Functional segregation in Parkinson’s disease

Abstract

The basal ganglia are characterized by somatotopic representation and are organized in parallel, functionally segregated corticostriatal circuits, but the impact of Parkinson’s disease (PD) on this architecture is unknown. We mapped task-evoked dopamine release using [¹¹C]raclopride positron emission tomography/magnetic resonance in 13 early PD and 15 healthy control (HC) subjects during the performance of motor, cognitive, and reward tasks. In PD, motor tasks elicited decreased relative dopamine release in dorsal putamen but increased in less affected caudate and ventral striatum compared to HC. HC showed distinct dopaminergic topographies for upper versus lower limb activation; this somatotopy was absent in PD. HC exhibited segregated patterns of dopamine release across all tasks, but both anatomical and functional segregation were lost in PD. Compensatory effects of altered dopamine release topography may help to maintain function in patients with PD, but loss of segregation may contribute to challenges in discrete selection of movement and in multitasking.

Task-evoked dopamine imaging reveals reorganization of corticostriatal signaling in Parkinson’s disease.

INTRODUCTION

The basal ganglia are responsible for the complex coordination of motor planning, task selection, and reinforcement learning (1). Striatal dopamine denervation, stemming from the progressive degeneration of dopaminergic neurons in the substantia nigra pars compacta, is the major pathological hallmark of Parkinson’s disease (PD) and is clinically associated with its cardinal motor features.

Previous studies suggested that the striatal dopaminergic deficit in PD follows a pattern that starts in the caudal and dorsal part of the putamen and advances toward rostral areas as the disease develops (2–4). However, while a caudal to rostral pattern of progression is supported by many molecular imaging studies, the possibility of an additional dorsal to ventral gradient is less established (5–7). Unlike the substantia nigra, there is clear evidence of somatotopic representation in the striatum (8).

By enhancing signal-to-noise ratio in striatal projection neurons (9), the localized release of dopamine, acting in conjunction with corticostriatal projections, is thought to result in discrete excitation of neurons whose activation (D1) or inhibition (D2) leads to the desired action. In addition to somatotopic representation within the striatum, there are thought to be parallel but segregated cortico-striatal-pallido-thalamo-cortical loops that subserve cognitive and motivational and motor functions (1, 10). Thus, the loss of dopamine seen in PD may be expected to result in loss of discrete somatotopic representation within the striatum and may additionally result in loss of segregation of motor, associative, and limbic functions.

We have therefore studied the topography of task-associated dopamine release in early-stage PD and compared it to that observed in healthy controls (HC) to determine whether there is functional reorganization of dopaminergic projections in early disease. Dynamic positron emission tomography (PET) imaging with [¹¹C]raclopride (RAC) was used to assess patterns of dopamine release induced by behavioral tasks involving the various corticostriatal loops in PD and HC subjects. PET-based estimates of dopamine release, while quantitative, are not absolute measures but are inferred through estimates of task-induced relative changes of synaptic dopamine levels. For convenience, we will use the term “dopamine release” when referring to such PET-based estimates of synaptic dopamine level changes. Our group recently introduced a framework known as residual space detection (RSD), which, based on simulations, demonstrated strong performance in detecting low-amplitude, localized dopamine release using only a single bolus of injected tracer (11). RSD is applicable to a variety of dopamine release scenarios and tracer binding conditions, including spatially heterogeneous receptor densities and low basal dopamine levels commonly observed in PD (11, 12). Prior studies have demonstrated that dopamine release evoked by transcranial magnetic stimulation is lower in magnitude but more diffuse in PD than in control subjects (13). We hypothesized that HC subjects would exhibit focal and segregated dopamine release in the striatum, predominantly in the left hemisphere during motor tasks involving the hands and feet, with a lesser degree of activation in right striatum. In contrast, PD would show relative dopamine release in larger overlapping clusters during the performance of both motor and nonmotor tasks, indicative of a loss of anatomical segregation of dopamine release.

RESULTS

Subjects in PD and HC groups were similar in age and had similar scores on Montreal Cognitive Assessment (MoCA) and Beck Depression Inventory (BDI) assessments. The disease duration for the patients was 3.54 ± 2.30 years, and the Movement Disorders Society–Unified Parkinson’s Disease Rating Scale III (MDS-UPDRS-III) score was 17.46 ± 10.00 points (Table 1) in the off medication state.

Table 1. Subject characteristics.

Demographics, disease duration, and clinical assessment scores for PD and HC groups. H&Y, Hoehn & Yahr; LEDD, Levodopa Equivalent Daily Dose.

	Parkinson’s disease (n = 13)	Healthy control (n = 15)	P
Age (years)	66.54 ± 8.67	64.20 ± 9.01	0.492
Sex (male/female)	8/5	7/8	0.684
Disease duration (years since symptom onset)	3.54 ± 2.30	NA	NA
Disease duration (years since diagnosis)	2.58 ± 1.66	NA	NA
H&Y, off	1.92 ± 0.28	NA	NA
MDS-UPDRS-III, off	17.46 ± 10.00	NA	NA
LEDD	455.82 ± 321.83	NA	NA
MoCA	26.31 ± 2.10	27.60 ± 1.96	0.106
BDI	4.85 ± 4.39	3.20 ± 2.37	0.220

The spatial distribution of dopamine release during the performance of motor tasks differs between PD and HC

Group-level three-dimensional (3D) images of mean RSD β values relating to striatal dopamine release during finger and foot tapping were produced for each group (Fig. 1, A and B). HC subjects demonstrated focal and segregated dopamine release in the striatum according to functional use of the hand or foot. This primarily manifested in the left hemisphere, with activation of the right striatum present but to a lesser degree. Conversely, PD subjects predominantly showed dopamine release in large clusters in the less affected striatum during tapping, with the spatial extent being largely comparable across both tasks. These visual results were further validated with the principal components analysis (PCA) analysis.

Fig. 1. Group mean maps of β values (%RAC signal change relative to baseline) in HC group and PD group for finger and foot tapping tasks.

(A) 3D surface view of striatum in HC group. (B) 3D surface view of striatum in PD group. (C)Three views of HC group. (D) Three views of PD group. The images highlight the contrast between discrete, localized dopamine release in HC compared to the more widespread, diffuse activation predominantly in the less affected striatum of PD subjects. L, left; R, right.

For the data obtained during finger tapping, the combination of PCs 1, 2, and 5 significantly separated PD and HC individuals’ pattern subject scores, with PD subjects consistently displaying higher subject scores than their HC counterparts (HC subject score mean = 0.00 ± 0.53, PD subject score mean = 1.58 ± 0.70; P < 0.0001) (Fig. 2A). This combination of PCs explained 43.5% of the total variance [leave-one-out 95% confidence interval (CI) = 35.71 to 42.12%, range: 18.48 to 47.035%]. In the voxel-based PCA analysis of β maps within the striatum, the right (less affected) caudate and ventral striatum exhibited strong positive weightings, while the left (more affected) caudate and bilateral dorsal putamen demonstrated relatively low weightings (Fig. 2, B and C). Figure S2 depicts a breakdown of each PC’s spatial distribution.

Fig. 2. Comparison of striatal dopamine release patterns between PD and HC groups during motor tasks.

(A) Comparison of z-scores of the combined PC pattern (PCs 1, 2, and 5) during finger tapping between PD and HC groups. Asterisk represents P < 0.001. Spatial topography of the combined PC pattern for finger tapping displayed on striatal cross sections (B) and 3D surface view displayed within the striatal brain mask (C). (D) z-scores of the combined PC pattern (PCs 1 and 6) during foot tapping between PD and HC groups. Spatial topography of the combined PC pattern for foot tapping displayed on striatal cross sections (E) and 3D surface view displayed within the striatal brain mask (F).

For the foot tapping dataset, the combination of PCs 1 and 6 separated PD and HC individuals’ pattern subject scores (HC subject score mean = 0.00 ± 0.56, PD subject score mean = 1.64 ± 0.57; P < 0.0001) (Fig. 2D). PCs 1 and 6 combined explained 26.3% of the total variance (leave-one-out 95% CI = 23.37 to 25.93%, range: 20.78 to 34.10%). Compared to HC, PD subjects demonstrated relatively higher dopamine release bilaterally in the caudate and ventral striatum and lower release bilaterally in the dorsal putamen (Fig. 2, E and F). Figure S3 depicts a breakdown of each PC’s spatial distribution.

Motor activation of different regions (upper versus lower limb) results in anatomically discrete dopamine release in HC but not in PD

To further investigate the different distributions of dopamine release between foot and finger tapping, we also conducted voxel-based PCA to compare finger and foot tapping β maps separately in HC and PD groups and separately for different sides. In the HC group, when analyzing the left striatum, the combination of PCs 1 to 4 effectively separated the distribution of dopamine release during foot and finger tapping. This combination of PCs consistently resulted in higher scores for data obtained from the foot tapping scans compared to finger tapping data (finger tapping score mean = 0.00 ± 0.59, foot tapping score mean = 1.33 ± 0.88; P < 0.0001) (Fig. 3A). This combination of PCs explained 55.6% of the total variance and was characterized by strong positive weightings in the left ventral striatum and negative weightings in the left caudate and putamen (Fig. 3, B and C). No difference between β maps obtained from the foot and finger tapping data was found in the right striatum of HC. In the PD group, neither side of the striatum showed differences in dopamine release patterns between foot and finger tapping. These results were further validated with Dice coefficient analysis.

Fig. 3. Comparison of striatal dopamine release patterns between foot tapping and finger tapping in the HC group.

(A) z-scores of the combined PC pattern (PCs 1 to 4) between foot and finger tapping tasks. Asterisk represents P < 0.001. (B) Spatial topography of the combined PC pattern discriminating foot from finger tapping, displayed at different striatal cross sections through all three planes. (C) 3D striatal surface view of the regions driving the combined PC pattern (displayed within the striatal brain mask).

In the analysis of the relationship between the scores of motor performance and dopamine release patterns during finger tapping and foot tapping in the PD group, we found that neither the total MDS-UPDRS-III score nor the subscores for bradykinesia and rigidity were correlated with the PCA scores derived from the finger tapping dataset. However, for foot tapping, the combined scores of PCs 1 and 6 were significantly correlated with the total MDS-UPDRS-III score for the more affected side (r = 0.57, P = 0.034), as well as the subscore for bradykinesia and rigidity on the more affected side (r = 0.60, P = 0.022) and bilaterally (r = 0.57, P = 0.034). In addition, a marginal correlation was noted with the bilateral MDS-UPDRS-III total score (r = 0.52, P = 0.07) (Fig. 4). This suggests that for foot tapping, the degree of alteration in the distribution of dopamine release is correlated with motor severity of disease.

Fig. 4. Correlation between dopamine release and motor function in the PD Group.

Linear regression analysis showing the relationship between z-scores of the combined PC (PCs 1 and 6) in foot tapping (x axis) and clinical motor scores (y axis). For foot tapping, PCs 1 and 6 correlated with total MDS-UPDRS-III (more affected side) and bradykinesia/rigidity subscores (more affected side and bilateral). Shaded areas represent the 95% CIs.

Performance of nonmotor tasks is associated with focal dopamine release in HC but diffuse release in PD

We further analyzed dopamine release in reward and cognitive tasks. HC subjects demonstrated discrete dopamine release in the caudate during performance of the Wisconsin card sorting task. Similarly, during the reward task, focal activation was observed in both the caudate and ventral striatum. In contrast, patients with PD primarily demonstrated widespread dopamine release across extensive striatal regions of both tasks (Fig. 5).

Fig. 5. Group mean maps of β (%RAC signal change relative to baseline) in HC group and PD group for reward and cognitive tasks.

(A) 3D surface view of striatum in HC group. (B) 3D surface view of striatum in PD group. (C)Three views of HC group. (D) Three views of PD group. The images highlight the contrast between discrete, localized dopamine release in HC and the diffuse, widespread dopamine release across extensive striatal regions in patients with PD.

During performance of the reward task, PC 1 significantly separated PD and HC individuals’ pattern subject scores, with PD subjects consistently displaying higher subject scores than their HC counterparts (HC subject score mean = 0.00 ± 0.40, PD subject score mean = 1.49 ± 0.85; P < 0.0001) (Fig. 6A). PC 1 explained 20.9% of the total variance in the dataset (leave-one-out 95% CI = 20.54 to 22.75%, range: 19.57 to 34.73%). Compared to HC, PD subjects demonstrated a higher degree of relative dopamine release bilaterally in the caudate and ventral striatum and lower relative release bilaterally in the dorsal putamen (Fig. 6, B and C). For the cognitive task, PC 2 effectively separated PD and HC individuals’ pattern subject scores, with PD subjects consistently displaying higher subject scores compared to HC (HC subject score mean = 0.00 ± 0.55, PD subject score mean = 1.58 ± 0.66; P < 0.0001) (Fig. 6D). PC 2 explained 15.7% of the total variance in the dataset (leave-one-out 95% CI = 16.04 to 24.54%, range: 14.34 to 45.20%). Similar to the reward task, the bilateral caudate and ventral striatum exhibited higher relative dopamine release in PD than in HC (Fig. 6, E and F).

Fig. 6. Comparison of striatal dopamine release patterns between PD and HC groups during nonmotor tasks.

(A) Comparison of z-scores of the single PC pattern (PC 1) during the reward task between PD and HC groups. Asterisk represents P < 0.001. Spatial topography of the single PC pattern (PC 1) for the reward task displayed on striatal cross sections (B) and 3D surface view displayed within the striatal brain mask (C). (D) Comparison of z-scores of the single PC pattern (PC 2) during the cognitive task between PD and HC groups. Spatial topography of the single PC pattern (PC 2) for the cognitive task displayed on striatal cross sections (E) and 3D surface view displayed within the striatal brain mask (F).

When comparing dopamine release during the reward and cognitive tasks in the HC group, PC 5, which explained 6.2% of the total variance in the dataset (leave-one-out 95% CI = 8.95 to 13.63%, range: 5.52 to 25.76%), effectively separated reward and cognitive dopamine release scores (reward task score mean = 0.00 ± 0.97, cognitive task score mean = 1.09 ± 0.70; P < 0.0001), with strong positive weightings in right caudate and negative weightings in left caudate and bilateral ventral striatum (Fig. 7, B and C). No difference in dopamine release between the reward task and cognitive task was found in the bilateral striatum for PD.

Fig. 7. Comparison of striatal dopamine release patterns between cognitive and reward tasks in the HC group.

(A) Comparison of z-scores of the single PC pattern (PC 5) between cognitive and reward tasks in the HC group. Asterisk represents P < 0.01. (B) Spatial topography of the single PC pattern (PC 5) discriminating cognitive from reward tasks in the HC group, displayed at different striatal cross sections through all three planes. (C) 3D striatal surface view of regions driving the single PC pattern (displayed within the striatal brain mask).

To further analyze the similarities and differences in striatal dopamine release across various tasks, we computed a four-way Dice coefficient for each subject using the β map output from Residual Space Detection (RSD)-Hybrid-Iterative Multilinear Reference Tissue Model (IMRTM) for each task (finger tapping, foot tapping, cognitive task, and reward task). Left/more affected β maps exhibited similar (low) overlap across tasks for both HC and PD groups. In contrast, right/less affected β maps exhibited stronger overlap for PD subjects across a range of sparsity thresholds (defined as the fraction of lowest positive β values removed), indicating common regions of dopamine release regardless of the task being performed (Fig. 8). Regions of common activation were notably seen in less affected anterior putamen and caudate clusters.

Fig. 8. Functional segregation of tasks.

Four-way Dice coefficient was computed for each subject using the β map outputs from RSD-Hybrid-IMRTM for each task (finger tapping, foot tapping, card sorting, and monetary gain). (A) Example maps for a single PD subject outlining the calculation of Dice coefficient, computed separately for more affected and less affected sides. (B) More affected β maps exhibit similar overlap across tasks for both HC and PD groups. (C) Less affected β maps exhibit stronger overlap for PD subjects across a range of sparsity thresholds (defined as the fraction of lowest positive β values removed) indicating common regions of dopamine release regardless of task performance. Common dopamine release regions across tasks are the large less affected putamen and caudate clusters seen in Fig. 3 (orange). Asterisks indicate significant differences (*P < 0.05; **P < 0.01).

DISCUSSION

As the central input region of the basal ganglia, the striatum plays a crucial role in the control of voluntary movement. Beyond its involvement in movement facilitation, the striatum is also implicated in a range of behavioral and cognitive functions, including reward processing and decision-making (10). These functions are thought to interact with various upstream and downstream brain regions through a series of parallel cortico-striatal-pallido-thalamo-cortical loops. In HC subjects, these loops are believed to be functionally segregated, with clear distinctions between anatomical boundaries of striatal regions associated with each specific function (10), while, based on our current findings, this segregation appears to be lost in PD. In PD and HC groups well matched for age, sex, cognitive function, and mood, notable differences were observed in the spatial distribution and localization of dopamine release within the striatum during the performance of motor and nonmotor tasks. Consistent with the initial hypothesis, HC participants exhibited discrete segregation of striatal function for hand versus foot movements, with prominent dopamine relative release in the left striatum relative to the tapping action and activation on the right side observed to a lesser extent. In contrast, the PD group exhibited a markedly different activation pattern, with predominant relative dopamine release observed in large clusters on the less affected side during tapping. The lack of the crossed pattern and increase on less affected side in dopamine release were also shown in a previous study (14).

It is widely accepted that functional decline of motor performance in early PD results from the degeneration of dopaminergic terminals within the striatum, especially the putamen. Our findings suggest that this degeneration of dopaminergic neurons in PD, which leads to hypokinesia, may trigger compensatory neuroplastic changes within the striatum and in corticostriatal connectivity as suggested by others (15), with resulting alterations in the regulation of dopamine release from striatal terminals. In contrast to HC, the spatial extent and localization of dopamine release was comparable across engagement of both upper and lower limbs in PD, consistent with findings from previous studies that assessed dopamine release in response to repetitive transcranial magnetic stimulation (13). Notably, this release of dopamine in PD was predominantly localized to the bilateral caudate and less-affected ipsilateral putamen rather than the more affected putamen and was virtually indistinguishable between finger and foot movements, further supporting the hypothesis of a loss of somatotopic segregation within the corticostriatal networks in PD.

Combining PCA with jackknife resampling and Akaike information criterion (AIC) identified PC combinations that optimally accounted for the overall group-related variance in motor-evoked dopamine release. In general, PD subjects exhibited a functional shift in dopamine release to the striatum on the less affected side and from sensorimotor striatal regions to more cognitive territories during motor tasks, relative to the pattern observed in HC subjects, consistent with prior functional magnetic resonance imaging (MRI) findings (16). In the finger tapping task, PD subjects showed prominent activation in the less affected striatum, whereas the more affected caudate and bilateral dorsal putamen exhibited lower relative dopamine release compared to HC subjects. PCA analysis of dopamine release data during foot tapping revealed higher relative dopamine release in the bilateral caudate and ventral striatum and lower relative dopamine release in the bilateral dorsal putamen in PD subjects compared to HC. Similar to the finger tapping task, greater relative dopamine release was observed in the caudate of PD subjects, relative to the expected sensorimotor activation, further supporting the hypothesis of a loss of functional segregation within the corticostriatal networks and a potential functional shift between sensorimotor and cognitive regions due to selective dopaminergic denervation.

These findings could be related to a previous observation of decreased connectivity between the posterior putamen and regions of the motor and somatosensory cortices and increased connectivity between the relatively preserved anterior putamen and these same cortical areas, particularly in the hemisphere least affected by the disease (17). Remapping of corticostriatal connectivity is seen in nonmanifesting carriers of pathogenic variants (18, 19). These processes could result in a “neural bottleneck” in the anterior putamen as degeneration progresses in the posterior striatum and impairs the ability to selectively allocate attention to various cognitive and motor tasks, thus disrupting normal functional segregation within the basal ganglia–thalamocortical circuits. In this regard, it may be relevant that in our prior work, a program of intensive exercise led to enhanced dopamine release in the caudate but a nonsignificant increase in the putamen, in response to transcranial magnetic stimulation applied over primary motor cortex (20). The functional shift is also in keeping with the observed shift from habit to goal-directed behaviors in PD (21).

Functional compensation has also been demonstrated within the framework of transfer of learning between two functionally independent, yet related, domains in the striatum (22, 23). The loss of functional segregation and the spillover of activity into unrelated striatal territories may occur when the striatal cognitive reserve is insufficient to meet the demands of a given task, such as during dual-task performance (24). Potential compensatory mechanisms may be inadequate to maintain proper functional performance under these conditions (24). Another hypothesis suggests that putaminal dopaminergic terminal loss is associated with impaired motor motivation in patients with PD, which consequently influences motor performance (25).

Notably, the patterns of dopamine release in PD subjects did not differ between foot and finger tapping, which is in marked contrast to HC subjects. In HC, during foot tapping, the left ventral striatum exhibited more relative dopamine release than during finger tapping, whereas the left caudate and putamen showed less relative dopamine release. In HC subjects, the activation areas of hand and foot movements in the striatum are distinctly separated (26), but in patients with PD, somatotopic segregation of dopamine release appears to be significantly diminished. Although structural presynaptic markers [such as dopamine transporter (DAT), vesicular monoamine transporter type 2 (VMAT2), or F-dopa] typically demonstrate focal dorsocaudal denervation in early PD, it is critical to note that our [¹¹C]RAC displacement paradigm measures synaptic dopamine release, which reflects not only the integrity of the dopaminergic nerve terminal but also the net result of corticostriatal input and dynamic compensatory mechanisms, including dopamine reuptake. It is well established that in early PD, compensatory mechanisms, such as up-regulation of dopamine synthesis and increased dopamine turnover, are activated to maintain striatal dopamine levels despite terminal loss (27, 28), while dopamine efflux in response to levodopa is actually greater when denervation is more severe (29, 30). Consequently, we propose that the “widespread” abnormality and loss of somatotopy observed here likely reflect this functional compensation: As focal denervation progresses, corticostriatal drive may become less selective or “spill over” to recruit broader striatal territories to maintain motor output. Thus, the geography of functional release does not linearly map onto the geography of structural denervation, a dissociation supported by prior multitracer studies (5). The loss of dopamine may alter the signal-to-noise ratio in striatal projection neurons, thereby reducing the capacity for discrete activation of striatal neurons and their connections (9).

While it might be argued that the lack of association between dopamine hypoactivity and focal clinical onset could be driven by the limited anatomical sensitivity of PET, the clear somatotopic segregation observed in our HC using the identical protocol argues against methodological insensitivity. Instead, we believe these findings reflect a genuine biological phenomenon specific to synaptic transmission. [¹¹C]Raclopride binding is affected by both receptor density and occupancy by endogenous dopamine, distinguishing it from presynaptic tracers. While amphetamine-induced dopamine release is directly related to the degree of striatal dopamine innervation (31), dopamine release in response to levodopa is inversely related to innervation, likely driven by increased dopamine turnover (32) and compensatory release in earlier disease stages (30). Limited prior work on motor task performance in PD similarly suggests a shift of activity to preserved striatal areas such as the caudate (33). Therefore, we propose that task-evoked dopamine release, in contrast to static structural markers, reflects altered corticostriatal input (20, 34) and synaptic dysfunction that may precede or extend beyond focal axonal degeneration.

One hypothesis suggests that, in PD, dopaminergic projections to the dorsal and caudal striatum degenerate first and most severely, followed by degeneration in more rostral and ventral striatal regions (3, 7). PD motor symptoms often initially manifest in the upper limbs. If PD indeed affects the dorsal motor striatum earlier than more ventral regions, then earlier involvement of upper limbs would seem to run counter to expectations based on the somatotopic organization of the striatum, as the lower limbs are thought to be represented more dorsally. However, if such gradients exist, then one might expect the initial motor symptoms of PD to appear in the lower limbs, based on the somatotopic mapping of the body within the striatum. This, however, does not align with typical clinical observations. Patients often present with early impairments in the arm or hand, which may be explained by increased sensitivity to functional deficits involving fine hand dexterity (26). The reduction in somatotopic segregation within the striatum may partly account for this phenomenon. Because of damage to dopamine-producing neuron terminals and compensatory mechanisms, the dopaminergic regions that govern arm and leg function lack clear segregation, and clinical deficits do not necessarily appear strictly according to functional divisions. Recent ¹⁸F-dopa PET imaging studies also challenge the dorsoventral pattern of progression in early PD (26, 35). Pineda-Pardo et al. (26) suggested that the dopaminergic loss primarily affects the intermediate caudal putamen, correlating with upper limb motor impairments, and then extends bidirectionally dorsally and ventrally, potentially due to synaptic connectivity demands. They also combined functional MRI group-activation maps during a self-paced motor task to characterize the decline in ¹⁸F-dopa uptake rate in the upper- and lower-limb territories. However, considering these are small studies based on a single dopaminergic tracer, the results need further validation.

The PCA z-scores of PD subjects (when comparing release in HC and PD during foot tapping) were significantly correlated with the more affected side of disease severity, as indicated by the MDS-UPDRS-III total score and its bradykinesia and rigidity subscores. However, no such correlation was found in the finger tapping task. Sustaining repetitive foot tapping while in the scanner may be more challenging than a simple finger tapping task, thus greater severity of disease may place greater stress on striatal function during execution of the former. In addition, although nigrostriatal denervation patterns may contribute to some of these findings, a more comprehensive model of Parkinsonian motor impairments likely involves additional extrastriatal dopaminergic and nondopaminergic mechanisms (36). These findings have important pathophysiological implications for understanding the motor impairments observed in PD.

In both reward and cognitive tasks, dopamine release in the caudate and ventral striatum plays a crucial role, consistent with the activation of corresponding nuclei observed in previous studies (37). Dopamine release in the caudate may be associated with the learning demands of new stimulus-response associations (38). However, our study further revealed that during performance of cognitive and reward tasks, individuals with PD exhibited greater relative dopamine release in the caudate and ventral striatum compared to HC, with a more diffuse pattern, which also corresponded to dopamine release during performance of motor tasks. While the dopamine release patterns in the striatum induced by reward and cognitive tasks significantly differ in HC, this distinction is absent in PD, suggesting a loss of both anatomical and functional segregation of dopamine release in PD.

The primary limitation of this study was its relatively small sample size, which is related to the demanding nature of the protocol, as well as the initiation of the study during the height of the COVID-19 pandemic. Despite this limitation and considering the substantial challenges associated with the duration and complexity of PET scanning, our analysis using within-subject cross-validation indicated robust and significant differences between the PD and HC groups. Thus, we believe that the current sample size is sufficient to support the conclusions drawn. In addition, given the nature of imaging techniques such as PET, there are inherent technical limitations, including relatively lower spatial resolution compared to other imaging modalities such as MRI.

We further performed an exploratory analysis to assess how predominant limb (i.e., upper versus lower limb impairment) might relate to different release patterns. We observed that predominant upper limb involvement on the more affected side tracked significantly with the pattern of striatal functional reorganization (i.e., greater clinical involvement of the upper limb compared to the lower limb was associated with a greater deviation from normal in dopamine release pattern in the upper limb compared to the lower limb). Given the limited sample size, we consider these findings preliminary; thus, future studies with larger, clinically diverse cohorts are required to definitively validate these trends.

MATERIALS AND METHODS

The pipeline for this study is displayed in Fig. 9. Twenty-eight participants were included (13 early-stage PD and 15 gender- and aged-matched HC). Subjects were recruited consecutively and studied (baseline assessment) from January 2021 until November 2024. The CONSORT diagram is shown in fig. S1. PD subjects were 50 to 85 years old and with disease duration less than 5 years from diagnosis. Disease duration was determined as years since receiving a PD diagnosis, as noted in patients’ clinical records. Neurological exams included the MDS-UPDRS, Purdue pegboard task, and BDI, all conducted in the medication-off state. The MoCA was assessed in the medication-on state. The study received approval from the University of British Columbia Research Ethics Board (certificate number: H19-03166) in accordance with the ethical guidelines outlined in the Helsinki Declaration. Before all study procedures, written informed consent was secured from each participant.

Fig. 9. Pipeline of the study including recruitments of subjects, imaging scans, and data analysis.

Demographics, disease duration, and clinical assessment scores for PD and HC groups.

Image acquisition and preprocessing

Patients with PD were scanned after overnight withdrawal of dopaminergic medication in the off state. All subjects underwent four separate scans conducted over 2 days on the GE SIGNA PET/MR Hybrid Scanner, each following a bolus injection of [¹¹C]RAC (550 megabecquerels). The scans were performed with a minimum of 2 hours (∼6 half-lives) between [¹¹C]RAC injections. Each scan began in a resting baseline state, followed by one of four types of task, aimed at inducing dopamine release: (i) performance of a complex sequence of finger tapping (33, 39); (ii) repetitive foot tapping to assess striatal dopamine release related to the motor function of the upper and lower limb, respectively; (iii) performance of the Wisconsin card sorting task to assess cognitive processing in the striatum; (iv) performance of a card selection task, with a 50% probability of monetary reward per trial, to evaluate dopamine release in the striatum in response to monetary incentive (reward task, to assess the limbic component of the striatum) (20, 40). HC subjects (all right-handed) performed all tasks using their right side. Patients with PD performed the motor task using the clinically more affected body side. PET scans for motor task conditions (i) and (ii) were 71 min in total and tasks began 36 min postinjection. For nonmotor task conditions (iii) and (iv), PET scans were 79 min long and tasks began 32 min postinjection due to the inclusion of a 10-min “control” portion of the tasks. For more detailed task descriptions, see the Supplementary Materials. Meanwhile, the 3D T1-weighted images were acquired using sagittal magnetization-prepared rapid gradient echo acquisition, which were used in the following normalization steps.

The acquired PET data were reconstructed with the 4D modified HighlY constrained backPRojection (HYPR4D) kernelized reconstruction method applied to time-of-flight PET data (i.e., PSF-HYPR4D-K-TOFOSEM), incorporating corrections for attenuation, scatter, randoms, radioactive decay, detector dead time, and normalization, as well as modeling of the scanner’s point spread function (41, 42). Following reconstruction, frame-to-frame realignment of the PET images was performed using SPM12 (43) to correct for potential head motion. Dynamic PET images were subsequently denoised with IHYPR4D (44). For the creation of group-level images, subject-specific striatal masks were normalized to a common-space striatal mask sourced from the Neurodevelopmental MRI Database (age 65 to 69) atlas (45), using Advanced Normalization Tools (ANTs) with a 2-mm isotropic voxel size (46).

Image analysis

To estimate task-related dopamine release during the acquisition of data following a single bolus of RAC, we applied voxel-wise RSD, making hybrid use of the measured data and the Iterative Multilinear Reference Tissue Model (RSD-Hybrid-IMRTM) to the subject-space dynamic PET images. Briefly, RSD-Hybrid-IMRTM predicts the baseline behavior of each striatal voxel’s RAC concentration across time [time-activity curve (TAC)], $\hat{C}\left(t\right)$, and compares it to the measured TAC data, $C\left(t\right)$. By computing “residuals,” defined as

$$R_{\text{pct}}\left(t\right)=100\%\left[1-\frac{C\left(t\right)}{\hat{C}\left(t\right)}\right]$$ (1)

Voxels containing dopamine release posttask initiation will exhibit a structured signal that can be extracted using a general linear model

$$R_{\text{pct}}\left(t\right)=\mathrm{\beta }P\left(t\right)+\mathrm{ϵ}\left(t\right)$$ (2)

where β represents the voxel TAC’s percentage change due to dopamine release, $P\left(t\right)$ is a predicted response shape [$\text{max}P\left(t\right)=1$], and $\mathrm{ϵ}\left(t\right)$ are residuals from the fit. The parameter of interest β is known to scale with relative increases in synaptic dopamine levels compared to baseline (relevant for PD, as discussed below) (11).

Thus, RSD-Hybrid-IMRTM generated voxelwise β maps for each subject, indicative of percentage RAC signal change due to dopamine release (11). Therefore, dopamine release was indexed by %RAC signal change relative to baseline. These parametric maps were subsequently transformed to a common space using the forward transformation from the normalization process (details are shown in the Supplementary Materials). For PD subjects whose clinically more affected side was on the left (i.e., those who performed tasks with their left limb), the spatially normalized images were flipped. The group mean map for each subject group was obtained by calculating the average RSD-Hybrid-IMRTM β map across HC and PD subjects, respectively.

To investigate group differences in spatial patterns of dopamine release within RSD-Hybrid-IMRTM, we applied the PCA using MATLAB (47), modified to restrict the analysis to striatal voxels. Voxel-wise PCA was applied to subjects’ Montreal Neurological Institute (MNI)-space RSD β results using a striatal MNI mask. PCs were linearly combined so as to maximize group differences of scores between PD and HC subjects, using AIC as the metric for inclusion (48). An additional PC inclusion threshold was set such that only PCs that described at least 5% of the data variance and separated PD and HC subject scores at P ≤ 0.2 were considered by the AIC model. Each calculated PC combination was confirmed to effectively separate the two subject cohorts via an independent sample t test on subject scores to confirm a statistically significant separation (P < 0.05). RSD-Hybrid-IMRTM β values were automatically demeaned in the PCA algorithm. Along with scores that separate subject groups, the PCA analysis generated 3D z-score images depicting the spatial distribution of voxels with relative positive and negative weightings that effectively separated the two cohorts. For consistency, in all figures in this work, orange voxels indicate regions of relatively increased relative dopamine release in PD, while blue voxels indicate regions of relatively decreased dopamine release. Jackknife resampling was applied to validate the PCA performance of the task-specific datasets. This technique estimated the variance of statistics derived from the entire task-specific dataset by systematically leaving out one observation from the dataset and estimating the mean and variance of the PCA regional and voxel-wise weights n times, where n is the total number of observations in the dataset (i.e., PCA is rerun on each dataset of n – 1 subjects) (49). A one-sample t test across all n – 1 voxel-wise PCA iterations was run to assess the reliability of the calculated PCA weights, with an alpha level of 0.0001 to minimize the false-positive rate (50).

Last, to examine functional segregation in HC and PD subjects, we compared β maps across all four tasks for each subject. Here, we computed the overlap of thresholded β maps across all tasks by way of a four-way Dice coefficient, defined as

$${\text{DCF}}_4=\frac{4\mid A\cap B\cap C\cap D\mid }{\mid A\mid +\mid B\mid +\mid C\mid +\mid D\mid }$$ (3)

where $A,B,C$, and $D$ are thresholded (binary) maps of striatal voxels predicted to contain dopamine release. A Dice coefficient equal to 1 indicates that an identical set of striatal voxels exhibited dopamine release in all four tasks, while a Dice coefficient of zero indicates that no common voxel exhibited dopamine release in all four tasks. Thus, a low Dice coefficient indicates a high degree of functional segregation, and vice versa. β was thresholded as a function of sparsity to enable intersubject comparisons, e.g., the top 20% of positive β values were kept in each map and fed into Eq. 3.

Statistics

All statistical analyses were completed using MATLAB (version 2019b) (47) and International Business Machines Corp. SPSS Statistics software (version 29) (51). We assessed the normality of data distribution using Shapiro-Wilk test. Within data passed normality tests (P > 0.05), independent sample t tests were conducted on all subject characteristics, cognitive function, and mood. Independent two-sample t tests were used for group comparisons (PD versus HC), and paired t tests were used for within-subject task comparisons. We used age as a covariate in partial correlation to examine the relationships between the PCA z-scores of finger tapping and foot tapping in the PD group and various motor function scores. These included the total MDS-UPDRS-III score, MDS-UPDRS-III scores for the more affected side, MDS-UPDRS-III scores for the less-affected side, scores for bradykinesia and rigidity in the MDS-UPDRS-III, as well as the bradykinesia and rigidity scores for both the more and less affected sides. Alpha level was set at 0.05 a priori for all statistical tests. Given the large number of correlations and limited number of subjects, correlational analyses were not corrected for multiple comparisons.

Supplementary Materials

This PDF file includes:

Supplementary Methods

Supplementary Results

Figs. S1 to S5

Tables S1 to S3