The goal of this nonhuman primate project was to assess whether TSPO PET imaging with [18F]FEPPA can be used in translational studies as a sensitive in vivo measure of microglial activation in the brain. As a study platform, we took advantage of an ongoing project in our lab in which hemiparkinsonian rhesus macaques were treated with allogeneic iPSC-mDA and we compared the in vivo [18F]FEPPA PET uptake to postmortem CD68 brain expression to further validate the PET imaging results. We chose to focus on [18F]FEPPA because, compared to the first-generation TSPO ligand [11C]PK11195, it has an increased signal-to-noise ratio due to greater specific binding [38–40], plus the fluorine-18 labelling improves image quality and increases half-life of the radioligand compared to carbon-11 .
Of significant translational interest is that [18F]FEPPA has been used to evaluate inflammatory cell recruitment in the brains of PD patients [41–43]. The imaging studies successfully captured the mild neuroinflammatory changes induced by the disease, yet they did not identify a direct relationship between the amount of microglial activation observed by [18F]FEPPA and PD progression. The results likely reflect stable activated microglia recruitment over time, although nucleotide polymorphism in the TSPO gene (rs6971) found in humans of European heritage may complicate the interpretation of the results . The monkeys in this study were rendered with stable hemiparkinsonism by intracarotid artery administration of the neurotoxin MPTP. Mild neuroinflammation is observed in the nigrostriatal system after a MPTP challenge  and persists for many years . The ongoing MPTP-induced neuroinflammation may have contributed to the clear unilateral inflammatory response triggered by the allogeneic grafts and detected by [18F]FEPPA signal in all animals.
A critical first step for preclinical application was to improve and standardize radioligand production. A method reported by Vignal et al (2018) produced high yields of [18F]FEPPA (34 ± 2% ndc) using an ethanol-based separation and 5 mg of precursor without an SPE step . A similar method using an ethanol-based separation but a lower reaction temperature (70C for 20 minutes) produced yields of 13 ± 8% ndc . Compared to these methods and the initial synthesis by Wilson et al (2008), our radiochemical synthesis optimization allowed high yield production of [18F]FEPPA (25 ± 8% ndc) with a 60% reduction in precursor mass, thus decreasing the cost of production and reducing the analyte mass load on the HPLC column. With our method, the use of an additional cartridge-based SPE step prior to semipreparative HPLC ensured no free [18F]-fluoride or residual kryptofix222 appeared in the final product, while further reducing the HPLC analyte mass load. The use of acetonitrile in the semipreparative HPLC mobile phase resulted in a clearly defined separation of product from impurities. Although the retention time of [18F]FEPPA increased with this semipreparative HPLC separation, the final product solution contained very low impurity mass, making it suitable for use in nonhuman primates and potential human studies.
The [18F]FEPPA data was analyzed using SUVs of static images averaged across the 90–120 minute image frames. Analysis was restricted to static images rather than use of kinetic modeling on the dynamic data since arterial metabolite sampling was not performed and [18F]FEPPA, among many other second-generation TSPO radioligands, does not have a validated reference region . Variability in the uptake of [18F]FEPPA in the monkeys was also observed, making direct comparisons of SUVs between the monkeys difficult to interpret. Thus, use of an asymmetry index was chosen as the metric to best interpret the data since each monkey contained a control region without the presence of iPSC-mDA grafts to compare against. To assess the SUVs at the graft sites, ROIs of the precise graft locations were generated to assess the levels of TSPO present around the grafts. The segmentation of these ROIs were performed on the MRI data acquired during the brain surgery, in which the iPSC-mDA delivery vehicle provided enough contrast in the image to allow for segmentation. The extent of contrast provided was indicative of the volume within the putamen that the iPSC-mDAs occupied.
Increased [18F]FEPPA SUV correlated with higher CD68 ratings, suggesting that increased TSPO expression measured by PET corresponds to increased recruitment of activated microglia due to the allogeneic nature of the grafts. Numerous mCherry positive rhesus iPSC-mDAs successfully survived 24 months after brain surgery without immunosuppression. The grafted cells were localized in the targeted putamen areas, forming tight clusters with short neurite extensions (a full description of the grafts is out of the scope of this paper and will be published elsewhere). These results differ from our previously reported poor survival of human-derived embryonic SC-mDAs in cyclosporine-immunosuppressed PD monkeys . In that study, three months after brain surgery, the xenograft recipients presented signs of neuroimmune rejection observed as copious amount of CD68-ir, with activated microglia in areas of necrosis and spread across the basal ganglia.
In the current study, the CD68 immunostaining was blindly analyzed using a semi-quantitative rating scale for microglial assessment. We chose this method because the inflammatory reaction was minimal and attempts to obtain optical density standardized measures of CD68 expression in the putamen were not sensitive enough to detect the subtle differences in cellular morphology and number. Since the CD68 ratings considered the accumulation of microglia with activated morphology, these results further suggest that [18F]FEPPA signal in these regions correspond to the presence of activated microglia rather than just an increase in microglial density. The neuroinflammation associated with the allogeneic grafts was subtle and mostly localized, yet [18F]FEPPA PET demonstrated sufficient sensitivity to detect it. An interesting finding was the discordance between R5’s high CD68-ir score and the relatively low [18F]FEPPA uptake. It should be noted the intense CD68-ir in this subject was distributed in a very narrow band (Fig. 1b) compared to, for example, the large tear drop clusters present in R4 (Fig. 4), suggesting that [18F]FEPPA uptake may be affected by the pattern of microglia accumulation.
The correlation between TSPO PET and CD68-ir is not unique to this study, as English et al (2014) showed that increased uptake of [11C]PBR28 corresponded with a higher CD68 rating in a rat model of aortic aneurysm . Hannestad et al (2012) reported that baboons treated systemically with E. coli lipopolysaccharide presented an increase in [11C]PBR28 uptake that positively correlated with serum cytokines IL-1β and IL-6 levels. This finding was associated with increased TSPO-ir that was mainly present in activated CD68 positive microglia . These studies provide further evidence that the increase in [18F]FEPPA uptake observed in the putamen with allogeneic grafts is a result of increased microglial activation.
With iPSC-derived lines taking center stage in translational studies aiming to replace cells lost to neurodegeneration, needs have emerged for in vivo methods to assess the host’s immune response against grafted cells. Kikuchi et al (2017) reported that human iPSC-derived dopaminergic progenitors survived and induced functional recovery in FK506-immunosuppressed monkeys with MPTP-induced parkinsonism . When evaluating the immune response to these cells via [11C]PK11195 PET, the authors described either no inflammation or mild inflammation in grafted areas . Additionally, use of S-[11C]KTP-Me PET to monitor cyclooxygenase-1 (COX-1) revealed no change in uptake, suggesting absence of microglial response. The postmortem analysis found none to minimal CD45 hematopoietic T cell immunoreactivity, but abundant MHC II expressing microglia, indicating the lack of sensitivity of [11C]PK11195, as well as S-[11C]KTP-Me to detect recruitment of microglia and its metabolic activity . In that regard, it has been reported that [11C]PK11195 has low brain uptake in both humans and nonhuman primates [38, 52], low sensitivity to detect TSPO, and high lipophilicity, resulting in nonspecific binding to lipids in the brain which can interfere with PET quantification . Compared to second generation TSPO radioligands such as [11C]PBR28, [11C]PK11195 specific binding in the brain is 80-fold lower in rhesus macaques . Our study has demonstrated that the second generation TSPO radioligand [18F]FEPPA has the sensitivity to detect changes in neuroinflammation following grafting of allogeneic iPSC-mDA while showing high uptake in the brain, holding promise for clinical applications in monitoring immune response.