Modern-day drug discovery and development is a multi-step approach, from finding hits to lead optimization. This approach starts with the identification of a "druggable" target for a particular disease and testing of its validity. Once the target is selected, the initial screening is done on a library of compounds to identify the hits that can bind to the target. Then the chosen lead compounds from the preliminary screening are subjected to further optimization based on the improvement of the properties of these lead compounds regarding better potency, metabolic stability, affinity, selectivity, oral bioavailability, and reduced side effects. The lead compounds that are shortlisted by the accomplishment of these properties are further subjected to clinical trials.

For exploring the drug discovery and development process, understanding of the structure of biological molecules at the atomic level is essential to use that knowledge to design drug candidates that can target them. Nuclear Magnetic Resonance (NMR) spectroscopy plays a vital role in attaining the desired goal, with a high success rate in screening compounds that can be used as potential drug candidates for curing diseases. During the past decade, NMR spectroscopy has been a very efficient and versatile tool in drug discovery and development as it can shed light on the molecular structure of the biomolecules, elucidate and verify the structure of the drugs, and provide structural information on the interaction of the biomolecules (target) with small molecule compounds (ligands); thus NMR spectroscopy proves to be a great tool in pharmaceutical research [1], for example, biocatalytic manufacture of drug islatravir [2]. Since the clinically used drugs are typically natural or synthetic compounds, quantitative analysis by solution-state NMR is quite useful in estimating the contamination profile of the drugs, describing the composition of drug products, and exploring the metabolites of drugs in body fluids, or study the dynamics and kinetics of proteins on solid surfaces [3, 4], or enzyme allostery [5]. Solid-state NMR methods can offer knowledge about polymorphism of drugs in powder form and their conformations in tablets or the active site of tryptophan synthase [6], or the structural properties of the toxic and non-toxic types of α-synuclein oligomers [7]. Microimaging is useful in studying the dissolution of tablets [8].

Figure 1. An overview on application of NMR spectroscopy in drug discovery and development.

This review will mainly focus on the solution state NMR methods implemented for target druggability, hit identification and validation followed by lead optimization in the field of drug discovery and development. In the recent past, many new NMR-based strategies have been introduced to increase the sensitivity of the methods to get higher throughput in hit identification, validation, and lead optimization as well as in evaluating target druggability. An overview of NMR screening methods implemented in drug discovery and development is shown in figure 1. The NMR-based screening methods mainly involve two modes of detection based on either target resonances or ligand resonances.

NMR screening methods constitute a very reliable and valuable tool for the identification of small molecules and hit-to-lead optimization. A library of compounds is screened for finding the hits that can bind to a specific target followed by their validation, for which NMR binding assays can be of high efficacy. These methods include the following.

In chemical shift perturbation approach [9], the information on the intermolecular interaction between the compound (ligand) and the protein (target), can be extracted based on the chemical shift differences between the free and bound protein. This chemical shift difference is generated due to the binding of the compound to the target protein, which causes perturbation in the chemical shift of the magnetic nuclei at the binding site. For this method, the target protein needs to be labeled with stable isotopes, ¹⁵N and/or ¹³C, for acquisition of the ¹⁵N/¹H and ¹³C/¹H two-dimensional hetero-nuclear correlation NMR spectra of the free and the bound forms of the protein [10, 11]. The uniform isotope labeling of the protein is essential to increase the sensitivity and resolution as well as reduce the complexity of the NMR spectra. Besides, it also helps in providing the sequential resonance assignment using the multidimensional triple resonance experiments [12], which together with chemical shift analysis can provide unequivocal identification of the location of the binding site of the ligand on the target protein.

Thus this method has many advantages. Apart from being very reliable and reproducible, it can provide crucial structural information on the ligand-binding site. It can detect compounds with low to high affinities for the target (generally compounds with dissociation constants on the mM to nM range); even extremely weak binding can be probed. The dissociation constant (Kd) can be measured by following the resonance of the target as a function of the ligand concentration in titration experiments and correlating the fractional chemical-shift change with the total ligand concentration [13]. If the structure of the target has been elucidated using NMR methods, the ligand-protein distance can be derived by nuclear Overhauser effect (NOE)-type experiments [14] that would allow more accurate binding mode determination of the ligand.

There is another NMR target-resonance based approach introduced by Fesik and his coworkers [15] for reducing the complexity of NMR screening methods and increasing the efficiency of high-throughput screening. In this NMR screening, methyl group chemical shift changes are observed. The methyl groups in individual amino acids are selectively labeled with ¹³C and monitored through ¹³C/¹H chemical shift changes of the methyl resonances in 2D ¹³C-¹H HSQC spectra [15]. This approach is advantageous in terms of increasing the sensitivity by threefold in comparison to experiments based on ¹⁵N/¹H chemical shift change and offers a complementary approach to such experiments especially when the ligands are located close to valine, leucine or isoleucine methyl groups. Further, screening of high molecular weight protein targets (MW > 50 kDa) is feasible with selective methyl group labeling in a perdeuterated solvent.

The ligand resonance-based NMR screening methods are more diverse than target-based approaches and therefore have additional benefits; for example, these experiments do not require isotope-labeled proteins, only a small quantity of protein is needed, less acquisition time is required for the experiments as they are generally one-dimensional and not the two-dimensional experiments required for a target-based approach, and there is no upper limit on the size of the target for screening. However, fluorine NMR spectroscopy has an reproducibility issue due to the spectrometer design and the scale system for the calibration of heteronuclear NMR spectra without internal referencing [16].

Saturation Transfer Difference (STD) NMR experiment involves selective irradiation of proton resonances of protein using a Gaussian pulse train. The saturation is rapidly propagated across the entire protein due to the spin diffusion effect. If ligand binds to a protein, saturation is also transferred to the bound ligand by cross relaxation at the ligand-protein interface, leading to the attenuation in the intensity of the ligand signal. The STD spectra containing signals of the binding ligands are obtained by subtracting the resulting spectrum from the reference spectrum without saturation [17]. This method helps in epitope mapping (determination of the pharmacophoric groups) of the direct binding segments of the ligand and identification of a ligand directly from a mixture of compounds.

Water-Ligand Observed via Gradient SpectroscopY (WaterLOGSY) is a valuable method for the identification of ligand that binds to a target biomolecule. In this NMR experiment, the magnetization is transferred from bulk water via the protein-ligand complex to free ligand. The resonances of the bound ligand appear with the opposite sign than those of the non-bound ligands, and hence are very useful for primary NMR screening [18]. However, this method has some disadvantages; for example, it does not provide information about the ligand-binding site. Also, it is unable to identify ligands having strong binding affinity for the target and slow dissociation rates. To overcome this problem, another binding experiment is designed, i.e., Competition Water-Ligand Observed via Gradient SpectroscopY (c-WaterLOGSY) [19] that allows the detection of strong binders. This experiment is applicable for the identification of potential strong inhibitors having slight solubility, as it requires low ligand concentration. The technique is predominantly appropriate for rapid screening of chemical mixtures and plant or fungi extracts. It is notable that similar competition-based approach could be applied to many other ligand-detected NMR screening methods. The basic principle for the competition-based experiments [20] involves, initially, identification of a reference compound with medium to low affinity for the target that binds to the target protein, followed by its displacement with another test ligand which binds to the same target with higher affinity. Thus, these competition-based screening methods have additional advantages concerning conventional methods, such as aiding in identifying high-affinity ligands which are not detected by conventional methods, detecting ligands that bind to the active site of the protein and hence avoiding non-specific binding ligands.

Spin Labels Attached to Protein Side chains as a Tool to Identify interacting Compounds (SLAPSTIC) is useful and sensitive for primary screening of compounds by NMR as the intermolecular interaction between the ligand and the protein can be identified and characterized by spin labels [21]. The spin labels (e.g., a paramagnetic organic nitroxide radical TEMPO) are covalently attached to the protein and cause the paramagnetic relaxation enhancement (PRE) of ligand resonances if they are near (in general up to 15-20 Å distance) to spin label groups. For instance, a small ligand in solution generally has sharp NMR resonances. When it bounds to the paramagnetic spin-labeled protein target, it experiences a significant reduction in the resonance intensity due to PRE effect by the spin label on the ligand [22].

Target Immobilized NMR Screening (TINS) process is rapid, sensitive, and identical for every target irrespective of their size and chemical composition. In this method, the ligands are screened based on their binding capability to the immobilized protein target. The mixture of compounds from a chemical library is applied to the immobilized protein target, and binding is detected by comparing the resulting 1D NMR spectrum with that of a suitable control sample [23]. This method has been validated for a variety of ligands that target proteins and nucleic acids with a K_D from 60 to 5000 μM. For fast characterization of the ligand-binding site, TINS can be used in competition mode. This method has several advantages. It requires a smaller amount of target protein in comparison to other fragment-based approaches, as the protein can be reused to screen the library of compounds. It is sensitive to the binding of a ligand to the target with a wide range of affinities and therefore not likely to overlook new hits. Besides this, weak binding interaction and binding of the ligands to other low-affinity allosteric sites can be eradicated. A very high level of specificity can be achieved by selecting the control sample cautiously. It may also be useful for screening of hard targets like membrane proteins, that are difficult to produce or insoluble.

Relaxation-edited NMR experiments are instrumental in monitoring a wide range of ligand affinity. Since binding of a ligand to target protein alters relaxation time, this allows an estimation of the affinity and the interacting functional groups are identified by the build-up curves. This relaxation-edited NMR approach [24] includes rotating frame nuclear spin longitudinal relaxation time (T₁ρ) and Transverse nuclear spin relaxation time (T₂) measurements as a mode of detection. For example, a target (protein) molecule is generally big, and therefore has slow translational diffusion, slow tumbling, fast relaxation, shorter T₂, broad linewidths and negative NOE, whereas ligands are small molecules and have rapid translational diffusion, fast tumbling, slow relaxation, longer T₂, narrow linewidths and positive NOE. When a ligand binds with the protein in fast exchange, diffusing less rapidly and relaxing quickly, negative NOE will be observed for the ligand, as it acquires NMR properties similar to that of the target, and the resonance lines will be broadened, giving a clear indication of binding.

Diffusion-edited NMR experiments involve the measurement of diffusion rates for ligands in the bound and free state. Since ligand in the free form diffuses more rapidly than bound form because of their size differences, this feature can be exploited in screening the ligands that can bind to the target. For this type of experiment, diffusion filters based on pulsed-field gradient (PEG)-stimulated echo (STE) experiments [24] are used. BT Falk et al, for example, used 1D and diffusion profiling methods to optimize ultra-rapid-acting insulin formulation [25].

Transferred Nuclear Overhauser effect (Tr-NOEs) is a useful screening tool to distinguish ligands from a mixture of compounds that can bind to a given target protein, or identify pairs of molecules that bind to a protein concurrently on allosteric sites (interligand NOEs). If the ligand has low-affinity binding or weak interactions with the target protein, transferred NOE (trNOE) spectroscopy method [26] can provide an excellent alternative. TrNOE renders information about the orientation of the ligand at the protein-binding site. Two of the above-mentioned techniques, STD and Water-LOGSY, are based on transfer-NOE type experiments.

NOE Pumping can give reliable knowledge about the protein-ligand interactions without the need for isotope labeling or chemical separation. In these experiments, the magnetization is transferred from protein to the bound ligands or vice versa, and the information of protein-ligand interactions are obtained through the dipole-dipole interactions between them. Magnetization transferred from the protein is accrued in the free ligands because of quick exchange between free (slow relaxation) and bound (fast relaxation) ligands, leading to a pumping effect for the NOE [27]. However, this method has some limitations, such as that significant difference in transverse relaxation times or diffusion coefficients between the target and ligands are essential, and the binding interaction between the free and bound states of the ligand must be in the fast exchange regime on an NMR time scale.

Affinity tags NMR binding assay can detect protein-protein interaction through the use of affinity tags [28]. In this approach, one of the protein-binding partners is attached to a ligand-binding domain having a medium affinity for the ligand. The interaction between the protein and its potential binding partner is probed via changes in the relaxation rate of the ligand, which is reversibly bound to the ligand-binding domain. The change in the relaxation rate of the ligand is determined by the molecular weight and molar ratio of the ternary protein-protein-ligand complex. The major advantage of this method is the relatively low quantity of unlabeled protein that is required.

Fluorine chemical shift Anisotropy and exchange for Screening (FAXS) is reliable, highly reproducible, large dynamic range, ligand-based ¹⁹F NMR screening method [29]. The basic principle of this method is that when the competitive ligand binds to the target protein, it displaces the fluorinated spy molecule that is already bound to it. The signal of the fluorinated spy molecule that was broad initially becomes sharp due to its displacement from the receptor proteins and is detected by ¹⁹F NMR. This method has significant advantages over the original competition-based approaches using ¹H NMR. For instance, the screening of a mixture of compounds becomes easier due to the absence of spectral overlap and requires low consumption of protein. A large chemical shift anisotropy of fluorine and a significant exchange contribution helps in selecting a weak-affinity spy molecule, thus resulting in a lower binding affinity threshold for the identified NMR hits.

Three Fluorine Atoms for Biochemical Screening (3-FABS) is a substrate-based ¹⁹F NMR method [30] mainly used for functional screening of enzyme inhibition reactions. In this method, a CF₃-labeled substrate is used to perform the enzymatic reaction in which a CF₃ moiety is used as a sensor group. The chemical modification of the substrate by the enzyme induces a change in the electronic cloud of the CF₃ moiety of the substrate, which leads to distinct ¹⁹F chemical shifts for the substrate and the enzymatically-modified reaction product. This technique is mainly used for monitoring enzymatic reactions. It allows rapid and reliable functional screening of compound libraries and accurate measurement of IC50 values [31].

NMR screening methods have broad application during the lead optimization process to improve the pharmacokinetic properties of the ligand. Since hits obtained are generally weak binders, their optimization is essential; this can be achieved by growing, merging or linking the ligands. NMR methods can provide high-quality structural information for complexes of weakly bound ligands, and also used to characterize the proteins that do not crystallize. NMR methods for lead optimization also have two modes of detection, target resonances, and ligand resonances.

Target resonance-based includes Structure-Activity Relationship (SAR) by NMR [32, 33] is a linked-fragment approach to increase the binding affinity of the ligand. In this method, ligands are identified from a library of compounds by ¹H-¹⁵N HSQC experiment, based on their binding affinity for the target, and are then optimized through chemical modification. In the presence of saturating amounts of the previously optimized first ligand, identification of the second ligand is performed along with its optimization for the second site. Finally, the two ligands are linked covalently to get the high-affinity ligand that can bind to two neighboring sites, and the ligand binding is tested by chemical shift mapping. This technique has broad application, as the structural information obtained for the two lead compounds are quite useful in the chemical synthesis of a bi-dentate ligand with higher affinity.

H₂O/D₂O exchange-rate measurement is useful in identifying the binding epitope [34]. In this method, the rate of exchange of a covalently bonded hydrogen atom (amide proton of the protein backbone) with a deuterium atom, or vice versa, is measured. For this purpose, the exchange reaction is carried out for the free protein and its complex with its ligand. If the ligand binds to the target protein, the amide protons at the binding site will be protected in the complex and exchange slowly. Thus the exchanging regions are compared, and the exchange rate can be measured.

For the optimization of a lead compound through a linking process, it is necessary that two ligands should bind within proximity to each other. Generally, ligands easily attach to the primary binding site of the protein. However, second-site binders have much weaker affinity than the first-site binders for the target, and therefore their detection becomes difficult. There are a few NMR-based methods, discussed below, that can robustly detect these weak interactions, (ii) Ligand resonance-based approach includes:

SLAPSTIC with first-site spin-labeled ligand method utilizes paramagnetic relaxation enhancement (PRE) effect and is very useful in identifying the second-site binders through paramagnetic spin labels on the ligand. In this method, a first-site ligand is labeled with a paramagnetic agent, e.g., TEMPO, and is used as a probe for the screening of the second-site binders. Ligands that bind to the second site are close to the first site ligand; hence line broadening is observed in NMR signals due to paramagnetic relaxation enhancement. This method is robust because PRE effect is detectable only if both the ligands bind to their respective sites, simultaneously or close to each other, which makes the screening more reliable [35].

Another method to identify second-site binders is based on inter-ligand NOE (ILOE) [36]. In this method, at a high concentration of first-site binders, screening is done for the second-site binders, and NOESY-type experiments detect intermolecular ligand-ligand NOEs. ILOEs are observed only if first- and second-site binders are close (~5 Å). However, this method has a disadvantage, in that weak ILOEs are observed since second-site ligands bind with low affinity, and first-site binders often have poor solubilities that prevent the saturation of the binding site. High concentrations of second-site ligands and NOESY with longer mixing time (200-600 ms) are recommended.

SAR by ILOEs detects the interaction between two ligands binding to the same protein but to adjacent sites [37]. It identifies first weak hits that are improved by chemical modification approaches into bi-dentate ligands with higher affinity, using information from the protein-mediated ligand-ligand NOEs (ILOEs) in complex mixtures. By this method, information on both the ligands that bind together and on the parts of the ligands that bind proximal to one another can be obtained from a single NOESY experiment. However, this experiment requires a longer mixing time (300-800 ms) to maximize ILOE detection.

Pharmacophore by ILOEs also detects protein-mediated ligand-ligand interaction and uses information for the pharmacophore-based search of possible linked molecules from large databases of commercially available compounds [38]. This pharmacophore-based search directed by experimental binding data of weakly interacting ligands could be of high efficacy in identifying or synthesizing the high-affinity, selective ligands.

INter-ligand NOEs for PHARmacophore MApping (INPHARMA) detects protein-mediated ligand-ligand interaction based on the competition between the two ligands for the same binding site [39]. In this method, inter-ligand NOEs are observed between two competitive ligands A and B that bind to the same target. Such inter-ligand NOE effects are mediated by spin diffusion from the first-site ligand through protein protons and back to ligands that compete with the first-site ligand for the same binding site. This information could be used to determine the relative orientation of competitive ligands in the binding pocket of the target protein.

In addition to the above-mentioned NMR screening methods, there are some additional methods, which apply slightly different approaches for lead optimization. These include

SHAPES Screening: In recent years, this screening method has been prevalent in pharmaceutical research. This strategy combines NMR screening of a library of small drug-like molecules with various complementary techniques, such as virtual screening, high throughput enzymatic assays, combinatorial chemistry, X-ray crystallography, and molecular modeling in searching for new leads. In this approach, binding of small but diverse drug-like molecules to the target is assessed to find a lead. The scaffold for small molecules is derived mainly from shapes or frameworks that are commonly found in successful drugs [40].

NMR Structurally Orientated Library Valency Engineering (NMR SOLVE) method involves a fragment-linking approach to identify ligands for enzyme families [41]. It is based on selective isotope labeling on specific amino acids of the protein to observe and assign only a few critical protons in a binding site. Therefore, this approach makes easier the derivation of the structural information of the protein-ligand complex even in the absence of a complete assignment. This technique can be applied to a large family of proteins having a common binding site, adjacent to the variable binding site, which is conserved throughout the family. A representative member of the protein family is selected, and selective isotope labeling is performed on specific amino acids present in the common binding site of that protein. The binding site of this protein is mapped with a common reference ligand to obtain resonance assignments for labeled residues in the binding site, especially those at the interface between the common ligand-binding site and the substrate site. A linker is designed based on the orientation of a standard reference ligand mimic in the binding site of the protein. This linker is directed into the adjacent substrate-binding site, and an object-oriented bi-ligand library is constructed. This resulting library is suitable for use on all members of the enzyme family. The technique is useful in the synthesis of a combinatorial bi-ligand library, which can be screened to identify specific high-affinity bi-ligand inhibitors.

NMR spectroscopy has evolved as a powerful tool in drug discovery and development. It has been widely applied for identifying and validating hits and in the process of lead optimization. NMR binding experiments are very efficient and versatile in discovering high-affinity ligands for biologically relevant macromolecules, elucidating ligand-binding sites and identifying small molecules with wide ranges of binding affinity. Since NMR provides structural information at atomic-level resolution, its interplay with other screening methods, such as HTS, mass spectroscopy, X-ray crystallography and in silico screening is quite advantageous in the rapid advancement of the hit-to-lead optimization process. Another potential application of NMR is as a tool to monitor protein-ligand interactions within living cells, i.e., ‘in vivo’ NMR screening, in drug discovery [42, 43]. With the fast-paced progression in NMR methodology development, sensitivity and resolution, NMR-based screening will lead to higher throughput and can be applied to even larger protein targets. NMR will be a robust tool for designing novel therapeutics that target the challenging "undruggable" protein targets. Further progress in innovative NMR screening methods can only add to an arsenal of tools for drug discovery and development.