Twelve male vervet monkeys (Chlorocebus aethiops sabaeus from the UCLA Vervet Research Colony), ranging from 5 to 9 years of age, were included in this study. Monkeys were individually housed in a climate-controlled vivarium, where they had unlimited access to water and received twice-daily portions of standard monkey chow (Teklad, Madison, WI). All of the subjects were able to see, hear and communicate with other individuals in the room. Monkeys received half of their daily portion of allotted chow in the morning after behavioral testing was conducted (approximately 1100 h) and their second half in the afternoon (approximately 1500 h); the total amount of chow received was never reduced during the experiment to facilitate task performance.
All monkeys were maintained in accordance with the ‘Guide for the Care and Use of Laboratory Animals’ of the Institute of Laboratory Animal Resources, National Research Council, Department of Health, Education and Welfare Publication No. (NIH) 85-23, revised 1996. Research protocols were approved by the UCLA Chancellor’s Animal Research Committee.
Discrimination Acquisition, Retention and Reversal Learning
Monkeys were trained to move from their individual cages into a transport cart, and were brought to a quiet testing room where the transport cart was aligned to a Wisconsin General Testing Apparatus, which has been described elsewhere (Lee et al., 2007
). It was equipped with an operable opaque screen that separated the monkeys from three equally spaced opaque boxes. Each box was equipped with a hinged opaque lid so that food rewards (small piece of apple, banana, grape or orange) could be concealed inside. Moreover, each box lid could be fitted with a unique visual stimulus, (clip art from the Microsoft Office®
library that consisted of colored objects unfamiliar to the monkey) that the monkeys could easily view when sitting at the apparatus.
Testing sessions began when the opaque screen was raised to present the three boxes (each fitted with a unique stimulus) to the monkey. Only one response, in which the monkey opened a box fitted with a stimulus, was allowed per trial. A trial ended after a correct choice, an incorrect choice or an omission (no response for 2 min), and a 20-s intertrial interval followed. The next trial ensued with a different spatial box sequence, but with the reward associated with the same visual stimulus. Up to 80 trials per session were given.
Monkeys were trained to acquire, retain and reverse novel visual discriminations. The first session of a discrimination problem was a discrimination-acquisition phase and was held on a Monday or Thursday. The monkey was presented with three novel stimuli and had to learn which one was associated with reward, solely on the basis of trial and error. After a performance criterion (seven correct choices within ten consecutive trials) was reached, the session was terminated and the monkey was returned to his home cage. If a monkey did not reach criterion within 80 trials, the session ended but the same discrimination problem was presented the following day(s) until the performance criterion was met.
One day after reaching criterion, subjects were assessed in the retention phase, during which stimulus-reward contingencies were unchanged, until a criterion of four correct choices in five consecutive trials was met. The reversal phase then began immediately with no explicit signal that the transition between retention and reversal had occurred, other than the change in feedback experienced by the subject. During the reversal phase, the stimulus that was previously rewarded was no longer rewarded, and one of the two previously non-rewarded stimuli was rewarded. The reversal phase continued until the monkey achieved criterion (seven correct choices in ten consecutive trials) or until 80 trials had been completed, whichever occurred first. The number of trials required to reach criterion in the acquisition, retention and reversal phases were the primary dependent measures. For the reversal phase, the number of responses directed at the previously rewarded stimulus (perseverative responses) and the number of responses directed at the never rewarded stimulus (neutral responses) were also measured. The probability of a monkey making each response type was also calculated by dividing the number of correct, perseverative or neutral responses by the total number of trials in the reversal phase.
Subjects acquired and reversed consecutive discrimination problems, each of which featured three novel visual stimuli. Due to technical delays in the acquisition of PET scans, the total number of discrimination problems completed and the number of days between completion of the last discrimination problem and the PET scans differed and are exhibited in ; therefore, the analysis described here focused on the averages of the dependent measures collected across the last three problems, as these were closest in time to the subsequent PET scans.
Table 1 The total number of discrimination, problems, the number of days between last completed discrimination session and assessment of D2-like receptor availability, and the average number of trials required to reach criterion in the last three reversal phases (more ...)
Feedback Sensitivity Measures
Because behavioral sensitivity to positive and/or negative feedback can affect learning performance, we examined choice behavior on a trial-by-trial basis during the reversal phase. Here, we categorized trials according to whether the subject experienced positive or negative feedback on the preceding trial. This allowed calculation of the probability that after experiencing positive feedback, a subject would make either: a) another correct response, b) a response directed to the stimulus that was previously rewarded or c) a response directed at the stimulus that was never rewarded. The response to negative feedback was assessed by calculating the probability that a negative feedback event would be followed with either: a) the same incorrect response or b) a response directed at a different stimulus, irrespective of whether this response was correct or incorrect. We also performed a similar analysis of choice behavior for the data gathered during the discrimination acquisition; however, because perseverative responses were not possible in this phase, behavioral sensitivity to positive feedback was calculated by examining the probability of following a correct response with either: a) another correct response or b) an incorrect response.
A variable number of days after behavioral performance had stabilized (), D2-like receptor availability was assessed using a microPET Model P4 scanner (Concorde Instruments, Knoxville, TN). Dopamine transporter (DAT) availability was assessed, using [11C]WIN-35,428, in the same subjects for a larger study; however, DAT availability measures were not included in the hypothesized mechanism for our primary analyses and, therefore, are not described here. Monkeys received an intramuscular injection of ketamine hydrochloride (10 mg/kg) and glycopyrrolate (0.01 mg/kg). After monkeys were immobilized, an endotracheal tube was placed to provide inhalation of 2-3% isoflurane (in 100% O2) anesthesia throughout the duration of the experiment. Vital signs (heart rate, respiratory rate, oxygen saturation and temperature) were monitored and recorded every 15 min throughout the scan. A tail-vein catheter was placed, and the monkey was positioned on the scanning bed such that the imaging planes were parallel to the orbitomeatal line and the top of the head at the front of the field of view. A 20-minute 68Ge transmission scan was acquired before administration of the radioligand for attenuation correction. All subjects received a bolus injection [11C]WIN 35428 (1.0 mCi/kg), followed by a 5-mL saline flush, and data were acquired for 90 min. When radioactivity had fallen to baseline levels (~3 h after [11C]WIN-35,428 administration), a bolus injection of [18F]fallypride was delivered (0.3 mCi/kg), followed by a 5-mL saline flush. Dynamic data were acquired in list mode for 180 min. After the scan, animals were removed from the gas anesthesia and allowed to recover overnight before being returned to their home cages.
Reconstruction of PET images
Three-dimensional sinogram files were created by binning the data into a total of 33 frames (six 30-sec frames, seven 60-sec frames, five 120-sec frames, four 300-sec frames, nine 600-sec frames, one 1200-sec frame and one 1800-sec frame). We applied a previously validated algorithm to the transmission scan list-mode data to generate attenuation maps (Vandervoort and Sossi, 2008
). This algorithm uses an analytical scatter correction, based upon the Klein-Nishina formula, for singles-mode transmission data. Following construction of the attenuation maps, emission list-mode files were reconstructed using Fourier rebinning and filtered back projection, and corrected for normalization, dead time, scatter and attenuation within software provided by the manufacturer (microPET Manager version 18.104.22.168). The resultant images had voxel dimensions of 0.949 mm × 0.949 mm × 1.212 mm and matrix dimensions of 128 × 128 × 63.
Structural magnetic resonance (MR) images were acquired to allow for anatomically based demarcation of regions of interest (ROI). MR images were acquired one week after the PET scans. The monkeys received an intramuscular injection of ketamine hydrochloride (10 mg/kg) and atropine sulfate (0.01 mg/kg). Once the monkey was immobilized, an endotracheal tube was inserted to provide inhalation of 2-3% isoflurane gas (in 100% O2
) for the remainder of the scan. Monkeys were positioned on the bed of a 1.5 T Siemens scanner, with the head in the gantry, surrounded by an 8-channel, high-resolution, knee-array coil (Invivo Corporation). Nine T1-weighted volumes with three-dimensional, magnetization-prepared, rapid-acquisition, gradient-echo (MPRAGE) images were acquired (TR=1900 ms TE=4.38 ms, FOV=96 mm, flip angle 15 degrees, voxel size 0.5 mm, 248 slices, slice thickness 0.5 mm). Individual images were aligned to each other using Statistical Parametric Mapping 5 (Institute of Neurology, University College London, London, England), averaged together and resliced according to a previously developed MR template (Fears et al., 2009
ROIs were drawn twice, referred to as replicates, on each subject’s structural MR image by a single experimenter blind to the subject identity using FSL View (FMRIB’s Software Library v4.0). ROIs included the whole caudate nucleus, putamen, ventral striatum and cerebellum.
- ROI-based determination of binding potential (BP): Reconstructed PET images were corrected for motion and coregistered to the subject’s MR image using the PFUS module within PMOD (version 3.15; PMOD Technologies). Using the ROIs, activity was extracted from the coregistered PET images and imported into the PMOD kinetic analysis program (PKIN). Time-activity curves were fit using the Simple Reference Tissue Model (SRTM) (Lammertsma et al., 1996) to provide an estimate of the k2’ value, the rate constant of tracer transfer from the reference region to plasma. The k2’ estimates of the high-activity areas in the caudate nucleus and putamen were averaged and time-activity curves refit using the SRTM2 model using the average, fixed k2’ value applied to all brain regions (Wu and Carson, 2002). BP was then calculated by subtracting 1.0 from the product of tracer delivery (R1) and tracer washout (k2’/k2a). BPs from the left and right brain structures were averaged to create a single BP measurement for the caudate nucleus, putamen and ventral striatum. The BP of the ROI replicates were highly correlated, so the BPs of the ROI replicates were averaged to obtain our final ROI-based measurements of D2-like receptor availability in each of the brain regions.
- Generation of Whole Brain BP Maps: Parametric binding maps, showing BP, were generated for each subject in PXMOD (PMOD), using the SRTM2 model with the same fixed k2’ values used above. This modeling requires time-activity data for low- and high-activity regions to generate the initial parameters for modeling. We used the activity in the putamen and cerebellum ROIs as the high- and low-activity references, respectively. In order to perform voxel-wise statistical analyses with BP maps, we realigned all BP maps to a study-specific MR template, which was created by sequentially registering each subject’s skull-stripped MR scan (Multitracer, AIR version 5.0) using affine registration (FLIRT, FMRIB’s Software Library v4.0), and creating an average of the registered images. Individual skull-stripped MR scans were then registered to the study-specific template space using affine registration (FLIRT), and the resultant transformation matrix was applied to each individual subject’s parametric binding map, which was previously registered to the individual subject’s MRI. No additional smoothing was applied to the images.
All statistical analyses were conducted using SPSS 15.0. Reliability of performance was examined by calculating Cronbach’s alpha, a coefficient of reliability, for the number of trials required to reach criterion in the acquisition, retention and reversal phases of the task during the first ten completed sessions. Paired-samples t-tests were conducted to examine the number of trials required to reach criterion in the acquisition and reversal phases, as well as the error types (neutral or perseverative) in the reversal phase of the task. Linear regressions were conducted to examine the relationships between D2-like receptor availability and our behavioral measures; though we found significant linear relationships (Y = a - bX), visual inspection suggested that for some relationships an inverse function (Y = a - b/X) was more appropriate for the data. The asymptote (a) and slope (b) of each curve were estimated using the curve-fitting tool in SPSS. Models were compared using the Akaike Information Criterion (AIC) to determine whether the linear or inverse function best fit the data. When the inverse function was identified as the AIC-preferred model, the independent variables were transformed accordingly and correlations performed with the transformed values to calculate the Pearson correlation coefficient and significance values.
To examine the anatomical distribution of the relationship between positive-feedback sensitivity and BP within the striatum, linear regressions were performed using the FSL RANDOMISE v2.1 tool (Permutation-based nonparametric inference, Oxford University, Oxford UK) with a variance smoothing of 5 mm (FWHM Gaussian). A binary, striatal mask was created and feedback-sensitivity measures transformed according to the model that best fit the data according to our initial ROI analysis (see above). Threshold-free cluster enhancement (TFCE) (Smith and Nichols, 2009) was used to detect significant clusters of activation; this method provides the ability to perform cluster-based inference without the need to specify an arbitrary cluster-forming threshold, as is necessary when using Gaussian random field theory. For each analysis, 10,000 randomization runs were performed. Statistical maps were thresholded at p<0.05 (two-tailed) and corrected for the search volume contained in the striatal mask.