Determination of ligand binding modes in weak protein-ligand complexes using sparse NMR data

Methyl probes in proteins for determining ligand binding mode in weak protein-ligand complexes

Structures of protein–ligand complexes provide critical information for drug design. Most protein–ligand complex structures are determined using X-ray crystallography, but where crystallography is not able to generate a structure for a complex, NMR is often the best alternative. However, the available tools to enable rapid and robust structure determination of protein–ligand complexes by NMR are currently limited. This leads to situations where projects are either discontinued or pursued without structural data, rendering the task more difficult. We previously reported the NMR Molecular Replacement (NMR²) approach that allows the structure of a protein–ligand complex to be determined without requiring the cumbersome task of protein resonance assignment. Herein, we describe the NMR² approach to determine the binding pose of a small molecule in a weak protein–ligand complex by collecting sparse protein methyl-to-ligand NOEs from a selectively labeled protein sample and an unlabeled ligand. In the selective labeling scheme all methyl containing residues of the protein are protonated in an otherwise deuterated background. This allows measurement of intermolecular NOEs with greater sensitivity using standard NOESY pulse sequences instead of isotope-filtered NMR experiments. This labelling approach is well suited to the NMR² approach and extends its utility to include larger protein–ligand complexes.

Selective binding of small molecules to Vibrio cholerae DsbA offers a starting point for the design of novel antibacterials

DsbA enzymes catalyze oxidative folding of proteins that are secreted into the periplasm of Gram‐negative bacteria, and they are indispensable for the virulence of human pathogens such as Vibrio cholerae and Escherichia coli. Therefore, targeting DsbA represents an attractive approach to control bacterial virulence. X‐ray crystal structures reveal that DsbA enzymes share a similar fold, however, the hydrophobic groove adjacent to the active site, which is implicated in substrate binding, is shorter and flatter in the structure of V. cholerae DsbA (VcDsbA) compared to E. coli DsbA (EcDsbA). The flat and largely featureless nature of this hydrophobic groove is challenging for the development of small molecule inhibitors. Using fragment‐based screening approaches, we have identified a novel small molecule, based on the benzimidazole scaffold, that binds to the hydrophobic groove of oxidized VcDsbA with a K_D of 446±10 μM. The same benzimidazole compound has ∼8‐fold selectivity for VcDsbA over EcDsbA and binds to oxidized EcDsbA, with K_D>3.5 mM. We generated a model of the benzimidazole complex with VcDsbA using NMR data but were unable to determine the structure of the benzimidazole bound EcDsbA using either NMR or X‐ray crystallography. Therefore, a structural basis for the observed selectivity is unclear. To better understand ligand binding to these two enzymes we crystallized each of them in complex with a known ligand, the bile salt sodium taurocholate. The crystal structures show that taurocholate adopts different binding poses in complex with VcDsbA and EcDsbA, and reveal the protein‐ligand interactions that stabilize the different modes of binding. This work highlights the capacity of fragment‐based drug discovery to identify inhibitors of challenging protein targets. In addition, it provides a starting point for development of more potent and specific VcDsbA inhibitors that act through a novel anti‐virulence mechanism.

NMR fragment screening reveals a novel small molecule binding site near the catalytic surface of the disulfide-dithiol oxidoreductase enzyme DsbA from Burkholderia pseudomallei

The presence of suitable cavities or pockets on protein structures is a general criterion for a therapeutic target protein to be classified as 'druggable'. Many disease-related proteins that function solely through protein-protein interactions lack such pockets, making development of inhibitors by traditional small-molecule structure-based design methods much more challenging. The 22 kDa bacterial thiol oxidoreductase enzyme, DsbA, from the gram-negative bacterium Burkholderia pseudomallei (BpsDsbA) is an example of one such target. The crystal structure of oxidized BpsDsbA lacks well-defined surface pockets. BpsDsbA is required for the correct folding of numerous virulence factors in B. pseudomallei, and genetic deletion of dsbA significantly attenuates B. pseudomallei virulence in murine infection models. Therefore, BpsDsbA is potentially an attractive drug target. Herein we report the identification of a small molecule binding site adjacent to the catalytic site of oxidized BpsDsbA. ¹H^N CPMG relaxation dispersion NMR measurements suggest that the binding site is formed transiently through protein dynamics. Using fragment-based screening, we identified a small molecule that binds at this site with an estimated affinity of K_D ~ 500 µM. This fragment inhibits BpsDsbA enzymatic activity in vitro. The binding mode of this molecule has been characterized by NMR data-driven docking using HADDOCK. These data provide a starting point towards the design of more potent small molecule inhibitors of BpsDsbA.

Determination of protein-ligand binding modes using fast multi-dimensional NMR with hyperpolarization

Elucidation of small molecule-protein interactions provides essential information for understanding biological processes such as cellular signaling, as well as for rational drug development. Here, multi-dimensional NMR with sensitivity enhancement by dissolution dynamic nuclear polarization (D-DNP) is shown to allow the determination of the binding epitope of folic acid when complexed with the target dihydrofolate reductase. Protein signals are selectively enhanced by polarization transfer from the hyperpolarized ligand. A pseudo three-dimensional data acquisition with ligand-side Hadamard encoding results in protein-side [13C, 1H] chemical shift correlations that contain intermolecular nuclear Overhauser effect (NOE) information. A scoring function based on this data is used to select pre-docked ligand poses. The top five poses are within 0.76 Å root-mean-square deviation from a reference structure for the encoded five protons, showing improvements compared with the poses selected by an energy-based scoring function without experimental inputs. The sensitivity enhancement provided by the D-DNP combined with multi-dimensional NMR increases the speed and potentially the selectivity of structure elucidation of ligand binding epitopes.

Automatic structure-based NMR methyl resonance assignment in large proteins

Isotopically labeled methyl groups provide NMR probes in large, otherwise deuterated proteins. However, the resonance assignment constitutes a bottleneck for broader applicability of methyl-based NMR. Here, we present the automated MethylFLYA method for the assignment of methyl groups that is based on methyl-methyl nuclear Overhauser effect spectroscopy (NOESY) peak lists. MethylFLYA is applied to five proteins (28–358 kDa) comprising a total of 708 isotope-labeled methyl groups, of which 612 contribute NOESY cross peaks. MethylFLYA confidently assigns 488 methyl groups, i.e. 80% of those with NOESY data. Of these, 459 agree with the reference, 6 were different, and 23 were without reference assignment. MethylFLYA assigns significantly more methyl groups than alternative algorithms, has an average error rate of 1%, modest runtimes of 0.4–1.2 h, and can handle arbitrary isotope labeling patterns and data from other types of NMR spectra.

The structures and dynamics of large proteins can be studied with methyl-based NMR but peak assignment is still challenging. Here the authors present MethylFLYA that allows automated assignment of methyl groups and apply it to five proteins with molecular weights in the range from 28 to 358 kDa.

Framework for and evaluation of bursts in random sampling of multidimensional NMR experiments

The grouping of data in bursts, also referred to as clusters, spikes or clumps, is a common phenomenon in stochastic sampling. There have been several reports that suggest that in NMR, the presence of such bursts is beneficial to spectral reconstruction where data are sampled nonuniformly. In this work, we seek to define a mode of sampling that produces bursts of randomly distributed data in a controlled manner. An algorithm is described for achieving this where the burst length and its uniformity is controlled - we refer to this type of sampling mode as clustered sampling. Measures are introduced for assessing the "burstiness" of nonuniformly sampled data in multiple dimensions and properties of the point-spread-function of these schedules are assessed. The clustered sampling method is applied to samples drawn from an exponentially weighted distribution either distributed randomly or pseudo-randomly by use of a jittering algorithm. The results reveal that bursts introduce characteristic sampling artifacts that are shifted to low frequencies (red shifted), with respect to the signal frequency, and that they produce artifact-reduced regions at frequencies related to the burst length. This observation is contrary to that observed for sampling methods that seek to evenly distribute NUS data, such as jittered or Poisson sampling. Extensive evaluation of simulated data with comparable inherent sensitivity, reveals that at high sampling coverage (25% in 1D), the distribution of the data has little impact on common spectral quality measures. Application of the introduced clustered sampling method to an experimental 3D NOESY experiment showed results consistent with that found for the simulated 1D data. However, in the extremes of very sparse sampling, the results suggest that there may be some advantages associated with incorporation of bursts in nonuniform sampling. The tools and theory presented will serve as a starting point to further explore this novel mode of sampling in NMR.

Determination of Ligand Binding Epitope Structures Using Polarization Transfer from Hyperpolarized Ligands

Drug discovery processes require the determination of the protein binding site structure, which can be achieved via nuclear magnetic resonance (NMR) spectroscopy. While traditional NMR spectroscopy suffers from low sensitivity, NMR signals can be significantly enhanced through hyperpolarization of nuclear spins. Here, folic acid is hyperpolarized by dissolution dynamic nuclear polarization (D-DNP). Polarization transfer to dihydrofolate reductase is compared to signal evolution predicted for docking-derived structures. The results demonstrate that a scoring function derived from the experimental data improves the ranking of structures. With data from six methyl groups, Spearman's correlation coefficient of the experimental scoring function to the root-mean-square deviation from a reference structure is 0.88 for five individually addressed ligand protons and 0.59 for the entire ligand, while the same correlation coefficient of the energy calculated from docking alone is 0.49. D-DNP NMR-derived ranking, therefore, is capable of determining the ligand structure with a small number of individually addressed source spins.

Selective Methyl Labeling of Proteins: Enabling Structural and Mechanistic Studies As Well As Drug Discovery Applications by Solution-State NMR

Escherichia coli expression protocols for selective labeling of methyl groups in proteins have been essential in expanding the size range of targets that can be studied by biomolecular NMR. Based on the initial work achieving selective labeling of isoleucine, leucine, and valine residues, additional methods were developed over the past years which enabled the individual and/or simultaneous combinatorial labeling of all methyl containing amino acids. Together with the introduction of new methyl-optimized NMR experiments, this now allows the detailed characterization of protein-ligand interactions as well as mechanistic and dynamic processes of protein-protein complexes up to 1MDa in size. In this chapter, we provide a general introduction to selective labeling of proteins using E. coli-based expression systems, describe the considerations taken into account prior to the selective labeling of a protein, and include the protocols used to produce such proteins. An overview of applications using selectively labeled proteins with an emphasis on examples relevant to the drug discovery process is then presented.

Online Optimization of Total Acceptance in Hamiltonian Replica Exchange Simulations

Replica exchange is a widely used sampling strategy in molecular simulation. While a variety of methods exist to optimize parameters for temperature replica exchange, less is known about how to optimize parameters for more general Hamiltonian replica exchange simulations. We present an algorithm for the online optimization of total acceptance for both temperature and Hamiltonian replica exchange simulations using stochastic gradient descent. We optimize the total acceptance, a heuristic objective function capturing the efficiency of replica exchange. Our approach is general and has several desirable properties, including: (1) it makes few assumptions about the system of interest, (2) optimization occurs online without the requirement of presimulation, and (3) most importantly, it readily generalizes to systems where there are multiple control parameters (e.g., temperatures, force constants, etc.) that determine the Hamiltonian of each replica. We explore some general properties of the algorithm on a simple harmonic oscillator system, and demonstrate its effectiveness on a more complex data-guided protein folding simulation.

Fragment library screening identifies hits that bind to the non-catalytic surface of Pseudomonas aeruginosa DsbA1

At a time when the antibiotic drug discovery pipeline has stalled, antibiotic resistance is accelerating with catastrophic implications for our ability to treat bacterial infections. Globally we face the prospect of a future when common infections can once again kill. Anti-virulence approaches that target the capacity of the bacterium to cause disease rather than the growth or survival of the bacterium itself offer a tantalizing prospect of novel antimicrobials. They may also reduce the propensity to induce resistance by removing the strong selection pressure imparted by bactericidal or bacteriostatic agents. In the human pathogen Pseudomonas aeruginosa, disulfide bond protein A (PaDsbA1) plays a central role in the oxidative folding of virulence factors and is therefore an attractive target for the development of new anti-virulence antimicrobials. Using a fragment-based approach we have identified small molecules that bind to PaDsbA1. The fragment hits show selective binding to PaDsbA1 over the DsbA protein from Escherichia coli, suggesting that developing species-specific narrow-spectrum inhibitors of DsbA enzymes may be feasible. Structures of a co-complex of PaDsbA1 with the highest affinity fragment identified in the screen reveal that the fragment binds on the non-catalytic surface of the protein at a domain interface. This biophysical and structural data represent a starting point in the development of higher affinity compounds, which will be assessed for their potential as selective PaDsbA1 inhibitors.