Antipsychotics with similar association kinetics at dopamine D2 receptors differ in extrapyramidal side-effects

The recent publication in Nature Communications by Sykes et al.1 made the interesting observation that forward binding rate constants (kon) at the dopamine D2 receptor (D2R) differ widely between antipsychotics, and moreover, that kon rather than koff (dissociation rate constant) values correlate with their liabilities to produce extrapyramidal symptoms (EPS). A high kon is postulated to cause a high degree of rebinding of the antipsychotic to D2R, conferring increased competition with endogenous dopamine, and thereby more EPS. As pointed out by the authors, their results are largely in agreement with our previous findings from a potassium channel (G protein-activated inward rectifier potassium; GIRK) activation assay2. However, there are some discrepancies between their work and ours which may have important consequences for the interpretation of the correlation between kon and EPS, and which we would like to point out below. 
 
Remoxipride, a clinically effective antipsychotic which was withdrawn due to the occurrence of aplastic anemia as a rare side-effect, has a very low rate of EPS compared to the typical antipsychotic, haloperidol3 (odds ratio for rigidity, 0.30; 95% CI, 0.22 to 0.41; n = 1104; P < 0.0001, Fisher’s exact test; Fig. 1). Importantly, unlike many other antipsychotics which target multiple types of monoamine receptors, both of these compounds preferentially bind to dopamine D2-like receptors4. Thus, the difference in EPS liability between remoxipride and haloperidol cannot easily be attributed to differential engagement of muscarinic or serotonergic receptors, which are thought to contribute to the low EPS potential of many of the newer antipsychotics, such as clozapine, quetiapine, and olanzapine4–6. This should make remoxipride and haloperidol particularly suitable for comparison based on their kon values at D2R. Interestingly, in our hands, remoxipride has a high kon, similar to that of haloperidol2 (Fig. 1), and a very rapid koff, more than 5 times faster than that of clozapine. In the study by Sykes et al.1, however, remoxipride kon is among the lowest in their dataset, thus fitting their kon hypothesis, while they report a koff for remoxipride which is similar to that of clozapine. It should be mentioned that remoxipride does not induce direct blockade of potassium currents up to a concentration of at least 300 µM, thus ruling out a potentially confounding effect of direct GIRK channel block on the measurement of D2R binding kinetics in our assay7. 
 
 
 
 
Open in a separate window 
 
 
Fig. 1 
 
 
Correlation plot showing the relationship between kon (from Sahlholm et al.2) and EPS odds ratio (from Leucht et al.8 and Lewander et al.3). The plotted antipsychotics are, from left to right: quetiapine, clozapine, olanzapine, amisulpride, asenapine, risperidone, paliperidone, chlorpromazine, haloperidol (Hal; turquoise), and remoxipride (Rem; red). ****p < 0.0001; p value calculated using Fisher’s exact test on the basis of data from Lewander et al.3 Vertical error bars represent credible intervals for the EPS odds ratios from the Bayesian hierarchical model by Leucht et al.8 (212 studies including 43049 patients), except for remoxipride, where the error bars represent a confidence interval calculated by extending the dataset of Leucht et al.8, by incorporating data from Lewander et al.3 (437 patients receiving haloperidol and 667 patients receiving remoxipride), as described in Methods. Horizontal error bars represent S.E.M. of the kon estimates (based on data from 9 to 30 oocytes for each antipsychotic) and reflect the uncertainty of the linear regression of data from Sahlholm et al.2

The recent publication in Nature Communications by Sykes et al. 1 made the interesting observation that forward binding rate constants (k on ) at the dopamine D 2 receptor (D 2 R) differ widely between antipsychotics, and moreover, that k on rather than k off (dissociation rate constant) values correlate with their liabilities to produce extrapyramidal symptoms (EPS). A high k on is postulated to cause a high degree of rebinding of the antipsychotic to D 2 R, conferring increased competition with endogenous dopamine, and thereby more EPS. As pointed out by the authors, their results are largely in agreement with our previous findings from a potassium channel (G protein-activated inward rectifier potassium; GIRK) activation assay 2 . However, there are some discrepancies between their work and ours which may have important consequences for the interpretation of the correlation between k on and EPS, and which we would like to point out below.
Remoxipride, a clinically effective antipsychotic which was withdrawn due to the occurrence of aplastic anemia as a rare side-effect, has a very low rate of EPS compared to the typical antipsychotic, haloperidol 3 (odds ratio for rigidity, 0.30; 95% CI, 0.22 to 0.41; n = 1104; P < 0.0001, Fisher's exact test; Fig. 1). Importantly, unlike many other antipsychotics which target multiple types of monoamine receptors, both of these compounds preferentially bind to dopamine D 2 -like receptors 4 . Thus, the difference in EPS liability between remoxipride and haloperidol cannot easily be attributed to differential engagement of muscarinic or serotonergic receptors, which are thought to contribute to the low EPS potential of many of the newer antipsychotics, such as clozapine, quetiapine, and olanzapine 4-6 . This should make remoxipride and haloperidol particularly suitable for comparison based on their k on values at D 2 R. Interestingly, in our hands, remoxipride has a high k on , similar to that of haloperidol 2 (Fig. 1), and a very rapid k off , more than 5 times faster than that of clozapine. In the study by Sykes et al. 1 , however, remoxipride k on is among the lowest in their dataset, thus fitting their k on hypothesis, while they report a k off for remoxipride which is similar to that of clozapine. It should be mentioned that remoxipride does not induce direct blockade of potassium currents up to a concentration of at least 300 µM, thus ruling out a potentially confounding effect of direct GIRK channel block on the measurement of D 2 R binding kinetics in our assay 7 .
Plotting our own k on data against EPS odds ratios taken from Leucht et al. 8 yields a similar positive correlation ( Fig. 1) as that reported by Sykes et al. 1 However, considering the kinetic properties of remoxipride found in our study, along with its low EPS liability, the correlation between k on and said side-effects becomes less striking. We have inserted data for remoxipride vs. placebo (odds ratio, 1.43; 95% CI, 0.96 to 2.13; see Methods), which was not included in the data set presented by Leucht et al. 8 , based on the odds ratio for rigidity given above. When analyzing data for the 9 compounds which were included both in our study 2 and that of Leucht et al. 8 , the correlation is strong (Spearman's R s = 0.93; P < 0.001), but is weakened by the inclusion of remoxipride (Spearman's R s = 0.70; P = 0.03).
The reasons for the discrepancies in remoxipride k on and k off between our study 2 and that of Sykes et al. 1 may relate to the differential kinetics of the probe ligands employed in the two studies-as pointed out in a previous publication by the same authors 9 , the kinetics of fast-dissociating ligands cannot by accurately measured by a much slower-dissociating probe ligand. It is thus relevant to note that the dissociation T 1/2 of PPHT-red, the fluorescent ligand used to probe receptor-ligand interactions, is about 80 s (k off = 0.52 min −1 ) 1 , whereas the kinetics of the dopamine response in our potassium channel assay is considerable faster, with a deactivation T 1/2 of just below 3 s and an activation T 1/2 below 1 s 10 . For comparison, the T 1/2 which we measured for remoxipride was 3.4 s 2 , whereas Sykes et al. 1 reported it as 0.36 min (21.6 s). Thus, estimates of the very rapid binding kinetics of remoxipride may have been distorted by the relatively slow kinetics of PPHT-red. In addition, the differences in kinetic parameters measured for remoxipride may also reflect the differences in receptor state-whereas our assay measures the functional, high-affinity state of D 2 R in living cells, the study by Sykes et al. 1 used GppNHp to convert all D 2 R in their membrane preparation to the non-signaling, low-affinity state. We consider that drug interactions with the functional state of the receptor might be more relevant to the occurrence of side-effects. Whereas dopamine binding to D 2 R is known to be sensitive to the transmembrane voltage, such that the EC 50 for dopamine-induced GIRK activation is about 4.2-fold higher at 0 mV compared to −80 mV 10 , we found the IC 50 of remoxipride to inhibit the GIRK response to 100 nM dopamine to be 4.8-fold lower at 0 mV (IC 50 202.2 nM; pIC 50 6.69 ± 0.11; n = 5) compared to −80 mV (IC 50 963.5 nM; pIC 50 6.02 ± 0.09; n = 5). The IC 50 value at −80 mV has been published previously 7 . This would be consistent with a competitive interaction between dopamine and remoxipride, where the affinity of remoxipride itself is not substantially altered by transmembrane voltage. Thus, a difference in remoxipride binding kinetics related to differential membrane potentials (the transmembrane voltage across isolated membranes, as used by Sykes et al. 1 , would be 0 mV, while the whole cells used in our study were clamped at a more physiological −80 mV 2 ), would seem unlikely, although a reciprocal change in both k on and k off cannot be ruled out. Lastly, we note that thioridazine, which had been placed in the uncertain typical/atypical category 1 , has among the highest k on values reported by Sykes et al. 1 , yet it has consistently been reported to have a favorable profile in terms of EPS 11 .
In summary, while we find the concept of rebinding, as discussed by Sykes et al. 1 , to be of great interest and of potential utility to the field of neuropharmacology, we believe that k on is not the only factor governing EPS liability, even when considering D 2 R-preferring compounds. This is illustrated by comparing haloperidol and remoxipride, which despite having similar k on at D 2 R in our hands, show very distinct EPS profiles in the clinic.

Methods
Electrophysiology data acquisition. The IC 50 of remoxipride at −80 and 0 mV was determined using two-electrode voltage clamp in Xenopus oocytes expressing the human D 2 R together with RGS-4 and GIRK1/4 channels, measuring the GIRK current amplitude as a readout of D 2 R occupancy by dopamine 2,7,10 . For concentration-response data, oocytes were randomly clamped at either −80 or 0 mV and exposed first to 100 nM dopamine, to evoke GIRK activation. Thereafter, four increasing concentrations (10 nM-10 µM) of remoxipride were applied consecutively, in the continued presence of 100 nM dopamine, at 50-s intervals. The current amplitude at the end of each antagonist application interval was normalized to the control response to 100 nM dopamine, within every oocyte. k on values were also obtained using the GIRK assay, and were taken from our previous work 2 . Oocyte harvesting from female Xenopus laevis toads was performed in accordance with the Guide for Care and Use of Laboratory Animals as adopted and promulgated by the United States National Institutes of Health. Prior approval for the procedure was granted by the Swedish National Board for Laboratory Animals and the local ethics committee; Stockholms Norra Djurförsöksetiska nämnd.
Odds ratio calculation. The odds ratio for remoxipride vs. haloperidol was obtained from EPS (rigidity) incidences and number of patients reported by Lewander et al. 3 The odds ratio for remoxipride vs. placebo was calculated as where R EPS ± , H EPS ± , and P EPS ± denote number of patients treated with remoxipride, haloperidol, and placebo, with our with or without EPS, respectively. Given the relationship above (1), the standard error SE log OR RjP ð Þ propagates according to which was used to calculate the confidence interval for remoxipride. This extension of the data set to include remoxipride did not use the Bayesian hierarchical model employed by Leucht et al. 8 However, the odds ratio calculated here should provide a good estimate for comparison with the antipsychotics in the original analysis.
Statistical analysis. The statistical significance of the differential proportions of patients experiencing EPS between haloperidol-treated and remoxipride-treated subjects (data from Lewander et al. 3