Synthesis and Structure–Activity Relationships of a Series of Pyrrole Cannabinoid Receptor Agonists

— We designed and synthesized a series of pyrrole derivatives with the aim of investigating the structure–activity relation- ship (SAR) for the binding of non-classical agonists to CB 1 and CB 2 cannabinoid receptors. Superposition of two pyrrole-con-taining cannabinoid agonists, JWH-007 and JWH-161, allowed us to identify positions 1, 3 and 4 of the pyrrole nucleus as amenable to additional investigation. We prepared the 1-alkyl-2,5-dimethyl-3,4-substituted pyrroles 10a – e , 11a – d , 17 , 21 , 25 and the tetrahydroindole 15 , and evaluated their ability to bind to and activate cannabinoid receptors. Noteworthy in this set of compounds are the 4-bromopyrrole 11a , which has an aﬃnity for CB 1 and CB 2 receptors comparable to that of well-characterized heterocyclic cannabimimetics such as Win-55,212-2; the amide 25 , which, although possessing a moderate aﬃnity for cannabinoid receptors, demonstrates that the 3-naphthoyl group, commonly present in indole and pyrrole cannabimimetics, can be substituted by alternative moieties; and compounds 10d , 11d , showing CB 1 partial agonist properties. # 2003 Elsevier Ltd. All rights reserved.


Introduction
Plant-derived and synthetic cannabimimetic agents such as Á 9 -tetrahydrocannabinol 1 (Á 9 -THC, 1, Fig. 1) bind to specific G-protein coupled cannabinoid receptors, which include the CB 1 subtype, 2 mainly present in the central and peripheral nervous systems, and the CB 2 subtype, 3 localized on immune cells. The identification, cloning, and biochemical characterization of cannabinoid receptors, 4 along with the discovery of their endogenous lipid ligands, 5 have fuelled a considerable interest in the physiology and pharmacology of the cannabinergic system. 6 Agents that modulate the activity of this system may have a broad therapeutic potential: beside acute and persistent pain conditions, additional therapeutic applications for CB 1 and CB 2 receptor agonists also may include stroke, glaucoma, multiple sclerosis and spinal cord injury. 7 Various classes of compounds active at cannabinoid receptors have been developed, and their structureactivity relationship (SAR) properties have been extensively investigated. 8 Agonists reported in the literature belong to two main classes: (i) dibenzo [b,d]pyrane derivatives (generally referred to as 'traditional' cannabinoids), for example, 1, HU 210 9 (2, Fig. 1) and structurally related molecules, for example CP 55,940 10 (3, Fig. 1); (ii) N-aminoalkyl indoles (AAIs), for example Win-55,212-2 11 (4, Fig. 2) and N-alkylindoles (non-AAIs), for example JWH-007 12 (5a, Fig. 2). A number of compounds having an indene or pyrrole nucleus as their basic feature has been also reported. 8b These comprise Huffman's derivatives JWH-030 (7a) and 7b (Fig. 2). 13 While the SAR of indole cannabimimetic agents have been extensively studied, much remains to be done in the area of pyrrole cannabinoids.
Molecular biology and theoretical studies have provided important insights on the pharmacophoric interactions occurring at cannabinoid receptors. The relatively high CB 1 affinity of the pentacyclic derivative JWH-161 (6, Fig. 2), a hybrid structure in which the elements of traditional and non-AAI cannabinoids are combined, 14 supports the hypothesis that a common pharmacophore exists for the two main classes of ligands. 12 According to this model, the 3-aroyl substituent of indole compounds may mimic the cyclohexene ring of Á 9 -THC (1), whereas the indole N-substituent and the 3-alkyl chain of traditional cannabinoids may engage in lipophylic interactions with the same region of the receptor. This hypothesis was questioned by experiments with site-directed mutated receptors, which suggest that a lysine in the third transmembrane (TM3) domain of CB 1 , K192, is essential for the binding of CP 55,940 (3), but not Win-55,212-2 (4). 15 However, as Huffman and co-workers pointed out, non-AAI were not included in these tests, leaving open the possibility that the N-alkyl group of these molecules may align to the side chain of traditional cannabinoids, whereas the morpholine group of Win-55,212-2 may bind to a different region of the receptor. 14 The CB 2 subtype offers a somewhat different scenario. In this case, K109, a lysine residue corresponding to K192 in CB 1 , may not be essential for the binding of either traditional or AAI ligands. On the other hand, the double mutation K109A/S112G abolishes the binding of Á 9 -THC (1) and 3, but not that of Win-55,212-2. Interestingly, the affinity of the non-AAI compound JWH-015 (5b, Fig. 2) is only partially affected in the singly and doubly mutated receptor. 16 Again, phenylalanine to valine mutation in the TM5 domain, F5.46, decreases the binding of Win-55,212-2. 17 It is worth noting that the CB 1 receptor affinity of Win-55,212-2 is enhanced by substitution of a valine, corresponding to CB 2 F5.46, for phenylalanine. Such mutations, however, did not affect the CB 1 and CB 2 binding of traditional cannabinoids such as HU 210 (2) and CP 55,940 (3). 17 Together with the results of docking experiments, 16,17 these observations suggest that hydrophobic aromatic interactions taking place in a region of the receptor not occupied by traditional ligands may play a crucial role in the binding of Win-55,212-2 and related compounds to CB 2 receptors, while polar interactions through K192 and S112 may contribute to the productive binding of 'traditional' ligands to CB 1 and CB 2 , respectively. The decisive role of aromatic stacking interactions for cannabinoid binding has been supported by two recent investigations. 18 In an attempt to extend Huffman's observations regarding the activity of pyrrole derivatives on CB receptors, we have prepared and tested, on CB 1 and CB 2 receptors, a series of pyrroles with chemical modifications on positions 1, 3, and 4 of the heterocycle. Our compounds were designed assuming that the pharmacophoric interactions occurring at cannabinoid receptors may be modelled by the superposition of the hybrid ligand 6 with JWH-007 (5a) (see Fig. 3a), so that the distal ring of the naphthoyl group of 5a corresponds to the 'ring A' of traditional cannabimimetics.
In particular, the replacement, in compounds 10a,d, 11a of the 3-naphthoyl, the standard C-3 substituent in Huffman's pyrrolic cannabimimetics (see compounds 7a,b, 13 Fig. 2), with a benzoyl group, may give information about the importance of the distal moiety of the naphthoyl group itself. In principle, this group could be replaced by fragments that fit the receptor in a similar manner. Therefore, an N-(2-acetylphenyl)carboxamido fragment, a planar pseudo-bicyclic substructure stabilized by an intramolecular H bond, is attached in position   3 of compound 21. 19 Moreover, it should be possible to place in position 3 substructures able to mimic the 'ring A' of classical cannabinoids, for example the N-cyclohexylcarboxamido group of compound 25 (see Fig. 3c). Lastly, we attempted to substitute the naphthalene moiety for a totally different structural element; therefore, in pyrrole 17 a linear alkyl chain is present bearing an alcoholic function, possibly interacting with the polar site of the receptor to which the 9-or 11-hydroxy residue of traditional cannabinoids are assumed to bind.
Concerning position 4, this could be substituted by a group like bromo (as in 11a-d); these compounds, along with the 4,5,6,7-tetrahydroindole derivative 15, in which an alkyl chain connects positions 4 and 5 of the pyrrole ring, are instrumental to further examine the concept that pyrrole derivatives may be as active as their indole congeners (5). It is commonly accepted that, compared to indole compounds, pyrrole cannabimimetics are endowed with a reduced affinity for CB 1 receptor. This conclusion, however, has been inferred from the analysis of a limited number of structures, namely 4,5unsubstituted pyrroles. 20 Another region of cannabinoid receptors, which is of crucial importance for ligand binding, is the one corresponding in our topographical model to the alkyl chain attached to the pyrrole nitrogen. Compounds in which a group bearing an aromatic ring replaces the typical alkyl chain substituent may provide useful information about the steric tolerance of this region. Therefore, we prepared compounds 10d,e and 11d, whose p-chlorobenzyl N-substitutent is a feature similar to that of diarylpyrazolic CB 2 antagonist SR144528. 8c However, most of our compounds retain an N-pentyl chain, a group known to afford optimal cannabinoid binding to non-AAI and pyrrole ligands; 20,21 an N-propylic chain was inserted in some cases (10c, 11c), since this shortened alkyl fragment is reported to confer CB 2 selectivity to AAI compounds. 22

Result and discussion
Compounds 10a-e and 11a-d were synthesized in a straightforward manner according to Scheme 1. Thus, aroylation of 2,5-dimethylpyrrole (8) in the presence of aluminum chloride, followed by N-alkylation, afforded compounds 10a-e; pyrroles 10a-d were then brominated with N-bromosuccinimide to give 11a-d. The synthesis of 4,5,6,7-tetrahydroindole derivative 15 (Scheme 2), also proceeding smoothly, followed the same route employed by Huffman and co-workers for the preparation of 1-pentyl-3-(1-naphtoyl)pyrrole (7a). 13 Thus, 12 23 was regioselectively acylated, 24 deprotected by alkaline treatment, and eventually alkylated in standard conditions to give 15. Compound 17 was obtained from 2,5dimethyl-1-pentylpyrrole 25 (16), prepared by the already mentioned N-alkylation procedure, and g-butyrolactone in polyphosphoric acid, according to the method of Moussavi et al. 26 (Scheme 3). Compounds 21 and 25 were prepared as outlined in Scheme 4. Thus, 2,5-dimethyl-3-pyrrole carboxylic acid ethyl ester 27 (18), prepared following a literature procedure for the synthesis of the methyl ester analogue of 18, 28 yielded, after hydrolysis, the carboxylic acid 19, 29 which was transformed into the amide 20 via acid chloride. Finally,  N-alkylation of 20 gave the desired compound 21. We originally conceived an alternative synthesis of this compound (Scheme 4); thus, compound 24, a putative immediate precursor of 21, was obtained from 18 by N-alkylation, hydrolysis, and treatment with dicyclohexylcarbodiimide. Even after prolonged heating, however, 24 did not react with 2 0 -aminoacetophenone, giving instead an intramolecular rearrangement to Ncyclohexylcarboxamido pyrrole 25. We also attempted to convert 23 to 21 via acid chloride. Unfortunately, by treating 23 with oxalyl chloride/DMF in CH 2 Cl 2 , followed by 2 0 -aminoacetophenone, we obtained, after workup, a mixture in which no trace of the desired amide 21, and of unreacted starting material 23, was present.
We used radioligand displacement assays to evaluate the affinity of compounds 10a-e, 11a-d, 15,17,21,25 for the native rat CB 1 receptor and the recombinant human CB 2 receptors. The former assay was conducted using rat cerebellar membranes (27,000g), and the latter using membranes of Chinese hamster ovary (CHO) cells that overexpress CB 2 receptors (Receptor Biology Inc. Perkin Elmer, Wellesley, MA, USA) using [ 3 H]Win-55212-2 (NEN-Dupont, Boston, MA, USA, 40-60 Ci/ mmol, 10 nM) as a ligand. 5a A summary of these results is provided in Table 1. N-Pentyl-3-naphthoylpyrroles 10a, 11a, 15, which can be regarded as congeners of Huffman's pyrrole derivative 7a (Fig. 2), retain a moderate to good affinity for CB receptors. In particular, the concomitant presence of a C-2 and a C-5 methyl substituent, a distinctive feature of the present series of pyrrole ligands influencing the conformational equilibrium of the N-alkyl group, appears to be tolerated at CB 1 , and to slightly increase the affinity at the CB 2 , receptor subtypes. This trend, illustrated by compound 10a, which is 20 times more potent than 7a in CB 2 binding affinity, persists over the entire group of our 1alkyl-3-naphthoyl pyrrole ligands which, therefore, albeit having only a moderate preference for CB 2 receptors, show a reversed CB 1 /CB 2 selectivity, compared with Huffman's derivative 7a. Substituents in position 4 cause different effects, depending on their nature: in agreement with the topographic model depicted in Fig. 3b, introduction of a bromo group produces a slight increase in binding affinity for both receptor subtypes (see 10a), whereas an unfavourable effect is produced by the tetramethylene chain linking positions 4 and 5 of compound 15. The substitution of the 3-(1-naphthoyl) group for a benzoyl one is detrimental for affinity (10b,e, 11b), and this is consistent with the previous observation that AAIs with monocyclic aroyl nuclei in position 3 are far less active than their 3-(1-naphthoyl) homologues. 11,30 The replacement of the C-3 naphthoyl substituent with groups of different structural characteristics provides compounds with reduced affinity for the cannabinoid receptors (21,25), or a complete loss of binding (17). In particular, the N-(2-acetylphenyl)carboxamido group of 21 is only in part, and only with CB 1 receptor subtype, able to reproduce the binding mode of the naphthoyl group. A slightly better binding affinity, at least for the CB 2 subtype, was obtained with compound 25. This supports our hypothesis of a cycloalkyl fragment mimicking the cyclohexene ring of classical cannabinoids, according to the model of Figure 3c. Finally, the attempt to reproduce the interactions afforded by the hydroxy group of classical cannabinoids by means of a hydroxyalkyl chain (17) was unsuccessful. The N-propyl derivatives 10c and 11c proved to be less potent than the corresponding N-pentyl analogues 10a and 11a. The decrease of affinity is less marked for the CB 2 receptor, resulting in a certain degree of CB 2 selectivity, which is consistent with literature data, 18,20 even though, in our pyrroles, the CB 2 /CB 1 affinity ratio enhancement caused by chain shortening is not as prominent as that found in prototypic 3-aroyl indole cannabimimetics. 31 Interestingly, the replacement of the N-linear alkyl chain by a substituent of different steric and electronic nature, that is, a p-chlorobenzyl group, yields compounds (10d, 11d) that retain a certain affinity for cannabinoid receptors. Indeed, such derivatives, especially relative to CB 1 binding, are more potent than the short alkyl chain analogues 10c, 11c. The binding mode of 10d and 11d may be somewhat different from that of the remaining N-alkylpyrroles, as suggested by the fact that introduction of a bromo group in position 4 does not increase potency (cfr. 10a vs 10d, and 11a vs 11d).
We also investigated the intrinsic activity of selected high-affinity ligands at CB 1 receptors by testing their ability to stimulate [ 35 S]GTPgS binding in rat cerebellar membranes. Figure 4 illustrates the effects of various pyrrole-based compounds on the binding of [ 35 S]GTPgS to rat cerebellar membranes. The compound 11a stimulates [ 35 S]GTPgS binding with an EC 50 value of 140.3 AE 8.2 nM (mean AE SEM, n=9) and a maximal degree of stimulation (238 AE 18%) identical to that of Win-55,212-2 (Fig. 4). We obtained similar results with compound 10a, which stimulates [ 35 S]GTPgS binding with an EC 50 value of 324.0 AE 20.8 nM. These findings suggest that 11a and 10a are full agonist ligands at rat CB 1 with an efficacy comparable to that of known cannabimimetic agents. 32 By contrast, the compounds 11d (Fig. 4)  In summary, our results extend previously established SARs for indole and pyrrole cannabimimetics. In particular, we identified a 3-naphthoyl pyrrole 11a, which displays a binding affinity and intrinsic activity comparable to that of 3-naphthoyl indoles. This suggests that a suitable lipophylic group, attached to position 4 of the pyrrole nucleus, can compensate for the lack of contribution of the benzo moiety present in indole compounds. The steric and electronic characters of this substituent seem to be strictly defined, as may be inferred from the reduced affinity of 4,5,6,7-tetrahydroindole 15. Compound 11a, the most active of our series, does not display a significant CB 1 /CB 2 selectivity; this parallels, however, the behaviour of N-pentyl cannabimimetic indoles. 31 Furthermore, the testing of several compounds that have structural elements unusual in  heterocyclic cannabinoid ligands provided further insights to the topography of cannabinoid receptor binding sites. Lastly, the fact that the N-chlorobenzyl compound 10d behaves as a partial agonist, indicates that the standard linear alkyl chain of non-AAI can be substituted with groups that interact with an area of the receptor involved in modulating ligand efficacy.

Conclusions
Some of the pyrrole derivatives described in this study exhibit interesting affinity and efficacy profiles for cannabinoid receptors. Functional tests indicate that these compounds may behave as full or partial agonists at the CB 1 receptor subtype, depending on the presence of specific substituents. Further investigations will be necessary to optimize the affinity and efficacy of the present class of compounds, and to explore in more detail their SAR properties.

Experimental Chemistry
All chemicals were purchased from Aldrich in the highest quality commercially available.

Molecular modeling
Three-dimensional models of the molecules were built with Sybyl 6.7 software package 33 and their geometry was optimized using the standard Tripos force field, 34 with the Powell method 35 to an energy gradient of 0.01 Kcal/mol Å , ignoring the electrostatic contribution.