NMR-based screening is a powerful experimental technique that can be exploited by employing either ligand- or receptor-based methods [V. Roldós, F.J. Cañada and J. Jiménez-Barbero, Chembiochem., 12, 990-1005 (2011)]. We have already described several protocols operating from the ligand’s perspective (see previous NMR Solutions). Now, we will illustrate the typical scheme employed when analysing binding from the protein’s point-of-view.
Receptor-based NMR methods detect and measure some specific NMR parameters of the protein signal resonances in the presence and lack of putative ligands [M. Pellecchia, D. S. Sem and K. Wuthrich, Nat. Rev. Drug Discovery, 1, 211–219 (2002)]. NMR parameters are very sensitive to changes in the chemical environment of the nuclei under analysis, from chemical shifts to relaxation times. Provided that the assignment of the resonances of the target receptor is known, ligand-induced changes in chemical shifts (δ) of different protein cross-peak signals resonances can be directly visualized on the primary sequence, or even in the secondary or tertiary structure of the macromolecular receptor. Indeed, the most commonly used parameter is the chemical shift (δ), usually monitored through heteronuclear 2D-methods. In fact, the protocol consists of following the chemical shift perturbations of key protein NMR signals in the presence of the added ligand. In principle, for protein-detected NMR studies of small proteins, chemical shift variations of specific proton resonance(s) of the receptor, as induced by the presence of the ligand, might suffice for proper monitoring of the binding event. However, signal overlapping in homonuclear 1H-1H 2D NMR spectra (TOCSY, NOESY) becomes a major problem for polypeptides above 50–60 residues. Therefore, heteronuclear experiments (HSQC or its variants) are the method of choice. In principle, either 1H-13C or 1H-15N correlation methods could be employed. In these cases, stable-isotope labelling (13C, 15N) is imperative. Isotope labelling enables monitoring of ligand binding by observation of the chemical shift perturbations on the detected 1H/13C/15N signals, being the observed perturbations associated to the ligand-binding event.
Different aspects should be taken into consideration when performing these experiments. The application of NMR methods to proteins demands specific physicochemical properties of the protein target that might pose difficult challenges. First, milligram quantities of soluble, non-aggregated protein must be over-expressed in the proper culture medium (bacteria, yeast, cells, etc) and purified. Therefore, suitable expression hosts must be found that permit the required isotope enrichment (e.g. 13C, 15N, 2H), which is critical for the resonance assignment process of large (>30000 Da) protein targets. Therefore, the experimental access to the required data might be rather expensive. After sufficient quantities of labelled receptor are available, it must be ensured that the sample is stable in solution for the time required for the sequential resonance signal assignment process. Typically, at least one week of measurement time might be necessary. Once the data are collected, the key cross peaks that will be monitored should be assigned. Although new data acquisition approaches promise to accelerate resonance assignment, it can still be a lengthy process (at least weeks) for relatively large proteins (>30000 Da) routinely encountered in research. Signal assignment is crucial since then, by identifying chemical shift perturbations in the assigned protein resonances, not only are ligands identified, but also the binding sites at the protein structure are localized in a straightforward manner. Moreover, localization of binding sites may also make it possible to distinguish immediately between specific and nonspecific binding. Unlike ligand-based methods, receptor-based NMR methods do not rely on fast exchange to retrieve bound state information. In this case, and contrary to ligand-based NMR techniques, observation of the interaction process may be possible for a larger range of affinities, from millimolar to nanomolar [T. Peters, B. Meyer, Angew. Chem., Int. Ed., 42, 890–918 (2003)]. Monitoring the chemical shift perturbations of the protein signals in the absence and in the presence of different ligand concentrations permits the characterization of both higher and lower affinity hits, including the estimation of the binding affinity, provided that a carefully designed titration experiment is carried out.
Preferably, HSQC experiments should be performed using a variety of ligand/protein ratios. Therefore, even having access to a good concentration (ca. 0.1 mM) of the receptor, at least one day of measurement is required to obtain good data. As for any 2D NMR experiment, for every ligand/protein ratio, a given number of fids (at least 128-256) have to be recorded to achieve the required digital resolution in the evolution dimension (f1), which can be further expanded using linear prediction algorithms. The number of scans is dictated by the amount of material available and by the sensitivity of the NMR instrument. A minimum magnet corresponding to a 600 MHz 1H NMR Larmor frequency is required (for proteins up to 20 kDa). For larger systems, bigger magnets are required. The NMR instrument to be employed depends on a variety of factors: first, and more importantly, on the accessibility to a high-field spectrometer, depending on the complexity and the degree of overlapping of the heteronuclear NMR spectrum of the protein. Therefore, the availability of very high-field magnets and especially the possible access to instruments equipped with cryo-probes allows NMR experiments to be performed with minute amounts of sample. For instance, nowadays, the use of cryo-probes allows protein amounts to be used in the micromolar range. In any case, there is a limit in the size of the protein, which can be easily analysed by heteronuclear NMR, of approximately 40 kDa. In this case, additional labelling of the backbone and side chain C-H with deuterium is advisable. Nevertheless for protein oligomers, due to symmetry reasons, receptor molecules displaying larger molecular sizes can also be monitored. For large molecules with this or beyond this size, relaxation is very fast. Thus, line broadening becomes important, and most of the NMR signals show very little intensity and may even disappear below the noise level. There are some heteronuclear experiments that have been devised to minimize the intensity losses due to relaxation. The most popular ones are TROSY (transverse relaxation-optimized spectroscopy, K. Pervushin, R. Riek, G. Wider, K. Wüthrich, Proc Natl Acad Sci USA, 94, 12366-12371 (1997) and CRINEPT-like techniques (cross-correlated relaxation enhanced polarization transfer, R. Riek, G. Wider, K. Pervushin and K. Wüthrich, Proc Natl Acad Sci USA, 96, 4918-4923 (1999)).
As standard, 0.5 mL of 0.05–0.2 mM of the receptor (protein) is used in this experiment. The ligand is usually present in 1–20 fold molar over the concentration of protein binding sites. The selected ligand/protein ratio depends on the binding affinity and on the kinetics of the exchange reaction. If possible, three or four molar ratios should be tested. First, specific heteronuclear NMR spectra of the free protein is separately recorded and analysed. Secondly, the protein sample is titrated using a concentrated stock solution of the ligand in the corresponding buffer. The possible existence of chemical shift perturbations are carefully monitored to assess binding for the different ligand/protein molar ratios.
As usual, the binding ability or the activity of the protein should be monitored before and after performing the NMR experiments. For extracting sound conclusions, the data collected after a significant loss of binding or specificity should be used with extreme caution, since the architecture of the binding could have changed. This fact is relatively easy to detect by the heteronuclear experiments.