Since the discovery in 1997 that the protein αSYN can be found in Lewy bodies in PD , much research on αSYN has been conducted. The aggregation of αSYN seems to be the central driver of the pathogenesis of synucleinopathies; however, the relationship between αSYN misfolding and cellular dysfunction or cell death is far from fully understood. The removal of αSYN aggregates holds considerable promise as a therapeutic strategy. In this context, imaging biomarkers are desperately needed for early and differential diagnosis, to follow disease progression and to determine the efficacy of potential disease-modifying treatments. An αSYN PET tracer would unquestionably be of high importance and a game-changing tool for diagnosis and therapy development.
Here, we aimed to develop a PET tracer based on the lead structure of anle138b by further modifying anle253b, which was previously reported by our group to have a good affinity towards human recombinant αSYN fibrils but poor pharmacokinetics in vivo, presumably caused by high lipophilicity with a calculated logP (clogP) of 5.21 . Ishikawa et al. performed studies on the improvement in polarity of drug candidates. They showed that exchange of the phenyl group with pyridine largely improved polarity in phosphate buffer at physiological pH, which consequently reduced the logP value . We designed a chemical derivative of anle253b, MODAG-001, in which one of the phenyl groups was exchanged with pyridine to reduce the lipophilicity of the compound to a clogP value of 3.85. After tritiation, [3H]MODAG-001 was tested in saturation binding experiments against human recombinant fibrils and showed very high affinity to αSYN fibrils, with a Kd value of 0.6 nM in vitro. Compared to [3H]MODAG-001, all the compounds, which have been described in the literature as potential αSYN PET ligands, showed an at least one order of magnitude lower binding affinity (higher Kd values) in vitro towards recombinant αSYN fibrils [14, 21, 24,25,26,27,28,29,30,31]. These compounds comprise [125I]SIL23, (Kd = 148 nM) and its derivative SIL26 (Ki = 15.5 nM),  [18F]BF-227 (Kd = 9.6 nM) , several fluorescent probes (Kds in the micromolar and elevated nanomolar range) [32, 33], and the 18F-labeled 3-(benzylidine)indolin-2-one derivative [18F]46a (Kd = 8.9 nM) .
Furthermore, we observed 30-fold selectivity over hTau46 and Aβ1–42 fibrils, which was suggested for an ideal CNS PET tracer, as both structurally similar proteins have also been shown to be highly abundant in many patients with synucleinopathies . However, the selectivity is determined not only by a lower Kd, but also by the difference of available binding sites. Bmax was approximately 7- and 50-fold higher for αSYN fibrils compared to hTau46 fibrils and Aβ1–42 fibrils, respectively. The measured fibril length of hTau46 (453.4 ± 261.9 nm) was three times larger compared to αSYN (152.5 ± 76.6 nm) and Aβ1–42 (139.4 ± 77.3 nm), which were very similar in size. Of note, not only the fibril length, but especially the amount of available binding sites per fibril length, which might differ due to differences in the 3D structure, may play an important role.
These encouraging in vitro binding data were the prerequisite for further 11C-labeling and in vitro and in vivo characterization of MODAG-001. The pharmacokinetic and metabolic profiles of [11C]MODAG-001 after i.v. injection into healthy mice revealed good BBB penetration with high uptake into the brain (SUV = 1.4) and relatively fast clearance from the brain. Our metabolite analysis revealed that [11C]MODAG-001 was degraded into the three metabolites, two of which were detectable in the brain, with 81% of the parent compound remaining at 15 min. We hypothesize that metabolite M3 might be the demethylated form of MODAG-001, as demethylation is a very common form of metabolic degradation. Fast metabolism resulting in BBB-penetrating radio-metabolites hampers tracer quantification. Previous studies have shown that improved metabolic stability can be obtained by incorporating deuterium into the molecule. In comparison to carbon-hydrogen bonds, carbon-deuterium bonds potentially decelerate metabolism by cytochrome P450 enzymes due to the primary kinetic isotopic effect . Since the NMe2 group of MODAG-001 is considered the main target of metabolism (M3 corresponds to the monodemethylated metabolite of MODAG-001), we fully deuterated the nonradioactive methyl group to reduce the formation of radio-metabolites and improve the metabolic stability of the tracer in vivo. Comparison of the pharmacokinetic and metabolic profiles of (d3)-[11C]MODAG-001 and [11C]MODAG-001 revealed that formation of the monodemethylated radio-metabolite M3 was reduced and that the ratio of M3 to M1 (which hypothetically represents the cleaved 11C-methyl group) was increased for (d3)-[11C]MODAG-001. Moreover, total metabolism of the parent compound was reduced, resulting in increased levels of the parent compound in the brain at 15 min. This is emphasized in Table 1 and Table 2 showing that the injected dose per gram for [11C]MODAG-001 was reduced from 10.5% at 5 min to 5.3% at 15 min, whereas the injected dose per gram for (d3)-[11C]MODAG-001 was reduced to a lower extent from 8.8% at 5 min to 7.2% at 15 min.
We next asked whether (d3)-[11C]MODAG-001 can detect aggregated αSYN in the brain and thus inoculated the right striatum in rats with the same batch of αSYN fibrils used for the in vitro binding experiments. (d3)-[11C]MODAG-001 binding properties in three fibril-inoculated rats and one non-injected control rat were assessed using in vivo PET imaging. In the three fibril-inoculated rats, we observed higher binding in the right striatum compared to the vehicle-inoculated contralateral side, with mean DVR-140-60min of 0.14 ± 0.1 (whole striatum) and 0.44 ± 0.21 (70% automatic isocontour detection) 4 days after injection, respectively, indicating in vivo binding of (d3)-[11C]MODAG-001 to inoculated αSYN fibrils. Furthermore, similar overall brain uptake was observed in fibril-inoculated rats (peak SUV, 2.1 ± 0.1) and the non-injected rat (peak SUV, 2.1), demonstrating that potential disruption of the BBB from the intracranial injection is unlikely to account for higher uptake into the inoculated striatum. A similar rat model was also used by Verdurand et al.  to select potential αSYN PET tracer candidates. Notably, they were able to detect the binding of [18F]BF227, [18F]2FBox, and [18F]4FBox to αSYN and Aβ1–42 fibrils using in vitro AR; however, no binding was observed in vivo despite good uptake into the cerebral tissue and high Am (68–543 GBq/μmol).
To further test binding of MODAG-001 in human brain slices with confirmed αSYN pathology, we performed in vitro AR using [3H]MODAG-001 in LBD, PSP, AD, and control cases . The advantage of using tritium AR over carbon-11 AR is its 10-fold higher spatial resolution of approximately 50 μm2 compared to the spatial resolution of carbon-11 AR of approximately 566 μm2 due to the positron range of the radionuclide [35, 36].
Despite the high affinity of [3H]MODAG-001 for recombinant αSYN fibrils and good selectivity for αSYN fibrils over hTau46 and Aβ1–42 fibrils, no strong binding was observed in brain sections from LBD cases. We observed slightly more intense binding in the cortical gray matter compared to the white matter, which could be blocked by the nonlabeled compound, but this signal was not significantly stronger than that in controls. Interestingly, clear, blockable binding corresponding to the distribution of Aβ plaques was observed in AD brain tissue. Whether this represents binding to Aβ fibrils or binding to aggregated αSYN species, which have been described in AD plaques, remains to be determined [37,38,39,40]. Notably, no evidence of binding to aggregated Tau was observed in the PSP and AD cases. A high SNR is an important requirement in brain binding studies with low target abundancy. Possible explanations why we did not observe stronger binding in LBD brain sections are likely related to low target availability in the brains of LBD patients, high nonspecific binding, and/or structural differences between the αSYN fibrils used in the screening assays and those in human brain tissue.
We determined the limit of detection of [3H]MODAG-001 using fibrils and brain homogenate at different concentrations. Quantification of specific binding was still possible at αSYN concentrations down to 5 nM in the presence of 100 μg protein/mL mouse brain homogenate, but not at higher homogenate concentrations of 500 μg protein/mL. When we used sucrose gradient centrifugation of human PD brain tissue to quantify aggregated αSYN, we found approximately 400 nM aggregated αSYN, corresponding to 4 nM aggregated αSYN and 830 μg protein/mL in homogenate at a 1:100 dilution (Supplemental Fig. S5c). At this fibril concentration, the homogenate concentration was approximately 8-fold higher than the limit for a similar fibril concentration determined in our assay. Therefore, we hypothesize that the nonspecific binding of [3H]MODAG-001 is likely responsible for the low signal-to-noise ratio in pathological human brain tissue. For MODAG-001, we calculated a clogP value of 3.85, which is rather high compared to the values calculated for successfully established PET ligands.
Structural differences between recombinant fibrils obtained in vitro and fibrils in the LB and LN of LBD patients may also account for the different binding behaviors of MODAG-001. Such structural differences were identified in a study in which αSYN fibrils from LBD patients were amplified from in vivo aggregates by protein misfolding cyclic amplification (PMCA) . The same study also revealed the heterogeneity of αSYN fibrils in different synucleinopathies using solution-state NMR spectroscopy and fluorescent probes. Furthermore, structural differences were also observed in cryo-EM studies in which artificially produced Tau fibrils were compared with Tau extracted from AD or Pick’s disease patients with Tau pathology [42,43,44,45].
Despite the presence of structural differences between recombinant fibrils and those found in human brain slices, homogenates, or αSYN brain extracts, the availability of the latter is very limited; as such, recombinant fibrils remain a widely used screening tool for the preselection of compounds before PET radiolabeling.
Many studies have shown that a large percentage of AD patients exhibit significant LB pathology in addition to Aβ plaques and neurofibrillary tangles (NFTs) and vice versa [46,47,48,49,50]. As an example, Hamilton et al.  investigated the postmortem tissue of 145 sporadic AD cases using immunohistochemistry for αSYN and observed LBs in 60.7% of all cases. Colo-Cadena et al.  found that Aβ deposition positively correlated with LBD pathogenesis. However, direct comparison of the results of AR and immunohistochemistry in the same AD brain slice revealed the clear colocalization of Aβ-positive plaques and bound MODAG-001. Therefore, the binding is either related to cross-β-sheet structures, a common feature shared by Aβ, Tau, and αSYN [51,52,53], or to the non-amyloid-β component (NAC) domain identified by Ueda et al., which is part of the αSYN protein in AD plaques . However, as we cannot provide experimental evidence for the latter, this remains speculative.
Determination of the αSYN fibril-binding sites of novel compounds is crucial for PET tracer development. Hsieh and Ferrie et al. identified three putative binding sites for fibrillar αSYN using a combination of in silico docking, photoaffinity labeling, and radiotracer binding studies . While styrene- and piperazine-based analogs showed a preference for sites 2 and 9, tricyclic compounds and an indolinone-diene analog showed a preference for site 3/13. We performed a [3H]MODAG-001 competition assay using the tricyclic compound SIL26 as a competitor and observed a Ki value of 21 nM, indicating preferential binding to site 3/13 at fibrillar αSYN. However, the interpretation needs to be done with care, as preference to sites 2 and 9 were not tested.