This communication describes our efforts towards the identification of a novel, drug-like class of compounds acting as positive allosteric modulators for GABA_B receptors.

In order to introduce molecular diversity in position 5 of the pyrimidine ring, a 4 steps procedure depicted in Scheme 1 was optimized in order to obtain compounds with a chlorine or with a hydrogen in position 6. 4,6-dichloro-2-methylpyrimidine was first substituted by cyclopentylamine and then iodinated to lead to compound 6. This scaffold was then used in a Suzuki cross coupling⁶ to give very efficiently a small focused library of substituted 4-amino-6-chloro-5-phenylpyrimidines (Cpds 7a–15a) which were then hydrogenated under standard conditions to give the desired 4-amino-5-phenylpyrimidines (Cpds 7b–15b). As a means to introduce molecular diversity at the last step in position 4 of the pyrimidine ring, a versatile way of synthesis was designed (Scheme 2). Starting from the commercially available 5-bromo-2,4-dichloropyrimidine, a regioselective nucleophilic substitution was performed in position 4 exclusively to give the 4-methylthio derivative 17⁷ which was then involved in a halogen exchange⁸ to give 2-iododerivative 18. The 2-methyl group and the 5-[(4-trifluoromethyl)phenyl] substituents were introduced by a Negishi cross coupling⁹ and a Suzuki cross coupling,¹⁰ respectively to afford 20. Our first attempt was to oxidize the methylsulfide moiety of 20 to the corresponding methylsulfonyl (21) by treatment with mCPBA¹¹ and then to use 21 as a template. Unfortunately, only few amines reacted cleanly with 21 probably because of the steric hindrance near the leaving group. Our second attempt was to hydrolyze the methylsulfide moiety of 20 with hydrochloric acid¹² to give 23 (Scheme 3). This compound was then reacted with POCl₃ to give the corresponding 4-chloropyrimidine 24. This useful intermediate was submitted to nucleophilic substitutions with a collection of amines to give compounds 25–34. Finally the third library with molecular diversity in position 2 was synthesized starting from compound 16, which was substituted first with exo-2-aminonorbornane¹³ and then with MeSNa to give 36 (Scheme 4). The phenyl moiety was introduced by a Suzuki cross coupling to afford 3737. On one hand, a desulfurization by treatment with Ni-Raney in EtOH¹⁴ leads to 39 and on the other hand 37 can also be oxidized into the corresponding 2-methylsulfonylpyrimidine 38 which was then reacted with various nucleophiles.

Thanks to a preliminary work with nitro-mimetics (Figure 2), we found that to replace the nitro group, we should have a lipophilic substituent (see Cpd 3) with an electron-withdrawing effect (see Cpds 1 and 2). One way to combine both of these effects is to substitute the position 5 of the pyrimidine ring by a substituted phenyl. Moreover, substantial efficacy in the biological assay¹⁵ was observed for compounds with only one cyclopentylamine substituent in position 4 of the pyrimidine ring (compare 1 with 2 and 3 with 4). Furthermore, it was also shown earlier that the replacement of the 2-methylthio substituent by a 2-methyl group is not detrimental for the activity.^3b After screening for GABA_B receptor positive modulatory activity of the first library of compounds (Table 1), we found that the 6-chloro substituent was detrimental to the efficacy at the receptor (compare for example 8a with 8b at 25 µM of compound, 12a with 12b and 15a) with 15b. Moreover, the introduction of a second cyclopentylamino substituent in position 6 led to only weakly active or inactive compounds (data not shown) confirming our hypothesis that the space in the receptor is very limited and that a proton is the best substituent for the position 6 of the pyrimidine ring. On the other hand, as postulated before, electron-withdrawing groups on the phenyl ring such as a 4-trifluoromethyl or a 4-trifluoromethoxy showed the best results (compare 9b with 12b and 10b with 15b). Indeed, the replacement of a 4-methylphenyl (9b) or 4-methoxyphenyl (10b) by a 4-trifluoromethylphenyl (12b) or a 4-trifluoromethoxyphenyl (15b) led to more efficacious positive modulators (Table 1). Other electron-withdrawing substituents were used on the phenyl ring but none of them had increased activity at the receptor compared to the trifluoromethyl or trifluoromethoxy groups (data not shown). The introduction of a substituent in position 3 of the phenyl ring led to a decrease of activity (compare 14b with 15b). To conclude, the best nitro-mimetic group identified in this series was a 4-(trifluoromethyl)phenyl substituent. Then we focused our attention on position 4 of the pyrimidine ring (Scheme 2 and Scheme 3).

This collection of compounds shows interesting structure activity relationships (Table 2). For this series it became obvious that an increased size of the cycloalkyl led to increased efficacy. This trend was first observed with compound 29 in which the cyclopentyl substituent was replaced by a cycloheptyl (compare 12bb and 22 with 29 at 2.5 µM). Interestingly, the introduction of an additional bulky substituent on the cyclohexyl ring (30) led to a less active product (compare 30 with 22). Moreover, the introduction of an aromatic ring onto the amino group was not tolerated by the receptor (compare 31, 32, 34 with 22) therefore we thought that a hydrophobic interaction with a spatially extended substituent was necessary at this position. Surprisingly, despite its 4-tert-butylamino substituent, 33 was found to be a weak positive modulator. The needs of bulky, cyclic aliphatic side chain was then confirmed when cycloalkyl chains were used such as a norbornyl (27), an adamantyl (25),(26) or a cycloheptyl (29). A substantial increase in activity was observed when comparing 29 to 12b and to 22. Despite its good efficacy, 29 was not investigated any further because of significant binding activities to other GPCRs (data not shown). We were then interested in compounds 25 and 26 bearing an adamantyl substituent on the amino group. Both compounds showed were potent positive allosteric modulators (Table 2) which led to the hypothesis that the substituents in position 4 of the pyrimidine ring bind into a large lipophilic pocket in the receptor. However, again these products were not considered for further evaluation despite an increased selectivity profile because of their high logPs (log P > 5.9) and low water solubilities (< 10 mg.L⁻¹). Finally, we focused our attention on 27 which bears a norbornyl group at its amino function. Both the exo isomer (27) and the endo isomer (28) were evaluated (Table 2). Compound 27 was more efficacious, and was of similar potency and efficacy compared to GS39783. Genotoxicity¹⁶ and mutagenicity¹⁷ assays were performed and 27 was found to be safe both in the micronucleus test and in the Ames test. For that reason 27 was considered for further in vitro and in vivo evaluations and the exo-2-norbornyl substituent was kept for the optimization of the position 2.

Finally the screening of the collection of compounds with molecular diversity in position 2 gave surprising results (Table 3). The replacement of the methyl moiety of 27 by an hydrogen led to a less active compound which indicated that a substitution at this position is mandatory. The introduction of a 2-SMe group (37), a methylsulfonyl (38) or a methylamino (42) were detrimental for activity (compare 37, 38 and 4242 with 27). In contrast, the replacement of a 2-methyl group by a cyano (40), a methoxy (41) or a dimethylamino (43) gave compounds with a similar potency than GS39783. Surprisingly, the introduction of a N-methylpiperazin-1-yl (44) reduced the biological activity suggesting that the space in the receptor is limited and that only small substituents are tolerated. Compounds 40, 41 and 43 are currently under evaluation for their physicochemical and pharmacokinetics properties as well as their full pharmacological characterization.