Introduction
The identification of specific biomarkers that allow for early detection of tau pathology in four-repeat tauopathies (4RTs) will become crucial for target engagement in tau targeting treatment trials. In this regard, imaging with the next-generation tau-PET tracer [
18F]PI-2620 facilitated discrimination of patients with a clinical diagnosis of the 4RTs progressive supranuclear palsy (PSP) [
1] and corticobasal syndrome (CBS) [
2] from healthy controls, non-tauopathy Parkinson syndromes and Alzheimer’s disease (AD). [
18F]PM-PBB3 also has the potential to differentiate 4RTs in vivo [
3]. Neurodegeneration of cortical and subcortical brain regions is a common feature of 4RTs [
4,
5], comprising a relevant objective parameter of disease progression [
6] in AD as the most frequent tauopathy [
7]. For AD, it has been proposed to classify the disease according to biomarkers for amyloid, tau, and neurodegeneration by the A/T/N scheme [
8]. In this classification scheme, neurodegeneration on a biomarker level can be determined in vivo by different diagnostic approaches: (i) atrophy in structural magnetic resonance imaging (MRI), (ii) levels of total tau in cerebrospinal fluid, (iii) hypometabolism in [
18F]fluorodeoxyglucose-(FDG)-PET, (iv) or hypoperfusion with several imaging techniques such as single-photon-emission-computed-tomography (SPECT). Brain atrophy in MRI as well as region-specific hypometabolism patterns in FDG-PET are well established in the diagnostic work-up in 4RTs [
9]. However, an equivalent concept to A/T/N of combining simultaneous visualization of tau pathology and neurodegeneration in 4RTs has not yet been established. Recently, we reported that early-phase imaging of β-amyloid PET closely matches the pattern of glucose uptake in CBS [
10] and we found that early-phase imaging of [
18F]PI-2620 tau-PET could also serve as a surrogate of brain perfusion in mixed neurodegenerative disorders [
11]. With respect to cost, radiation exposure, and patient burden, such “one-stop shop” protocols provide the opportunity to examine two important diagnostic and potentially also prognostic biomarkers simultaneously with one procedure. We hypothesized that early-phase imaging with [
18F]PI-2620 PET mirrors the known neurodegeneration pattern in the brain of 4RT patients when compared to controls. Furthermore, we hypothesized that 4RT-related perfusion expression patterns may facilitate the discrimination of 4RT patients against other neurodegenerative diseases. Since the utility of imaging biomarkers for diagnosis and disease progression may well differ [
12], we directly correlated perfusion patterns as well as the amount and pattern of tau pathology with clinical and functional scores in 4RTs.
Discussion
In this cross-sectional study, we found that early-phase [18F]PI-2620 imaging yielded a valuable surrogate biomarker for perfusion alterations in 4RTs. The observed pattern of hypoperfusion in patients with 4RTs, as compared to healthy controls, matched the known topology of neuronal dysfunction in PSP and CBS. However, our study indicated that only consideration of combined brain regions has potential to facilitate discrimination of 4RT patients from patients with other neurodegenerative disorders that underwent an equal clinical workup, since single region hypoperfusion was not specific enough in 4RTs. Furthermore, we observed that combining perfusion and tau pattern information may have an additive value for the discrimination of 4RTs from other neurodegenerative disorders compared to each pattern alone. Finally, we observed stronger associations between 4RT-related perfusion pattern expression with clinical severity scales when directly compared to corresponding tau deposition. This implies that perfusion imaging could facilitate an objective read-out of disease progression of neurodegeneration in 4RTs and needs to be tested in longitudinal studies with the goal of validation as an endpoint for clinical trials.
The first goal of this study was to validate [
18F]PI-2620 perfusion imaging for detection of regional neuronal dysfunction in 4RTs. Our previous study found a strong correlation between early static SUVr and R1 of [
18F]PI-2620 with FDG-PET in a mixed population of neurodegenerative disorders [
11]. Thus, we hypothesized that early static SUVr of [
18F]PI-2620 facilitates the detection of known neuronal injury patterns of 4RTs against healthy controls. We selected 0.5–2.5 min SUVr since this methodology can be achieved by a simple dual-phase [
18F]PI-2620 protocol, readily providing images for clinical interpretation without high sophisticated reconstruction and analysis methodology. In line with the known patterns of neuronal injury that were detected by perfusion imaging or FDG-PET in 4RTs [
6,
21‐
23], we found a fronto-temporal and subcortical hypoperfusion with predominance in the thalamus, the caudate nucleus, and the anterior cingulate cortex at the group level of 4RTs against healthy controls. The putamen and the globus pallidus indicated a non-significant hyperperfusion, which was consistent with the regionally elevated time-activity-curves in the basal ganglia of patients with PSP within the perfusion phase [
1]. This general pattern of perfusion in a mixed cohort of patients with 4RTs likely represents the least common denominator of perfusion alterations regardless of distinct clinical features among subgroups. Future studies should interrogate the associations between varying phenotypes of patients with 4RTs and resulting deviations from this general perfusion pattern.
On the group level, statistical analysis indicated satisfactory sensitivity of [
18F]PI-2620 perfusion imaging for detection of 4RTs. We challenged the methodology by a mixed sample of 4RTs and other neurodegenerative diseases and used a threshold-based multiregion classifier. Here, we found only low specificity of [
18F]PI-2620 perfusion imaging and very limited PPVs and NPVs for detection of 4RTs (average PPV/NPV < 60%). In line, low specificity of perfusion imaging and FDG-PET were consistently reported when different neurodegenerative disorders were evaluated against each other instead of comparing against healthy controls [
24]. Our findings support regional similarity of hypoperfusion among diseases with partially similar clinical phenotype such as PSP-F and FTD or CBS and asymmetric AD. Thus, our findings were not surprising and emphasized the need for more detailed analyses of neuronal injury patterns [
25]. Indeed, several studies successfully investigated data-driven metabolic network-based classification algorithms for discrimination of atypical Parkinsonian syndromes [
25‐
27]. Here, sensitivity, specificity, PPV, and NPV for differential diagnosis of different parkinsonian syndromes were consistently > 80% in an automated image-derived classification procedure [
25]. Importantly, one of these studies found that metabolic expression patterns did not differ between patients with PSP and patients with CBS [
26] which supports pooling of 4RTs [
28]. This was also justified since the majority of patients with CBS of our sample also fulfilled the MDS PSP criteria [
9]. Interestingly, in our clinically pre-diagnosed cohort, the perfusion 4RT-related pattern expression showed potential for discrimination of patients with 4RTs from patients with mixed neurodegenerative diseases (AUC: 0.850). This suggests that consideration of whole brain patterns facilitates improved discrimination when compared to consideration of single regions with strongest hypoperfusion in 4RTs. This indicated the presence of disease-specific pattern apart from the regions with severe neurodegeneration and our validation cohort substantiated the usefulness of the determined networks. Furthermore, midbrain glucose hypometabolism to FDG-PET and midbrain atrophy in structural MRI were already acknowledged as supportive imaging biomarkers for diagnosis of PSP [
9,
29]. In conclusion, perfusion pattern expression shows promise for 4RT discrimination in comparison to the multi-region classifier discussed above. Interestingly, a recent [
18F]FP-CIT study similarly indicated that the early-phase of a brain PET ligand facilitates quantification of a metabolic network expression surrogate [
30].
Subsequently, we tested if the combination of early and late phase 4RT pattern expressions of [
18F]PI-2620 provide an additive value. Assuming that early-phase [
18F]PI-2620 imaging provides the neuronal injury pattern [
11] and late-phase [
18F]PI-2620 imaging delivers information on tau aggregation [
1,
31], we hypothesized a complementary gain of information. As a limitation it needs to be considered that [
18F]PI-2620 binding in patients with 4RTs [
1,
2] was not yet confirmed by autopsy in patients that underwent PET. Nevertheless, our data suggest an additive value for the combination of pattern expression in comparison to stand-alone perfusion or tau for the discrimination of 4RTs against other neurodegenerative disorders. Higher sensitivity of perfusion and higher specificity of tau pattern expression fit into the concept of “(N)” and “T” biomarker information, meanwhile well established for AD [
32]. A strength of this comparison is the head-to-head evaluation of perfusion and assumed tau information in a relevant number of cases with clinically diagnosed 4RTs, according to current diagnosis criteria. As a limitation, it needs to be considered that we used 20–40-min static SUVr for assessment of tau pattern expression [
15], and not the gold standard kinetic modeling approach. The focus of this research aimed to generate data that can be used in a clinical routine setting which is easier to accommodate by static windows. Therefore, it needs to be considered that the used 20–40-min static SUVr can be influenced by altered cerebral blood flow [
33].
Additionally, in this study, AUC values are assessed in an already clinically diagnosed cohort with clinical evaluation being the current standard for diagnosis, which limits the value of the individual AUC values to hypothesis generating data. Therefore, here we primarily compare the additional value of combined tau and perfusion expression pattern against each pattern on their own, while prospective studies in a cohort of patients suspected to suffer from neurodegenerative disease will be needed to properly test the AUC values of tau and perfusion pattern expression against clinical diagnosis as gold standard for identification of 4RT patients. Our assessment of AUC values in an already prediagnosed cohort strongly support the hypothesis of pattern expressions being valuable biomarkers in 4RT but need to be followed up on and tested in the prospective study design.
We observed a significant association between 4RT-related perfusion pattern expression and clinical severity in our patient cohort with 4RTs. This association was specifically observed with PSP rating scale scores and activities of daily living (SEADL) but not apparent for cognitive screening (MoCA). Therefore, our findings indicate that the regional networks involved in 4RT-related perfusion pattern expression have stronger associations with gait, bulbar, limb motor, and ocular motor features than cognitive domains captured by MoCA. We note that MoCA was not developed as a dedicated screening tool for 4RTs which might limit its interpretation. Fitting to patterns of atrophy, we observed congruent decreases of early- and late-phase [
18F]PI-2620 PET in regions near to the ventricles (i.e. caudate). Thus, it is likely that the detected lower tracer binding in these regions is not only related to hypoperfusion but also to partial volume effects, which was not entirely recoverable by PVEC. Based on our findings, we hypothesize that 4RT-related perfusion pattern expression could be a relevant biomarker for clinical progression in 4RTs which deserves testing in longitudinal studies. This could be relevant for monitoring of therapy trials since associations between different tau-related biomarkers and clinical progression in 4RTs were lower or not present in earlier cross-sectional [
1,
3,
34] and longitudinal [
35] studies. We note that an a priori available 4RT-related expression pattern of the [
18F]PI-2620 perfusion phase was not available. Thus, it was necessary to use our large cohort as a training set with only a small validation set available. Longitudinal studies will aid deciphering the pathophysiology underlying the association of detected 4RT-related perfusion pattern with clinical symptoms and symptom progression. The data presented here suggests, that not tau depositions but rather resulting neuronal cell loss, i.e. perfusion, predicts symptom development and progression. Prospective investigations will be needed to understand the interplay of tau pathology, perfusion deficits, and clinical disease presentation in the 4RT disease spectrum. As a limitation, we acknowledge that autopsy confirmation of clinical diagnosis was only available in few patients. Thus, the analyses of the manuscript rely on clinical diagnosis, supporting biomarkers, and confirmation during clinical follow-up, which implies a limited number of wrong diagnoses, given by the nature of an observational study.
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.
German Imaging Initiative for Tauopathies (GII4T)
LMU Munich, Dept. Neurology: Johannes Levin, Jonathan Vöglein, Urban Fietzek, Sonja Schönecker, Georg Nübling, Catharina Prix, Kai Bötzel, Adrian Danek, Carla Palleis, Endy Weidinger, Sabrina Katzdobler.
LMU Munich, Dept. Nuclear Medicine: Matthias Brendel, Mengmeng Song, Alexander Nitschmann, Maike Kern, Gloria Biechele, Anika Finze, Leonie Beyer, Peter Bartenstein, Stefanie Harris, Julia Schmitt, Florian Eckenweber, Simon Lindner, Franz-Joseph Gildehaus, Emanuel Joseph, Maximilian Scheifele, Christian Zach.
LMU Munich, Dept. Psychiatry and Psychotherapy: Robert Perneczky, Jan Häckert.
LMU Munich, Dept. Radiology: Boris-Stephan Rauchmann, Sophia Stöcklein.
Hannover Medical School, Dept of Neurology: Günter Höglinger, Gesine Respondek.
University of Leipzig, Dept. Nuclear Medicine: Henryk Barthel, Marianne Patt, Andreas Schildan, Osama Sabri, Michael Rullmann.
University of Leipzig, Dept. of Neurology: Joseph Classen, Dorothee Saur, Jost-Julian Rumpf.
Max-Plank-Institute of Human Cognitive and Brain Sciences Leipzig: Matthias L. Schroeter,
Technical University of Munich, Dept. Neurology: Matthias Höllerhage.
University of Cologne, Dept. Nuclear Medicine and Forschungszentrum Jülich: Alexander Drzezga, Thilo van Eimeren, Jochen Hammes, Bernd Neumaier.
University of Cologne, Dept. Neurology: Michael T. Barbe, Oezguer Onur.
DZNE Munich/Bonn: Estrella Morenas-Rodriguez, Jochen Herms, Sigrun Roeber, Thomas Arzberger, Christian Haass, Frank Jessen.
Life Molecular Imaging: Andrew Stephens, Norman Koglin, Andre Mueller.
Publisher's note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.