Visualization of regional tau deposits using 3H-THK5117 in Alzheimer brain tissue

Introduction The accumulation of neurofibrillary tangles, composed of aggregated hyperphosphorylated tau protein, starts spreading early in specific regions in the course of Alzheimer’s disease (AD), correlating with the progression of memory dysfunction. The non-invasive imaging of tau could therefore facilitate the early diagnosis of AD, differentiate it from other dementing disorders and allow evaluation of tau immunization therapy outcomes. In this study we characterized the in vitro binding properties of THK5117, a tentative radiotracer for positron emission tomography (PET) imaging of tau brain deposits. Results Saturation and competition binding studies of 3H-THK5117 in post-mortem AD brain tissue showed the presence of multiple binding sites. THK5117 binding was significantly higher in hippocampal (p < 0.001) and temporal (p < 0.01) tissue homogenates in AD compared to controls. Autoradiography studies with 3H-THK5117 was performed on large frozen brain sections from three AD cases who had been followed clinically and earlier undergone in vivo18F-FDG PET investigations. The three AD cases showed distinct differences in regional THK5117 binding that were also observed in tau immunohistopathology as well as in clinical presentation. A negative correlation between in vivo18F-FDG PET and in vitro3H-THK5117 autoradiography was observed in two of the three AD cases. Conclusions This study supports that new tau PET tracers will provide further understanding on the role of tau pathology in the diversity of the clinical presentation in AD. Electronic supplementary material The online version of this article (doi:10.1186/s40478-015-0220-4) contains supplementary material, which is available to authorized users.


Introduction
Alzheimer's disease (AD) is the most common form of dementia among the elderly. The two pathological hallmarks of AD are fibrillar amyloid beta (Aβ) deposition in the brain in the form of senile plaques (SPs), and accumulation of hyperphosphorylated tau protein in the form of neurofibrillary tangles (NFTs). Although the relationship between SPs and NFTs remains unclear, both appear to precede the onset of clinical symptoms of AD [1]. Several biomarkers have been developed to target the pathological features in vivo in order to help the clinical diagnosis of AD, showing increased reliability with time. Quantification of Aβ and tau from cerebrospinal fluid (CSF) appears consistent with quantification of the lesions [2,3]. However this invasive method is indirect, as the relationship between CSF biomarkers and AD lesions is still unclear [4]. Positron emission tomography (PET) imaging has been developed over recent decades and is now a very important tool for diagnosing AD, with particular regard to in vivo assessment of the cerebral amyloid load. However, the pathophysiological link between formation of the amyloid plaques and neurodegeneration is still unknown, and amyloid-targeting drug therapies have so far failed to show conclusive results.
In order to visualize and increase our understanding of the time course of NFT deposition during the development of AD and other dementia disorders, there is currently an attempt to develop PET tracers that are specific for tau protein [5]. Postmortem studies on patients with confirmed AD have shown a correlation between NFTs (but not SPs) and cognitive impairment [6][7][8]. Since tau aggregates are mainly intracellular, the radiotracer has to cross the blood-brain barrier as well as the cell membrane. Moreover, the tau ligand needs to be very specific for NFTs; in some regions of AD affected brains, amyloid plaques and NFTs occur together, although the occurrence of amyloid plaques is 5-20 times greater than that of NFTs [9]. The PET tracer 18 F-FDDNP, initially introduced as an Aβ PET tracer, also binds to NFTs [10]. A negative correlation between 18 F-FDDNP binding and episodic memory has been reported in patients with AD and patients with mild cognitive impairment (MCI) [11], but this tracer showed limitations with low levels of specific binding [12]. Recently, several PET tau tracer candidates from three different chemical classes have been developed, including the THK analogues 18 F-THK523, 18 F-THK5105, and 18 F-THK5117, as well as the ligands 11 C-PBB3, 18 F-T808 and 18 F-T807 [5,13]. Clinical studies with 18 F-THK5117, 11 C-PBB3, and 18 F-T807 are currently in progress [14][15][16][17].
In this study, we characterized the in vitro binding properties of THK5117 in large frozen sections from autopsied AD brains, and compared its binding pattern with measured tau histopathology as well as with previous in vivo measurement of 18 F-FDG and 11 C-Pittsburgh compound B (PIB) PET imaging in the same patients.
In this study, we focused on the 18 F-THK5117 tracer that was reported to have good binding properties in AD tissue [17]. Here we aimed at investigating in greater details the THK5117 binding using large frozen sections from autopsied AD brains, and compared its binding pattern with measured tau histopathology as well as with previous in vivo measurement of 18 F-FDG PET imaging in the same patients. In vitro binding assays Human brain tissues from eight Alzheimer patients and eight controls were obtained from the Netherlands Brain Bank (demographic details in Table 1), and homogenized in PBS containing 0.1 % BSA and protease/phosphatase inhibitor.

Chemicals
Optimal conditions for the binding assay study were determined in preliminary experiments by varying the time of incubation (1 h, 2 h or 3 h), the tissue concentration (100 μg/mL, 200 μg/mL, 500 μg/mL or 750 μg/ mL) and the buffer (PBS, PBS + 0.1%BSA or Tris buffer).
Saturation binding assays were carried out with incubation of temporal cortex and hippocampus homogenates (0.1 mg tissue) with 3 H-THK5117 (0.04-270 nM) for 2 h at room temperature. Non-specific binding was determined with 1 μM THK5117. The binding reaction was terminated by filtering through glass fiber filters (pre-soaked for 3 h in 0.3 % polyethylenimine), and rinsing three times with cold binding buffer, and counted with a beta scintillation counter (Beckman Coulter LS6500). The results were analyzed with GraphPad Prism software (version 5.0f for Mac OS X, GraphPad Software, San Diego California USA). The equilibrium dissociation constant (Kd) and maximum number of binding sites (Bmax) were then determined. The specificity and selectivity of 3 H-THK5117 binding was determined by incubating homogenates (0.1 mg tissue) from Alzheimer's patients' temporal cortex and hippocampus with 3 H-THK5117 (3 nM), in the presence of THK5117 (10 −5 -10 −13 M), FDDNP (10 −6 -10 −14 M) or BTA-1 (10 −5 -10 −13 M). The results were analyzed with GraphPad Prism software, and the resulting inhibitory constants (Ki) were determined.
Regional binding studies with 3 H-THK5117 were carried out on the brain homogenates from temporal cortex, frontal lobe, parietal lobe, caudate nucleus, hippocampus and cerebellum that were incubated with a single concentration of 3 H-THK5117 (3 nM). Non-specific binding was determined with 1 μM of THK5117. Comparisons between AD and control tissues were made using Mann-Whitney non-parametrical tests, with a threshold for significance of p < 0.05.
Autoradiography binding studies in Alzheimer's disease and control autopsy brain tissue Postmortem brain tissue for autoradiography studies was obtained from three patients with AD that all three had been clinically followed (by A.N.) at the Department of Geriatric Medicine, Karolinska University Hospital, Huddinge, Sweden. The left hemisphere of each brain was frozen in large block sections. The right hemispheres were fixed in formaldehyde, brain tissue was sampled according to the BrainNet Europe protocol and histopathological investigations were undertaken (by I.N.) to confirm AD pathology (see Additional file 1).
AD case one was referred for memory assessment at 49 years of age due to complaint of progressive memory impairment over 2 years. The patient was diagnosed with AD, and received galantamine (cholinesterase inhibitor) and memantine (NMDA antagonist) treatment. The patient was followed clinically and underwent two 18 F-FDG PET at MMSE 18/30 and 12/30, respectively. After 4 years in a nursing home, the patient died at the age of 60, and autopsy (post-mortem delay of 25 h) confirmed AD pathology, with Braak stages V-VI. General atrophy of the gyri was observed, especially the frontal lobe, and severe atrophy of the amygdala and the anterior part of the hippocampus.
AD case two was referred at the age of 71 with a 7year history of memory complaint. Neuropsychological testing showed impairment in several cognitive domains with a MMSE score of 26/30, and CSF biomarkers were pathological. The patient was diagnosed with AD, treated with phenserine and rivastigmine (cholinesterase inhibitors) and underwent 18 F-FDG PET. The patient died aged 79 years, after being in a nursing home for 2 years, and autopsy (post-mortem delay of 16 h) confirmed AD with at least Braak stage V. The brain presented with generally slightly smaller gyri than normal, and atrophic anterior part of the hippocampus.
AD case three was referred at 70 years of age with a 7year history of memory complaint. Neuropsychological testing showed disturbances in episodic memory, spatial orientation, attention, and MMSE was at 26/30. CSF biomarkers were pathological, and MRI showed frontal atrophy. The patient was diagnosed with AD, treated with rivastigmine (anticholinesterase inhibitor) and memantine (NMDA antagonist, and underwent three 18 F-FDG PET examinations at 70, 72 and 75 years of age. The patient died aged 81 years after four years in a nursing home, and autopsy (post-mortem delay of 17 h) confirmed Braak stages V-VI with slight atrophy of the frontal lobe. Frozen brain sections (80 μM) from the left hemispheres of the three AD cases and the healthy control case were cut with a cryostat at −20°C and thaw-mounted on chrome alum gelatin-subbed slides following a previously described protocol [18].
For each case, frozen sections were dried, preincubated for 10 min with PBS + 0.1%BSA, and were then incubated for one hour with 4 nM 3 H-THK5117. The non-specific binding assay was performed on adjacent sections in the presence of 10 μM THK5117. After incubation, sections were washed 3 times for 5 min with cold incubation buffer, dipped into cold distilled water, and allowed to dry for at least 24 h. A Fujifilm BAS-2500 phosphor imaging plate was then exposed to the section, along with a tritium standard, for three days (Larodan Fine Chemical AB, Malmö, Sweden) before being scanned by a phosphoimager Fujifilm BAS-2500.
Binding of the tracer was quantitatively assessed using MultiGauge Fujifilm software (V3.0). Two independent raters manually segmented different regions of interest, as defined by Duvernoy in "The Human Hippocampus" [19], resulting in regional mean uptake values for each segment. Inter-rater agreement was assessed for all uptake values using a Bland-Altman plot, with limits of agreement set to 95 %. When agreement was satisfactory, regional uptake values from both raters were averaged for subsequent analyses.
Autoradiography with enantiomerically pure (S)-18 F-THK5117 was also performed on adjacent sections for each case. The sections were pre-incubated for 10 min with PBS + 0.1%BSA and incubated for 15 min with 0.5 MBq/mL (S)-18 F-THK5117. Non-specific binding was determined in adjacent sections by adding 10 μM THK5117. After incubation, sections were washed three times 5 min with cold buffer and then dipped into 50 % v/v cold distilled water and ethanol before drying in an incubator at 37°C. A phosphor imaging plate was exposed to the sections overnight, before scanning with a phosphoimager (Cyclone plus, Perkin Elmer).

Autoradiography and immunostaining on small paraffin AD brain sections
Routine pathology with AT8 has been performed on anterior and posterior hippocampus paraffin sections from the right hemisphere of the three AD cases. Autoradiographies with 3 H-THK5117 on adjacent sections have been performed after deparafinization following the same protocol as cited above.
A paraffin section from hippocampus of a patient with confirmed AD (70 years old at death, Braak VI, from the Netherlands brain bank) were quenched and stained with THK5117 following the procedure used by Okamura and colleagues [17,20] for autofluorescence studies. An adjacent section was labeled with AT8 anti-tau and assessed using the EnVision + System HRP (DAB) (Dako Denmark) detection method.

In vivo imaging
Image Acquisition: PET Data: All 18 F-FDG PET investigations were carried out at the Uppsala PET centre using ECAT EXACT HR+ scanners (Siemens/CTI). The mean injected dose was 229 ± 49 MBq. Acquisition was performed during 45 min after injection. All reconstructed frames were realigned to correct for patient motion during each PET scan. The sum image from 25 to 45 min was used for analyses. MRI Data: Case two and case three underwent a structural T1 MRI sequence using a 3 T Siemens Trio scanner. Images were acquired with a slice thickness of 1.5 mm and were reconstructed to 1.0 x 1.0 x 1.0 mm isometric voxels.
Regional quantification of 18 F-FDG PET images for correlation with in vitro autoradiography: In order to compare the in vitro quantitative autoradiography binding measurements with the in vivo 18 F-FDG PET images, the coronal section of the T1 MRI judged to correspond best to the relevant autoradiography section was manually segmented by two independent raters. The atlas used for segmentation of autoradiographies (see above [19]) was used to define regions of interest on the T1 section. Mean uptake values were calculated for each of these regions on the registered 18 F-FDG PET images.
Inter-rater agreement was assessed for all resulting regional uptake values using a Bland-Altman plot, with limits of agreement set to 95 %. When agreement was satisfactory, regional uptake values from both raters were averaged for subsequent analyses.
The relationship between 18 F-FDG regional uptake values measured by PET and the in vitro binding results from autoradiography was assessed using a Spearman correlation test.

THK5117 multiple binding sites
Saturation binding assay performed with increasing concentrations of 3 H-THK5117 (0.04-270 nM) in post-mortem AD temporal brain tissue homogenates showed saturation with a Bmax of 1416 fmol/mg and Kd 2 of 24nM (Fig. 1a). A second binding site (Bmax 250fmol/mg and Kd 1 of 2.2 nM) was determined manually from the Scatchard plot (Fig. 1b). Saturation binding data from the temporal cortex were averaged from all seven experiments.
Saturation binding in the hippocampus (one Alzheimer brain) revealed two binding sites, with Kd values of 3.1 nM and 34 nM and Bmax values of 250 fmol/mg and 1226 fmol/mg, respectively (Additional file 2).
Competition studies with increasing concentration of unlabeled THK5117 (10 −5 -10 −13 M) and 3 H-THK5117 (3 nM) in AD temporal cortices (Fig. 1c) showed two binding sites (Ki 2 = 16 nM, Ki 3 = 750 nM). Results were averaged from ten competition experiments. Setting the constraints in the statistical model to >100 for the start of the curve indicated the presence of a potential third site (Ki 1 = 1 pM).

Selectivity of THK5117
Increasing concentrations of FDDNP showed a trend for competition with 3 H-THK5117 for a super-high affinity binding site (picomolar range) and a low affinity site (micromolar range; Additional file 3a). No competition was observed neither between 3 H-THK5117 and BTA-1 nor between 3 H-PIB and unlabeled THK5117 (Additional file 3b).

Regional binding distribution of 3 H-THK5117 in tissue homogenates
The regional binding distribution of 3 H-THK5117 in the AD cases was highest in the hippocampus (198 ± 67 fmol/mg), followed by the caudate nucleus (198 ± 30 fmol/mg), the temporal lobe (110 ± 47 fmol/mg) and the parietal lobe (105 ± 46 fmol/mg), and was very low in the cerebellum (Fig. 2). Statistical group comparisons revealed a significantly higher binding in the AD than in the controls in the hippocampus (p < 0.001) and the temporal region (p = 0.005) (Fig. 2).
Binding in caudate nucleus was found higher in the three patients with early onset AD (EOAD<65 years old) cases than in the three with late onset AD (LOAD>65 years old).

Regional distribution of 3 H-THK5117 evaluated by autoradiography
The total and non-specific binding autoradiograms for AD case one and the control case are shown in Fig. 3, while autoradiograms for AD case two and three are shown in Additional file 4. From those autoradiograms, fourteen regions of interest could be defined for case one, and twelve for case two and three. Quantitative results are presented in Table 2 for each case.
Inter-rater agreement for manual segmentation was within the 95 % limit for agreement.
Specific 3 H-THK5117 binding in case one was intense in the hippocampus, the temporal cortex and the frontal cortex. 3 H-THK5117 binding in AD case two and three was less intense than in AD case one; specific binding was high in the hippocampus and the cingulate for AD case two, and high in the hippocampus and the frontal cortex for AD case three. Furthermore, the binding intensities differed within a single brain region, as illustrated by the values in temporal gyri for case one. Low and regional binding was found in the control case compared to the three AD cases (Fig. 3).
As shown in Table 2, there were large differences among the three patients with respect to 3 H-THK5117 binding in the regions studied. Regional 3 H-THK5117 binding was highest in case one and lowest in case three.
The differences between the three cases were greatest in the entorhinal cortex, inferior temporal cortex, middle temporal cortex and fusiform gyrus. In all three cases, hippocampus was one of the regions with highest 3 H-THK5117 binding.
Autoradiography results for (S)-18 F-THK5117 indicated a similar binding pattern to that for 3 H-THK5117 (Additional file 5).

Validation of the selectivity of THK5117 towards TAU
Autoradiography with 3 H-THK5117 and immunostaining with AT8 performed on adjacent sections of anterior and posterior part of hippocampus of AD case one showed identical binding pattern (Fig. 4).
Autofluorescence performed on the hippocampus section from another AD case revealed THK5117 binding on the neurofibrillary tangles confirmed by tau immunostaining with the AT8 tau antibody on the adjacent section (Fig. 5).

Correlation of in vivo 18 F-FDG PET with in vitro 3 H-THK5117 autoradiography binding results
Regional in vivo 18 F-FDG PET data were available for all three AD cases. A hypometabolism pattern characteristic of AD was observed in all cases. The inter-rater agreement for manual segmentation was within the 95 % limit of agreement for all defined regions of interest in all patients. For the three cases, hippocampus showed the lowest 18 F-FDG PET uptake and one of the highest 3 H-THK5117 binding results (see Additional file 6 for 18 F-FDG uptake values).
A significant negative correlation was found in AD case one and two between in vivo uptake of 18 F-FDG and in vitro 3 H-THK5117 autoradiography binding while no significant correlation was observed for AD case three (Fig. 6).

Discussion
In this study we explored the binding properties and regional distribution of the novel PET tau tracer THK5117 in autopsy brain tissue homogenates and autoradiography on whole frozen brain hemisphere. 3 H-THK5117 showed a high affinity for tau, as observed in competition binding studies where 3 H-THK5117 demonstrated up to three binding sites. In saturation binding assays, we could observe two binding sites, one being in the nanomolar range, in agreement with Okamura et al. [17].
Binding assays performed in brain homogenates should reflect the actual binding situation more accurately and are supposed to be more reliable than binding assays performed on tau fibrils, but they also reflect pathological variability between subjects, possibly leading to more heterogeneity in the results.
Since the selectivity of a PET tracer for its target is of crucial importance, THK5117 binding was assessed in competition studies with other compounds. No competition was observed with the amyloid tracers 3 H-PIB and BTA-1. More interestingly, a trend for competition was observed between 3 H-THK5117 and FDDNP at the superhigh affinity binding site. FDDNP was initially introduced as Aß PET tracer [12,21] but has then been reported to also bind to NFTs [22,23], and to correlate with cognitive decline [11]. As suggested previously [24], the THK compounds family (such as THK523 and now THK5117) could share some binding properties with FDDNP; all of them are of interest for targeting tau pathology.
The 3 H-THK5117 binding in AD tissue homogenates was highest in both the hippocampus and temporal cortex which are regions that are expected to be more severely affected earlier by tau pathology in AD [6,25]. Consistent findings were also observed in brain hemisphere sections used in autoradiography, with the highest binding in the temporal regions for the three AD cases. Interestingly, those three cases, all with different clinical presentations, differed markedly in regional 3 H-THK5117 binding intensity.
In AD case one, presenting with an early onset of AD, 3 H-THK5117 binding in the hippocampus and the entorhinal cortex was about six times higher than in the other two AD cases. This difference could also been observed on AT8 staining for tau, in particular in hippocampus. Indeed case one's brain was severely affected by the disease, in particular in those temporal regions, and was very shrunken, due to neuronal loss. This atrophy could partly explain the higher extent of binding compared to case two and three, in whom the superficial layers were more compact and dense. One could also assume that the high level of 3 H-THK5117 binding could be related to the early onset of the disease. It has been suggested that EOAD and LOAD might have different mechanisms for disease progression [26]. Some of the AD patients included in the binding assays also had EOAD. However, there were no differences between EOAD and LOAD in 3 H-THK5117 binding except in the caudate nucleus homogenates, where it was higher in EOAD tissue than in LOAD tissue. Although tau pathology has been reported in the striatum of both EOAD and LOAD brains [27], no comparisons between the two clinical entities in this respect have been reported to our knowledge. Interestingly, control tissue also showed binding in caudate, that appeared similar to the one observed in LOAD tissue. In recent in vivo Tau PET imaging studies, high signal has been reported in striatum of both controls and AD patients. Although it has been suggested that this binding would be nonspecific, no evidence has been provided yet. Both our binding assays and autoradiography in AD case one showed to some extent specific binding in striatum of both control and AD cases. Our limited sample prevents Fig. 6 Correlations between in vitro 3 H-THK5117 autoradiography binding and in vivo 18 F-FDG uptake in case one (a), case two (b), and case three (c). All 18 F-FDG uptake values are reported as standard uptake value ratios (SUVR) with reference to the grey matter of the cerebellum. Temporal regions are shown in blue, frontal regions in orange, the rest in grey. r and p values from the Spearman test are reported for significant correlations (significance threshold: p < 0.05). Regions were defined by manual segmentation. hippo = hippocampus; inf = inferior; mid = middle; sup = superior us from drawing any conclusion, and specific investigation addressing this issue should be performed.
Compared to case one and three, case two showed an in-between binding of 3 H-THK5117 in hippocampus and frontal cortex, which is in agreement with the numerous tau lesions reported on pathological examination. 3 H-THK5117 binding was lowest in case three, who had only few tangles in the frontal cortex despite a lot of tau in the granular cell layer of the hippocampus. This interesting pattern is in agreement with the clinical presentation of the patient, who showed dramatically pronounced episodic memory impairment, but was less impaired in global cognition than the other two patients. One limitation of this study is the sample size. The in vitro investigations need to be performed on a larger sample to address the variability of both clinical and pathological presentation in AD patients. Furthermore, a next step would be to investigate THK5117 binding distributions in other tauopathies. Since tau exists in different forms, it would be important to assess the form of tau to which THK5117 is binding, as this may have implications for our understanding of the different tauopathies.
Autoradiography binding studies with (S)-18 F-THK5117 indicated binding patterns that were very similar to those of 3 H-THK5117 in all three patients, although autoradiography with tritiated compounds is associated with higher resolution than autoradiography with fluorinated compounds. This finding is very promising for use in in vivo PET scans.
One interesting characteristics of this study was that autoradiography was performed on hemisphere sections from three patients with confirmed AD who had previously undergone 18 F-FDG during the early course of their disease. Thus, it was possible to compare the post-mortem THK5117 binding patterns with the ante-mortem patterns for 18 F-FDG in the same subjects. Although these findings should be viewed with caution, since the PET scans had been performed several years before death, a negative correlation between in vivo 18 F-FDG uptake and postmortem THK5117 binding was observed in two AD cases. More severe tau pathology is observed in regions with lower cerebral glucose metabolism while less tau pathology is found in regions with preserved metabolism, which may support the assumption that NFTs trigger neuronal dysfunction. No significant correlation was found for case three. Since α-synucleinopathy was reported in this case, although little is known about the relationship between αsynuclein and tau in AD, pathological processes involving α-synuclein may have influenced pathological progression in this patient.

Conclusions
In conclusion, the tau tracer THK5117 binds with high affinity and good selectivity to cerebral tau lesions. The binding patterns for 3 H-THK5117 not only showed significant differences between AD patients and controls, but also appeared to reflect the different tau pathology as well as clinical symptoms of the patients. The use of different in vivo tau tracers in PET imaging will shed further light on the role of tau pathology in the time course of AD and its concurrent clinical features.