Selective Inhibitors of Bacterial t-RNA-(N1G37) Methyltransferase

Article pubs.acs.org/jmc

Selective Inhibitors of Bacterial t‑RNA-(N1G37) Methyltransferase (TrmD) That Demonstrate Novel Ordering of the Lid Domain Pamela J. Hill,*,† Ayome Abibi,‡ Robert Albert,† Beth Andrews,‡ Moriah M. Gagnon,† Ning Gao,‡ Tyler Grebe,† Laurel I. Hajec,‡ Jian Huang,‡ Stephania Livchak,‡ Sushmita D. Lahiri,‡ David C. McKinney,† Jason Thresher,‡ Hongming Wang,† Nelson Olivier,§ and Ed T. Buurman‡ †

Departments of Chemistry, ‡Biosciences and §Discovery Sciences, Infection Innovative Medicines Unit, AstraZeneca R&D Boston, 35 Gatehouse Drive, Waltham, Massachusetts 02451, United States S Supporting Information *

ABSTRACT: The tRNA-(N1G37) methyltransferase (TrmD) is essential for growth and highly conserved in both Gram-positive and Gram-negative bacterial pathogens. Additionally, TrmD is very distinct from its human orthologue TRM5 and thus is a suitable target for the design of novel antibacterials. Screening of a collection of compound fragments using Haemophilus influenzae TrmD identified inhibitory, fused thieno-pyrimidones that were competitive with S-adenosylmethionine (SAM), the physiological methyl donor substrate. Guided by X-ray cocrystal structures, fragment 1 was elaborated into a nanomolar inhibitor of a broad range of Gram-negative TrmD isozymes. These compounds demonstrated no activity against representative human SAM utilizing enzymes, PRMT1 and SET7/9. This is the first report of selective, nanomolar inhibitors of TrmD with demonstrated ability to order the TrmD lid in the absence of tRNA.

■

INTRODUCTION Although all tRNAs are composed of nucleosides A, C, U, and G, a total of 81 different modifications have been described.1 Methylation at the N1-position of guanosines immediately 3′ to the anticodon of seven tRNAs (m1G37) is present in virtually all species in the three kingdoms of life, suggesting an ancient and important function in translation.2,3 The presence of the modification prevents Watson−Crick base-pairing of this guanosine with cytosine in mRNA and translational frameshifting that would result as a consequence.4 The methylation of G37 in bacteria is performed by tRNA(N1G37) methyltransferase (TrmD)5 and has been shown to be essential for growth in a range of bacterial species.6−11 A TrmD orthologue in Archaebacteria could not be identified by sequence homology in genome databases. However, a functional, yet structurally distinct, homologue of TrmD was found by selection of a genome library of Methanococcus vannielii for complementation of a growth defect due to a temperature sensitive trmD allele in Salmonella typhimurium.2 This led to the identification of TrmD in Methanococcus jannaschii and subsequently the Saccharomyces cerevisiae orthologue TRM5.2 X-ray crystal structures of TrmD of Escherichia coli12 and Haemophilus influenzae13 showed the protein as a homodimer with the S-adenosylmethionine (SAM) binding site at the interface. This suggests that under physiological conditions TrmD exists as a dimer, a fact that was corroborated in Streptococcus pneumoniae.14 Each monomer consists of two domains held together by an ill-defined linker that showed improved ordering upon binding of SAM. Given the proximity to © 2013 American Chemical Society

both a large pocket and the methionine moiety of the substrate, further ordering of the linker would probably require binding of the tRNA substrate. However, such structures have not yet been published. Although TRM5 performs an identical function to TrmD, the two seem to have evolved independently. The most striking difference for drug discovery purposes is the conformation with which SAM binds to the respective enzymes. In TrmD, the trefoil-shaped binding site binds the substrate in a bent conformation, whereas the substrate binds in an open conformation with TRM5.15 This strongly indicates that TrmD inhibitors that bind in the SAM binding site will be selective, without showing mechanism-based toxicity toward mammalian cells. Therefore, we set out to identify chemical starting points that inhibit TrmD by binding in the SAM binding pocket and thereby preventing methylation of tRNA. Co-crystallization and structure elucidation of one such fragment bound to H. influenzae TrmD demonstrated binding in the adenosine portion of the pocket. The ensuing structure-guided effort allowed extension of the fragments and additional binding in the methionine portion of the SAM pocket, thus yielding broad-spectrum nanomolar inhibitors of TrmD. Received: May 14, 2013 Published: August 27, 2013 7278

dx.doi.org/10.1021/jm400718n | J. Med. Chem. 2013, 56, 7278−7288

Journal of Medicinal Chemistry

■

Article

RESULTS AND DISCUSSION Fragment Screening. An AstraZeneca fragment collection (14 000 compounds) was screened at a compound concentration of 100 μM to identify inhibitors of H. influenzae TrmD using a biochemical assay monitoring methylation of tRNA. The resulting hits were then screened for activity using a similar assay with Staphylococcus aureus as a representative Grampositive isozyme. This secondary screen selected for compounds with the potential to be broad spectrum inhibitors. Sinefungin, a natural product derivative of SAM, was used as a positive control. All fragments that displayed IC50 < 50 μM in both assays were cocrystallized with H. influenzae TrmD to determine the binding location. X-ray cocrystal structures revealed that several fragments bound in the SAM binding pocket (Figure 1). All

fragment, as this position was occupied by carbon C2. A hydrogen bond not present between SAM and H. influenzae TrmD was observed between the Oγ of S132 and N7 of the pyrimidone ring in compound 1. Lead Generation. The X-ray cocrystal structure of 1 with H. influenzae TrmD shows a very tight binding pocket around the pyrimidone core and suggests few options for substitution other than at the 3-position. Maintaining this substitution trajectory, a series of alternate heterocycles was fused to the pyrimidone core to probe the H-bond donor−acceptor tolerability of the binding site (Table 1). A hydrogen bond donor at position 1 (4, 5) was not optimal for inhibition of Gram-negative activity, while the Gram-positive isozyme appeared more forgiving. A fivemembered ring was preferred to a six-membered ring, likely due to the resulting vector with which the phenyl substituent is forced to adopt in the six-member example. Both 1 and 2 demonstrated good potency against all isozymes tested. However, the isoxazole was not as amenable to varying the substitution at the 3-position. The larger atomic radius of the sulfur atom in compound 1 results in the substituent at the 3position, taking on a nonplanar orientation relative to the core. This lowest energy conformation is identical to the binding conformation. In the isoxazole ring, the 3-position substituent takes on a planar geometry relative to the core. For the phenyl substitution there was little impact; however, for most other substitutions, this resulted in a significant entropic penalty and resulting loss in potency to adopt the binding conformation. Therefore, 1 was the lead compound chosen for SAR studies. Synthesis and Biochemical Structure−Activity Relationship (SAR) of Thienopyrimidones. The thienopyrimidones were prepared in three steps (Scheme 1) beginning with the condensation of disubstituted ketone 6 with ethylcyanoacetate to form intermediate 7, which was then reacted with elemental sulfur to form the desired amino thiophene ester 8. Cyclization with formamidine acetate afforded the desired thienopyrimidones 9−16. Utilizing this sequence, a variety of substituents at the R1 and R2 positions of the thiophene ring were introduced and assessed for biochemical activity to establish SAR around the thiophene ring. The cocrystal structure of 1 shows the phenyl ring at the 3position of the thiophene moiety extending into the ribose portion of the binding pocket. To gather a better understanding of the SAR associated with occupying the ribose pocket, a series of analogues was prepared where the phenyl ring was substituted or replaced (Table 2). The des-phenyl compound (10) showed almost an order of magnitude loss in potency against E. coli with an IC50 of 120 versus 17 μM for compound 1. This demonstrated the importance of occupying the ribose pocket for activity. Substituting off the para position of the phenyl ring with a small lipophilic group improved potency 2-fold (11). The phenyl ring was replaced with a variety of aromatic heterocycles, of which the most potent was the pyrazole variant 12 at 2.1 μM. Saturated ring systems (13−15) resulted in a significant loss in activity. A series of acyl linked substituents was also prepared, of which the amide 16 appeared most promising with an IC50 of 6.5 μM. The phenyl and pyrazole substituted compounds both displayed good potency; however, they resulted in an undesirable vector to further extend substitution into the methionine pocket. Substitution off the meta or para position would require the phenyl ring to rotate out of plane disrupting the binding orientation. Therefore, compound 16 was the focus for further analogue design.

Figure 1. Hits from fragment screening.

bound fragments were heteroaromatic systems which mimicked the key acceptor−donor−acceptor hydrogen bonding motif present in the adenine core of SAM. The low molecular weight (MW < 300 Da) and good aqueous solubility (>500 μM) of these fragments made them ideal starting points for SAR studies. It was apparent that a fused pyrimidone motif was common to several of the fragments. This structural motif bound in the SAM pocket in an orientation that should permit chemical modification to access the methionine portion of the pocket. As such, this core structure was chosen as the starting point for a structure-guided chemistry optimization effort. The cocrystal structure of compound 1 (Figure 2A,B) in complex with TrmD from H. influenzae was solved at 1.6 Å resolution in the space group H32 with one protein molecule in the asymmetric unit (see Supporting Information for data collection and refinement statistics). By superimposing this structure with the published H. influenzae structure in complex with SAM (PDB: 1UAK13), it was found that 1 occupied the same space as the adenine and ribose of SAM. It was apparent that the thienopyrimidone ring system of 1 did not superimpose with the purine ring system of SAM by aligning the adenine N6 with the O6 of the thienopyrimidone. Instead, the binding mode showed the respective nine-member rings offset by 90°, placing the phenyl group of 1 in the vicinity of the SAM ribose moiety. The thienopyrimidone ring system formed key hydrogen bonds to three of the four protein residues known to interact with the adenosine base of SAM: I133, Y136, and L138. Additionally, the hydrophobic contacts between adenine and P89 and P144 confirmed binding. In contrast, N6 from the adenine base of SAM formed a hydrogen bond with the carbonyl oxygen of G134. This interaction was not possible for the pyrimidone 7279



Article

Figure 2. Inhibitor-bound co-crystal structures of H. influenzae TrmD. (A,C,E) Compounds 1, 27, and 38, respectively. Protein monomers of the biologically relevant dimer are colored yellow and cyan. Dotted lines represent possible hydrogen bonds between protein atoms and inhibitors, labels indicate residue code and H. influenzae TrmD sequence position of the amino acids. (B,D,F) Superposition of inhibitor-bound H. influenzae TrmD cocrystal structures against SAM-bound H. influenzae TrmD cocrystal structure (1UAK13), inhibitors are colored yellow, SAM is colored in gray, and protein atoms have been removed for clarity. (E) 2Fo − Fc electron density map contoured to 1.0σ of compound 38. Hydrogen bonds between inhibitor and labeled amino acids from both protein monomers of the biologically relevant arrangement are indicated with dashed lines, and hydrogen bonds between the thienopyrimidone core and protein are omitted for clarity. Amino acid D177 is modeled with its carboxylate in alternate conformations, and only one conformation is within hydrogen bond distance to the imidazole ring on the compound.

Synthesis and Biochemical SAR of Amide-Linked Thienopyrimidones. The amide-linked compounds (16, 21−29) were synthesized as shown in Scheme 2. Condensation of ethylacetoacetate and ethylcyanoacetate followed by cyclization with elemental sulfur yielded the desired thiophene amino diester 17. Subsequent cyclization with formamidine acetate yielded the fused thienopyrimidone ethyl ester (18) which resulted from transesterfication with ethanol during the

cyclization. Finally, amide 20 was arrived at by heating with the desired amine or via a two-step procedure of saponification followed by T3P mediated amide coupling. Replacement of the phenyl ring of 16 with a pyridyl ring caused a loss in potency, 20-fold for the ortho isomer (21, Table 3) and 2-fold for the meta and para isomers (22 and 23). Methylation of the amide N−H (24) caused a significant loss of activity as did alkylation of the α-methylene (25 and 26). The 7280



Article

Table 1. Biochemical Potency, in IC50s (μM), of Fused Core Variations against Various TrmD Isozymesa

a

ND = not determined.

Scheme 1. Synthesis of Fused Thienopyrimidonesa

a

Conditions: (a) NH4OAc, acetic acid, toluene, reflux; (b) sulfur, Et2NH, EtOH, reflux; (c) formamidine acetate, EtOH, reflux.

exists for an inhibitor exploiting the spatial electrostatic characteristics of SAM. The X-ray cocrystal structure of the TrmD with compound 27 (Figure 2C,D) was solved to determine the binding interaction of the phenyl methylamine moiety. The amide link extends out of the adenine pocket, forming an intramolecular hydrogen bond between the cognate nitrogen and the pyrimidone oxygen. The trajectory of the phenyl ring superimposed with that of the methylene groups from the ribose of SAM and the methylamine group terminates in a region similar to where the amino end of the substrate binds (Figure 2D). Although S170 from the opposing monomer has been shown to interact with both SAM and S-adenosylhomocysteine (AdoHcy),13 this residue was disordered in the cocrystal structure of 27. A hydrogen bond between the terminal amine of 27 with the carboxylate side chain of E116 was evident. This specific interaction was not observed between SAM and TrmD. E116 is strictly conserved among TrmD orthologues and has

cocrystal structure suggests that the para position of the phenyl ring offers the best vector for substitution to occupy the space of the physiological substrate. To this end, compound 27 bearing a p-methylamine substituent was made. While both 27 and 28 showed a significant boost in potency, compound 28 was slightly less active, perhaps indicating some value in hydrogen bond donor capacity at this position. Saturation of the phenyl ring in 27 led to the cyclohexyl variant (29) which, while still active, was 6-fold less potent. The decrease in activity may be attributed to a steric clash between the ring system of P89 and the nonplanar geometry of a saturated cyclohexyl group. The analysis of the SAM-bound X-ray crystal structure (PDB: 1UAK13) suggests the methionine portion of the binding pocket would allow binding of the methylene groups of SAM in an extended conformation. Comparison of the apo- and SAMbound X-ray crystal structures suggests little change in conformation upon substrate binding and, thus, the potential 7281



Article

Table 2. Biochemical Potency, in IC50s (μM), of Variations of the Ribose Pocket Group against Various TrmD Isozymes

Scheme 2. Synthesis of Amide Substituted Thienopyrimidonesa

a

Conditions: (a) (i) NH4OAc, acetic acid, toluene, reflux, (ii) sulfur, Et2NH, EtOH, reflux; (b) formamidine acetate, EtOH, reflux; (c) RNH2, EtOH, 70 °C; (d) LiOH, MeOH/water, 80 °C; (e) Et3N, T3P, RNH2, DCM.

been proposed to be involved in the recognition of G37 of the tRNA substrate.13 Accessing Beyond the Methionine Pocket: Synthesis and Biochemical SAR of the Substituted Amines. A series of analogues was prepared substituting on the para position of the phenyl ring of 27 to build into the methionine portion of the binding pocket (Scheme 3, Table 4). Ahn et al.13 reported the ternary complex of H. influenzae TrmD with AdoHcy and a phosphate ion. In this X-ray crystal structure, an ordering of residues G161 to D169 from an opposing monomer was observed, forming a so-called “lid,” which covered the substrate binding cleft. Although the lid contains the acidic and catalytically relevant residue D169, none of our cocrystal X-ray structures indicated an interaction between the inhibitor and D169. As such, compound 27 was extended beyond the confines of the SAM pocket to probe the disordered region of the protein and hopefully establish a hydrogen bond interaction with D169.

Our hypothesis was that engaging with D169 could facilitate stabilization of the residue and concomitantly improve potency. Replacing the aminomethyl of 27 with a hydroxymethyl (30) or phenyl (31) resulted in a significant decrease in potency (Table 4). Changing the electron donating character of the amine by converting to the amide (32) also showed a decrease in potency while the amidine (33) was equipotent with 27. To build into the methionine pocket, substituted amines were investigated. Propylamine (34) was equipotent to the primary amine as was methoxyethylamine (35). Incorporating an additional hydrogen bond donor, like aminoethyl (37) or imidazole (38), improved potency approximately 2-fold. Following up on the potent amidine 33, substituted guanidine (39) was prepared and significantly improved potency to an IC50 of 0.19 μM. Compounds 40−51 represented a series of cyclic amines designed to explore the effect of removing the hydrogen bond donor proximal to the core and extending this group deeper into the pocket to target the D169 interaction. Compared to 27, 7282



Article

Table 3. Biochemical Potency, in IC50s (μM), of Variations of the Amide Substituents against Various TrmD Isozymes

Scheme 3. Synthesis of the Substituted Phenyl Amidesa

Conditions: (a) ethylene glycol, pyridinium-p-toluenesulfonic acid, toluene 130 °C, Dean−Stark; (b) LAH, THF, 0 °C to room temperature; (c) Et3N, EtOH, reflux; (d) 6 M HCl, 1,4-dioxane; (e) R1R2NH, Ti(OPri)4, NaBH(OAc)3, THF, 45 °C.

a

amino-azetidine (40) showed a 3-fold improvement in potency. An exploration of ring size demonstrated that the pyrrolidine as well as the piperidines and piperazines were well tolerated.

Incorporation of hydrogen bond donating groups onto the pyrrolidine resulted in compounds with IC50s of 0.2−0.4 μM (42−44). The unsubstituted piperazine was equipotent to 7283



Article

Table 4. Biochemical Potency, in IC50s (μM), of Substituted Phenyl Amides

compound 27, while altering the basic character of the free N−H with the piperizinone 46 showed more than a 3-fold loss of activity. Building off the piperazine nitrogen to explore the methionine pocket resulted in a modest improvement in potency (47−48). Switching to the piperidine and moving the donor

outside of the ring resulted in a significant improvement in potency (51) to 0.18 μM. Thienopyrimidone TrmD Inhibitors Have Broad Spectrum Enzymatic Activity. A representative set of the most potent compounds were tested for broad spectrum TrmD 7284



Article

confirm or refute mode of action via TrmD were not successful due to a low resistance frequency of less than 10−9. While the inability to isolate resistant mutants did not enable the confirmation of the mode of action, it is an indication of a low likelihood of developing resistance. Given their potent enzymatic activity but weak antibacterial activity, including against the efflux mutant strains, these compounds presumably penetrate slowly into the cytosol. This is likely due to the complex nature of the bacterial cell wall and membrane that can act as a barrier to charged compounds such as these. This may indicate a dichotomy for this series between what is required for binding to the target and what is desirable to penetrating bacterial cells. Given the nature of the binding pocket, the most potent compounds were those that contained dibasic groups to extend into the methionine pocket. However, these same dibasic groups make penetrating the bacterial cell more difficult.17 Modulation of compound physical properties would be necessary to enhance cell penetration. TrmD Inhibitors Demonstrate Novel Ordering of the “Lid” Region. To further analyze the structural implications of substituted amines binding to TrmD, the cocrystal structure of TrmD with 38 (Figure 2E,F) was solved at a resolution of 1.9 Å. The position of the thienopyrimidone and phenyl ring of 38 in the SAM binding pocket is virtually identical to the binding mode for 27. Consistent with our hypothesis, the hydrogen bond donating nitrogen (N1, pKa = 14.4) of the imidazole ring was within binding distance to D169 carboxylate (2.9 Å). Additionally, the resulting electron density map supports an alternate conformation of the D177 side chain on the opposing monomer. The refined occupancy state for the carboxylate orientation pointing toward or away from the main chain amide was 0.56/ 0.44, respectively. The conformer with the carboxylate pointing toward the main chain amide was 2.8 Å from the second imidazole nitrogen (N3, pKa = 6.5) of 38, a reasonable distance for a hydrogen bond contact to be established. The conformer that placed the carboxylate away was beyond hydrogen bond distance to the inhibitor. Given that crystallization was

activity employing a variety of Gram-positive and Gram-negative isozymes (Table 5). Inhibitory activity against E. coli, Table 5. Biochemical Spectrum, in IC50s (μM), of Selected Inhibitors against a Wide Range of TrmD Isozymes

27 37 38 40 43 44 51

E. coli

A. baumannii

H. influenzae

K. pneumoniae

P. aeruginosa

S. aureus

0.91 0.44 0.52 0.33 0.29 0.40 0.18

0.61 0.74 0.23 1.3 0.24 0.63 0.58

0.56 0.33 1.5 0.33 0.42 0.60 0.33

0.83 0.61 0.31 1.3 0.51 0.63 0.35

0.098 0.044 0.067 0.25 0.063 0.12

Selective Inhibitors of Bacterial t-RNA-(N1G37) Methyltransferase

Recommend Documents