Figure 1.
Optimization of inhibitor scaffolds.
For the fragment hit and inhibitor candidates C1, C2, C3 and C4, identical cutaway views of solvent accessible surface representations of the active-site pockets of E. faecalis GyrB from the crystal structures of complexes of the inhibitors with the 24 kDa N-terminal fragment of GyrB from E. faecalis GyrB are shown. The bound inhibitors are drawn with green bonds, the conserved ATP-binding aspartate is drawn with blue bonds and the structural water molecule that plays a key role in substrate binding in GyrB and ParE is shown as a red sphere. Potential hydrogen-bonds between the inhibitors, aspartate and water molecule are depicted as dotted lines. Optimization of the pyrrolopyrimidine scaffold led to inhibitors like C1 with good enzyme potency but only moderate Gram-negative antibacterial activity [16]. Expansion of the bicyclic pyrrolopyrimidine scaffold to a tricyclic pyrimidoindole scaffold (C2) fills an interior lipophilic pocket and offers superior optimization vectors to improve enzyme potency. Subsequent elaboration of the tricyclic scaffold with a fluorine atom at R6 and an aminomethyl moiety at R8 dramatically improved inhibitor potency and ligand efficiency. The 6-fluoro-N-methyl-9H-pyrimido[4,5-b]indol-8-amine scaffold quantitatively fills the lipophilic interior sub-pockets of the GyrB/ParE active-sites and adds a new hydrogen-bond. C3 and C4 demonstrate sub-nanomolar enzyme potency versus GyrB and ParE enzymes from a broad range of Gram-positive and Gram-negative pathogens; inhibition constants (Ki values) are shown for a representative enzyme panel that includes the full length GyrB and ParE enzymes from E. faecalis, Francisella tularensis, and E. coli.
Figure 2.
Summary of inhibitor optimization strategies.
A) Side view of the “salt-bridge” pocket from the crystal structure of the complex of E. coli GyrB with C3, with key interactions highlighted. C3 is drawn with green bonds. Potential hydrogen-bonds are depicted as dotted lines. The residues comprising the salt-bridge pocket are drawn with blue bonds and a semi-transparent surface representation of the pocket is shown. The salt-bridge pocket residues curl around the R2 pyrimidine, forming a U-shaped pocket. R2 substituents were designed to address the complex structural and electronic features of the salt-bridge pocket. Extensive ab initio and binding free energy calculations of the R2 methyl pyrimidine of C3 show significant binding energy from a π–cation interaction with the salt-bridge Arg. The methyl pyrimidine also engages the Arg on the outer rim of the salt-bridge pocket through a water-mediated hydrogen-bond. Van der Waals interactions are observed between a conserved proline that defines the face of the salt-bridge pocket opposite the Glu-Arg salt-bridge pair.
B) Alternate view of the E. coli GyrB complex structure with C3, highlighting key polar interactions between the R4 diamine of C3, active-site residues and an ordered solvent network. The Asn residue shown in the figure (N46 from the E. coli structure) and salt-bridge residues are conserved in all bacterial topoisomerases, while the residues comprising the “pocket floor” (blue for the residues in E. coli GyrB, tan for the residues from the overlaid F. tularensis ParE/C3 complex structure differ between GyrB and ParE enzymes. The R4 diamine sits at the protein-solvent interface at the outer rim of the lipophilic interior pocket that binds the (A) ring of the inhibitor. The upper face of the inhibitor occupies a highly conserved polar pocket while the lower face occludes a lipophilic shelf (the pocket floor) that is structurally heterogeneous between GyrB and ParE due to sequence differences in the enzymes, as highlighted. The R4 diamine adopts a low energy conformation that does not impinge on the structurally diverse pocket floor and directs a basic amine out of the pyrimidoindole plane to interact via hydrogen-bonds with the conserved Asn at the mouth of the interior pocket (N46). The basic amine complements the negative electrostatic potential in this region of the active-site; the same anionic pocket captures the terminal amine from a conserved lysine residue involved in phosphate binding in the dimeric complex of E. coli ParE with ADPNP [17]. The R4 diamine also hydrogen-bonds with an ordered solvent network above the pocket floor. The water molecules from this network shown in the figure were observed in similar positions in GyrB and ParE crystal structures from multiple orthologs (as shown in C), and molecular dynamics simulations showed this water network remained throughout all simulations, and each water molecule had significantly long residence times. Thus, these water molecules were treated as conserved structural elements during inhibitor optimization.
C) Electron density maps showing the positions of C3 and the water network that was conserved across all GyrB and ParE enzymes that were structurally characterized in this study. Final 2|fo-fc| electron density maps contoured at 1.3σ for: i) the 1.6 Å E. coli GyrB complex with C3, ii) the 1.3 Å E. faecalis GyrB complex with C3, and iii) the 2.4 Å F. tularensis ParE complex with C3. The water network that interacts with the R4 diamine of the inhibitor is conserved in the three protein orthologs. R4 groups were designed to position an amine that simultaneously hydrogen-bonds with the water network and a conserved Asn residue in the active-site pocket.
Figure 3.
Selectivity of TriBE inhibitors versus eukaryotic ATP-binding proteins.
Inhibitory activities of C3 and C4 against a panel of divergent human kinases, and human topoisomerase II. All compounds were assayed at 10 μM concentration. Level of inhibition is color-coded as indicated in the inset.
Figure 4.
The effects for C3 (A) and C4 (B) on macromolecular synthesis in E. coli (BAS849) imp.
Incorporation of [3H]-precursors of DNA (●), RNA (○), protein (▲) and cell wall (▽) was examined. The MIC value for each compound is indicated by a vertical dashed line. Both compounds exert a primary effect on DNA synthesis and a secondary effect on RNA synthesis.
Figure 5.
Reduction in bioburden after 24 hrs in a neutropenic mouse thigh infection model.
The MIC values for C3, C4 and levofloxacin against the E. coli strain used in the study are 0.13, 0.13 and 0.03 μg/mL, respectively.