Nonamyloid PET Biomarkers and Alzheimer's Disease: Current and Future Perspectives

Lucas Porcello Schilling; Antoine Leuzy; Eduardo Rigon Zimmer; Serge Gauthier; Pedro Rosa-Neto


Future Neurology. 2014;9(6):597-613. 

In This Article

Pet Biomarkers of Neurodegeneration

Imaging Tau Pathology

Misfolding and aggregation of hyperphosphorylated tau into NFTs is known to occupy a central mechanistic role in the pathogenesis of AD.[49,50] A growing body of evidence suggests that tau pathology is thought to spread to distant brain regions following a shift toward self-propagation in a prion-like manner, disrupting neuronal function and leading, ultimately, to neuronal loss and cognitive decline.[51] While increased cerebrospinal fluid (CSF) levels of tau have been shown to correlate with disease severity[52] – with the presence of tau in the CSF of AD patients thought to reflect neurodegeneration[53] – the use of CSF tau as a biomarker for AD remains controversial.[54] Moreover, in addition to the invasive nature of lumbar puncture, CSF measurements cannot provide topographic information, and are prone to variation across centers.[55] Noninvasive methods for quantification of tau pathology are therefore highly desirable, particularly given the increasing efforts directed towards the development of tau based therapeutics. In this context, several PET research groups have recently developed novel classes of compounds characterized by suitable kinetics and high affinity/selectivity for tau fibrils, including benzimidazole pyrimidine derivatives-[18F]T-807 and [18F]T-808-phenylquinoline derivatives-[18F]THK-523, [18F]THK-5105 and [18F]THK-5117 and the benzothiazole derivative [11C]PBB-3.

Following a screening of more than 900 compounds, several benzimidazole pyrimidine derivatives were identified as potential tau ligands, including [18F]T-807 and [18F]-T808. In vitro autoradiography using human AD brain sections showed that [18F]T-807 exhibits strong binding to NFTs, with a selectivity estimate of 29-fold for tau, relative to Aβ.[14] Moreover, comparison between double immunohistochemical staining of NFTs and Aβ on adjacent tissue sections and [18F]T-807 autoradiography showed the colocalization of [18F]T-807 binding with immunoreactive NFTs, but not with Aβ1-42 plaques.[56] In the case of [18F]T-808, in vitro autoradiographic-binding assays revealed high affinity and good selectivity for NFTs over Aβ,[15] with in vivo assessment in wild-type mice and rats using micro-PET showing fast brain uptake followed by a rapid washout, suggesting low nonspecific binding,[15] in line with in vivo finding obtained using [18F]T-807.[56] Finally, the first human brain images obtained using [18F]T-807 revealed an elevated standardized uptake value ratio to the cerebellum in AD – as compared with subjects with MCI and healthy controls across temporal, parietal and frontal cortices, as well as in the hippocampus/entorhinal area, with the differential patterns of tracer accumulation observed in keeping with Braak staging.[14]

Phenylquinoline derivatives exhibiting high affinity and selectivity for tau aggregates were pioneered by the Tohoku group, with [18F]THK-523[57] the first tau radiotracer candidate for PET. Initial in vitro binding assays showed [18F]THK-523 to possess binding affinity for tau fibrils, with follow up autoradiographic work using AD medial temporal brain sections showing accumulation of [18F]THK-523 in the hippocampal Sommer's sector, as well as in the pre- and pri-β layers of the entorhinal cortex. These findings were consistent with the density of paired helical filament (PHF)-tau deposition, as confirmed via immunohistochemistry.[58] Subsequent autoradiographic and histofluorescence studies showed that [18F]THK-523 binding to NFTs colocalized with tau immonoreactivity, with no detectable binding to Aβ plaques.[16] Similar findings were obtained in transgenic mice (rTg4510), with [18F]THK-523 found to colocalize with NFTs in tissue immunostained with tau antibodies, corroborating in vivo findings obtained using micro-PET. By contrast, no colocalization was noted in a model harboring human amyloid precursor protein and presenilin 1 mutations (APP/PS1), a model displaying amyloidosis in the absence of tau pathology, with micro-PET showing no difference in [18F]THK-523 between APP/PS1 mice and wild-type littermates.[16] More recent immunohistochemical and histofluorescence studies, however, suggest that [18F]THK-523 does not bind tau inclusions in non-AD tauopathies,[59] with preliminary clinical data casting doubt on [18F]THK-523′s future in both research and clinical settings owing to very high nonspecific WM binding.[60]

Following optimization to improve specificity, second generation quinoline derivatives were introduced, in the form of [18F]THK-5105 and [18F]THK-5117. In vitro binding assays conducted using [18F]THK-5105 and [18F]THK-5117 showed high binding affinity to synthetic truncated tau (K18ΔK280) fibrils – comprising the four repeat regions (244-372) in the absence of lysine 280 (ΔK280) – with both tracers proving superior to [18F]THK-523. Further examination of the selective binding capacity of these compounds-performed using in vitro autoradiography and AD mesial temporal brain sections-showed elevated tracer accumulation within the parahippocampus and subiculum, with particularly high binding in the Sommer's sector of the hippocampus. These findings, confirmed with Gallyas–Braak staining and immunohistochemistry, were reduced among healthy controls.[17] Further assessment of [18F]THK-5105, conducted using AD hemibrain sections and [11C]PIB, showed dense accumulation of [18F]THK-5105 in tau rich areas – including the insula, inferior and middle temporal gyri, cingulate gyrus and hippocampus/parahippocampus-with the pattern of tracer retention corresponding to the known distribution of tau pathology but not to that of Aβ or areas showing elevated retention of [11C]PIB. Furthermore, biodistribution studies conducted in normal mice showed abundant and rapid brain uptake and fast clearance, with the kinetics of both tracers superior to those reported for [18F]THK-523.[17]

Following a screening of several fluorescent chemicals capable of binding to β-sheet conformations, a phenyl/pyridinyl-butadienyl-benzothiazoles/benzothiazoliums (PBBs) class of tau ligands was developed for visualization and quantitative assessment of diverse structural forms of phosphorylated tau. Two-photon microscopic data using the most promising of these probes, [11C]PBref-3-a pyridinated PBB radiolabeled with carbon-11-provided strong evidence that [11C]PBref-3 rapidly transits both the blood–brain barrier and neuronal plasma membranes, binding to intraneuronal tau inclusions. Moreover, accumulation of PBref-3 in AT8-positive, NFT-like lesions in Tg mice expressing a single human four-repeat tau isoform with the P301S FTDP-17 mutation (PS19 line) was confirmed using ex vivo microscopy. Additionally, in vitro and ex vitro autoradiographic studies showed that [11C]PBref-3 produced selective, high-contrast labeling of neuronal tau inclusions in the brain stem of PS19 mice. Similar findings were obtained using in vitro autoradiography and AD tissue, with marked radiolabeling of fibrillar aggregates in the frontal cortex, as well as the Sommer's sector and subiculum of the hippocampus. Finally, an exploratory clinical PET study using [11C]PBref-3 in patients with probable AD revealed elevated tracer retention in lateral temporal and frontal cortices – consistent with distribution of tau pathology at Braak stage V/VI – with increasing standardized uptake value ratios correlating with decreased scores on the mini mental state examination (MMSE). In addition, a slight increase in [11C]PBref-3 retention was noted around the hippocampus of a control subject who showed a decline in MMSE, consistent with Braak stage III/IV or earlier.[18]

Imaging Brain Glucose Metabolism

[18F]FDG PET has long been used to investigate the neurodegenerative aspects of AD, with cerebral glucose metabolism taken as a proxy for neuronal activity,[61] and as a marker of synaptic density.[62] In AD, metabolic deficits are found to follow a specific regional pattern, with declines in glucose metabolism noted in parietotemporal areas, precuneus,[19] posterior cingulate cortex[63] and the medial temporal lobe (see Figure 3).[64] These hypometabolic areas are noted to extend toward frontal association areas with progression of the disease, with relative preservation of the visual cortex, primary sensory motor cortices, basal ganglia, thalamus and cerebellum.[19,64,65] While this in vivo pattern of hypometabolism is found in the vast majority of clinically diagnosed AD patients and in over 85% pathologically confirmed cases,[19] the extent and topography of hypometabolism has been found to vary across atypical focal cortical presentations of AD. Specifically, relative to typical AD, patients with posterior cortical atrophy have been shown to exhibit selective hypometabolism in occipito-parietal regions – as well as in the frontal eye fields[66] – with logopenic primary progressive aphasia associated with disproportionate left temporoparietal hypometabolism.[67] Greater hypometabolism has likewise been noted in patients with early onset AD – with metabolic reductions among those with mild dementia comparable to that seen in late onset cases with severe dementia[68] – consistent with studies showing more rapid progression among patients with early onset AD[69] and, potentially, with cognitive reserve theory.[70]

Figure 3.

Representative [18F]FDG images in a cognitively normal subject and in a patient with Alzheimer's disease. Metabolic imaging with [18F]FDG PET in AD shows the characteristic pattern of reduced tracer emission, comprising bilateral parietal, temporal and posterior cingulate regions. In contrast, metabolism within these regions is preserved in the cognitively normal subject.
AD: Alzheimer's disease; CN: Cognitively normal.

Among patients with MCI, metabolic abnormalities are usually noted in brain areas typically affected in AD,[71] albeit of inferior magnitude,[72–74] with the anterior hippocampal formation of particular value for differentiating subjects with MCI from healthy controls.[73] In the context of conversion studies, MCI subjects presenting a more pronounced or 'AD-like' pattern have been found to decline to AD at higher rates,[75,76] with reductions in glucose metabolism shown to predict future AD with accuracies in the range of 75–100%.[77,78] In addition, [18F]FDG has proven of use in differentiating progressive from nonprogressive MCI, particularly when combined with memory scores,[79] with progressors showing the typical AD functional pattern comprising hypometabolism in the parietal and posterior cingulate cortices, and more severe memory impairment.[79,80] Among nonprogressors, memory impairment was found to be less severe, with hypometabolism confined to the dorsolateral frontal cortex, consisted with this regions role in episodic memory processes such as encoding and retrieval.[81,82]

[18F]FDG-PET has likewise been used to study the progression to MCI and AD among cognitively normal older individuals, and has been shown to predict cognitive decline within this population with an accuracy approaching 80%.[83,84] In a clinicopathological study incorporating longitudinal [18F]FDG among cognitively normal individuals followed through MCI to pathologically confirmed AD, progressive reductions in glucose metabolism were noted years in advance of clinical symptoms, with reductions in the hippocampus preceding declines in cortical regions.[85] Similar functional changes have been observed in cognitively normal individuals homozygous for the APOE ε4 allele[79,86] – a susceptibility gene-asymptomatic carriers of mutations causative for early onset familial AD,[77,78,87] and among those with a maternal family history of AD, as compared with those with a paternal history or no family history of AD.[88] Finally, [18F]FDG may prove of use in the characterization of a subset of patients exhibiting neurodegeneration in the absence of Aβ deposition.[20,89] Placed in the category of 'suspected nonamyloid pathophysiology' (SNAP), these patients have led to the hypothesis that the onset of neurodegeneration in AD may not depend of the accumulation of Aβ.[90]