Determinants of the Inhibition of DprE1 and CYP2C9 by Antitubercular Thiophenes

Abstract Mycobacterium tuberculosis (Mtb) DprE1, an essential isomerase for the biosynthesis of the mycobacterial cell wall, is a validated target for tuberculosis (TB) drug development. Here we report the X‐ray crystal structures of DprE1 and the DprE1 resistant mutant (Y314C) in complexes with TCA1 derivatives to elucidate the molecular basis of their inhibitory activities and an unconventional resistance mechanism, which enabled us to optimize the potency of the analogs. The selected lead compound showed excellent in vitro and in vivo activities, and low risk of toxicity profile except for the inhibition of CYP2C9. A crystal structure of CYP2C9 in complex with a TCA1 analog revealed the similar interaction patterns to the DprE1–TCA1 complex. Guided by the structures, an optimized molecule was generated with differential inhibitory activities against DprE1 and CYP2C9, which provides insights for development of a clinical candidate to treat TB.

Mtb Decaprenylphosphoryl-b-d-ribofuranose 2-oxidase (DprE1) is af lavin adenine dinucleotide (FAD)-dependent enzyme,w hich together with decaprenylphosphoryl-d-2ketoerythropentose reductase (DprE2), converts decaprenylphosphoryl-beta-d-ribose (DPR) to decaprenylphosphorylbeta-d-arabinofuranose (DPA), an essential cell wall component. [1] Conditional knockdown studies showed that loss of dprE1 results in as trong bactericidal effect in vitro. [2] Multiple covalent [3] and non-covalent [4] DprE1 inhibitors have been identified and showed in vitro and in vivo activities against Mtb,f urther validating DprE1 as an attractive anti-tuberculosis (TB) drug target. [5] We previously reported the identification of TCA1f rom ac ell-based phenotypic screen, and demonstrated it is aD prE1 inhibitor. [6] Since then we have generated new derivatives of TCA1 to improve its in vitro potency,P K properties,a nd in vivo efficacy. When analyzing toxicity profiles,wenoticed that many TCA1analogs show inhibition of CYP2C9, one of six major cytochrome P450 enzymes which determine the clearance of 75 %o fm arketed drugs.T herefore,i nhibition of P450 enzymes such as CYP2C9 can potentially lead to drug-drug interactions,w hich may cause serious issues for combinatory drug regimen for TB treatment. [7] Here we report the X-ray crystal structures of WT DprE1, aT CA1-resistant DprE1 mutant, and CYP2C9 in complex with TCA1 analogs.Along with structure activity relationship studies,t he mechanism of resistance by the mutant DprE1 enzyme and the tight correlation between DprE1 and CYP2C9 inhibition was elucidated. We further generated analogs with excellent in vitro and in vivo activities against Mtb,and selectively targeting DprE1 over CYP2C9.
In the DprE1-TCA1 structure,T CA1s howed ap lanar conformation and tight fit in the DprE1 active site.T he benzothiazole core is oriented parallel to the FADi soalloxazine ring and the thiophene carboxamide resides in small hydrophobic pocket. Besides substantial hydrophobic interactions,m ultiple hydrogen bonding interactions are formed between TCA1a nd residues Lys418, His132, and Ser228 ( Figure S1 in the Supporting Information, SI). Guided by the structure,o ur medicinal chemistry effort focused on the benzothiazole core (red), thiophene carboxamide (green) and the acylcarbamate (blue) of TCA1( Figure 1).
TheGln334 amide group and the Ty r60 hydroxy group are only 3.7 away from C4 and C5 of the TCA1benzothiazole ring, respectively,w hich make them two potential H-bond donors if H-bond acceptors are introduced at these positions (Figure 1and Figure S1). Indeed, the introduction of nitrogen to the benzothiazole ring (TCA007) improved the IC 50 by 8fold (5 AE 2nm)c ompared to TCA1 (48 AE 14 nm)( Figure 1, Table 1, SI methods and Figure S2-S4).
Thet hiophene ring is located in ah ydrophobic pocket formed by Gly133, Lys367, Phe369 and Asn385 and accepts aH -bond from His132 Ne2( 3.2 ). Replacement of thiophene with other heterocycles or adding bulky substitutions resulted in reduced activity against Mtb (data not shown). Only afluoro substituent on the thiophene afforded an analog (TCA481) that is 5-fold more potent than TCA1b ased on MIC (Figure 1and Table 1). To better understand the basis of the activity increase,wesolved the crystal structure of DprE1 in complex with TCA481 (Table S1). Thes tructure reveals that the fluoro group forms aH -bond with Asn385 Nd2 (3.4 ). Thef luorine is accommodated in the pocket with stronger hydrophobic interactions with residues Phe369, His132, and Lys367 than TCA1( Figure 2B and Figure S1).
Modification of the carbamate group was carried out; however, most of the analogs such as amides and esters are not active,d emonstrating the importance of hydrogen bonding between Ser228 and the carbamate group (3.0 to 3.5 ) ( Figure 2C and Figure S1). When replacing the carbamate with various heterocycles,o nly TCA020 with pyrimidine showed moderate activity (MIC = 2.5 mm,I C 50 = 370 AE 133 nm)( Figure 1a nd Table 1). To help optimizing the pyrimidyl substituent, we crystallized DprE1 in complex with TCA020 (SI methods). TheD prE1-TCA1 structure revealed that the two nitrogen atoms of the pyrimidine ring occupy the positions of the two oxygens of the carbamate group as H-bond acceptors.Based on these structural insights, we introduced amethyl group at 4-position of the pyrimidine to gain ah ydrophobic interaction with W230 (TCA787), which has similar activity to TCA1( Figure 1a nd Table 1).
Y314 resides close to the carbamate moiety and it was previously found that as pontaneous TCA1r esistant mutant harbors aY 314C mutation. [6] We measured the inhibitory activities of TCA1a nd its analogs against DprE1 and the Y314C variant and observed 5-to 40-fold shift on IC 50 values under the same assay condition, which is consistent with the MIC shift between the wildtype (WT) and the mutant strain (Table 1). These data support the notion that growth inhibition of Mtb by the TCA1 series of inhibitors is via DprE1 inhibition. We also generated aY 314A mutant and it is sensitive to TCA1inhibition, suggesting that the side chain of Y314 does not directly contribute to TCA1b inding (Table S2). To elucidate the molecular basis of the resistance mechanism, the structure of DprE1 Y314C bound with TCA1 was determined to 2.2 ( Figure 2D, Figure S1 and Table S1).   Although the overall fold of the mutant enzyme is identical to wildtype,consequential structural changes in the active site of DprE1 were evident. Them ost striking difference between the WT and mutant DprE1 complexes is the 1808 8 flipped orientation of the boomerang-shaped ligand in the active site ( Figure S1 and Figure S5). Thea lternative orientation of the ligand can be linked to the effects of cysteine 314 substitution on the orientation of Lys134, which positions the e-ammonium group in the void created by substituting the tyrosyl ring with at hiol. Ther eorientation of Lys134 is driven by formation of as trong H-bond (2.75 )b etween its eammonium and the thiol of Cys314 and results in anoticeable shape change of the active site around Lys134. Consequently, the acylcarbamate keto-oxygen in the original orientation would clash with the alkyl moiety of Lys134 (with the shortest distance at 2.7 ), which likely weakens the binding of TCA1 ( Figure S5). Due to its superior activity against Mtb and improved PK properties,T CA007 was selected as al ead compound for further investigation. We evaluated the efficacy of TCA007 against Mtb Erdman in acute and chronic BALB/c mouse low dose aerosol infection models. [8] After treatment for 3w eeks during the acute phase of infection with TCA007 at adose of 50 mg kg À1 and 100 mg kg À1 body weight, ad ose-dependent in vivo efficacy was observed with reduction of bacterial load in lung by 1.7 and 2.3 log 10 CFU,respectively ( Figure 3A and SI methods). This excellent in vivo bactericidal activity is comparable to the first-line drug rifampin. Following 4weeks of treatment during the chronic phase of infection, TCA007 at 200 mg kg À1 achieved areduction of > 1.7 log 10 CFU in mouse lungs ( Figure 3B).
Atypical treatment of TB lasts for more than six months, and therefore,t he safety profile of anti-TB drugs is of great importance.T CA007 is not cytotoxic against Vero and HepG2 cell lines at up to ac oncentration of 100 mm. TCA007 does not inhibit hERG potassium channel in ap atch-clamp assay (IC 50 > 30 mm), suggesting al ow risk of cardiotoxicity.Inaddition, we observed no genetic toxicity in amini-Ames mutagenicity test. TCA007 showed no inhibition of four major CYPs,1 A2, 3A4, 2D6, 2C19, but as trong inhibitory effect for CYP2C9 (SI methods).
Thep otent inhibition of P450 2C9 could reflect oxygenation of thiophene moiety of TCA analogs to an electrophilic S-oxide or epoxide [9] that reacts with the enzyme to form an irreversible inhibitor as seen for tienilic acid. [10] However,the inhibition of 2C9 by TCA analogs was not time dependent indicating ar eversible mode ( Figure S6). To better understand the interactions of TCA007 with 2C9, the structure of the complex was determined by X-ray crystallography to 2.0 resolution [11] (PDB:5 W0C) (SI methods,T able S3 and Figure S7). Similar to the binding of TCA1t oD prE1, the thiophene group of TCA007 binds in ahydrophobic pocket of the 2C9 substrate binding cavity that is far from the iron of the heme cofactor where the reactive oxidant is formed, thus preventing oxygenation of the thiophene moiety.T he pocket is formed by the inner surface of helix B' and adjacent residues ( Figure 4A)a nd hydrophobic residues on helix F  across from B' and helix Ga bove TCA007. Mimicking the role of DprE1 Lys418, Arg108 donates H-bonds from Ne to the carbonyl oxygen of the thiophene carboxamide (3.0 ) and from Nh2tothe thiazole nitrogen (3.4 ). Thethiophene carbonyl also accepts an H-bond from awater molecule.T he Arg108 Ne also contacts the thiophene sulfur. Arg108 plays ac ritical role in recruiting negatively charged or partially charged substrates such as diclofenac and warfarin, respectively,t othes ubstrate binding site of 2C9. Substitution of Arg108 by alanine, [12] histidine or phenylalanine [13] leads to aloss catalytic activity for the two substrates.
Phe100, Leu102, and Phe114 form ahydrophobic pocket together with Phe476 where the carbamate group of TCA007 is bound ( Figure 4A). In contrast to DprE1, protein H-bond donors for the carbamate are not present in this pocket, but the cavity is open to solvent and an ordered water molecule donates an H-bond to the carbonyl of the acyl group.T he pyridyl nitrogen of the pyridothiozole group is oriented toward the heme but is too distant (6.6 )for coordination to the heme iron. Consistent with this observation, TCA007 does not significantly alter the visible absorption spectrum of 2C9 in solution. In contrast, miconazole binds to the heme iron of 2C9 and alters the spectrum of the heme co-factor, and competitive displacement of iron bound miconazole by TCA007 reverts the spectrum of the miconazole complex to that of the apo enzyme ( Figure S8). This portion of the active site is also hydrophobic, but am olecule of water is evident 2.88 from the pyridyl nitrogen of TCA007 ( Figure 4A). The water molecule is hydrogen bonded to as econd water molecule that donates al one pair of electrons to the heme iron and aH -bond to the backbone carbonyl of Ala297. In general, the extensive hydrophobic interactions and hydrogen bonding interactions with Arg108 are likely to form the basis for the high affinity,reversible binding of TCA007 to 2C9, but these interactions also sequester TCA007 in position that reduces the likelihood of metabolism, which is consistent with absent or low rates of metabolite formation or time dependent inhibition.
Removal of these aforementioned heteroatoms will likely reduce the CYP2C9 inhibition. However,i tw as shown that the thiophene group and carbonyl group are also essential for DprE1 binding.Therefore,weinstead modified the ring Ato reduce CYP2C9 inhibition while gaining extra interactions in ring Bt om aintain DprE1 binding ( Figure 4B). Both sulfur and nitrogen in the ring Aw ere removed to reduce the interaction in CYP2C9, and two nitrogens were introduced into ring Bw hich potentially form H-bonds with Ty r60 and Glu334 to compensate for the reduced interaction in DprE1 ( Figure 4C). TheI C 50 ratio (CYP2C9 over DprE1) for TCA582 and TCA711 were 33 and 1515, respectively, which, compared with TCA007 (ratio = 1.8), represents asignificant improvement in selectivity ( Figure 4D). Them odeling of CYP2C9 with both TCA582 and TCA711 revealed that the carbamate group could gain favorable hydrophobic interactions with Phe100 (3.6 ), Phe114 (4.1 )a nd Phe476 (3.8 ). These interactions would be weakened by an unfavorable hydrophobic interaction with Phe100 (2.9 and 3.0 )w hen replacing the carbamate with pyrimidyl substituent, which could explain the nearly eliminated inhibition of CYP2C9 by TCA711 (IC 50 of CYP2C9 > 50 um)( Figure 4D). Thei nvivo studies on the advanced analog TCA711 are ongoing.
In summary,w ed emonstrate the mode of inhibition of DprE1 by the thiophene scaffold via enzymatic and structural studies.T he SAR studies not only helped us to optimize the potencyo ft he inhibitors but revealed an unconventional resistant mechanism of DprE1. Theo ptimized lead compound TCA007 showed excellent in vitro and in vivo activities,PKproperties,and low risk of toxicity profile except for the strong inhibition of CYP2C9. Guided by the structures of CYP2C9 in complex with TCA1a nalogs,a no ptimized molecule,TCA711, was generated with differential inhibitory activities against DprE1 and CYP2C9, which paved the road for further development of ac linical candidate to treat TB.