Introduction: A Brief History of Amyloid Imaging

Until recently, the assessment of the molecular basis of neurological disorders was largely limited to postmortem tissue. This changed with the development of amyloid radioligands such as [11C]Pittsburgh compound B (PiB) (Table 1) [1]. PiB binds fibrillar amyloid with nanomolar affinity in postmortem specimens [24]. As a result, it avidly stains neuritic plaques, a hallmark of Alzheimer disease (AD). In addition, it also stains many diffuse plaques, depending on their content of fibrillar amyloid [46]. In contrast, PiB does not appear to significantly label soluble oligomers or monomers of β-amyloid (Aβ). At brain concentrations achieved in human imaging, PiB has negligible off-target binding for Lewy bodies [7], neurofibrillary tangles (NFTs) [8], or other protein aggregates in postmortem brain tissue. However, PiB also labels vascular Aβ, the defining feature of cerebral amyloid angiopathy [4, 9, 10]. As required for a successful biomarker of Aβ, amyloid imaging in AD reliably reports significant amyloid deposition, and reliably reports its absence in young healthy control subjects (HCS) [11, 12] and in patients with other neurodegenerative diseases, including autopsy-proven Creutzfeld–Jacob disease [13] and frontotemporal dementia (FTD) on the basis of semantic dementia, albeit without autopsy confirmation [14].

Because the short half life of 11C limits the use of PiB, a number of 18F-labeled amyloid ligands have been developed as well, including [18F]florbetaben [15], [18F]florbetapir [16, 17], and [18F]flutametamol [18, 19]. White matter uptake complicates the use of the fluorinated molecules, but this disadvantage has been offset by their significantly longer half-lives compared with [11C]PiB, which permits their use at sites that lack radiochemistry facilities. Like PiB, these amyloid ligands demonstrate specific affinity for amyloid plaques in postmortem tissue and show increased retention in AD patients compared with HCS [17, 1922] and neurodegenerative diseases not associated with Aβ accumulation, such as FTD [23•]. Head-to-head comparisons of PiB with these fluorinated ligands have demonstrated high correlations of cortical uptake of these tracers [24•]. Other amyloid ligands have also been developed, including 2-(1-{6-[(2-[18F]fluoroethyl)(methyl)amino]-2-naphthyl}ethylidene)malononitrile ([18F]FDDNP), which labels amyloid plaques, NFTs, and Lewy bodies [25, 26].

Amyloid imaging of older HCS has demonstrated a broad distribution of amyloid burden, with some HCS harboring amyloid loads well into the AD range [11, 12, 2731]. In fact, cortical deposits of amyloid can be detected in approximately 25 % of HCS [12, 27, 28, 30, 31]. Consistent with neuropathology studies, amyloid burden measured with PiB PET increases with age [12, 29] and with the presence of the apolipoprotein ε4 haplotype [12, 29, 30, 32]. These findings have called into question the interpretation of a “positive” amyloid scan. However, even in HCS, the presence of significant amyloid burden is associated with small but significant impairments of global cognition and episodic memory [33]. Furthermore, recent longitudinal studies have consistently detected a contribution of baseline amyloid burden to the subsequent development of cognitive impairment and dementia [31, 3437]. Thus, although amyloid burden in HCS may have only subtle repercussions for current cognitive function, it may nonetheless portend future cognitive decline.

Over the last 7 years, amyloid imaging has been applied fruitfully not just to AD but also to Parkinson disease (PD) dementia (PDD) and dementia with Lewy bodies (DLB). In this review, we will explore these efforts to relate imaged amyloid burden to clinical phenotype and to clinical course in PD and DLB.

Table 1 Characteristics of amyloid imaging

PD Dementia and DLB

Cognitive impairment is common in PD, increasing in severity with advancing age and with progression of motor impairment [3840]. As a result, dementia is overrepresented in PD relative to the general population, by up to a factor of 6 [41], with a prevalence of 30–80 %, depending on the motor stage and the duration of the disease [42]. A diagnosis of PDD is made when dementia develops in the setting of well-established idiopathic PD, which by consensus is taken to be at least 1 year of extrapyramidal motor symptoms [43]. PDD is clinically similar to DLB, a dementia associated with concomitant parkinsonism, visual hallucinations, and fluctuations of wakefulness and cognition [44]. In fact, these diseases share additional clinical features, including an association with REM sleep behavioral disorder, sensitivity to dopamine receptor antagonists, and response to acetylcholinesterase inhibitors. DLB and PDD share neuropathological features as well, including characteristic Lewy bodies and neurites (with involvement of cortical neurons frequent in PDD and requisite in DLB) [45]), and the loss of both nigral dopamine neurons [46] and basal forebrain cholinergic populations [47]. For these reasons, DLB and PDD are considered to be different clinical expressions of the same fundamental neuropathological processes. Their clinical differences, which consist primarily of the relative timing of dementia and parkinsonism, have been the focus of intense investigation [45].

Neuropathology Studies in PD and DLB

The biological basis for cognitive impairment in PD and DLB appears to be multifactorial. Aggregation of α-synuclein [48, 49] into oligomers, Lewy neurites and Lewy bodies and the involvement of cortical neurons contribute [50, 51], perhaps through secondary synapse dysfunction [52, 53]. Loss of monoamine neurotransmitter systems, including tegmental dopamine neurons and basal forebrain cholinergic neurons, which project to cognitive and limbic brain regions, also appears to play a role [54]. Finally, concomitant pathology characteristic of AD has been associated with both PDD and DLB [5557].

Although Lewy-body-associated neuropathologic changes comprise the dominant findings in PD, PDD, and DLB, these need not occur in isolation. Indeed, neuropathology studies in PD report a range of concomitant neuropathological findings characteristic of AD, including Aβ-containing plaques and tau-containing NFTs. Approximately 35 % of PD cases are estimated to fulfill Consortium to Establish a Registry for Alzheimer's Disease criteria for a pathological diagnosis of AD [58, 59]. Diffuse plaques are more common than cored plaques in PD [60]. When dementia is present in PD, higher densities of cortical and striatal (cored and diffuse) plaques have been observed, compared with nondemented PD [60], although diffuse plaques remain more prevalent [61]. Several clinicopathologic studies of PDD have also found an association between higher levels of Aβ and a faster transition to dementia [60, 62, 63].

Concomitant Aβ pathologic findings are more common in DLB than in PDD [55, 64, 65]. In fact, deposition of extracellular Aβ into neuritic and diffuse plaques is present in approximately 85 % of cases of DLB [50, 66, 67]. The presence of amyloid plaques in the striatum has been reported to differentiate DLB and PDD from PD [68]. Interestingly, in that study, striatal cored plaques were much more common in DLB than in PDD. In PDD and DLB, higher levels of Aβ have also been associated with greater risk of mortality [69, 70]. The capacity for amyloid burden to accelerate cognitive decline in PDD holds when DLB and PDD are studied together [64, 65].

In addition to Aβ, the intracellular accumulation of NFTs has been linked to both PDD [49] and DLB [57]. NFTs are frequently present but rarely marked. Thus, AD pathologic changes are commonly observed in PD and DLB, most often at levels insufficient for a pathological diagnosis of AD, but their presence appears to have clinical implications.

Amyloid Imaging in PDD and DLB

Several cross-sectional imaging studies have explored amyloid deposition in PD, PDD, and DLB. One general finding is that amyloid burden tends to be modest in PD, overlapping with that in HCS [27, 7175, 76•, 77]. Although cortical Aβ levels are often low in PD, they are not necessarily negligible. Indeed, specific cortical retention of PiB can be observed in some cases on direct visual assessment of the parametric images [27]. In addition, consistent with the broad distribution of amyloid burden in HCS described earlier, occasional PD cases demonstrate increased amyloid deposition as well, into the AD range [74, 75, 76•, 77]. Results obtained with florbetapir [23•, 78•] have been similar. Despite its capacity to label Lewy bodies, [18]FDDNP binding has also been reported to overlap in PD and HCS [79]. Together, these data demonstrate that the range of amyloid deposition in cognitively normal PD is similar to that seen in the healthy elderly population.

Amyloid burden is generally low in PD subjects with mild cognitive impairment (MCI), comparable to that in HCS [74, 76•, 80, 81•], and PiB uptake does not distinguish cognitively normal PD subjects from PD subjects with MCI. Low levels of cortical amyloid deposition have also been reported in most cases of PDD, overlapping with levels seen in HCS and frequently with levels observed in PD subjects as well [27, 71, 7375, 76•, 77, 82]. Visual assessment of PDD patients has been notable for the frequent presence of focal cortical deposits of amyloid [27]. Because aggregate scores of global cortical ligand binding are insensitive to small foci of specific cortical retention, quantitative assessment of small foci may prove useful in relating amyloid burden to clinical phenotype and prognosis.

A minority of PDD subjects demonstrate elevated cortical amyloid deposition, in the AD range [71, 7375, 76•, 77, 82]. The clinical characteristics of PDD patients with high and low amyloid burden appear to overlap [71, 76•, 77, 83•], although small sample sizes have limited these evaluations.

In contrast to the modest levels of amyloid observed in most PiB studies of PDD, amyloid deposition in DLB subjects is often high relative to that in HCS and PD subjects, in the AD range [23•, 27, 71, 73, 74, 76•, 78•, 84, 85, 86•, 87•, 88, 89]. Some DLB patients are clearly amyloid-negative, however. These negative scans presumably denote those with “pure” diffuse Lewy body changes without accompanying AD neuropathological changes, although one PiB-negative DLB patient proved to have substantial regional levels of Aβ despite low plaque burden at autopsy [90•]. These results are broadly consistent with the neuropathology studies described previously [55, 64, 65, 68]. Interestingly, higher cortical amyloid burden in a small cohort of DLB and PDD subjects has been associated with greater cortical and medial temporal lobe atrophy [83•], suggestive of superimposed classic AD. As observed in HCS, risk factors for higher amyloid burden in DLB and PDD subjects include age and the presence of the apolipoprotein ε4 allele [73, 76•, 84].

The clinical significance of Aβ deposits in Lewy body disorders remains a topic of active investigation. In some cross-sectional studies, global amyloid deposition has been found to relate to cognitive performance within diagnostic groups [23•, 74, 76•, 80]. In addition, in a small study, the absence of cortical amyloid deposition was associated with a better response to acetylcholine esterase inhibitors [88]. The higher amyloid burden of DLB relative to PDD has been offered as one potential explanation for the major clinical distinction between these related entities: the timing of dementia relative to parkinsonism [27, 64, 71, 84]. In this model, early and significant cortical amyloid burden may accelerate cognitive decline in patients with Lewy body disease. When this effect drives cognitive symptoms within 1 year of motor symptoms, this would result in a diagnosis of DLB.

One longitudinal study of nondemented PD subjects has been published to date. In that study, in which PD subjects were followed up for up to 5 years, those with higher amyloid burden progressed to cognitive impairment and dementia faster than those with lower amyloid burden [81•]. The hazard ratio for conversion was approximately 9 per unit PiB (distribution volume ratio). Not all individuals with positive PiB scans developed cognitive impairment, however, and some of the original cohort remained cognitively intact after 5 years of follow-up. These findings are similar to the results of longitudinal PiB studies of HCS (see earlier) [31, 3437].

Amyloid imaging thus identifies a contribution of brain amyloid deposition to both DLB and PDD that is detectable in life (Fig. 1). Higher amyloid burden in DLB may contribute to its accelerated cognitive phenotype relative to parkinsonism, possibly as a result of synergistic Aβ–α-synuclein neurotoxicity. Cortical amyloid burden in PD, although lower than in DLB, does not appear to be harmless, however, as it has been associated with longitudinal cognitive decline.

Fig. 1
figure 1

Representative [11C]Pittsburgh compound B (PiB) scans of a healthy older subject and patients with Alzheimer disease, Parkinson disease (PD), PD with dementia, and dementia with Lewy bodies. Amyloid deposition in PD is usually mild and comparable to that seen in normal control subjects. In PDD, amyloid deposition is common, but is not usually marked. Amyloid deposition in DLB is often significant, to levels that approach those observed in AD. Note that focal cortical deposits of amyloid may not be reflected in aggregate measures of PiB binding derived over large cortical regions

Amyloid Deposition and Motor Features of Parkinsonism

PD can present with distinct motor phenotypes, including a classic asymmetric tremor extrapyramidal syndrome, a postural instability gait disorder (PIGD) syndrome, and an akinetic rigid syndrome. Although all of these presentations are associated with the presence of Lewy bodies at autopsy, presentations that deviate from the classic asymmetric tremor syndrome carry a higher risk of cognitive decline [91, 92]. A small number of neuropathology and biomarker studies have explored the relation between regional amyloid burden and motor features of parkinsonism. Higher amyloid plaque density postmortem has been associated with the PIGD phenotype [63]. One PiB imaging study on this issue has reported a positive correlation between striatal amyloid burden and PIGD motor features in PD subjects with MCI [93•]. A similar result was found using CSF [94]. This possible relation between amyloid burden and PD motor phenotype is intriguing, as it would extend the influence of Aβ to the motor domain. However, these results should be viewed as preliminary and would benefit from confirmation in larger studies.

Relation of Amyloid Imaging to CSF Biomarkers

CSF biomarkers of amyloid burden provide an independent assessment of levels of Aβ. In AD, CSF levels of Aβ-42 fall early and levels of the intracellular protein tau and its phosphorylated form phospho-tau increase [95]. The reduction in CSF levels of Aβ-42 has been attributed to its aggregation into amyloid plaques, whereas the increase in the levels of tau species has been interpreted to reflect of neuronal injury. In AD, CSF levels of Aβ-42 and amyloid imaging findings are strongly correlated [96•].

To date, amyloid imaging and CSF assessment of Aβ in PD, PDD, and DLB have generated similar results. CSF levels of Aβ and tau do not reliably differentiate PD subjects from HCS cross-sectionally [97], although some PD subjects show levels of Aβ-42 that are intermediate between those of HCS and AD subjects [98]. In PDD, the level of Aβ can fall, with relative preservation of the level of tau [99]. Two longitudinal studies of cognition in PD have found that low CSF levels of Aβ-42 are associated with a greater rate of cognitive decline [100•, 101•]. One of these studies identified a tenfold greater risk associated with low CSF levels of Aβ-42 for developing dementia in PD; in contrast, normal Aβ-42 levels were associated with a very low risk for developing dementia [101•]. In DLB subjects, CSF levels of Aβ are frequently low, relative to those in PD subjects, PDD subjects, and HCS, whereas tau levels are more variable [97, 102]. As described above, one CSF study has also explored the relation of Aβ to motor function. Consistent with PiB imaging results, that study reported that low CSF levels of Aβ-42 were associated with PIGD clinical features in PD [94]. Together, the amyloid imaging and CSF Aβ-42 reports corroborate one another.

Concordance of Amyloid Imaging and Postmortem Neuropathology in DLB and PDD

As a bulk measure of brain amyloid accumulation, amyloid PET has had remarkable success. However, it is important to keep in mind that, as described already, amyloid imaging is neither a selective marker of neuritic plaques nor a measure of total Aβ burden. Rather, amyloid imaging is a highly specific marker of fibrillar amyloid [8]. If the goal were to measure total Aβ burden, PiB would necessarily fall short, given its inability to label all Aβ species, independent of their fibrillar content [7, 90•]. In addition, because PiB stains cored and neuritic plaques reliably, but also labels those diffuse plaques that contain fibrillar amyloid in a dose-dependent manner [46, 103], as well as cerebral amyloid angiopathy [9, 10, 103, 104], variation in the proportions of neuritic and diffuse plaques in PD and DLB subjects could conceivably contribute to differences in their amyloid imaging characteristics. These caveats constrain interpretation of amyloid imaging, in general and in PD, PDD, and DLB. Nonetheless, the accumulating data reviewed here suggest that the amyloid species that PiB detects are detrimental, and the ability to measure them antemortem is proving valuable.

Outstanding Issues

The observation that brain amyloid burden is often high in DLB, overlapping with that in AD, along with the high frequency of elevated amyloid deposition in HCS, constrains the specificity of amyloid imaging as a marker of clinical or preclinical AD. As DLB is the second most common neurodegenerative dementia, after AD [105, 106], positive amyloid imaging in dementia may signify DLB. Similarly, it will be important to determine the extent to which positive amyloid imaging in MCI patients or in healthy patients may portend future DLB rather than AD.

In addition, the possible contribution of even modest levels of cortical amyloid deposition to cognitive decline in PD is noteworthy. Some data suggest that the cognitive repercussions of amyloid burden in PD and DLB should be more severe than in the general population. On the one hand, amyloid burden may be an independent contributor to cognitive decline in PD and DLB, and amyloid deposition may occur sporadically in PD and DLB as it does in the general population. However, a number of studies have demonstrated a synergistic relationship between α-synuclein and Aβ deposition. For example, Aβ-42 and α-synuclein increase each other’s toxicity and aggregation [107, 108•]. Aβ deposition may therefore contribute to cognitive impairment in PD both directly, as it would in the general population, and indirectly, by enhancing α-synuclein toxicity. As a result, Aβ may be more toxic in the synucleinopathies, and patients with Lewy body disease may be unable to tolerate even modest levels of Aβ. Amyloid imaging may therefore carry particularly important prognostic information in PD and DLB.

Although PiB has been used in numerous in vitro and in vivo studies, experience with the fluorinated agents is more limited. The capacity of florbetapir and its sister compounds to label different fibrillar amyloid species in PD and DLB requires further quantification. Additional studies with these new ligands will be necessary in order to understand and interpret the imaging characteristics of these ligands in PD and DLB, as they make their way into clinical practice.

Lastly, the specific molecular cascades in which amyloid participates, and the relation of those cascades to the development of cognitive and motor symptoms in PD and DLB remain to be fully elucidated.

New Directions

Amyloid imaging is one of several imaging and nonimaging biomarkers that are contributing to our understanding of PDD and DLB. For example, fluorodeoxyglucose PET, PET and single photon emission computed tomography of the dopamine system and the cholinergic system, structural and functional MRI, and meta-iodobenzylguanidine (MIBG) imaging, which assesses sympathetic cardiac denervation, are providing important insights as well [109•, 110•]. Nonimaging biomarkers of the CSF and blood are also promising [109•, 110•]. Increasingly, applying these tools together is proving informative. For example, amyloid imaging and dopamine terminal (vesicular monoamine transporter) imaging have been acquired together to differentiate DLB from AD and FTD [86•] and from AD subjects, PD subjects, and HCS [78•]. This approach may be useful for distinguishing the causes of MCI as well [111].

By expanding imaging to multimodal PET, researchers are beginning to dissect molecular cascades associated with cognition in PD and DLB. For example, brain inflammatory changes that are characteristic of AD are increasingly amenable to PET measurement. Using the PET ligand [11C]PK11195 as a marker of microglial activation, along with PiB imaging, one study has found evidence for voxel-level correlations between microglial activation and amyloid burden in PD and PDD [77]. Nonspecific and variable binding of [11C]PK11195 has complicated its use, however, and next-generation studies with more sensitive ligands such as [11C]PBR28 are under way. These tools may help to determine the relative timing of amyloid deposition and microglial activation, and their relation to cognitive impairment in PD and DLB.

Tau aggregates are also becoming an important molecular imaging target. Paired helical filaments of tau (PHF-tau) progressively accumulate in the medial temporal lobe in AD and spread to neocortex as the disease progresses [112]. Recent reports have demonstrated propagation of tau aggregates in cell culture and animal models, with trans-synaptic spread [112], raising the possibility that tau propagation may contribute to progression of disease. Interestingly, PHF-tau deposition has also been reported in PDD [49, 55] and DLB [55, 57], with greater deposition and spread than PD, reflected in higher Braak NFT stage [55]. These results suggest that the molecular processes associated with PHF-tau deposition may also contribute to cognitive impairment in PD and DLB. It is also noteworthy that NFTs are the dominant abnormality associated with the parkinsonian tauopathies—progressive supranuclear palsy and corticobasal ganglionic degeneration [112]—and that the regional distributions of NFTs in progressive supranuclear palsy and corticobasal ganglionic degeneration are distinct from those in AD, PDD, and DLB.

In this context, PET ligands for PHF-tau molecular imaging are being developed and characterized [25, 113118]. Preliminary results for some ligands are quite promising, with greater affinity for tau than for Aβ in vitro and with a distribution of retention in human AD subjects that is reminiscent of Braak NFT staging [115118]. If successful, PHF-tau PET will have broad applicability for the study of the parkinsonian dementias. Such ligands will be useful not only to explore the relation of PHF-tau burden to cognition and clinical course in the Lewy body diseases but also to differentiate the synucleinopathies and the parkinsonian tauopathies.

A PET ligand for α-synuclein is not yet available. The future development of such a marker will have widespread application in understanding and diagnosing PD, PDD, and DLB.

Conclusion

Amyloid imaging provides a noninvasive means to assess brain amyloid burden antemortem. To date, PD and DLB studies using amyloid imaging are broadly consistent with studies of CSF Aβ-42 and neuropathology. Together, these results suggest that Aβ deposition contributes to cognitive impairment in both PD and DLB. More work is needed to confirm these early observations, however. Notably, cortical amyloid deposition is common in DLB and is not specific to AD. Therapeutic trials in AD can now leverage amyloid imaging both to identify and to stratify subjects and as a surrogate measure of therapeutic success. The development of anti-amyloid therapeutics, if successful, would have a large impact on research in PD and DLB, as well as immediate application in the treatment of these patients.