How to tackle protein structural data from solution and solid state: An integrated approach

Machine Learning-Enhanced Quantum Chemistry-Assisted Refinement of the Active Site Structure of Metalloproteins

Understanding the fine structural details of inhibitor binding at the active site of metalloenzymes can have a profound impact on the rational drug design targeted to this broad class of biomolecules. Structural techniques such as NMR, cryo-EM, and X-ray crystallography can provide bond lengths and angles, but the uncertainties in these measurements can be as large as the range of values that have been observed for these quantities in all the published structures. This uncertainty is far too large to allow for reliable calculations at the quantum chemical (QC) levels for developing precise structure-activity relationships or for improving the energetic considerations in protein-inhibitor studies. Therefore, the need arises to rely upon computational methods to refine the active site structures well beyond the resolution obtained with routine application of structural methods. In a recent paper, we have shown that it is possible to refine the active site of cobalt(II)-substituted MMP12, a metalloprotein that is a relevant drug target, by matching to the experimental pseudocontact shifts (PCS) those calculated using multireference ab initio QC methods. The computational cost of this methodology becomes a significant bottleneck when the starting structure is not sufficiently close to the final one, which is often the case with biomolecular structures. To tackle this problem, we have developed an approach based on a neural network (NN) and a support vector regression (SVR) and applied it to the refinement of the active site structure of oxalate-inhibited human carbonic anhydrase 2 (hCAII), another prototypical metalloprotein target. The refined structure gives a remarkably good agreement between the QC-calculated and the experimental PCS. This study not only contributes to the knowledge of CAII but also demonstrates the utility of combining machine learning (ML) algorithms with QC calculations, offering a promising avenue for investigating other drug targets and complex biological systems in general.

The evolution of paramagnetic NMR as a tool in structural biology

Paramagnetic NMR data contain extremely accurate long-range information on metalloprotein structures and, when used in the frame of integrative structural biology approaches, they allow for the retrieval of structural details to a resolution that is not achievable using other techniques. Paramagnetic data thus represent an extremely powerful tool to refine protein models in solution, especially when coupled to X-ray or cryoelectron microscopy data, to monitor the formation of complexes and determine the relative arrangements of their components, and to highlight the presence of conformational heterogeneity. More recently, theoretical and computational advancements in quantum chemical calculations of paramagnetic NMR observables are progressively opening new routes in structural biology, because they allow for the determination of the structure within the coordination sphere of the metal center, thus acting as a loupe on sites that are difficult to observe but very important for protein function.

Characterizing conformational ensembles of multi-domain proteins using anisotropic paramagnetic NMR restraints

It has been over two decades since paramagnetic NMR started to form part of the essential techniques for structural analysis of proteins under physiological conditions. Paramagnetic NMR has significantly expanded our understanding of the inherent flexibility of proteins, in particular, those that are formed by combinations of two or more domains. Here, we present a brief overview of techniques to characterize conformational ensembles of such multi-domain proteins using paramagnetic NMR restraints produced through anisotropic metals, with a focus on the basics of anisotropic paramagnetic effects, the general procedures of conformational ensemble reconstruction, and some representative reweighting approaches.

Integrating cryo-EM and NMR data

Single-particle cryo-electron microscopy (cryo-EM) is increasingly used as a technique to determine the atomic structure of challenging biological systems. Recent advances in microscope engineering, electron detection, and image processing have allowed the structural determination of bigger and more flexible targets than possible with the complementary techniques X-ray crystallography and NMR spectroscopy. However, there exist many biological targets for which atomic resolution cannot be currently achieved with cryo-EM, making unambiguous determination of the protein structure impossible. Although determining the structure of large biological systems using solely NMR is often difficult, highly complementary experimental atomic-level data for each molecule can be derived from the spectra, and used in combination with cryo-EM data. We review here strategies with which both techniques can be synergistically combined, in order to reach detail and understanding unattainable by each technique acting alone; and the types of biological systems for which such an approach would be desirable.

Magnetic susceptibility and paramagnetism-based NMR

The magnetic interactions between the nuclear magnetic moment and the magnetic moment of unpaired electron(s) depend on the structure and dynamics of the molecules where the paramagnetic center is located and of their partners. The long-range nature of the magnetic interactions is thus a reporter of invaluable information for structural biology studies, when other techniques often do not provide enough data for the atomic-level characterization of the system. This precious information explains the flourishing of paramagnetism-assisted NMR studies in recent years. Many paramagnetic effects are related to the magnetic susceptibility of the paramagnetic metal. Although these effects have been known for more than half a century, different theoretical models and new approaches have been proposed in the last decade. In this review, we have summarized the consequences for NMR spectroscopy of magnetic interactions between nuclear and electron magnetic moments, and thus of the presence of a magnetic susceptibility due to metals, and we do so using a unified notation.

Design and applications of lanthanide chelating tags for pseudocontact shift NMR spectroscopy with biomacromolecules

In this review, lanthanide chelating tags and their applications to pseudocontact shift NMR spectroscopy as well as analysis of residual dipolar couplings are covered. A complete overview is presented of DOTA-derived and non-DOTA-derived lanthanide chelating tags, critical points in the design of lanthanide chelating tags as appropriate linker moieties, their stability under reductive conditions, e.g., for in-cell applications, the magnitude of the anisotropy transferred from the lanthanide chelating tag to the biomacromolecule under investigation and structural properties, as well as conformational bias of the lanthanide chelating tags are discussed. Furthermore, all DOTA-derived lanthanide chelating tags used for PCS NMR spectroscopy published to date are displayed in tabular form, including their anisotropy parameters, with all employed lanthanide ions, C_B-Ln distances and tagging reaction conditions, i.e., the stoichiometry of lanthanide chelating tags, pH, buffer composition, temperature and reaction time. Additionally, applications of lanthanide chelating tags for pseudocontact shifts and residual dipolar couplings that have been reported for proteins, protein-protein and protein-ligand complexes, carbohydrates, carbohydrate-protein complexes, nucleic acids and nucleic acid-protein complexes are presented and critically reviewed. The vast and impressive range of applications of lanthanide chelating tags to structural investigations of biomacromolecules in solution clearly illustrates the significance of this particular field of research. The extension of the repertoire of lanthanide chelating tags from proteins to nucleic acids holds great promise for the determination of valuable structural parameters and further developments in characterizing intermolecular interactions.

What are the methodological and theoretical prospects for paramagnetic NMR in structural biology? A glimpse into the crystal ball

NMR spectroscopy is very sensitive to the presence of unpaired electrons, which perturb the NMR chemical shifts, J splittings and nuclear relaxation rates. These paramagnetic effects have attracted increasing attention over the last decades, and their use is expected to increase further in the future because they can provide structural information not easily achievable with other techniques. In fact, paramagnetic data provide long range structural restraints that can be used to assess the accuracy of crystal structures in solution and to improve them by simultaneous refinements with the X-ray data. They are also precious for obtaining information on the conformational variability of biomolecular systems, possibly in conjunction with SAXS and/or DEER data. We foresee that new tools will be developed in the next years for the simultaneous analysis of the paramagnetic data with data obtained from different techniques, in order to take advantage synergistically of the information content of all of them. Of course, the use of the paramagnetic data for structural purposes requires the knowledge of the relationship between these data and the molecular coordinates. Recently, the equations commonly used, dating back to half a century ago, have been questioned by first principle quantum chemistry calculations. Our prediction is that further theoretical/computational improvements will essentially confirm the validity of the old semi-empirical equations for the analysis of the experimental paramagnetic data.

Joint X-ray/NMR structure refinement of multidomain/multisubunit systems

Data integration in structural biology has become a paradigm for the characterization of biomolecular systems, and it is now accepted that combining different techniques can fill the gaps in each other's blind spots. In this frame, one of the combinations, which we have implemented in REFMAC-NMR, is residual dipolar couplings from NMR together with experimental data from X-ray diffraction. The first are exquisitely sensitive to the local details but does not give any information about overall shape, whereas the latter encodes more the information about the overall shape but at the same time tends to miss the local details even at the highest resolutions. Once crystals are obtained, it is often rather easy to obtain a complete X-ray dataset, however it is time-consuming to obtain an exhaustive NMR dataset. Here, we discuss the effect of including a-priori knowledge on the properties of the system to reduce the number of experimental data needed to obtain a more complete picture. We thus introduce a set of new features of REFMAC-NMR that allow for improved handling of RDC data for multidomain proteins and multisubunit biomolecular complexes, and encompasses the use of pseudo-contact shifts as an additional source of NMR-based information. The new feature may either help in improving the refinement, or assist in spotting differences between the crystal and the solution data. We show three different examples where NMR and X-ray data can be reconciled to a unique structural model without invoking mobility.

Pseudocontact shifts and paramagnetic susceptibility in semiempirical and quantum chemistry theories

Pseudocontact shifts are traditionally described as a function of the anisotropy of the paramagnetic susceptibility tensor, according to the semiempirical theory mainly developed by Kurland and McGarvey [J. Magn. Reson. 2, 286-301 (1970)]. The paramagnetic susceptibility tensor is required to be symmetric. Applying point-dipole approximation to the quantum chemistry theory of hyperfine shift, pseudocontact shifts are found to scale with a non-symmetric tensor that differs by a factor g^T/g_e from the paramagnetic susceptibility tensor derived within the semiempirical framework. We analyze the foundations of the Kurland-McGarvey pseudocontact shift expression and recall that it is inherently based on the Russell-Saunders (LS) coupling approximation for the spin-orbit coupling. We show that the difference between the semiempirical and quantum chemistry pseudocontact shift expressions arises directly from the different treatment of the orbital contribution to the hyperfine coupling.

Metal-ion Binding to Host Defense Peptide Piscidin 3 Observed in Phospholipid Bilayers by Magic Angle Spinning Solid-state NMR

Cationic antimicrobial peptides (AMPs) are essential components of the innate immune system. They have attracted interest as novel compounds with the potential to treat infections associated with multi-drug resistant bacteria. In this study, we investigate piscidin 3 (P3), an AMP that was first discovered in the mast cells of hybrid striped bass. Prior studies showed that P3 is less active than its homolog piscidin 1 (P1) against planktonic bacteria. However, P3 has the advantage of being less toxic to mammalian cells and more active on biofilms and persister cells. Both P1 and P3 cross bacterial membranes and co-localize with intracellular DNA but P3 is more condensing to DNA while P1 is more membrane active. Recently, we showed that both peptides coordinate Cu2+ through an amino-terminal copper and nickel (ATCUN) motif. We also demonstrated that the bactericidal effects of P3 are linked to its ability to form radicals that nick DNA in the presence of Cu2+ . Since metal binding and membrane crossing by P3 is biologically important, we apply in this study solid-state NMR spectroscopy to uniformly 13 C-15 N-labeled peptide samples to structurally characterize the ATCUN motif of P3 bound to bilayers and coordinated to Ni2+ and Cu2+ . These experiments are supplemented with density functional theory calculations. Taken together, these studies refine the arrangement of not only the backbone but also side chain atoms of an AMP simultaneously bound to metal ions and phospholipid bilayers.

Treating Biomacromolecular Conformational Variability

The function of a biomacromolecule is related not only to its structure but also to the different conformations that its structural elements can sample. It is therefore important to determine the extent of the structural fluctuations and to identify the states that are actually populated as a result of the rearrangement. However, this accomplishment is undermined by an intrinsic limitation: the amount of experimental data is by and large inferior to the number of the states that a biomacromolecule can actually sample. This means that additional, a priori information must be applied in order to derive the most from the available experimental data but not to run into overinterpretation. In this chapter we will give a summary of the experimental observables that can be used towards the reconstruction of structural ensembles, how the data can be profitably combined and to what extent the data are affected by error; finally we will give an overview of the computational methods that have been developed to model structural ensembles, highlighting their difference and similarities, advantages and disadvantages.

Reactions in NMR Tubes as Key Weapon in Rational Drug Design

NMR spectroscopy is a powerful technique suitable for obtaining detailed structural and dynamic data at atomic resolution. Progress in NMR instrumentation has led the scientific community to produce novel techniques which provide valuable information to resolve demanding and crucial questions of molecular biology and rational drug design. This chapter outlines the progress of NMR spectroscopy in the rational drug design. In addition, it offers an example of a reaction in NMR tube for achieving rational drug design.

Farseer-NMR: automatic treatment, analysis and plotting of large, multi-variable NMR data

We present Farseer-NMR ( https://git.io/vAueU ), a software package to treat, evaluate and combine NMR spectroscopic data from sets of protein-derived peaklists covering a range of experimental conditions. The combined advances in NMR and molecular biology enable the study of complex biomolecular systems such as flexible proteins or large multibody complexes, which display a strong and functionally relevant response to their environmental conditions, e.g. the presence of ligands, site-directed mutations, post translational modifications, molecular crowders or the chemical composition of the solution. These advances have created a growing need to analyse those systems' responses to multiple variables. The combined analysis of NMR peaklists from large and multivariable datasets has become a new bottleneck in the NMR analysis pipeline, whereby information-rich NMR-derived parameters have to be manually generated, which can be tedious, repetitive and prone to human error, or even unfeasible for very large datasets. There is a persistent gap in the development and distribution of software focused on peaklist treatment, analysis and representation, and specifically able to handle large multivariable datasets, which are becoming more commonplace. In this regard, Farseer-NMR aims to close this longstanding gap in the automated NMR user pipeline and, altogether, reduce the time burden of analysis of large sets of peaklists from days/weeks to seconds/minutes. We have implemented some of the most common, as well as new, routines for calculation of NMR parameters and several publication-quality plotting templates to improve NMR data representation. Farseer-NMR has been written entirely in Python and its modular code base enables facile extension.

Overview of refinement procedures within REFMAC5: utilizing data from different sources

Refinement is a process that involves bringing into agreement the structural model, available prior knowledge and experimental data. To achieve this, the refinement procedure optimizes a posterior conditional probability distribution of model parameters, including atomic coordinates, atomic displacement parameters (B factors), scale factors, parameters of the solvent model and twin fractions in the case of twinned crystals, given observed data such as observed amplitudes or intensities of structure factors. A library of chemical restraints is typically used to ensure consistency between the model and the prior knowledge of stereochemistry. If the observation-to-parameter ratio is small, for example when diffraction data only extend to low resolution, the Bayesian framework implemented in REFMAC5 uses external restraints to inject additional information extracted from structures of homologous proteins, prior knowledge about secondary-structure formation and even data obtained using different experimental methods, for example NMR. The refinement procedure also generates the `best' weighted electron-density maps, which are useful for further model (re)building. Here, the refinement of macromolecular structures using REFMAC5 and related tools distributed as part of the CCP4 suite is discussed.

Paramagnetic NMR as a new tool in structural biology

NMR (nuclear magnetic resonance) investigation through the exploitation of paramagnetic effects is passing from an approach limited to few specialists in the field to a generally applicable method that must be considered, especially for the characterization of systems hardly affordable with other techniques. This is mostly due to the fact that paramagnetic data are long range in nature, thus providing information for the structural and dynamic characterization of complex biomolecular architectures in their native environment. On the other hand, this information usually needs to be complemented by data from other sources. Integration of paramagnetic NMR with other techniques, and the development of protocols for a joint analysis of all available data, is fundamental for achieving a comprehensive characterization of complex biological systems. We describe here a few examples of the new possibilities offered by paramagnetic data used in integrated structural approaches.

Atomic Scale Structural Studies of Macromolecular Assemblies by Solid-state Nuclear Magnetic Resonance Spectroscopy

Supramolecular protein assemblies play fundamental roles in biological processes ranging from host-pathogen interaction, viral infection to the propagation of neurodegenerative disorders. Such assemblies consist in multiple protein subunits organized in a non-covalent way to form large macromolecular objects that can execute a variety of cellular functions or cause detrimental consequences. Atomic insights into the assembly mechanisms and the functioning of those macromolecular assemblies remain often scarce since their inherent insolubility and non-crystallinity often drastically reduces the quality of the data obtained from most techniques used in structural biology, such as X-ray crystallography and solution Nuclear Magnetic Resonance (NMR). We here present magic-angle spinning solid-state NMR spectroscopy (SSNMR) as a powerful method to investigate structures of macromolecular assemblies at atomic resolution. SSNMR can reveal atomic details on the assembled complex without size and solubility limitations. The protocol presented here describes the essential steps from the production of 13C/15N isotope-labeled macromolecular protein assemblies to the acquisition of standard SSNMR spectra and their analysis and interpretation. As an example, we show the pipeline of a SSNMR structural analysis of a filamentous protein assembly.

Perspectives on paramagnetic NMR from a life sciences infrastructure

The effects arising in NMR spectroscopy because of the presence of unpaired electrons, collectively referred to as "paramagnetic NMR" have attracted increasing attention over the last decades. From the standpoint of the structural and mechanistic biology, paramagnetic NMR provides long range restraints that can be used to assess the accuracy of crystal structures in solution and to improve them by simultaneous refinements through NMR and X-ray data. These restraints also provide information on structure rearrangements and conformational variability in biomolecular systems. Theoretical improvements in quantum chemistry calculations can nowadays allow for accurate calculations of the paramagnetic data from a molecular structural model, thus providing a tool to refine the metal coordination environment by matching the paramagnetic effects observed far away from the metal. Furthermore, the availability of an improved technology (higher fields and faster magic angle spinning) has promoted paramagnetic NMR applications in the fast-growing area of biomolecular solid-state NMR. Major improvements in dynamic nuclear polarization have been recently achieved, especially through the exploitation of the Overhauser effect occurring through the contact-driven relaxation mechanism: the very large enhancement of the ¹³C signal observed in a variety of liquid organic compounds at high fields is expected to open up new perspectives for applications of solution NMR.

A protocol for the refinement of NMR structures using simultaneously pseudocontact shift restraints from multiple lanthanide ions

The binding of paramagnetic metal ions to proteins produces a number of different effects on the NMR spectra of the system. In particular, when the magnetic susceptibility of the metal ion is anisotropic, pseudocontact shifts (PCSs) arise and can be easily measured. They constitute very useful restraints for the solution structure determination of metal-binding proteins. In this context, there has been great interest in the use of lanthanide(III) ions to induce PCSs in diamagnetic proteins, e.g. through the replacement native calcium(II) ions. By preparing multiple samples in each of which a different ion of the lanthanide series is introduced, it is possible to obtain multiple independent PCS datasets that can be used synergistically to generate protein structure ensembles (typically called bundles). For typical NMR-based determination of protein structure, it is necessary to perform an energetic refinement of such initial bundles to obtain final structures whose geometric quality is suitable for deposition in the PDB. This can be conveniently done by using restrained molecular dynamics simulations (rMD) in explicit solvent. However, there are no available protocols for rMD using multiple PCS datasets as part of the restraints. In this work, we extended the PCS module of the AMBER MD package to handle multiple datasets and tuned a previously developed protocol for NMR structure refinement to achieve consistent convergence with PCS restraints. Test calculations with real experimental data show that this new implementation delivers the expected improvement of protein geometry, resulting in final structures that are of suitable quality for deposition. Furthermore, we observe that also initial structures generated only with traditional restraints can be successfully refined using traditional and PCS restraints simultaneously.

Variability and conservation of structural domains in divide-and-conquer approaches

The use of protein building blocks for the structure determination of multidomain proteins and protein-protein complexes, also known as the "divide and conquer" approach, is an important strategy for obtaining protein structures. Atomic-resolution X-ray or NMR data of the individual domains are combined with lower-resolution electron microscopy maps or X-ray data of the full-length protein or the protein complex. Doing so, it is often assumed that the individual domain structures remain invariant in the context of the superstructure. In this work, we show the potentials and limitations of NMR to validate this approach at the example of the dodecameric DnaB helicase from Helicobacter pylori. We investigate how sequentially assigned spectra, as well as unassigned spectral fingerprints can be used to indicate the conservation of individual domains, and also to highlight conformational differences.