Is there a relation between novelty seeking, striatal dopamine release and frontal cortical thickness?

Background Novelty-seeking (NS) and impulsive personality traits have been proposed to reflect an interplay between fronto-cortical and limbic systems, including the limbic striatum (LS). Although neuroimaging studies have provided some evidence for this, most are comprised of small samples and many report surprisingly large effects given the challenges of trying to relate a snapshot of brain function or structure to an entity as complex as personality. The current work tested a priori hypotheses about associations between striatal dopamine (DA) release, cortical thickness (CT), and NS in a large sample of healthy adults. Methods Fifty-two healthy adults (45M/7F; age: 23.8±4.93) underwent two positron emission tomography scans with [11C]raclopride (specific for striatal DA D2/3 receptors) with or without amphetamine (0.3 mg/kg, p.o.). Structural magnetic resonance image scans were acquired, as were Tridimensional Personality Questionnaire data. Amphetamine-induced changes in [11C]raclopride binding potential values (ΔBPND) were examined in the limbic, sensorimotor (SMS) and associative (AST) striatum. CT measures, adjusted for whole brain volume, were extracted from the dorsolateral sensorimotor and ventromedial/limbic cortices. Results BPND values were lower in the amphetamine vs. no-drug sessions, with the largest effect in the LS. When comparing low vs. high LS ΔBPND groups (median split), higher NS2 (impulsiveness) scores were found in the high ΔBPND group. Partial correlations (age and gender as covariates) yielded a negative relation between ASTS ΔBPND and sensorimotor CT; trends for inverse associations existed between ΔBPND values in other striatal regions and frontal CT. In other words, the greater the amphetamine-induced striatal DA response, the thinner the frontal cortex. Conclusions These data expand upon previously reported associations between striatal DA release in the LS and both NS related impulsiveness and CT in the largest sample reported to date. The findings add to the plausibility of these associations while suggesting that the effects are likely weaker than has been previously proposed.


Results
BP ND values were lower in the amphetamine vs. no-drug sessions, with the largest effect in the LS. When comparing low vs. high LS ΔBP ND groups (median split), higher NS2 (impulsiveness) scores were found in the high ΔBP ND group. Partial correlations (age and gender as covariates) yielded a negative relation between ASTS ΔBP ND and sensorimotor CT; trends for inverse associations existed between ΔBP ND values in other striatal regions and frontal CT. In other words, the greater the amphetamine-induced striatal DA response, the thinner the frontal cortex.

Introduction
Impulsive novelty seeking (NS) traits have been proposed to reflect individual differences in meso-striatal dopamine (DA) transmission [1], [2] and aspects of cortical morphometry, including grey matter volume and cortical thickness (CT) [3], [4]. The model is compelling. Limbic DA transmission potently influences the incentive salience of rewarding and potentially rewarding (novel) stimuli [5], [6], [7], and cortical projections densely innervate the striatum [8], [9] further modulating the planning of approach and avoidance behaviors [1], [10]. In humans, individual differences in CT, striatal DA D 2 receptor availability and striatal DA responses to a drug challenge (e.g., amphetamine) have been reported to account for a substantial proportion of the variance in temperamental features such as NS (Tables 1-3). However, given the noise inherent to both neuroimaging data and personality trait measurements, it seems surprising that a single snapshot of a neurobiological feature can account for so much variation in temperament [11]. Since much of this work has been conducted in small samples, and rarely with the same measure of impulsivity, in the current study, we investigated-for the first time-the relation between NS, CT and amphetamine-induced striatal DA release (via regression analyses) in the largest sample of healthy adults reported to date. Based on the previous studies, we predicted, a priori, that greater DA release would be associated with higher impulsivity-related NS scores and thinner frontal cortices. Additionally, since the association between CT and DA release has only been previously reported once in a smaller cohort [12], we sought to replicate this finding in a substantially larger sample. The same is true for reproducing the association between striatal DA release and NS, which was reported only once previously [13].

Participants
Participants were healthy men (n = 45) and women (n = 7). Following a telephone screen, volunteers underwent an in-person physical exam, standard laboratory tests and a psychiatric assessment (Structured Clinical Interview for DSM-IV-TR [SCID-IV-TR], non-patient edition [36]). Inclusion criteria were as follows: a) absence of axis I disorder, b) no first-degree relative

Neuroimaging acquisition
PET scans were carried out on two separate days (inter-scan time: 23.6 days±26.9; range: 1-104 days). Participants were asked to abstain for a minimum of 7 h and 24 h from tobacco  and alcohol, respectively. Preliminary evidence suggests that there is not an effect of menstrual phase on [ 11 C]raclopride BP ND values [37], but females were tested during the follicular phase of their cycle. All participants tested negative on a urine drug screen (Triage Drugs of Abuse Panel, Biosite Diagnostics) prior to each scan. PET scans were acquired, with and without ingesting a capsule filled with amphetamine (0.3 mg/kg, p.o.) 60 min prior to tracer injection, with a Siemens ECAT HR+ PET scanner (CTI/Siemens, Knoxville, TN, USA; 63 slice coverage; maximum resolution of 4.2 mm; fullwidth half-maximum [FWHM] in the center of the field of view). In three of the studies, a placebo capsule was administered, while in two studies a baseline scan was obtained (i.e., no placebo capsule; study was used as a covariate, when appropriate, and details are provided in the following sections). Immediately after the transmission scan (for attenuation correction), [ 11 C]raclopride (8-10 mCi) was injected as a bolus into the antecubital vein. Emission data were collected over 60 min in 26 time frames of progressively longer durations.

PET analysis
Region of interest (ROI) analysis. The striatum was divided into 6 functional ROIs, defined on each participant's MRI (automated segmentation using in-house Automatic Nonlinear Image Matching and Anatomical Labeling [ANIMAL]; ROI-MRI overlap was visually inspected). ROIs were the left/right sensorimotor striatum (SMS; post-commissural putamen), associative striatum (ASTS; pre-commissural dorsal caudate and dorsal putamen, post-commissural caudate) and ventral limbic striatum (LS) [38]. Parametric images were generated by deriving [ 11 C]raclopride BP ND values from each ROI using a simplified kinetic model that uses the cerebellum as a reference tissue devoid of DA D 2/3 receptors to describe the kinetics of the free and specifically bound ligand [39]. Mean BP ND values were extracted from each ROI in the conditions with and without amphetamine. During rest, BP ND is proportional to the  [40]. Voxel-wise analysis. PET images were co-registered to MRIs and parametric [ 11 C]raclopride BP ND data were generated at each voxel using a simplified reference tissue model (cerebellum). Images were normalized into standardized (MNI) space. An 8 mm Gaussian smoothing kernel (FWHM) was applied to PET images (2 individuals excluded due to co-registration problems between BP ND images/conditions; N = 47). A t-map was created to assess changes in [ 11 C]raclopride BP ND between scans with and without amphetamine using a residual t-statistic (BP ND baseline>BP ND amphetamine) [41]. In this approach, residuals are used to calculate the variance in parameter estimates; this method has been used extensively and validated using Monte Carlo simulations and real PET data [41], [42]. Significant voxels were identified by thresholding the t-map at t>4.1 (p<.05, Bonferroni corrected for multiple comparisons, search volume of the entire striatum).

Cortical thickness (CT) analysis
MRIs were processed using the CIVET 2.0.0 pipeline with CBRAIN (https://cbrain.mcgill.ca/). The pipeline involves: a) Normalizing each image to standardized space (MNI ICBM-152 template). b) Correction of intensity non-uniformity artifacts. c) Tissue type classification. d) Fitting images with a deformable mesh model to extract 2D inner (white/gray matter interface) and outer (pial) cortical surfaces using the CLASP algorithm [43]. This generates CT measurements at 40,962 vertices/hemisphere (distance between outer CSF/gray matter and gray/white matter interfaces). e) Surfaces are registered to the MNI ICBM-152 surface template. f) Reverse linear transformations allow CT estimations in native space. g) CT is calculated at each cortical point using the t link metric [44]. h) Smoothing with a 20 mm FWHM kernel.

Statistics
Unless stated otherwise, means, standard deviations (SD) and partial eta-squares (η 2 partial : effect sizes) are presented. To characterize the effect of amphetamine-induced DA release, repeated-measures analysis of covariance (rmANOCVA; study as covariate since preliminary analyses revealed a main effect of study on BP values) was carried out on BP ND values with drug condition (no amphetamine, amphetamine) and striatal ROI (LS, ASTS, SMS) as withinsubject factors. Since no hemisphere effects were seen in preliminary analyses, it was not included as a factor. rmANOVAs were carried out on ΔBP ND values, with striatal ROI as the within-subject factor (study was not used as a between-subject factor since preliminary analyses yielded no main effect of study on ΔBP ND values). Finally, we carried out a median split of ΔBP ND in the LS, yielding high vs. low amphetamine responders, and univariate ANCOVAs tested for group (high vs. low responders) differences in NS scores (age as covariate). Since NS and other impulsivity traits generally decrease as adults grow older, age was used as a covariate [48] (although our sample was relatively young, we noted an inverse correlation between NS3 and age: r = -.33, N = 52, p = .017).
Partial correlations were carried out between BP ND values per drug condition in each striatal ROI (6 total) and NS scores (5 total: 4 subscales and NS Total ; age and study controlled for), only p<.005 results were reported (i.e., .05/11 = .0045). Partial correlations (age controlled for) were also carried out between ΔBP ND in each striatal ROI (3 total) and NS scores (5 total); only p<.006 results were reported (i.e., .05/8 = .0063).
rmANCOVAs were carried out on CT measures with frontal region (DLPFC, SM, vmPFC/ limbic) as the within-subject factor; age and gender were used as covariates [49]; preliminary analyses indicated no hemispheric effects, thus, it was not included as a factor). Study was not used as a factor as preliminary analyses yielded no effect of study on CT. Partial correlations (controlling for age and gender) were carried out between CT measures in the frontal regions (averaged across hemispheres) and NS scores (5 total); only correlations with p<.006 were reported (i.e., .05/8 = .006).
Partial correlations (controlling for age, gender and study) were also carried out between frontal CT values (3 regions) and BP ND measures (3 ROIs) in each drug condition (amphetamine, baseline/placebo); correlations between CT values and ΔBP ND were also carried out (controlling for age and gender). Significance was set at p = .006 (i.e., .05/8) and p = .008 (i.e., .05/6), respectively.
Finally, we carried out stepwise multiple regressions to assess the influence of CT measures (3 regions) and NS scores (5 total) on ΔBP ND in each striatal ROI. Age and gender were also included as predictor variables (stepwise). Multicollinearity was assessed using the variance inflation factor (VIF; <1.5 deemed acceptable); autocorrelations in the residuals were assessed using the Durbin Watson statistics (acceptable: 1.5-2.5). The significance of the ANOVA was p<.005 (i.e., p<.05/11 factors). Unstandardized beta coefficients are presented.

BP ND & ΔBP ND results
No sphericity violations existed, and there was no relation between inter-scan time duration (days) and ΔBP ND . A main effect of study existed for BP ND 47; Fig 1).
The rmANOVA assessing ΔBP ND values yielded a main effect of ROI [F(2,96) = 7.84, p = .001, η 2 partial = .14, study was not used as a covariate as there was no study effect on ΔBP ND ]. Follow-up comparisons indicated that ΔBP ND values in the LS were larger than those in the ASTS (p = .017) and SMS (p = .001; Fig 2). Voxel-wise analysis confirmed reduced BP ND following amphetamine administration in four clusters, with the greatest effect in the right LS (Table 5; Fig 3).

CT measures & correlations with NS scores
A rmANCOVA (age and gender as covariates) yielded a main effect of region [F(2,98) = 4.05, p = .02, η 2 partial = .08], with cortical thickness greatest in the limbic region and thinnest in the SM region (ps<.001). Study was not used as a covariate since preliminary analyses yielded no main effect of study on CT. No significant partial correlations between CT and NS scores emerged (gender and age were used as covariates, significance set at p<.006). When age and gender were not controlled for, exploratory Spearman's correlations indicated that there was an inverse correlation between thickness in the vmPFC/limbic cortex and NS3 scores (extravarelation also existed between LS ΔBP ND and CT in the SM region (r = -.36, p = .013) and between ASTS ΔBP ND and CT in the DLPFC (r = -.35, p = .015). For each association, the thinner the cortex, the greater the amphetamine-induced DA response. No significant associations between baseline BP ND values and CT were observed (age, gender and study controlled for).

Multiple regressions
No autocorrelation or multicollinearity violations were noted. Stepwise multiple regression analysis yielded no regressions at the p<.005 level. LS ΔBP ND was weakly associated with CT in the SM cortex

Discussion
In the current study, we examined the relation between NS, frontal CT and striatal DA responses to an amphetamine challenge in a large sample of young healthy adults. The primary findings were that the greatest amphetamine-induced changes in BP ND occurred in the ventral Novelty seeking, dopamine release & cortical thickness limbic striatum (LS). Individuals with high vs. low LS ΔBP ND responses (high vs. low drug responders) had higher impulsiveness (NS2) scores. Partial correlations indicated that greater amphetamine-induced striatal DA responses are associated with thinner frontal cortices. As expected, BP ND values decreased following amphetamine administration. Thanks to the greater statistical power afforded by the large sample size, this study was able to convincingly demonstrate regional differences in the magnitude of this effect, with larger ΔBP ND responses in the LS vs. other striatal regions, consistent with smaller PET studies in humans [13], [34], [46], [50], [51] and microdialysis studies in laboratory animals [52]. Via our voxel-wise analysis, we also confirmed that, with the d-amphetamine dose and route of administration used, there are no statistically significant changes in [ 11 C]raclopride binding in voxels outside the striatum.
The mechanism accounting for the different magnitude of DA responses within striatal sub-regions requires more study, but limbic striatal inputs from the ventral tegmental area (VTA) and dorsal substantial nigra (SN) express fewer D 2 autoreceptors and DA transporters (DAT) than projections to other striatal regions [53]. Decreased D 2 autoreceptor expression may be associated with greater amphetamine-induced DA release in the LS (vs. other striatal regions) [28], [54], though the effect of fewer DATs is less clear [38], [55]. The LS also sends non-reciprocal GABAergic projections to the ASTS, as such, ASTS activity is regulated by LS GABAergic input (similarly, the ASTS appears to regulate the SMS via GABA projections) [38], [55]. Finally, the LS has extensive connections with the orbitofrontal cortex (OFC), vmPFC, and aspects of the anterior cingulate cortex (ACC) [46] as well as the amygdala and hippocampus [55]. Together, these systems play a pivotal role in the initiation and inhibition of approach toward rewarding and potentially rewarding stimuli [56].
The ΔBP ND values in the current study are smaller than what has been reported previously with smaller samples [12], [13], [29], [33], [34]. This likely reflects that, in the current study, we used an automated PET analysis approach (i.e., an in-house PET analysis pipeline, involving few manual data corrections specifically with respect to co-registration between structural MRI and BP images; the pipeline employs an iterative co-registration process) that integrates Turku PET centre tools (http://www.turkupetcentre.net/)) for ROI analysis, which relies on traditional non-linear fitting to estimate Simplified Reference Tissue Model [SRTM] parameters. This approach is different from those used by the individual studies, which calculated SRTM parameters for ROI analysis using a basis function [39]. Perhaps most importantly, the automated pipeline employed ROIs that were larger than those used in some of the other studies. As a result, the correction for multiple comparisons was greater. Finally, we applied double-erosion (2 voxels), which may have excluded peak activation in the inferior and ventralmost aspects of the LS. These methodological differences may have influenced our BP and ΔBP ND ROI values. However, given the large sample and different methods used previously, the automated pipeline improved the objectivity of the analyses, and the use of more stringent significance detection approaches in the ROI analyses minimized false positives.
Consistent with our a priori hypothesis [28], [13], [30] high vs. low amphetamine responders (i.e., those with greater DA release to amphetamine) exhibited higher NS2 (impulsiveness) scores. That this effect was subtle should not be surprising. There is inherent variability in neuroimaging indices, and a snapshot of a single aspect of function in a particular neural system (or morphometry) is unlikely to capture even an unvarying trait with minimal noise [11]. Measures of externalizing traits suffer from the same limitations. Moreover, any one measure of personality is unlikely to encapsulate the aspect most closely related to the neuroimaging metric being assessed; indeed, different aspects of externalizing traits may reflect various aspects of meso-striatal DA system function [57]. Nevertheless, the consistency with which measures of striatal DA transmission have been found to be associated with impulsive/externalizing traits ( Table 2), along with the current study, suggests that individuals with higher impulsiveness scores are characterized by a more pronounced DA response to an amphetamine challenge. Based on the extensive connections with cortical regions implicated in emotion regulation [46], greater LS DA responses may facilitate more disinhibitory, impulsive-like behaviours to novelty and rewards.
The most novel aspect of this study was our attempt to relate striatal DA activity, frontal CT and NS traits in a large, healthy population. Inconsistent with our hypotheses, however, regression analyses indicated that the combination of NS and CT did not yield a stronger statistical prediction of striatal DA responses, as compared to either variable alone (i.e., although CT was associated with ΔBP ND , inclusion of NS did not improve the model). We found that greater DA responses in the ASTS were correlated with thinner sensorimotor cortices; similar tendencies existed between LS DA response and sensorimotor CT as well as between ASTS DA response and DLPFC CT. This is consistent with previous work by our group in a smaller sample, which was included in the current study [12]. To our knowledge, no other published data exist regarding the relation between striatal DA activity and cortical morphometry in healthy individuals. However, disease states characterized by striatal DA dysfunction, including Parkinson's disease [58], [59], schizophrenia [60] and substance use disorders [61], [62], are typically associated with widespread cortical thinning as well as volume loss of the grey matter. Although brain morphometric changes in such disorders can be accounted for by numerous factors, and are certainly not solely related to striatal DA activity modifications, these findings suggest, at least to a certain extent, a relation between cortex morphometry and striatal DA activity.
Previous work found that OFC metabolism was inversely associated with drug-induced striatal DA release in controls [63], consistent with the idea that the OFC plays an integral role in reward valuation by way of LS activity regulation. Additionally, thicker PFCs are generally associated with improved cognition [64]. As such, a thicker PFC may index functional integrity, including more effective engagement of regulatory processes in the presence of rewards and other salient cues. Given the significant overlap between cortico-striatal loops, it is perhaps not surprising that we did not see specific correlations between ΔBP ND in distinct striatal regions (ASTS, LS, SMS) and CT in corresponding fronto-cortical regions.

Conclusion & limitations
This is the first known study to simultaneously assess in vivo striatal DA release in relation to CT and NS in humans. This noted, the results should be considered in light of the following. First, the no drug condition contained a placebo in some studies but not all. However, there were no differences in ΔBP ND between studies with vs. without a placebo capsule, and study was used as a covariate in the statistical analyses when appropriate. Second, although our sample size is large for a PET study, it is relatively small for CT assessments, and this may have decreased the probability of identifying a correlation between CT and NS. Large populations are generally required to account for the maturational variability that occurs throughout late adolescence and early adulthood (i.e. the age group we examined) [65]. Further, even in large healthy populations, the association between externalizing traits (and personality traits in general) and brain morphometric features is generally subtle [3], [4]. Third, CT can be adjusted by total brain volume (TBV)/intracranial volume (ICV), grey matter volume or not at all. In the current study, we adjusted CT by TBV, following the recommendations of Zhou et al., 2014 [47]. By not including TBV as a covariate or control variable (in the partial correlations), the degrees of freedom were affected. We compensated for the potential influence of this decision on significance by setting stringent p-value thresholds for the CT analyses. In comparison, significant associations were not observed when using absolute CT values or CT values with TBV added as a covariate. It is plausible that the addition of TBV as a covariate decreased our power to detect effects, or that adjusted CT values are more sensitive in revealing a relation between striatal DA responses and cortical morphometry. Indeed, there is little consensus as to whether TBV/ICV should [66] or should not [4], [67], [68] be included as a covariate in CT analyses in the context of substance use research. More broadly, there is limited agreement as to how CT should be analyzed (i.e., absolute vs. corrected values). Thus, such methodological differences must be kept in mind when comparing and replicating future research. Further, we focused on TPQ-assessed NS, however, different externalizing measures may lead to different results, and should be investigated in future work. Finally, associations between NS, striatal DA release and CT may be nonlinear, and non-linear statistical approaches might reveal more complex associations. However, in our sample, exploratory non-linear correlations were not significant, and the current results strengthen the evidence for associations between striatal DA release, NS related impulsiveness and frontal cortical thickness in the largest known sample to date.
Supporting information S1 Table. Excel table containing