Quantification of P-glycoprotein function at the human blood-brain barrier using [¹⁸F]MC225 and PET

The blood-brain barrier (BBB) is a physical dynamic boundary between the capillaries of the brain vasculature and brain tissue that is composed of endothelial cells connected by tight junctions. The architecture of the BBB prevents paracellular diffusion of hydrophilic compounds and only small molecules can cross by passive transmembrane diffusion. Transport across the BBB is possible due to the presence of multiple transporters, which regulate the influx and efflux of both exogenous and endogenous compounds across the BBB [1]. P-glycoprotein (P-gp) or ABCB1 is one of those transporters and is responsible for the transport of a variety of compounds, like neurotoxins and drugs. It is an efflux transporter and member of the ATP Binding Cassette (ABC)-transporter family and it plays an important role in maintaining brain homeostasis and protection of the brain through efflux of harmful substances [2]. P-gp expression and function can be altered due to several factors, including administration of pharmaceuticals, stress, ageing, diet, inflammatory responses, and genetic differences [3, 4].

Changes in P-gp function are proposed as possible aetiology of several neurological and psychiatric disorders, including Alzheimer’s disease and Parkinson’s disease [5‐8]. In Alzheimer’s disease, decreased P-gp function is associated with a reduced efflux of neurotoxic amyloid, which itself is a P-gp substrate. This causes an imbalance between production and clearance of amyloid, leading to higher levels of neurotoxic amyloid in the brain [9]. On the other hand, several pharmaceuticals are classified as P-gp substrates and the efflux of those therapeutic drugs across the BBB is a potential mechanism of drug-resistance and drug-drug interactions. In treatment-resistant depression, schizophrenia, epilepsy, and oncology, an increase in P-gp function is associated with low bio-availability of pharmaceuticals in the brain [10, 11].

PET imaging is an established method to measure P-gp transporter function in vivo and, as such, it allows for a better understanding of transporter-related pathophysiological processes. Moreover, PET imaging of P-gp transporter function may be useful in selecting the best treatment option for patients with an altered P-gp function in case of drug resistance, identifying patients who are at risk for developing a disease, and evaluating new potential treatment strategies, including P-gp inducer therapy [12‐15]. Several P-gp substrates have been labelled in order to image P-gp function at the BBB. These compounds are transported out of the brain by P-gp and therefore P-gp function can be assessed indirectly by measuring uptake of the tracer in the brain. At present, the P-gp substrate (R)-[¹¹C]verapamil is the most widely used PET tracer to measure P-gp function in a variety of preclinical and clinical studies [16]. (R)-[¹¹C]verapamil shows high selectivity for the P-gp transporter and, using this tracer, a decrease in P-gp function of patients with Alzheimer’s disease compared with healthy controls has been demonstrated [17]. However, one of the major drawbacks of (R)-[¹¹C]verapamil is the fact that is a so-called avid or strong substrate. This means that its brain uptake in a normal or healthy P-gp state is very low and therefore reliable quantification is very challenging [18].

To overcome this challenge, Savolainen et al. developed 5-(1-(2-[18F]fluoroethoxy))-[3-(6,7-dimethoxy-3,4-dihydro-1H-isoquinolin-2-yl)-propyl]-5,6,7,8-tetrahydronaphthalen ([¹⁸F]MC225), a novel PET tracer for measuring P-gp function at the BBB [19]. Pharmacokinetic evaluation showed that [¹⁸F]MC225 acts as a weak P-gp substrate and that it is selective for P-gp, having no affinity for the BCRP transporter at the BBB as well [20]. In a head-to-head comparison performed in non-human primates (NHP), Garcia Varela et al. showed that brain uptake of [¹⁸F]MC225 at P-gp baseline levels is higher than that of [¹¹C]verapamil, making this tracer better suitable for measuring both increases and decreases in P-gp function [18]. In the first-in-human study of [¹⁸F]MC225, Toyohara et al. evaluated the tracer in three healthy volunteers and concluded that [¹⁸F]MC225 has an acceptable radiation dose together with pharmacological safety at the dose that is required for PET imaging. The volume of distribution (V_T), used as indicator of P-gp function, was compared with previous P-gp PET tracers, and it was shown that baseline uptake of [¹⁸F]MC225 was indeed higher than that of any other existing P-gp PET tracer [21]. However, test-retest variability of [¹⁸F]MC225 has not been studied yet, nor have adequate studies been performed to identify the best kinetic model for quantifying [¹⁸F]MC225 data.

The aim of the present study was twofold: First, [¹⁸F]MC225 behaviour in healthy volunteers was analysed to identify the pharmacokinetic model that best describes the kinetic behaviour of this novel P-gp PET tracer in the brain based on accuracy, precision, and biological assumptions. Second, the selected model was used to evaluate test-retest variability of [¹⁸F]MC225.

In this study, several compartment models were investigated to select the optimal model for describing [¹⁸F]MC225 kinetics in humans based on accuracy, precision, and biological assumptions. The present study shows that the 2T4K model with V_B as model parameter is the most accurate model to describe [¹⁸F]MC225 kinetics in the human brain, and therefore it is the optimal model for quantifying P-gp function at the BBB. Previous preclinical [¹⁸F]MC225 studies in both rats and non-human primates also have shown that the reversible two tissue compartment model provides the best fit to the PET data, at least for longer scan durations [20, 24]. As [¹⁸F]MC225 is a substrate of P-gp instead of a tracer that binds to a receptor, the tracer will not follow the same rules of pharmacokinetic quantification as receptor-bound tracers. Therefore, absolute quantification of [¹⁸F]MC225 is relatively complicated. Based on a biological approach, it could be assumed that k₂ would be the parameter of choice to quantify P-gp function, as it represents the transport from the first tissue compartment back to plasma. On the other hand, Lubberink and colleagues showed that, in general, K₁ is the preferred parameter to describe P-gp function [25]. This could be explained by the fact that pharmacokinetic models are an oversimplified mathematical description of the complex underlying biological processes. If transport by P-gp results in tracer efflux before the tracer actually enters the brain, changes in P-gp function will be reflected in changes in K₁ rather than in k₂. However, K₁ may depend on perfusion and requires a correction for changes in regional cerebral blood flow (CBF) with a separate CBF measurement. In the current study, in addition to the ‘standard’ reversible and irreversible models, a reversible model with a fixed k₃/k₄ parameter was evaluated, to explore the assumption of a stable binding (k₃/k₄) across patients.

The present results show a relatively high standard error for both K₁ and k₂, resulting in low precision for these parameters. Moreover, since K₁ is the product of flow and extraction fraction, changes in K₁ may be due to changes in CBF. In contrast to the high standard errors for K₁ and k₂, the volume of distribution showed a low standard error and therefore, it was selected as the optimal parameter for evaluating BBB P-gp function. The volume of distribution is defined as the ratio of the radioligand concentration in target tissue to that in plasma at equilibrium. This ratio is inversely related to P-gp function.

In cerebral PET studies, the input function needs to be corrected for labelled plasma metabolites, as plasma metabolites usually do not cross the BBB. In the present study, [¹⁸F]MC225 was relatively stable over time with a parent fraction of more than 60% after 60-min scan duration. For comparison, the widely used P-gp tracer (R)-[¹¹C]verapamil showed a parent fraction of 45% after 60 min [16]. This would imply that the signal measured with [¹⁸F]MC225 is less affected by the formation of radio-metabolites, improving the accuracy of P-gp measurements in vivo.

[¹⁸F]MC225 is a weak P-gp substrate and was developed to measure both increases and decreases in P-gp function. [¹⁸F]MC225 V_T for the whole brain grey matter region was 6.59, which is substantially higher than the 0.78 reported for (R)-[¹¹C]verapamil [26]. This is in line with the fact that verapamil is a more avid P-gp substrate than MC225. In addition, in preclinical studies and in a head-to-head comparison with (R)-[¹¹C]verapamil, higher baseline uptake was found for [¹⁸F]MC225 [18, 20]. This higher baseline uptake enables measurements of increased P-gp function, which give rise to a reduction in uptake (and V_T), as already has been shown in preclinical studies using the P-gp inducer MC111 [27]. Future human studies in, for example, a patient population suffering from treatment-resistant epilepsy are needed to assess the ability of [¹⁸F]MC225 to measure increases in human P-gp function as well.

One important finding of the present study is the relatively large TRT variability of 28% for V_T, which is higher than the 9% TRT variability reported for (R)-[¹¹C]verapamil [28]. For [¹⁸F]MC225, TRT variability may be affected by physiological fluctuations of P-gp function and expression, environmental factors, and technical and methodological factors. One other explanation for the large TRT variability is that the lower reproducibility of [¹⁸F]MC225 might be a result of [¹⁸F]MC225 being a weaker P-gp substrate, which is transported more gradually across the BBB and, therefore, is more affected by subtle changes in P-gp function [29]. Physiological small changes in P-gp function are present during the day due to the circadian cycle and hormonal influences and although patients were scanned at the exact same time of the day, minor changes in P-gp function could have contributed to the lower reproducibility of [¹⁸F]MC225 [30]. Furthermore, interindividual outcomes were highly variable (S5-7) and one of the subjects showed an increase in whole brain grey matter V_T of 88%. When excluding this subject from the analysis, the TRT variability decreased from 28.3 ± 20.4 to 20.2 ± 10.1%, which is similar to the test-retest variability of 19% that was previously found in a preclinical [¹⁸F]MC225 study under controlled conditions [31]. When excluding this subject from the analysis, the TRT variability decreased from 28.3 ± 20.4 to 20.2 ± 10.1%, which is similar to the test-retest variability of 19% that was previously found observed in a preclinical [¹⁸F]MC225 study under controlled conditions (32). In this preclinical study, the high TRT variability was attributed to the use of isoflurane anaesthesia. However, in the current study, no anaesthesia was used and the high TRT variability can therefore not be explained by these pharmaceutical effects. When comparing the results of [¹⁸F]MC225 with the current ‘gold standard’ (R)-[¹¹C]verapamil, it should be noted that in the study of van Assema and colleagues in which a TRV of 9% was found for (R)-[¹¹C]verapamil, patients were scanned twice on the same day [28]. In the current study, scanning of both test and retest scans on the same day was not possible due to the longer half-life of the radionuclide. There was a minimum interval of 2 weeks between scans, possibly contributing to the higher TRT variability.

Although TRT variability of [¹⁸F]MC225 V_T in humans is relatively high, under pathophysiological conditions, changes in P-gp function usually are larger, including a nearly 50% reduction in P-gp activity in mild Alzheimer’s disease and a 40% increase in P-gp protein levels in patients with epilepsy [32, 33]. In an ongoing study in which differences in brain [¹⁸F]MC225 uptake at baseline and after inhibition of the P-gp function are assessed, a difference of 25% was found in the first study subject. Although the high TRT variability may raise concerns about the interpretation of quantitative PET data in individual patients, it does most likely not affect the outcome of this study at group level, as in that case changes are expected to be unidirectional [34] When taken these findings into account, pathophysiological changes in P-gp function might still be measurable using [¹⁸F]MC225, at least at a group level. Further studies in neurological and psychiatric patient populations are needed to assess whether these changes in P-gp function are large enough (compared with normal subjects) to be measured with [¹⁸F]MC225.

In the current study, both male and female subjects in the age group 50–80 years were included to ensure that the study population could serve as a representative control group for future studies in Alzheimer’s disease and Parkinson’s disease. Although the main reason for including both males and females was to obtain a representative control group for future studies, this inclusion also allowed for an assessment of possible gender differences in P-gp function at the BBB. However, in contrast to previous literature in which gender differences in P-gp function were present, in this study, no differences in P-gp function were found between male (V_T = 6.6 ± 1.1) and female (V_T = 6.2 ± 1.3) subjects (p = 0.595) [35, 36]. This might be due to the relatively small study population of 14 subjects. Future studies are necessary to study the effect of age with [¹⁸F]MC225. Toyohara et al. stated in the first clinical study with [¹⁸F]MC225 that P-gp function at the BBB decreases with increasing age. In our study, only a very low correlation (R² = 1.25*10⁻⁴) was found between V_T values of the whole brain grey matter and the increasing age of the study subjects (supplementary data S9). This might be explained by the inclusion of only elderly subjects for which the variability is not as large as in other studies in which ageing and P-gp function was studied.

[¹⁸F]MC225 V_T, derived using a reversible two-tissue compartment model, is the preferred parameter to describe P-gp function at the human BBB. This outcome parameter has an average test–retest variability of 28%. Although this TRTV is relatively high, changes in P-gp function in pathophysiological conditions are even more extensive and adequate measurements of those alterations in future clinical studies should be feasible with [¹⁸F]MC225.