Understanding the role of carbamate reactivity in fatty acid amide hydrolase inhibition by QM/MM mechanistic modelling.

QM/MM modelling of FAAH inactivation by O -biphenyl-3-yl carbamates identiﬁes the deprotonation of Ser241 as the key reaction step, explaining why FAAH is insensitive to the electron-donor eﬀect of conjugated substituents; this may aid design of new inhibitors with improved selectivity and in vivo potency. Carbamate-based compounds are widely used as covalent inhibitors of serine hydrolases of therapeutic interest, including fatty acid amide hydrolase (FAAH). 1 This enzyme is characterized by an uncommon Ser-Ser-Lys catalytic triad. It is the main enzyme responsible for the hydrolysis of the endocannabinoid anandamide. Despite its unusual catalytic mechanism, 2–4 FAAH is inhibited by classical serine hydrolase inhibitors, 1 and by O -biphenyl-3-yl carbamates, which are promising clinical candidates for the treatment of central nervous system and peripheral disorders. 5

This prediction has been recently confirmed by the crystallographic structure of the FAAH-URB597 carbamoylated adduct, 11 showing the reliability of this QM/MM approach for FAAH.
In vitro potency of FAAH inhibitors depends on a combination of covalent binding to the catalytic Ser241 and noncovalent interactions with specific recognition elements in the active site of FAAH, 7 while in vivo potency is also affected by plasmatic stability. 12 Structure-activity relationship (SAR) studies showed that conjugated, electron-donor groups on the biphenyl scaffold of URB524 (which increase electron density around the carbamate carbon) are associated with enhanced carbamate stability both in alkaline buffer and in rat plasma. 13 Conversely, these groups do not affect FAAH inhibitor potency in vitro. 7 To understand this apparent discrepancy, we here apply a QM/MM approach to model the carbamoylation reaction of FAAH by the reference inhibitor URB524 (IC 50 = 25.6 nM) and by its p-OH (URB694, IC 50 = 30.0 nM) and p-NH 2 (URB618, IC 50 = 27.2 nM) derivatives (Fig. 1). Furthermore, differential transition state stabilization (DTSS) analysis, 14 an emerging approach for identifying important interactions in enzyme-catalysed reactions, 15 was also performed to identify crucial residues involved in FAAH inhibition.
The QM/MM calculations show that carbamoylation of Ser241 takes place with a similar mechanism, and similar energetic barriers, for the three inhibitors, with formation of the tetrahedral intermediate (TI) being the rate-limiting step of the inhibitory process. Electron donor substituents do not significantly affect the key transition state (TS) of the reaction, in agreement with experimental findings. This provides an explanation of the activity of URB524 analogues, as discussed below.
Michaelis complexes of FAAH with carbamate inhibitors were built according to the energetically preferred 9 and  Steps of Ser241 carbamoylation in FAAH by biphenyl-3-yl carbamates. A, C, E, G are significant intermediates along the reaction pathway. Labels refer to the reaction coordinates defined in the text. experimentally observed 11 binding orientation II. 16 The resulting structures were solvated by a 25 Å -radius sphere of TIPS3P water molecules and equilibrated by 250 ps of molecular dynamics. 9 The equilibrated complexes were employed for QM/MM reaction modelling at the PM3-CHARMM22 level [details in the ESIw], with DFT energy corrections. 17 A coordinate driving approach 18 was used to explore potential energy surfaces (PESs) for Ser241 carbamoylation. The reaction was modelled in three steps (Fig. 2) starting from the Michaelis complex (A): (1) formation of the TI (C); (2) m-biphenate expulsion with formation of carbamoylated Ser241 (E); (3) m-biphenate protonation and formation of neutral Lys142 (G). Other mechanisms were also tested (e.g. TI protonation followed by biphenyl-3-ol expulsion) but as found previously for URB524, 9 only the mechanism reported here was found to be energetically and structurally reasonable.
Formation of the TI was modelled by restraining two reaction coordinates : )], to transfer H 2 from Lys142 to Ser217. Rx, Ry, Rz, Rr, and Rs were increased in steps of 0.1 Å , applying harmonic restraints of 5000 kcal mol À1 Å À2 . For each value of the reaction coordinates, the energy was computed by single point calculation at the PM3-CHARMM22 level, removing the energy contribution due to reaction coordinate restraints. Higher level calculations (B3LYP/6-31+G(d)) 17 were used to correct the energies of crucial stationary points (intermediates and approximate TS structures) from the PM3-CHARMM22 PESs, similarly to refs. 3, 9 and 19. The optimized structures of the reactants (Michaelis complexes) indicate that the three inhibitors bind similarly in the FAAH active site, with the O-biphenyl moiety within the cytoplasmic access (CA) channel and the N-cyclohexyl group occupying the acyl-chain-binding (ACB) pocket of FAAH. The inhibitor carbonyl oxygen is placed in the oxyanion hole, undertaking one hydrogen bond with Ile238, and three weaker polar interactions with backbone NHs of Gly239, Gly240 and Ser241. The hydroxyl of Ser241 is close to the carbamate carbon, and forms a hydrogen bond network with Ser217 and Lys142. For URB694 (Fig. 3) and URB618, the substituent in the para position forms a hydrogen bond with the backbone NHs of Cys269 and Val270 via a water molecule (here named Wat627), conserved in crystallographic structures. 20 Fig . 4 shows the resulting B3LYP/6-31+G(d)//PM3-CHARMM22 energy profiles for carbamoylation of Ser241 by URB524 (blue), URB694 (red) and URB618 (green), starting from the corresponding Michaelis complexes (A). The reaction starts with a double proton transfer, in which Lys142, in cooperation with the bridging Ser217, 3 deprotonates Ser241, triggering the nucleophilic attack and formation of the TI (C).
Consistent with our previous findings, 9 formation of the TI, rather than its protonation and collapse, is the rate limiting step (largest barrier) for FAAH inactivation. The B3LYP/ 6-31+G(d)//PM3-CHARMM22 barriers for this crucial step are similar for the three inhibitors (31.5 kcal mol À1 for URB524, 32.7 kcal mol À1 for URB694 and 33.6 kcal mol À1 for URB618), despite their different chemical structure and reactivity. Thus, introduction of conjugated, electron-donor groups (OH, s p = À0.37 and NH 2 , s p = À0.66) 21 does not significantly affect the overall shape of the energy profiles, nor the energy of the important TS of the reaction in FAAH. This indicates that the electronic distribution of the carbamate has a negligible effect on Ser241 carbamoylation, in spite of its profound importance for chemical and plasmatic stability of these compounds. 12 This can be understood as this TS structure is characterized by deprotonation of Ser241 (H 1 is found between O 1 and O 2 ), with the nucleophile O 1 still approaching the carbonyl carbon of the inhibitor ( Table 1).
The TI (C) is a transient configuration along the carbamoylation pathway, significantly less stable than the reactants (A). It lies 18.0 kcal mol À1 above the reactants for the p-hydroxy analogue, compared to 16.8 kcal mol À1 for the p-amino variant, and 15.8 kcal mol À1 for URB524. Expulsion of the biphenate leaving group gives carbamoylated Ser241 (E) overcoming small energy barriers (lower than 2 kcal mol À1 ), suggesting that  This journal is c The Royal Society of Chemistry 2011 Chem. Commun., 2011, 47, 2517-2519 2519 formation and collapse of TI are highly concerted. Calculations also show that, in contrast to FAAH substrates (e.g. oleamide and oleoylmethylester), 22 the expulsion of the leaving group occurs effectively without prior protonation. However, protonation of the leaving group is required to conclude the catalytic cycle. It is facilitated by Ser217, which is well oriented both to deprotonate the positively charged Lys142 and to protonate the oxygen of the biphenate anion. Energy barriers for this step are small, B3 kcal mol À1 for URB524, B2 kcal mol À1 for URB694 and B3 kcal mol À1 for URB618, with respect to E. The final product (G) is the most stable configuration of the pathway. All the covalently inhibited complexes are relatively very stable (17 kcal mol À1 lower in energy than A for URB524, and by 13 and 11 kcal mol À1 for URB618 and URB694, respectively).
The QM/MM results presented here provide a mechanistic explanation for the observed lack of correlation between inhibitory potency and reactivity of these carbamates. The rate-limiting step for FAAH carbamoylation is the activation of Ser241, with the TS structure dominated by the proton transfer between Ser241 and Ser217. Conversely, for reactions with ''off-target'' carboxylesterases, reactivity of O-biphenyl-3-yl carbamates correlates well with inhibitor potency. 13 DTSS calculations (details in ESIw) indicate that TS structures for Ser241 carbamoylation are greatly stabilized, compared to Michaelis complexes, for all three inhibitors. This stabilization is mainly related to residues involved in the proton transfer from Ser241 to Lys142, through Ser217. FAAH active site (Fig. 5 and  Table S1, ESIw), provides comparable DTSS for all inhibitors (-25.1, À24.0, and À25.4 kcal mol À1 for URB524, URB694, and URB618, respectively). The highest contribution to the lowering of activation barriers is from Thr236 (BÀ7.5 kcal mol À1 ). This residue accepts a H-bond from Lys142, and favours its protonation, 23 critical for Ser241 activation. 3 Ser218 (which also accepts a H-bond from Lys142), Ile238, Gly239 and Gly240 contribute comparable amounts of DTSS for the three inhibitors. Other amino acids show negligible contributions, with the exception of Asp237, which shows a small destabilizing effect on the TS of URB618. The water molecule (Wat627, Fig. 3), involved in the recognition of p-OH and p-NH 2 substituents, 7 has a negligible contribution to TS stabilization (DTSS of À0.4 and À0.7 kcal mol À1 for URB694 and URB618, respectively).
The calculations reported here show that the mechanism of FAAH inhibition by URB524 and its derivatives gives similar barrier heights for carbamates that have significantly different reactivity. As intrinsic reactivity of carbamatebased inhibitors strongly affects their pharmacokinetics and selectivity, QM/MM calculations may be useful for the design of new inhibitors with improved pharmaceutical properties e.g. improved in vivo potency and better selectivity vs. putative off-targets.
AJM and EC thank EPSRC for support. JS thanks the Royal Thai Government for funding. a Atom labels consistent with Fig. 2 and 4.