Structure-Kinetic Profiling of Haloperidol Analogues at the Human Dopamine D2 Receptor.

Haloperidol is a typical antipsychotic drug (APD) associated with an increased risk of extrapyramidal side-effects (EPS) and hyperprolactinemia relative to atypical APDs such as clozapine. Both drugs are dopamine D2 receptor (D2R) antagonists, with contrasting kinetic profiles. Haloperidol displays fast association/slow dissociation at the D2R whereas clozapine exhibits relatively slow association/fast dissociation. Recently, we have provided evidence that slow dissociation from the D2R predicts hyperprolactinemia, whereas fast association predicts EPS. Unfortunately, clozapine can cause severe side-effects independent of its D2R action. Our results suggest an optimal kinetic profile for D2R antagonist APDs that avoids EPS. To begin exploring this hypothesis, we conducted a structure-kinetic relationship study of haloperidol and reveal that subtle structural modifications dramatically change binding kinetic rate constants, affording compounds with a clozapine-like kinetic profile. Thus, optimisation of these kinetic parameters may allow development of novel APDs based on the haloperidol scaffold with improved side-effect profiles.

2 other typical APDs, are associated with severe on-target side effects including EPS (e.g., Parkinsonian symptoms such as bradykinesia and tremor) and hyperprolactinemia. 3,4 These symptoms are mediated by blockade of D 2 R signalling in the nigrostriatal and tuberoinfundibular DA pathways, respectively. 2-7 Tardive dyskinesia is also associated with long-term exposure to typical APDs such as 1. 8 Atypical APDs display a diminished incidence of EPS and hyperprolactinemia relative to typical APSs. [9][10][11] While the primary distinction between typicality and atypicality is based on such clinical observations, the mechanism(s) that might drive this distinction remain unclear. Clozapine (2, Figure   1) is a prototypical atypical antipsychotic. It has a complex pharmacological profile with high affinity for other members of the biogenic amine receptor family and, in particular, a relatively high affinity for the serotonin 2A receptor (5-HT 2A R). 12 Many atypical APDs have similar pharmacology leading to the hypothesis that a relatively high affinity for the 5HT 2A R as compared to the D 2 R confers atypicality. [13][14][15][16] However, not all atypical APDs share this profile suggesting that this theory cannot account for all examples of atypicality. 13,17 Unfortunately, this lack of selectivity across aminergic receptors is associated with off-target side-effects, including sedation, metabolic disorders, weight gain, urinary incontinence and constipation. 18 2 can also cause acute agranulocytosis, a potentially life-threatening white blood cell disorder. The relatively fast rate at which 2, and other related APDs, dissociate from the D 2 R has also been suggested to be the basis for an atypical profile. 17 Rapid dissociation of an antagonist might allow a fraction of D 2 Rs to be occupied by transiently high concentrations of DA released into the synapse whereas an antagonist with a slow dissociation rate would cause insurmountable antagonism. Central to this hypothesis was the consensus that APDs exhibit similar association rates (k on ) for the D 2 R meaning that affinity is largely mediated by differences in dissociation rate (k off ). 19 Olanzapine, however, which has a similar high affinity for the D 2 R as many typical antipsychotics, displays an atypical profile. 4 The incorporation of drug-receptor kinetic binding parameters into drug discovery programs is seen as increasingly important for the development of next generation therapeutics. [20][21][22][23][24][25][26][27][28] Previous efforts to derive estimates of APD kinetic rate constants have used radiometric detection methods with limited 3 assay throughput. 19,29,30 We have recently developed a competition association assay using TR-FRET to determine ligand kinetic parameters of unlabelled D 2 R agonists, 31,32 and profiled an extensive series of APDs in order to explore the kinetic basis for on-target side effects. 33 We found that the association rates of the APDs varied over three orders of magnitude and that association rates, rather than dissociation rates, correlated with EPS. These observations led us to propose a revised kinetic hypothesis whereby rapid association rate leads to drug rebinding at the D 2 R, maintaining a higher concentration of APD in the synaptic compartment. This causes increased competition with DA leading to EPS. In contrast, hyperprolactinaemia was correlated with APD dissociation rate. 33 Optimising D 2 R binding kinetics may permit the design of novel tools to test this kinetic hypothesis, as well as facilitate the generation of new APDs with an improved therapeutic profile.
Although clozapine (2) appears to possess the desired slow on/fast off kinetic profile for reduced ontarget side effects (k on = 8.23 ± 1.42 × 10 7 M -1 min -1 , k off = 1.67 ± 0.25 min -1 ), 33 it displays affinity for many aminergic GPCRs, contributing to its off-target side effects. Haloperidol (1), in contrast, has a better off-target selectivity profile, but an undesirable fast on/slow off kinetic profile at the D 2 R (k on = 1.29 ± 0.21 × 10 9 M -1 min -1 , k off = 0.61 ± 0.04 min -1 ) that contributes to its on-target side effects.
The aim of the current study was to optimise the kinetic binding parameters of the more selective scaffold of haloperidol towards a slow on/fast off profile. To this end, we herein describe the design and synthesis of 50 analogues of 1, focusing on structural modification of four key moieties (figure 2) and use competition association kinetic binding methodology to determine their association and dissociation rates, and equilibrium affinities at the D 2 R. We reveal that both the association and dissociation kinetics of this scaffold can vary considerably with subtle structural modification.
Interestingly, we have identified previous analogues of 1, among others, that may have been overlooked on the basis of affinity-driven scaffold optimisation, that possess favourable kinetic profiles. These data reveal the structure-kinetic relationships (SKR) of 1, as well identify novel tool compounds with which to interrogate the relationship between APD kinetic binding parameters and on-target side-effect profiles. Although the structure-activity relationships (SAR) surrounding the butyrophenone scaffold of APDs have been extensively studied in previous years, [34][35][36][37][38][39] to our knowledge, these data represent the first reported SKR relating to analogues of 1.
The synthesis of analogues containing 2,3-difluorophenyl (14c) and 2,6-difluorophenyl (14f) substituents were problematic. When attempting to N-alkylate key intermediate 7b with the corresponding alkyl halides (13c, 13f) major side-products due to a competing S N Ar reaction were observed, making purification of the target compounds by FCC and preparative HPLC extremely challenging. These side-products are believed to arise due to activation of the position ortho to the ketone moiety, when a fluoro-substituent is present. To circumvent the S N Ar reaction, syntheses of the affected analogues were modified to incorporate ketal protection/deprotection of the ketone, permitting nucleophilic displacement of the alkyl halide only (scheme 4). Beginning with ketal protection of 13c, we employed a pTsOH-catalysed reaction with trimethyl orthoformate in MeOH at room temperature, to afford the corresponding dimethyl ketal (15c). Alternatively, 13f was reacted with 1,2-ethanediol in the presence of catalytic pTsOH in toluene under Dean-Stark conditions, to afford the corresponding 1,3-dioxolane (15f). These compounds were then subjected to nucleophilic displacement using 7b to furnish 16c and 16f, followed by acid-catalysed hydrolysis in acetone at reflux, affording final compounds 14c and 14f. Variation of ketone and linker moiety of 1. We focused on replacement of the ketone group of 1 with a range of moieties, including ether, thioether and the corresponding carbinol (racemic). The etherand thioether-variants of 1 were accessed using a literature procedure in three steps 45 (17a-b, Figure   3, Supplementary Scheme 1), whilst the corresponding secondary alcohol was afforded in two-steps also through literature procedure (18, Figure 3) 46 Figure 3. Literature analogues of 1 synthesised using various methodologies. Ether-and thioether analogues 45 (17a-b, respectively); racemic alcohol analogue 46 (18) ; tropanyl analogue 34 (42); 9 piperazinyl analogue 34 (43); phenyl-and p-chlorophenylpiperidine analogues 38,47 (47-48, respectively); reverse substitution analogue 48 (53) ; des-halo analogue 41 (54).
To assess analogues of 1 incorporating internal aromatic alkynes (scheme 9), 1-fluoro-4-iodobenzene (37) was initially subjected to modified Sonagashira conditions 54 using commercially available alcohols (38a, 38c), affording aryl alkynes (39a, 55 39c). Next, the alcohols were converted to their corresponding mesylates (40a, 56 40c), followed by N-alkylation of 7b with the appropriate mesylate to afford the corresponding final propynyl and pentynyl analogues (41a and 41c, respectively). The butynyl analogue 41b was accessed via the cross-coupling reaction between key intermediate 10 and Variation of the piperidinol moiety of 1. Modification to the piperidinol moiety of 1 was another key interest in our SKR investigation. To observe the kinetic effect of introducing an ethylene bridge on the piperidinol, we synthesised tropanyl analogue 42 according to a literature procedure 36  Removing the tertiary alcohol within 1 to generate the corresponding 3,6-dihydropyridine (45) 58 were used to understand the ligand binding pathways of 1 and 2 at the D 2 R/D 3 R. The final stable pose of 1 was shown to occupy the same space as predicted in a number of molecular docking studies; 59-61 however, the molecular orientation was contradictory to these data by 180º, with the butyrophenone moiety buried most deeply in the receptor. Therefore, and due to confounding studies regarding the orientation of 1 at the D 2 R, it was of interest to investigate the kinetic effects of modifying both phenyl moieties of 1 simultaneously. Accordingly, we synthesised a further two structural analogues of 1 ( Figure 3). These modifications included swapping both aromatic termini (53), as well as removal of these aromatic substituents (54). 53 was synthesised according to a literature procedure following Friedel-Crafts acylation and N-alkylation ( Figure 3, Supplementary Scheme 6). 43 Compound 54 was similarly accessed through literature methods ( Figure 3, . 41 14 association rates at six different ligand concentrations (Supplementary Figure 1B). The observed rate of association was related to PPHT-red concentration in a linear fashion (Supplementary Figure 1C).
Kinetic rate parameters for PPHT-red were calculated by globally fitting the association time courses, resulting in a k on of 9.21 ± 0.24 × 10 6 M -1 min -1 and k off of 0. 35 Tables 1-5, and representative competition curves are presented in Figure 4A. In these tables we have separated the analogues into five groups, those that have been modified at the para-chlorophenyl, para-fluorophenyl, piperidinol, ketone/alkyl linker, and concurrent phenyl ring moiety modification, as indicated in Figure 2. Representative kinetic competition curves for selected analogues are in  Tables 1-5. To validate the rate constants, we compared the kinetically derived dissociation constant (K d ) values (k on /k off ) with the dissociation constant (K i ) obtained from equilibrium competition binding experiments ( Figure 5). There was a good correlation between these two values for all compounds tested (two-tailed Pearson's correlation r 2 = 0.99, p < 0.0001), indicating that the parameters determined in the kinetic assay were in agreement with those determined at equilibrium.
Characterisation of the kinetic profile of 1 at the hD 2L R. The equilibrium affinities and kinetic rate constants of 1 and 2 have recently been determined using the aforementioned TR-FRET assay. 33 Prior to initiating an investigation into 1, we also assessed its parameters and determined similar estimates in agreement with literature 33 (k off = 0.61 ± 0.04 min -1 , k on = 1.29 ± 0.21 × 10 9 M -1 min -1 , pK d = 9.31 ± 0.05, Table 1), validating our experimental conditions and further demonstrating that 1 is indeed a high affinity, fast k on /slow k off compound at the hD 2L R. Kinetic estimates for 1 are outlined in Tables   1-5 and all experimental structure-kinetic data will make specific reference to these data as a comparison. Furthermore, compounds with fast k off values approaching >1.0 min -1 were reassessed using a modified injection protocol, whereby the hD 2L R membrane homogenates were introduced 15 using an online injector whilst simultaneously measuring TR-FRET binding. This is to avoid any delay between membrane addition and initial TR-FRET measurement, improving the quality of the non-linear fit for compounds with rapid equilibration kinetics and thus increasing our confidence in the rate parameter estimate. Characterisation of 1 using this methodology returned comparable estimates to the offline injection protocol. Additional data acquired for selected compounds using this methodology are located in Tables 2 and 3.  Table 1). All compounds antagonised the effect of dopamine to a basal (unstimulated) level except for 47 which reduced the effect of dopamine to a level consistent with the intrinsic activity determined in the agonist assay protocol (Supplementary Table 1).
Kinetic effects of variation of the para-chlorophenyl moiety of 1. Initially focusing on modification of the para-chlorophenyl moiety of 1, we sought to assess the kinetic effect of all possible mono (8a, 8n) and di-chlorophenyl substituents (8b-g), as well as variation of the para-substituent (8h-m) through the synthesis of 14 structural analogues (Table 1). These compounds exhibited a 17-fold variation in affinity, which was driven by interesting changes in kinetic parameters, spanning a >10fold variation in association rate (k on = 1.22 ± 0.20 × 10 8 M -1 min -1 to 2.95 ± 0.30 × 10 9 M -1 min -1 ), and a ~4-fold variation in dissociation rate (k off = 0.30 ± 0.01 min -1 to k off = 1.25 ± 0.09 min -1 ).
The data show that analogues lacking an electron withdrawing group (EWG) (chloro) substituent at the meta-and para-positions have reduced binding affinity, and this loss is mirrored by a decrease in k on and an increase in k off relative to 1. For example, the ortho-Cl analogue (8n) displayed an ~8-fold reduction in affinity resulting from a decreased k on and increased k off (k on = 3.54 ± 0.16 × 10 8 M -1 min -1 , k off = 1.16 ± 0.11 min -1 ). This was also evident for the 2,6-diCl analogue (8e) losing ~6-fold affinity, also mediated by a slowed association and increased dissociation rate (k on = 5.07 ± 0.47 × Addition of a strong electron donating group (EDG) (N,N-dimethylamino, 8k) results in a >10-fold decrease in affinity (pK d = 8.12 ± 0.04) and again appears to be driven by a decrease in k on and an increase in k off (k on = 1.64 ± 0.12 × 10 8 M -1 min -1 , k off = 1.25 ± 0.09 min -1 ). Furthermore, other analogues bearing weakly electron donating substituents (e.g. para-tolyl analogue 8i) saw a smaller decrease in affinity (~3-fold), similarly mediated by a change in both rate constants towards a slow on, fast off profile (k on = 1.00 ± 0.06 × 10 9 M -1 min -1 , k off = 0.98 ± 0.02 min -1 ).

18
All compounds bearing an ortho-substituent (8n, 8c, 8d, 8e, 8m), with the exception of 8b, displayed a reduced on-rate, indicating potential sensitivity to steric bulk at this position through resulting rotation of the phenyl group relative to the piperidinol. Interestingly, the 2,3-diCl analogue (8b), contains the privileged 2,3-dichlorophenylpiperidine pharmacophore known to confer high affinity in other molecules at both the D 2 -like and 5HT receptors. This particular substitution pattern may therefore support a different binding mode. Both increased lipophilicity and steric bulk are preferred at the meta-and para-positions of the ring, with the 4-position being optimal, which is supported by 8h (4-fluoro) and 8l (4-H) being less favoured. For the off rate, the substituent effect is reversed in terms of increasing k off (o>m>p). This parameter appears to be less impacted by steric factors, and instead the electronics may play a greater role (8k, 8i, 8h). In summary, these initial data provide insight into how structural modifications of haloperidol (1) impact upon individual kinetic parameters, demonstrating the potential for differential modification of rate constants towards a slow on, fast off profile, depending on the position and nature of the aryl substituents of the 4phenylpiperidin-4-ol moiety.  to pK d = 8.75 ± 0.02 (14l), and is associated with a wide range of association and dissociation rate constants. These losses in affinity are mediated through concurrent changes in both k on and k off . This applies to all but the para-Cl analogue (14k), as it lost affinity 10-fold relative to 1, but this was largely mediated by a decreased rate of association (k on = 1.38 ± 0.05 × 10 8 , k off = 0.70 ± 0.03 min -1 ).

20
The des-fluoro analogue (14l) maintained the highest affinity, and similar to the previous series, this was facilitated by a shift in both rate constants (pK d = 8.75 ± 0.02, k on = 6.33 ± 1.06 × 10 8 M -1 min -1 , k off = 1.12 ± 0.18 min -1 ). Of the three ortho-substituted analogues (14a (m-F), 14i (m-Cl)), 14j (m-CH 3 )), the fluoro substituent was the least favourable in terms of affinity, decreasing ~13-fold relative to 1, whereas the ortho-tolyl substituent only reduces affinity by 6-fold. However, these changes are likewise mediated by a decreased association rate and increased rate of dissociation. Notably, the m-Cl (14i) and m-CH 3 (14j) substituents have similar Van der Waals radii, but very different electronic effects, thus highlighting a steric factor as being important. The meta-fluoro substituted analogue (14b) also dramatically reduced the affinity and was similarly driven by a decreased k on and increased k off (k on = 2.55 ± 0.27 × 10 7 M -1 min -1 , k off = 1.08 ± 0.21 min -1 ). Di-fluoro substitution of the phenyl ring revealed no clear SKR and commonly caused substantial losses in binding affinity. However, unlike the previous chloro series, greater increases in the rate of dissociation were observed. Interestingly, using our online injection protocol, we identified compounds with even slower k on values relative to 2, coupled with equal to or faster k off values, despite their affinities being lower than 2. For example, the 2,3-(14c), 2,4-(14d) and 2,5-difluoro (14e) analogues of 1 (pK d = 7.28 ± 0.04, 6.92 ± 0.05 and 6.85 ± 0.04, respectively) showed dissociation rates faster than any compound identified in the previous series (k off = 1.70 ± 0.09 min -1 , k off = 1.36 ± 0.21 min -1 and k off = 1.49 ± 0.36 min -1 , respectively). In conclusion, these preliminary data suggest that different fluorine substitution patterns dramatically reduce binding affinities, mediated through changes in both kinetic parameters. However, the relationship between the nature of substituents, the substitution pattern and the corresponding kinetic profile is unclear. *Completed using online injection protocol.

22
Kinetic effects of variation of ketone and linker moieties of 1. We next examined the effect of modification to the linker and ketone moieties of 1 through synthesis of a further 15 analogues.
All compounds in this series lost binding affinity relative to 1 and, for the most part, this was mediated through a decrease in k on and an increase in k off . Converting the ketone to its corresponding secondary alcohol (18) (racemic), whilst engendering a 13-fold reduction in affinity compared to 1, was exclusively caused by a slowed k on (pK d = 7.04 ± 0.01, k on = 6.19 ± 0.41 × 10 6 M -1 min -1 ). Replacement of the carbonyl moiety with sulfur (17a) or oxygen (17b) modulated both kinetic binding parameters, though their respective association rates varied ~6-fold (k on = 4.99 ± 0.59 × 10 8 M -1 min -1 and k on = 1.22 ± 0.35 × 10 9 M -1 min -1 , respectively). This difference may be due to a number factors, including the electronegativity and size difference between the sulfur and oxygen atoms, the relatively longer S-C bond length compared to that of the O-C bond and the orbital arrangement around each heteroatom (resulting in considerably smaller bond angles in the thioether compared to the ether).
The trans alkene (23) lost ~10-fold affinity relative to 1, and this was again predominantly due to a decreased k on (k on = 6.39 ± 0.88 × 10 7 M -1 min -1 ). Interestingly, the cis-isomer (26) saw a further 5fold reduction in affinity (pK d = 7.49 ± 0.12); however, this was predominantly due to a change in association rate, displaying a k on almost 20-fold slower and a k off 2-fold faster than 1 (k on = 3.35 ± 0.72 × 10 7 M -1 min -1 , k off = 1.25 ± 0.16 min -1 ). Analysis of the racemic cycloalkane diastereomers was also interesting; introduction of the trans-cyclopropane (29b) resulted in a ~10-fold increase in affinity relative to the parent trans-olefin 23, which was predominantly due to a ~10-fold increase in association rate (k on = 6.07 ± 0.87 × 10 8 M -1 min -1 ). Conversely, introduction of the cis-cyclopropane (30b) had no effects on affinity relative to the parent cis-olefin 28; however, this substituent marginally decreased k on whilst increasing the k off (k on = 4.47 ± 0.47 × 10 7 M -1 min -1 , k off = 1.37 ± 0.09 min -1 ). These data indicate that cis-geometry is preferred as opposed to trans-with respect to this sub-set of compounds in reference to tuning the kinetic profile towards "slow on, fast off" characteristics, and demonstrates the importance of geometry in the corresponding pharmacological profile of APDs. Analysis of the propiophenone and valerophenone analogues of 1 returned further intriguing results. Decreasing the linker length by just one carbon (34a) relative to 1 resulted in 23 dramatic changes in both association and dissociation rate constants (k on = 1.33 ± 0.17 × 10 7 M -1 min -1 , k off = 1.95 ± 0.32 min -1 ), resulting in a loss of affinity at the D 2 R by >20-fold (pK d = 6.84 ± 0.05).
Perhaps the most exciting compound to arise from our study was the valerophenone analogue (34c).
Despite losing affinity by >10-fold relative to 1, this compound displayed a ~10-fold slower k on and a >3.5-fold faster k off than 1. Both the kinetic profile and affinity are similar to that of 2, which our previous studies predict would confer a low propensity to cause extrapyramidal side effects. 33 The alkane analogues of 1 (36a-c) exhibited a 10-fold variation in affinity with respect to one another, with the butylene analogue (36b) found to be optimal in terms of affinity conservation relative to 1 (pK d = 8.17 ± 0.03), despite all having >10-fold losses in affinity relative to 1. Despite having a >20fold lower affinity compared to 1, the propylene analogue (36a) was found to have a "slow on, fast off" kinetic profile (pK d = 7.18 ± 0.06, k on = 2.29 ± 0.15 × 10 7 M -1 min -1 , k off = 1.54 ± 0.07 min -1 ).
Finally, analysis of the 3-5-carbon alkyne analogues (41a-c) saw a 10-fold variation in affinity, with the pentyne analogue (41c) being optimal (pK d = 7.75 ± 0.03), as well as displaying the largest change in both rate constants towards a slow on, fast off profile (k on = 6.41 ± 0.80 × 10 7 , k off = 1.15 ± 0.15 min -1 ). Kinetic effects of variation of the piperidinol moiety of 1. The kinetic effect of structural modifications to the 4-phenylpiperidin-4-ol moiety of 1 was explored through the synthesis of eight additional analogues. We observed the effects of introducing an ethylene bridge (42), as well as modification primarily to the tertiary alcohol through methyl ether formation (52) and its subsequent removal, generating a variety of compounds containing piperazinyl (43), dihydropyridinyl (45), cyclopropyl (46), and piperidinyl (47,48) functionalities. We observed a wide range of affinities that spanned a ~30-fold difference, and unlike the previous chemical series, modification to the piperidinol moiety 25 for the most part had relatively negligible effects on the k off , with the majority maintaining similar values to that of 1 (Table 3). Instead, a decrease in affinity relative to 1 was largely facilitated by a decreased k on . Notably, of the two analogues with higher affinities relative to 1, these were instead largely mediated by an increase in k on and decrease in k off . For example, introducing the tropanyl moiety (42) conferred a ~10-fold increase in affinity which was equally driven by an increase in k on and decrease in k off (pK d = 10.26 ± 0.06, k on = 3.68 ± 0.64 × 10 9 M -1 min -1 , k off = 0.19 ± 0.02 min -1 ).
The cyclopropane variants (46) 5-fold improved affinity relative to 1 was also mediated by an increased k on and decreased k off (pK d = 9.84 ± 0.02, k on = 2.03 ± 0.09 × 10 9 M -1 min -1 , k off = 0.30 ± 0.01 min -1 ). The improved affinities and decreased dissociation rates of 42 and 46 (tropanyl and cyclopropane analogues, respectively) can perhaps be rationalised through a major conformational difference induced by these substituents, resulting in a more entropically favourable binding event.
From these preliminary data, it appears that modification to the piperidinol moiety is not particularly amenable to significant increases in the corresponding compounds rate of dissociation. Dual modifications to both phenyl moieties of 1. Finally, we assessed the effect of swapping the halogen substituents on each end of haloperidol (1) through compound (53), as well as their simultaneous removal as exemplified by the des-halo analogue 54 (Table 5). These structural changes all decreased affinity, which was reflected by decreases in the corresponding k on , with only minor effects on k off relative to 1. Swapping the halogen atoms on each ring (53) caused a 16-fold loss in affinity (pK d = 7.73 ± 0.02), which was predominantly driven by a 16-fold decrease in k on (k on = 4.17 ± 0.28 × 10 7 M -1 min -1 ). Finally, removal of both halogen atoms (54) simultaneously caused a ~18fold loss in affinity (pK d = 7.51 ± 0.01), driven by a sole ~18-fold decrease in k on relative to 1 (k on = 2.28 ± 0.13 × 10 7 M -1 min -1 ). This effect is unlike that of previous analogues bearing a para-halo substituent on only one of the two phenyl rings (8h and 14l), whereby both k on and k off are altered (tables 1 and 2, respectively). Our studies show that modifying the scaffold of 1 produces compounds with a wide range of both association rates (spanning ~3 orders of magnitude, from k on = 3.37 ± 0.62 × 10 6 M -1 min -1 to 3.68 ± 0.64 × 10 9 M -1 min -1 ) and dissociation rates (spanning >10-fold, from k off = 0.19 ± 0.02 min -1 to 2.35 ± 0.19 min -1 ), which constituted large variations in hD 2L R affinities (spanning over three orders of magnitude from K d = 288 nM to 0.0549 nM). To further understand the relationship between kinetic rate constants and the affinity of D 2 R ligands, we have correlated the kinetic binding data of these 50 compounds (k on , k off ) with the derived equilibrium affinity estimates (pK d ) ( Figure 6A). Our data confirms that pK d is robustly correlated with association rate (see Figure 6A, Spearman's r 2 = 0.96, p > 0.0001), whereas pK d is, to a much lesser extent, correlated with dissociation rate ( Figure 6B).
These data are in contrast to previous studies claiming the differences in APD affinities are determined entirely by how fast they dissociate from the D 2 R. 19 This is due to the fact that association rates have widely been assumed to be diffusion limited. Indeed, studies conducted at other systems, namely the M 3 muscarinic acetylcholine and A 2A adenosine receptors, have found correlations between k off values and affinity. [62][63][64] However, the association rate constants of a series of metabotropic glutamate receptor 2 positive allosteric modulators were found to be strongly correlated to affinity, whereas dissociation rate constants were not. 22 This correlation has also been observed at the orexin OX 2 receptor and β 2 -adrenoreceptors for ligands with distinct chemotypes. 65,66 It is evident that modification to the scaffold of 1 and the corresponding changes in affinity are principally mediated by a change in the rate of association ( Figure 6A). Though, our study highlights that particular structural moieties of 1 are more appropriate for the modification of both kinetic parameters towards a "slow on, fast off" profile. For example, when modification to the piperidinol moiety caused a loss in binding affinity relative to 1, this was predominantly k on mediated, whilst  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 28 having negligible effects on k off . However, modification of the p-fluorophenyl or linker moieties and subsequent losses in affinity saw greater changes in both kinetic rate constants, highlighting these areas as a focal point for future SKR investigations. In addition, we were able to derive preliminary SKR for the p-chlorophenyl moiety of 1. From our kinetic data obtained from a limited amount of compound structural/chemical diversity, we determined that both the electronic nature and position of substituents on the aromatic ring dictate the corresponding kinetic profile. We found that metaand para-EWG groups (depending on compound affinity), can either slow the k on whilst having no effect on k off (8a, 8c, 8d), or equally, slow the k on whilst increasing k off (8b, 8f, 8g). Conversely, compounds bearing ortho-Cl substituents and that are not meta-or para-substituted, act to slow the k on but increase the k off (8e, 8n). This is also true for para-EDG substituents at these positions (8h, 8i, 8k). It may be possible to use such molecules as templates in an attempt to further increase affinity via decoration of the aromatic termini, whilst maintaining an 'attractive' or slow on, fast off kinetic profile.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 29 correlation r 2 = 0.96, p = < 0.0001) between these two variables. (B) Conversely, a plot of pK d vs. log k off demonstrates a much poorer correlation (two-tailed Pearson's correlation r 2 = 0.34, p = < 0.0001) despite the traditional scientific consensus that APD affinity is solely driven by changes in k off . (C) The observed association rate (log k on ) and calculated partition coefficient ( The derived association rate of all compounds was further assessed for any potential correlation with physicochemical parameters such as clogP ( Figure 6C) and topological polar surface area (tPSA) (Supplementary Figure 2), to which there was found to be no relationship. This is unsurprising as this study places particular emphasis on the kinetics of not only positional isomers between subsets of compounds, but close structural analogues which display very similar properties of size, lipophilicity and polarity. This further provides evidence that the observed changes to affinity and kinetic profile are not simply due to modification of physicochemical properties. These data are in contrast to previous observations at the D 2 R reporting that compounds with fast dissociation rates are less lipophilic and have lower molecular weights. 67 This is notable as additional micropharmacokinetic/pharmacodynamic mechanisms, such as ligand binding to the cell membrane, are known to play a role in target binding kinetics. 68 Although it is widely accepted that increasing lipophilicity results in increased affinity, this study shows that for this subset of compounds this is not the case, highlighting that careful analysis of kinetic parameters is essential and also likely to be context/target dependent.
Our recent proposal to expand the kinetic hypothesis for APD side effects considers not only the dissociation rate (and therefore the propensity to display insurmountable antagonism), but the association rate and subsequent potential for receptor rebinding. 33 Based on this hypothesis, we proposed three broad classes of APDs in an attempt to explain how different kinetic characteristics have the potential to influence on-target side effects. Class 1: fast on/slow off compounds exemplified  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 30 by haloperidol (1), Class 2: fast on/fast off compounds, namely chlorpromazine, an early typical APD and Class 3: slow on/fast off compounds exemplified by clozapine (2). A fast association rate will result in a higher D 2 R rebinding potential in the striatum and consequently high EPS potential. In contrast, slow dissociation from D 2 Rs expressed on pituitary lactotrophs results in insurmountable antagonism leading to increased prolactin release (e.g. 1). These data suggest that the profile of 1, i.e.
slow k on /fast k off kinetics as exhibited by 2, is optimal for APDs targeting D 2 Rs. Using the scaffold of 1, we have shown that single structural modifications to one of four moieties produces structurally similar molecules with a spectrum of association and dissociation kinetic rate constants ( Figure 6D   Recently, MD simulations have been carried out, attempting to explore the ligand binding journeys of both haloperidol (1) and clozapine (2) at the D 2 R/D 3 R. 58 Interestingly, the binding of 1 at the D 2 R  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 32 has been proposed to arise via a "handover" mechanism, whereby an initial key π-stacking interaction with Tyr 7.35 (Ballesteros-Weinstein numbering scheme) 72 allows this residue to act as a pivot point from which the ligand can explore the extracellular vestibule, followed by formation of a salt-bridge with Asp 3.32 . 58 This mechanism appears to be reliant upon an optimal intramolecular distance between Notably, comparison of the D 2 R and D 3 R crystal structures reveals a considerably different arrangement of the extracellular domains. 69 In order to further our understanding about the SKR reported in our study, our future work will focus on conducting advanced MD simulations using the recently reported D 2 R structure, with a view to correlating how subtle structural changes in the haloperidol analogues (imbuing distinct kinetic profiles) described above might influence interaction with specific residues which line the entry to and exit from the ligand binding site.

 CONCLUSIONS
In this study, we report the chemical synthesis and extensive kinetic profiling of 50 analogues of haloperidol (1) at the hD 2L R, using a TR-FRET competition association kinetic binding assay, permitting the derivation of multiple equilibrium and kinetic parameters (pK i , pK d , k on and k off ). All analogues retained the hD 2L R antagonist action of 1 apart from 47 that gave a partial response relative  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 33 modified variants, incorporating cis-and trans-olefins and their corresponding cyclopropanes, together with numerous alkanes and alkynes. In addition, we investigated the effect of modification to the tertiary alcohol, as well as incorporation of piperazinyl, tetrahydropyridinyl and other piperidinyl moieties. Importantly, we show that there is no correlation between k on and the physicochemical parameters clogP and TPSA, meaning that differences in kinetic profiles and corresponding compound affinities are not simply due to non-specific effects such as cell membrane binding. Moreover, we reveal that k on is significantly correlated with pK d , and is contrary to previous reports at the D 2 R. Thus, we found that a loss in binding affinity is generally associated with a decrease in k on . However, preliminary SKR derived for the p-chlorophenyl moiety of 1, demonstrates that particular substitution patterns and the nature of aromatic substituents are more likely to concurrently decrease k on whilst increasing k off . For example, chloro substituents at the ortho-position modulate the kinetic parameters toward a slow k on /fast k off profile, whereas meta and/or para-chloro substituents can either decrease the k on , whilst having no effect on k off , or, equally, they may also simultaneously decrease k on /k off . The p-fluorophenyl and ketone/alkyl linker structural moieties of 1 were found to be important for mediating changes in both kinetic rate parameters, particularly the k off , whilst the piperidinol moiety was more linked to changes in k on only. For example, converting the aryl ketone to a cis-cyclopropane group or increasing/decreasing the linker length, significantly modulates both rate constants, whereas most modifications to the piperidinol ring simply modulate the k on . We show that with minimal variation this scaffold can be converted to the slow on, fast off kinetic profile that we hypothesise is characteristic of APDs with reduced on-target side effect profiles (e. g. 14c, 30b, 34c, 36a). These compounds may be used as tools to further explore the influence of  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 35 Et 2 O (3  30 mL), and the aqueous phase made alkaline with the addition of 2 M NaOH solution.

General Procedure C.
NaHCO 3 (2 equiv.) followed by toluene. This suspension was then heated at reflux temperature for 24 h. The reaction was filtered and evaporated to dryness followed by direct chromatographic purification using an appropriate eluent as indicated.
2 to derive a single best-fit estimate for k on and k off as described under data analysis. The expression level of the hD 2L R recombinantly expressed in CHO cells was assessed, using [ 3 H]-spiperone saturation binding and determined to be 1.13 ± 0.11 pmol mg −1 protein. 33 Competition binding kinetics. To determine the association and dissociation rates of D 2 R ligands, we used a competition kinetic binding assay recently described to profile the kinetics of a series of D 2 R agonists 31 and antipsychotic drugs. 33  ).