Can Neuroimaging Predict Dementia in Parkinson's Disease?

Juliette H. Lanskey; Peter McColgan; Anette E. Schrag; Julio Acosta-Cabronero; Geraint Rees; Huw R. Morris; Rimona S. Weil


Brain. 2018;141(9):2545-2560. 

In This Article

Radionuclide Imaging

Metabolic Activity

Changes in brain metabolism can be measured using fluorodeoxyglucose (FDG) PET, which is sensitive to glucose uptake and also with single photon emission computed tomography (SPECT), which detects changes in cerebral blood flow. Areas of hypometabolism are seen using FDG-PET in patients with PD-MCI (González-Redondo et al., 2014) with larger areas in PDD (Jokinen et al., 2010), particularly in posterior regions (Garcia-Garcia et al., 2012; González-Redondo et al., 2014; Shoji et al., 2014; Tang et al., 2016) (Figure 1A). This reduction in temporo-parietal metabolism is also seen in some patients with Parkinson's disease without dementia, potentially reflecting early posterior cortical involvement in these patients (Hu et al., 2000).

Figure 1.

Cerebral hypometabolism and dementia in Parkinson's disease and grey matter atrophy in Parkinson's with cognitive involvement. (A) Regions of cerebral hypometabolism in patients with PDD overlap with regional atrophy. Adapted from González-Redondo et al. (2014). (B) Statistical maps of baseline 18F-FDG-PET data comparing patients with Parkinson's disease that later develop dementia with controls. Hypometabolism is seen in posterior brain regions particularly in cuneus and precuneus. Image adapted from Bohnen et al. (2011). (C and D) Vertex-wise comparisons of cortical thickness between patients with PD-MCI and Parkinson's disease without cognitive involvement. Atrophy patterns differ between studies, although atrophy in the precuneus is frequently reported. (C) Greatest atrophy seen in left precuneus. Modified with permission from Pereira et al. (2014). (D) Greatest atrophy seen in precuneus and bilaterally in superior parietal regions. Figure adapted from Segura et al. (2014). Lh = left hemisphere; Rh = right hemisphere.

SPECT studies show a similar picture, with reduced cerebral blood flow in patients with PD-MCI (Osaki et al., 2005; Derejko et al., 2006; Nobili et al., 2009), and even greater reductions in PDD (Kawabata et al., 1991; Sawada et al., 1992; Mito et al., 2005; Ma et al., 2008).

Longitudinal studies consistently show involvement of posterior regions in earlier stages of cognitive decline in Parkinson's disease, with reductions in baseline FDG-PET metabolism in posterior cortical regions in patients that later convert from cognitively-normal to PDD (Bohnen et al., 2011; Tard et al., 2015; Firbank et al., 2017; Baba et al., 2017; Homenko et al., 2017) (Figure 1B).

Cognitive changes may also be preceded by metabolic increases in other areas. Using a principal components analysis approach to FDG-PET data, frontal as well as parietal metabolic reductions were seen, alongside increases in other areas, including cerebellar vermis and dentate nucleus (Huang et al., 2007; Meles et al., 2015).

Dopaminergic Function

Dopamine projections can be probed in vivo with PET or SPECT using markers of dopaminergic terminal integrity and may relate to cognitive involvement in Parkinson's disease. Decline in cognitive function, particularly executive dysfunction, is associated with loss of caudate uptake on dopamine transporter (DAT) SPECT imaging (Nobili et al., 2010; Arnaldi et al., 2012; Ekman et al., 2012; Lebedev et al., 2014; Siepel et al., 2014; Pellecchia et al., 2015) and with reduced caudate dopaminergic function, as assessed using PET (Brück et al., 2001). Caudate uptake on DAT-SPECT imaging may even predict cognitive decline, especially when combined with other measures including age and CSF (Schrag et al., 2016).

Studies using PET radioligands that bind to dopamine D2 receptors, show reduced D2-receptor availability in the striatum (Monchi et al., 2006; Sawamoto et al., 2008) and orbitofrontal cortex (Ko et al., 2013) of neurologically normal people when performing executive tasks (Monchi et al., 2006; Sawamoto et al., 2008). These reductions are not seen in people with Parkinson's disease (Sawamoto et al., 2008; Ko et al., 2013), suggesting that the normal striatal and orbitofrontal release of endogenous dopamine in the striatum and orbitofrontal cortex during executive processes is impaired in Parkinson's disease.

Patients with PD-MCI have reduced availability of D2 receptors in the bilateral insula, compared to patients with Parkinson's disease and normal cognition, and this availability is positively correlated with executive function (Christopher et al., 2014). Importantly, no between-group differences in cortical thickness are seen in any region, suggesting that loss of D2 receptors in the insula contributes to executive dysfunction in Parkinson's disease, and that these changes are seen before structural alterations take place.

These studies suggest that executive deficits in Parkinson's disease are associated with dopaminergic dysfunction in both the striatum and cortex. In comparison to measures of atrophy, PET measures may be more sensitive to cognitive impairment (Christopher et al., 2014). However, as D2-receptor availability is most sensitive to executive dysfunction (Christopher et al., 2014), measures of D2-receptor availability may have less power to identify those individuals that actually progress to Parkinson's dementia (Kehagia et al., 2012).

Cholinergic Function

Radioligands of cholinergic enzymes allow in vivo assessment of cholinesterase activity. Several studies show lower cholinesterase activity in PDD than in Parkinson's disease without dementia (in patients who were not taking cholinesterase inhibitors) (Kuhl et al., 1996; Hilker et al., 2005; Shimada et al., 2009; Klein et al., 2010), particularly in parietal (Kuhl et al., 1996; Hilker et al., 2005; Shimada et al., 2009; Klein et al., 2010) and occipital regions (Kuhl et al., 1996; Klein et al., 2010). Decreased cortical cholinergic activity is also associated with poorer scores on cognitive testing (Bohnen et al., 2006, 2015; Lorenz et al., 2014). Whether early reductions in cholinesterase activity in parieto-occipital regions is associated with later dementia has not yet been explicitly shown, but in light of converging evidence for the importance of posterior dysfunction and cholinergic deficits as a precursor for Parkinson's dementia, these are likely to become important future measures.

Phosphodiesterase 4 Expression

11C-rolipram PET can measure expression of phosphodiesterase 4, an intracellular enzyme involved in synaptic plasticity and memory. Reduced expression of phosphodiesterase 4 in people with Parkinson's disease, specifically in caudate, thalamic and frontal regions, correlates with impaired spatial working memory (Niccolini, 2017). Notably, these reductions are observed in the absence of cortical and subcortical atrophy in regional analyses.

Amyloid-β and tau Imaging

α-Synuclein deposition is a key pathological hallmark of Parkinson's disease and related PDD but there is currently no radioligand that binds to α-synuclein. Other pathological substrates, especially amyloid-β and tau, are strongly linked with PDD (Compta et al., 2011; Irwin et al., 2013) and PET can be used to detect these. In PET studies, Pittsburgh compound B (PiB) binds to amyloid-β. A recent systematic review identified increased amyloid positivity in PDD compared with Parkinson's disease patients without dementia, where amyloid positivity was defined as those exhibiting Alzheimer's-range cortical amyloid deposition on PET imaging performed with PiB (Petrou et al., 2015). Across the 11 studies included in the meta-analysis, 21/74 people with PDD were amyloid positive, compared with only 3/60 patients with PD-MCI. Amyloid binding is also negatively correlated with cognition (Gomperts et al., 2013; Akhtar et al., 2017).

In a longitudinal study (Gomperts et al., 2013), increased baseline amyloid burden in Parkinson's disease without dementia was associated with higher risk of developing cognitive symptoms. However, in the same study, participants with highest baseline PiB-amyloid-β did not develop PDD, and other studies found no difference in intensity (Foster et al., 2010; Gomperts et al., 2012) or pattern (Campbell et al., 2013) of amyloid-β burden between PDD and controls, suggesting that PiB binding alone has low specificity for dementia prediction in Parkinson's disease.

Tau has been found to co-localize with α-synuclein (Compta et al., 2011) and MAPT polymorphism is associated with increased risk of dementia in Parkinson's disease (Williams-Gray et al., 2013). The radioligand 18F-AV-1451 binds strongly to tau (Dani et al., 2016) and a recent cross-sectional study found a correlation between 18F-AV-1451 uptake in the precuneus and inferior temporal gyrus with cognitive performance (Gomperts et al., 2016), but this has not yet been confirmed longitudinally.

Unlike PiB or 18F-AV-1451, which selectively bind to amyloid-β plaques and neurofibrillary tau tangles, respectively, 18F-FDDNP binds to both tau and amyloid-β aggregates. Buongiorno and colleagues (2017) found that binding of 18F-FDDNP globally and in lateral temporal regions was higher at baseline in people with Parkinson's disease who converted to PDD at follow-up than in patients who did not develop dementia and that baseline lateral temporal 18F-FDDNP binding correlated with worse performance at later cognitive testing. Given that post-mortem evidence suggests it is the combination of pathological proteins that is most discriminatory for PDD (Compta et al., 2011), it is of particular relevance that the best evidence for pathological protein imaging, in the absence of a specific α-synuclein ligand, is for a ligand sensitive to both amyloid and tau.

Neuroimaging of Neuroinflammation

Increased microglial activation is seen in Parkinson's disease both within regions with Parkinson's disease-related pathology and at distant regions (Imamura et al., 2003). Changes in microglial morphology could be in response to local, microenvironment cues (Mrdjen et al., 2018) and may precede the spread of pathology in neurodegeneration (Streit et al., 2009). As microglial activation is linked to increased expression of the mitochondrial translocator protein (TSPO), ligands that bind to TSPO indicate areas of neuroinflammation. A study using a TSPO ligand demonstrated increased cortical microglial activation in patients with Parkinson's disease (with and without dementia) and identified increased left parietal neuroinflammation in patients with Parkinson's dementia compared to Parkinson's patients without dementia (Edison et al., 2013). Importantly, even in Parkinson's patients without dementia, this measure of microglial activation in temporo-parietal, occipital and frontal areas negatively correlated with cognitive performance. Measures of microglial activity are also sensitive to the earliest stages of dementia with Lewy bodies (Iannaccone et al., 2013), suggesting that imaging measures of neuroinflammation such as these may have a role in predicting Parkinson's dementia.