Introduction

NMR-based ligand screening (NMR screening, see Glossary) is now a well-established methodology for drug discovery. As a valuable complement to other available techniques, such as high-throughput screening (HTS), combinatorial chemistry and structure-based drug design, NMR screening is being broadly applied in pharmaceutical research. A commonly used strategy for fragment-based drug discovery consists in building up high-affinity ligands in a modular way, starting from small scaffolds or “fragments” and growing, linking or merging them into high-affinity ligands. Because the individual fragments generally bind with low affinity, they are difficult to identify with conventional assays. NMR spectroscopy with its intrinsic high sensitivity to detect weak interactions becomes the method of choice to pick up such low-affinity fragments by screening of a compound library. This process has led to the discovery of numerous high affinity ligands for several biologically relevant protein targets (numerous practical examples are presented in the Related Articles).

NMR screening methods can be divided into two categories: methods that observe target resonances and methods that observe ligand resonances. In screening methods that observe the macromolecular target, the parameters that are typically monitored are the chemical shifts (see Glossary). The advantages offered by these target observation methods are substantial: the NMR assay is applicable to any class of compound, with no upper limit in affinity (typically, compounds with dissociation constants from mM to nM and lower can be monitored), identification of the ligand binding site is possible, and measurement of the dissociation constant (K_d) can be pursued for millimolar and micromolar ligands [1]. The main disadvantages of these techniques are the requirement of large amounts of well soluble and isotopically labeled protein, the need to deconvolute compound mixtures, and the upper limit for protein size. The “SAR-by-NMR” technique (see Glossary) belongs to the target-observe methods [2].

For experiments that rely on observation of ligand resonances, the choice of NMR parameters is more diverse. These include longitudinal, transverse (e.g. T_1ρ relaxation experiments, see Glossary) and double-quantum relaxation, diffusion coefficients, and intramolecular and intermolecular magnetization transfer (including transferred NOE, NOE pumping, saturation transfer and WaterLOGSY experiments; see Glossary) (for detailed reviews, see 3, 4 and 5). These experiments do not require isotopically labeled proteins, the one-dimensional (1D) experiments used by these methods typically require a shorter acquisition time than the 2D experiments used in the screening methods with target observation, the hits can be identified directly without the need to deconvolute them from compound mixtures, and there is no upper-limit to the size of target that can be screened (in fact, many of them work best for larger proteins). However, a general drawback of the ligand-observation methods is their inability to detect high-affinity ligands. In strongly bound ligands, slow dissociation rates prevent transfer of the properties of the relative small fraction of bound ligand molecules to the bulk unbound ligand molecules, which are in effect the ones observed by the NMR experiments. Competition-based experiments (see below) can eliminate this drawback.

In the past few years, several new screening methods have been developed. Some of them have been devised to combine the advantages from both, ligand-detected and target-detected techniques, and to overcome intrinsic limitations of available methods. A general trend is observed to design NMR experiments and instrumentation that yield increased sensitivity, leading to higher throughput and/or decrease of protein consumption. At the same time, novel NMR techniques that allow measurements on larger molecular targets have been recently developed 6, 7, 8 and 9. This has extended the applicability of NMR to studies of soluble proteins with molecular weights well above 50 kDa [10], of integral membrane proteins in micelles 11 and 12 and of macromolecular assemblies [13]. A brief description of recent advances in NMR screening techniques is given in the next sections.

New methodologies for NMR screening

Reporter screening

As discussed above, the main issue with most ligand-detected experiments is their limited affinity range, which is often restricted to 3–4 orders of magnitude in K_d, within the range 1 mM to 100 nM. One approach that has recently been used to overcome this constraint is competition-based experiments, such as NMR reporter screening 14, 15 and 16. By this method, not the binding to the protein target of the test compounds is directly observed, but the ability of a test compound to displace a known ligand that is added to the mixture of protein and test compounds as a “reporter ligand” or “spy molecule”, which binds to the protein with medium affinity (Fig. 1). Bound reporter ligands can be readily distinguished from unbound ligands by their resonance signals as seen in conventional 1D ¹H or ¹⁹F spectra, proton T_1ρ relaxation, WaterLOGSY, or saturation transfer difference (STD) spectra.

(33K)

Figure 1. Principle of NMR reporter screening. In this experiment, the resonance signals of a reporter ligand (triangles) are detected in the presence of the target protein and test compounds in 1D proton spectra. (a) Without protein target, the signals of the reporter ligand are sharp, as commonly observed for small, unbound molecules. (b) In the presence of the target protein but in the absence of test compounds, the reporter ligand is bound to the target protein with moderate affinity, and its relaxation rate is increased, as evidenced by the severe line broadening of the reporter ligand resonances in the NMR spectrum. (c) After adding test compounds that bind to the same site as the reporter ligand with higher affinity, the reporter ligand is displaced from the binding site, so that it becomes unbound. Unbound reporter ligand can be readily distinguished from bound reporter ligand by its sharp resonances in the NMR spectrum (peaks colored blue, compare with spectrum in (a)). If test compounds are added that do not bind to the protein target, bind more weakly than the reporter ligand, or bind to a different, non-competitive binding site, the test compounds would not displace the reporter ligand from its binding site, and therefore it would remain bound to the target. This would be manifested by broad lines in the NMR spectrum of the reporter ligand, as seen here in (b).

Reporter screening has several additional advantages when compared with conventional NMR screening methods based on detection of ligand signals (Table 1). First, it allows observation of high-affinity ligands that are commonly detected as “false negative” by other ligand-observation NMR methods. Moreover, it only detects ligands that bind to the protein active site, while not detecting unspecific binding. As an alternative, if fluorine-containing reporter compounds are used, ¹⁹F NMR spectroscopy can be applied, which can increase the throughput and significantly reduce signal overlap in the NMR spectra 17 and 18.

Table 1.

Overview of five NMR screening technologies

Technology	Reporter screening	Spin labels	NMR screening with methyl group resonances	3-FABS	Affinity tags
Pros	- Affinity range extended to high-affinity ligands	- Relatively high sensitivity	- Relatively high sensitivity	- Relatively high sensitivity	- Relatively low protein concentrations needed
	- Possibility to discriminate specific from unspecific binding	- Very robust in identifying compounds that bind simultaneously at neighboring binding sites	- Works very well with larger proteins (MW > 50 kDa)	- Possibility to determine IC50 values from measurements at variable inhibitor concentrations
	- Possibility to rank ligand affinities using titration series	- Possibility to obtain distance information and orientation of ligands with respect to each other and with the protein	- Valuable complement to experiments based on amide group chemical shift changes, especially when the ligands bind to hydrophobic surface areas of the protein
Cons	- Hit identification requires deconvolution of the mixture components	- Chemistry is required to introduce spin labels in compounds or in protein side chains	- 13C-labeling of the protein methyl groups is required	- Chemistry is required to introduce CF3 groups in compounds	- The sequence of one of the protein partners has to be modified (addition of a ligand binding domain)
	- Need to identify suitable reporter ligand(s)			- Only applicable to enzymes
References	14, 15 and 16	19, 20 and 21	[22]	24 and 25	[26]

Spin labels for NMR screening

Spin labels can be used to identify and characterize intermolecular interactions. Recently, several applications of spin labels for NMR screening have been reported [19]. They are based on observation of relaxation enhancement of ligand resonances caused by its proximity (typically up to 15–20 Å apart) to a spin-labeled group, for instance, a paramagnetic organic nitroxide radical such as TEMPO. The spin labels can be covalently attached either to the protein, or to a first ligand.

The SLAPSTIC method (spin labels attached to protein side chain as a tool to identify interacting compounds) (Fig. 2) is an extension of the traditional T_1ρ relaxation experiments, where one forces amplification of the bound state relaxation properties via covalently attached spin labels to protein side chains, such as lysine, tyrosine, cysteine, histidine and methionine 19 and 20. As a condition, at least one residue of this type should be as close as possible to the binding site, but must not interfere with ligand binding. SLAPSTIC is a very sensitive NMR screening technique: it allows a reduction of protein demand by 1–2 orders of magnitude compared to T_1ρ relaxation experiments on non-modified protein targets.

(40K)

Figure 2. Principle of the SLAPSTIC experiment. Small, unbound organic compounds in solution have typically sharp resonances (spectrum at the top). When they bind to a paramagnetic spin-labeled protein target, their signal intensity is drastically reduced or completely quenched by paramagnetic relaxation enhancement (spectrum at the bottom).

Second-site NMR screening can be efficiently performed with the use of spin labels 19 and 21. This method utilizes a spin-labeled compound as a first-site ligand. Screening this complex with a library of compounds allows identification of compounds that bind simultaneously with the first, spin-labeled ligand, in a neighboring binding site (second site). Second-site ligands can then be identified from quenching of their NMR signals by the spin-labeled first ligand. It is important to remark that this relaxation enhancement will be manifested, if and only if both ligands bind simultaneously to the target protein and at neighboring binding sites. This is of special interest during the drug discovery and optimization process, because linking two “fragments” under these conditions can generate compounds with significantly higher affinity than the affinities of the two individual compounds. The principal advantages of this approach are its robustness to identify second-site ligands and its high sensitivity.

NMR screening based on methyl group chemical shifts

As an alternative to NMR screening by observation of protein target resonances in 2D [¹⁵N,¹H]-HSQC spectra, Fesik and coworkers suggested to monitor¹³C/¹H chemical shift changes of methyl group resonances in 2D [¹³C,¹H]-HSQC spectra [22]. Application of this method demands labeling of the methyl groups with ¹³C. Because uniform ¹³C-labeling of proteins is relatively expensive, a protocol for cost-effective selective labeling of valine, leucine and isoleucine(δ¹) residues in proteins has been proposed by the authors [22].

In NMR binding studies that monitor methyl group chemical shift changes, the sensitivity is increased about threefold compared with the corresponding experiments based on ¹⁵N/¹H chemical shift observation. In addition, selective methyl group labeling on a perdeuterated background is advantageous for screening high molecular weight protein targets (MW > 50 kDa). Finally, NMR screening or binding assays that monitor methyl group chemical shifts offer a valuable complement to experiments based on amide group chemical shift observation, especially in cases were the ligands are located close to valine, leucine or isoleucine methyl groups.

3-FABS

All NMR screening techniques discussed so far are binding assays that detect ligand binding rather than enzyme inhibition. Functional screening of enzyme inhibition can be performed by NMR spectroscopy if substrate and product concentrations are monitored as a function of time. This is a well-established NMR technique (reviewed in [23]). Recently, an improvement was proposed in a method called 3-FABS (three fluorine atoms for biochemical screening) 24 and 25. This method allows rapid and reliable functional screening of compound libraries, performed at protein and substrate concentrations comparable to the ones utilized by standard HTS techniques. The experiment is only applicable to enzymes and allows measurement of accurate IC₅₀ values.

3-FABS monitors ¹⁹F signal intensities rather than ¹H signals. Signal overlap with test compounds is thus drastically reduced. 3-FABS requires labeling of the substrate with a CF₃ moiety (Fig. 3). During the assay, the enzymatic reaction is performed with the CF₃-labeled substrate and quenched after an established delay that depends on the enzyme and the reaction conditions. Fluorine NMR spectroscopy is then used to monitor the substrate and the enzymatically modified reaction product. Because of the high intrinsic sensitivity of ¹⁹F chemical shifts to the chemical environment, modifications of the substrate during the enzymatic reaction are readily detectable, even when the CF₃ moiety is distant from the reaction site. The high sensitivity of ¹⁹F NMR spectroscopy, the 100% natural abundance of the isotope ¹⁹F and the presence of three fluorine atoms result in ¹⁹F signals of high intensity.

(24K)

Figure 3. Schematic diagram showing the principle of the 3-FABS method. First, a fluorine-containing moiety, like CF₃, is introduced in the substrate as a “sensor” or reporter group. The chemical modification of the substrate by the enzyme (here represented as addition of a chemical fragment denoted “A”) induces changes in the electronic cloud of the CF₃ moiety. This results in distinct chemical shifts for the product and the substrate ¹⁹F signals, and therefore chemical shift changes in the ¹⁹F NMR spectra (bottom).

Affinity tags

Protein–protein interactions can be detected by a novel NMR reporter system based on affinity tags [26]. In this approach, one of the binding partners is fused to a ligand-binding domain, where a medium-affinity, low-molecular weight reporter ligand is bound. Protein–protein interactions are then monitored via changes in the NMR relaxation of the reporter ligand: because the parameters of the reporter ligand spectra depend on the molecular weight of the protein–ligand complex (among other factors, e.g. affinity constant, and protein and ligand concentration), changes of these spectral parameters can be used to probe changes in the molecular composition of the ternary protein–protein–ligand complex. One of the principal advantages of this technique is the relatively low consumption of unlabeled proteins.

NMR instrumentation and experimentation

All NMR techniques currently applied for drug discovery, including the ones discussed above, take advantage of recent advances on NMR experimentation and instrumentation. Particularly, technologies that improve throughput or allow reduction of sample consumption are very attractive. Some of them are discussed briefly in the following paragraphs.

The introduction of cryogenic probes (see Glossary) (http://www.bruker-biospin.com; http://www.varianinc.com) has been a significant step ahead in NMR technology. By cooling the radio-frequency coils and preamplifiers to 30 K, the electronic noise can be significantly reduced, which results in a signal-to-noise ratio increase by a factor of 2–4 (depending mainly on sample conditions). This translates in reduction of measurement times by factors of 4–16, or reduction of protein consumption by factors of 2–4, to achieve equivalent signal-to-noise ratios. In the context of NMR screening, these numbers represent considerable saving of resources, because NMR screening involves preparation of a large amount of NMR samples. By contrast, cryoprobe technology opens the possibility to perform measurements at even lower concentrations, which allows detection of ligands with lower affinities and compounds with low solubility.

A further approach to increase the probe sensitivity has come from microcoil technology. Although microcoils have been commercially available for several years for work with small molecules, they have been so far scarcely used for biomolecules. A recent publication showed that a newly introduced CapNMR microcoil probe (MRM Corp.; http://www.protasis.com), working with sample volumes as low as 5 μl, can provide useful NMR spectra of biomolecules [27]. Despite of the higher mass sensitivity of this probe (approximately 7.5 times higher than a conventional 5 mm probe), they need much higher protein concentrations than 5 mm probes, which is a drawback for practical work with typical protein samples. Future developments, perhaps the further improvement of cryogenic microcoil probes, will probably extend the application scope of these probes for routine NMR work.

Recent advances in the NMR field, particularly in techniques that speed up NMR data acquisition, have a considerable potential for NMR screening. During the past few years, extensive activity has been going on with the aim to speed up multidimensional experiments, provided that the inherent sensitivity is high enough. Among the most promising “fast multidimensional NMR techniques” [28], as they are coined, are GFT-spectroscopy [29], single-scan multidimensional spectroscopy [30], Hadamard spectroscopy [31] and projection reconstruction techniques [32]. To date, the majority of these techniques have been tested using small, well-behaving proteins or peptides, where they could increase data collection speeds from 1 to several orders of magnitude. However, it remains to be seen whether these new methods will indeed dramatically speed up the characterization of larger protein molecules and have a significant impact in the process of drug discovery. In the future, with upcoming advances in NMR instrumentation, some of these techniques might become very useful tools for NMR screening.

Comparison of technologies

The technologies compared include reporter screening, spin labels for NMR screening, NMR screening monitoring methyl group resonances, 3-FABS and affinity tags for study of protein–protein interactions. All these methods have advantages and disadvantages and should be regarded as complementary. Table 1 lists the most relevant characteristics of these techniques.

Conclusions and outlook

NMR screening is nowadays widely applied for validation of hits identified in other assays and for the discovery of high-affinity ligands for biologically relevant macromolecules. NMR spectroscopy offers several advantages, like applicability to any soluble protein, without the need for prior development of a target specific set-up, robustness against detection of false positives, and versatility, which is shown by the vast number of experiments or “assays” that are possible and the large number of spectral parameters that can be monitored. Moreover, NMR screening techniques can be applied at any point in a drug discovery program, be it for hit finding very early in the program, long before an HTS enzymatic assay is developed, or for hit validation, or during the process of lead optimization.

In the drug discovery process, the interplay of NMR with other screening methods is likely to be of increasing importance, because these techniques complement each other very well. Particularly powerful are methods that combine NMR screening with HTS, mass spectrometry, X-ray crystallography and in silico screening, to exploit the benefits of each technique. Particularly, X-ray-based screening has the unrivalled advantage of providing direct structural information at atomic resolution of the protein–ligand complex, which can be immediately used for further optimization of the compound. Usually, collection of this information by NMR spectroscopy, if possible at all for the system studied, is very time-consuming. By contrast, NMR screening is relatively fast in delivering a hit list for the target, whereas X-ray-based screening requires availability of good-quality crystals that are suitable for soaking or co-crystallization. Furthermore, X-ray screening can miss numerous ligands if the particular protein–ligand complex does not crystallize.

A challenging and exciting frontier for NMR screening lies with those macromolecules that resist structure determination by conventional methods. These targets include integral membrane proteins, particularly G-protein coupled receptors (GPCRs), macromolecular complexes, and molecular machines, such as the ribosome. Another potential application of NMR in drug discovery will probably be “in vivo” NMR screening, where NMR is used as a tool to monitor protein–ligand interactions within living cells 33 and 34. An attractive feature of this method is that proteins can be studied in the presence of other proteins and endogenous small molecules, which represents more appropriate physiological conditions.

Further developments in NMR instrumentation and methodologies will lead to higher throughput and application of NMR screening technologies to even larger protein targets. As NMR hardware and techniques continue to improve sensitivity and resolution, NMR screening will become a powerful tool for designing novel therapeutics that target the most challenging biological systems. Without any doubt, further technical progress and innovative methods in the NMR field will have a profound impact on drug discovery in the near future.