Binding kinetics of cariprazine and aripiprazole at the dopamine D3 receptor

The dissociation behaviours of aripiprazole and cariprazine at the human D2 and D3 receptor are evaluated. A potential correlation between kinetics and in vivo profiles, especially cariprazine’s action on negative symptoms in schizophrenia, is investigated. The binding kinetics of four ligands were indirectly evaluated. After the receptor preparations were pre-incubated with the unlabelled ligands, the dissociation was initiated with an excess of [3H]spiperone. Slow dissociation kinetics characterizes aripiprazole and cariprazine at the D2 receptor. At the D3 receptor, aripiprazole exhibits a slow monophasic dissociation, while cariprazine displays a rapid biphasic behaviour. Functional ß-arrestin assays and molecular dynamics simulations at the D3 receptor confirm a biphasic binding behaviour of cariprazine. This may influence its in vivo action, as the partial agonist could react rapidly to variations in the dopamine levels of schizophrenic patients and the ligand will not quantitatively dissociate from the receptor in one single step. With these findings novel agents may be developed that display rapid, biphasic dissociation from the D3R to further investigate this effect on in vivo profiles.

. D 2 R k off and equilibrium dissociation constants (±s.d.) of the unlabelled ligands. pK i are determined by radioligand displacement assays. D 2s R membrane preparations were incubated with [ 3 H]spiperone and the compound. k off are determined by radioligand dilution assays. D 2s R membrane preparations were preincubated with the compound. Afterwards dissociation was initiated with an excess of [ 3 H]spiperone. Results are means ± s.d. at least performed with n = 3 independent experiments at room temperature.  Table 2. D 3 R k off and equilibrium dissociation constants (±s.d.) of the unlabelled ligands. pK i are determined by radioligand displacement assays. D 3 R membrane preparations were incubated with [ 3 H]spiperone and the compound. k off are determined by radioligand dilution assays. D 3 R membrane preparations were pre-incubated with the compound. Afterwards dissociation was initiated with an excess of [ 3 H]spiperone. Results are means ± s.d. at least performed with n = 3 independent experiments at room temperature. k off displays the k off derived by the monophasic fit, while k off1 and k off2 are the two k off derived by the biphasic fit.  14 . Spiperone, aripiprazole and cariprazine (slow k off ) share a large, lipophilic aromatic motif, connected through a linker to a second motif with different hydrogen-bond acceptor and donor groups (Fig. 2), while rotigotine is missing this diverse motif. In the presented study the antagonist and the partial agonists display a slow dissociation, while the full agonist rotigotine shows fast dissociation from the D 2 R, which is supported by their structural and binding properties. D 3 R kinetic experiments. The evaluation of k off at the D 3 R reveals different dissociation behaviours of aripiprazole and cariprazine ( Table 2, Fig. 3a,b). Cariprazine displays a rapid biphasic dissociation behaviour (p < 0.0001) that is not sufficiently described by a monophasic fit. Dissociation of aripiprazole is slower and better described by a monophasic fit (p = 0.0925). Evaluation of the antagonist spiperone also reveals a slow, but biphasic (p < 0.0001) dissociation, while the agonist rotigotine displays a monophasic (p = 0.0688) dissociation, not significantly faster than aripiprazole or spiperone. The biphasic dissociation of cariprazine is not altered by the  Functional binding assays at the D 3 R. ß-arrestin 2 recruitment assays result in EC 50 values of 23.7 nM for aripiprazole and 10.2 nM for cariprazine (monophasic fit). However, a biphasic fit of cariprazine is more appropriate (p = 0.0004) with EC 50 values of 5.52 nM and 4.19 µM (Fig. 3c,d). Both are partial agonist in the tested system with ca. 30% of E max . A statistically non-significant biphasic binding is observed for cariprazine in radioligand displacement studies with pK i of 8.71 and 11.91. In case of aripiprazole, the biphasic fit is not applicable.  ligand proceeds till it reaches the OBP and forms the H-bond interaction with Asp3.32, which remains intact (10-30 ns). The orientation of the ligand (Fig. 5d) varies compared to cariprazine, but similar interactions were also detected (Fig. 5f). Both H-bonds (Asp3.32, Ile183 ECL2 ) and hydrophobic interactions (e.g. Trp6.48, Phe6.51, Tyr7.43) stabilize its binding mode. Stability assessments and unbinding simulations suggest that aripiprazole might also reorient in the binding pocket to a similar pose as cariprazine.

Molecular
Even though the unbinding simulations offer less reliable structural insight, the trajectories display some characteristic dissociation features. Cariprazine relocates into the SBP after 17 ns (Fig. 4c) and fully dissociates after 21 ns. During the unbinding of aripiprazole (Fig. 4d) we could not observe well defined stages, the higher fluctuations in the RMSD values show less stabilization and the shape of the graph resembles more a one-phase dissociation kinetics.

Discussion
Cariprazine and aripiprazole display similar dissociation behaviours at the D 2 R, but differ at the D 3 R. Aripiprazole exhibits similar RT at the D 3 R and the D 2 R, whereas cariprazine shows the shortest RT and the most pronounced biphasic behaviour of the tested ligands at the D 3 R. This could influence cariprazine's in vivo action, as it can react rapidly to variations in the dopamine level. Recent research revealed that metastable receptor states may determine binding orientation of bivalent drugs in MD simulations 22 and may influence their binding kinetics 23 . Our findings suggest that cariprazine's biphasic properties are connected to a metastable binding pose. Different binding poses may influence a biased agonism at the D 3 R, by stimulating different signalling pathways depending on the natural ligand's concentration. It has to be mentioned that there may be other explanations for the biphasic in vitro behaviour. Given the high affinity and the fast k off from the D 3 receptor, the k obs should be fast, which could result in rebinding effects in the assay. To check this possibility, different concentrations of cariprazine (0.2 and 40 times K i value, n = 1) were evaluated, however, the biphasic nature of its dissociation remained. Although other models may also describe the obtained data, the combination of the experimental data and the MD simulations strengthen a biphasic in vitro dissociation profile of cariprazine and a pronounced interaction with different receptor sites. Various studies have shown that the ß-arrestin pathway at the D 2 R plays an important role in clinical efficacy of antipsychotics. Genetic ß-arrestin depletion led to a loss in antipsychotic activity of tested agents with an increase in motoric ADE 24 . A BRET assay revealed the antagonism of antipsychotic drugs, including aripiprazole, on the D 2 R mediated ß-arrestin recruitment 25 . Given the biphasic nature of cariprazines ß-arrestin recruitment and dissociation from the D 3 R, increases in local dopamine concentration may not lead to a complete displacement of the ligand, but a certain amount will be able to act at the target for a longer period.
A recent review 26 criticizes the use of the RT as sole explanation of a drug's profile and claims for individual evaluation of its impact. In the case of antipsychotic agents however, the correlation between RT and drug profiles seems to be a reasonable hypothesis for different in vivo profiles 9 . Cariprazine displays a different dissociation profile than other tested ligands and shows concurring properties in functional assays and MD simulations at the D 3 R. Its profile differs especially from aripiprazole, also a highly active antipsychotic, only lacking the effectivity on the negative symptoms. Therefore, the characteristic receptor interaction may be one possible cause behind the efficacy of cariprazine on the negative symptoms of schizophrenia.
Although the clinical effects of cariprazine seem to correlate to its binding behaviour at the D 3 R, there may be additional factors. The translation of an in vitro/in silico biphasic behaviour to in vivo effects remains to be elucidated and requires more research on the origin of negative symptoms. However, with these findings, it is possible to rationally develop compounds with biphasic kinetics and characterize them in vivo regarding their effects on the negative symptomatic. Within this work, we show that: • Cariprazine and aripiprazole share similar dissociation properties at the D 2 R • Cariprazine and aripiprazole display different dissociation profiles at the D 3 R • Cariprazine dissociates faster from the D 3 R than from the D 2 R • Cariprazine exhibits biphasic kinetics at the D 3 R • The biphasic behaviour is also observed in functional assays with cariprazine • MD simulations support the experiments with structural interpretation.
Recent work 13 emphasizes the significance of evaluating and comparing binding modes to understand and advance drug development. By elucidating the binding profile of cariprazine the insights on receptor interaction at the D 3 R were broadened and future evaluation of novel antipsychotic drugs with special focus on the negative symptoms has gained an interesting aspect. By this, a next step in the understanding and improvement of antipsychotic therapy was made and future approaches may benefit from specialized agents for this diverse symptom complex.

Methods
Cell culture and membrane preparation of CHO cells expressing the hD 2s R and the hD 3 R. Cell culture and membrane preparations were performed as reported previously with modifications 27 . CHO cells stably expressing the human dopamine D 2short R or D 3 R were cultured in DMEM (with 1% glutamine, 10% FBS, and 1% penicillin/streptomycin for D 2; 1% glutamine, 10% dialysed FBS for D 3 ). CHO-D 2 cells were collected in PBS buffer, CHO-D 3 cells in medium and centrifuged at 3,000 × g for 10 min at 4 °C. The pellet was resuspended in binding buffer (1 mM MgCl 2 , 1 mM CaCl 2 , 5 mM KCl, 120 mM NaCl and 50 mM Tris, pH 7.7), disrupted and centrifuged at 23,000 × g for 30 min (4 °C). The resulting pellet was stored in binding buffer at −80 °C.
Radioligand displacement assays at the hD 2 R and the hD 3 R. Displacement assays were performed as reported previously 28 with modifications. Briefly, membrane preparations (D 2s R: 25 μg/well; D 3 R: 20 µg/well) were co-incubated with [ 3 H]spiperone (0.2 nM) and the test ligand. Nonspecific binding (NSB) was measured with haloperidol (10 µM) and separation of bound radioligand was performed using VE-water. Assays ran in triplicates at least in three independent experiments. Data was analysed using non-linear regression and equation "one site competition". The K i values were calculated from the IC 50 values using the Cheng-Prusoff equation 29 .
Determination of the [ 3 H]spiperone dissociation rate constants at the D 2 R and the D 3 R. The k off of [ 3 H]spiperone (RL) was measured, using an excess of haloperidol 30 . Cell preparations (D 2s R: 25 μg/well; D 3 R: 20 µg/well) were incubated (120 min, 250 rpm) with 0.2 nM RL in 0.2 mL. The dissociation was initiated at different time points with an excess of haloperidol (0.4 mM stock, 15 µL to prevent dilution). Assays ran in quadruplicates with eleven time points with n = 4 at the D 2 R and n = 7 at the D 3 R. NSB was determined with haloperidol (10 μM), total binding with 0.2 nM RL for the total time of the experiments. Bound RL was separated as described above. Binding data was analysed using non-linear regression and fitting to "one phase exponential decay".
Determination of unlabelled ligands dissociation rate constants at the D 2 R and the D 3 R. The k off of unlabelled ligands (UL) were measured indirectly by the dilution method, similar as described previously 30,31 . 50 µL of the membrane preparations (D 2s R: 25 μg/well; D 3 R: 20 µg/well) were incubated with the UL at four times its K i value for 120 min. Afterwards the dissociation was initiated at different time points with an excess of RL (150 µL, 20 times its K D value). This dilution results in the majority of the receptors being occupied by RL at the end of the experiment. As the RL may only bind when the UL has dissociated, k obs of the RL reflects the k off of the UL (Supplementary Fig. S1). To ensure, that the association of the RL occurs without delay to the dissociation of the UL, the concentration of the RL has to be very high (e.g. 20 times K D value). This model assumes that RL and UL bind to the same binding site. Assays ran in triplicates with eight time points with at least n = 3 independent experiments. Total binding was measured in the absence of UL; at the end of the experiments 80-100% of the total binding were achieved (spiperone and rotigotine 70% at the D 3 R). Rotigotine was measured at 5 times its K i value at the D 3 R. NSB and separation are described above. The k off were obtained by applying non-linear regression and fitting to "one phase exponential association" or "two phase exponential association". As this indirect method underlies model theories and assumptions the resulting k off are approximations that serve mainly as comparison criteria. The effects of Gpp(NH)p, the omitting of sodium-ions in the buffer and of using [ 3 H]raclopride as RL were evaluated with the same method (n ≥ 2).

Measurement of ß-arrestin 2 activation at the D 3 R by aripiprazole and cariprazine. The
PathHunter ® ß-arrestin eXpress GPCR assay kit was used to measure ß-arrestin recruitment at the D 3 R (protocol of agonists for aripiprazole and cariprazine). Briefly, U2OS cells were incubated at 37 °C for 48 h. Agonist dilutions were added to the respective wells and incubated for 90 min at 37 °C. Afterwards the detection reagent was added and incubated for 60 min at room temperature in the dark. Luminescence was read with the Infinite 1000 Reader (Tecan). Assays ran in duplicates with n = 2 independent experiments. Data was analysed using non-linear regression and fitting to equations "log(agonist) vs. normalized response -variable slope" and "biphasic". For biphasic fitting "bottom" was constrained to zero, "top" to 100 and nH1/nH2 to 1.
Molecular modelling approaches at the D 3 R. MD simulations were performed on cariprazine and aripiprazole. The initial structures were constructed from the D 3 R X-Ray structure (Protein database entry code: 3PBL) 32 , mutated residues were transformed to their natural form, eticlopride was removed. The protein was embedded into a 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) lipid bilayer, the complexes were solvated in TIP3P waters and neutralized with chloride ions. Parameterization was based on Amber ff14SB force field 33 and restrained electrostatic potential (RESP) 34 charges were calculated for the ligands. All preparatory steps were carried out with the default parameters in the BiKi Life Sciences suite 35 . The structures were minimized and equilibrated following the default protocol of the program package (see details in the Supplementary  Information). Binding studies against the D 3 R were carried out applying BiKi Life Sciences MD Binding tool implemented in GROMACS-4.6.1 36 as published recently 37 . This approach is based on adaptive forces attracting the ligand into the predefined binding site, which residues were identified by NanoShaper 38 and refined by visual inspection (Supplementary Table S1). The additional attractive forces between the ligand and the binding site facilitate the ligand to overcome the barrier of entering the binding pocket; therefore reduce the timescale of the full binding event into a few 10 ns. The bias was switched off, when the ligand reached the 4 Å distance criteria from the backbone heavy atoms of Ser193. During the 20 replica production runs an 0.2 gaining factor was applied in the 30 ns long simulations in isothermal-isobaric (NPT) ensemble at 300 K and 1.013 bar pressure. The binding poses lacking hydrogen bond interaction with the conserved Asp110 were ruled out, the remaining unbiased trajectories were clustered with GROMACS g_cluster tool by the single-linkage method (1.5 Å cut-off) and their stability was tested with scaled MD. The most stable pose was identified as final binding mode and the SCIENTIfIC REPORTS | (2018) 8:12509 | DOI:10.1038/s41598-018-30794-y interaction fingerprints were calculated by IChem 39 with default settings based on every 10 th frame of the corresponding trajectories. The unbinding studies were based on scaled MDs with a scaling factor of 0.4 applying BiKiNetics tool 40 . The final binding mode of the MD binding simulations was the starting pose using structural restraints in canonical (NVT) ensemble at 300 K without the explicit membrane environment. The applied parameters are summarized in Supplementary Table S1. The ligands were considered fully dissociated, when the 6 Å solvation shell around them contained only water molecules.
Data and statistical analysis. In vitro assays were analysed with Prism 6 (GraphPad Software Inc., San Diego, CA) and are given as means ± standard deviation (s.d.). In some instances the number of experiments was increased to characterize binding properties into detail (e.g. biphasic dissociation of cariprazine). This is marked in the tables. The signal of the indirect kinetic experiments and the ß-arrestin assays were normalized (zero was set as zero) to allow comparison of different curve shapes. Significance of biphasic fitting over one-phasic fitting or "biphasic" over "log(agonist) vs. normalized response" was tested with the globalized data sets using the "extra sum-of-squares F Test" provided by GraphPad. Comparisons were considered significant if p-value was <0.05 (α = 0.05). All in vitro assays were performed at room temperature. Cariprazine was synthesized at RCNS, Hungary. The authors confirm the identity and purity of the given compound. The ß-arrestin kit was purchased from DiscoverX (Birmingham, UK).