The shortcoming of using glibenclamide in exploratory clinical headache provocation studies

Introduction

The vascular ATP-sensitive potassium (K_ATP) channel is recognized as a promising therapeutic target for the treatment of migraine. This channel is expressed in migraine-related anatomical structures and basic neuroscience has revealed that stimulation of vascular K_ATP channels activate and sensitize trigeminovascular neurons. Furthermore, recent human research has demonstrated that K_ATP channel activation triggers headache and migraine attacks associated with dilation of cephalic arteries (1). These observations have generated hypotheses about potential molecular mechanisms of action that implicate vascular K_ATP channels as a downstream target in the calcitonin gene-related peptide (CGRP) and pituitary adenylate cyclase-activating peptide (PACAP) signaling pathways, and hereby as a mediator of migraine pathogenesis. Preclinical studies have reported that the K_ATP channel inhibitor glibenclamide reverses vasodilation caused by CGRP and PACAP and attenuates trigeminal pain transmission in models of provoked migraine-like pain (2). To test whether these findings translate to humans, glibenclamide has been tested in a series of exploratory clinical studies. However, from these intervention studies, it was reported that there were no cerebrovascular effects of glibenclamide (3) and, furthermore, that glibenclamide had no effect on the triggered headache induced by levcromakalim, CGRP, or PACAP administration in healthy volunteers (4–7).

Glibenclamide is a second-generation sulfonylurea drug widely used in the treatment of type II diabetes mellitus. It is a K_ATP channel inhibitor that targets the regulatory sulfonylurea receptor (SUR) subunits, but with a higher affinity for the SUR1 subtype that is expressed in the pancreas. It is well established that sulfonylureas such as glibenclamide bind tightly to plasma proteins (8) and, as a consequence, the effective (free) concentration in plasma is much lower than the total concentration. It could therefore be speculated whether the free plasma concentration of glibenclamide obtained in the above-mentioned clinical headache provocation studies was sufficiently high to ensure target engagement. The assessment of in vivo target engagement plays an important role in drug discovery and development. It reflects the fraction of an endogenous target population that is occupied by the drug at a specific dose and time after administration to patients. To evaluate the degree of K_ATP channel inhibition following dosing of glibenclamide as used in the clinical headache provocation studies, we have applied simple mathematical modeling of glibenclamide pharmacokinetics and receptor potencies to simulate K_ATP channel occupancy levels of the pancreatic Kir6.2/SUR1 channel and of the migraine relevant K_ATP channel subtype Kir6.1/SUR2B, which is expressed in the smooth muscle cells of the vascular system, where it is involved in controlling the vascular tone.

Methods

Pharmacokinetics model

We applied a standard one-compartmental model of a single PO drug administration:

C = F * D * k a V d * k a − k e * exp ⁡ − k e * t − exp ⁡ − k a * t

(1)

where C is the total plasma concentration, F is the bioavailability, D is the dose, Vd is the volume of distribution, ka is the absorption constant, ke is the elimination constant and t is the time after dosing. The model was parameter-adjusted (Table 1) to obtain the best possible representation of the literature pharmacokinetics data following single oral dosing of 10 mg of glibenclamide to non-insulin dependent diabetic patients, as reported by Coppack et al. (9). The model was based on the following reported data: peak concentration in plasma (C_max; 245 ng/ml), the time to C_max (t_max; 3.55 h) and the area under the glibenclamide concentration vs. time curve from time 0 to 24 h (AUC_0–24; 1999 ng/ml/h). Glibenclamide is extensively bound to plasma proteins with literature reporting human plasma protein binding values that range between 98.5 and 99.9% (i.e. free fractions equal to 0.015 and 0.001, respectively). Simulations were performed using both values to cover the maximal and minimal free fractions reported in the literature.

OPEN IN VIEWER

Table 1.

Parameters used for the human receptor occupancy models.

Variable	Value
V d	0.31 l/kg
F	100%
ka	0.40 h–1
ke	0.18 h–1
MW	494 g/mol
IC50 (Kir6.2/SUR1)	0.87 nM
IC50 (Kir6.1/SUR2B)	21 nM

V_d, volume of distribution; F, bioavailability: ka, absorption rate constant, ke, elimination rate constant.

Receptor occupancy models

We recently performed a pharmacological profiling of K_ATP channel modulators at different human K_ATP channel subtypes and found that glibenclamide inhibited the pancreatic Kir6.2/SUR1 subtype with an IC₅₀ value of 0.87 nM and the vascular Kir6.1/SUR2B subtype with a somewhat higher IC₅₀ value of 21 nM (10). The Hill-equation was used to convert the plasma concentrations to receptor occupancy (RO) values:

RO = f p * C I C 50 + f p * C

(2)

where C is the total plasma concentration generated by Equation 1, f_p is the free fraction in plasma and IC₅₀ is the free concentration that results in 50% channel inhibition (i.e. 50% RO). The model was parameter-adjusted (Table 1) to fit concentration–response inhibition of both the pancreatic K_ATP channel subtype Kir6.2/SUR1 and the migraine relevant K_ATP channel subtype Kir6.1/SUR2B.

All simulations were performed using a custom written program based on IGOR Pro (WaveMetrics Inc., Portland, OR, USA).

Results

All clinical headache provocation studies with glibenclamide have been performed using a single oral dose of 10 mg, which is the highest tolerated dose in non-diabetics to avoid uncontrollable hypoglycemia. In these studies, target engagement at the pancreatic K_ATP channel (Kir6.2/SUR1) is evident because glucose administration was necessary to stabilize blood glucose levels during the experiments. However, despite comprehensive analysis of multiple vascular parameters that might serve as biomarkers for Kir6.1/SUR2B target engagement (i.e. cerebral artery velocity, superficial temporal and radial artery diameter, facial skin blood flow, heart rate and mean arterial blood pressure), none of these parameters were sensitive to glibenclamide administration (6). We therefore applied mathematical modeling to evaluate target engagement at both Kir6.2/SUR1 and Kir6.1/SUR2B, using quantitative pharmacokinetic and pharmacological properties of glibenclamide. Figure 1 shows the simulated RO profiles at Kir6.2/SUR1 and Kir6.1/SUR2B following administration of a single oral dose of 10 mg of glibenclamide. Simulations were performed using plasma free fractions of both 0.1 and 1.5% to cover the dynamic range of the values reported in the literature for glibenclamide. These values are in good agreement with experimental findings showing a strongly reduced inhibition of Kir6.2/SUR1 by application of serum albumin during electrophysiological studies (8). Assuming a free fraction of 0.1%, the model predicts a Kir6.2/SUR1 occupancy that is above 18% between 1 and 12 h after drug administration, and with a maximal RO of 36% (Figure 1a). Simulations performed using a free fraction of 1.5% yielded occupancy values in the time span between 1 and 12 h that were above 77% and with a maximal RO of 90%. The relatively high predicted Kir6.2/SUR1 RO levels are compatible with the decreased glucose levels observed in the clinical headache provocation studies. Performing the same simulations with Kir6.1/SUR2B resulted in the range of occupancies shown in Figure 1b. In strong contrast to the RO profile for the pancreatic subtype, only low levels of occupancy were obtained at Kir6.1/SUR2B. Hence, a maximal RO value of 2.3% was obtained following a simulation using a free fraction of 0.1%. For the simulation using a free fraction of 1.5%, the corresponding maximal RO was 26% and with occupancy levels above 12% in the 1–12-h time span.

OPEN IN VIEWER

Figure 1.

Predicted human receptor occupancy (RO) vs. time profiles at (a) Kir6.2/SUR1 and (b) Kir6.1/SUR2B, following oral dosing of 10 mg of glibenclamide. The lower dashed lines and upper solid lines represent RO levels using a free fraction of 0.1 and 1.5%, respectively. The shaded areas denote predicted RO levels for all intermittent values of the free fraction.

Discussion

Vascular K_ATP channel inhibition has been suggested as potential novel treatment of migraine based on evidence from preclinical studies of CGRP release, cranial arterial dilation, behavioral models and genetically modified mice (2). The K_ATP channel inhibitor glibenclamide has been tested against levcromakalim- (4,7), CGRP- (5) and PACAP-induced (6) headache in healthy volunteers. However, glibenclamide was found to be ineffective against both hemodynamic changes and headache induction after infusion of all three migraine and headache triggering compounds. The 10-mg dosage of glibenclamide that was used in all the clinical headache provocation studies is the highest well-tolerated dose in non-diabetics because higher doses could lead to severe hypoglycemia as a result of pancreatic SUR1-regulated K_ATP channel inhibition. An important question is therefore whether target engagement was obtained for the migraine relevant K_ATP channel subtype Kir6.1/SUR2B using this dosage.

Despite the advancement of positron emission tomography and single photon emission computed tomography, direct assessment of occupancy levels in humans can be challenging. For glibenclamide, the high unspecific binding to albumin may represent a special problem, as may also the specific binding to other SUR1-complexes, such as SUR1-TRPM4 (3). As an alternative, mathematical modeling can be used to connect in vitro target potency data with in vivo information such as drug pharmacokinetics and downstream pharmacology. This simplified approach is not sufficient to provide an integrated overview of the complex reactions of a biological system but offers a first rationalized estimate of the impact of drug administration on the targeted receptor. In the present study, we simulated RO vs. time profiles for glibenclamide inhibition of the pancreatic K_ATP channel subtype Kir6.2/SUR1 and the migraine relevant K_ATP channel subtype Kir6.1/SUR2B. Our models predicted maximal RO levels of 36–90% at Kir6.2/SUR1 after glibenclamide administration. By contrast to this, a substantially lower range of occupancy levels were predicted for the vascular K_ATP channel subtype Kir6.1/SUR2B with maximal RO levels between 2 and 26%.

Free Drug Theory (11) indicates that the free (or unbound) fraction of a drug is an important determinant of drug potency because (in general terms) only the free concentration of a drug is available to interact with pharmacological target proteins. For compounds such as glibenclamide, where plasma protein binding is high, medical efficacy can be dramatically reduced. The ability of serum albumin to reduce glibenclamide block of K_ATP channel activity has previously been demonstrated by means of electrophysiological recordings of recombinant Kir6.2/SUR1 channels expressed in Xenopus oocytes (8). In the present study, an IC₅₀ value of 1.2 mM was obtained in the absence of serum albumin, which is in good agreement with that reported for inhibition of whole-cell K_ATP currents in β-cells (12), and also with potency findings that we previously reported using a functional fluorescence-based assay on cloned K_ATP channels (10). However, in the presence of 0.9 mM bovine serum albumin, the IC₅₀ value increased to 1.6 µM, suggesting that the free concentration of glibenclamide was reduced by more than a 1000-fold in the presence of serum albumin.

Underlying all this work is the assumption that RO levels are pharmacologically related to measurable changes in disease status. Also, it is recognized that different receptors will require different levels of RO for efficacy. However, with an estimated maximal RO level at the vascular K_ATP channel subtype that did not exceed 26% for glibenclamide, it is unlikely that sufficient target engagement was reached to abolish the pathways leading to headache in the clinical provocation studies. Also, because levcromakalim acts via a direct activating effect on Kir6.1/SUR2B, we consider that these modeling data are in good accordance with the absence of effect of glibenclamide on levcromakalim-induced pharmacodynamics (3) and headache (4), further substantiating the lack of glibenclamide-mediated Kir6.1/SUR2B target engagement.

Conclusions and perspectives

The opening of K_ATP channels by systemic administration of levcromakalim can induce headache and migraine attacks with and without aura. A convincing amount of preclinical evidence, based mainly on glibenclamide experiments, suggests K_ATP channel inhibition as a promising mechanism for the treatment of migraine, although translation to patients is pending better pharmacological tools. Hence, novel selective Kir6.1/SUR2B inhibitors with good bioavailability and low plasma protein binding are required to realize the full potential of K_ATP channels in this devastating neurological disorder. Because the importance of Kir6.1/SUR2B channels in triggering vs. maintaining episodes of migraine is unclear, we consider that both preventive and acute reversal paradigms (as tried with glibenclamide) need to be revisited.

Clinical implications

Simulated RO vs. time profiles following glibenclamide dosing in clinical headache provocation studies predict low occupancy levels of the vascular K_ATP channel subtype Kir6.1/SUR2B.