As shown in Fig. 1, we used PET and fMRI, along with behavioral tests, to investigate the relationships between neurotransmitters, brain activity, and cognitive performance. Thirty-seven participants were enrolled in a single-blind, placebo-controlled, crossover study. Each participant underwent imaging sessions before and after oral administration of 60 mg MP or placebo in a randomly counterbalanced order. We analyzed changes in dynamic brain features (e.g., fractional occupancy, dwell time, and appearance rate) associated with MP administration. In addition, we examined the relationship between these dynamic changes, DA receptor availability, and performance on a visual attention task.
Figure 1. Study design and pipeline for functional dynamic analysis. |
Participants and study design
Data from 37 healthy adults (24 males, 13 females, aged 22–64 years) were included. All participants provided written informed consent, and the study was approved by the Institutional Review Board Committee of the National Institutes of Health. Participants were excluded if they had a history of substance misuse or dependence (other than nicotine), psychiatric disorders, neurological disease, medical conditions that may alter cerebral function (i.e., cardiovascular, endocrinological, oncological, or autoimmune diseases), current use of prescribed or over-the-counter medications, and/or head trauma with loss of consciousness > 30 minutes. Detailed demographic information is shown in Table 1.
Table 1. The demographic characteristics of the participants
|
MP administration and PET acquisition: PET scans were used to measure the availability of D1R using [11C]NNC-112 and D2R using [11C]raclopride. Participants were imaged using two scanners based on availability: a Siemens AG High Resolution Research Tomography (HRRT) scanner used for 17 individuals (including 7 females), and a Siemens AG Biograph PET/CT scanner used for 20 individuals (including 6 females). Specifically, all [11C]NNC-112 scans were scheduled at 10 AM under baseline conditions. [11C]raclopride scans were conducted on two separate occasions for each participant: one performed one hour after oral placebo pill and the other performed one hour after 60 mg oral MP, in a single-blind manner with counterbalanced session order. The two [11C]raclopride scans were uniformly scheduled at 1 pm and performed on the same scanner for each individual to ensure consistency. The technical details for scan procedures were as previously published (19). Briefly, the [11C]NNC-112 imaging began immediately after injection of a maximum dose of 555 MBq and was followed by a series of 21 dynamic emission scans from the time of injection to 90 minutes post-injection. The [11C]raclopride imaging was started after an injection of a maximum dose of 370 MBq followed by a series of 22 dynamic emission scans performed from the time of injection to 60 minutes post-injection. Before analysis, all dynamic emission scan images underwent a rigorous evaluation by one of the investigators (SBD) to ensure the exclusion of any images compromised by motion artifacts or misplacement.
PET modeling, harmonization, and striatum extraction
The MAGIA toolbox(20) was employed for PET modeling. MAGIA consisted of frame-alignment (motion-correction, using the middle image of the image volume time series as master image for the alignment) and co-registration with the individual brain MRI (or a PET template if MRI image is missing), and a quality report for image quality evaluation. To mitigate scanner-specific effects between HRRT and PET/CT, we used ComBat (21), a data harmonization approach, to perform voxel-level harmonization independently for each tracer in the PET assessments to ensure data consistency between the two PET scanners. The ComBat model included age, gender, and race as covariates. For voxel-wise analysis, a subcortical atlas (22) consisting of 32 subcortical regions was used to extract voxels in the striatum, including voxels in the putamen, nucleus accumbens (NAc), and caudate. For ROI analysis, the same subcortical atlas was used to extract and average striatal ROIs. Details of the 32-ROI subcortical atlas and the striatum mask are displayed in Supplementary File S1.
MRI acquisition: Approximately 180 to 300 minutes after MP or placebo administration, participants underwent structural and resting-state fMRI using a 3.0T 32-channel Siemens Prisma scanner. To acquire resting fMRI time series, a multi-echo, multiband EPI sequence was used: multiband factor = 3, anterior-posterior phase encoding, TR = 891 ms, echo times = 16, 33, and 48 ms, flip angle = 57 deg, 45 slices with 2.9 × 2.9 × 3.0mm voxels and 520 time points while the participant relaxed with their eyes open (total acquisition time = 8 min). A fixation cross was presented on a black background under dimmed room lighting using a liquid-crystal display screen (BOLDscreen 32, Cambridge Research Systems; UK). The 3D MP-RAGE (TR/TE = 2400/2.24 ms, FA = 8 deg) and variable flip angle turbo spin-echo (Siemens SPACE; TR/TE = 3200/564 ms) pulse sequences were used to acquire high-resolution anatomical brain images with 0.8mm isotropic voxels field-of-view (FOV) = 240 × 256 mm, matrix = 300 × 320, and 208 sagittal slices.
Functional MRI processing: Resting-state fMRI was preprocessed using fMRIPrep (23) and in-house codes. Specifically, the fMRIPrep pipeline was used for multi-echo-multi-band optimization(24), gradient distortion correction, rigid body realignment, field map processing, and spatial normalization to standard MNI space. For post-processing steps, spatial smoothing was applied with a Full Width at Half Maximum (FWHM) of 5 mm. The fMRI data were filtered with a bandpass filter ranging from 0.01 to 0.08 Hz. The data also underwent a rigorous detrending process and regression analysis to mitigate the influence of six motion-related and three anatomical-related nuisance variables. Time points were excluded if the volume-to-volume BOLD signal met the following criteria: DVARS > 150 or framewise displacement (FD) > 0.5 mm. Time series for 200 cortical regions (25) and 32 subcortical regions (22) were then extracted and demeaned. Details of the two atlases are shown in Supplementary File S1.
Dynamic functional analysis
We followed the methodology established by Cornblath et al. (16, 26) in which the fMRI time courses for all subjects under the MP and placebo were concatenated in time. A k-means clustering algorithm was then used to identify clusters of brain activation patterns or states. The Pearson correlation coefficient was used as a metric to measure the distance in the k-means. The clustering process was repeated 50 times with random initialization to select the best partitions of the data. The optimal number of clusters (k) was selected based on the elbow criterion. Specifically, we plotted the explained variance curve from k = 3 to k = 22 and identified the inflection point or "elbow". The inflection point was consistently observed between 4 and 6 for different partitions (Supplementary file S2). Given the increased k beyond 6 resulted in less than 1% variance gain, the k in this study was set at 6, taking into account its balanced and interpretable nature. To strengthen the robustness of the partitions, we independently replicated the clustering process ten times and compared the Adjusted Mutual Information (AMI) across the ten resulting partitions. The partition with the highest cumulative AMI relative to the others was selected for subsequent analysis (Supplementary file S3).
N-ball-track task
We used a blocked visual attention paradigm (27) focusing on sustained attention and visual indexing. Specifically, a limited set of visual objects is marked for rapid attentional processing. Each TRACK epoch, lasting one minute, consists of five tracking and response intervals (Supplementary File S4). Briefly, a subset of balls (2 or 3 out of 10) is highlighted, followed by their random movement in the visual field. Participants were required to fixate on the central cross and to track these target balls as they moved. At the end of the tracking periods, the balls stopped moving, a new set of balls was highlighted, and participants were instructed to press a button if the highlighted balls were the target balls. Each "DO NOT TRACK" or control epoch consisted of similar one-minute intervals with no ball highlighting. Participants passively watched the balls move and stop. Each task variant (with 2 or 3 balls) contains three cycles of "TRACK" and "DO NOT TRACK" epochs, totaling 6 minutes and 10 seconds. The tasks were displayed using MRI-compatible goggles connected to a computer. We recorded response time and hitting accuracy via button presses and synchronized the paradigm with the fMRI acquisition using a scanner trigger signal. Of note, the n-ball-track task was performed during fMRI, but here we only use the behavioral data from this task. All dynamic functional analysis was performed on resting state fMRI data that were acquired immediately after the VA task.