Early-phase [18F]PI-2620 tau-PET imaging as a surrogate marker of neuronal injury

Purpose Second-generation tau radiotracers for use with positron emission tomography (PET) have been developed for visualization of tau deposits in vivo. For several β-amyloid and first-generation tau-PET radiotracers, it has been shown that early-phase images can be used as a surrogate of neuronal injury. Therefore, we investigated the performance of early acquisitions of the novel tau-PET radiotracer [18F]PI-2620 as a potential substitute for [18F]fluorodeoxyglucose ([18F]FDG). Methods Twenty-six subjects were referred with suspected tauopathies or overlapping parkinsonian syndromes (Alzheimer’s disease, progressive supranuclear palsy, corticobasal syndrome, multi-system atrophy, Parkinson’s disease, multi-system atrophy, Parkinson's disease, frontotemporal dementia) and received a dynamic [18F]PI-2620 tau-PET (0–60 min p.i.) and static [18F]FDG-PET (30–50 min p.i.). Regional standardized uptake value ratios of early-phase images (single frame SUVr) and the blood flow estimate (R1) of [18F]PI-2620-PET were correlated with corresponding quantification of [18F]FDG-PET (global mean/cerebellar normalization). Reduced tracer uptake in cortical target regions was also interpreted visually using 3-dimensional stereotactic surface projections by three more and three less experienced readers. Spearman rank correlation coefficients were calculated between early-phase [18F]PI-2620 tau-PET and [18F]FDG-PET images for all cortical regions and frequencies of disagreement between images were compared for both more and less experienced readers. Results Highest agreement with [18F]FDG-PET quantification was reached for [18F]PI-2620-PET acquisition from 0.5 to 2.5 min p.i. for global mean (lowest R = 0.69) and cerebellar scaling (lowest R = 0.63). Correlation coefficients (summed 0.5–2.5 min SUVr & R1) displayed strong agreement in all cortical target regions for global mean (RSUVr 0.76, RR1 = 0.77) and cerebellar normalization (RSUVr 0.68, RR1 = 0.68). Visual interpretation revealed high regional correlations between early-phase tau-PET and [18F]FDG-PET. There were no relevant differences between more and less experienced readers. Conclusion Early-phase imaging of [18F]PI-2620 can serve as a surrogate biomarker for neuronal injury. Dynamic imaging or a dual time-point protocol for tau-PET imaging could supersede additional [18F]FDG-PET imaging by indexing both the distribution of tau and the extent of neuronal injury. Electronic supplementary material The online version of this article (10.1007/s00259-020-04788-w) contains supplementary material, which is available to authorized users.


Introduction
Tauopathies consist of neurodegenerative diseases including, among others, Alzheimer's disease (AD), frontotemporal dementia and atypical parkinsonian syndromes such as progressive supranuclear palsy (PSP) and corticobasal syndrome (CBS), with the connective This article is part of the Topical Collection on Neurology Electronic supplementary material The online version of this article (https://doi.org/10.1007/s00259-020-04788-w) contains supplementary material, which is available to authorized users. characteristic of misfolded and accumulated tau protein in different parts of the brain [1,2]. Visualization of tau deposits in vivo has become possible with various tautargeting ligands for use with positron emission tomography (PET) [3,4]. While first-generation tau tracers suffered from off-target binding [5,6] and a resulting large inter-and intra-case variability [7], secondgeneration tau-PET tracers showed promising first results in vivo with specific binding in affected regions in patients with mild cognitive impairment and AD compared to healthy controls [8][9][10][11].
Tau-PET imaging complements an important biomarker for the characterization of neurodegenerative diseases. For AD, it has been proposed to classify the disease according to the biomarkers for amyloid, tau and neuronal injury by the A/T/N scheme [12]. In this classification scheme, neuronal injury in the pathological definition can be determined in vivo by three different diagnostic approaches. Atrophy in structural magnetic resonance tomography and total tau in cerebrospinal fluid are considered as well as hypometabolism in 18 F-fluorodeoxyglucose ([ 18 F]FDG)-PET. This is underpinned by a combined study in prion disease, indicating that metabolic imaging via [ 18 F]FDG correlates with neuropathologic changes including neuronal loss [13].
For first-generation tau tracers and several tracers for amyloid-PET imaging, it has been shown that the perfusion-phase images as obtained by these tracers is comparable to glucose metabolism as assessed by [ 18 F]FDG-PET and can therefore be used as surrogate biomarker of neuronal injury [14][15][16][17][18][19][20][21]. With respect to cost and radiation exposure, such "one-stop-shop" protocols have the opportunity to examine two important biological markers with one procedure.
In contrast to amyloid imaging where the subjects can only be classified as amyloid-positive or amyloidnegative, tau-PET imaging shows characteristic patterns for several different neurodegenerative entities [3]. Therefore, tau-PET imaging could probably also be used to discriminate a range of tauopathies beyond AD. Due to the lack of specificity of first-generation tau tracers, it is of great interest whether and how the perfusionphase images as obtained by second-generation tracers can be used as a marker of neuronal injury.
Thus, we aimed to investigate the potential of the secondgeneration tau-PET ligand [ 18 F]PI-2620 [22] as an additional (in addition to detecting tau pathology) biomarker of neuronal injury. We validated early-phase [ 18 F]PI-2620 data against [ 18 F]FDG-PET and focused on optimizing dynamic or coffee break acquisition protocols for tau-PET imaging with dual biomarker information.

Study design and patient enrolment
Patients with different suspected tauopathies were referred by dementia or movement disorder experts to [ 18 F]PI-2620 tau-PET imaging. We selected all subjects with an additional [ 18 F]FDG-PET acquired < 12 months before/after tau-PET imaging. The cohort consisted of patients with different suspected clinical diagnoses and represented a true clinical cross-section in a tertiary centre with a focus on dementia/ movement disorders. The most likely clinical differential diagnosis was recorded before [ 18 F]PI-2620 tau-PET imaging. All patients provided informed written consent to PET imaging. The study was conducted in accordance with the principles of the Declaration of Helsinki, and approval for scientific data analysis was obtained from the local ethics committee (application number 17-569).

Radiosynthesis
Radiosynthesis of [ 18 F]PI-2620 was achieved by nucleophilic substitution on a BOC-protected nitro precursor using an automated synthesis module (IBA, Synthera). The protecting group was cleaved under the radiolabelling conditions. The product was purified by semipreparative HPLC. Radiochemical purity was 99%. Non-decay corrected yields were about 35% with a molar activity of 8•10 6 GBq/mmol at the end of synthesis. [ 18 F]FDG was purchased commercially.

Image processing
All image data were processed and analysed using PMOD (version 3.5, PMOD Technologies Ltd., Zurich, Switzerland). For spatial normalization, tracer-specific templates in the MNI space were created for [ 18 F]PI-2620 (30-60 min) and [ 18 F]-FDG as described previously [14]. Dynamic [ 18 F]PI-2620 images were coregistered to the MNI space by applying the 30-60-min transformation (brain normalization settings: nonlinear warping, 8-mm input smoothing, equal modality, 16 iterations, frequency cut-off 3, regularization 1.0, no thresholding). All images were analysed in MNI space. The regional cerebral blood flow estimate R 1 was computed from the dynamic [ 18 F]PI-2620 images by applying the simplified reference tissue model 2 (SRTM2) as described previously [24,25], using the cerebellum (excluding the dentate nucleus and superior layers) as a reference region. A total number of ten predefined cortical volumes (bilateral frontal, central region, parietal, temporal, occipital) of interests (VOIs) deriving from the Hammers atlas [26] were delineated in the MNI space and standardized uptake value ratios (SUVr) of all VOIs were extracted for the different images (and different time frames) used for the analysis.
In preparation for visual analyses of all images (after selection of the appropriate time frame), dynamic data frames #2-#5 (0.5-2.5 min) were summed. To account for the lower count statistics of the early-phase images, an additional 6.0mm Gaussian filter was applied for both ( F]PI-2620-PET SUVrs for the ten cortical VOIs were extracted and correlated with the SUVrs of the corresponding [ 18 F]FDG-PET data after normalization of uptake to global mean (GBM) or by use of a cerebellar reference region (CBL). Pearson's correlation coefficients (R) were compared between different time frames and the selection of the optimal early-phase time window was based on the extent and significance of single frame agreement.

Regional comparison of optimized early-phase [ 18 F] PI-2620 0.5-2.5min and [ 18 F]PI-2620 R1 versus [ 18 F]FDG
The SUVr values of the optimized summed [ 18 F]PI-2620 early-phase image (0.5-2.5 min), the [ 18 F]PI-2620 R 1 image and the [ 18 F]FDG image were correlated for all ten cortical regions to investigate the regional relationship between tau-PET perfusion and glucose metabolism.

Visual analysis of stereotactic surface projections
For visual interpretation of early-phase [ 18  Three more experienced nuclear medicine physicians (H.B., T.v.E., M.B.) and three less experienced nuclear medicine interns (L.B., M.U., J.S.) visually assessed the 3D-SSP images using the Z-score maps (GBM scaling) and rated cortical regions used in the clinical routine (bilateral frontal, parietal, temporal, occipital cortex areas) from 0 to 3 (0 = no reduction, 1 = low reduction, 2 = intermediate reduction, 3 = severe reduction 0.5 steps were allowed). Furthermore, whole-brain 3D-SSP images were rated binary (0/1) and according to the severity of present neurodegeneration (0 = no/1 = mild/2 = intermediate/3 = severe neurodegeneration; 0.5 steps were allowed). A significant neuronal injury of the patient (A/T/N: N+) was defined by the majority read of binarized [ 18

Demographics
A total of 26 subjects (age = 66 ± 11 years, 17 female) were included in the analysis. The cohort consisted of seven subjects with a most likely diagnosis of AD, 13 subjects with movement disorders and most likely diagnosis of PSP or CBS, one case with most likely frontotemporal dementia, two cases with most likely multi-system atrophy, two cases with most likely Parkinson's disease, and one case with cognitive impairment of unknown reason. The mean time interval between both PET investigations was 1.2 ± 1.7 months. For  Fig. 1). The correlation was statistically significant for all cortical regions for frames #2-#7 for global mean scaling (0.5-4 min after injection), but frames #6-#7 showed overall lower correlation coefficients compared to frames #2-#5 (especially with a cerebellar reference region). Because of the fast wash-out and sufficient count statistics for frames #2-#5, the mean values of the 0.     Fig. 3. In all cases, the regional pattern of hypoperfusion in [ 18 F]PI-2620 0.5-2.5min and [ 18 F]PI-2620 R1 3D-SSP resembled the hypometabolism pattern in [ 18 F]FDG-PET. Furthermore, exemplary distribution volume ratio images of those cases with the specific tau-PET binding pattern are shown in Supplementary Fig. 1. Visual interpretation revealed moderate to high regional correlations for the majority of regions between early-phase tau-PET (both PI-2620 0.5-2.5min and PI-2620 R1 ) and [ 18 F]FDG-PET (see Table 3).

Discussion
Tau-PET imaging is of great interest as tauopathies make up the majority of neurodegenerative diseases [28]. However, not only information on tau-positivity but also on the presence of neuronal injury is considered in current diagnosis or research criteria of AD [12,29] and non-AD tauopathies [30]. In accordance with previous amyloid-PET studies [14][15][16][17][18]21], and considering the known limitations of first-generation tau radiotracers [19,31], we aimed to evaluate the potential of the early-phase of the second-generation tau-PET radiotracer [ 18 F]PI-2620 [11] as a potential surrogate biomarker of neuronal injury. Our results demonstrate a strong quantitative and visual agreement between reduced perfusion in tau-PET and reduced glucose metabolism in [ 18 F]FDG-PET imaging. Therefore, early-phase tau-PET images may potentially eliminate the need for an additional [ 18 F]FDG-PET. The ability to obtain two different biomarkers reduces costs, time and radiation exposure, and enables improved differential diagnosis through a one-stop-shop procedure, which is especially important for disabled patients.
The semiquantitative comparison of [ 18 F]PI-2620-PET single frames with individual [ 18 F]FDG-PET images revealed the highest agreement at 0.5-2.5 min after injection and decreased steadily until it was no longer significant for all cortical regions later than 4 min after injection. Previous studies evaluating the agreement between early-phase amyloid-PET acquisitions and [ 18 F]FDG-PET suggested time frames up to 10 min after injection, often sparing the first minute after injection [14,15,17]. Compared to these studies, we can conclude that the acquisition of early-phase [ 18 F]PI-2620 must include a very early and relatively short time frame to achieve a high agreement between tau-PET and [ 18 F]FDG-PET, likely related to the fast washout of [ 18 F]PI-2620 (Fig. 1a). Although the frames between 0.5 and 4.0 min p.i. indicated a significant correlation with [ 18 F]FDG-PET, we recommend the use of 0.5 to 2.5 min p.i. for early-phase [ 18 F]PI-2620 acquisition as this time window was more robust across different scaling methods. For implementation in study or clinical routine workflows, dynamic or coffee break protocols (included break between early-and late-image acquisitions with the possibility for the patient to rest) could be considered. In this regard, dynamic protocols will have the advantage of being able to perform non-invasive kinetic modelling for the tau-PET readout, which will be important to account for blood flow changes in longitudinal studies [32]. We also propose to exclude the first half minute of acquisition, which showed more variability, likely related to variance in blood flow (i.e., reduced cardiac ejection fraction), site of injection, and variance of manual injection. Therefore, we additionally calculated the regional cerebral blood flow estimate R 1 of [ 18 F]PI-2620 to take such individual differences into account and found similar results by this approach. The VOI-based semiquantitative regional correlation of both the summed early-phase (0.5-2.5 min) and R 1 of [ 18 F]PI-2620 showed a highly significant correlation in all assessed cortical regions and no significant advantage for one of the methods. For clinical implementation, the reconstruction of a summed image seems to be more suitable as it can be easily added to conventional reconstruction algorithms of common PET systems.
All semiquantitative analyses were performed with global mean normalization and whole cerebellum scaling. In accordance with our previous investigation on early-phase amyloid-PET [14], the cerebellar normalization approach showed lower correlations for all investigated cortical regions. Since cerebellar perfusion tends to be highly variable, e.g., as a consequence of significant gender differences [33], a global mean normalization is likely superior to cerebellar scaling for interpretation of early-phase tau-PET images at the level of individual patients. Nevertheless, in subjects with an overall high load of neuronal injury or other reasons of a reduced whole brain perfusion (e.g., after ischemic stroke), the cerebellar scaling proved to be a viable alternative approach for normalization.
Based on the semiquantitative results, the visual analysis of summed [ 18 F]PI-2620 perfusion-like images was also performed with 3DSSP images generated by global mean normalization. Both tau-PET early-phase images (summed 0.5-2.5 min and R 1 ) showed an overall high visual agreement for the detection of neuronal injury with the corresponding metabolic imaging. Less and more experienced nuclear medicine physicians had no or only minor disagreement between early-phase [ 18 F]PI-2620-PET and metabolic images in the majority of cases, indicating broad clinical applicability. Different patient preparation prior to injection of [ 18 F]FDG (eye patch and noise cancelling according to the EANM protocol [23]) and [ 18 F]PI-2620 (room noise, no eye patch) needs to be considered as a limitation.
In our population of subjects referred with suspected tauopathies, half of all patients showed a significant neuronal injury pattern. This highlights the potential of this dual-phase protocol for evaluating two different biomarkers with one examination. While the detailed additive value of early-phase tau-PET imaging ultimately needs to be evaluated in larger samples, we already obtained promising results in individual cases. The patient with a final clinical diagnosis of frontotemporal dementia shown in Fig. 3 was initially referred with suspected AD, and the combination of tau-negativity together with a frontal hypoperfusion pattern was in line with the clinical presentation after 1-year follow-up. For the evaluation of perfusion-like images as a surrogate of neuronal injury, the comparison with metabolic imaging has the restriction that even a metabolic correlate might not unequivocally represent the actually existing neuronal injury. Temporary neuronal dysfunctions, for example, in the Both reduced metabolism in [ 18 F]FDG-PET and reduced perfusion in perfusion-phase images of different radiotracers are influenced and partially caused by partial volume effects and might not purely reflect the underlying neuronal injury. Nevertheless, a close correlation with atrophy in magnetic resonance imaging [34] or neuropathology findings [13] was demonstrated for metabolic imaging via [ 18 F]FDG. Therefore, most classification schemes, such as ATN [12], list [ 18 F]FDG-PET as a surrogate of neuronal injury together with atrophy in structural magnetic resonance imaging and total tau in cerebrospinal fluid. Our findings clearly demonstrate that dualphase [ 18 F]PI-2620-PET cannot only provide information on "T" [22] but has also the potential to facilitate assessment of "N".

Conclusions
The present study demonstrates that early-phase imaging of second-generation [ 18 F]PI-2620 tau-PET can serve as a surrogate biomarker for neuronal injury as it shows excellent semiquantitative and visual agreement with metabolic imaging using [ 18 F]FDG-PET. Dynamic imaging or a dual time-point protocol for tau-PET could replace additional [ 18 F]FDG-PET imaging by indexing two biomarkers in neurodegenerative disease, the distribution of tau and the amount and regional pattern of neuronal injury. The short time required for recording perfusion-like images is a great advantage in terms of patient comfort, examination time, radiation safety and costeffectiveness.
Acknowledgements The authors thank the staff of the departments of nuclear medicine and neurology at the University Hospital LMU Munich. We thank the patients and their families. Ethical approval All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee (LMU Munich-application number 17-569) and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.
Informed consent Informed consent was obtained from all patients. All patients provided informed written consent to PET imaging. The study was conducted in accordance with the principles of the Declaration of Helsinki, and approval for scientific data analysis was obtained from the local ethics committee (application number 17-569).
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated Fig. 4 Disagreement between visual rating of early-phase a [ 18 F]PI-2620 0.5-2.5min and b [ 18 F]PI-2620 R1 3D-SSP images in contrast to visual rating of [ 18 F]FDG-PET 3D-SSP images for all evaluated brain regions of all cases (n = 624). Less experienced (LER) and more experienced (MER) readers are displayed separately otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.