Subtle modifications to a thieno[2,3- d ]pyrimidine scaffold yield negative allosteric modulators and agonists of the dopamine D2 receptor

scaffold agonists of the D2 Abstract: We recently described a structurally novel series of negative allosteric modulators (NAMs) of the dopamine D 2 receptor (D 2 R) based on thieno[2,3- d ]pyrimidine 1 , showing it can be structurally simplified to reveal low molecular weight, fragment-like NAMs that retain robust negative cooperativity, such as 3 . Herein, we report the synthesis and functional profiling of analogues of 3 , placing specific emphasis on examining secondary and tertiary amino substituents at the 4-position, combined with a range of substituents at the 5/6-positions (e.g. aromatic/aliphatic carbocycles). We identify analogues with diverse pharmacology at the D 2 R including NAMs ( 19fc ) with sub- µ M affinity ( 9h ) and, surprisingly, low efficacy partial agonists ( 9d and 9i ). three- and Our data reveal that relatively subtle structural changes can cause a change in pharmacology from that of a NAM to that of a weak D 2 R agonist. With two exceptions, the actions of the ligands described within this study are mediated through interaction with the D 2 R. The phenomenon whereby subtle modifications to a small molecule allosteric scaffold act to modulate modes of pharmacology, presumably via a change in receptor conformation, have been coined as “molecular switches”. 29 This phenomenon has been documented for allosteric ligands targeting multiple GPCR and non-GPCR targets, including muscarinic acetylcholine receptors (mAChRs), 30-32 as well as kinase 33 and phospholipase 34,35 allosteric ligands. Molecular switches have been reported to encompass a number of subtle structural changes, for example, stereochemistry, ring size and simple aryl substitution (i.e. fluoro vs methyl) to afford compounds with diverse pharmacology (e.g. positive and negative alloseric modulators, partial antagonists, and agonists). Indeed it is not surprising that such changes to allosteric ligands can cause dramatic changes in pharmacology given that similarly subtle changes to orthosteric ligands can convert agonists to antagonists. However, for those compounds that displayed agonism rather than the NAM activity of 1 , the nature of this agonism i.e. whether it is non-competitive (allosteric) or competitive (orthosteric), cannot be conclusively confirmed due to their low D 2 R affinity. Thus it is not clear whether such compounds bind to the same allosteric site as 1 , and are examples of molecular switching, or whether the relatively subtle structural changes investigated within this study confer the ability to engage the orthosteric site. 18 Indeed, our recent paper proposed that the allosteric binding site of 1 was in close proximity to the orthosteric site. multiplet doublet doublets doublet triplets (dt).


M A N U S C R I P T
A C C E P T E D ACCEPTED MANUSCRIPT

Introduction
The D 2 R is a class A G protein-coupled receptor (GPCR) implicated in the pathophysiology and treatment of a number of central nervous system (CNS) disorders, including schizophrenia (SCZ).

M A N U S C R I P T
A C C E P T E D ACCEPTED MANUSCRIPT 2 As such, drugs targeting this receptor have traditionally targeted the orthosteric site -where the endogenous ligand DA binds. Clinically used antipsychotic drugs (APDs) targeting this site act as D 2 R antagonists or low efficacy partial agonists, and are termed typical or atypical APDs based on their propensity to cause extrapyramidal side effects. Unfortunately, the efficacy of these drugs is largely limited to the positive symptoms of this disorder. Typical APDs are associated with extrapyramidal motor symptoms (EPS) and hyperprolactinaemia, which are mediated by blockade of D 2 R signalling in the nigrostriatal and tuberoinfundibular DA pathways, respectively. 1-6 Atypical APDs show a reduced incidence of EPS, but display other off-target side-effects mediated through the interaction with other aminergic receptors, including metabolic disorders and weight gain. 7 Alternative approaches to target the D 2 R have emerged via the identification of homo-and heterodimeric complexes. 8 Such complexes in specific tissues may provide novel pharmacological targets for compounds with distinctive functional profiles and improved therapeutic windows. [9][10][11] Another proposition which entails targeting binding sites topographically distinct to the orthosteric site of GPCRs might also be advantageous. Negative allosteric modulators (NAMs) of the D 2 R may represent a safer therapeutic approach for the treatment of SCZ symptoms. A D 2 R NAM with limited negative cooperativity with DA may display antipsychotic efficacy but avoid EPS through partial blockade of the D 2 R akin to the action of atypical partial agonist APDs such as aripiprazole. 12 Furthermore, as a NAM will allow DA to bind, the spatiotemporal pattern of DA signalling is more likely to be maintained. Finally, NAMs may display greater selectivity for the D 2 R over other targets via the targeting of less evolutionarily conserved sites and consequently reduced off target toxicities. Drug-like (NAMs) of DA receptors have recently been identified including the scaffold that is the focus of the present study. [13][14][15][16][17][18][19] Our previous study described the pharmacological validation of a virtual ligand screen hit (1, Figure 1), confirming that it binds with low µM functional affinity and exerts its effect via negative allosteric cooperativity, primarily by modulation of dopamine signalling efficacy. 18 Based on the scaffold of 1, we synthesised a small library of structural analogues to further understand key M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 3 molecular features that were responsible for changes in affinity and negative allosteric cooperativity. Of several promising analogues identified, 2 ( Figure 1) maintained low µM affinity (K B = 1.92 µM) and significantly attenuated DA signalling efficacy (β = 0.001). This compound was devoid of the morpholinomethyl moiety present on 1, and also incorporated replacement of the ((3trifluoromethyl)phenyl)amino-substituent with an N,N-diethylamino motif. In addition to this, our previous work focused on assessing various substituents at the 5/6-positions (e.g. phenyl, cyclohexyl) in place of the fused cyclohexane system. Moreover, we discovered that the scaffold could be structurally simplified, to reveal a low molecular weight fragment-like starting point that maintained µM affinity and negative cooperativity (MW = 207, K B = 4.56 µM, β = 0.13) (3, Figure   1). Thus, we identified an appropriate starting point for further structural interrogation and elaboration of the core scaffold using 2 and 3 as our lead compounds with the aim to identify higher affinity NAMs. To this end, we sought to investigate the influence of varying the nature (both aliphatic and aromatic) of the 4-amino moiety of 2 (series 1, Figure 2). In addition, a second series of compounds was designed to further explore the effect of 5/6-thiophene substitution with respect to altering M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 4 functionality at the 4-position (series 2, Figure 2). This would permit further refinement of the structural determinants of D 2 R affinity and negative cooperativity. Through the parallel synthesis of an additional thirty-seven structural analogues of 2, we have identified molecules that exert a range of modulatory behaviour including, surprisingly, derivatives that display agonism. To confirm a D 2 R-mediated mechanism of action, all compounds with agonist profiles together with selected NAMs were assessed for their ability to modulate our functional readout in the absence of the D 2 R.

Analogues of 2 varied at the 4-(N,N-diethylamino) substituent
Compound 2 was originally synthesised as an analogue of 1 to observe the effect of simultaneously removing the morpholinomethyl moiety whilst incorporating a 4-(N,N-diethylamino) group in in the context of retaining the tricyclic structure of the 5,6,7,8-tetrahydrobenzo [4,5]thieno [2,3d]pyrimidine. As this molecule was found to be a potent NAM of DA efficacy at the D 2 R, an additional series of compounds was synthesised to investigate further changes to the 4-position of the pyrimidine ring (Scheme 1). Preliminary modifications included incorporating 6-membered secondary aromatic and aliphatic amino substituents (e.g. anilino (9a) and cyclohexylamino (9b) together with 6-membered cyclic aliphatic tertiary amine moieties, including piperidino (9c), morpholino (9d) and piperazino (10), followed by the smaller pyrrolidino (9e) to assess the influence of a five-membered cyclic aliphatic system. The piperazino analogue 10 was obtained via M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 6 the S N Ar reaction between commercially available N-Boc piperazine and 8 under microwave conditions as outlined previously (Scheme 1) to afford 9j as a crystalline solid requiring no further purification. Standard trifluoroacetic acid-mediated N-Boc de-protection followed by an alkaline work-up furnished 10. This molecule is unique relative to its counterparts in that it will impart a positive charge at physiological pH. Moreover, we investigated both monocyclic (cyclopropylamino (9f), cyclopropylmethylamino (9g), cyclobutylamino (9h)) and acylic aliphatic secondary amine substituent (tert-butylamino (9i)). The tert-butylamino substituent was installed to assess the effect of an acylic aliphatic secondary amine as a comparison to N,N-diethylamino. These syntheses were all achieved using chemistry as previously outlined (Scheme 1), and successfully afforded ten analogues (9a-i, and 10) in yields varying from 68-88% following S N Ar of 8 with the appropriate amine.

Fragment analogues of 3
We previously reported a low molecular weight fragment (N,N-diethylthieno[2,3-d]pyrimidin-4amine, (3)) that retains low µM functional binding affinity and negative allosteric cooperativity at the D 2 R (K B = 4.56 µM, β = 0.13, Figure 1). 18 This molecule was originally synthesised using a deletion strategy to examine the effect of switching from a 5,6,7,8-tetrahydrobenzo [4,5] [4,5]thieno [2,3-d]pyrimidine moiety of 2 whilst varying the nature of the 4-substituent Our previously reported SAR study of 1 focused on structural modifications to the fused cyclohexane system, including switching from a 5,6,7,8-tetrahydrobenzo [4,5]thieno [2,3d]pyrimidine to the corresponding thieno[2,3-d]pyrimidine, 5-and 6-aryl substitution, 5-and 6cyclohexyl substitution, 5,6-dimethyl substitution, as well as incorporation of fused cyclopentane, cycloheptane, and cyclooctane systems. 18  thiophene, whilst concurrently incorporating the aforementioned 4-amino substituents (Series 2, Figure 2). These compounds would further refine the SAR and provide insight into the design of further high affinity NAMs. These modifications are depicted in Figure 2 and their chemical synthesis is outlined in Scheme 3. All compounds were accessed using established chemistry as detailed in our previous work, beginning with construction of the appropriate thiophene via Gewald 20 chemistry (16a-h), 18 formation of the corresponding pyrmidinones (17a-h) 18 and chloropyrimidines (18a-h), 18 followed by S N Ar with the appropriate amine to furnish a further twenty-four analogues (19aa-hc).  Table 1). 18 The depression in the DA dose-response curve caused by increasing concentrations of 2 is characteristic of the action of a NAM of agonist efficacy and β was fixed to 0.001 when fitting these data to signify high negative cooperativity ( Figure 3A, Table 1).
To our surprise, we identified molecules based on this NAM scaffold that now appear to display agonism. It was important, then, to confirm that these effects are D 2 R mediated. Luciferase assays may identify false positive 'hits' through a variety of mechanisms, for example, by inhibition of the luciferase enzyme. [23][24][25] To allow us to discriminate between D 2 R-mediated activities and any non-specific effects (Supplementary Figure 1) were found to display effects independent of any action at the D 2 R. Importantly, the original VLS hit (1), representative NAMs (2, 9h, 19bb, 19fb, and 19fc), as well as all other agonists exert an effect dependent on the presence of the D 2 R.

Results and discussion
4.1. Functional analysis of 5,6,7,8-tetrahydrobenzo[4,5]thieno [2,3-d]pyrimidine analogues of 2 We first examined the effect of introducing various monocyclic/acyclic aliphatic amines to the 4position of the core scaffold whilst still bearing the fused cyclohexane system. This was based on our finding that 2, bearing a 4-(N,N-diethylamino) substituent, displayed high negative cooperativity. As outlined in Table 1, anilino (9a) and cyclohexylamino (9b) substituents rendered the corresponding analogues functionally inactive. Substitution with the smaller cyclopropylamino substituent (9f) resulted in a modest increase in functional affinity (K B = 0.88 µM) and maintenance of a negative modulatory effect on DA efficacy (β = 0.03, Figure 3C). The cyclopropylmethylamino analogue (9g) maintained a functional affinity (K B = 1.19 µM) similar to that of 2, but more modest modulatory effects upon dopamine efficacy (β = 0.29). Introduction of a cyclobutylamino substituent (9h) not only maintained this negative modulatory action, but acted to increase affinity Figure 3D). Conversely, the tert-butylamino analogue (9i) lost ~50-fold functional affinity relative to 2 (K B = 92.5 µM), but now, surprisingly, exerted agonism (τ B = 0.91). It is interesting to note that introduction of the tert-butylamino moiety (9i) abolished negative cooperativity, whereas the N,N-diethylamino substituent (2) confers a high degree of negative allosteric modulatory effects upon dopamine efficacy. These data demonstrate that the ring size and nature of the substituent at the 4-position are crucial for both affinity and negative modulatory action. We found that larger 6-membered cyclic substituents bearing secondary amines, both of aromatic and aliphatic nature, are not tolerated, whereas three-and four-membered cyclic substituents (cyclopropylamino and cyclobutylamino, respectively) increase affinity and maintain modulatory properties in conjunction with the fused cyclohexane system. We next examined the effect of substituents bearing tertiary amines by incorporating various cyclic amines. The piperidino analogue (9c) decreased the functional affinity by 10-fold (K B = 59.3 µM), and was devoid of any modulatory effect on DA but, surprisingly, appeared to exert an intrinsic response in its own right.
However, our assay using FlpIn CHO cells that do not express the D 2 R revealed that this is effect is most likely an off target effect (Supplementary Figure 1). The morpholino analogue (9d) further decreased the functional affinity, some ~65-fold (K B = 126 µM), but also displayed agonism (τ B = 1.66) with negligible modulatory action on DA an effect that required the expression of the D 2 R ( Figure 3B). Interestingly, isosteric replacement of the morpholine O with NH to give the piperazino analogue (10), significantly changed the pharmacology as concentrations of 10 caused no decrease in DA maximal response, whilst acting to cause a limitless rightward shift in the DA dose-response curve. Such a pattern could be consistent with either very high negative cooperativity with DA affinity, or conversely, a competitive mode of action. Accordingly, these data could also be fit with a model of competitive antagonism (pA 2 = 4.67 ± 0.09, Schild slope: 1.06 ± 0.12, Figure   3E). If this molecule is indeed competitive with DA we speculate that the presence of the piperazino ionisable nitrogen atom at physiological pH may potentially confer greater affinity for the orthosteric binding pocket, consequently converting the pharmacology from allosteric to competitive. Decreasing the ring size by one carbon relative to 9c to give pyrrolidine analogue (9e), maintained functional affinity (K B = 4.23 µM), and displayed robust negative allosteric modulatory effects upon DA efficacy (β = 0.17).

M A N U S C R I P T
A C C E P T E D ACCEPTED MANUSCRIPT Table 1 Functional parameters derived from cAMP BRET assay for analogues of 1 modified with various amines at the 4-position   Table 2). However, this pattern of ligand action is also consistent with that of a competitive low efficacy partial agonist. Furthermore, introduction of a cyclobutylamino substituent (15b) maintained affinity but abolished modulatory effects and engendered agonism (K B = 7.38 µM, τ B = 0.48, Table 2, Figure 3F). Thus, relatively subtle structural modifications significantly change pharmacology, from compounds which negatively modulate the behaviour of DA from a secondary binding site, to molecules that activate the receptor with no cooperativity. Of note, the D 2 R orthosteric agonists pramipexole and sumanirole are both low molecular weight heterocyclic compounds bearing acyclic secondary amines.

M A N U S C R I P T
A C C E P T E D ACCEPTED MANUSCRIPT Table 2 Functional parameters derived from cAMP BRET assay for fragment analogues of 2 with modifications to the 4-position Values represent the mean ± S.E.M. from at least three independent experiments performed in duplicate.

Functional analysis (cAMP) of fused cyclohexane modified analogues of 2
Our previous study revealed that incorporating various substituents (fused cyclopentane,   18,28 However, these interactions were predicted for compounds bearing a fused hydrophobic ring in conjunction with additional substituents located at the 2-position (morpholinomethyl) and 4-position (3-(trifluoromethyl)anilino). As these groups were now altered, we wanted to re-examine the importance of this ring system for such compounds. The N,Ndiethylamino analogue (19aa) displayed an increase in functional affinity (K B = 0.74 µM) relative to 1 and acted to negatively modulate DA signalling efficacy (β = 0.14, Figure 4A). The cyclopropylamino analogue (19ab) maintained low µM affinity (K B = 6.01 µM) and similarly acted as a NAM of DA efficacy (β = 0.28). Conversely, the cyclobutylamino analogue (19ac) displayed a >15-fold decrease in functional affinity (K B = 102 µM) and acted as an agonist (τ B = 0.57). These data were surprising as 9h, the corresponding fused cyclohexane variant of 19ac, displayed sub-µM affinity at the D 2 R and acted as a NAM of DA efficacy.

Larger homologous fused rings (cycloheptane and cyclooctane).
Increasing the hydrophobic ring size by one carbon relative to 1 was previously shown to abolish D 2 R activity, potentially due to the hydrophobic allosteric pocket failing to accommodate such extended ring sizes. 18 Likewise, further expansion of the hydrophobic ring by one additional carbon in the presence of the 3-(trifluoromethyl)anilino substituent also previously rendered the corresponding analogue inactive. 18 In the presence of a    Table 3 and are presented as mean ± SEM from three independent experiments performed in duplicate.

M A N U S C R I P T
A C C E P T E D ACCEPTED MANUSCRIPT 23 Our data reveal that relatively subtle structural changes can cause a change in pharmacology from that of a NAM to that of a weak D 2 R agonist. With two exceptions, the actions of the ligands described within this study are mediated through interaction with the D 2 R. The phenomenon whereby subtle modifications to a small molecule allosteric scaffold act to modulate modes of pharmacology, presumably via a change in receptor conformation, have been coined as "molecular switches". 29 This phenomenon has been documented for allosteric ligands targeting multiple GPCR and non-GPCR targets, including muscarinic acetylcholine receptors (mAChRs), 30-32 as well as kinase 33 and phospholipase 34,35 allosteric ligands. Molecular switches have been reported to encompass a number of subtle structural changes, for example, stereochemistry, ring size and simple aryl substitution (i.e. fluoro vs methyl) to afford compounds with diverse pharmacology (e.g. positive and negative alloseric modulators, partial antagonists, and agonists). Indeed it is not surprising that such changes to allosteric ligands can cause dramatic changes in pharmacology given that similarly subtle changes to orthosteric ligands can convert agonists to antagonists.
However, for those compounds that displayed agonism rather than the NAM activity of 1, the nature of this agonism i.e. whether it is non-competitive (allosteric) or competitive (orthosteric), cannot be conclusively confirmed due to their low D 2 R affinity. Thus it is not clear whether such compounds bind to the same allosteric site as 1, and are examples of molecular switching, or whether the relatively subtle structural changes investigated within this study confer the ability to engage the orthosteric site. 18 Indeed, our recent paper proposed that the allosteric binding site of 1 was in close proximity to the orthosteric site.

Conclusions
In this study, we report the functional characterisation of a small library of structural analogues based on a thieno [2,3-d]pyrimidine scaffold that we have previously shown to act as D 2 R NAMs. 100% solvent B over 20 min at a flow rate of 1 mL/min. All compounds subjected to biological testing were found to be >95% pure by HPLC at two wavelengths (λ of 254 and 214 nm).

General procedure A for the Synthesis of 9a-j, 15a-b, 19aa-hc
In a suitable microwave reaction vessel the required chloropyrimidine (1 equiv.) was taken up in i-PrOH. To this was added the required amine (1.1 equiv.) and the mixture irradiated under stirring at 120 ˚C for 1-2 h. Upon completion of the reaction, the mixture was directly purified using FCC to afford the compound. Similarly, any precipitate could be collected under vacuum and washed several time with cold i-PrOH to afford the desired compound.

cAMP measurement (interaction & off-target assay)
The cellular cAMP levels were measured with the CAMYEL BRET-based biosensor for cAMP. 26 One day after transfection, cells were trypsinised and seeded in white 96-well microplates.

Data analysis
Computerized nonlinear regression, statistical analyses and simulations were performed using Prism 6.0 (GraphPad Prism 6.0b Software, San Diego, CA).

Analysis of functional data
All concentration-response data were fitted to the following modified four-parameter Hill equation where E is the effect of the system, nH is the Hill slope and EC 50 is the concentration of agonist [A] that gives the midpoint response between basal and maximal effect of dopamine or other agonists (E max ), which are the lower and upper asymptotes of the response, respectively.
To determine the mode of interaction of 1 and analogues of 1 at the D 2 R in relation to the agonist dopamine, data were fit to both a competitive and allosteric model and the best fit compared   & 19ca). Reverse-phase analytical HPLC traces and high-resolution mass spectra are included for key compounds highlighted in the TOC graphic (9d, 9h, 9i, 19fc).

Author information
Corresponding Authors