Background
The main goal of early phase drug development is the demonstration of substantial target engagement at doses which are safe and tolerable and the determination of a therapeutic dose range. Positron emission tomography (PET) provides a unique technology for establishing the relationship between the plasma concentration of a compound and the degree of target engagement in tissues such as the brain, which are not accessible by other methods. The method of choice for defining this relationship is the use of a well-understood probe (or radioligand) which allows the quantification of target availability at baseline and following the administration of various doses of a novel drug. This approach allows the construction of a realistic pharmacokinetic model relating the kinetics of a novel drug in blood and plasma to that in tissue and enabling the prediction of target occupancy in future clinical studies [
1]. The practical limitation of this approach is the requirement for a well-characterised PET radioligand suitable for the quantification of the target examined. While a range of radioligands are available for the central nervous system (CNS), a large number of targets do not have a suitable radioligand. An alternative, indirect, method to obtain target occupancy information in these cases is to estimate the ‘free’ or unbound concentration of the drug in the tissue of interest (
CFT) by directly labelling the drug molecule. The total concentration of a drug in tissue is of limited utility, as the non-specifically bound fraction is pharmacologically inactive and thus the information is restricted to providing confidence in Blood-brain barrier (BBB) penetration as a binary phenomenon. On the other hand,
CFT can be combined with an estimate of the affinity of the drug for its target (
KD), obtained from suitable
in vitro or pre-clinical studies, to obtain an estimate of target occupancy. While such an approach makes several assumptions about the shape of the dose-occupancy relationship and the equivalence between
in vitro-derived
KD measures and the true
in vivo KD, it provides a practical method for estimating target occupancy information in the absence of a well-characterised PET radioligand. The theory underpinning this approach has been fully described in a recent manuscript [
2]; hence, we present a short summary only.
Passive transport of a drug across the BBB determines that at equilibrium, the
CFT (given by the product of the total non-displaceable tissue concentration [
CND] and its free or unbound fraction [
fND]) will be equal to the free concentration of the drug in the plasma,
CFP (=
C
p
f
p
) which implies,
(1)
The fractional occupancy of the target (Occ
T) can then be calculated knowing
C
p
f
p
and
KD as,
(2)
The main goal of a ‘biodistribution’ PET study is thus to test the assumption that the drug crosses the BBB and to assess whether the transport is consistent with passive diffusion. The distribution of a drug in the tissue of interest can be evaluated following its labelling with a radionuclide. The substitution of a
12C or
19 F atom by the positron emitting isotopes
11C or
18 F enables the quantification of the drug distribution in tissue over time, without changing its physiochemical or pharmacological characteristics. In principle, one can estimate the total volume of distribution (
VT[
3]), for the drug in the brain, which is equivalent to the equilibrium partition coefficient for the drug between brain and plasma.
If the
VT of a labelled drug is not reduced following the administration of pharmacologically relevant dose of the same drug, the total brain concentration of the drug (
CT) can be taken to measure the sum of the free and the non-specifically bound components. In this case, the
VT will be independent of the dose of labelled compound administered and measurements made at tracer concentrations will provide an adequate estimate of drug distribution at pharmacological doses.
(3)
It is worth noting that if the specific binding component of a labelled drug is not negligible compared to the free and non-displaceable components, the labelled drug may be used as a radioligand to enable a direct examination of target occupancy.
While
VT,
CP, and
f P can be measured directly in the course of a PET study in humans, it is not possible to measure
f ND directly, and hence,
CFT cannot be obtained without other information being available. If a value for
f ND can be obtained, then the assumption of passive diffusion across the BBB can be tested. Under the conditions of passive diffusion, at equilibrium,
CFT =
CFP, and Eq.
3 is simplified to:
A
VT measurement from a PET biodistribution study can be compared to the
f P/
f ND ratio, and passive diffusion can be assumed if the two are similar. Equilibrium dialysis provides a practical method for measurements of brain tissue
f ND (and also
f P if necessary) [
2].
We have applied the methodology above to examine the distribution of GSK1034702 in the primate and human brain. GSK1034702 is a selective muscarinic-1 (M1) receptor allosteric agonist, which offers a potential therapy for the treatment of cognitive dysfunction in neurodegenerative disorders. It belongs to the series of novel N-substituted benzimidazolones recently described [
4]–[
6].
Muscarinic acetylcholine receptor (mAChR) agonists such as xanomeline have produced some efficacy in the treatment of cognitive dysfunction in patients with Alzheimer's disease (AD) and schizophrenia (SZ) [
7],[
8]. The therapeutic potential of the cholinergic agents tested thus far (cholinesterase inhibitors and muscarinic agonists) is modest and is thought to be limited by peripheral m
2AChR and/or m
3AChR-related side effects, such as gastrointestinal disturbances [
9].
The discovery of an allosteric (or ectopic) site for m
1AChR that is not conserved across mAChR subtypes has provided the opportunity to develop m
1AChR agonists with true receptor selectivity [
10]. Allosteric m
1AChR selective agonists are postulated to display reduced side effects, and thus greater utility, compared with the existing non-selective orthosteric agonists. In an initial study, in healthy smokers using the nicotine abstinence model of cognitive dysfunction, GSK1034702 improved episodic memory [
6].
GSK1034702 [
11] is a potent allosteric agonist at human recombinant m
1AChR (pEC
50 = 8.1 ± 0.1, intrinsic activity (IA) = 0.78 ± 0.02) with 100-fold selectivity over human m
2-5AChR receptors and with >90 over other molecular targets from a variety of classes. GSK1034702 is a weak substrate for P-glycoprotein (PGP) (efflux ratio of 2.6:1 in MDCK cell line expressing human MDR1) and has displayed some species variability in brain-to-plasma ratios (brain/blood ratios of 0.4, 0.6 and 2.0:1, in mouse, rat and marmoset, respectively; GlaxoSmithKline (GSK) data on file). Therefore, the delivery of GSK1034702 into human brain could have been adversely affected. Initially, the primate study was carried out in order to demonstrate that [
11C]GSK1034702 crossed the BBB in primates, before proceeding to expensive human studies. The human study aimed to use PET with radioactively labelled GSK1034702 to help ascertain the role PGP plays in limiting brain penetration in the presence and absence of a pharmacological relevant oral non-labelled dose of 5 mg.
Here, we investigated the distribution of GSK1034702 in the living human brain. These data provide vital information to determine whether GSK1034702 can be used at doses that offer therapeutic benefit to patients with AD and SZ without inducing intolerable muscarinic-related side effects.
Discussion
The exploding costs of drug development, and in particular in the neurosciences, make it imperative that decisions about compound progression are made as early as possible, in order to terminate unsuitable candidates before they reach later, more expensive phases of development. The fundamental requirement of any new molecule is that it is able to reach its site of action in pharmacologically meaningful quantities, at doses which are generally safe and tolerable. The quantification of target engagement in the CNS is hampered by our inability to directly assay the target in the living human brain and the frequent species differences encountered. The direct quantification of drug interaction with its target via the use of PET with a selective radioligand is the method of choice in deriving the relationship between the plasma concentration of a compound and the occupancy of its target. This technique can provide information on target occupancy following single-dose administration that can be extrapolated to provide accurate estimates of occupancy in the clinically meaningful setting of repeat-dose administration [
1].
A large fraction of CNS targets of interest are not suitable for examination by this method, due to the lack of suitable radioligands. In such situations, the traditional approach has been to label the molecule of interest and derive the partition coefficient between the labelled drug in plasma and in the brain, as a semi-quantitative measure of drug availability at the target. Such measures are useful, but more limited than direct estimates of target occupancy, since the partition coefficient (known in the PET literature as the total volume of distribution, VT) provides quantification of the total concentration of the drug in the brain. The total concentration is composed of the drug specifically bound to the target (VS) and the non-displaceably bound drug (VND), which comprises drug-bound non-specifically as well as the free drug. If the VS can be estimated robustly from kinetic data, direct target engagement may be evaluated, using the methods developed for target-specific radioligands. However, this situation is rare for drugs in development, primarily due to the presence of high levels of non-specific binding (high VND). Hence, for a large proportion of labelled drugs, the measured VT is essentially equivalent to VND.
The pharmacologically relevant fraction of the drug is the free concentration, CFT, which is able to interact with the target, and the fND is needed to estimate CFT when only the total tissue concentration is available. If the drug transport across the BBB is by passive diffusion, then at equilibrium, the free concentration of the drug in plasma would be equal to CFT. A demonstration that the VT for a compound is approximated by the ratio fP/fND provides confidence in drug transport by passive diffusion and hence the estimation of CFT from peripheral plasma data.
We did not see conclusive evidence for a change in regional VT values for GSK103702 following a pharmacological dose that was statistically significant, suggesting that any specific binding component of [11C]GSK1034702 is negligible compared to the free and non-specific components. The whole brain VT in human was 4.9, which is broadly comparable to the measured fP/fND of 2.63, implying passive diffusion across the BBB. The difference between VT and fP/fND is most likely due to experimental noise in the PET and equilibrium dialysis assays, but even if this difference represents a deviation from passive diffusion across the BBB, this deviation would be in the direction of an active transport of the molecule into the brain. Such a situation, were it to be clinically relevant would lead to a higher brain CFT than the plasma-free concentration, and thus would minimise the incidence of peripheral side effects for a given central m1AChR occupancy.
Similarly, there was significant uptake of GSK1034702 into the brain at tracer quantities and following oral dosing of 5 mg in the human study. As there was some regional heterogeneity in the binding of GSK1034702 in the human brain, the cortex was used to derive a VT assumed to be the most representative of the VND, as this region produced VT in the lower range of those examined. The whole brain VT was 4.9, which is in broad agreement with primate VT and the fP/fND ratio (3.97 and 2.63, respectively).
The results are broadly consistent across species, although the primate and human estimates are somewhat higher than the fP/fND ratio. Thus, if active transport plays a role in the BBB passage of GSK1034702, it would appear to be an active influx into the brain (leading to VT > fP/fND), rather than an extrusion of the compound from the brain (which would lead to VT < fP/fND). Thus, if anything, therapeutic levels of m1AChR occupancy may be achieved with even lower plasma concentrations and hence lower incidence of adverse events, than those in the passive diffusion scenario. The presence of an active transport mechanism would make extrapolation from tracer studies to pharmacological doses more problematic, due to possible changes in the transport kinetics at higher doses (e.g. due to saturation of the transporters). However, we do not think that this is an issue for GSK1034702, as our examination of a therapeutically relevant dose (5 mg p.o.) produced results similar to those at tracer dose. The inclusion of a therapeutic dose requires extra safety information and complicates the study design, but may be worth including in situations where active transport is suspected.
A regionally heterogeneous distribution of [
11C]GSK1034702 was observed in humans, with the highest
VT observed in the medial temporal lobe (MTL), consistent with the known distribution of m
1AChR in the human brain [
16]. This finding is supported by a small (but non-significant) reduction in the MTL
VT following the administration of a 5-mg oral dose of GSK1034702, in both subjects for whom pre- and post-dose data was available. Our data are consistent with a small specific binding component for [
11C]GSK1034702 in the human brain. The data from humans are in contrast to those in the baboon brain, where homogenous distribution was observed. A potential explanation for the between species differences in distribution could be a masking of high affinity states of m
1AChR by ketamine [
17],[
18] in the baboon brain. However, further studies would be needed to properly understand these differences.
The magnitude of the specific component may be too low for [11C]GSK1034702 to be a useful PET ligand for the m1AChR. This may be due either to a relatively low in vivo affinity of [11C]GSK1034702 for the m1AChR in relation to the target density or due to a relatively low specific activity (SA) and hence a high mass of GSK1034702, leading to partial blockade of the specific binding component being used in this study. The radiolabelling method used precludes the achievement of higher SA than reported here, and hence, [11C]GSK1034702 may not be suitable for the quantification of m1AChR density with PET. However, it may provide a useful lead in the search for related compounds with higher affinity for this target or compounds that may be labelled using methods leading to higher SA.
This study type has several limitations. For the PET data, the presence of brain-penetrant metabolites cannot be ruled out for all compounds using these methods. In this case, it is unlikely because there was little metabolism of the compound and because the data was well described by a two-tissue compartment model. For the equilibrium dialysis data, the duration of dialysis, temperature and buffer conditions and the fact that different primate brains were used for the in vitro and in vivo data of the primate study, as well as primate brain instead of human brain tissue for the human study, could all introduce errors into the measurements.
Conclusions
In primate and human PET studies designed to evaluate the transport across the BBB of a novel selective M1 receptor allosteric agonist GSK1034702, we have demonstrated good brain uptake. Although there was variability in this limited dataset, BBB kinetics were consistent with passive diffusion or active influx.
The single oral dose of 5 mg GSK1034702 was well tolerated by healthy subjects in this PET study, and the pharmacokinetic parameters were consistent with those observed in a separate First-Time-In-Human safety and tolerability study (GSK study number 110623 and study number NCT00743405 on CT.gov).
In conclusion, an examination of the BBB kinetics of GSK1034702 in phase 0 and phase I of development has discharged some of the perceived development risks for GSK1034702 and provided information to progress the molecule into the next stage of clinical development.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.
The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
Competing interests
The study was sponsored by GlaxoSmithKline (GSK) pharmaceuticals [study number 110771 and ClinicalTrials.gov Identifier NCT00937846]. All authors except AAD and MS were employees of GSK and held shares in the company at the time of this study. AAD and MS had previously received funding by GSK.
Authors’ contributions
KR carried out the human PET studies and drafted the manuscript. VC performed quantitative analysis of the human PET study, helped design the human PET study and helped draft the manuscript. MH designed and optimised the method to radiolabel [11C]GSK1034702 for use in humans and synthesised [11C]GSK1034702 used in human studies. LM designed and optimised the method to radiolabel [11C]GSK1034702 for use in primate studies and helped draft the primate studies section of the manuscript. SPM performed quality control (QC) and the QC analytical method development for [11C]GSK1034702 for use in human studies. JP performed metabolite analysis and the metabolite analytical method development for in the human PET study. RNG assisted with quantitative analysis of human PET data and helped draft the manuscript. GS helped design the analysis pipeline and performed the image processing for the human PET study. AAD participated in the design and coordination of the primate PET study. MS performed image processing and quantitative analysis of the primate PET study. JW helped obtain brain protein binding data for GSK1034702 and provided input into interpretation of the results in relation to the biology of the target. ML conceived of the study and participated in its design. EAR designed the PET studies and drafted the manuscript. All authors read and approved the final manuscript.