Knowledge of biomolecular interactions is of importance to our understanding of biological processes such as enzyme catalysis and inhibition. Biophysical techniques enable sensitive detection and accurate characterization of binding and are therefore powerful tools in enzymology and rational drug design. The applications of NMR spectroscopy and isothermal titration calorimetry to study enzyme–ligand interactions will be discussed. Recent work on ketopantoate reductase, which catalyses an important step on the biosynthetic pathway to vitamin B5, is used to illustrate the potential of this approach.
- drug design
- enzyme–ligand interaction
- ketopantoate reductase
- vitamin biosynthesis
Many enzymes bind non-protein groups known as cofactors and require them to carry out catalysis. Most of the organic cofactors, which are called coenzymes, are vitamins or their derivatives. Vitamins are synthesized in vivo by bacteria, fungi and plants and are essential for their survival. Consequently, enzymes involved in the biosynthetic pathways to vitamins are potential targets of novel antibiotics and herbicides. And because these enzymes are absent from humans, such drugs are more likely to be selective and to have minimal side effects .
Enzyme inhibitors are useful chemical tools to study biological systems and are commonly identified by random screening or designed from a knowledge of the enzyme structure and of the reaction mechanism. In the last few years, there has been an emergence of new approaches to drug discovery based on the identification of molecular fragments binding weakly to enzymes and their subsequent rational development into more potent inhibitors driven by structural information . Biophysical methods that directly monitor protein–ligand interactions are used to study binding of small ligands such as coenzymes, substrates and inhibitors to enzymes and are often more suitable than conventional bioassays to screen and detect fragments . The applications of X-ray crystallography  and NMR spectroscopy  are now well established in fragment-based drug discovery to provide structural information on potential protein-binding sites for small molecules. Other biophysical techniques such as MS, ITC (isothermal titration calorimetry) and surface plasmon resonance are also becoming increasingly popular to characterize biomolecular interactions in proteomics  and drug discovery . Although these methods can be slow, material-intensive and therefore are not generally suitable for high-throughput ligand screening, they enable the direct detection of binding in solution, which minimizes any interference from non-specific associations, and allow access to weak affinities and detailed characterization of the thermodynamics and kinetics of the interaction. Biophysical methods are therefore powerful tools in enzymology and are now being developed as alternative lead discovery tools.
Model system studies: KPR (ketopantoate reductase)
Pantothenate (vitamin B5) is the precursor to CoA and the phosphopantetheine moiety of acyl-carrier proteins. It is an essential nutrient in animals and the biosynthetic pathway is limited to bacteria and plants . Four enzymes are responsible for pantothenate biosynthesis – KPHMT (ketopantoate hydroxymethyltransferase; EC 22.214.171.124) encoded by the panB gene, KPR (EC 126.96.36.199) encoded by the panE gene, ADC (aspartate decarboxylase; EC 188.8.131.52) encoded by the panD gene and finally PS (pantothenate synthetase; EC 184.108.40.206) encoded by the panC gene. The three-dimensional structures of all Escherichia coli enzymes have been recently solved by X-ray crystallography .
We are currently developing novel approaches to inhibitor design using a range of biophysical methods and the pantothenate enzymes as targets. Three enzymes in the pathway – KPHMT, KPR and PS – require coenzymes to catalyse their reactions – Me-THF (methylene tetrahydrofolate), NADPH and ATP respectively. Coenzyme-utilizing enzymes and, more generally, multisubstrate systems are particularly suitable model systems for biophysical methods and to apply fragment-based approaches. Coenzyme and substrate binding occur at separate pockets of the enzyme active sites to form the intermolecular complexes and the interactions can be studied in the absence of catalytic turnover under equilibrium conditions. Drugs can be designed to bind either the more conserved coenzyme-binding site, the more variable substrate-binding site or both. This flexibility can be reflected in the mode of action of the drug, allowing broad-range targeting of subfamilies of targets as well as more specific action on single members within the family.
KPR is used here as a model enzyme to illustrate the potential of biophysical approaches. KPR catalyses the NADPH-dependent reduction of ketopantoate to pantoate (Scheme 1).
The E. coli enzyme has been characterized both biochemically  and structurally . We have recently solved the structure of the binary complex between KPR and NADP+ to 2.1 Å (1 Å=0.1 nm) resolution and proposed a detailed catalytic mechanism of the enzyme . KPR was chosen as a model system because (i) it is monomeric, stable over a long time and highly soluble, (ii) its activity is conveniently monitored by following the decrease in absorbance at 340 nm due to the conversion of NADPH to NADP+, and (iii) it proceeds through the formation of a ternary complex, hence allowing coenzyme and substrate binding to be investigated separately. The enzyme was expressed as a His6-tag construct and purified in high yield in a single chromatographic step . The applications of NMR spectroscopy and ITC to monitor enzyme–ligand interactions are described here.
Biophysical methods I: NMR spectroscopy
A variety of NMR methods are available to investigate protein–ligand interactions and their applications to drug discovery have been recently reviewed [13,14]. They can be divided into ligand-based and protein-based methods, depending on which signals are monitored. Fesik and co-workers  pioneered protein-based methods in which NMR spectroscopy is used to screen libraries of compounds. Perturbations on the two-dimensional spectrum of a protein are used to indicate that ligand binding is taking place. Although this approach allows information to be obtained on the location of the binding site, large quantities of protein are required, isotope labelling (typically 15N but in some cases also 13C) is mandatory and two-dimensional protein NMR spectra must be recorded and the peaks assigned.
Ligand-based NMR techniques are generally faster, require a small amount of protein and enable direct identification of binders even in mixtures using one-dimensional NMR spectra. CPMG (Carr–Purcell–Meiboom–Gill) relaxation-edited  and WaterLOGSY (water–ligand observed via gradient spectroscopy)  are rapid and powerful one-dimensional 1H NMR binding experiments. An NMR spectrum of the ligand or the ligand-mixture is recorded in the presence of the protein and compared with the one recorded in its absence under identical experimental conditions. The ligands are present at a total concentration typically in the mM range, in excess with respect to the protein. If a ligand binds to the protein receptor, chemical exchange (generally fast on the NMR timescale) between bound and free ligand occurs. This allows the properties of the bound ligand to be ‘transmitted’ to the free ligand and hence to be detected in the NMR spectrum. In CPMG experiments, a relaxation delay of 100–400 ms is employed which causes disappearance of the NMR signals of the ligands that bind to the protein due to a decrease in the transverse relaxation time (T2). In WaterLOGSY experiments, the signals recorded are generated by transfer of magnetization from water molecules to the ligand. When the ligand is free in solution (non-binders) the effect due to bulk water gives rise to negative signals, whereas when the ligand is bound to the protein (binders) the effect of water molecules present at the protein-binding site is opposite to that of bulk water and gives rise to a positive contribution to the signal intensity.
Biophysical methods II: ITC
The strength of ITC lies in its ability to directly measure the heat associated with a chemical process . The ITC instrument consists of two identical cells made of highly conductive material placed inside an adiabatic jacket and a small temperature offset is maintained between them . A solution of the ligand contained in a syringe is titrated into a solution of the protein contained in the sample cell. Changes in the sample cell temperature due to the heat associated with an injection from the syringe are detected by sensitive circuits and the time-dependent power (μcal/s; 1 cal=4.184 J) required to restore the equilibrium temperature offset between the two cells is recorded. The peaks obtained are integrated over time to yield a plot of molar heat (kcal/mol) as a function of the number of injections. In the first few injections, most of the ligand will bind to the protein, allowing enthalpy ΔH of the binding to be measured. As the experiment proceeds and the macromolecule saturates with the ligand, the signal diminishes, allowing the estimation of the binding constant and of the stoichiometry. At the end of the titration, full saturation is achieved and hence the background heat of dilution is observed.
The shape of the resulting titration curve is dictated by the relative ratio between the protein concentration and the dissociation constant Kd of the protein–ligand complex. Therefore careful consideration about experimental conditions is required to obtain reliable thermodynamic parameters, depending on the binding affinity and on the available protein concentration. Values of Kd in the range of 10−8 M<Kd<10−3 M are directly accessible by ITC, although indirect methods that allow this range to be extended for both weak (Kd>10−3 M) and tight (Kd<10−9 M) affinities have been reported . More recently, Turnbull and Daranas  reported a theoretical treatment and practical considerations about the feasibility of using ITC to study weak affinity systems. When correctly designed, one ITC experiment provides values of both the dissociation constant (Kd) and the enthalpy of binding (ΔH). The entropy of binding (ΔS) is then directly calculated using the equations ΔG=RT ln(Kd) and ΔG=ΔH–TΔS. Simultaneous measurements of the free energy ΔG, the enthalpy ΔH and the entropy ΔS make ITC the only technique currently available that allows full characterization of the thermodynamics of the interaction in a single experiment.
The implications of ITC for drug design have been reviewed by Ward and Holdgate . Potential high-throughput applications of biocalorimetry in drug discovery have also been discussed recently . Amongst the limitations of ITC are the requirement for high quantities of materials and the long times necessary to run experiments, which have prevented its spread and implementation along the drug-discovery process. This has led to the recent development of new automated instrumentation that includes an integrated autosampler capable of running up to 288 samples unattended (http://www.microcal.com).
To validate the approach, binding of NADPH and NADP+ was monitored using NMR spectroscopy experiments and quantified by ITC. Coenzymes are ideal for testing the techniques as they are large molecules, hence likely to bind with high affinities, are highly charged, hence likely to have large and negative ΔH of binding, and are also highly soluble in aqueous solutions. Binding of NADPH and NADP+ was detected by NMR spectroscopy. NMR spectra of NADP+ binding to KPR by both CPMG (Figure 1, left) and WaterLOGSY (Figure 1, right) experiments are shown. The intensities of the NMR signals of the ligand in the absence of enzyme (Figure 1a) are changed upon addition of the enzyme (Figure 1b).
In CPMG, decreases in signal intensities are observed, consistent with the ligand binding to the protein. In WaterLOGSY, positive ligand signals in the presence of KPR are indicative of binding. It is clear from Figure 1 that binders (positive signals) are unambiguously distinguished from non-binders (negative signals) by WaterLOGSY. This allows direct detection of binding without the requirement of control experiments, making WaterLOGSY suitable for screening mixtures of ligands. In contrast, the changes in the intensities of ligand signals in CPMG can be subtle, as shown by the peaks at δ 8.1 and 8.4 ppm. Reference spectra in the absence of the protein are hence needed as controls for CPMG experiments. The thermodynamics of coenzyme binding to KPR was characterized by ITC. Dissociation constants of 6 and 0.3 μM were found for NADP+ and NADPH respectively .
Binding specificity by competition experiments
Binding at specific sites on the protein can be proved by performing competition experiments with coenzymes, substrates or inhibitors known to interact with high affinity. Two different approaches using NMR spectroscopy and ITC are illustrated here, which involve the displacement of the ligand of interest by a high-affinity competitor. ADP was used as a test ligand as it is a fragment of NADPH and hence is likely to bind at the coenzyme-binding site with weak affinity. First, CPMG and WaterLOGSY experiments were conducted to detect the binding of ADP to KPR using the same conditions as described above for NADP+. No changes in the NMR signals of ADP were seen (results not shown). The NMR experiments were repeated by increasing both ligand and protein concentrations (Figures 2a and 2b). NADPH was subsequently added to displace the ligand from the active site (Figure 2c).
Binding of ADP to KPR was detected and successful displacement by NADPH proved specific interaction at the coenzyme-binding site. The weak affinity of ADP for KPR was later quantified by ITC and a Kd of 0.4 mM was measured, corresponding to an approx. 1000-fold increase relative to NADPH (A. Ciulli, C.M.C. Lobley, K.L. Tuck, G. Williams, A.G. Smith, T.L. Blundell and C. Abell, unpublished work). WaterLOGSY proved to be a more sensitive method for identifying weak interactions than CPMG, as exemplified by Figure 2. It is difficult to identify changes in the CPMG spectra of ADP, whereas binding of ADP and displacement by NADPH are clear from the WaterLOGSY spectra.
After the NMR analysis, competition ITC was also used to confirm interaction of ADP at the coenzyme-binding site of KPR. Two ITC titrations of KPR with NADPH were performed in the absence (Figure 3a) and in the presence (Figure 3b) of ADP in the cell under otherwise identical conditions. The thermodynamic parameters for NADPH binding were affected by the presence of ADP, consistent with the displacement of a weakly bound ligand from the enzyme active site during the titration (Figure 3).
Analysis of the data using the competition model described by Sigurskjold  provided a Kd of 440 μM and a ΔH of –2.7 kcal/mol for ADP. The thermodynamic parameters obtained from competition ITC were in good agreement with those obtained from direct ITC of KPR with ADP (A. Ciulli, C.M.C. Lobley, K.L. Tuck, G. Williams, A.G. Smith, T.L. Blundell and C. Abell, unpublished work).
Biophysical methods enable sensitive detection and accurate characterization of specific enzyme–ligand interactions and are therefore powerful tools in enzymology and for the design of inhibitors. Despite their intrinsic low throughput, biophysical techniques are now gaining momentum as alternative lead discovery tools, particularly in fragment-based approaches.
Coenzymology: the biochemistry of vitamin biogenesis and cofactor-containing enzymes: Independent Meeting held at King's College, Cambridge, U.K., 4–7 April 2005. Organized and Edited by A.G. Smith (Cambridge, U.K.) and A.W. Munro (Leicester, U.K.).
Abbreviations: CPMG, Carr–Purcell–Meiboom–Gill; ITC, isothermal titration calorimetry; KPR, ketopantoate reductase; WaterLOGSY, water–ligand observed via gradient spectroscopy
- © 2005 The Biochemical Society