PET Imaging in Movement Disorders

Abstract

Positron emission tomography (PET) has revealed key insights into the pathophysiology of movement disorders. This paper will focus on how PET investigations of pathophysiology are particularly relevant to Parkinson disease (PD), a neurodegenerative condition usually starting later in life marked by a varying combination of motor and non-motor deficits. Various molecular imaging modalities help to determine what changes in brain herald the onset of pathology; can these changes be used to identify presymptomatic individuals who may be appropriate for to-be-developed treatments that may forestall onset of symptoms or slow disease progression; can PET act as a biomarker of disease progression; can molecular imaging help enrich homogenous cohorts for clinical studies; and what other pathophysiologic mechanisms relate to non-motor manifestations. PET methods include measurements of regional cerebral glucose metabolism and blood flow, selected receptors, specific neurotransmitter systems, post synaptic signal transducers, and abnormal protein deposition. We will review each of these methodologies and how they are relevant to important clinical issues pertaining to PD.

Introduction

Positron emission tomography (PET) permits measurements of regional brain blood flow and metabolism, various brain receptors, neurotransmitter systems, post-synaptic signal transduction and abnormal protein deposition. Each of these measures can reveal various insights into the pathophysiology of the clinical manifestations of movement disorders including Parkinson disease (PD) and other related parkinsonian syndromes. This review will focus on PD since each of these PET imaging modalities have been used to address several key issues: what changes in brain herald the onset of pathology; can these changes be used to identify presymptomatic people who may be appropriate for to-be-developed treatments that may forestall onset of symptoms or slow disease progression; can PET act as a biomarker of disease progression; can molecular imaging help enrich homogenous cohorts for clinical studies; and what other pathophysiologic mechanisms relate to non-motor manifestations.

Some clinical and pathology clues provide some insights into these issues. PD begins commonly once a person hits the 60’s. However, the risk increases with age and yet as many as up to 5.8% may begin before age 40¹. Pathologically, deposition of misfolded α-synuclein in intracytoplasmic inclusions called Lewy bodies or in neuronal processes called Lewy neurites begins in pigmented nuclei in caudal brainstem (with or without olfactory tubercle involvement) and then progresses rostrally². Once the pathologic process reaches the dopaminergic neurons of the substantia nigra in the midbrain, striatal dopamine loss occurs. However, the loss of nigrostriatal dopaminergic neurons must cross a threshold to produce motor dysfunction related to PD. This review will address how imaging can provide clues as to when the pathophysiologic process begins and how this relates to clinical manifestations.

Such studies have direct relevance for identifying when the pathophysiologic process begins in the brain, potentially to identify pre-symptomatic patients for possible treatment with to-be-developed therapies that may forestall onset of symptoms or slow disease progression. Many PET methods have been applied to such studies, and we critically review the relevant literature. Along the same lines, those people with selected co-morbid conditions may be at greater risk for development of PD. For example, those with rapid eye movement (REM) sleep behavior disorder (RBD) may have as much as a 50–60% risk of developing PD³. Applying PET to a higher risk group of people could be particularly revealing.

PET also may provide a measure of disease progression. A variety of PET methods have been used for this in clinical studies to determine efficacy of interventions to slow disease progression. These techniques have been fraught with challenges, with frequent application of PET methods prior to adequate validation of the technique. Of course, interpretation of such studies depends in large part on how accurately one can diagnose PD. Clinically, this can be challenging since many of the clinical features of PD overlap with other degenerative disorders such as multisystems atrophy (MSA) and progressive supranuclear palsy (PSP) as well as non-degenerative conditions (drug-induced parkinsonism).Many studies have tried to use PET or single photon emission computed tomography (SPECT) to distinguish these various conditions. This has clinical implications for potential response to treatment and research implications for selection of cohorts for clinical trials. We will address how well molecular imaging can help in this endeavor.

Finally, molecular imaging has begun to reveal new insights into the pathophysiology associated with non-motor manifestations of PD. These non-motor manifestations can be quite varied from autonomic deficits affecting multiple systems to cognitive and psychiatric dysfunction. Cognitive dysfunction is quite pervasive in PD with about 10% progressing to dementia every year; and ultimately affecting more than 80% of those who survive for two decades or more^4,5. Cortical synucleinopathy with Lewy pathology accompanies this cognitive impairment. Thus PET imaging of α-synuclein could be highly informative and may identify cortical progression prior to symptom onset. Development of such a radiopharmaceutical is currently underway but one is not yet available. Other imaging approaches will be discussed.

Motor Manifestations of PD

Onset of PD

Several key questions revolve around the onset of brain pathologic changes that underlie PD. The initial studies in humans suggested that striatal dopamine deficiency must reach 70% to produce motor dysfunction^6,7. These studies used postmortem measures and extrapolated clinical onset from retrospective review of patients’ charts^6,7. Furthermore, the focus included striatum (caudate and putamen) rather than the putamen which preferentially affects motor function^8,9. Postmortem studies in people with incidental Lewy pathology and no history of parkinsonism revealed about 50% to 70% striatal dopamine reduction⁸. However, these postmortem studies may overestimate dopamine loss as post-mortem interval may reduce subsequent striatal dopamine measurements¹⁰. Thus, an in vivo technique to measure the integrity of nigrostriatal dopaminergic neurons could help address these limitations.

The most direct PET based measure to address this question would use a radiotracer specific for α-synuclein, however, such a PET radiotracer is yet to be developed. Alternatively, multiple investigators have used PET measures of nigrostriatal neurons targeting presynaptic sites in the striatum. The integrity of these presynaptic terminals has been evaluated primarily with three classes of radiotracers: 6-[¹⁸F]fluorodopa (FD; reflects aromatic amino acid decarboxylase (AADC) that converts levodopa to dopamine and its uptake and storage in presynaptic terminals), [¹¹C]dihydrotetrabenazine (DTBZ; reflects vesicular monoamine transporter type 2 (VMAT2) in the membranes of presynaptic vesicles that store dopamine in preparation for synaptic release), and 2-beta-[¹¹C]carbomethoxy-3-beta-4-fluorophenyltropane (CFT; reflects membranous dopamine transporter (DAT) specific for dopaminergic neurons important for reuptake of synaptic dopamine back into the presynaptic neuron). Multiple different DAT radiotracers including radiopharmaceuticals for SPECT (mostly tropane derivatives) have been used. CFT is a commonly used PET radiotracer. We will critically evaluate the results of some key prior PET studies using these radiotracers and discuss their potential roles from a clinical and research perspective.

The potential clinical utility of PET to identify early changes in PD was promising as images revealed decreased striatal accumulation in people with PD^11–14. Studies with the other classes of radiotracers targeting presynaptic nigrostriatal neurons also reported reduced striatal uptake in either people with PD or in animal models of parkinsonism^15–18. This also set the stage for trying to determine when pathophysiologic changes begin in the brain with PD. All three radiotracer types gave similar answers - linear back extrapolation of rate of change in early PD suggested that onset of pathologic changes began less than seven years before diagnosis¹⁹.

Another entirely different approach yielded very similar result. Using PET scans of [¹⁸F]fluorodeoxyglucose to obtain regional measures related to glucose metabolism, a covariance pattern was found in a principal components analysis (PCA) of the regional metabolism. This pattern of covarying regions has been called the Parkinson disease related pattern (PDRP), and a single subject score can be calculated to indicate how strongly this covariance pattern exists in an individual. The strength of this single subject score correlated with severity of motor ratings of PD, thereby providing face validity for the approach. Back extrapolating the strength of these single subject scores would predict that the abnormal covariance pattern began about five years prior, close to the same range as found with the presynaptic markers²⁰. This same group also noted abnormal PDRP expression in the ipsilateral presymptomatic hemisphere in hemiparkinsonian patients and concluded that the PDRP pattern expression becomes abnormal about two years before the onset of motor symptoms suggesting its potential as a prodromal marker of PD²¹. PDRP expression also was elevated in people with idiopathic REM sleep behavior disorder (RBD) at levels comparable to the early PD^22,23. Progression to PD could be predicted by the baseline PDRP expression pattern²⁴. Thus FDG PET can aid in identifying presymptomatic individuals especially in an at-risk population.

The limitation of all of these methods is the assumption of a linear degeneration of either the nigrostriatal pathway, a subsequent disturbed covariance pattern of brain metabolism and a linear change in motor severity. Subsequent studies have shown the limitations of these assumptions. In fact, applying a multivariable, nonlinear approach to presynaptic markers based upon longitudinal scans in people with PD produced an estimate of onset of striatal abnormality about 17 years before diagnosis²⁵.

It remains to be determined whether the first pathologic changes in the brain that indicate onset of PD involve a widespread network dysfunction, that could in fact reflect early deficits in caudal brainstem pigmented nuclei that project to many cortical areas or is the initial deposition of abnormal α-synuclein. If the latter is true, then measurements of abnormal uptake of α-synuclein radiotracer may shed light on this remaining question.

Identification of presymptomatic PD patients

Identification of subclinical pathologic defects in the nigrostriatal pathway is critically important for research and potentially for clinical treatment. The importance for clinical treatment will be relevant once we actually have a treatment that can forestall disease onset in a population at risk or even retard the tempo of progression after onset. Hence an objective PET imaging marker that can detect the earliest pathophysiologic changes has substantial value. Molecular imaging measures of presynaptic nigrostriatal markers may be helpful in this regard.

The motor symptoms of PD usually develop following the loss of a substantial proportion of putaminal dopamine terminal function whereas a lesser degree of nigrostriatal loss may not cause any clinical deficits. Thus, loss of nigrostriatal neurons begins before illness develops and presents a window of opportunity for presymptomatic identification of early neuronal damage in people with PD. Early studies provide support for this notion. An FD PET study showed abnormal uptake in four MPTP (a neurotoxin that selective destroys dopaminergic neurons) exposed individuals even before the onset of clinical parkinsonism¹². Another FD PET study of 17 asymptomatic monozygotic and dizygotic twins of patients with PD reported significantly reduced putaminal FD uptake (83–87% of normal) but no significant difference was evident in the caudate²⁶. Interestingly, one of the monozygotic twins in this study developed PD symptoms within two years highlighting the potential of FD PET to detect nigrostriatal deficits in presymptomatic individuals. But this type of study must be interpreted with caution. This subject had a postural tremor at the time of the PET, which strictly speaking can be considered to be a forme fruste initial manifestation of PD, thus this may have been a symptomatic patient rather than one without any manifestations²⁶.

The data from animal models are somewhat clearer. MPTP treated cynmologous monkeys demonstrated significant reduction of striatal FD uptake even in the absence of clinical signs²⁷. This was further corroborated by another group where sequential MPTP treated monkeys at 3–6 week interval demonstrated significant reduction of uptake of radiolabeled β-[¹¹C]DOPA and a DAT radioligand [¹²³I]PE2I even before the emergence of tremor. The tremor coincided with the reduction in [¹²³I]PE2I binding to about 15% and β-[¹¹C]DOPA uptake to about 34% of pretreatment levels²⁸. In a separate, large scale longitudinal study using all the three presynaptic dopaminergic markers, only DTBZ but not FD or another DAT marker [11C]d-threo-methylphenidate (MP) was noted to be significantly reduced in PD patients with younger disease onset and estimates based on extrapolated calculations (without actual data points) suggested the ability of DTBZ, MP and FD to identify presynaptic dopaminergic deficits about 17, 13 and 6 years respectively prior to actual onset of motor symptoms^19,29,30 while some others have projected disparate estimates ranging up to 50 years based on linear calculations³¹. Reduced DAT binding (using MP) was noted to be one of the earliest imaging markers in asymptomatic carriers of the autosomal dominant LRRK2 (leucine-rich repeat kinase 2) mutation. The reduced DAT binding was observed in the setting of normal FD PET, but the onset of clinical symptoms coincided with abnormal FD uptake³². A recent longitudinal study showed 14 out of 21 (67%) hyposmic individuals with an abnormal (uptake <65% of predicted) DAT SPECT converted to PD at 4 years compared to 2 of 22 hyposmic individuals with indeterminate and 3 of 109 with a negative DAT SPECT scan who went on to develop PD³³. Thus molecular imaging can be more sensitive than clinical findings and this could be of particular value in an at-risk population, although caution must be exercised as the line of demarcation between normal and abnormal findings can be somewhat murky especially with minimal defects in the nigrostriatal pathway or in individuals with only subtle clinical deficits. The clinical utility of presymptomatic identification remains limited until development of a treatment which could forestall the disease manifestation.

On the other hand, such presymptomatic identification has research implications. For example, potential identification of an “affected” individual could act as an endophenotype for genetics research. PET potentially could “convert” a normal to an “affected” for genetic investigations of pedigrees with familial parkinsonism. However, some patients with overt early PD manifestations including asymmetric resting tremor with bradykinesia and rigidity may have entirely normal FD PET and yet go on to develop typical PD manifestations³⁴. Thus, a normal PET does not exclude PD and may misidentify a person with presymptomatic PD.

Measurements of PD progression

An objective measure of PD progression can be a critical part of assessment of a disease modifying therapy³⁵. Many have suggested that PET or SPECT measures of striatal uptake of presynaptic radiotracers may provide such an objective measure. This assumes that progressive loss of nigrostriatal neurons reflects disease progression²¹. As noted, quantification of abnormal deposition of α-synuclein may be a more direct measure but that is not yet available.

Multiple clinical trials of PD used molecular imaging to measure effects of an intervention. However, these presynaptic nigrostriatal molecular imaging biomarkers did not match clinical measures of disease progression. The ELLDOPA study prospectively tested three different doses of levodopa and placebo in a large cohort of PD patients (n=361)³⁶. Clinical progression determined by a standardized clinical rating scale of motor parkinsonism yielded a completely different result from a SPECT-based measure of striatal DAT. The clinical measure suggested that those treated with the highest dose of levodopa had reduced disease progression over nine months whereas the SPECT DAT suggested in a large subgroup that this group fared worse. The clinical measures were thus discordant with the imaging measures. Several other interventional studies also reported discordant results^37–41. These discrepancies raise questions about the validity of the neuroimaging measures and how they had been validated^42–44. The bottom line is that the striatal uptake of any of the three presynaptic tracers correlated well with striatal dopamine but not with nigral dopaminergic cell bodies⁴⁵.

To address these issues, we performed a meticulous analysis of striatal uptake using the three PET tracers (FD, DTBZ, CFT) in MPTP treated non-human primates^45,46. All animals had PET imaging at baseline and at 8 weeks after unilateral intracarotid infusion of varying doses of MPTP. We previously demonstrated that this approach produces a stable, chronic hemiparkinsonism by 3–4 weeks⁴⁷. Animals had blinded video motor ratings, were euthanized after the second PET for striatal dopamine measurements and unbiased counts of tyrosine hydroxylase stained nigral cells (this identifies dopaminergic cells). The relevant finding for this part of the review is that parkinsonism developed once animals had 25–35% loss of either striatal dopamine or unbiased stereologic counts of dopaminergic cell bodies in substantia nigra. Striatal PET measures of all three radiotracers types correlated strongly with striatal dopamine and with severity of parkinsonism. However, striatal uptake of any of the tracers no longer correlated with parkinsonism or nigral dopaminergic cell counts once nigral cell loss exceeded 50%. These data suggest that nigrostriatal reserve is substantially less than previously thought, in other words, less nigrostriatal injury is required to produce parkinsonism, at least in nonhuman primates after unilateral MPTP-induced injury.

This means that motor parkinsonism correlated with nigral cell counts but not with striatal dopamine. Once nigral cell counts decreased to about 50% striatal dopamine were near zero, yet motor parkinsonism continued to progress as nigral cell counts further decreased. One could speculate that this relationship occurred only in the MPTP (toxin) induced injury. However, similar results were found in people with PD. Postmortem measures from humans that had PD revealed that nigrostriatal neurons appear to reach a nadir when a person had only mild to relatively moderate PD⁴⁸. Similarly, PET measures of striatal uptake of a presynaptic radiotracer in a longitudinal study in PD demonstrated that posterior putaminal uptake of a DAT marker reached a nadir within five years of PD diagnosis⁴⁹. Thus two studies in human PD confirm the findings in MPTP-treated nonhuman primates.

Thus surprisingly, striatal uptake of any of the presynaptic markers only correlates with motor progression during the early course of PD⁴⁵. Interestingly, PET measures of midbrain uptake of either a DAT or VMAT2 radiotracer did correlate well with nigral dopaminergic cell counts and severity of parkinsonism throughout the full range of parkinsonism severity and has potential to act as a neuroimaging biomarker of nigral dopaminergic cell loss and severity of parkinsonism⁵⁰.

These findings also likely explain many of the reported discrepancies in studies of PD progression. Most of the studies recruited unmedicated, newly diagnosed patients and then followed them in the study. During the course of the follow up, many may have reached the nadir in striatal measures—leading to striatal molecular imaging measures that no longer correlate with severity of parkinsonism and is likely fraught with substantial noise^35,45.

FDG and principal components analysis have been proposed as a measure of disease progression. A longitudinal analysis spanning 4 years was carried out in 15 early PD patients within 2 years of diagnosis⁵¹. The patients underwent scanning with FDG as well as PET imaging with a presynaptic dopaminergic terminal marker [¹⁸F]-fluoropropyl βCIT (FP-CIT), a marker of presynaptic DAT at multiple time points including baseline, 24 and 48 months. UPDRS motor measures also were collected at each time point. Statistical parametric mapping (SPM) analysis revealed longitudinal increase in glucose utilization over time in subthalamic nucleus, globus pallidus interna, dorsal pons, primary motor cortex with concomitant decrease in the prefrontal and inferior parietal regions. This was associated with a significant increase in both PDRP and the PD related cognitive pattern (PDCP); the latter progressed at a much slower pace reaching abnormal values only at the 4 year mark. As had been noted in multiple prior cross sectional FDG PET studies in PD patients of varying motor severity^52,53, the PDRP expression correlated well with UPDRS at each time point. The PDRP also correlated with the declining striatal FP-CIT uptake. The PDCP did not correlate with the UPDRS or the DAT binding and no cognitive measures were available for correlation. The trajectory of progression of the disease specific spatial covariance patterns and their behavioral correlations has not been assessed in longitudinal studies with a bigger population or over a longer time interval.

Selection of participants for clinical trials of PD

Since parkinsonian manifestations can overlap with other types of parkinsonism, a means to reliably identify those with underlying PD could potentially enhance the homogeneity of recruited participants for clinical trials of PD. Can molecular imaging help in this regard? Even if this works, the generalizability of the study will be limited. If a molecular imaging is used to screen participants in an interventional trial then the findings in the trial may only apply to those who are screened in the same manner. Thus, if an intervention demonstrates slowing of progression in this type of study, then application of the intervention may require similar molecular imaging screening.

Approaches to this differential diagnostic process can be considered at two levels: the ability to reliably differentiate PD from healthy individuals and at the second level to differentiate PD from atypical parkinsonism (APs that includes MSA, PSP, etc.). Multiple PET imaging studies with presynaptic terminal markers have reasonably distinguished PD from controls (Fig 1) but as noted some false negatives do occur^13,54,55. Nevertheless, presynaptic terminal imaging markers have great value from a research perspective to help define a more homogeneous group with abnormal scans as inclusion criteria for enrollment, but in the same vein it would reduce the generalizability of the results of the study to only those who meet the scanning criteria.

Figure 1.

6-[¹⁸F]fluorodopa (FD) PET in a normal individual (A) and two PD patients with mild (B) and moderate (B) motor impairments showing asymmetric, gradient loss of putaminal FD uptake in PD patients compared to controls.

But the ability of PET with presynaptic dopaminergic terminal markers to distinguish PD from APs remains to be proven. The pattern of striatal FD uptake in PD lacks specificity; other conditions including MSA, post encephalitic parkinsonism, spinocerebellar ataxia type 2 can produce similar changes⁵⁶. The striatal FD uptake in clinically diagnosed MSA was heterogeneous ranging from putaminal sparing identical to PD to a more uniform striatal involvement that has been described as the characteristic pattern of PSP⁵⁷. One study initially suggested the ability of DAT SPECT to pick up a statistically significant reduction in DAT uptake in the caudate in PSP⁵⁸. Another larger cross sectional DAT SPECT study reliably differentiated parkinsonian syndromes (57/61 had abnormal scans) from ET (none of 11 had abnormal DAT uptake) but failed to differentiate PD from APs ^59,60. Similar limitations occur with DAT SPECT scans and this been reviewed recently^35,61.

PET imaging could rationally influence clinical decision making process. PET-assisted diagnosis could help select appropriate candidates for deep stimulation surgery or help identification of implantation sites for deep brain stimulation (DBS). Whether this targeting strategy produces clinical benefit remains to be determined. It is also important to reiterate that optimal clinical management still heavily relies on meticulous clinical assessment and close follow-up. Hence PET imaging with presynaptic dopaminergic terminal imaging markers can reasonably distinguish PD (especially with moderate motor impairment) from controls but has not yet proven its worth in differentiating PD from atypical parkinsonisms or in their accurate differential diagnosis.

PCA based disease specific spatial covariance pattern of FDG may help differentiate AP from PD. An initial FDG PET analysis identified relative striatal and thalamic hypometabolism in 75% of AP cases when compared to PD⁶². Specific patterns of regional hypometabolism were also noted in MSA marked by reductions in putamen and caudate compared with disease duration/severity matched PD patients or controls⁶³ while another separate study on an independent cohort found regional hypometabolism additionally in the cerebellum and brainstem in nine early stage clinically diagnosed MSA patients⁶⁴. The autonomic symptoms correlated with brainstem glucose metabolism while the striatal FD uptake correlated with the motor symptoms⁶⁴. Analysis of 41 clinically diagnosed probable or possible PSP subjects found relative reductions in glucose utilization quite diffusely in the frontal, parietal association, thalamus, caudate and cerebellum⁶⁵. Cortico-thalamic metabolic asymmetry that was greater than control subjects or asymmetric PD patients were noted in FDG PET scans from five patients with corticobasal degeneration (CBD)⁶⁶. Another study used statistical parametric imaging (SPM) tools to create disease related templates from FDG PET scans from a random selection of eight patients from each of PD, MSA, CBD and PSP patient groups. Application of this template either by visual assessment or a blinded SPM based computer assisted interpretation of individual FDG PET scans showed a diagnostic accuracy (compared to clinical diagnosis at 2 year follow-up) of 85.4% and 92.4% respectively⁶⁷. While the numbers were not drastically different from prior clinico-neuropathologic correlations^68,69, it is interesting to note that the computer assessment agreed with the clinical assessment for only 86.5% of healthy control subjects while it was about 90.9% for the visual inspection. Specific disease related spatial covariance patterns may exist for MSA^70,71, PSP⁷¹ and CBD⁷² but their utility in accurate differential diagnosis is not entirely clear. In a recent study of 51 PD with 127 AP patients, the PDRP in nondemented PD patients while significantly different from controls was not significantly different compared to the MSA-P, PSP, or CBD subgroups. Considerable between groups overlap was also evident with MSA, PSP or CBD related patterns, although it is important to note that dopaminergic medications were not held prior to scans which could have eroded some of the group differences in this analysis⁷³. Another group utilizing a distinct methodological approach of automated voxel-based multivariate supervised machine learning reported even lower diagnostic accuracy of specific AP subtypes, varying between 45–62%⁷⁴. A more recent study utilizing the previously delineated metabolic patterns and a two level logistic classification algorithm reported the ability to distinguish PD from AP with 94% specificity and 96% PPV in the first step. The indeterminate cases (~19%) were not included in the second step of analysis, where the AP cases were further differentiated into either MSA or PSP. This further revealed 79% sensitivity, 90% specificity and 85% PPV for MSA while the sensitivity, specificity and PPV for PSP was reported at 100%, 94% and 94% respectively. 60% of the subjects had disease duration of less than 2 years and a greater rate of false negatives was noted for classifying AP patients with short duration of symptoms. Also, important to note that the PPVs reported here were more reflective of the study population with 37% AP patients, that was significantly higher than the general population^23,75. Of note, the clinical diagnosis was selected as the gold standard and there was no neuropathologic confirmation of the inferences made with imaging in any of the abovementioned studies. As previously discussed the positive predictive value of a clinical diagnosis of the AP subtypes hovers around 80% and could be even worse early in the disease process, so drawing critical inferences based on the above results might be premature. FDG-PET could have a role in the differential diagnosis of AP subtypes, but further studies and preferably pathologic correlations are required at this stage.

Thus, overall, molecular imaging may help identify a more homogenous cohort for clinical trials but at the expense of loss of some of the generalizability of the study findings.

Non-motor manifestations

Measure of cognitive impairment

Several molecular imaging approaches have been used to investigate pathophysiology underlying cognitive impairment that commonly occurs in people with PD. In this section, we will review FDG based measures, then discuss the role of receptor measures and finally address proteinopathy based approaches. The value of such studies is to try to identify new targets of engagement for therapies aimed at addressing this poorly treated aspect of PD.

FDG PET:

PCA based approaches of FDG PET has been a major focus of multiple studies. The PDCP spatial covariance pattern in FDG PET images is characterized by relative metabolic reductions in frontal and parietal association areas with concomitant increases in cerebellar vermis and dentate and was noted even in cognitively normal PD patients⁷⁶. The strength of the individual subject PDCP expression predicted memory performances, visuospatial function and executive functioning. Comparison of the PDCP pattern in cognitively intact PD versus single domain mild cognitive impairment (MCI) and multiple cognitive domains MCI revealed a stepwise increment in brainstem/cerebellar metabolism and concurrent decrements in the prefrontal and parietal metabolism with worsening cognition⁷⁷. The PDCP pattern has some topographic resemblance to the frontoparietal control network. Levodopa mediated changes in verbal learning occurred only in a subset of PD patients with cognitive impairment; these patients had a higher baseline PDCP expression and the cognitive response correlated with the levodopa mediated reduction of PDCP expression. PDCP expression was not different in the non-responders or in the subset of patients with improvement in cognitive measures after placebo⁷⁸. Analysis of FDG and FD PET in 106 PD subjects showed correlation of PDRP expression with FD uptake in caudate and putamen (and only putamen after controlling for PDCP expression). In contrast, PDCP expression correlated with the anterior striatum, but only with modest effect sizes suggesting the metabolic patterns did not only reflect underlying nigrostriatal denervation⁷⁹. Another longitudinal FDG-PET study by a separate group utilizing a volume-of-interest and stereotactic surface projection (SSP) based analysis revealed hypometabolism in the visual association cortex, posterior cingulate cortices and to a lesser extent in the caudate that could predict subsequent dementia onset in cognitively intact PD patients⁸⁰. Follow-up scans in a subset of these patients who progressed to dementia revealed interval reduction of metabolism in multiple cortical areas including the thalamus, posterior cingulate, occipital, parietal and frontal cortices compared to baseline with relatively preserved metabolism in the temporal cortex. A recent longitudinal study with 79 newly diagnosed PD patients and 20 controls reported hypometabolism in the occipital and inferior parietal lobes in the PD cohort; the parietotemporal and frontal metabolism correlated with memory based and attentional tasks; while the baseline parietal hypometabolism predicted the MMSE and MoCA at 18 months⁸¹. Prior cross-sectional and longitudinal SPECT studies revealed regional hypoperfusion in PD patients with dementia in those same regions as compared to PD patients with intact cognition^82,83. Interestingly the cuneus has the most severe cholinergic denervation in demented PD patients and the reduced glucose metabolism could possibly reflect an ongoing cholinergic deafferentation⁸⁰.

Cholinergic system:

Measures of cholinergic neurons also may be particularly pertinent to non-motor features of PD. Initial PET studies using N-[¹¹C]methyl-piperidin-4yl-propionate (PMP), an acetylcholine analogue and marker of acetylcholinesterase (AChE) activity showed significantly reduced cortical AChE activity in both PD with dementia (PDD) (−20.9%) and PD without dementia (−12.7%) compared to controls and the AChE activity correlated with attentional and executive functioning as well as working memory but not with severity or duration of motor symptoms⁸⁴. Another PMP PET study reported similar findings additionally showing maximal reduction in medial occipital cortex (BA18) in PDD and dementia with Lewy bodies (DLB) patients; no group difference in AChE activity was noted between early or advanced PD patients or between DLB and PDD⁸⁵. PET study utilizing PMP, DTBZ and VMAT2 compared PD patients with versus without falls; the PD fallers had significantly reduced cortical and thalamic AChE activity compared to nonfallers. But more interestingly, the thalamic AChE activity was significantly reduced only in the fallers but not the nonfallers compared to controls; this was also noted to be decreased in fallers even after adjusting for nigrostriatal dysfunction⁸⁶. This highlights the importance of the cholinergic outputs from the PPN to the thalamus in regulating gait and postural stability. AChE activity also correlated with gait speed in PD patients without dementia and no significant difference in gait speed was noted with isolated nigrostriatal dysfunction in the absence of cholinergic denervation⁸⁷. Another PET study using radiopharmaceuticalN-methyl-[¹¹C]2-(4’-methylaminophenyl)-6-hydroxybenzothiazole also known as Pittsburgh compound B (PiB), DTBZ and PMP in 143 PD patients of which 20 had freezing of gait (FoG) found frequency of FoG was the highest (41.7%) in PD patients who had both increased amyloid burden (PiB+) as well as reduced AChE activity, the frequency was 30.4% if they were just PiB+ and 4.8% if they were neither PiB+ nor had any cholinergic denervation⁸⁸. Even though the PD patients with FoG had significantly reduced DTBZ uptake, the role of extra nigral pathology in FoG was highlighted by this study. Studies with radiotracers targeting the postsynaptic muscarinic receptor showed increased muscarinic acetylcholine receptor (AChR) density in the frontal cortex of PD patients but without any behavioral correlates^89,90. Postsynaptic nicotinic AChR density was also analyzed in PD patients with 2-[¹⁸F]fluoro-3-(2[S]-2-azetidinylmethoxy)-pyridine (2FA); this showed reduced uptake in multiple cortical and subcortical regions in PD patients compared to controls, and the binding correlated with both depressive and cognitive symptoms⁹¹.

PET studies using radiotracers targeting vesicular acetylcholine transporter (VAChT) is a substantial advance that enables qualitative and quantitative analysis of presynaptic cholinergic nerve terminals. The initial human PET study in healthy controls using (−)-5-[¹⁸F]-fluoroethoxybenzovesamicol (FEOBV) tracer, a vesamicol derivative that selectively binds VAChT showed highest binding in the striatum, followed by thalamus, vermis and cerebral cortex⁹². The original developers of this radioligand are currently applying this to a large longitudinal cohort of people with people but no studies in PD have been reported to date. We also characterized a VAChT tracer in nonhuman primates^93,94 and studies in PD patients are currently ongoing. Thus cholinergic PET imaging has immense potential to delineate the pathophysiologic bases of two key causes of morbidity and mortality in PD patients that are refractory to levodopa, namely gait and cognition. With the advent of superior radiotracers that are able to better demonstrate cholinergic pathology in subcortical areas of interest especially cerebellar vermis, lot remains to be told in this area.

Proteinopathy:

PET assays with radiotracers targeting abnormal proteins have enhanced our understanding of the underlying pathophysiology of neurodegenerative disorders. The prodromal pathology or disease progression and response to therapy could be best assessed by a radiotracer specific for α-synuclein. This still continues to elude the field because of technical challenges of finding a lipophilic ligand with a low background signal that would effectively cross the blood brain barrier and then specifically bind to the intracellular oligomeric α-synuclein⁹⁵. Radiolabeled benzoxazole compounds or phenothiazines lack specificity given their affinity for β-amyloid and/or tau and high signal to noise ratio precludes their use for invivo imaging⁹⁶.

PiB PET amyloid imaging has enabled the in vivo measurement of fibrillary β-amyloid (Aβ)⁹⁷. This has been particularly important in delineating the pathophysiology of cognitive impairment in Alzheimer disease (AD) where Aβ deposition precedes the abnormal deposition of tau and onset of dementia; in fact the increased uptake of PiB predicts an early onset of dementia^98,99. The presence of amyloid pathology has long been reported in neuropathologic studies of PD especially in the presence of dementia but the prevalence of classical AD pathology or its role in dementia in PD was far from clear^100–102. A PET study with PiB in 13 DLB, 12 PDD, 10 PD with normal cognition and 41 age matched controls showed significantly increased amyloid load in over 80% of subjects with DLB and in only two of the subjects with PDD, although there was no significant difference in the regional pattern of uptake. None of the nondemented PD subjects had an increased cortical PiB uptake¹⁰³. Another study found a similar proportion of DLB and PDD subjects with increased PiB uptake (PiB+) but additionally noted reduced cerebrospinal fluid (CSF) Aβ42 level, lower MMSE and higher ApoE4 prevalence in PiB+ patients. Motor symptoms did not correlate with the PiB binding; this led them to contemplate the predominance of AD pathology in the subjects with increased PiB uptake¹⁰⁴. But it was far from clear whether a positive PiB scan actually indicated an AD pathology and the true answer lay in the pathologic validation of the PET findings.

Initial data from a large scale ongoing longitudinal study at our center has helped clarify this. Analysis with only a limited number of subjects found an equal proportion of subjects in each of the groups based on cognitive status (PD with normal cognition, PD-MCI, PDD, DLB and age matched healthy controls) with elevated PiB binding potential (BP) for predefined cortical regions and mean cortical binding potential (MCBP). Elevated PiB BP was associated with worse global cognition but did not correlate with earlier onset or faster rate of progression of cognitive impairment¹⁰⁵. Initial pathologic correlation from 3 individuals with PDD who had a PiB PET imaging within 15 months of death showed the two PiB+ PET scans had abundant diffuse Aβ plaques but only scant neuritic plaques and intermediate neurofibrillary tangle pathology and hence did not meet the NIA-Raegan criteria for the diagnosis of AD. All three patients had abundant cortical Lewy bodies (Braak PD stage 6), thus the PiB-PET scan in the third individual who had only rare diffuse Aβ plaques proved the specificity of PiB-PET for fibrillary Aβ amyloid and not α-synuclein (Fig 2). More importantly this study showed that PiB+ PET was not sufficient to indicate coexisting AD pathology in a PD patient¹⁰⁶. In a subsequent autopsy series of 32 cases, it was clearly evident that classical AD pathology (with moderate tauopathy) was present in only about 3% of the PD cases. But further analysis showed substantial accumulation of Aβ in addition to cortical synucleinopathy predicted significantly shorter survival rates. Although there was a trend towards it, it did not necessarily predict an earlier onset of dementia¹⁰⁷. Interestingly even though Aβ pathology was noted in 59% of the PD patients, only about 20% had PiB+ PET scans in this study. While the time lag between initial PiB PET and autopsy could contribute to this discrepancy, the specificity of the PiB tracer for labeling only the fibrillar Aβ seems to be the major determinant in this case¹⁰⁸. A study with baseline PiB PET and longitudinal clinical measures showed the increased PiB uptake predicted rapid decline in cognition but did not alter the tempo of motor progression in PD patients¹⁰⁹. A cross sectional PiB PET study of 40 patients of PD with MCI showed only 6 (15%) subjects with PiB binding at levels comparable to AD, but a significant correlation between cortical PiB binding and global composite cognitive function was noted¹¹⁰. This pattern of higher PiB+ studies in DLB compared to PDD, and the lower prevalence of PiB+ scans in PD-MCI as compared to non-PD related MCI was also evident in a recent systematic review of the literature¹¹¹. The pooled prevalence of PiB+ studies was 0.68 in the DLB group, 0.34 in the PDD group and 0.05 in the PD-MCI group. Interestingly this was lower than the 25–30% range reported in prior studies of unaffected elderly subjects. Another study found higher cortical amyloid burden in the DLB group, comparable to AD patients, but amyloid deposition in the PDD group was low and comparable to the PD and normal controls. The amyloid deposition in the lateral parietal/precuneus and posterior cingulate regions in the DLB, PDD and PD groups correlated with visuospatial impairment¹¹². A PCA analysis of PiB PET scans from AD patients and PD patients with cognitive impairment showed contrasting PiB binding pattern in the two groups possibly alluding to a different pathophysiologic role of Aβ deposition in cognitive impairment in PD¹¹³. CSF-Aβ was found to be lower in nondemented PD patients compared to controls and the PiB MCBP inversely correlated with the CSF-Aβ level. Interestingly, the CSF-α-synuclein level was also found to correlate with the CSF-Aβ level thus suggesting a pathophysiologic connection between their metabolisms in PD¹¹⁴. Thus a PiB PET scan in PD patients reflects fibrillary amyloid burden but this does not necessarily translate to a coexisting AD pathology; and this may predict a more rapid cognitive decline and possibly shorter survival.

Figure 2.

[¹¹C]-PiB PET images from three PD patients with dementia (Case 1–3) and a normal individual. Case 1 and Case 2 both have increased PiB signal in multiple cortical areas signifying diffuse amyloid burden, while Case 3 had only minimal cortical PiB signal like the normal individual. The signal in the white matter areas is likely secondary to non-specific PiB binding. (Image adapted from Burack et al.¹⁰⁶)

Tau depositions reflecting neurofibrillary tangle (NFT) pathology have been detected with [18F]-AV-1451 (AV) radiotracer in non-PD patients with MCI¹¹⁵. A recent PET study with 26 PD patients (nine had MCI) and 25 controls reported no significant difference between the PD-MCI and PD patients with normal cognition, in keeping with our prior pathologic analysis¹¹⁶. Another cross sectional study of AV-PET and PiB PET in 7 DLB, 8 PD with cognitive impairment and 9 PD patients with normal cognition reported somewhat contrasting findings. AV cortical uptake was significantly elevated in DLB patients compared to controls particularly in the inferior temporal gyrus and precuneus and this correlated with cognitive measures¹¹⁷. Quite interestingly, the PiB uptake did not correlate with AV PET measures. Also, 17 out of the 21 patients with Lewy body disease had a low amyloid burden (distribution volume ratio <1.15) but 4 of these 17 patients had elevated AV uptake in spite of a low amyloid burden. This was in stark contrast to our prior pathologic findings¹⁰⁷. The nascent data with AV PET needs to be interpreted with caution given the important lesson from our experience with PiB that further validation especially pathologic correlation could be the key missing piece; also binding specificity of the AV tracer in an environment with PD pathology may be different from AD. So this remains to be sorted out.

Dopamine receptors:

Dopamine receptors belong to two broad categories: the D₁—like family (D₁, D₅; usually presynaptic; stimulates adenylyl cyclase and generates cAMP), and D₂—like family (D₂, D₃ and D₄; inhibits adenylyl cyclase and decrease cAMP levels; D₂ receptors can be both pre or postsynaptic). The striatum harbors an equal density of D₁ and D₂ receptors¹¹⁸. No pathophysiologic correlates of striatal D₁-like receptor density were found in PD patients^90,119 perhaps due to the low selectivity of the radiotracers¹²⁰. The D₂-like ligands can either be antagonists of high affinity that bind irreversibly to the D₂-like receptor, or of low affinity like ([¹¹C]raclopride, (RAC)). RAC because of its low affinity to the D₂-like receptor can be subject to competition by endogenous striatal dopamine concentration and striatal changes in RAC binding potential can reflect dynamic changes in the synaptic concentration of dopamine¹²¹. Multiple studies have highlighted the role of RAC PET in dissecting the pathophysiology of levodopa induced dyskinesia^90,122 and in predicting onset of motor fluctuations in PD patients¹²³. But more recent RAC PET studies have shed light on the pathophysiology of certain non-motor manifestations of PD. A RAC PET study showed significantly reduced dopamine release in the dorsal caudate but not in the anterior cingulate cortex of PD patients compared to controls during the performance of an executive task implicating the association of nigrostriatal dopaminergic deficit with executive dysfunction in PD¹²⁴; but detection of changes in anterior cingulate could be challenging. Another RAC PET study revealed greater synaptic release of dopamine in the ventral striatum of PD patients with pathological gambling both during gambling and during performance of a control task as compared to PD patients without pathological gambling¹²⁵. Significant reduction in the extrastriatal D₂ receptor density was noted in the hypothalamus of PD patients compared to controls¹²⁶.

Few studies have explored the role of D₃ receptors, in large part due to inadequate specific radiotracers. A modestly selective D₃ radioligand [¹¹C]-PHNO found low levels of hypothalamic D₃ receptors that correlated with excessive daytime sleepiness in PD patients¹²⁷.

Serotonergic receptors:

PET studies with [¹¹C]DASB, a marker of serotonin transporter have highlighted the pathophysiologic role of serotonergic dysfunction in multiple non-motor symptoms in PD including depression¹²⁸, fatigue¹²⁹, visual hallucinations and sleep problems^90,130. An inverse relationship between cortical serotonergic innervation and the amyloid burden in PD patients has also been reported; the relationship in the striatum is less clear^131,132.

Phosphodiesterases:

Phosphodiesterases (PDE) are intracellular enzymes that catalyze hydrolysis of the second messengers cAMP and cGMP; PDE10A is almost exclusively found in the axons of striatal medium spiny neurons and plays an important role in the dopaminergic signaling cascade¹³³. A PET study using [¹¹C]-IMA107, a PDE10A specific radioligand showed lower expression of PDE10A in caudate, putamen and globus pallidus and this correlated with disease duration and motor severity¹³⁴. A more recent study with [¹¹C]rolipam, a marker of PDE4 showed reduced expression of PDE4 in multiple subregions of the striato-thalamo-cortical circuit including caudate, thalamus, hypothalamus, dorsolateral prefrontal cortex, medial frontal cortex and supplementary motor area in PD patients compared to controls that correlated with deficits of spatial working memory¹³⁵.

Conclusion

In summary, multiple molecular imaging methods have provided exciting new critical insights into the pathophysiology of PD. These methods can help detect the onset of brain pathology, identify presymptomatic people destined to develop PD; provide a measure of disease progression - at least during the early milder stages of PD; can help to enrich cohorts of PD for clinical trials and investigate non-motor aspects of PD. However, the approaches used for these studies have important limitations that, in some cases, may restrict application or interpretation of findings, yet careful investigations can lead to significant advances in our understanding of PD.