18F-Flortaucipir in TDP-43 associated frontotemporal dementia

Retention of 18F-Flortaucipir is reportedly increased in the semantic variant of primary progressive aphasia (svPPA), which is dominated by TDP-43 pathology. However, it is unclear if 18F-Flortaucipir is also increased in other TDP-43 diseases, such as bvFTD caused by a C9orf72 gene mutation. We therefore recruited six C9orf72 expansion carriers, six svPPA patients, and 54 healthy controls. All underwent 18F-Flortaucipir PET and MRI scanning. Data from 39 Alzheimer’s Disease patients were used for comparison. PET tracer retention was assessed both at the region-of-interest (ROI) and at the voxel-level. Further, autoradiography using 3H-Flortaucipir was performed. SvPPA patients exhibited higher 18F-Flortaucipir retention in the lateral temporal cortex bilaterally according to ROI- and voxel-based analyses. In C9orf72 patients, 18F-Flortaucipir binding was slightly increased in the inferior frontal lobes in the ROI based analysis, but these results were not replicated in the voxel-based analysis. Autoradiography did not show specific binding in svPPA cases or in C9orf72-mutation carriers. In conclusion, temporal lobe 18F-Flortaucipir retention was observed in some cases of svPPA, but the uptake was of a lower magnitude compared to AD dementia. C9orf72-mutation carriers exhibited none or limited 18F-Flortaucipir retention, indicating that 18F-Flortaucipir binding in TDP-43 proteinopathies is not a general TDP-43 related phenomenon.

PET and MR Imaging. T1-weighted magnetization-prepared rapid gradient echo (MPRAGE) and Fluid-attenuated inversion recovery (FLAIR) images were acquired for all patients on a 3T Siemens Skyra scanner (Siemens Medical Solutions, Erlangen, Germany) and processed along with the PET images using an in-house developed pipeline, described previously 31 . Based on the FreeSurfer 5.3 (https://surfer.nmr.mgh.harvard.edu) regions derived using the Desikan-Killiany atlas larger bilateral composite ROIs (frontal and inferior frontal; medial and lateral parietal; medial and lateral temporal; and occipital cortex) were created, and to assess lateralization in patients with svPPA, a left and a right ROI was created for each lateral temporal lobe. The exact composition of these ROIs is described in the Supplementary Methods.
The radiosynthesis procedure, radiochemical purity, and scanning methods for 18 F-Flortaucipir have been described in detail previously 32 . Subjects in this study underwent a simplified protocol including an 18 F-Flortaucipir PET scan 80-100 min (4 × 300 s frames) post injection on a GE Discovery 690 PET scanner (General Electric Medical Systems, Milwaukee, WI, USA). PET data was processed using our in-house developed pipeline 31 . In brief, the PET data was motion corrected, summed and co-registered to the MRI data. Using the FreeSurfer segmentation of the MRI, standardized uptake value ratio (SUVR) calculations were performed using the inferior cerebellar grey matter as reference region 33 . For voxel-based analyses SUVR PET images were warped into Montreal Neurological Institute 152 standard space, the images were smoothed with an 8 mm full-width at half maximum (FWHM) Gaussian kernel. Calculations were performed using SPM12 (Wellcome Department of Cognitive Neurology, London, UK; http://www.fil.ion.ucl.ac.uk/spm) in MATLAB_R 2017b. Mean SUVR-images for AD patients and controls were calculated from non-smoothed SUVR-images in MNI-space using SPM12 and images prepared using Pmod 3.711 (Pmod Llc. Zurich, Switzerland). Individual MRI and SUVR PET-images of svPPA and C9orf72 patients were fused and prepared using the Fusion tool in Pmod 3.711. Partial volume error correction using the Geometrical Transfer Matrix method 34 was applied to the main data, results are presented in Supplementary Fig. 2. Autoradiography and immunohistochemistry. Fresh frozen blocks from temporal and frontal cortex from two patients with TDP-43 related semantic dementia and two patients with bvFTD due to expansions in the C9orf72-gene were kindly provided by the Dutch Brain Bank. Semantic dementia patients had TDP-43 pathology, type C, and in the C9orf72 mutation carriers TDP-43 pathology of type B. The cases selected for autoradiography showed no or minimal tau pathology using immunohistochemistry. Ten µm sections were cut on a Leica CM3050 cryostat at −17 °C chamber temperature and −15 °C object temperature. Sections were mounted onto Histobond + microscope slides (Marienfeld Laboratory Glassware) and dried for 3 hours at room temperature before storage at −20 °C. 3 H-Flortaucipir was tritiated at Roche with a specific activity of 33 Ci/mmol and a radiochemical purity higher than 99%. The brain sections were incubated in assay buffer (50 mM Tris buffer, pH 7.4), containing 10 nM radioligand at room temperature for 30 min. After three ten min washes at 4 °C in assay buffer and 3 quick dips in H 2 O dist. at 4 °C, the sections were dried at 4 °C for 3 h. The sections were then exposed to a FujiFilm Imaging Plate (BAS-IP TR 2025), placed in a FujiFilm Cassette (BAS 2025), for 5 days and subsequently scanned with a FujiFilm IP reader (BAS-5000) with a resolution of 25 µm per pixel. The autoradiograms were visualized with www.nature.com/scientificreports www.nature.com/scientificreports/ the software MCID analysis (version 7.0, Imaging Research Inc.). Non-specific binding of 3 H-Flortaucipir was assessed by co-incubation with 10 µM unlabelled T-808. The experimental protocol was previously set up and optimized to give a robust signal in tissue sections from positive controls.

statistics.
Statistical analyses for ROI-based comparisons were performed using GraphPad Prism 7.0a for Macintosh. For comparisons between multiple groups Kruskal-Wallis tests were used, followed by between group comparisons using Mann-Whitney tests where appropriate. Statistical significance was assumed at p < 0.05. For voxel-wise analyses of 18 F-Flortaucipir uptake patterns between respective patient groups and control subjects we employed a voxel-wise two-sample t-test as implemented in SPM12 (http://www.fil.ion.ucl.ac.uk/spm). The voxel-wise comparisons were thresholded using family-wise error (FWE) correction with a p-value of < 0.05.

Results
study participants. Six bvFTD C9orf72 expansion carriers, six svPPA patients (three svPPA, three R-SD) and fifty-four age-matched, neurologically healthy, controls were recruited to the study.
Clinical details of the participants are given in Table 1. Structural MRI in svPPA shows various degree of atrophy in the anterior and lateral temporal lobes supporting the clinical diagnosis ( Fig. 1). SvPPA cases all suffered from clinically established disease, with median symptom duration of 6 years and a CDR of 0,5 or 1. All C9orf72 cases presented with the clinical syndrome of bvFTD, with psychotic symptoms present in 4 of the cases, and none www.nature.com/scientificreports www.nature.com/scientificreports/ with concomitant motor neuron disease. Transversal and coronal T1-weighted MRI images and 18 F-Flortaucipir images for all participating svPPA and C9orf72 patients are shown in Fig. 1.
For subjects with C9orf72-mutations the only statistically significant difference was a slight increase of the retention in the inferior frontal lobes. The magnitude of the increase in the inferior frontal lobes was relatively www.nature.com/scientificreports www.nature.com/scientificreports/ small (Controls: 1.06 (range 0.94-1.26) vs C9orf72: 1.14 (range 1.07-1.20), p = 0.02; Fig. 2H). There were no significant differences in other brain regions. Since it has been shown that patients with C9orf72-mutations also have cerebellar dipeptide repeat and p62-pathology 35,36 , and since this pathology may interfere with the results using an inferior cerebellar reference region, we performed the same analysis using a brainstem reference. When using a brainstem reference no differences were seen in the inferior frontal lobes, but the remaining results in the C9orf72-mutation group were not changed (data not shown). Using partial volume error corrected data the numerical differences between groups increased and we found statistically significant differences between controls and patients carrying the C9orf72 mutation ( Supplementary Fig. 2). However, except for the inferior frontal ROI all or all-but-one of the C9orf72 values were within the control mean + two standard deviations (SD) values (dotted lines), indicating that the increases in absolute values were very small.
To compare the regional patterns of atrophy between the two disease groups the cortical thickness was calculated in each ROI. These values were normalized to z-scores using the control mean and SD and are presented in Fig. 3. We find the most severe atrophy in the anterior temporal lobes of subjects with svPPA, whereas C9orf72-mutation carriers instead show a more modest fronto-temporal atrophy.

Voxel-based analyses. We further performed voxel-based analyses in the svPPA and C9orf72
groups compared to controls. In the svPPA > control contrast we could identify a large significant cluster (k E = 18886; p FWE-corr < 0.001) in the anterior right temporal lobe and a smaller cluster in the left temporal pole (k E = 2693; p FWE-corr < 0.001) after applying family-wise error (FWE) correction (Fig. 4, upper panel). In the C9orf72-mutation carriers > control contrast there was one small cluster located in the right posterior limb of the internal capsule (k E = 45; p FWE-corr = 0.03; Fig. 4, lower panel). No suprathreshold clusters were identified using the Control > svPPA or the Control > C9orf72 contrasts. Similar results were obtained if the svPPA patient with the highest SUVR values (case 1) was excluded from the analysis (data not shown). www.nature.com/scientificreports www.nature.com/scientificreports/ Autoradiography. Fresh frozen tissue samples from frontal and temporal cortex from two patients with svPPA and two with C9orf72-mutations, with type C and type B TDP-43 pathology respectively, were analysed using 3 H-Flortaucipir autoradiography. No specific binding of 3 H-Flortaucipir to TDP-43 pathology could be detected in the cortical tissue sections in these tissue samples (Fig. 5A-E). A grainy autoradiography pattern was sometimes seen (Fig. 5D), but this signal could not be blocked by high concentration unlabeled T-808 and the grains did not colocalize with TDP-43 positive structures using TDP-43 immunohistochemistry ( Supplementary  Fig. 3). In a positive control from AD cortex (superior temporal gyrus) a clear positive 3 H-Flortaucipir signal, that was blocked after addition of cold T-808 compound, was seen in several cortical layers (Fig. 5F).

Discussion
We present data showing increased retention of 18 F-Flortaucipir in the lateral temporal lobes of left-sided svPPA and in the homologous right-sided SD using a ROI-based analysis. These results are in line with previously reported findings of increased binding potential (BP ND ) of 18 F-Flortaucipir in patients with svPPA/R-SD 19,20 , and consistent with the expected distribution of pathology in svPPA. At a voxel-wise level, significantly increased clusters in svPPA were located in the anterior temporal poles bilaterally, confirming the location of the ROI-based data. The underlying molecular pathology of svPPA/SD is most often type C TPD-43 pathology, but the syndrome can also be caused by AD pathology or 3R tau pathology (Pick's Disease) 1,3,37 . The Aβ status among the svPPA patients shows that three of the patients in this study have pathological levels of Aβ (Table 1). The β-amyloid-PET values are, however, borderline positive despite the patients having had disease symptoms for many years. We believe that the patients should have shown more aberrant values of Aβ, if the patients would have had their symptoms primarily due to AD pathology. Other notions that would support other causes than AD being the underlying pathology in these cases is that, apart from case 1, the magnitude of increase in 18 F-Flortaucipir SUVR is low in comparison to the retention seen in symptomatic AD (see Supplementary Fig. 4 for comparison); and, patterns of cortical atrophy showing a relatively isolated anterior temporal lobe atrophy in svPPA-cases. Further, the previously published findings of increased anterior lateral temporal 18 F-Flortaucipir retention in four 19 and five 20 Aβ-negative subjects from two independent research groups support the idea that there is an Aβ-independent increase in 18 F-Flortaucipir retention in svPPA-cases. Another theoretically possible explanation is the phenomenon of mixed TDP-43 and tau pathology, or concomitant "secondary" tau pathology, however this appears to be a very rare phenomenon 37 .
The finding of increased 18 F-Flortaucipir retention in the svPPA disease group has led to the questioning of the selectivity of the 18 F-Flortaucipir tracer for tau pathology, and off-target binding to TDP-43 has been suggested 19 . We therefore recruited six patients with bvFTD due to hexanucleotide expansions in the C9orf72 gene. www.nature.com/scientificreports www.nature.com/scientificreports/ The C9orf72 gene mutation is strongly associated to TDP-43 pathology of subtypes A and B 4,38 and the likelihood of these patients having TDP-43 pathology is thus high. In the ROI-based analysis of C9orf72 mutation carriers we did find a significantly increased retention of 18 F-Flortaucipir in the inferior frontal cortex, that could potentially fit with a behavioural variant FTD, but the magnitude of 18 F-Flortaucipir retention increase was low. This effect was not visible on voxel-based analysis and disappeared when using the pons as an alternative reference region. In the voxel-wise analysis of C9orf72-mutation carriers the only area that was statistically higher than in controls was a small cluster in the posterior limb of the internal capsule, the significance of this cluster is uncertain, but it colocalizes to a region with atrophy on MRI-based measures in C9orf72 patients 36 . Apart from these regions no regions or voxels with increased 18 F-Flortaucipir retention were detected, including the cerebral cortex and hippocampus, where TDP-43 pathology is prominent in C9orf72 carriers 35,36,38 . A previous case report has  www.nature.com/scientificreports www.nature.com/scientificreports/ shown an increased temporal retention of Flortaucipir in a subject with a C9orf72 mutation and a three year history of personality change 21 . The disease duration of the published subject is similar to the disease duration of the participants in this study. We could not reproduce this finding in our six patients. The Aβ-status of the published subject is unfortunately unknown and a concomitant AD can not be excluded.
Previous autoradiography studies have shown no 13 or minimal 12,15 binding of 18 F-Flortaucipir to post mortem tissue from TDP-43 related FTD cases. In these studies only cases with TDP-43 pathology of types A and C were included. In our autoradiography results we show no specific binding of 3 H-Flortaucipir to cortical tissue from C9orf72-mutation carriers having a type B TDP-43-pathology. In a similar fashion we found no specific binding to TDP-43 type C pathology in tissue from svPPA cases.
Taken together, our results do not support the view that the increased retention of 18 F-Flortaucipir seen in svPPA would be due to a general off-target binding to TDP-43 pathology. It is possible that 18 F-Flortaucipir might have an off-target binding to TDP-43 type C pathology, but the finding of no, or only minimal, binding of 3 H-Flortaucipir to type C TDP-43 pathology in autoradiography experiments would argue against that notion. The possibility of 18 F-Flortaucipir binding to monoamine oxidase B (MAO-B), similar to the MAO-B-binding of THK5351 39 has been suggested 19,20 , but recent studies have shown that there is no reduction of 18 F-Flortaucipir retention in patients taking MAO-B inhibitors 40,41 . Since the regions where the most intense 18 F-Flortaucipir retention is seen are regions with a pronounced atrophy a possible cause for the increased retention could be an off-target binding to another neurodegenerative process that parallels the atrophy in svPPA. Possibly, the severe atrophy seen in the temporal poles in svPPA may damage the blood-brain barrier and cause an increased retention of the 18 F-Flortaucipir tracer. Since the tracer kinetics of Flortaucipir differs between off-target binding regions, such as the putamen, and binding to AD tau pathology in the cerebral cortex 32 future kinetic studies may shed light on the nature of the cortical tracer retention in svPPA. Apart from the retention in the temporal lobes of svPPA-cases no other regions of unexpected off-target binding were seen in svPPA or C9orf72-mutation carriers.
A limitation of the present study is the low number of participants in the respective disease groups, and the results should be interpreted with this in mind. Another limitation is the absence of neuropathological confirmation of the diagnoses and direct neuropathological correlations of pathology to PET-retention. In svPPA the majority of cases are related to type C TDP-43 pathology, but despite determination of Aβ-status using CSF or PET other underlying pathologies, such as AD and 3R tau, cannot fully be excluded in our study. Further, the C9orf72-cases have a shorter disease duration and a lower degree of atrophy compared to the svPPA cases. We cannot exclude the possibility that there could be an increased retention in more advanced disease also in C9orf72 mutation carriers.
In conclusion, we find an increased 18 F-Flortaucipir retention in vivo in the lateral temporal cortices of patients with svPPA/R-SD. In patients with mutations in the C9orf72-gene, only very limited 18 F-Flortaucipir retention could be detected, indicating that binding of 18 F-Flortaucipir in TDP-43 proteinopathies is not a general TDP-43 related phenomenon.