Synthesis, Radiosynthesis and Biological Evaluation of Buprenorphine‐Derived Phenylazocarboxamides as Novel μ‐Opioid Receptor Ligands

Abstract Targeted structural modifications have led to a novel type of buprenorphine‐derived opioid receptor ligand displaying an improved selectivity profile for the μ‐OR subtype. On this basis, it is shown that phenylazocarboxamides may serve as useful bioisosteric replacements for the widely occurring cinnamide units, without loss of OR binding affinity or subtype selectivity. This study further includes functional experiments pointing to weak partial agonist properties of the novel μ‐OR ligands, as well as docking and metabolism experiments. Finally, the unique bifunctional character of phenylazocarboxylates, herein serving as precursors for the azocarboxamide subunit, was exploited to demonstrate the accessibility of an 18F‐fluorinated analogue.


Introduction
Opioid receptors (OR) play an important role in medicinal chemistry [1][2][3][4][5] and continuous efforts are made to develop new drug candidates [1,[6][7][8] as well as ligands designed for investigations in the chemical biology of these receptors. Among the major subtypes μ, [9] k, [10] δ, [11] and the nociceptin receptor [12] the selective addressing of the μ subtype is of interest for a number of applications. [13][14][15][16][17][18] For example, the selective μ-OR agonist PZM21 has been proposed for pain treatment with reduced side effects. [19] PZM21 is not only subtype selective, but also displays biased signaling with minimal β-arrestin recruitment. [20,21] Further analogues of PZM21 were recently reported by Shi and co-workers. [22] The partial μ-OR agonist NAP, which contains a morphinan scaffold, also shows biased signal transduction. [23] NAP was developed for the treatment of opioid-induced constipations. A first-in-human clinical trial was already performed applying the biased μ-OR agonist TRV130. [24,25] Although pain management with reduced side effects such as respiratory depression was observed within this clinical trial, the value of TRV130 has proven to be controversial. [26] Upon optimization of a piperidine benzimidazolone scaffold, promising μ-OR agonists with exceptionally high bias factors have recently been discovered by Bohn and Bannister. [27] Regarding the recently developed OR ligands from a structural point of view, non-morphinan-derived [19,22,24,27] as well as morphinan-derived [23,[28][29][30][31] scaffolds have been used. Among the latter ones, scaffolds based on morphine (1), [32] diprenorphine (2) [33] and β-funaltrexamine (β-FNA) (3) [34,35] still represent valuable starting points in the development of opioid receptor ligands (Figure 1). [36] With regard to subtype selectivity, the particular importance of the side chain on the diprenorphine scaffold can be derived from the binding data summarized in Table 1. [33,34,37,38] When aiming at μ subtype selectivity, the 3-hydroxy substituted ligands diprenorphine (2), 4 and 5 (R 1 = H) do not appear as preferred lead structures since these compounds show limited subtype discrimination. Methylation of the 3hydroxy group (R 1 = Me), however, changes the influence of the substituent R 2 . As demonstrated by the diprenorphine derivatives 6-8, [38] and in particular by the biphenyl derivative 7, a suitably chosen side chain R 2 can now lead to preferred binding to the μ subtype.
For this study, we thought to combine the high binding affinity of the cinnamide 5 [34] towards the μ-OR (K i = 0.7 nM, Table 1) with methylation at the 3-hydroxy group to shift the selectivity towards the μ subtype (Scheme 1). Moreover, the cinnamide substructure should be replaced by a phenylazocarboxamide so that ligands of the general structure 9 would be obtained.
Besides our general interest in the question whether phenylazocarboxamides can serve as bioisosteric replacements for cinnamide substructures, [39,40] the successful preparation of the target compounds 9 from amine 10 and azo esters 11 would further enable a straightforward access to the corresponding 18 F-fluorinated analogue (R = 4-18 F; Scheme 1). This is due to the good availability of the 18 F-labeled tert-butyl phenylazocarboxylate 11 (R = 4-18 F), [41][42][43] whereat the corresponding 18 F-labeled azocarboxamide 9 (R = 4-18 F) could then serve as a μ-OR radioligand for in vivo PET imaging studies. [42][43][44][45][46] For all OR ligands of the general structure 9, non-radiolabeled as well as radiolabeled derivatives, we envisaged the coupling of the amine 10 with the corresponding tert-butyl phenylazocarboxylate 11 as final synthetic step. [41] In addition to chemical and radiochemical syntheses, computational docking studies were carried out to give insights into the factors responsible for binding affinity and subtype selectivity. Besides the potential cytotoxicity, we further investigated the metabolic stability of the novel ligands using an in vitro rat microsomal stability assay.

Results and Discussion
The synthetic route to the azocarboxamide and cinnamide ligands 9 a-d and 16 a-b evaluated in this work is depicted in Scheme 2. For some reactions, the conditions are based on those previously established by Kok et al., [47] and for other steps, the reaction conditions were derived from the work by Derrick et al. [34] The ligand synthesis started with the demethylation of thebaine (12) using ferroceneacetic acid, mCPBA and hydrochloric acid to yield a secondary amine, followed by an Nalkylation of this amine with (bromomethyl)cyclopropane in the presence of NaHCO 3 to obtain the alkylated compound 13. Afterwards, ethyl acrylate was used as dienophile for a hetero Diels-Alder reaction with diene 13 to give ester 14. The ester 14 was hydrolyzed under acidic conditions, the resulting carboxylic acid was activated with oxalyl chloride, and the acid chloride was trapped with benzylamine to furnish the amide 15. The primary amine 10 was prepared from 15 by reduction of the amide moiety by LiAlH 4 and subsequent hydrogenation of the double bond which was accompanied by cleavage of the benzyl protecting group. At this point, it is worth to note that the carbon-carbon double bond was already partially reduced under the LiAlH 4 conditions of step vii). For step viii), the procedure by Derrick [34] recommended the use of H 2 and Pd/C at 45°C under a H 2 pressure of 30 psi (ca. 2.1 bar) to give a yield of 76 %. We initially conducted the hydrogenation under Table 1. Binding affinities of known ligands derived from diprenorphine (2). [33] H 0.14 2.0 0.73 4 [37] H 2.19 � 0.65 4.15 � 1.3 3.66 � 0.92 5 [34] H 0.7 � 0.25 2.6 � 0.0 0.7 � 0.05 6 [38] Me 14.4 0.15 89.3 7 [38] Me 1.91 25.8 1753 8 [38] Me 1.87 0.74 -Scheme 1. Structural modifications of cinnamide ligand 5 and synthetic approach to azocarboxamides 9 from amine 10 and tert-butyl phenylazocarboxylates 11.

ChemMedChem
Full Papers doi.org/10.1002/cmdc.202000180 atmospheric pressure conditions (H 2 balloon), leaving all other reagents and conditions unchanged, and obtained the primary amine 10 in 68 % yield. Even after extended reaction times, only partial conversion of the starting material had taken place, and after the addition of a larger excess of acidic acid, we detected the opening of the cyclopropyl ring via LC/MS analysis. Thus, the conditions were changed to Pd/C and ammonium formate at a reaction temperature of 90°C. Under these conditions, the primary amine 10 was obtained in 81 % yield after 2.5 h, and the desired product could now be easily purified by column chromatography. The synthesis of the azocarboxamide ligands 9 a-d was accomplished by a coupling of the primary amine 10 to the corresponding tert-butyl phenylazocarboxylates 11 a-d in the presence of sodium bicarbonate or triethylamine. Among the four different azocarboxylic esters used in step ix), slightly modified conditions were required for the tert-butyl 4-fluorophenylazocarboxylate (11 b), as the long reaction time of three days led to a partial substitution of the 4-fluoro substituent by an ethoxy group originating from the solvent ethanol. Since ethanol turned out to be the best choice compared to other solvents, the synthesis of the fluorinated ligand 9 b was conducted at increased concentrations and with a larger excess of the azocarboxylate 11 b. In this way, the reaction time could be reduced from three days to 2.5 h and the side reaction to the aryl ethyl ether could be suppressed. In step x), the cinnamide ligands 16 a and 16 b were prepared from amine 10 and the related cinnamic acid chlorides 17 a,b in the presence of sodium bicarbonate. The binding affinities of the azocarboxamides 9 a-d and the cinnamides 16 a,b to the μ-, k-and δ-OR subtypes were determined by radioligand competition binding assays using [ 3 H]diprenorphine and membrane preparations derived from HEK293T cells transiently expressing the related receptor subtypes. Displacement curves resulted in K i values for 9 a-d, 16 a,b and the reference β-FNA (3) ( Table 2).
In the previous biological evaluation of cinnamide 5, [34] which served as a starting point for the design of our azocarboxamide ligands 9 a-d, β-FNA (3) was used as a reference compound (  Table 2. Binding affinities of compounds 3, 5, 9 a-d and 16 a,b towards the μ-, k-and δ-OR subtypes. Compound  [34] K i values determined for β-FNA (3) were all increased by factors in the range of three to 11 (× 4 for the μ-OR, × 11 for the k-OR and × 3.5 for the δ-OR) thereby suggesting that our assay is less sensitive for all three OR subtypes. At this point it is important to remember that β-FNA (3) is known to be an irreversible ligand at the μ-OR, so that the equilibrium binding affinity to this particular subtype can be highly sensitive to deviations in the assay conditions. As the K i values for all three OR subtypes are however altered by comparable factors, and binding of β-FNA (3) to the k-OR and the δ-OR subtype is reversible, a certain comparison of our data with that available for 5 [34] appears possible. Taking into account the above mentioned sensitivity factors (× 4 for the μ-OR, × 11 for the k-OR and × 3.5 for the δ-OR), one can assume that cinnamide 5 and our ligands 9 a-d and 16 a,b show comparable binding affinities at the μ-OR and k-OR subtypes. At the δ-OR, in contrast, the binding of 5 is significantly stronger. Methylation at the 3-hydroxy group, which constitutes the general structural difference between the reference compound 5 and all of our ligands 9 a-d and 16 a,b, thus leads to an improved selectivity profile for the μ towards the δ subtype, but not for μ compared to k. This analysis is also reflected by the selectivity ratios k/μ and δ/μ reported in Table 2.
Among our ligands 9 a-d and 16 a,b neither di-aza substitution nor ring substitution on the azocarboxamide (c.f. Scheme 1) led to strong changes in binding affinity at any of the three OR subtypes, thereby indicating that these structural modifications are tolerated and that no major ligand-receptor interactions are changed. Upon comparison of the newly prepared azocarboxamides 9 a-d and cinnamides 16 a,b, the 4fluoro and the 4-bromo azocarboxamide derivatives 9 b and 9 c showed the most favorable biological profiles with regard to the μ-OR. The highest binding affinities were determined for 9 b and 9 c and the subtype selectivity ratios k/μ and δ/μ for these ligands were also among the best values.
Because a covalent binding mode to the μ-OR was originally considered for the reference compound 5, [34] radioligand depletion assays were performed for the novel azocarboxamide 9 b and its corresponding cinnamide 16 b. Through these assays it turned out that a covalent binding to the μ-OR is neither likely for 9 b, nor for 16 b, whereas 16 b is more closely related to reference compound 5 due to the common cinnamide substructure. Figure 2. Functional investigation of the selected test compounds applying an IP accumulation assay (IP-One ® ) for testing G-protein signaling and an arrestin-2 recruitment assay (Path Hunter ® ). A, C) G-protein signaling was determined in HEK-293T cells transiently co-transfected with μ-OR and the hybrid G protein Gα qi (Gα q protein with the last five amino acids at the C terminus replaced by the corresponding sequence of Gα i ). Functional assays at μ-OR were performed with the selected test compounds 9 b and 16 b applying an inositol phosphate (IP) accumulation assay for G-protein mediated signaling ( IP-One assay®) and an arrestin recruitment assay for receptor stimulated recruitment of β-arrestin-2 (PathHunter assay). In both signaling pathways the cinnamide 16 b revealed as a neutral antagonist while the azocarboxamide 9 b showed antagonist properties for arrestin recruitment but a weak partial agonist effect with an efficacy of 19 % for G-protein mediated signaling ( Figure 2).
These antagonist/partial agonist results could be confirmed by inhibition experiments when the activity of a fixed concentration of the reference agonist DAMGO was fully Due to the potential reactivity of the unsaturated azocarboxamide substructure of 9 b and the cinnamide entity of 16 b we investigated both compounds on cytotoxicity in comparison to the μ-OR reference naloxone. For that we incubated HEK293T cells with 100 nM of 9 b, 16 b or naloxone for 24 hours and determined the number of cells indicating any influence on cell growth resulting in cell densities of 92 � 6 % (n = 6, mean � SEM) for 9 b, 106 � 6 % (n = 6) for 16 b and 84 � 5 % (n = 6) for naloxone relative to the effect of vehicle (DMSO). Complemented by the verification of viability of the cells by optical controls these results reveal no cytotoxic effect of the ligands.
In the next step, the binding modes of the ligands 5, 9 a-d, 16 a and 16 b within the μ-OR, the k-OR the δ-OR subtype were compared (Figures S5-S7 in the Supporting Information) by using available X-ray structures (PDB IDs μ-OR: 4DKL; δ-OR: 4 N6H; k-OR: 4DJH) as templates. All three structures were obtained by co-crystallization with antagonists, which ensures that the receptor geometries represent the inactive state and are therefore directly comparable by geometric superposition. Within the μ-OR subtype, the morphinan scaffold of the ligands 5, 9 a-9 d, 16 a and 16 b adopts the same orientation, whereas the side chain flips depending on the attached substituents, in particular for 9 c and 9 d bearing larger substituents on the azocarboxamide ( Figure S3). The preferred binding modes found for 5, 9 a-9 d, 16 a and 16 b in the k-OR do not show a significant dependence on the structural variations ( Figure S4). In comparison to the binding modes predicted for the μ-OR subtype ( Figure S3), those at the k-OR ( Figure S4) however suggest a completely different position of the morphinan scaffold, which could be due to the fact the k-OR binding pocket is narrower than those of the μ-OR and δ-OR. All ligands 5, 9 a-d and 16 a,b are also likely to present very similar binding modes within the δ-OR subtype ( Figure S5), whereat the overall orientation of the ligands is comparable to binding mode 2, which was only observed for the two bulkier ligands 9 c and 9 d in the μ-OR subtype ( Figure S3).
The key interactions of azocarboxamide 9 b, which is one of the two ligands 9 b and 9 c with the most favorable binding profile (Table 2), to the μ-OR, k-OR and the δ-OR subtype were analyzed and are shown in Figure 3. The interactions of the protonated amine of 9 b to Asp149 3.32 (μ-OR, Figure 3A) and Asp128 3.32 (δ-OR, Figure 3C) as well as between the bridging ether oxygen of 9 b to Tyr150 3.33 (μ-OR, Figure 3A) and Tyr129 3.33 (δ-OR, Figure 3C) are comparable in both subtypes. Although the distances between the ether oxygen of 9 b and the hydroxy group on the adjacent Tyr150 3.33 or Tyr129 3.33 (3.7 Å for the μ-OR and 3.4 Å for the δ-OR) appear as too long for a hydrogen bond, it is important to note that amino acid side chains are not flexible in our docking setup, so that geometric relaxation of the docked complex by energy minimization could indeed enable this interaction. Additionally, the hydrophobic environment is very similar in both subtypes, leading to an almost identical placement of the morphinan scaffold with its cyclopropyl side chain within the binding pocket. Interestingly, the 4-fluoro substituent on the phenylazocarboxamide side chain is likely to form a σ-hole-based halogen bond to the carbonyl unit of Asn129 2.62 (TM2) within the μ-OR subtype, which could induce subtype selectivity. The δ-OR possesses a bulkier Lys108 3.32 without halogen bond capability at position of Asn129 2.62 in the μ-OR subtype. This results in a different binding mode of the flexible 4-fluorophenyl side chain due to steric restrictions. Within the k-OR subtype, the geometry of the binding pocket does not allow an orientation of the morphinan scaffold as in the other subtypes (see also Figure S4). Especially the residues Ile290 6.51 , Ile294 6.55 , and Tyr312 7.35 are responsible for these geometric restrictions. In the alternative binding mode, compound 9 b is able to form a hydrogen bond from the CONH group to Asp138 3.32 . With only minor reorientation of this Asp side chain, an additional hydrogen bond to the protonated amine nitrogen is possible, characterizing Asp138 3.32 as a key residue for the ligand recognition. Despite the different orientation of 9 b within the other subtypes, the 4-fluorophenyl side chain finds itself in a stabilizing hydrophobic environment formed by the residues Trp287 6.48 , Ile290 6.51 , Ile316 7.39 , and Tyr320 7.43 . Unfortunately, the bridging oxygen atom of the morphinan scaffold is not directly involved in interaction to the protein. However, this deficiency is compensated by the strong polar interactions of the ligand to Asp138 3.32 .
To get an impression about the metabolic stability of the newly synthesized azocarboxamide and cinnamide ligands, we treated azocarboxamide 9 b and its cinnamide analogue 16 b with rat liver microsomes. To enable a largely independent evaluation of the phenylazocarboxamide and the cinnamide motif, the simple piperidine-derived azocarboxamide 18 and its corresponding cinnamide 19 were also prepared. In Table 3, the determined half-lives (t 1/2 ) and intrinsic clearances (CL' int ) of the azocarboxamides and cinnamides are summarized together with the values obtained for the reference compound imipramine (20).
The positive control imipramine (20) showed the shortest half-life (52 � 6 min) and corresponding to that also the highest clearance with 132 � 4 μL × min À 1 × mg À 1 ( Table 3). The two piperidine-derived compounds 18 and 19 displayed slightly higher, but comparable stabilities (18: t 1/2 = 73 � 12 min; 19: 85 � 18 min), thus demonstrating that the bioisosteric replacement of a cinnamide unit by a phenylazocarboxamide does only result in a weakly reduced life-time under the chosen assay conditions. For each compound 18 and 19, LC/MS analysis of the reaction mixture revealed the formation of two oxidized species, whereat the mass difference of m/z + 16 points to hydroxylation or, is the case of the cinnamide, possibly also to epoxidation. The trend that the cinnamide derivative shows a higher metabolic stability than the corresponding phenylazocarboxamide also turned out to be true for the ligands 16 b and 9 b, albeit with larger relative deviation than previously found for 18 and 19. For the azocarboxamide 9 b displaying the favorable binding profile (Table 2), a half-life of 166 � 16 min and an intrinsic clearance of 42 � 4 μL min À 1 mg À 1 were determined, suggesting sufficient metabolic stability of [ 18 F]9b for invivo use as a PET imaging agent. The metabolite which was primarily detected by LC/MS for 9 b is likely to be the Ndealkylated species (m/zÀ 54, cleavage of cyclopropylmethyl group) along with small amounts of a reduced derivative (m/z + 2) that could result from reduction of the azo unit to the corresponding hydrazine. For the cinnamide analogue 16 b showing the slightly higher metabolic stability (t 1/2 = 252 � 94 min, CL' int = 28 � 5 μL min À 1 mg À 1 ), only one metabolite corresponding to N-dealkylation (m/zÀ 54) was found. Interestingly, oxidized species (m/z + 16), as they were observed as main metabolites for 18 and 19, were neither detected for 9 b nor for 16 b.
The radiosynthesis shown in Scheme 3 takes advantage of the excellent availability of the 18 F-labeled tert-butyl phenylazocarboxylate [ 18 F]11b, which can readily be prepared in a single step from the quaternary ammonium triflate 21 in only 30 seconds and in high radiochemical yield. [41] Based on our previously developed radiosyntheses, [40][41][42][43] 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 was confirmed through co-injection with the reference compound 9 b by HPLC methods (see Supporting Information). A concentration of amine 10 of 54 mM led to a radiochemical yield of [ 18 F]9b of 11 � 2 % (n = 3) after 5 minutes reaction time, which did not increase any further after 10 minutes. The concentration of amine 10 was therefore increased to 108 mM, which led to an only slightly higher radiochemical yield of 13 � 2 % (n = 3) after 5 minutes. F]11b in larger amounts for further characterization by in vitro rat brain autoradiography or extended in vivo small animal PET imaging studies. As the opioid receptor, especially the μ subtype, is of high interest in the field of brain research concerned with addiction or pain processing, [48] subtype selective μ-OR radioligands for imaging studies by positron emission tomography (PET) are valuable tools to study the regulation of the μ subtype in vivo. The currently available 11 C-labeled OR receptor ligands for PET, such as 11 C-carfentanil or 11 C-PEO, [44,49] suffer from the short half-life (20.3 min) and their use is clearly restricted to institutions with cyclotrons. The radiosyntheses of most 18 Flabeled OR ligands, such as 18 FÀ FEÀ DPN or 18 FÀ FEÀ PEO, [46,50] are rather laborious and give low yields. Therefore, an alternative 18 F-labeled μ-OR ligand, available by a more straightforward radiosynthesis is still desirable.

Conclusion
In summary, a novel type of buprenorphine-derived opioid receptor ligands was obtained upon targeted structural modification. Besides the introduction of a methyl group on the 3hydroxy functionality of the buprenorphine core, the cinnamide side chain of the lead structure 5 was exchanged for various phenylazocarboxamide moieties, also with the aim to explore future suitability for 18 F-radiosynthesis. Ligand binding studies at the three OR subtypes μ, k and δ revealed an even increased subtype selectivity for ligands 9 compared to 5 at a basically unchanged affinity to the μ subtype. Evaluation of the azocarboxamide 9 b in functional assays showed a weak bias for 9 b, as this compound acts as an antagonist in the β-arrestin pathway, but as a weak partial agonist in G protein activation. Its cinnamide analogue 16 b, in contrast, behaved as a neutral antagonist in both pathways. Docking studies gave an impression on the possible binding modes thereby supporting the experimental observation that methylation at the 3-hydroxy group of the buprenorphine core is not decisive for ligands of type 9 or 16. Further assays revealed a reasonable metabolic stability of azocarboxamide 9 b compared to 16 b and related reference compounds. Finally, the course of the 18 F-radiosynthesis of an opioid receptor radioligand candidate for PET was demonstrated by the straightforward synthesis of [ 18 F]9b.

Experimental Section
General experimental. Reactions which are sensitive to air or water were evacuated under traditional Schlenk conditions while heating and under an inert argon atmosphere. Solvents and reagents for sensitive reactions classified as "extra pure", "dry" or "extra dry" or with water contents lower than 0.1 % were applied. For reactions which were insensitive towards water, chemicals classified as "pure" or "for synthesis" were used. The applied chemicals were used as purchased by commercial sources (Alfa Aesar, Acros Organics, Merck, Sigma Aldrich  13 C NMR: 101 MHz) instruments from Bruker at 300 K in deuterated solvents from Deutero GmbH. The chemical shifts δ are presented in ppm and are calibrated either in accordance to tetramethylsilane (TMS) or to the used deuterated solvents: [51] Coupling constants J are listed as experimentally determined differences of the frequen- cies. Coupling between fluoro and carbon atoms in 13 C spectra are presented as J CF . The determination of the spectra was performed using Mestre-C and TopSpin.