Effective in silico prediction of new oxazolidinone antibiotics: force field simulations of the antibiotic-ribosome complex supervised by experiment and electronic structure methods

Abstract

We propose several new and promising antibacterial agents for the treatment of serious Gram-positive infections. Our predictions rely on force field simulations, supervised by first principle calculations and available experimental data. Different force fields were tested in order to reproduce linezolid's conformational space in terms of a) the isolated and b) the ribosomal bound state. In a first step, an all-atom model of the bacterial ribosome consisting of nearly 1600 atoms was constructed and evaluated. The conformational space of 30 different ribosomal/oxazolidinone complexes was scanned by stochastic methods, followed by an evaluation of their enthalpic penalties or rewards and the mechanical strengths of the relevant hydrogen bonds (relaxed force constants; compliance constants). The protocol was able to reproduce the experimentally known enantioselectivity favoring the S-enantiomer. In a second step, the experimentally known MIC values of eight linezolid analogues were used in order to crosscheck the robustness of our model. In a final step, this benchmarking led to the prediction of several new and promising lead compounds. Synthesis and biological evaluation of the new compounds are on the way.

Keywords: compliance constants; computational chemistry; drug design; molecular recognition; relaxed force constants.

Figure 1

Two-dimensional structure and nomenclature of the oxazolidinone linezolid. The different rings are highlighted with capital letters A, B and C. The acetamidic substituent is denominated as C5 side chain. The pharmacophore portion is colored in blue. Only the S-enantiomer is active.

Figure 2

Superposition of all low energy minima of linezolid applying AMBER (left) and OPLS-AA (right) force field.

Figure 3

Superposition of the found global linezolid gas phase minimum (AMBER on the left and OPLS-AA on the right) with the linezolid bioactive conformation (green). The OPLS-AA force field reproduces the conformation linezolid adopts in the solid state [38] or complexed by a transporter protein [39].

Figure 4

RMSD/Potential energy plots for linezolid in the gas phase. The OPLS-AA plot is characterized by an absence of conformers roughly between 25–34 kJ/mol.

Figure 5

Model building process. a) linezolid bound to the 50S subunit from Haloarcula marismortui, code 3CPW, ca. 100 000 atoms; b) 30 Å radius mother-shell pulled from the crystal structure, linezolid in green; c) optimizing the mother-shell; d) 10 Å radius working-shell used for the conformational searches throughout the paper.

Figure 6

Superposition of all 43 minima of the ribosome–linezolid complex. Linezolid is shown in yellow and CCA-N-Phe in green. The nucleotide nomenclature is the same as the one reported in the crystal structure original paper [12]. Note the bent conformation of linezolid in its bioactive state and the flexibility of the morpholine ring.

Figure 7

Comparison between ribo/S-lzd and ribo/R-lzd (quasi) global minima. On top, a 2D interaction diagram is displayed. A zoom of the found minima superimposed each other that illustrate the conformation and interactions made by S-lzd (yellow) and R-lzd (green).

Scheme 1

Experimentally characterized linezolid analogues that were used as test cases for our simulation protocol. Compound 7 is the active form of the prodrug torezolid. For references, see the text.

Scheme 2

Predicted new linezolid-like candidates.