3304
J. A. Rose et al. / Bioorg. Med. Chem. Lett. 25 (2015) 3301–3306
side (RHS) fragments (compound 11 and 12, respectively) showed
that both sides of the original backbone are required for enzyme
inhibition. LHS modification to compound 13 showed that the R-
stereochemistry of the phenyl substituent was required for
potency. Modifications of the amide linker also resulted in loss of
inhibition, including changing to the gem-dimethylamino acid 14,
removal of the amide hydrogen to give tertiary amide 15, and
change to include a bulky side chain like valine 16. Interestingly,
an intermediate side chain such as in serine analog 17 allowed
for retention of modest potency, likely reflecting a steric constraint
in that region of the binding pocket. All analogs demonstrated
limited specific interactions to the LHS group other than to the
backbone peptide. However, the outer lining of this pocket is
formed of aromatic groups such as Tyr168, Trp221, and Tyr62,
which provides favorable interactions to the phenyl substituent
of the LHS. Additionally, the structure clearly shows that the inhi-
bitor binds to the pocket in a bent conformation such that the phe-
nyl group of LHS is brought in close proximity to the difluorobenzyl
ring of RHS. An intramolecular hydrophobic contact provides some
favorable interactions to the otherwise exposed phenyl group of
the RHS.
As shown in the crystal structure, the nitrogen of the benzyl
amide is shown to be interacting with active site residues Asp97,
Asp108, and Glu204 through a water bridge W1 (Figs. 2a and 3a).
These residues are theorized to interact with the nitrogen of
methionine at the N-terminus of the natural substrate, allowing
proper stabilization for peptide hydrolysis.13,23 Therefore it was
hypothesized that the inhibitor resembles a substrate analog and
that addition of a benzyl amino group to the right-hand side would
mimic the methionine nitrogen. It was hypothesized that this
would result in improved potency through displacement of the
water molecule and direct hydrogen-bonding with these three
active site residues (Fig. 3b). Consistent with this hypothesis, syn-
thesis of compound 18 showed improved biochemical potency in
all three isozymes with sub-micromolar potency against S. pneu-
moniae (Table 2). A small library of benzyl amine compounds
showed consistent improvement in potency as compared to the
methylene analogs, and delivered the most active compound for
this series (22) with sub-micromolar inhibition against E. coli and
S. pneumoniae (Table 2). Unfortunately even with this improved
potency, compounds 10, 18, 21, and 22 failed to show antibacterial
aqueous solubility >200 lM with other physical properties of cer-
tain analogs similar to 1 (see Supplementary material).
An X-ray crystal structure of 1 in complex with E. coli MAP was
obtained at 1.8 angstrom resolution (PDB ID 4Z7M). The structure
showed one metal bound to the active site and a clear density for
compound 1 was in the vicinity (Fig. 2a). Several strong interac-
tions were observed between 1 and the peptide backbone, with
no direct interaction with the divalent metal cofactor. However,
the compound does not displace the primary metal, but instead
interacts with the water molecule that is bound to the metal
(Fig. 2b). There are multiple polar interactions observed in the
structure. The alanine carboxy oxygen was found to form a hydro-
gen bond with His79. The amino group of the azepinone ring made
a hydrogen bond with the carboxy oxygen of Cys169 and the car-
boxy oxygen of the azepinone ring made a hydrogen bond with
the backbone amide of Cys169. The carboxy oxygen of the benzyl
amide formed hydrogen bonds with the imidazole nitrogens of
His171 and His178, while the nitrogen of the benzyl amide formed
an indirect interaction with active site residues Asp97, Asp108,
Glu235 and Glu204 through a water bridge (W1). These interac-
tions are summarized in Figure 2a.
activity at concentrations up to 64 lg/ml against any of the diverse
collection of Gram-negative and Gram-positive pathogens in our
In addition to the polar interactions, there are also hydrophobic
residues that contribute to the binding mode. Most notable is the
benzyl ring interactions that are positioned in a pocket formed
by aromatic residues Phe177, Trp221, Tyr65, and Tyr62. This
pocket positions the N-terminal methionine that allows recogni-
tion and cleavage of this amino acid, and thus is hydrophobic
and enclosed (Fig. 2c). In contrast, the LHS azepinone ring is more
solvent exposed, as this part of the pocket binds to the rest of the
peptide chain with very little specificity for the amino acid
sequence other than the backbone peptide. As a result, there are
typical screening panel.
The structural modifications were able to improve biochemical
potency against E. coli and S. pneumoniae MAP, thus allowing the
design of a broad spectrum compound. However, the compounds
always lacked measurable S. aureus potency. Crystal structures of
S. aureus MAP are available in the public domain and provide a
basis for rationalizing this difference. When the E. coli structure
bound to compound 1 was overlaid on a previously-published S.
aureus MAP structure (1QXZ, Oefner et al),24 it was clear that all
key polar interactions and the residues surrounding the
Figure 3. Structure/substrate-based design. (a) X-ray crystal structure of compound 1 in E. coli MAP showing interaction of the benzyl amide nitrogen with Asp97, Asp108,
Glu235 through a water bridge. (b) Substrate-based design of lead compound hypothesizing that addition of a benzyl nitrogen would mimic the methionine nitrogen of the
natural substrate and interact with directly with active site residues.