BY NMR FUNNEL METADYNAMICS

23 downloads 0 Views 2MB Size Report
high-throughput molecular dynamics (MD) at the microsecond time scale. The ... using funnel metadynamics (FD), and solution NMR (Nuclear Magnetic.
Laura TROUSSICOT*, Florence Guilliere*, Vittorio Limongeli$, Olivier Walker* and Jean-Marc Lancelin* * Institut des Sciences Analytiques, Université Claude Bernard Lyon1, Villeurbanne, France $ Department of Pharmacy, University of Naples Federico II, I-80131 Naples, Italy

Human peroxiredoxin 5 (hPrx5) is a member of thiol-active peroxidases. These enzymes catalyze hydroperoxides decomposition through a complex catalytic pathway and their active sites are differently shaped around an active thiolate. Peroxiredoxins were found to be involved in key signaling pathways and the design of specific ligands would afford new tools to better understand their biological roles. We study here the binding of recently discovered ligands using

high-throughput molecular dynamics (MD) at the microsecond time scale. The affinity of two closely related ligands homolog, pyrocatechol (Catechol) and 4methyl-pyrocathecol (MetCat), was for the first time evaluated at the atomic resolution comparing the absolute binding free-energy (ΔG°bind) precisely calculated using funnel metadynamics (FD), and solution NMR (Nuclear Magnetic Resonance).

CHEMICAL SHIFT PERTURBATIONS (CSPs) BY NMR

UNBIASED MOLECULAR DYNAMICS (MD) Material and method: Simulations were carried out with the AMBER99SB-ILDN force field and MD is set up with ACEMD1. NVT production of around 200 ns to analyse the binding/unbinding events. Oxidized DTT of the crystal complex was kept in one of the two active sites of the homodimer while the second active site was free of ligands. We positionned 6 ligands (equivalent to 15mM) arbitrary near the two actives sites. During the equilibration-relaxation of the system, constrains where applied to both the heavy atoms of the protein and the ligands (Figure 1).

Material and method: method Spectra were acquired at 301K with a Varian Inova 600MHz NMR spectrometer equipped with a 5 mm standard triple resonance (1H/13C/15N) inverse probe with a z z-axis field gradient. Samples contained 400uM of the reduced 15N-labeled protein in PBS, and an increasing volume of unlabeled ligands was added until reaching saturation CSPs were plotted vs [L]/[P] and KD calculated using MatLab software. saturation. Results Results:

Figure 1: hPrx5 with cristallographic H2O, DTT in one active site, and 6 molecules of methylcatechol

Results: MD is used to have a first approach of the binding. The chemical space is randomly explored thanks to the high concentration of ligands. At least one of the active site is visited by one molecule of methylcatechol during the simulation, replacing the DTT (Figure 2). We then used FM then to quantify the affinity of the ligand with this active site of hPrx5 by calculating the ΔG°bind. Figure 2: binding of one molecule of methylcatechol to one of the active cysteine, replacing oxidized DTT into the catalytic site

FUNNEL METADYNAMICS (FM) Figure 1: HSQC spectra of hPrx5 400uM and addition of catechol until a final [Catechol] = 9mM. Ratio [Lt]/[Pt] = 25

Material and methods A funnel-shaped restraint potential is applied to the system. It combines a cone restraint which includes the binding site, and a cylindric part directed toward the solvent (Figure 3). The sampling of ligand-bound and -unbound states is highly enhanced, leading to an accurate estimation of the ΔG°bind within a reasonable simulation time2 (50ns/days for a system of 55000atoms).

Δδ = [(ΔδH)2 + (ΔδN/5)2]1/2 CATECHOL

METHYLCATECHOL

We used FM to study the binding process of two homologs to hPrx5: Catechol and Methylcatechol. Three atoms were fixed at the base of the protein: Cα of G6, G31 and K65. Figure 3: structure of hPrx5 with catechol, and the funnel restraint potential used in FM

A)

B)

C)

Catechol and Methylcatechol affinity for hPrx5 A) Position projection on z axis [Å] vs the potential mean force (PMF) : W(Z) [kcal.mol-1] B) Plot ΔG°bind [kcal.mol-1] vs the simulation time [ns] C) Free Energy Surface (FES) 2D [kcal.mol-1] : Z axis position projection vs the distance from the Z axis, collecting all of the configurations sampled. MetCat has a wider basin than Catechol and a slighty better ΔG°bind : ΔG°bind = + RT ln KD Catechol: ΔG°bind = - 2.0 ± 0.6 kCal/mol Methylcatechol: ΔG°bind = - 3.6 ± 0.4 kCal/mol

KD = 4.0 ± 0.5 mM ΔG°bind = - 3.3 ± 0.4 kCal/mol

KD = 0.9 ± 0.1 mM ΔG°bind = - 4.2 ± 0.5 kCal/mol

Figure 2: KD fitted curves for both ligands and structure of the binding with residues in interaction coloured

Chemical Shift Perturbation at saturated concentration of ligand

Binding conformations Two energetic basins are visible for both ligands, so two binding conformations are exchanging during the run :  Catechol (Figure 4): The major one (A) with the single H-bond is comparable to the positions described to mimic the transition state in all crystal structures of hPrx53. The energetic barrier to overcome to obtain the double H-bonds conformation (B) is of 1kcal.

Despite its better KD, MetCat induces less CSP than Catechol.

 Methylcatechol : each basin could correspond to two different conformations because of the non symmetry of the MetCat. Figure 4: 3D FES of Catechol [kcal.mol-1]. As function of the distance from Z axis [Å] and a protein-ligand angle (S-O-O) [rad]. Structure of the two binding conformations A and B in exchange

A

B

References 1. M. J. Harvey et al. ACEMD: Accelerating Biomolecular Dynamics in the Microsecond Time Scale. J. Chem. Theory Comput., 2009, 5, 1632-1639 2. V. Limongelli et al., Funnel metadynamics as accurate binding free-energy method, Proc. Natl. Acad. Sci., 2013, 110, 6358 3. A. Hall et al., Structural evidence that peroxiredoxin catalytic power is based on transition-state stabilization, J. Mol. Biol. 2010, 402, 194

Conclusion To study the affinity of ligands with hPrx5 we used solution-state NMR and molecular dynamics in a coupled way. MD without bias is used to have a first microscopic approach of the binding, then FM is used to study the binding and to calculate with accuracy the value of ΔGbind of the protein-ligand complex. NMR is used to get an average analysis that match the MD approach. The calculated KD and ΔGbind’s of those systems were both in the mM range, and where shown very closed to the experimentally determined data. MetCat KD is better than Catechol KD. Catechol show a more localized interaction (narrower energetic basin) while MetCat samples larger interaction possibilities in the active site.