Despite the development of numerous new anticonvulsant drugs, pharmacoresistance still affects about 20–30% of epilepsy patients.1, 2 The etiology of different types of epilepsy includes various causes like developmental disorders, brain insults, or genetic factors.3 As drug resistance itself may also be caused by various mechanisms, a huge amount of variation between individual patients becomes obvious. This variation explains the need of diagnostic tools to evaluate potential mechanisms of drug resistance in order to provide adequate individual treatment. Preclinical molecular imaging provides the opportunity to evaluate proposed mechanisms of drug resistance in animal models of pharmacoresistant epilepsy with a high translational potential. Furthermore, new imaging concepts can be established and evaluated using imaging in combination with histology and molecular biology.
14.2 Animal Models of Drug-Resistant Epilepsy
Animal models of drug-resistant epilepsy are still crucial for evaluating mechanisms of drug resistance, as availability of diseased human brain tissue and brain tissue of corresponding healthy controls is extremely limited. A consensus definition of drug resistance4 requires unsuccessful treatment with at least two antiepileptic drugs (AEDs) with different targets and at maximum tolerable dose before the diagnosis of drug resistance. Regarding the translational value of animal models, they should fulfil at least this requirement. One can distinguish two major types of rodent models of drug resistance: (1) models in which all animals are per se resistant to a variety of AEDs and (2) models in which certain percentages of animals are resistant (nonresponders) and others are responsive to AEDs (responders).5, 6 In principle, both types of models are valuable for evaluation of mechanisms of drug resistance. However, an inevitable requirement for imaging studies is that animals underwent an epileptogenic process leading to brain alterations responsible for pharmacoresistance. Models that exhibit a priori drug-resistant induced seizures, like the 6-Hz psychomotor seizure model in mice,7 have only limited value for imaging studies. In contrast, models with 100% pharmacoresistant chronically epileptic animals can serve for evaluation of potential new mechanisms of pharmacoresistance in comparison to control animals. Examples are lamotrigine-resistant kindled rats, showing also resistance to carbamazepine, phenytoin, and topiramate,8 or methylazoxymethanolacetate-exposed rats, which are refractory to valproate, ethosuximide, or carbamazepine.9 Nevertheless, this approach can give clear evidence that a certain mechanism is involved in pharmacoresistance only if a treatment strategy is successful and able to reverse drug resistance. The second type, i.e., models with responder and nonresponder subgroups, in contrast, gives the unique opportunity to directly compare both subgroups and to evaluate whether a certain resistance mechanism is differently expressed. Still, proof that a certain mechanism is responsible for drug resistance is given only if drug resistance can be counteracted in the nonresponder group. The criterion of resistance to at least two AEDs in the nonresponder group has been met by the amygdala-kindling rat model, in which pharmacoresistance to phenytoin extends to various other AEDs,10, 11 as well as post-status-epilepticus rat models in which resistance to phenobarbital extends to phenytoin (BLA model) or levetiracetam (pilocarpine model).12–14 Notably, most published preclinical imaging studies did not use animal models that fulfil the requirements mentioned above. This means that results from these studies will give insight into brain changes during epileptogenesis or chronic epilepsy, but a direct relation to drug resistance remains vague and is strengthened only in combination with results of other, nonimaging studies.
14.3 Mechanisms of Drug Resistance in Epilepsy
Various mechanisms underlying drug-resistant epilepsy have been proposed. Still, the two major mechanistic concepts of drug-resistant epilepsies are represented by (1) the target hypothesis and (2) the transporter hypothesis.15, 16 The target hypothesis explains drug resistance by changes in drug target affinity and expression. For example, both in patients and in animal models, changes in drug sensitivity of voltage-gated sodium channels have been shown,17, 18 which might be caused by down-regulation of β1 and β2 subunits.19 Another proposed mechanism of drug resistance is a changed GABAA receptor, the target of GABA-mediated inhibition. Changes in the subunit composition of GABAA receptors of epileptic rats correlated with profound alterations in pharmacosensitivity and receptor function.20–22 For both targets, the GABAA receptor and voltage-gated sodium channels, molecular imaging approaches are available and first studies are published (see the “Imaging Drug Target Alterations” section).
However, a single change in one drug target alone cannot explain why a patient is resistant to different AEDs with different mechanisms of action, although it might be possible that multiple drug targets are changed within the same patient. In this context, the transporter hypothesis might provide further explanation for the existence of multiresistant patients. Over the last two decades, the transporter hypothesis has been the most extensively studied explanation for drug-resistant epilepsy.15, 23 Drug efflux transporters are expressed in various organs all over the body and particularly at tissue barriers like the blood-brain barrier. Here, they have a major influence on passage of their substrates. At the blood-brain barrier, several so-called multidrug transporters with partially overlapping substrate spectrum are physiologically expressed, of which P-glycoprotein has been by far most intensively studied until today. Furthermore, transport of various AEDs by P-glycoprotein has been shown both in vitro and in vivo.24 Strong evidence is given for overexpression of drug transporters in pharmacoresistant epilepsy patients as well as in animal models of drug-resistant epilepsy.25–28 As a proof of concept for the transporter hypothesis, Brandt et al.29 showed that pharmacoresistance toward the AED and P-glycoprotein substrate phenobarbital in chronically epileptic rats can be reversed by application of the P-glycoprotein inhibitor tariquidar. In parallel, van Vliet et al.30 described that coadministration of phenytoin and tariquidar markedly increased seizure control.
14.4 Imaging Changes in Multidrug Transporter Expression
Over the last years, imaging of multidrug transporter expression has been a major research focus in preclinical molecular imaging of epilepsy models.31, 32 A large part of these studies has been performed within the European Union–funded research consortium EURIPIDES (European Research Initiative to Develop Imaging Probes for Early in Vivo Diagnosis and Evaluation of Response to Therapeutic Substances). For radio-isotope imaging of efflux transporters at the blood-brain barrier, two principal approaches were proposed: (1) radio-labeled transporter substrates for imaging transporter function and (2) radio-labeled nontransported inhibitors for imaging transporter expression.33 Despite some evidence that apart from P-glycoprotein, other brain efflux transporters might be involved in pharmacoresistant epilepsy, imaging approaches for these transporters are still very limited and no studies in epilepsy models or patients are published.
The still most widely used transported radiotracer remains the P-glycoprotein substrate [11C]verapamil, which was first evaluated with regard to P-glycoprotein function at the mouse blood-brain barrier by Hendrikse et al.34 using the racemic formulation. They could demonstrate that in mdr1a(-/-) mice [11C]-verapamil brain uptake was 9.5 times higher than in the wild type mice. Furthermore, after injection of a dose of 50 mg/kg of the P-glycoprotein inhibitor cyclosporine A, [11C]verapamil brain uptake increased 10.6-fold in wild type mice whereas mdr1a(-/-) mice did not show any significant increase. However, many substrate tracers like (R)-[11C]verapamil, described to be the superior enantiomer of [11C]verapamil,35 like the serotonin 5-HT1A antagonist [18F]MPPF36 or like the more recently developed [11C]-N-desmethyl-loperamide37 are effectively transported (high-affinity substrates) resulting in very low brain uptake in rats (Figure 14.1). Brain uptake values for (R)-[11C]verapamil in rats are well below 0.1% injected dose per gram tissue (%ID/g) at 60 min after tracer injection and <0.1 mean standardized uptake value (SUV) was reported for [18F]MPPF and [11C]-N-desmethyl-loperamide, whereas complete P-glycoprotein inhibition leads to up to 12-fold, 10-fold, and 5-fold increase, respectively (Figure 14.1).36, 38, 39 Notably, the translational value of [18F]MPPF for imaging transporter function in patients is rather limited as it appears not to be transported by human P-glycoprotein.40
Figure 14.1. Horizontal (A and C) and sagittal (B and D) (R)-[11C]verapamil PET summation images (0–60 minutes) of healthy rats recorded before (A and B) and 120 minutes after (C and D) administration of the P-glycoprotein inhibitor tariquidar. Complete inhibition of P-glycoprotein by 15 mg/kg tariquidar leads to a 12-fold increase in (R)-[11C]verapamil brain uptake. Originally published in J Nucl Med. Bankstahl JP, Kuntner C, Abrahim A, et al. Tariquidar-induced P-glycoprotein inhibition at the rat blood–brain barrier studied with (R)-11C-verapamil and PET. J Nucl Med. 2008;49:1328–35.
As in epilepsy only increased expression of transporters will contribute to drug resistance, and thereby leads to even more reduced tracer brain uptake, any quantification is extremely challenging. Multiple approaches have been followed to overcome this limitation. First, less effectively transported substrates would lead to higher brain uptake, making increased transporter function more easily quantifiable. In this case decreased brain uptake would be indicative of increased transport activity. In this regard, the GABAA ligand [11C]mephobarbital, a methylated analog of the AED phenobarbital, was evaluated for potential transport by P-glycoprotein,41 as shown for phenobarbital both in vitro and in vivo.24 Interestingly, insertion of a methyl function into phenobarbital completely abolished transporter substrate properties in [11C]mephobarbital.41 Furthermore, the sodium channel ligand [11C]phenytoin was evaluated resulting in brain uptake levels of about 0.2 SUV in rats.42 After maximal P-glycoprotein inhibition, an increase by only 45% is described,42 confirming low-affinity substrate properties. The usefulness of this approach to identify disease-related increases in P-glycoprotein function has yet to be shown. Second, a prodrug-approach was suggested to make quantification of drug efflux easier. In this approach, a nontransported radiolabeled prodrug that can easily cross the blood-brain barrier is metabolized in the brain into a radiolabeled substrate, and then transported out of the brain. For P-glycoprotein, Sander et al.43 could demonstrate high initial brain uptake and increased tracer levels after transporter inhibition by using such a radiolabeled prodrug, but no differences in brain clearance using pharmacokinetic modeling. More work is needed to prove the practicability and usefulness of this approach. Notably, some preclinical imaging studies using high-affinity substrates like (R)-[11C]verapamil in epilepsy models were also performed after maximal P-glycoprotein inhibition. Without inclusion of baseline scans within the same subjects for data analysis, these studies will only allow to evaluate P-glycoprotein-transport-independent brain changes. Therefore, the third approach includes only partial P-glycoprotein inhibition in order to increase brain uptake of effectively transported substrates.44 In this proof-of-concept study, the authors could demonstrate that P-glycoprotein expression levels correlate with efflux rate constant k2 obtained by pharmacokinetic modeling after half-maximal P-glycoprotein inhibition (Figure 14.2). In a preceding dose finding study using the third generation P-glycoprotein inhibitor tariquidar in combination with (R)-[11C]verapamil, the ED50 value for tariquidar of 3.0 mg/kg was identified in rats.45 Most strikingly, this dose leads to almost identical tariquidar plasma levels like the half-maximal inhibiting dose in human volunteers.46 This partial-inhibition approach has already been successfully translated into clinical research.47 Finally, Moerman et al.48 described the very interesting idea of combining [11C]-N-desmethyl-loperamide with the injection of different doses of AEDs in healthy rats for evaluation of potential transporter modulation. The extension of this approach to animal models of pharmacoresistance might be valuable.
Figure 14.2. Changes in P-glycoprotein function and expression two days after status epilepticus (SE) in rats. (A) Coronal, horizontal, and sagittal (R)-[11C]verapamil PET brain images (0–60 minutes) recorded 120 minutes after administration of 3 mg/kg tariquidar, i.e., half-maximal P-glycoprotein inhibition, in control and post-SE rats. White arrows indicate obvious differences in brain radioactivity uptake in cerebellum and cortical regions. (B) Representative examples of immuno-stained brain sections of a control rat (a, c, e) and a rat 48 hours after SE (b, d, f). P-glycoprotein expression is shown in brain capillaries of third cerebellar lobule (a, b), thalamus (c, d), and hippocampus (e, f). (C) Correlation analysis between SE-induced changes in P-glycoprotein expression and changes in compartmental-model-derived efflux rate constant k2 relative to control group in five different brain regions after partial P-glycoprotein inhibition.
Imaging of transporter expression rather than transporter function requires the availability of nontransported radio ligands. If such a tracer could be developed, increased brain signal would be indicative for increased transporter expression. P-glycoprotein inhibitors like tariquidar, elacridar, and laniquidar were described to bind to the transporter molecule at a nontransport binding site.49 Surprisingly, during preclinical evaluation of both [11C]tariquidar and [11C]elacridar it became apparent that both inhibitors are transported in tracer levels.50 The same is true for [11C]laniquidar, which also shows rather low brain uptake values that increase after P-glycoprotein inhibition by cyclosporine A, but, interestingly, not by the second-generation P-glycoprotein inhibitor valspodar.51 This result is indicative of transport of [11C]laniquidar, whereas it is unclear which other transporters might be involved. For all radio labeled transporter inhibitors, efforts to evaluate their potential for imaging of transporter expression are ongoing. Considering their unclear interaction with transport proteins, it is doubtful whether they can provide significant additional value over other transported tracers.
Most of the studies in epilepsy models have been performed using high-affinity substrate tracers. Two studies were performed as proof-of-concept studies using (R)-[11C]verapamil PET shortly after a status epilepticus, which often acts as epileptogenic insult in rodent epilepsy models.44, 52 Syvänen et al.52 performed PET scans with and without transporter inhibition 7 days after kainic-acid-induced status epilepticus in rats. Unfortunately, at this time point P-glycoprotein levels were not changed, and therefore it was not very surprising that (R)-[11C]verapamil uptake levels were also unchanged. Due to the known time profile of early P-glycoprotein overexpression in the pilocarpine post-status-epilepticus rat model,53 Bankstahl et al.44 performed their study at a time point more close to status epilepticus, i.e., after 48 hours, and after half-maximal inhibition of P-glycoprotein (Figure 14.2). During early epileptogenesis, an overexpression of P-glycoprotein in hippocampus, thalamus, and cerebellum measured by immunohistochemical analysis significantly correlated with increased efflux rate constant k2. Of course, at this time point one cannot gain information about pharmacoresistance in chronic epilepsy. Therefore, more elaborate models are needed, in which AED responders and nonresponders undergo PET imaging. Bartmann et al.54 performed PET scans in chronically epileptic phenobarbital responders and nonresponders (BLA model) using [18F]MPPF. This study could not reveal any differences in influx rate constant K1 or efflux rate constant k2 without transporter inhibition. After administration of 5 mg/kg tariquidar, clear differences between responders and nonresponders became apparent. Interestingly, K1 was significantly increased and k2 significantly decreased in nonresponders as compared to responders while decreased influx and increased efflux would be expected. Although the authors explained this with a “greater tariquidar sensitivity of K1,” other factors like changes in 5HT1A receptor expression, the target of [18F]MPPF, might have influenced the results. Syvänen et al.55 performed another study in the same responder/nonresponder rat model using [11C]quinidine and [11C]laniquidar without and with complete transporter inhibition. Both [11C]quinidine, another radiolabeled inhibitor for P-glycoprotein and [11C]-laniquidar revealed, without inhibition, significantly higher hippocampal levels in nonresponders early after tracer injection as compared to controls, but not in responders. It was suggested that transporter inhibitors at tracer levels might bind at least to one transporter and another nontransport binding site at transport proteins.50 This could explain why in addition to obvious transport of the radiolabeled inhibitors, increased signal without transporter inhibition might still be indicative for increased transporter expression, and is in very good agreement with data from Müllauer et al.56 showing a negative correlation between uptake of the radiolabeled P-glycoprotein inhibitor [11C]tariquidar and volume of distribution Vt of (R)-[11C]verapamil after half-maximal inhibition of P-glycoprotein. Notably, the difference shown by Syvänen et al.55 disappeared at 40 minutes after tracer injection. Furthermore, after complete P-glycoprotein inhibition, [11C]quinidine revealed significant differences between responders and nonresponders suggesting the involvement of other, not P-glycoprotein-related mechanisms. Further studies including alternative approaches for imaging transporter function as described above are still needed to elucidate the most promising method for translation to patients.
14.5 Imaging Drug Target Alterations
Changed expression of one or more drug targets might contribute to insufficient drug response in part of epilepsy patients. For most of the AEDs, the main seizure-preventing mechanism is relatively well understood.2 While various radio tracers have been shown to visualize epilepsy-related brain changes in rodent models, like in dopamine receptors using [18F]fallypride,57 in glucose utilization using [18F]fluor-deoxy-D-glucose,58, 59 or of neuroinflammation using TSPO ligands,60, 61 only few radioligands are available for AED targets. These include [18F]flumazenil and [11C]mephobarbital targeting the GABAA receptor,41, 62 and [11C]phenytoin targeting voltage gated sodium channels.42 Importantly, both [18F]flumazenil and [11C]phenytoin have been recently described as P-glycoprotein substrates in rodents.42, 63 Therefore, data analysis has to be performed very carefully, as brain uptake might be confounded by epilepsy-associated changes in transporter activity at the blood-brain barrier. To date, only [18F]flumazenil has been evaluated in epilepsy models with regards to neuro-receptor expression. Liefaard et al.64 could show a reduction of GABAA receptor density of 36% measured by [18F]flumazenil PET in the amygdala kindling rat model, but did not distinguish between AED-responding and nonresponding animals. In parallel, the pharmacological total volume of distribution VBr in the brain was increased by 78% suggesting a kindling-induced increase of transport at the blood-brain barrier. Syvänen et al.65 used [18F]flumazenil PET during epileptogenesis 7 days after a kainate-induced status epilepticus and before the development of spontaneous seizures. They could show a 12% decrease in GABAA receptor density in kainate-treated rats (Figure 14.3). They also evaluated potential epileptogenesis-associated differences in P-glycoprotein-mediated transport of [18F]flumazenil at the blood-brain barrier, but did not find any changes,52 which is consistent with unaltered P-glycoprotein expression at this time point described by the same group.52 As yet, studies in chronically epileptic rats with or without selection of AED responders and nonresponders are not published.
Figure 14.3. [11C]Flumazenil PET quantification after a dose of approximately 4 mg of flumazenil in hippocampus in control rats (diamonds) and rats 7 days after kainate-induced status epilepticus (triangles). Open and closed symbols represent scans before and after P-glycoprotein inhibition by tariquidar treatment, respectively. Error bars indicate SD. Flumazenil concentrations were calculated by pharmacokinetic modeling and were lower in kainate-treated rats than in controls, both before and after tariquidar treatment. Originally published in J Nucl Med. Syvänen S, Labots M, Tagawa Y, et al. Altered GABAA receptor density and unaltered blood-brain barrier transport in a kainate model of epilepsy: an in vivo study using 11C-flumazenil and PET. J Nucl Med. 2012;53:1974–83.
Multiple approaches for radio-isotope imaging of mechanisms of drug resistance are available. Many of these protocols have been tested in rodent models of experimental epilepsy. Nevertheless, to date, only limited evidence is reported evaluating their suitability to distinguish between AED responding and nonresponding animals, but further studies are currently performed. Furthermore, there is an urgent need to further characterize available models of drug-resistant epilepsy, as complex mechanisms finally leading to drug resistance are only partially understood. For example, Bogdanovic et al.61 showed higher brain uptake of the neuro-inflammation tracer [11C]PK11195 in phenobarbital-resistant rats as compared to responsive animals, which might be related to a higher seizure frequency in nonresponders. To better judge the translational value of animal models of pharmacoresistance, ideally, data from pharmacoresistant and pharmacoresponsive patients would be needed for comparison.