Advances in CNS PET: the state-of-the-art for new imaging targets for pathophysiology and drug development

An obvious requirement for a CNS PET tracer is the accumulation within the CNS; however, achieving this in tracer design is non-trivial. Multiple factors play a key role in determining the success or failure of a tracer in this facet. Some published ‘rule of thumb’ criteria for passive diffusion into the brain are highlighted in Box 1 [11, 12]. High-molecular-weight compounds often struggle to cross the tight junction in the blood-brain barrier leading to no or very slow accumulation within the brain and rendering them unsuitable for PET imaging [11]. The lipophilicity of a compound is essential for accumulation and availability within the brain. This is often determined from the partition coefficient between octanol and aqueous phases at physiological pH, quantified as the LogD_7.4. If the LogD_7.4 is too low, then the tracer will be unable to passively cross lipid membranes preventing accumulation in the brain, unless there is active transport. However, if it is too high, the compound will preferentially remain within lipid bilayers, increasing non-specific binding and decreasing the availability and dynamic range.

Other parameters such as charge and polarity play a part in the lipophilicity, but have also been linked to increasing susceptibility for being efflux transporter substrates [13, 14]. Efflux transporters are responsible for the inability of a large proportion of drugs and pharmaceuticals to accumulate in the brain, shuttling the compounds back into the bloodstream too fast to allow accumulation [11]. These efflux transporters, which include P-glycoprotein (P-gp), multidrug resistance–associated protein (MDR), and breast cancer–resistant protein (BCRP), vary considerably between species and often render substrates useless for CNS applications [15].

The ability to cross the BBB is an essential criterion for all CNS PET tracers, but the pharmacokinetics and selectivity ultimately determine a PET tracer’s usefulness. Factors impeding or reducing its ability to accurately report on its target can severely limit its applicability or render it unusable. Some of these key parameters are listed in Box 2.

While an ideal PET tracer would be outstanding in all of the criteria listed in Box 2, in practice, this is often unachievable. Fully characterising the limitations of a tracer allows informed decisions on its applicability to answer the proposed question, and importantly when it is not. For example, consider a hypothetical tracer which has overall good properties, but shows additional high off-target binding within the brain. In brain areas where both target and off-target sites are present, this will often render the tracer unable to answer the desired question. However, in brain regions with low or no off-target site, it may be possible to accurately quantify the target and gain an accurate answer. Therefore, if the off-target site is known and well-characterised, then a tracer with off-target binding in some brain regions may, nevertheless, be useful for studies where the focus is the regions with low off-target binding.

To effectively interpret the results presented in this review, an understanding of the various outcome parameters used in these studies is necessary. Below is an overview of some of the most common outcome parameters used in PET studies, both pre-clinically and in-human, summarised in (Table 1). For a comprehensive review, see reference [11] or refer to the Turku PET centre website (http://​www.​turkupetcentre.​net).

SUV_target is the standardised uptake value of the target region, SUV_other is the standardised uptake value of other regions, C_T is the concentration of PET tracer in tissue, C_P is the concentration of PET tracer in plasma, V_{T (tissue)} is the tissue volume of distribution, V_{T (ND)} is the non-displaceable volume of distribution, BP_{ND (baseline)} is the non-displaceable binding potential at baseline, BP_{ND (drug)} is the non-displaceable binding potential at after drug administration, OP_test is the outcome parameter measurement from an initial test scan, OP_retest is the outcome parameter measurement from a repeated scan, SD is standard deviation, BSMSS is between subject mean sum of squares, WSMSS is the within-subject mean sum squares, K, in this case, is the number of repeated observations.

The most simplistic outcome measures quote the proportion of radiotracer in the target region at a designated time, such as the percentage injected dose per gram of tissue (%ID) or the injected dose corrected for subject weight, standardised uptake value (SUV). Ratios of uptake between areas (SUVr) provide easy to obtain and useful outcome parameters in early characterisation of a PET tracer and can be justified in human studies when assumptions can be made regarding constant radiotracer delivery and brain non-displaceable binding. However, for many tracers and applications in humans, they do not provide sufficient characterisation of the target to be useful. A useful parameter is the ratio of tracer in a target region in comparison with the tracer in the blood plasma. For reversible tracers, this ratio becomes constant at equilibrium and is quoted as the volume of distribution, V_T [16]. V_T is commonly used as an outcome parameter in human studies of reversible tracers. As a parameter, V_T is a measure of both specific (displaceable) and non-specific (non-displaceable) signal and, as such, can be insensitive to change or differences, especially if background signal is relatively high (where a large change in target availability/density may only cause small changes in observed V_T). Additionally, to calculate V_T, blood sampling, and metabolite correction is required, increasing the time, effort, and invasiveness of PET procedures. Correcting for background non-specific binding (non-displaceable binding seen upon blocking the target, denoted ND) across the brain allows a more sensitive measure of target alterations.

Often for a given tracer, regions of the brain contain no or negligible quantities of the target protein and represent only non-specific binding. This region can thus serve as a reference region to account for non-specific binding in the region of interest. As non-displaceable binding is generally assumed to be constant across the brain, the very useful ratio of V_T/V_ND (DVR) can be easily obtained when a reference region is present. The SUVr between the region of interest and the reference region at equilibrium gives DVR (and BP_ND) without the necessity of plasma input methods (as plasma component cancels out) [16]. However, this calculation should always be initially validated against full plasma input methods in humans to determine suitability and bias. DVR and BP_ND are parameters which are more sensitive to alteration of target availability/density for reversible tracers than other outcomes, with a decrease of 100% BP_ND representing full block of the target. As DVR or BP_ND are intrabrain comparison outcomes, they give no information on the overall brain uptake of a tracer.

An important parameter to consider in the design of studies is the variability of the measurement, as this influences the sample size required to sufficiently power a study. Variability can be considered as within-subject (measured by test-retest value (TRV)) and between-subjects, which is often expressed as the coefficient of variation (COV). For tracers where these values are high, delineating small alterations in outcome parameters becomes increasingly more difficult as measurement variability obscures effects. As a rule of thumb, alterations in outcome parameters similar or less than inherent variability will not be accurately quantified on small-scale studies. The intraclass correlation coefficient (ICC) is a measure of reliability comparing intra- and inter-subject variability.

Lastly, it is important to remember that multiple factors may contribute to the alteration of an outcome parameter in vivo. Some of these include differences in target expression, alteration in target affinity (i.e. high or low-affinity states), internalisation of a target, and changes in endogenous occupancy [17].

The list of tracer and target systems was based on the national institute of mental health research priorities list (https://​www.​nimh.​nih.​gov/​research-priorities/​therapeutics/​cns-radiotracer-table.​shtml) and supplemented by hand-searching of references, including of recent review articles [18, 19]. PET tracers with the first-in-human peer-review publication between 2013 and 2018 were considered. This was refined with further inclusion criterion of PET tracers for targets that had not previously been imaged in man, or where the tracer(s) published within this timeframe represented a ‘significant advancement’ over those published prior to 2013. A significant advancement was defined as the potential of a tracer to answer questions about the targets that were previously unobtainable or with much greater accuracy. Additionally, PET tracers with first-in-human studies reported at conference between 2013 and 2018, but without associated first-in-human peer-review publication, were also included if other PET tracers for that target met the inclusion criterion of the article.

PubMed literature search terms including (CNS and PET), (brain and PET and first in human), (‘target’ and PET), and (‘target’ and imaging) were used to identify tracers translated into human. Tracers meeting selectivity criteria were systematically reviewed via PubMed literature search of all articles containing the [‘tracer’] term, including previous names and isotopologues, including non-radioactive molecule. The published literature for each in-human tracer was collated and compared with other tracers using the evaluation criteria outlined below.

In-human tracers were assessed via two categories of criteria: the selectivity of the tracer in vivo and its pharmacokinetic profile. Selectivity was primarily assessed from in vivo blocking and occupancy data. In vitro techniques were also considered, especially in cases where specific off-target sites were investigated. Where a known off-target specific binding site was found for a tracer, discussion of the potential impact on quantification of the desired target is also conducted.

In vitro studies were deemed insufficient to extrapolate to proof of in vivo selectivity. Self-blocking and structurally dissimilar heterologous blocking with selective agents in vivo allow assessment of total specific binding and total selective binding, respectively, at full occupancy. Tracers were deemed to have proven high specificity or high selectivity if the outcome parameter approached saturation value upon relevant blocking experiment. At full occupancy, theoretical alteration of BP_ND=0, DVR=1, SUVr=1, RO=100%. For situations where full occupancy is unobtainable, i.e. due to toxicity, Lassen plots can provide a suitable alternative; however, they become less accurate at lower occupancies [20]. For parameters without correction for non-displaceable binding, such as V_T and SUV, the magnitude of decrease depends on the proportion of non-displaceable binding present and therefore is expected to show lower relative alterations than other outcome parameters.

Studies which specifically investigate off-target binding were also investigated and potential impact on tracer assessed. Compounds which bind to sites other than the target site can be used in blocking studies to determine if the tracer also binds to these off-target sites. Additionally, self-blocking experiments where a reduction in signal is observed in regions where no specific signal is expected (i.e. in brain regions where no target is present, in target knockout models, or in healthy controls (HC) tissues not expressing the target of interest) highlight areas of off-target specific binding which can perturb target quantification.

The pharmacokinetic profile of a tracer encompasses multiple parameters essential for tracer performance in vivo. The ability to efficiently cross the BBB is fundamental to a CNS PET tracers’ success. Outcome measures related specifically to brain signal, such as SUV or V_T, were assessed for evidence of this. Further observation of a PET tracers’ regional brain distribution provides circumstantial evidence of selectivity when correlating with known distribution of the target.

The accuracy and reliability of modelling techniques to produce outcome parameters depend partially on the kinetics of a tracer. Slow kinetics generally requires longer scan times and produces more variable outcome measures, reducing the usefulness of a tracer.

The dynamic range of a tracer is the proportion of signal alteration that can occur under a perturbed system (i.e. during an occupancy study or altered expression in disease state). A high dynamic range allows smaller alterations in a system to be accurately detected, thus increasing sensitivity. Tracers with low dynamic range (for example due to high non-specific binding) may not be able to accurately determine even large alterations in target availability, rendering the tracer incapable of quantifying target accurately.

The presence of a reference region, an area of the brain with no or very low specific signal, allows simplified calculation of outcome parameters such as DVR and BP_ND without invasive arterial input functions. This allows a simplified scanning procedure, reducing invasiveness and potentially improving the accuracy of outcome parameters. A reference region was deemed validated if there was evidence of a strong correlation between the outcome parameters calculated from full arterial input function and reference tissue methods. Any evidence of consistent bias of reference region models compared to plasma input methods was reported (Tables 2, 3, and 4).

The metabolism of a tracer can play a key role in the success or failure of a tracer. Rapid metabolism in vivo reduces the availability of tracer to bind to the target, reducing signal magnitude. Additionally, the presence of radiometabolites within the brain can have a huge impact on the quantification of the desired target. Evidence relating to rapid metabolism, radiometabolite formation within the brain, or peripheral formation of brain penetrant metabolites is discussed along with potential impact on tracer performance.

Intra- and inter-subject variability was assessed from published in human data quoting TRV and ICC or COV, respectively. Higher variability reduces the usefulness of a PET tracer, requiring larger sample size studies to delineate alterations between populations. Therefore, the clinical utility of tracers with moderate to high variability is more limited than those with low variability. TRV <10%, COV <10%, and ICC >0.8 were deemed as low variability and a high intraclass correlation coefficient, respectively.

Lastly, the radiochemical parameters of the PET tracer can also have an impact on outcome parameters. For human studies, it is assumed that high radiochemical purity is a minimum requirement and is not discussed within. However, the ratio of radioactive tracer to non-radioactive isotopologue can vary substantially between tracers and individual syntheses. The common measurements of this are either specific activity, with units of GBq/μg or molar activity GBq/μmol [153]. As a directly comparable term between tracers and targets, molar activity is used throughout this article. Low molar activity can result in high injected mass of non-radioactive compound and can cause non-negligible self-blocking, with a magnitude dependant on tracer and target.

The CNS targets are discussed in three sections: proteinopathies, which focusses on imaging of misfolded protein aggregates; receptors and transporter proteins; and enzymatic targets. The evidence for tracer selectivity and pharmacokinetic parameters is summarised in tables at the beginning of each section along with a short summary of the current advantages and disadvantages of each tracer. Within this table blocking study, results are quoted from the region of highest alteration, at the largest target occupancy dose. Studies with highly structurally related blocking agents are included within homologous blocking studies due to the high likelihood of displacing the tracer from all specific binding sites. Reference regions listed have been validated in humans unless otherwise stated.

Each target is introduced and its relevance in human disease state is summarised. The in-human tracers for the target are discussed collectively, outlining the evidence for selectivity followed by pharmacokinetic suitability. Where tracer limitations or lacking evidence is apparent, this is also highlighted within these sections. When available, results from in-human disease states are also succinctly summarised. Overall evaluation of the target and the available tracers’ applicability for imaging it is summarised in the final discussion section of each target.

¹⁸F-AV-1451, ¹⁸F-THK5351, ¹⁸F-MK-6240, and ¹⁸F-RO-948 all showed rapid brain delivery and fast kinetics suitable for imaging, with ¹⁸F-THK5351 having the fastest washout from the cerebellum of the THK series [21, 47, 55, 58]. Where results have been published, inter- and intra-subject variability reported is low (Table 2), meaning the repeatability of the outcome parameter(s) of these tracers within patients is high, and the differences in uptake across HC subjects is low. Additionally, the cerebellum appears a suitable reference region for tau imaging and is validated as such for multiple tau tracers (Table 2).

Retention in areas of the brain not expected to contain tau or in HC can indicate off-target binding and perturb quantification of tau in vivo. ¹⁸F-AV-1451 shows sites of high uptake in HC with the most prominent being the basal ganglia, mid-brain, and choroid plexus [163]. Initial compounds of the THK series ¹⁸F-THK523 and ¹⁸F-THK5117 showed high white matter retention in HC [40, 42], but this was progressively improved over the series, through ¹⁸F-THK5317 and finally ¹⁸F-THK5351 [45, 47]. For ¹⁸F-MK-6240, the ethmoid sinus, clivus, meninges, and substantia nigra had increased uptake, outlining these as sites of off-target binding, as well as some skull uptake, indicative of de-fluorination [55]. Sites with reported off-target binding of ¹⁸F-RO-948 include the substantia nigra, cerebellar vermis, meninges, and in the retina [58]. The presence of tracer retention in brain areas of HC raises issues of tracer quantification. A high off-target signal will reduce the relative dynamic range of the tracer in that area and may differ in magnitude between individuals or patient groups, making correction for off-target binding difficult. In extreme cases, signal from areas adjacent to sites of high off-target binding may be perturbed due to partial volume effect. Therefore, for example, white matter retention is a substantial barrier to quantification due to widespread distribution and variability between patients. However, for areas removed from expected distribution and spread of tau, such as the retina, or regions of uptake which remain constant across study groups may not present substantial quantification issues.

Several additional tau tracers have been translated into humans and presented at conference including ¹⁸F-GTP-1 [61], ¹⁸F-PI-2620 [169] (replacing weaker candidate ¹⁸F-MNI-815, which has also been translated into humans [60]), ¹⁸F-AM-PBB3, and ¹⁸F-PM-PBB3 (also known as ¹⁸F-MNI-958) (Table 2) [62]. The initial presented data from these tracers appears promising. However, no in vivo data have been published in peer-reviewed journals, preventing objective comparison with the more established tracers. Therefore, the field eagerly awaits the emergence of clinical trial data and associated publications.

The large number of recent tracers for tau highlights the impetus associated with the development and translation of tau PET probes within the medical imaging community. The ability to distinguish AD patients from HC and people with MCI on the basis of tau load has been shown in multiple studies. Expansion of tau imaging into other disease states and clinical populations is well underway. Over 60 ‘PET + tau’ clinical studies are currently active or recruiting to investigate tau load in many disease states including dementias, motor neuron diseases, traumatic brain injury, and depression, amongst others (clinicaltrails.​gov, data obtained 30 October 2018). There are currently no promising in-human tau-based therapeutics. The ability to track tau load within human subjects is of huge importance for tau-based drug development programmes. Assessment of an anti-tau drugs effect on tau load in vivo, and direct comparison to cognitive performance/disease progression, would provide invaluable information of a drug’s efficacy and the merits of tau reduction as a therapy in humans.

Of all in-human tau tracers, ¹⁸F-MK-6240 is currently the most promising. The selectivity profile, to date, is the most robust, with minimal off-target specific binding apparent in NHP, and strong correlation to cognitive scores in AD reported. However, for this, and all tau tracers published, there are still many open questions to be addressed. In the following section, we discuss the issues that would be useful to address.

The lack of in vivo blocking data represents a major drawback for the confidence of selectivity for ¹⁸F-FEOBV in vivo. In rodents, dose escalation of homologous blocking studies was deemed unethical due to adverse effects of blocking VAChT in vivo; therefore, only a partial self-block was achieved [66]. No heterologous blocking studies have been reported for ¹⁸F-FEOBV in vivo.

In contrast, large displacement of ¹⁸F-VAT was observed upon administration of vesamicol in NHP (Table 3) [69, 70]. Vesamicol is not selective for VAChT with well-known binding to sigma receptors [71] and therefore represents a non-selective block. The in vitro characterisation of ¹⁸F-VAT determined high selectivity of VAT over sigma receptors [69], but this cannot be assumed to translate into in vivo selectivity of a radiotracer, as discussed with ¹⁸F-AV-1451 above. Nevertheless, substantial binding to sigma seems unlikely as no decrease in SUV was observed in the cerebellum upon blockade, where sigma receptors are prevalent, as well as showing contradictory distribution [190].

For imaging α7-nAChR with ¹⁸F-ASEM, substantial reduction in uptake upon heterologous α7-nAChR specific block, with DXMB-A and SSR180711 in rodent and NHP respectively has been shown [74, 191], and no response to multiple negative control blocking studies in rodents [74]. These studies imply high selectivity and dynamic range of ¹⁸F-ASEM for α7-nAChR in vivo.

¹⁸F-FEOBV, ¹⁸F-VAT, and ¹⁸F-ASEM all show good peak brain activity, heterogeneous distribution in line with target distribution, and moderate to fast kinetics in most brain regions [65, 73]. The exception is the slow kinetics of ¹⁸F-FEOBV in basal ganglia structures requiring long scan durations or delayed scanning protocols for quantification in this region [65]. Outcome parameter variability interpatient was moderate for ¹⁸F-FEOBV and ¹⁸F-ASEM, as well as moderate intrapatient variability reported for ¹⁸F-ASEM (Table 3). Peer-reviewed in-human, ¹⁸F-VAT data is not yet available.

As α7-nACh is present across the entire brain, there is no available reference region for ¹⁸F-ASEM, and therefore, full plasma input methods will be required for accurate quantification. For ¹⁸F-FEOBV, the cerebellar grey matter has been validated as a suitable reference region and provided lower variability than arterial input measures, with high correlation, however, had a lower dynamic range [65]. In NHP, the cerebellum may also provide a reference region for ¹⁸F-VAT quantification [70].

¹⁸F-FEOBV and ¹⁸F-ASEM both show pharmacokinetic profiles suitable for imaging and have been successfully utilised in disease state imaging. ¹⁸F-FEOBV was shown to have the highest sensitivity compared with both ¹⁸F-FDG (metabolism) and ¹⁸F-NAV4694 (amyloid) for distinguishing between HC and AD patients in a small-scale study [67]. ¹⁸F-ASEM V_T was found to be significantly decreased in schizophrenia patients in some brain regions, although an outlier was excluded in order to achieve this [75]. Additionally, an occupancy of up to 49% was determined when schizophrenic patients were treated with α7-nAChR selective agonist DXMB-A. The significant alteration observed with these tracers in disease state clinical trials shows the prospective applicability of these cholinergic targets in disease state imaging.

The lack of reference region available for α7-nACh increases the practical complexity of scanning with ¹⁸F-ASEM as quantification will require full arterial input function. The slow kinetics of ¹⁸F-FEOBV in the basal ganglia structures may result in higher demands on equipment for quantification in these regions, due to long scan duration. The in vivo selectivity data for ¹⁸F-ASEM appears comprehensive and robust, whereas further work is required for ¹⁸F-FEOBV.

Blocking experiments in NHP with two A_2A antagonists, preladenant (structurally similar to ¹⁸F-MNI-444) and tozadenant (structurally dissimilar to ¹⁸F-MNI-444), showed a dose-responsive decrease in SUV, with TAC in striatal regions similar to that of the cerebellum at maximum dose for both blocking agents (RO ≈100%) [77]. This represents approximately full blockade of specific signal in these regions providing strong evidence of high A_2A selectivity. Preladenant was treated as a homologous block due to the high structural similarity to ¹⁸F-MNI-444.

In both NHP and human brains, ¹⁸F-MNI-444 distribution is heterogeneous and matched that of known A_2A receptor distribution [76, 77], with high retention in the striatum and fast washout in the cerebellum, where the concentration of A_2A is very low [211]. A limitation is the relatively slow kinetics which may require long scan durations. In blocking studies, a small (<15%), non-dose-responsive decrease in cerebellar activity was observed, making this a non-perfect reference region. However, comparison between arterial input function and reference region analysis methods showed a high correlation, indicating cerebellum may be able to be used as a reference region in both human and NHP, with minimal error [76, 77]. Utilising this method, high intrapatient repeatability was observed [76].

There are currently no published results in patients with ¹⁸F-MNI-444. However, it appears superior for imaging A_2A in the CNS to the other evaluated tracers, with a robust selectivity profile and suitable pharmacokinetics for imaging A_2A. Further studies with ¹⁸F-MNI-444 are currently underway and will help determine its full potential.

Both ¹⁸F-UCB-H and ¹¹C-UCB-J show a substantial reduction in V_T upon heterologous block experiments in rat and NHP, respectively, with similar reduction observed with homologous blocking of ¹¹C-UCB-J (Table 3). Additionally, screening of ‘cold’ UCB-H and UCB-J showed no activity (<50% effect or inhibition at 10 μM) across a wide range of brain receptors, transporters, enzymes, and ion channels in vitro [79, 83]. This data supports a high degree of SV2A-specific binding and large dynamic range for both ¹⁸F-UCB-H and ¹¹C-UCB-J in vivo.

Due to the near-ubiquitous nature of SV2A across the brain, no suitable reference region is clearly established, although the centrum semiovale is being evaluated as a pseudo-reference region [217]. Both ¹¹C-UCB-J and ¹⁸F-UCB-H have fast kinetics. However, ¹¹C-UCB-J has a higher dynamic range and significantly higher BP_ND in NHPs and human and has calculated target density (B_max) and K_d closely matching that of ex vivo data [81, 82]. As such, ¹¹C-UCB-J has been pursued as the lead tracer in this series and showed high stability of V_T in HC test-retest scans, although ICC values were low indicating high interpatient variability (Table 3) [84].

Of the two tracers, ¹¹C-UCB-J appears to be the superior SV2A tracer, with higher affinity in vivo. However, for distribution purposes, the longer half-life of ¹⁸F will make ¹⁸F-UCB-H the option available to sites without an on-site cyclotron. One disease state imaging proof of concept study with the SV2A tracers has been carried out. Patients with medically refractory temporal lobe epilepsy showed higher levels of ¹¹C-UCB-J asymmetry (>50%) in the hippocampus compared with that usually observed for ¹⁸F-FDG (<20%), with controls showing little asymmetry [82]. Direct head-to-head studies on larger cohorts are required; however, this is a very promising initial study into the use of ¹¹C-UCB-J in epilepsy. Multiple investigations are ongoing into other diseases using SV2A B_max as an index for synaptic density. While a decrease in ¹¹C-UCB-J signal may represent a loss in synaptic density, it is important to remember that alterations in synaptic vesicle concentration or SV2A regulation or availability may also cause alterations in signal without necessarily correlating to synaptic density and will need to be explored.

Nevertheless, SV2A imaging is one of the most exciting new areas for CNS imaging with scope in many disease states. Clinical trials are currently active in AD and addiction (clinicaltials.​gov, accessed 07 November 2018) as well as schizophrenia [218].

Ex vivo and in vivo studies with, structurally dissimilar, I2BS selective ligand BU224 showed large decrease in SUV of ³H-BU99008 in rats and ¹¹C-BU99008 in NHP [86, 87]. The spatial colocalisation and reported affinity of imidazoline ligands for MAO identify this as a potential high-risk site for off-target binding [225]. No significant alteration was observed in NHP when treated with MAO-A and MAO-B ligands, suggesting off-target binding to MAO is negligible [87]. The majority of signal in NHP appears to be due to specific binding to I2BS.

Distribution of ¹¹C-BU99008 within the brain followed similar patterns throughout species with basal ganglia structures > cortex > cerebellum, in line with I2BS expression [85]. The kinetics of the tracer are relatively slow, requiring long scan durations in humans, with greater implications to image quality for a ¹¹C-labelled tracer due to the shorter half-life. Partial blockade with non-selective I2BS in humans reduced SUV across all brain regions showing lack of available reference region [85].

Test-retest variability in humans showed some discord with variation ranging from 5 to 25% across regions with higher variations found in high uptake regions [85]. The authors postulated the slow kinetics of ¹¹C-BU99008 and, therefore, high V_T could be a contributing factor. Another possible factor could be the relatively low and highly variable molar activity used (quoted as 35.3±17.5 GBq/μmol), with test scans having over twice the injected mass on average than re-test (3.8±3.2 μg and 1.8±1.1 μg, respectively) [85]. A study in rats investigated the effect of ¹¹C-BU99008 molar activity, finding significant increase (+28%) in hypothalamus SUV area under curve integrals using ultra-high molar activity ¹¹C-BU99008 (>5000 GBq/μmol) in comparison with standard molar activity samples (55–220 GBq/μmol) [226]. Therefore, the high variability test-retest, and higher average uptake in the retest scans, may be due in part to the molar activity of the radiotracer range used. It should be noted that the method used to produce ultra-high molar activity ¹¹C-BU99008 had an order of magnitude lower radiochemical yield than conventional methods.

¹¹C-BU99008 appears to be a selective tracer for I2BS with appropriate properties for imaging in vivo. However, further studies on the dynamics of the tracer, the biology of the target, and relevance in disease states are required. Factors which may improve the reproducibility of V_T measurements include higher (and more consistent) molar activity of ¹¹C-BU99008, a tracer with faster kinetics, or a ¹⁸F-labelled tracer.

¹¹C-ITMM and ¹⁸F-FIMX are the two PET tracers successfully translated into humans for mGluR1 and are structurally related. Pre-clinical characterisation provides strong evidence of selectivity for both tracers, with large decrease in signal and removal of brain heterogeneity upon both heterologous (JNJ-16259685 [237]) and homologous blocking protocols [89, 95]. Additional characterisation of ¹¹C-ITMM in mGluR1 KO mice showed substantial decrease in uptake compared with wild-type, although only whole-brain data of knockouts, rather than individual regions, were reported [89]. No decrease in uptake of ¹⁸F-FIMX upon administration of mGluR5 selective ligands was observed, showing a lack of specific binding to this site [95].

The robust pre-clinical characterisation of ¹¹C-ITMM and ¹⁸F-FIMX shows a high level of selectivity for mGluR1 in vivo and gives confidence in these tracers for subsequent studies.

Both ¹¹C-ITMM and ¹⁸F-FIMX show brain distribution in humans concordant with mGluR1 expression; however, pharmacokinetic profiles of the tracers differ substantially. ¹¹C-ITMM has relatively slow kinetics, requiring long scan times, relatively low brain uptake, and slow metabolism [88]. In contrast, ¹⁸F-FIMX has higher brain uptake, fast kinetics, and much faster decrease in plasma parent fraction [94]. Currently, neither tracer has published TRV, although for ¹⁸F-FIMX, it has been outlined as an outcome of a current clinical trial (clinicaltrials.​gov, trial identifier: NCT02230592). The pons and the medulla are areas of low mGluR1 expression and have been proposed for use in reference region models [93], although no reference region has been validated in humans to date.

An ultra-high molar activity study with ¹¹C-ITMM in rodents saw a significant increase in SUV and V_T compared with standard molar activity, highlighting molar activity as a potential risk for introducing variability [238]. A study investigating the relationship of ¹¹C-ITMM V_T with age and gender reported increases in V_T in older controls [92]. However, due to significantly higher molar activity of tracer in the older population (average increase of 50% over young controls, P<0.01) and significantly lower injected activity, the V_T alterations quoted in the paper could be influenced by the lower injected dose of ITMM in older HC, in line with the pre-clinical work.

¹¹C-ITMM and ¹⁸F-FIMX have shown excellent performance in pre-clinical characterisation studies for imaging of mGluR1. The higher uptake, faster kinetics, and longer half-life of ¹⁸F-FIMX suggest it will provide higher quality images and better quantification of mGluR1 in patients over ¹¹C-ITMM, although further studies are required to verify this. Significant alterations of ¹¹C-ITMM in ultra-high molar activity rodent studies highlight molar activity as a potential confound in clinical studies.

So far, ¹¹C-ITMM has shown promise for cerebellar ataxia, where it has been shown to have a larger dynamic range than MRI methods and potentially more sensitive than ¹⁸F-FDG [90, 93]. However, no other disease states have been investigated with either tracer. Given the current interest in the glutamatergic system, studies are expected to emerge on a variety of disease states over the next few years.

Both ¹¹C-GR103545 and ¹¹C-LY2795050 showed promising in vitro selectivity over μ- and δ-OR [99, 248]. As an agonist tracer for κ-OR, ¹¹C-GR103545 is expected to have low dose tolerance before negative pharmacological effects are observed. As such, self-blocking protocols are very limited, with changes in V_T values obtained in NHP too low for an accurate Lassen plot to be constructed [97]. Brain distribution of ¹¹C-GR103545 was reported to align with expected κ-OR density in NHP [96]. Blocking with naltrexone showed substantial decreases in V_T in human studies; however, naltrexone is known to bind to κ-, μ-, and δ-OR [99, 249].

¹¹C-LY2795050 is an antagonist tracer for κ-OR. In vivo blocking studies showed good response to both homologous block and heterologous block, with the non-selective opioid ligand naloxone reducing BP_ND to approximately 0 in NHP, showing approximately all specific binding signal is opioid-related [99]. Further studies in NHP were conducted where non-radioactive LY2795050 was used to block the signal of ¹¹C-LY2795050 and μ-OR tracer ¹¹C-carfentanil to determine the selectivity of LY2795050 for κ-OR over μ-OR. The in vivo results showed 7.6 fold selectivity for κ-OR over μ-OR, much lower than the reported in vitro selectivity of 36 fold [102].

In humans, the kinetics of ¹¹C-GR103545 are slow, leading to complications in the calculation of kinetics and outcome parameters, compounded by the use of ¹¹C. This will most likely be a substantial factor in the moderate-high degree of variability observed, with TRV on average 41% for the highest uptake region [96].

For ¹¹C-LY2795050, the in-human pharmacokinetic properties observed are good. It had relatively high initial brain uptake and fast kinetics suitable for accurate kinetic modelling, although no reference region is available in humans [98]. Occupancy studies largely removed regional uptake differences showing almost full occupancy of receptor [98], and the test-retest repeatability was good [101].

¹¹C-GR103545 suffers from a number of drawbacks including lack of in vivo selectivity studies, due to target related toxicity and poor pharmacokinetics in humans. The binding affinity of ¹¹C-GR103545 determined in vitro was very high, with a K_i of 0.02 nM [248]. As discussed by the authors of this work, a lower affinity derivative may have faster kinetics and be more suitable for imaging in vivo, highlighting that too high affinity can hinder kinetic analysis.

The moderate selectivity of ¹¹C-LY2795050 for κ-OR over μ-OR in vivo is a manageable limitation of this tracer. In vivo, the vast majority of signal is expected to arise from κ-OR due to the moderate selectivity and higher natural abundance of κ-OR, especially in regions of high κ-OR to μ-OR density. ¹¹C-LY2795050 appears the more promising of the two tracers due to its larger dynamic range of occupancy reporting and lower variability. However, experiments must be carefully designed in order to reduce, or control for, potential perturbation of signal by μ-OR alterations. There is therefore a scope for more selective tracers to make an impact on this field.

In vitro characterisation showed high binding affinity to all three 5-HT₂R subtypes (A–C) [258]. In NHP, selective blocking of 5-HT_2CR was achieved with SB 242084 and showed approximately 100% decrease in BP_ND in areas of high 5-HT_2CR density such as the choroid plexus [104]. In both human and NHP blocking, studies with 5-HT₂R antagonist ketanserin (non-selective between 5-HT₂ subcategories A–C) showed significant reduction across the brain (Table 3) [103, 104]. Pre-clinical studies have shown a significant response to pharmacologically increased and decreased levels of serotonin in vivo in NHP and rodents respectively showing the potential sensitivity of ¹¹C-Cimbi-36 to endogenous neurotransmitter concentration [257, 259].

In humans, the uptake of ¹¹C-Cimbi-36 was relatively high in cortical regions although the kinetics were moderately slow. Metabolism of the tracer was relatively fast; however, the reported metabolites were more polar than ¹¹C-Cimbi-36 and are not likely to cross the BBB. The lack of signal alteration upon blocking in the cerebellum highlights this as a suitable reference region, although a consistent negative bias was found when using this, and the human test-retest showed very low variability across brain regions [104].

¹¹C-Cimbi-36 is the first 5-HT₂R agonist PET tracer translated into man with pharmacokinetics acceptable for in-human imaging. While it appears not to be selective for 5-HT₂R subtypes, the distribution of these within the brain is substantially different. 5-HT_2BR is expressed in low concentrations in the human brain and is therefore unlikely to significantly contribute to observed signal [260]. For 5-HT_2AR and 5-HT_2CR, regional distribution differences may allow selective determination of subtype signal due to some areas of high 5-HT_2AR being almost devoid of 5-HT_2CR and vice-versa (i.e. cortical regions and choroid plexus, respectively). Therefore, in these regions, ketanserin can be treated as a pseudo-selective block in HC. However, caution must be expressed when interpreting results from regions of subtype colocalisation or disease states.

Further studies comparing agonist and antagonist tracers will hopefully provide a more comprehensive report on the role the 5-HT₂ receptor plays in neuropsychiatric disorders.

Pre-clinical selectivity data on the current P2X7 tracers is limited. No small animal selectivity studies are published for ¹⁸F-JNJ-64413739; however, slides on pre-clinical data are available online (www.​nas.​edu). While the data available appears promising, in-depth review of this must wait until peer-review publications are available.

¹¹C-JNJ-54173717 and ¹¹C-SMW139 have both been characterised with a viral vector model which causes expression of human P2X7 in one brain hemisphere of rats, with control viral vector injected into the other. Both tracers showed increased uptake in the hemisphere with human P2X7 expressing viral vector and homogeneity upon blocking with P2X7 antagonists [107, 109]. For ¹¹C-JNJ-54173717, the block compound was structurally similar whereas for ¹¹C-SMW139, this was structurally dissimilar. Further studies in NHP with ¹¹C-JNJ-54173717 showed a substantial decrease in observed SUV upon self-block and blocking with chemically distinct P2X7 antagonist (JNJ-42253432) [109].

¹¹C-GSK1482160 has been characterised in a lipopolysaccharide model of inflammation where V_T was significantly increased upon treatment, with 97% of the increased signal displaced in lipopolysaccharide and homologous block–treated animals [111]. The in vivo selectivity profile for ¹¹C-GSK1482160 is currently lacking heterologous P2X7 blocking data. In vitro response to P2X7 expressing cells and correlation to expression of P2X7 and activated microglia in an ex vivo multiple sclerosis rodent model both increase confidence in selectivity [111, 112], but are not in themselves proof of P2X7 binding in vivo.

¹⁸F-JNJ-64413739, ¹¹C-JNJ-54173717, and ¹¹C-SMW139 appear to have relatively fast kinetics and high brain uptake suitable for imaging in the species investigated [106, 107, 109]. ¹¹C-GSK1482160, however, has slow kinetics in rodent and NHP making accurate quantification more difficult. In rodents, ¹¹C-SMW139 suffered from radiometabolites within the brain as well as limited dynamic range. Twenty-one percent of the brain radioactivity was observed as metabolites at 15 min, increasing to 34% by 45 min [107]. The presence of radiometabolites in the brain may be detrimental to the accurate quantification of P2X7 in human studies. ¹¹C-JNJ-54173717 studies found no evidence of radiometabolites in rodent brain; however, upon self-blocking in NHP, an increase in SUV towards the end of the scan was observed, possibly indicative of brain penetrant metabolites, although this increase was not present in other protocols [109].

Human pharmacokinetics of ¹⁸F-JNJ-64413739 showed moderate TRV and high ICC suitable for PET imaging studies, although high COV indicates intrapatient variability may be high [106]. No data is available for metabolites of ¹⁸F-JNJ-64413739.

The rapid translation of P2X7 tracers into humans signifies both academic and commercial interest in the area. Publication of pre-clinical and clinical data on these recently developed tracers will allow comparison and critical analysis and allow advancement of the quickly progressing P2X7 imaging field. For ¹⁸F-JNJ-64413739 and ¹¹C-GSK1482160, the lack of published, heterologous blocking studies is a limitation in confidence going forward. Additionally, brain penetrant metabolites may be a constraint to one or multiple of the discussed tracers.

Self-block and blocking with Me-KTP-ME (prodrug selective for COX-1 over COX-2) in NHP both showed similar decreases in V_T (≈−55%), compared with baseline scans [114]. A negative control blocking study with a COX-2 selective compound (MC1) showed no significant alterations in uptake. Additionally, in vitro screening found K_i constants of >10 μM for a wide range of human recombinant receptors, indicating potential selectivity against those tested [114]. To date, no in-human data has been published in peer-reviewed journals; however, initial scans are currently being carried out in a clinical trial (www.​clinicaltrial.​gov, trial identifier: NCT03324646). Some of the early scan data has been presented at the Society of Biological Psychiatry conference (2018) [113].

In NHP, ¹¹C-PS13 had the highest brain uptake and V_T values of the series tested, as well as suitable kinetic profile, although there appears to be no suitable reference region [114]. The proportion of non-specific binding in the NHP brain is relatively high and free fraction in plasma low, reducing the dynamic range of the tracer. Early data presented on human brain distribution appears similar to that observed in NHP, although more comprehensive analysis will have to wait until peer-reviewed publication [113].

The most apparent disadvantage of ¹¹C-PS13 is the high proportion of non-specific binding in the NHP brain. This is most likely related to the relatively high lipophilicity of ¹¹C-PS13 (LogD=4.3). Therefore, future tracers may reduce this with less lipophilic derivatives. However, the data presented so far, while limited, appears promising and may mark the first successful ligand for COX translated into humans.

Early pre-clinical characterisation of ¹⁸F-BCPP-EF showed moderate response upon dosing with selective mitochondrial complex 1 inhibitor, rotenone, in both rodents and NHP. However, due to the cardiac toxicity of rotenone, only very low doses were administered in vivo [116]. In vitro studies on rat brain slices allowed higher doses of rotenone and signal intensity of ¹⁸F-BCPP-EF was decreased by 96% upon maximum block, implying a larger dynamic range and greater selectivity than shown in vivo [118]. The moderate reduction of V_T in NHP upon blocking with rotenone may allow the estimation of V_ND and occupancy using a Lassen plot analysis, therefore giving greater information on dynamic range in vivo.

High uptake and fast kinetics were observed in rat and NHP brain. Due to the ubiquitous nature of mitochondrial complex 1, no reference region is available for this target. Implementation of ¹⁸F-BCPP-EF for use in humans has recently been undertaken and 8 HC scanned. Initial data presented at conference showed high brain uptake with V_T values around 30 in the striatum [115].

¹⁸F-BCPP-EF is the first tracer to measure the availability of any mitochondrial complex within the CNS and may also have potential as a reporter for mitochondrial density. Since its discovery, multiple preclinical studies have used ¹⁸F-BCPP-EF and shown significant alterations in mitochondrial complex 1 availability in models of ischemic injury and PD [119, 121], as well as showing negative correlations with age and amyloid load in NHP [120]. However, no in-human patient data are currently available. The toxicity of mitochondrial complex 1 inhibitors is a drawback of the field, as full in vivo blockade is unfeasible.

¹¹C-Martinostat is reported to bind with high affinity to HDAC 1, 2, and 3 and to a lesser extent 6, from in vitro assays. In vivo self-block of ¹¹C-Martinostat largely reduced V_T in NHP (Table 4), showing high level of specific binding in this species [123, 124]. In rodents, similar decreases in ‘% whole brain uptake normalised to uptake at 6 min’ was observed for both self-blocking at 2 mg/kg and pre-treatment of CN54, a structurally dissimilar HDAC inhibitor [123]. In vitro screening revealed that at 50 nM of Martinostat resulted in 24% inhibition of dopamine transporter, highlighting this as a potential off-target binding site. However, additional studies using dopamine transporter tracer ¹¹C-β-CFT showed no alteration of signal upon dosing with Martinostat (1 mg/kg) showing that Martinostat does not compete to measurable quantities with ¹¹C-β-CFT in vivo [123].

¹¹C-Martinostat has been shown in all investigated species to have high brain uptake, but slow kinetics [122‐124]. Upon self-blocking protocol in NHP, ¹¹C-Martinostat was decreased in all brain regions, highlighting the lack of reference region available [123]. In HC, intrapatient reproducibility via test-retest was high and the interpatient variability was moderately high for V_T measurements. Interpatient variability was improved using SUV_60-90 as an outcome parameter; however, to be applied to patient populations would require validation against arterial input methods. In a comparison with ¹¹C-Martinostat uptake in schizophrenic and schizoaffective cohorts compared with HC, SUVR was used as an outcome parameter. The activity in each region was referenced to the activity in the whole brain. This method showed significant alterations in schizophrenic patients in multiple regions; however, the use of SUVR in this way requires validation in patient populations with arterial input functions, as highlighted by the authors [125].

¹¹C-Martinostat represents the first HDAC tracer to be useable for brain imaging in humans. The pre-clinical selectivity data for this compound appears robust, with good response to blocking protocols. The slow kinetics of the tracer is a disadvantage; however, for the purposes of determining brain penetration and occupancy of HDAC-targeted drugs, this tracer appears suitable. From the two published in-human studies, two different outcome parameters were proposed in SUV_60-90 and SUVR (using the whole brain uptake as a reference region). Both of these methods would require further validation in patient populations. Standardisation of outcome parameter in future studies would allow direct comparison between results.

In NHP, brain distribution concordant with PDE2A expression was found, and occupancy studies in NHP with PDE2A inhibitor PF-05180999 showed a decrease in V_T of 32% and a maximum BP_ND reduction of 72%, in high-binding regions (Table 4) [127]. However, in lower uptake regions (including the cerebellum as the reference region), V_T increased upon blocking, with no hypothesis from the authors discussed [127]. Due to high structural similarities, PF-05180999 would be expected to act as a homologous block. In vitro screening of PF-05270430 indicates selectivity over other PDE and CNS targets; however, no heterologous blocking studies have been reported with structurally diverse PDE2A ligands in vivo, reducing confidence in selectivity.

Pre-clinical characterisation ¹⁸F-PF-05270430 showed high uptake, fast kinetics, and low TRV in NHP [127]. In rodents, low levels of non-parent brain metabolites (5–7%) were observed indicating potential for BBB penetrant metabolites in clinic [127]. The occupancy and decrease in V_T reported indicate that the V_T signal is predominantly non-displaceable binding, restricting its dynamic range. Moreover, the metabolism of ¹⁸F-PF-05270430 varied substantially between NHP subjects. The percentage of ¹⁸F-PF-05270430 in plasma was very variable, ranging from approximately 10 to 45% across the three NHP subjects at 120 min post-injection in baseline scans. The percentage of ¹⁸F-PF-05270430 in plasma for intra-animal metabolism rate remained constant under test-retest conditions and upon blocking [127].

In the only published human data on ¹⁸F-PF-05270430, similar brain distribution to pre-clinical studies was observed, suggesting binding to PDE2 and low test-retest variability across brain regions reported [126]. For this study, the cerebellum was used as a reference region, although additional studies are required to validate this. Both V_T and BP_ND were low, decreasing dynamic range and reducing signal-to-noise ratios [126].

From the data presented to date, ¹⁸F-PF-05270430 may be adequate for some applications; however, it appears non-optimal for imaging PDE2A in humans and is lacking critical selectivity data. Therefore, while ¹⁸F-PF-05270430 may represent the first PDE2A tracer to be translated into humans, there is the need to develop new, more specific tracers in this field.

The selectivity data for the in-human tracers of PDE10A (Table 4) generally show a high degree of selectivity across tracers, with the exception of ¹⁸F-MNI-654, for which no data are available, and ¹¹C-T-773 which shows substantial off-target, specific binding. Two blocking studies have been published with ¹¹C-T-773 in NHPs, both using female rhesus monkeys. The first used structurally dissimilar, PDE10A selective, MP-10 which showed a moderate decrease in striatal region V_T (caudate −36%, putamen −47%), indicating binding to PDE10A in these regions, with little effect on other regions investigated including the cerebellum [145]. The second used a structurally related drug candidate (TAK-063) which showed large decreases in V_T across the whole brain (approximately 75% on average) [144]. The large reduction of V_T in areas of negligible PDE10A concentration with TAK-063 that showed no response when blocked with PDE10A selective compound MP-10 strongly indicates that ¹¹C-T-773 has substantial off-target specific binding to the unknown site(s) across these regions. There is therefore no reference region for this tracer and interpretation of results is complicated by potential alteration in off-target site expression/availability.

¹⁸F-JNJ42259152 is a structural derivative of selective PDE10A ligand MP-10; studies using this inhibitor are expected to act as a self-block. In the mouse KO model used for ¹⁸F-JNJ42259152 characterisation, homogenous brain distribution was observed, giving confidence of a high level of selectivity in the mouse. All other tracers are structurally dissimilar to MP-10 and utilised this compound in selectivity studies across a number of species, showing promising results (Table 4). For ¹¹C-IMA107, structurally related compound IMA102 was used assumed to act as a self-block.

Kinetic profiles of the PDE10A PET tracers were generally reported to be suitable for PET imaging in human. ¹⁸F-JNJ42259152 showed similar brain distribution as pre-clinical models; however, V_T and BP_ND values were globally low with max V_T approximately 1.5 in the putamen and <1 for other regions [130]. Contrasting to the other PDE10A tracers, the frontal cortex was used as a reference region due to consistently showing the lowest uptake across brain regions. For all remaining tracers, the cerebellum was validated as an appropriate reference region (Table 4), with the exception of ¹¹C-T-773 for which no reference region is available.

¹⁸F-MNI-654 displayed slower kinetics, higher variability, and lower uptake than ¹⁸F-MNI-659 and was therefore deemed an inferior PDE10A tracer [132]. All other reported tracers showed low to moderate inter- and intrapatient variability where reported (Table 4).

From preclinical characterisation, two tracers have evidence of brain penetrant radiometabolites. ¹⁸F-JNJ42259152 showed high quantities of radiometabolite in brain tissue of rats (53% of cerebellum activity, 60 min), which could make quantification of this tracer difficult. However, the identified metabolite is not expected to bind to PDE10A and authors of human studies claim this does not impact the quantification of PDE10A with ¹⁸F-JNJ42259152 [130]. ¹¹C-Lu AE92686 underwent fast metabolism in NHP, with slow increase in cerebellum V_T observed throughout the scan, indicating a brain-penetrating metabolite may be present in small quantities [141]. The cerebellum V_T increase over time has not yet been reported in human, preventing the assessment of the possibility of brain-penetrating metabolite accumulation.

Research in HC with¹⁸F-MNI-695 have investigated mapping PDE10A, showing a significant negative correlation of striatal BP_ND with age, as well as increased BP_ND in women over men [135, 283]. ¹⁸F-MNI-659 has also been successfully applied to receptor occupancy studies [133, 135].

The first investigation into PDE10A in human disease was a pilot study with ¹⁸F-JNJ42259152. Significant decrease in BP_ND (alongside MRI volume) was observed in the putamen and caudate nucleus of HD patients compared with controls [131]. No significant correlation with disease severity was observed, although due to the small sample size (n=5), this is unsurprising. BP_ND images published did, however, hint at a trend towards significance.

In HD, ¹⁸F-MNI-659 BP_ND of the basal ganglia structures was around 50% of HC and showed a strong inverse correlation to both genetic burden (burden of pathology) and clinical disease rating (motor subscale score) [134]. Additionally, while both studies had small cohorts, and further studies are required, these results imply PDE10A is a major feature in HD pathology. They also support the applicability of ¹⁸F-MNI-659 for PDE10A imaging in HD.

In a cohort of 12 early premanifest HD patients, predicted several years until onset, ¹¹C-IMA107 BP_ND was significantly decreased in multiple areas, with the largest in the sensorimotor striatum (−33.0% caudate, −30.5% putamen) as well as substantial increase in motor thalamic nuclei (+34.5%) compared with HC [139]. Significant correlations between ratios of motor thalamic nuclei BP_ND/striatal BP_ND compared with predicted disease onset [139]. The authors claim this to be the strongest correlation with risk of symptomatic HD conversion of any technique, and the earliest in vivo pathophysiological mechanism identified. No comment was published on correlation directly between striatal BP_ND and risk of symptomatic HD conversion. The motor thalamic nucleus has BP_ND values substantially lower than close proximity structures of the basal ganglia. As a region, this may, therefore, be susceptible to partial volume correction errors and show substantial differences dependent on the analysis method used. Nevertheless, this study correlates well with that published for ¹⁸F-MNI-659, discussed above, with PDE10A emerging as a potential therapeutic target and ¹¹C -IMA107 as a suitable tool for tracking it in clinical trials. Longitudinal studies will, however, be required to assess the predictive power of PDE10A imaging in symptomatic conversion.

In 24 PD patients, a significant reduction in ¹¹C -IMA107 BP_ND was found in multiple basal ganglia regions, with the largest decreases observed in caudate (−28.4%) and putamen (−25.5%) compared with that in HC [138]. Significant inverse correlations were found between ¹¹C -IMA107 BP_ND, compared with disease duration and severity of motor symptoms [138]. As the only published PET study of PDE10A in PD patients, this work provides substantial justification for further investigations into the role of PDE10A in PD.

In a study investigating the role of PDE10A in schizophrenia, ¹¹C -IMA107 showed no significant difference between patients and HC in any brain region, with all average differences within the variability of the tracer from previous test-retest data [282]. In contrast, ¹¹C-Lu AE92686 showed significantly lower BP_ND in the putamen and caudate nucleus compared with that in HC, but no correlation to symptoms was found [142].

Multiple tracers appear suitable for imaging PDE10A in humans; therefore, the selection of lead tracer to use in studies is not straightforward.

No head-to-head studies have been carried out on the tracers in humans (except MNI-654 and MNI-659), so conclusions on the applicability of tracers are subjective. However, comparison of the HC data collected so far allows forerunners for clinical imaging purposes to be identified. Due to its slow kinetics and high variability, ¹⁸F-MNI-654 appears unsuitable for clinical imaging. ¹¹C-T-773 has the highest quoted V_T in HC, fastest kinetics, and low test-retest variability. However, the presence of an unknown off-target binding site, and the unsuitability of the cerebellum as a reference region, may restrict use in preference of other available tracers. ¹¹C-Lu AE92686 also shows high brain uptake and low variability in high PDE10A regions. The kinetics of ¹¹C-Lu AE92686 are slower than ideal and the indication of a possible brain-penetrating metabolite, confounding cerebellum V_T, is the drawback of this tracer. Of the remaining tracers, ¹⁸F-JNJ42259152, ¹⁸F-MNI-659, and ¹¹C-IMA107 all have appropriate kinetics for clinical imaging, have suitable test-retest, and have successfully been applied for disease state imaging. Of the three, ¹⁸F-MNI-659 has the highest V_T and BP_ND values; however, it currently has no self-bock in vivo data available. ¹⁸F-JNJ42259152 and ¹¹C-IMA107 both have convincing selectivity data from in vivo pre-clinical studies. The low brain uptake of ¹⁸F-JNJ42259152 results in low V_T values, potentially complicating quantification, although BP_ND values remain moderately high. The reverse is the case with ¹¹C-IMA107 which shows moderate V_T values but the lowest BP_ND values of the in-human tracers discussed. None of the in-human tracers is without flaws; however, multiple tracers appear suitable for clinical studies, with the current forerunners being ¹⁸F-JNJ42259152, ¹¹C-IMA107, and ¹⁸F-MNI-659.

Both FAAH tracers showed promising pre-clinical blocking profiles in pre-clinical models and humans. Ex vivo data from rats showed ¹¹C-CURB had regional heterogeneity in accordance with the known distribution of FAAH with high uptake in the cerebral cortex, cerebellum, and hippocampus, compared with that in the hypothalamus [147]. A high degree of specific binding was found in the regions investigated upon blocking with cold reference compound (also known as URB694) and the highly structurally related compound URB597 [147]. In humans, studies with the structurally dissimilar and selective FAAH inhibitor PF-04457845 showed an excellent response, decreasing the outcome parameter, λk₃ (see further explanation below), by over 90% and removing heterogeneity of this parameter across the brain, giving further evidence to the selectivity of the tracer [148].

¹¹C-MK-3168 is the other FAAH tracer to be evaluated in humans and binds reversibly. Pre-clinical data show good brain uptake and response to blocking with a compound of the same series, as well as high occupancy with the structurally dissimilar FAAH inhibitor JNJ-42165279 [150, 151]. Additionally, an occupancy study in humans, with JNJ-42165279, showed high FAAH occupancy, correlating with dose, and decreased the V_T of ¹¹C-MK-3168 by over 90% in multiple brain regions [150].

Both tracers showed high brain uptake and distributions in line with FAAH distribution and similar TRV. With the different binding pharmacokinetics, analysis methods and outcome parameters for the two tracers are substantially different. Modelling of ¹¹C-CURB revealed an irreversible two-tissue compartment model (2-TCMi) gave the most accurate fit, giving the outcome parameter of scans as λk₃ [146]. Here, k₃ is the rate constant into the irreversibly bound compartment and λ is K₁/k₂, equivalent to the equilibrium distribution volume of the ligand in the free and non-specifically bound compartment. ¹¹C-MK-3168 has the more intuitive outcome parameter of V_T measurements, but suffers from slow kinetics at baseline and rapid metabolism in humans (<5% parent after 30 min), with no published data on the nature of the metabolites. However, from the occupancy data reported, this does not appear to cause issue. Originally, white matter was proposed as a reference tissue for ¹¹C-MK-3168 but a decrease in white matter signal was associated with blocking (<15%) and poorer repeatability reported [152]. Additionally, no reference tissue method was used in the subsequent clinical trial employing ¹¹C-MK-3168 [150]. These observations suggest that white matter is not a suitable reference tissue for ¹¹C-MK-3168.

The binding of ¹¹C-CURB has been found to be linked to variation in a common nucleotide polymorphism on the FAAH gene (C385A, associated with increased risk of addiction); the λk₃ of ¹¹C-CURB was found to be significantly lower (−23%) in those with the A allele (A/C+A/A) than those with the C/C alleles [149]. While the authors suggest the change may represent a decrease in B_max, the possibility of decreased ¹¹C-CURB binding affinity has not been ruled out, especially as the polymorphism alters a nucleotide close to the catalytic site [149]. Either way, controlling for this polymorphism in further studies employing ¹¹C-CURB will be necessary. A further study by the group investigated the effect of cannabis and found significantly lower (−14% ¹¹C-CURB λk₃ values) in cannabis users vs HC, and a negative correlation with clinical impulsiveness scale assessments [284]. In both reported studies, one subject with the A/A polymorphism was included and had the lowest λk₃ value from each cohort, implying an A-allele dose-response may be present. For ¹¹C-MK-3168, no work has been published on the effect of C385A polymorphisms on binding properties.

These two tracers represent the first successful strategies to image FAAH in humans, proving feasibility of PET studies for this target. While the kinetics of ¹¹C-MK-3168 are relatively slow, reversible tracers are preferred as quantification is easier and outcome parameters are more intuitive than their irreversible counterparts, such as for ¹¹C-CURB. Nevertheless, both tracers have shown repeatability of their outcome parameters in humans and appear suitable for occupancy studies. The effects of the C385A polymorphism are a complicating factor for ¹¹C-CURB studies; however, characterisation of the polymorphism impact will now be expected for all FAAH tracers.

It is well known that ‘low’ molar activity will cause occupation of the target site due to self-block, effecting outcome parameters. However, the quantity of injected mass that would cause a non-negligible self-block is dependent on multiple factors including the target density and affinity of tracer for the target and is often not well characterised in humans. Therefore, the validity of results from studies using poor molar activity, or significantly different injected masses between groups, may be questioned, reducing confidence in results.

Multiple tracers discussed have evidence of off-target binding sites. For some, such as ¹¹C-LY2795050 and ¹¹C-Cimbi-36, off-target sites are known and characterised, allowing assessment of when this limitation will and will not impede study outcomes. For others, however, the off-target site(s) are unknown, preventing informed decisions on potential outcome parameter impact, raising concerns on biological outcomes, and ultimately reducing tracer usefulness.

Addressing these issues will increase the chances of developing useful PET tracers for the CNS and help prevent late-stage issues arising. Notwithstanding these issues, the large number of tracers for new CNS targets identified in this review indicates there is considerable potential to advance understanding of brain function across disorders and informs the development of new therapies.