covariate, ruling out partial volume confound. Another consideration is the absence of negative D2/3-BIS-11 correlations in the amygdala or ventral striatum in HCs (Buckholtz et al., 2010). Age differences and ROI delineation methods may be important, yet findings linking increased ventral striatal dopamine transporter expression to lower BIS support our results (Smith et al., 2018). Differences may also relate to low sample sizes because of the relative expense of PET. Data collection on two scanners is a concern partially assuaged by consistent findings despite subcohort exclusion and use of a “scanner” covariate (although only PD subjects were divided between scanners, potentially limiting covariate effectiveness). As a cross-sectional study, the current work cannot speculate on longitudinal D2/3 changes. Furthermore, PET methods do not allow disentanglement of D2/3 expression from competitive synaptic dopamine binding.
Overall, our findings emphasize that PD is associated with divergent patterns of D2/3-related behavioral regulation in a diverse set of limbic regions. These changes implicate a disease-related change in the balance between cortical and subcortical limbic dopamine networks. Future studies might assess these networks with rigorous dopaminergic system interrogation (e.g., amphetamine-evoked dopamine release) or multimodal imaging (e.g., [18F]fallypride PET with surrogate measures of cerebral metabolism [arterial spin-labeling MRI] or functional connectivity within mesocorticolimbic systems).
Footnotes
This work was supported by the National Institutes of Health/National Institute of Neurological Disorders and Stroke R01NS097783 and K23NS080988; National Institute on Aging R01AG044838; National Center for Advancing Translational Science CTSA Award UL1TR000445; and the Vanderbilt Medical Scholars Program. We thank the volunteers who participated in this study; and the Vanderbilt University Institute of Imaging Sciences, including the Human Imaging Core and Radiochemistry Core.
The authors declare no competing financial interests.
- Correspondence should be addressed to Daniel O. Claassen at daniel.claassen{at}vumc.org
