OUP user menu

18F-THK523: a novel in vivo tau imaging ligand for Alzheimer’s disease

Michelle T. Fodero-Tavoletti, Nobuyuki Okamura, Shozo Furumoto, Rachel S. Mulligan, Andrea R. Connor, Catriona A. McLean, Diana Cao, Angela Rigopoulos, Glenn A. Cartwright, Graeme O’Keefe, Sylvia Gong, Paul A. Adlard, Kevin J. Barnham, Christopher C. Rowe, Colin L. Masters, Yukitsuka Kudo, Roberto Cappai, Kazuhiko Yanai, Victor L. Villemagne
DOI: http://dx.doi.org/10.1093/brain/awr038 1089-1100 First published online: 24 March 2011


While considerable effort has focused on developing positron emission tomography β-amyloid imaging radiotracers for the early diagnosis of Alzheimer's disease, no radiotracer is available for the non-invasive quantification of tau. In this study, we detail the characterization of 18F-THK523 as a novel tau imaging radiotracer. In vitro binding studies demonstrated that 18F-THK523 binds with higher affinity to a greater number of binding sites on recombinant tau (K18Δ280K) compared with β-amyloid1–42 fibrils. Autoradiographic and histofluorescence analysis of human hippocampal serial sections with Alzheimer's disease exhibited positive THK523 binding that co-localized with immunoreactive tau pathology, but failed to highlight β-amyloid plaques. Micro-positron emission tomography analysis demonstrated significantly higher retention of 18F-THK523 (48%; P < 0.007) in tau transgenic mice brains compared with their wild-type littermates or APP/PS1 mice. The preclinical examination of THK523 has demonstrated its high affinity and selectivity for tau pathology both in vitro and in vivo, indicating that 18F-THK523 fulfils ligand criteria for human imaging trials.

  • tau
  • imaging
  • Alzheimer's disease
  • dementia
  • PET


The clinical diagnosis of neurodegenerative diseases such as Alzheimer's disease is typically based on progressive cognitive impairments while excluding other diseases. However, clinical diagnosis is often challenging, with patients presenting with mild and non-specific symptoms attributable to diverse and overlapping pathologies that present as similar phenotypes (van der Zee et al., 2008). Consequently, definitive diagnosis of neurodegenerative diseases is still reliant on post-mortem examination.

Post-mortem examination of the Alzheimer's disease brain is characterized by gross cortical atrophy (Wenk, 2003). Microscopically, Alzheimer's disease is characterized by the presence of extracellular β-amyloid plaques and intracellular neurofibrillary tangles (Wisniewski et al., 1989; Ho et al., 1994). There has been much progress in developing PET imaging radiotracers for the non-invasive detection of β-amyloid deposition (Shoghi-Jadid et al., 2002; Klunk et al., 2005; Rowe et al., 2007, 2008; Choi et al., 2009). Recent reports indicate that the best characterized and successful imaging agent Pittsburgh Compound-B (PiB), preferentially binds to fibrillar β-amyloid contained within cored and compact plaques (Klunk et al., 2004; Maeda et al., 2007; Ikonomovic et al., 2008) and with much lower affinity to the oligomeric forms of β-amyloid (Maezawa et al., 2008) that are thought to be the toxic species of β-amyloid in Alzheimer's disease (Lambert et al., 2001; Walsh et al., 2002; Ferreira et al., 2007; Cairns et al., 2009).

While amyloid imaging PET studies confirmed that β-amyloid deposition occurs well before the onset of symptoms (supporting the hypothesis that this represents preclinical Alzheimer's disease), these studies also showed the lack of correlation between β-amyloid plaque deposition and cognitive impairment in Alzheimer's disease; suggesting that markers for different and downstream effects of β-amyloid may be better suited to assess disease progression (Jack et al., 2010). Therefore, new ligands are needed to explore alternative biomarkers as specific indicators of neurodegeneration. Such agents may prove invaluable in the diagnosis, follow-up and therapeutic monitoring of Alzheimer's disease and other dementias.

An obvious biomarker is tau and in particular, abnormal deposits of hyperphosphorylated tau as neurofibrillary tangles, neuropil threads and as dystrophic neurites surrounding β-amyloid plaques (a pathological hallmark of Alzheimer's disease); however, tau deposits are also characteristic of a larger group of neurodegenerative diseases termed tauopathies [i.e. sporadic corticobasal degeneration, progressive supranuclear palsy, Picks disease, as well as frontotemporal dementia and parkinsonism linked to chromosome 17 (FTDP-17)] (Lee et al., 2001). Unlike β-amyloid plaque deposition, human post-mortem studies indicate that neurofibrillary tangle density correlates with neurodegeneration and cognitive impairment (Duyckaerts et al., 1987, 1990; Delaere et al., 1989; Arriagada et al., 1992; Dickson, 1997; McLean et al., 1999). Futhermore, abundant neurofibrillary tangles are not observed in cognitively unimpaired individuals, in contrast to β-amyloid plaques that are present in some non-demented people (Katzman et al., 1988; Delaere et al., 1990; Rowe et al., 2007, 2008). Moreover, CSF-tau and phospho-tau (ptau181) have been proven useful biomarkers in the diagnosis of Alzheimer's disease (Blennow and Hampel, 2003; Ganzer et al., 2003; Hampel et al., 2009a, b).

Despite the quantitative assessment of CSF levels of tau and phospho-tau being reliable biomarkers of neurodegeneration (Jack et al., 2010), lumbar puncture is an invasive procedure for the widespread screening of the at-risk population. Additionally, CSF measures do not provide information on regional brain tau deposition that may have clear correlates with cognition (i.e. hippocampus) and therefore, might not be able to provide important information on the therapeutic outcomes or response to current drugs aimed at modulating tau/neurofibrillary tangles (Gozes et al., 2009; Hampel et al., 2009a, b; Wischik and Staff, 2009).

Molecular neuroimaging with tau-specific radiotracers may provide highly accurate, reliable and reproducible quantitative statements of global and regional brain tau burden, essential for the evaluation of disease progression, therapeutic trial recruitment and the evaluation of tau-specific therapeutics (for both Alzheimer's and non-Alzheimer's disease taupathies); where tau plays a central role. Certainly, the viability of imaging disease-specific traits has been demonstrated in recent years by PET ligands such as 11C-PiB (Klunk et al., 2004) and 18F-FDDNP, used for imaging β-amyloid deposition. Unlike PiB, it has been suggested that FDDNP also binds to neurofibrillary tangles (Agdeppa et al., 2001), which may contribute to 18F-FDDNP retention in the mesial temporal cortex where β-amyloid-specific tracers such as 11C-PiB scarcely bind (Kepe et al., 2006; Ng et al., 2007; Pike et al., 2007; Rowe et al., 2007).

Okamura and colleagues (2005) screened over 2000 small molecules to develop novel radiotracers with high affinity and selectivity for tau pathology/neurofibrillary tangles. Consequently, they identified a series of novel quinoline and benzimidazole derivatives that bind neurofibrillary tangles and, to a lesser extent, β-amyloid plaques. Serial analysis of those compounds led to the design and synthesis of a novel imaging agent, 18F-THK523. The purpose of this study was to utilize a series of in vitro, ex vivo and in vivo techniques to determine whether 18F-THK523 satisfied a number of radioligand criteria, assessing its suitability for the quantitative imaging of tau pathology in the human brain.

Materials and methods


All reagents were purchased from Sigma, unless otherwise stated. Human β-amyloid1–42 was purchased from the W. M. Keck Laboratory (Yale University).


Mice were housed in conditions of controlled temperature (22 ± 2°C) and lighting (14:10 h light–dark cycle) with free access to food and water. rTg(TauP301L)4510 and their wild-type (CamKII) littermates were a kind gift from Jada Lewis (Dept Neuroscience, Mayo Clinic, Florida, USA) and APP/PS1 [B6C3-Tg(APPswe, PSEN1dE9)85Dbo/J] and wild-type littermates were purchased from JAX® Mice and Services. MicroPET studies employed 6-month-old rTg(TauP301L)4510 mice and 12-month-old APP/PS1 [B6C3-Tg(APPswe, PSEN1dE9)85Dbo/J] mice and their respective wild-type littermates.

Tissue collection and characterization

Human brain tissue was collected at autopsy. The sourcing and preparation of the human brain tissue was conducted by the Victorian Brain Bank Network. Alzheimer's disease pathological diagnosis was made according to standard NIA-Reagan Institute criteria (1997). Determination of age-matched control cases were subject to the above criteria. Three Alzheimer's disease and three healthy, age-matched control cases were examined in this study.

18F-labelling of THK523

Unlabelled THK523 and 2-(4-aminophenyl)-6-(2-tosyloxyethoxy)quinoline (BF241; the precursor for 18F-THK523) were custom synthesized by Tanabe R&D Service Co. and confirmed for purity by reverse phase high-performance liquid chromatography, 1D nuclear magnetic resonance and mass spectrometry. 18F-THK523 (Fig. 1) was synthesized by nucleophilic substitution of the tosylate precursor (BF-241). Following a 10-min reaction at 110°C, the crude reaction was partially purified on an activated Sep-Pak tC18 cartridge before undergoing semi-preparative reverse phase high-pressure liquid chromatography purification. Standard tC18 Sep-Pak reformulation produced 18F-THK523 in >95% radiochemical purity. The radiochemical yield was 24% (non-decay corrected) and at end of synthesis, the average specific activity was 100 GBq/µmol (2.7Ci/µmol).

Figure 1

Chemical structure of 18F-THK523 [2-(4-aminophenyl)-6-(2-fluoroethoxy)quinoline)].

Measurement of octanol/water partition coefficient

18F-THK523 (37 MBq) was added to a mixture of 3 ml 1-octanol and 3 ml of 1 M potassium phosphate buffer (pH 7.4). The mixture was shaken for 30 min, followed by centrifugation for 3 min. Aliquots (0.5 ml) were carefully taken from each phase for assay. The partition coefficient was calculated as follows: (count per minute/0.5 ml l-octanol)/(count per minute/0.5 ml buffer). Measurements were done in triplicate.

Generation and protein purification of K18Δ280 K-tau

K18Δ280 K-tau is a fragment of the full length protein, htau40 (Barghorn et al., 2004; von Bergen et al., 2006) comprising the four repeat regions of tau including residues 243–372. Polymerase chain reaction was implemented to generate K18Δ 280 K-tau from plasmids kindly provided by the Mayo Clinic. Δ280K refers to the deletion of the lysine residue at position 280. DNA encoding this region was cloned into expression vector pET15b at the NcoI and XhoI sites and transfected into BL21DE3 Escherichia coli. Ampicillin selected E. coli were lysed in buffer comprising 50 mM PIPES pH 6.9, 1 mM EDTA, 5 mM dithiothreitol and protease inhibitor cocktail (Roche), sonicated on ice (6 × 30 min, with 30 s rest intervals) and the lysate was then spun at 18 000g at 4°C for 15 min. The supernatant was removed and added to a solution of NaCl at a final concentration of 0.5 M. The sample was then boiled for 20 min prior to centrifugation using the abovementioned conditions. The supernatant was then applied to a PD10 column (Amersham) and equilibrated in equilibration buffer (50 mM Tris pH 8.2, 20 mM NaCl, 1 mM EDTA, 5 mM dithiothreitol) and filtered prior addition to a SP sepharose column. Protein fractions were then analysed by Coomassie staining and western blot and appropriate fractions containing a single tau band were pooled, buffer exchanged into water (PD10), lyophilized and stored at −80°C.

Preparation of β-amyloid1–42 and tau fibrils

Synthetic β-amyloid1–42 was dissolved in 1× phosphate buffered saline pH 7.7 to a final concentration of 200 µM. K18Δ280K-tau was dissolved in 1× phosphate buffered saline pH 7.4 buffer to a final concentration of 20 µM. The solutions were then incubated at 37°C for 2 and 3 days, respectively, with agitation at 220 and 800 rpm, respectively (Orbital mixer incubator, Ratek). β-amyloid1–42 fibril aggregation was confirmed via thioflavin T fluorescence spectroscopy and tau aggregation was confirmed by thioflavin S fluorescence spectroscopy; both fibril preparations were examined by transmission electron microscopy.

Thioflavin S/thioflavin T fluorescence

Aggregation of β-amyloid1–42 fibril was confirmed using thioflavin T fluorescence (LeVine, 1999). Reactions (100 µl) comprising 20 µM β-amyloid1–42 fibrils, 10 µM thioflavin T, 50 mM phosphate buffer were analysed at 444 nm (excitation) and 450–550 nm (emission), with an integration time of 1 s. K18Δ280K-tau fibril formation was confirmed by thioflavin S fluorescence whereby reactions comprising K18Δ280K-tau fibrils, 0.005% thioflavin S in 1× phosphate buffered saline pH 7.4 were analysed at 440 nm (excitation) and 480 nm (emission), with an integration time of 1 s. Measurements were recorded using a Varian fluorescence spectrophotometer.

Transmission electron microscopy

Fibril formation of β-amyloid1–42 and K18Δ280K-tau was further confirmed by transmission electron microscopy following staining with uranyl acetate. Carbon-coated copper electron microscopy grids were coated with K18Δ280K-tau or β-amyloid1–42 fibrils, as described previously (Smith and Radford, 2001). Grids were viewed on a Siemens 102 transmission electron microscope, operating at a voltage of 60 kV.

In vitro18F-THK523 binding assays

Synthetic β-amyloid1–42 or K18Δ280K-tau fibrils (200 nM) were incubated with increasing concentrations of 18F-THK523 (1–500 nM). To account for non-specific binding of 18F-THK523, the reactions described above were duplicated in the presence of unlabelled 1 µM THK523. The binding reactions were incubated for 1 h at room temperature in 200 µl of assay buffer [phosphate buffered saline, minus Mg2+ and Ca2+ (JRH Biosciences); 0.1% bovine serum albumin]. Separation of bound from free radioactivity was achieved by filtration under reduced pressure (MultiScreen HTS Vacuum Manifold; Multiscreen HTS 96-well filtration plates; 0.65 µm, Millipore). Filters were washed three times with 200 µl assay buffer and the radioactivity contained within the filters was counted in a γ-counter (Wallac 1480 Wizard 3”; Perkin Elmer). Binding data were analysed with curve fitting software that calculates the KD and Bmax using non-linear regression (GraphPad Prism Version 1.0, GraphPad Software). All experiments were conducted in triplicate.

Immunohistochemistry and fluorescence analysis

Brain tissue from Alzheimer's disease and healthy control cases [three Alzheimer's disease (two female, one male), age range 75–83 years; three healthy controls (three female), age range 72–85 years], as well as mice (rTg4510, APP/PS1 and wild-type littermates) was fixed in 10% formalin/phosphate buffered saline and embedded in paraffin. For immunohistochemistry, 5 µm serial sections were deparaffinized and treated with 80% formic acid for 5 min and endogenous peroxidase activity was blocked with 3% hydrogen peroxide. Sections were then treated with blocking buffer (20% foetal calf serum, 50 mM Tris–HCl, 175 mM NaCl pH 7.4) before incubation with primary antibodies to β-amyloid (1E8; 1:50) or tau pAb (DAKO), for 1 h at room temperature. Serial 5 µm tissue sections were stained as follows: the first and third sections were immunostained with tau or 1E8 antibodies to identify tau tangles or β-amyloid plaques, respectively. The second serial section was stained with unlabelled THK523 to assess whether THK523 staining co-localized with the immunodetected tau tangles and/or β-amyloid plaques. Visualization of antibody reactivity was achieved with the LSAB kit (labelled streptavidin-biotin, DAKO) and sections were then incubated with hydrogen peroxidase-diaminobenzidine (H2O2-DAB) to visualize the tau tangles or β-amyloid-positive deposits. Sections were counterstained with Mayer's haematoxylin. To detect THK523 fluorescence, quenching was first performed whereby sections were first deparaffinized and tissue autofluorescence minimized by treatment of sections with 0.25% KMnO4/phosphate buffered saline for 20 min prior to washing (phosphate buffered saline) and incubation with 1% potassium metabisulphite/1% oxalic acid/phosphate buffered saline for 5 min. Following autofluorescence quenching, sections were blocked in 2% bovine serum albumin/phosphate buffered saline pH 7.0 for 10 min and stained with 100 µM THK523 for 30 min. Washed (phosphate buffered saline) sections were then mounted in non-fluorescent mounting media (DAKO). Epifluorescence images were visualized on a Zeiss microscope [47CFP; filter set 47 (EM BP 436/20, BS FT 455, EM BP480/40)]. Co-localization of the THK523 and antibody signals was assessed by overlaying images from each of the stained serial tissue sections.


For autoradiography, the hippocampal brain section of a patient with Alzheimer's disease (90-year-old female) was incubated with 2.2 MBq/ml of 18F-THK523 at room temperature for 10 min and then washed briefly with water and 50% ethanol. After drying, the labelled section was exposed to a BAS-III imaging plate (Fuji Film) overnight. Autoradiographic images were obtained using a BAS-5000 phosphor imaging instrument (Fuji Film) with a spatial resolution of 25 × 25 μm. Neighbouring sections were immunostained using AT8 anti-tau monoclonal antibody (Innogenetics; diluted 1:20) or 6 F/3D anti-Aβ antibody (DAKO; diluted 1:50).

Ex vivo biodistribution of 18F-THK523

18F-THK523 (0.68–1.32 MBq) was administered into the tail vein of ICR mice (n = 20, male, average weight 28–32 g). The mice were then sacrificed by decapitation at 2, 10, 30, 60 and 120 min post injection. The brain, blood and other organs were removed and weighed, and the radioactivity was counted with an automatic γ-counter. The percentage injected dose per gram (%ID/g) was calculated by comparison of tissue count to tissue weight. Each %ID/g value is an average ± SD of four separate experiments.

Small animal positron emission tomography imaging

All PET scans were conducted using a Philips MOSAIC small animal PET scanner with a transaxial spatial resolution of 2.7 mm full-width at half-maximum. Mice [n = 8 rTg4510 (four females, four males), n = 7 wild-type (four females, three males) mice and n = 3 APP/PS1 (all females) and three of their wild-type littermates (all females)] were intravenously injected with 100 µl of radiotracer comprising 3.7 MBq (0.35 μg/kg) of 18F-THK523 via the tail vein. Mice were then anaesthetized using an isoflurane vaporizer with oxygen flow metre set to 5 l/min/5% isoflurane. Anaesthesia was maintained in a Veterinaire MINERVE anaesthetic assembly with the oxygen flow metre set to 2 l/min and vaporizer setting at 2%. A series of 6 × 5-min dynamic emission scans were acquired starting at 5 min after injection. All images were reconstructed using a 3D row action maximum likelihood algorithm (RAMLA). Summed 25–35 min post-injection images were used for comparison between transgenic and wild-type mice. Image analysis was conducted using Wasabi v.2.0 software.

Statistical analysis

Normality of distribution was tested using the Shapiro–Wilk test and visual inspection of variable histograms. Statistical evaluations to assess differences in 18F-THK523 binding were performed with analysis of variance (ANOVA) and a Tukey–Kramer Honestly Significant Difference test to establish differences between group means. Data are presented as mean ± SD unless otherwise stated.


18F-THK523 exhibits high affinity and selectivity for recombinant tau fibrils

To determine whether 18F-THK523 satisfied the criteria of high affinity and selectivity for tau, the binding properties of 18F-THK523 to tau fibrils was investigated and compared with β-amyloid1–42 fibrils. A previously described truncated mutant of human tau, termed K18Δ280K-tau (Barghorn et al., 2004; von Bergen et al., 2006) that comprises the C-terminus of tau, including the four repeat regions and the FTDP-17 tau gene deletion resulting in the omission of lysine at position 280 (denoted Δ280K) was used for the studies. K18Δ280K-tau aggregates at low micromolar concentrations into paired helical filaments and straight filaments in the presence and absence of heparin (Perez et al., 1996). Prior to conducting the binding assays, K18Δ280K-tau was formed into fibrillar structures (as monitored by thioflavin S fluorescence and transmission electron microscopy) by incubating 20 µM protein over 3 days at 37°C. On day 3, K18Δ280K-tau showed a thioflavin S fluorescence signal at ∼480 nm (Fig. 2A), indicative of positive fibril formation. Fibril formation was confirmed by transmission electron microscopy with uranyl acetate staining (Fig. 2B). The β-amyloid1–42 fibrils were generated as previously described (Fodero-Tavoletti et al., 2007).

Figure 2

Characterization of K18Δ280K-tau fibrils. Recombinant K18(Δ280K)-tau was incubated with agitation (800 rpm) for 3 days at 37°C. (A) Graph depicting thioflavin S fluorescence, excitation/emission 440/480 nm, for K18Δ280K-tau (solid line) and no tau buffer control (dotted line). The graph for K18Δ280K-tau is indicative of positive amyloid fibril formation. (B) Electron microscopy image of K18Δ280K-tau fibrils. TechaniG2 electron microscope; ×59 000 magnification. Scale bar: 50 nm. These data are representative of three independent experiments. a.u. = arbitrary units.

In vitro saturation studies were conducted using equimolar concentrations (200 nM, ∼4.0 × 10−11 moles) of either K18Δ280-tau or β-amyloid1–42 fibrils. While two classes of binding sites were identified on K18Δ280-tau fibrils (Fig. 3A) only one class of 18F-THK523 binding sites was identified on β-amyloid1–42 fibrils (Fig. 3B). Furthermore, there was a 10-fold higher affinity of 18F-THK523 for the first class of K18Δ280-tau binding sites compared with β-amyloid1–42 fibrils (Table 1). Overall, there was a ∼5-fold higher number of 18F-THK523 binding sites (Bmax) on K18Δ280-tau fibrils, compared with β-amyloid1–42 fibrils (Table 1).

Figure 3

In vitro binding studies indicate two classes of 18F-THK523-binding sites on K18Δ 280K-tau fibrils. Scatchard plots of 18F-THK523 binding to synthetic K18Δ280K-tau (A) or (B) β-amyloid1–42 fibrils. (A) Scatchard analysis identified two classes of THK523 binding sites on K18Δ280K-tau fibrils (KD1 and Bmax1 of 1.67 nM and 2.20 pmol THK523/nmol K18Δ280 K-tau, respectively; KD2 and Bmax2 of 21.7 nM and 4.46 pmol THK523/nmol K18Δ280 K-tau, respectively). (B) Scatchard analysis identified one class of THK523 binding sites on β-amyloid1–42 with KD and Bmax of 20.7 nM and 1.25 pmol THK523/nmol β-amyloid1–42. Binding data were analysed using GraphPad Software (Version 1.0). These data are the mean of three experiments for K18Δ280 K-tau and four experiments for β-amyloid fibrils.

View this table:
Table 1

Binding parameters of 18F-THK523 binding to fibrils

K18Δ280K-tau fibrils1.672.2021.744.46
β-amyloid142 fibrils20.71.25
  • KD are in nM and Bmax are in pmol 18F-THK523/nmol fibrils.

THK523 demonstrates selectivity for tau pathology in sections of human hippocampal tissue

As a qualitative measure of its selectivity for tau pathology, THK523 recognition of tau pathology was assessed by histofluorescence and autoradiography. 19F-THK523 and 18F-THK523 share the same chemical structure, although 19F is substituted for 18F in the radiolabelled compound. For histofluorescence, unlabelled THK523 binding to fixed serial sections from the hippocampus of subjects with Alzheimer's disease and age-matched controls was assessed. Contiguous sections were immunostained for β-amyloid and tau pathology with anti-β-amyloid and anti-tau antibodies, respectively. In all tissue sections examined, positive THK523 staining co-localized with tau pathology as detected in the contiguous tau immunostained section assessed (Fig. 4). THK523 failed to bind to diffuse β-amyloid plaques as indicated by the lack of co-localization with immunodetected β-amyloid pathology (Fig. 4). Likewise, autoradiography analysis in Alzheimer's disease hippocampal sections demonstrated that 18F-THK523 bound to tau pathology with no 18F-THK523 co-localization with immunodetected β-amyloid plaques (Fig. 5).

Figure 4

Histofluorescence analysis indicates that THK523 binds specifically to tau tangles with no detectable binding to β-amyloid plaques. Microscopy images of three serial sections (5 µm) from the hippocampus of a patient with Alzheimer's disease (AD) (top and middle) and a healthy control (HC) (bottom), immunostained with antibodies against tau (DAKO) and β-amyloid (1E8), to identify tau tangles and β-amyloid (Aβ) plaques, respectively; or stained with 100 µM THK523. Arrows indicate the location of tau tangles, while circles indicate the location of β-amyloid plaques. Positive THK523 staining appears to co-localize with tau immunostaining of neurofibrillary tangles in the hippocampus sections examined, but not to plaques. Tissue sections were imaged using a Zeiss microscope and Axiocam digital camera. Scale bars: 100 µm (top) and 200 µm (middle and bottom). These figures are representative of three subjects with Alzheimer's disease (two females, one male, age range 75–83 years) and three healthy controls (all female, age range 72–85 years).

Figure 5

Autoradiography analysis indicates that 18F-THK523 binds specifically to tau tangles with no detectable binding to β-amyloid plaques. (Top) 18F-THK523 autoradiogram of Alzheimer's disease hippocampus (90-year-old female) serial section (low magnification). Bottom: microscopy images and autoradiogram (higher magnification) images of three serial sections (5 µm) from the hippocampus of the same Alzheimer's disease brain, immunostained with antibodies to tau (AT8, Innogenetics) and β-amyloid (6F/3D, DAKO), to identify tau tangles and β-amyloid (Aβ) plaques, respectively; or labelled with 2.2 MBq/ml 18F-THK523. Positive 18FTHK523 labelling appears to co-localize with tau immunostaining of neurofibrillary tangles in the hippocampus sections examined, but not to plaques. Scale bars: 500 µm. Autoradiographic images were obtained using a BAS-5000 phosphor imaging instrument (Fuji Film).

18F-THK523 crosses the blood–brain barrier in mice

As well as being of low molecular weight (282.31 g/mol) and amenable to labelling with 18F at high specific radioactivity [100 GBq/µmol (2.7 Ci/µmol)], a tau radiotracer should be adequately lipophilic to be able to cross the blood–brain barrier. The octanol/water coefficient (logPoct) of 18F-THK523 as a measure of lipophilicity, was calculated to be 2.91 ± 0.13. Ex vivo biodistribution studies of 18F-THK523 in ICR mice, measured at 2, 10, 30, 60 and 120 min post injection, showed brain peak uptake of 2.75 ± 0.25% ID/g at 2 min post-intravenous injection (Fig. 6), indicating that 18F-THK523 has adequate lipophilicity to cross the blood–brain barrier.

Figure 6

Ex vivo biodistribution studies of 18F-THK523 in ICR mice. Initial uptake was highest (∼6.2%ID/g) in the heart and kidney followed by a fast clearance. Liver radioactivity peaked (∼5.0%ID/g) at 10 min after injection and was followed by a slow clearance, which mirrored a steady and substantial rise in radioactivity in the intestine (11%ID/g at 120 min after injection), suggesting that most of the tracer and/or its metabolites are eliminated through biliary excretion. There was a slow but steady increase in bone radioactivity, reaching a 2.1%ID/g at 120 min after injection, probably indicative of some degree of defluorination. The insert shows in better detail the brain and blood curves. Maximal 18F-THK523 brain uptake (2.75%ID/g) was observed at 2 min after injection of the radiotracer, followed by a rapid clearance from the brain. Radioactivity in blood showed a different kinetic behaviour than the one observed in the brain, with a steady rise in radioactivity reaching an apparent plateau at about 60 min after injection. Uptake at each time point is expressed as percentage of injected dose per body weight (%ID/g) of 18F-THK523. Curve represents the mean ± SD from four independent experiments. A total of 20 mice were examined.

In vivo retention of 18F-THK523 is significantly higher in tau transgenic mice brain compared with control and APP/PS1 mice

To further characterize 18F-THK523 as a tau imaging radiotracer, in vivo microPET studies were performed to compare the retention of 18F-THK523 in tau transgenic mice (rTg4510), versus their wild-type littermates (CamKII). Four independent studies were undertaken with 15 mice (n = 8 rTg4510 and n = 7 CamKII). Representative microPET images are depicted in Fig. 7A and 18F-THK523 time activity curves are depicted in Fig. 7C. Brain retention at ∼30 min post injection of 18F-THK523 was significantly higher (48%; P < 0.007) in the rTg4510 mice compared with their wild-type littermates (Fig. 7B). Analysis of bone, liver and intestine showed no significant differences in 18F-THK523 retention (Fig. 7B), indicating a specific difference in brain uptake. Following microPET scanning, each mouse was euthanized and brains were harvested for biochemical and histofluorescence analysis. All rTg4510 mice brains examined were positive for tau overexpression as determined by western blot and immunohistochemical analysis (data not shown). Histofluorescence analysis of the same rTg4510 mice assessed by microPET identified positive THK523 staining that co-localized with immunopositive tau deposits (Fig. 8).

Figure 7

In vivo 18F-THK523 microPET studies of tau and β-amyloid overexpressing transgenic mice. (A) Representative microPET scans at 30-min post injection of 18F-THK523. rTg4510 mice (top, left) exhibited higher 18F-THK523 brain retention compared with their wild-type (WT) littermate (bottom, left). Low 18F-THK523 retention was observed APP/PS1 (top, right) versus their wild-type littermates (bottom, right). (B) Analysis of the 18F-THK523 brain microPET data (30-min post injection) in rTg4510, APP/PS1 mice and their respective wild-type littermates revealed significantly higher (*) retention of 18F-THK523 in the brain (top) of rTg4510 mice compared with APP/PS1 mice as well as their respective wild-type littermates. No significant differences in 18F-THK523 retention were observed in the liver, intestine and bone (bottom). Data are presented as mean ± SD. (C) Brain time-activity curves of 18F-THK523 microPET data expressed as percentage of injected dose per body weight (%ID/g) of 18F-THK523 at each time point. Curve represents the mean ± SD of four independent studies employing n = 8 rTg4510 (four females, four males), n = 7 WT (four females, three males) mice and n = 3 APP/PS1 (all females) and three of the wild-type (all females) mice. Data are presented as mean ± SD.

Figure 8

Co-localization of THK523 staining with tau pathology. Microscopy images of two serial sections (5 µm) from brains of rTg4510 and APP/PS1 mice immunostained with either tau (DAKO) or β-amyloid (1E8) antibodies, to identify tau tangles and β-amyloid (Aβ) plaques respectively; or stained with 10 nM THK523. Arrows indicate the location of tau tangles while circles indicate the location of β-amyloid plaques. Positive THK523 staining co-localize with tau immunostaining of neurofibrillary tangles, but not with β-amyloid plaques. Tissue sections were imaged with a Zeiss microscope and Axiocam digital camera. Scale bars: 100 µm. These data are representative of four independent studies employing eight rTg4510 and three APP/PS1 mice.

To further characterize the in vivo selectivity of 18F-THK523 for tau pathology, microPET studies were conducted using the same experimental procedure in APP/PS1 transgenic mice (n = 3), exhibiting cerebral β-amyloid pathology but no tau deposits (Holcomb et al., 1999). MicroPET analysis demonstrated that there was significantly lower retention of 18F-THK523 in the brains of APP/PS1 mice, no different from the retention in their wild-type littermates (n = 3; Fig. 7B). Importantly, histofluorescence evaluation of rTg4510 and APP/PS1 brain tissue with 10 nM THK523 (a concentration that is achieved in the brain during PET studies), showed binding of THK523 to tau deposits in rTg4510 mice brains with negligible binding to β-amyloid plaques in the brain of APP/PS1 mice (Fig. 8).


With the recent advances in instrumentation, image analysis and the development of new brain radiotracers, molecular neuroimaging with PET is rapidly expanding our knowledge base of neurodegenerative disease progression, improving early and accurate diagnosis, while promising to be effective in therapeutic monitoring and aiding in drug discovery and development. To date, much success has been achieved with β-amyloid radiotracers, in particular PiB being the best characterized radiotracer both in vitro and in vivo; showing selectivity for β-amyloid pathology resulting in a robust difference in 11C-PiB brain retention in Alzheimer's disease compared with healthy aged-matched individuals in PET studies (Klunk et al., 2004, 2005; Fodero-Tavoletti et al., 2007, 2009). In addition to β-amyloid plaques, Alzheimer's disease brains are also pathologically characterized by the presence of tau pathology. Therefore, tau imaging may improve the specificity of diagnosis, allowing early detection of Alzheimer's disease and Pick's disease, where tau plays a role.

The identification and development of suitable PET radiotracer(s) is a demanding task especially given the considerable number of requirements that a radiotracer should fulfil to be deemed suitable for in vivo quantitative brain imaging. This study is the first to report a tau imaging radiotracer (18F-THK523), that satisfies a number of criteria required for quantitative imaging of tau pathology in the human brain (Laruelle et al., 2003; Nordberg, 2004; Pike, 2009). This study has shown that 18F-THK523 has high affinity for recombinant tau fibrils and selectivity for tau fibrils/pathology over β-amyloid fibrils/pathology in vitro. Furthermore, it penetrates the blood–brain barrier, selectively highlighting tau pathology in the brains of rTg4510 tau transgenic mice in vivo.

In vitro saturation binding studies demonstrated that 18F-THK523 binds to recombinant tau fibrils with high affinity in the low nanomole range. Typically ligands displaying affinities between 0.01–1.00 nM are deemed useful for in vivo quantitative PET studies. The high affinity 18F-THK523-binding site (KD1; 1.7 nM) exhibited >10-fold higher affinity compared with β-amyloid1–42 fibrils (20.7 nM). Moreover, the number of high affinity 18F-THK523-binding sites (KD1) was almost 2-fold higher than the number of sites on β-amyloid1–42 fibrils. In comparison to previous 3H-PiB studies (Klunk et al., 2005; Fodero-Tavoletti et al., 2007), the affinity of 3H-PiB for β-amyloid1–42 (KD1, 0.71–0.91 nM) is similar to the affinity of 18F-THK523 for tau fibrils (KD1, 1.7 nM). However, tau fibrils exhibit a larger number of 18F-THK523 binding sites (Bmax1, 2.20 pmol 18F-THK523/nmol K18Δ280K-tau), compared with what has previously been reported for 3H-PiB and β-amyloid1–42 (1.01 pmol PiB/nmol β-amyloid1–42) (Fodero-Tavoletti et al., 2007). As the concentration of imaging radiotracers typically achieved during PET studies is in the low nanomole range, these findings strongly suggest that 18F-THK523 will bind with high affinity and selectively to tau pathology under PET imaging conditions. Furthermore, as the brain area occupied by plaques is larger in comparison to neurofibrillary tangles, a >10-fold higher affinity and a larger number of 18F-THK523-binding sites on tau/neurofibrillary tangles over β-amyloid plaques may prove essential in ascertaining a high tau signal over background in human PET studies (Laruelle et al., 2003).

Further evidence of 18F-THK523 selectivity for tau pathology was demonstrated by autoradiography and histofluorescence with positive THK523 staining, co-localizing with tau pathology and not with β-amyloid plaques in human Alzheimer's disease hippocampal sections. Importantly, even at THK523 concentrations 10 000-fold higher than those typically achieved under PET studies, THK523 failed to bind to diffuse plaques in the histofluorescence studies. There was some inconsistent staining of cored/compact plaques, suggesting that there might be some 18F-THK523 binding to cored β-amyloid plaques, but only under non-PET radiotracer conditions. Similarly, variable staining of neurofibrillary tangles at high concentrations of PiB, has been reported by Ikonomovic and colleagues (2008).

In addition to high affinity and selectivity, a suitable tau radiotracer must be able to cross the blood–brain barrier to reach its target in vivo. The small size (molecular weight <450) (Laruelle et al., 2003) and lipophilic nature of 18F-THK523 [log POCT value of 2.9 ± 0.1; −logPOCT values in the range of 0.9 and 3.0, show optimal entry into the brain (Dishino et al., 1983)] indicates that 18F-THK523 is able to penetrate the blood–brain barrier. This was confirmed in both ex vivo biodistribution and in vivo microPET imaging studies. Additionally, microPET imaging demonstrated that 18F-THK523 retention was significantly higher (48%; P = 0.007) in the brains of rTg4510 tau transgenic mice compared with their control littermates, devoid of tau pathology; in agreement with the in vitro saturation and histofluorescence studies. Moreover, selectivity of THK523 for tau pathology was further supported by the 18F-THK523 microPET assessment of APP/PS1 mice. These mice possess substantial cerebral β-amyloid plaque load; however, the retention of 18F-THK523 in these mice was significantly lower than in rTg4510 tau transgenic mice and not different from the retention in CamKII mice or their own wild-type littermates; suggesting that THK523 does not significantly bind to β-amyloid plaques and is selective for tau pathology in vivo.

Analysis of 18F-THK523 biodistribution in the microPET studies showed no significant differences in 18F-THK523 retention in the liver, intestine or bone between rTg4510 tau transgenic and wild-type mice. 18F-THK523 retention in bone is indicative of some degree of defluorination (Van Dort et al., 1995). In vitro stability testing showed that 18F-THK523 was stable in vitro, suggesting that defluorination most likely occurs post-injection (data not shown). However, as the degree of free 18F-bone retention is similar in both transgenic and control mice, the free 18F does not contribute differentially to the retention of 18F-THK523 in the mouse brain. Similarly, as was observed in the ex vivo biodistribution studies, accumulation of radioactivity was observed within the intestine and liver of both rTg4510 and their control littermates indicating that most of the tracer and/or its metabolites were eliminated rapidly from the body through biliary excretion. Both tau transgenic and control littermates exhibited similar, low expression levels of tau in the liver (data not shown), further suggesting that 18F-THK523 liver retention was due to the metabolic processing of 18F-THK523 and not attributable to tau expression.

In conclusion, 18F-THK523 is a novel tau radiotracer that fulfils the major criteria necessary for an ‘ideal’ PET radiotracer (Laruelle et al., 2003; Nordberg, 2004). In addition to the abovementioned properties, THK523 was successfully labelled with 18F with high specific activity. The relatively longer half-life of 18F (110 min) precludes the need for an onsite cyclotron, allowing widespread distribution.

The clinical application of 18F-THK523 as a selective tau imaging biomarker will provide important information regarding tau pathophysiology in Alzheimer's disease and non-Alzheimer's disease tauopathies, allowing correlation of brain tau load with cognitive function, monitoring disease progression and evaluation of therapeutic efficacy of newly developed drugs; especially aimed at modulating tau pathology (Gozes et al., 2009; Hampel et al., 2009a, b; Wischik and Staff, 2009). This study provides an important and critical step in defining the role of 18F-THK523 as a tau specific PET radiotracer.


National Health and Medical Research Council of Australia (in part); Neurosciences Victoria and the Ministry of Health, Labour and Welfare, Japan (in part); Industrial Technology Research Grant Program in 2009 from New Energy and Industrial Technology Development Organization (NEDO) of Japan(in part); AAR Viertel Fellowship (to M.T.F.-T.); NHMRC Senior Research Fellowship (to R.C. and K.J.B.). Perpetual Trustees H & L Hecht Trust.


We thank Fairlie Hinton and Geoff Pavey from the Victorian Brain bank Network for sourcing and preparation of the human brain tissue.

 Pittsburgh Compound-B


View Abstract