Methods for identifying ligands using competitive binding 1H NMR experiments

Abstract

A method of identifying a ligand to a target molecule that includes: providing a reference compound that interacts with the target molecule; collecting a 1 D 1H nuclear magnetic resonance spectrum of the reference compound in the presence of the target molecule; providing a plurality of test samples, each test sample including at least one test compound; collecting a 1 D 1H nuclear magnetic resonance spectrum of the reference compound in the presence of each test sample and the target molecule; comparing the spectrum of the reference compound in the presence of the target molecule to the spectrum of the reference compound in the presence of each test sample and the target molecule to determine a change in one or more of the reference compound resonances; and identifying at least one test compound that interacts with the target molecule, wherein the test compound displaces the reference compound.

Description

BACKGROUND OF THE INVENTION

[0002] Many drugs currently on the market were developed from leads identified from high throughput screening (HTS). Targets of therapeutic interest used in HTS are often recombinant proteins produced from cloned genes which can be expressed in different ways. A large compound collection is typically screened against these proteins for the identification of inhibitors.

[0003] During the last ten years the size of the proprietary compound collection has increased exponentially as a result of systematic application of combinatorial chemistry to different projects. Combinatorial chemistry nowadays generates large compound libraries that complement other compound libraries available from traditional medicinal chemistry and natural sources. The development and application of robotics and automation have made it feasible to test large numbers of compounds in a short period of time. Several new detection systems are used for the identification of potential lead molecules.

[0004] Recently, nuclear magnetic resonance (NMR) has emerged as a powerful method for the detection of small molecules that interact with targets of pharmaceutical interest. Although NMR is not a sensitive technique, it has the advantage that it is less subject to artifacts observed with other systems of detection. Recent developments in cryogenic NMR probe technology have reduced the period of time or the amount of protein necessary for the screening.

[0005] NMR methods have been used for screening a large compound collection against isotopically labeled proteins. Chemical shift changes of cross peaks in a 15N—1H HSQC spectrum of the target protein are monitored in the presence of a compound mixture. Deconvolution of the mixture then results in the identification of the molecule interacting with the protein (i.e., the compound responsible for the chemical shift changes). When the three dimensional structure of the protein is known and the sequence specific NMR assignments of the protein backbone resonances have been obtained, the method provides important structural information of the ligand binding site and ligand binding mode.

[0006] Another method for performing NMR screening is based on the detection of the ligand resonances. Several NMR parameters have been proposed in the literature as a tool for ligand identification. These methodologies permit rapid deconvolution of the screened mixtures and are particularly suited for the identification of medium to low affinity ligands.

[0007] However, these techniques suffer from some drawbacks. First, no structural information regarding the binding site is directly available. Second, high affinity ligands and molecules that bind covalently to the receptor escape detection because of the large excess of the test compound over protein typically used in these experiments. That is, compounds interacting tighter to the protein or compounds that have a slow on-rate will not be detected because the residence time of these compounds within the protein is longer than the window of the mixing time (e.g., 1 to 2 seconds) employed in the NMR experiments. Third, compounds with poor solubilities that are potential ligands are difficult to detect since the method requires the observation of the ligand signals.

[0008] Thus, what is needed are additional NMR methods that can be used to detect ligands to target molecules, such as proteins, without the drawbacks associated with typical ligand-observed screening experiments.

SUMMARY OF THE INVENTION

[0009] The present invention is related to rational drug design. Specifically, the present invention provides a nuclear magnetic resonance (NMR) method of screening for compounds that interact with a target molecule (e.g., typically a protein). The method involves the use of 1H NMR competition binding experiments to detect the binding interaction.

[0010] Competition binding experiments involve the displacement of a reference compound in the presence of a competitive molecule. Preferably, the reference compound binds to the target molecule with a binding affinity in the micromolar range. Preferably, the test compound interacts with the target molecule with a binding affinity stronger than 1 micromolar (e.g., in the nanomolar range), although compounds binding with a binding affinities of weaker than (i.e., more than) 1 micromolar can also be evaluated using the methods of the present invention.

[0011] The present methodology involving competition binding experiments can be used for performing efficient high throughput screening (HTS) based on properly set-up competition binding experiments without the drawbacks associated with typical ligand-observed screening experiments. In addition, the methods provide an estimation of the KD of the identified ligand using a single point measurement. With this approach it is possible to screen thousands of compounds in a short period of time against protein or DNA and RNA fragments, for example.

[0012] The present invention could also find useful applications for rapid screening of chemical mixtures (i.e., mixtures of two or more test compounds) such as plant and fungi extracts. Rapid screening techniques typically involve providing a plurality of test samples, each test sample comprising a mixture of two or more test compounds.

[0013] Methods of the present invention involve identifying a ligand to a target molecule using at least the following steps: providing a reference compound that interacts with the target molecule; collecting a 1 D 1H nuclear magnetic resonance spectrum of the reference compound in the presence of the target molecule; providing a test sample (preferably a plurality of test samples) comprising at least one test compound; collecting a 1 D 1H nuclear magnetic resonance spectrum of the reference compound in the presence of each test sample and the target molecule; comparing the spectrum of the reference compound in the presence of the target molecule to the spectrum of the reference compound in the presence of each test sample and the target molecule to determine a change in one or more of the reference compound resonances; and identifying at least one test compound that interacts with the target molecule, wherein the test compound displaces the reference compound.

[0014] Preferably, methods of the present invention include a step of identifying the reference compound comprising: collecting a WaterLOGSY nuclear magnetic resonance spectrum of a potential reference compound in the absence of the target molecule; collecting a WaterLOGSY nuclear magnetic resonance spectrum of the potential reference compound in the presence of the target molecule; and comparing the WaterLOGSY spectra to identify whether the potential reference compound interacts with the target molecule.

[0015] For certain embodiments of the methods of the present invention, collecting a 1 D 1H nuclear magnetic resonance spectrum includes collecting a 1 D 1H selective or multi-selective spectrum.

[0016] For certain embodiments of the methods of the present invention, collecting a 1 D 1H nuclear magnetic resonance spectrum includes collecting a selective T2 weighted Total Correlated Spectroscopy (TOCSY) spectrum, a multi-selective T2 weighted TOCSY spectrum, a selective T2 weighted Correlation Spectroscopy (COSY) spectrum, or a multi-selective T2 weighted COSY spectrum. Herein, the COSY experiments can involve either anti-phase or in-phase COSY experiments, as demonstrated in the Examples Section.

[0017] For certain embodiments of the methods of the present invention, the reference compound is provided in combination with a non-interacting compound. For these methods, collecting a 1 D 1H nuclear magnetic resonance spectrum of the reference compound in the presence of the target molecule includes collecting a spectrum of the reference compound and the non-interacting compound in the presence of the target molecule; and collecting a 1 D 1H nuclear magnetic resonance spectrum of the reference compound in the presence of each test sample and the target molecule includes collecting a spectrum of the reference compound and the non-interacting compound in the presence of each test sample and the target molecule.

BRIEF DESCRIPTION OF THE DRAWINGS

[0018] FIG. 1. Expanded region of the 1 D 1H spectra recorded for 40 μM Compound A in the presence (a) and in the absence (b) of the kinase (2 μM in PBS). The data were multiplied with a cosine window function prior to Fourier transformation. An arrow indicates the resonance undergoing significant broadening in the presence of the protein.

[0019] FIG. 2. ITC binding experiments on the binding of Compound A to the kinase. Upper panel: Injection heat effects measured during a titration of 100 μM of the protein into buffer (A) as well as into 8 μM Compound A dissolved in the same buffer (B). 50 mM Tris/Cl pH 7.0, 100 mM NaCl, 1 mM DTT was used as a buffer in this experiment. Lower panel: Integrated and normalized binding heats derived from the raw data trace shown in the upper panel. Binding enthalpies measured on Compound A are shown as filled squares. The corresponding fitted function is indicated as a dotted line. Dilution heat effects determined from the blank titration are shown as open circles. The binding thermodynamics are KB=1.4±0.4 105 M−1, ΔHobs=−3.3±0.8 Kcal/mol, TΔS=3.6 Kcal/mol, N=0.8±0.2.

[0020] FIG. 3. Expanded region of the 1 D 1H spectra recorded at different Compound A concentration in the presence of the kinase (1.5 μM in PBS). The spectra were acquired with 128 scans and 2.82 s repetition time. The ligand concentration was 40 μM (a), 80 μM (b) and 140 μM (c).

[0021] FIG. 4. Plot of the signal intensity ratio of two ligand resonances I7.13/I6.65 (a) and I7.13/I7.53 (b) of Compound A as a function of the ratio [EL]/[LTOT]. The resonance at 7.13 ppm is the singlet (1 proton) that undergoes significant exchange broadening in the presence of the protein, the resonance at 7.53 ppm is a triplet (2 protons) and the resonance at 6.65 is a doublet (1 proton) with a small J. The ratio [EL]/[LTOT] was calculated using the ITC derived KD of 7.1 μM for Compound A. The first point on the left corresponds to the value in the absence of the protein. The concentration of the protein kinase was 1.5 μM in PBS. The curves represent the best fit of the experimental points.

[0022] FIG. 5. NMR screening and deconvolution performed with 50 μM Compound A in the presence of the kinase (2 μM in PBS). Expanded region of the 1 D spectra recorded in the absence (a) and in the presence (b) of a 20 μM seven compound mixture containing the molecules SPECS AB-323/25048456 (supplied by SPECS, Rijswijk, the Netherlands) ethyl 2-quinoxalinecarboxylate, methyl isoquinoline-3-carboxylate, 7-phenyl-4-pteridinol, 2-amino-6-methylquinazolin-4-ol, 5-methylbenzimidazole and Compound B, (c) spectrum recorded in the presence of the chemical mixture without Compound B, (d) spectrum recorded in the presence of only Compound B. The spectra were acquired with 128 scans and 2.82 s repetition time. An arrow indicates the resonance undergoing significant broadening in the presence of the protein.

[0023] FIG. 6. Selective R1 spectra as a function of τ acquired with selective inversion of the C2-H resonance of Trp (100 μM) in the absence (a) and in the presence (b) of HSA (8 μM in PBS). Only the spectral region containing the C2-H resonance is displayed. The spectra were acquired with 8 (a) and 32 (b) scans. The repetition delay was 10.82 s (a) and 8.82 s (b). An array of 16 (a) and 12 (b) τ values was used starting from 0.005 s (left) and with an increment of 0.475 s. For (a) only the first 12 Ε values are displayed for a direct comparison with (b). The longer repetition time and the larger array in (a) was necessary because of the small R1, s for Trp in the absence of the protein.

[0024] FIG. 7. (Top) Plot of the signal intensity ratio of the inverted Trp C2-H resonance in two R1, s filtered experiments recorded with τ=0.48 s and τ=1.91 s as a function of the ratio [EL]/[LTOT]. The magnetization is negative with the first τ value and positive with the second τ value. (Bottom) Plot of the signal intensity ratio for the Trp C2-H and C4-H resonances in a R1, s filtered experiment (inversion of the C2-H resonance) recorded with Ε=1.91 s as a function of the ratio [EL]/[LTOT]. The L-Trp concentration was kept constant at 100 μM and the protein (HSA) concentration was varied from 0 μM to 10 μM. The ratio [EL]/[LTOT] was calculated using the equilibrium dialysis-derived KD value of 23 μM for Trp measured at 20° C. (U. Kragh-Hansen, Biochem. J, 273, 641-644 (1991)). The first point on the left corresponds to the value in the absence of protein. The curves represent the best fits of the experimental points.

[0025] FIG. 8. Plot of the signal intensity I(τ) as a function of the τ period in a R1, s experiment. Decays with R1 of 0.4 s−1 and 1 s−1 were used in the simulation. The signal intensity difference between the two decays is shown with a dashed line. The two rectangles correspond to the time regions that can be selected for the R1,s filtered experiments.

[0026] FIG. 9. NMR screening and deconvolution performed with a 1 D R1, s filtered experiment for 100 μM Trp in the presence of the protein HSA (8 μM in PBS). Expanded region containing only the selectively inverted Trp C2-H resonance recorded in the absence (a) and in the presence (b) of a 30 μM four compound mixture containing sucrose, D-5-CH3 Trp, L-5-CH3 Trp and diazepam, (c) spectrum recorded in the presence of the chemical mixture without diazepam, (d) spectrum recorded in the presence of only diazepam, (e) spectrum recorded in the absence of HSA and the mixture. The spectra were acquired with 128 scans and 5.82 s repetition time. The τ value was 0.955 s, corresponding closely to the null point for the spectrum in (a). Only a very weak positive signal is observed in (a) and (c).

[0027] FIG. 10. Expanded region of the spectra recorded with two R1,s filtered experiments (Trp C2-H selective inversion) acquired with τ=0.96 s (left) and τ=1.91 s (right). Trp and HSA concentrations were 100 and 8 μM, respectively. (a) Spectra for only Trp, (b) spectra for Trp in the presence of 30 μM Compound C, (c) spectra for Trp in the presence of 30 μM Compound D. A total of 48 scans were recorded with a repetition time of 8.8 s. The τ value of 0.96 s corresponds closely to the null point in the spectrum (a) left. The asterisks indicate the Trp C2-H resonance.

[0028] FIG. 11. Signal intensity of the Trp C2-H resonance in R1,s filtered spectra recorded with selective inversion of the Trp C2-H resonance and τ=0.96 s. Trp and HSA concentrations were 100 and 8 μM, respectively. Only the spectral region containing the Trp C2-H resonance is displayed. The spectra were acquired with 48 scans and with a repetition time of 8.8 s. The spectrum on the left (indicated with an asterisk) is the control experiment for Trp without the competing molecules. The other seven spectra were recorded in the presence of 30 μM of each lead molecule. The spectra have been arranged according to the binding affinity of the competing molecules. The binding constant was derived using the diagrams of FIG. 10 and following the same procedure as described in the Examples Section. The number associated at each spectrum identifies the analyzed molecule.

[0029] FIG. 12. Doubly-selective R1-filtered experiments recorded as a function of the HSA concentration. The methyl group resonances of 5-CH3 Trp (2.43 ppm) and warfarin (2.21 ppm) were selectively inverted with a 25 ms long 180° Gaussian SLP. A total of 64 scans were recorded with repetition time and filter delay of 6.83 and 0.4 s, respectively. The concentration of the two molecules was kept constant at 25 μM whereas the concentration of HSA was from left to right, 0, 0.5, 1.0, 1.5 and 2.5 μM, respectively. The dashed horizontal line indicates the lack of signal intensity change for the Trp derivative signal.

[0030] FIG. 13. Titration of 3.0 μM sodium warfarin (left panel) and 4hydroxy-3-[1-(p-iodophenyl)-3-oxobutyl] coumarin (4) (right panel) with HSA in PBS at 23° C. Data points are the relative fluorescence values of the compounds versus the concentration of HSA. The solid line represents the theoretical fit to the experimental data using the nonlinear least-squares method as described.

[0031] FIG. 14. (Top) Plot of the signal intensity ratio of the two selectively inverted resonances of FIG. 13 as a function of the fraction of bound warfarin ([EL]/[LTOT]). (Bottom) Plot of the signal intensity ratio of the same two resonances extracted from spin-echo experiments recorded with a total delay of 256 ms (4×64 ms) as a function of the fraction of bound warfarin. The concentration of the two molecules was kept constant at 25 micromolar (μM) whereas the concentration of HSA was varied from 0 to 2.5 μM. The first point on the left corresponds to the value in the absence of protein. The ratio [EL]/[LTOT] was calculated using the fluorescence-derived KD value of 6.1 μM. The racemic form of warfarin was used in these studies because, as reported in the literature, the KD for the S and R forms are very similar (T. Wiseman et al., Anal. Biochem., 179, 131-137(1989)). The curves represent the best fits of the experimental points.

[0032] FIG. 15. 1H ligand-based competition binding NMR screening performed with 25 μM of (1) and (2) against HSA. (Top) Doubly-selective R1-filtered experiments recorded with a filter delay of 0.4 s. The methyl group resonances of (1) and (2) were selectively inverted with a 25 millisecond (ms) long 180° Gaussian SLP. A total of 64 scans were recorded with repetition delay of 6.83 seconds(s). (Bottom) Spin-echo experiments recorded with a delay of 256 ms (4×64 ms). A total of 128 scans were recorded with repetition delay of 3.83 s. The spectra were acquired in the presence of 1.5 μM HSA and in the absence (left) and presence of 150 μM Tolbutamide (middle). The corresponding spectra in the absence of HSA and Tolbutamide are shown on the right. The signal intensity ratio of the two resonances is from left to right, on the top, (1.94, 1.09, 0.71) and on the bottom (3.13, 1.56, 0.86). The signal at 2.38 ppm in the spectra in the middle originates from the aromatic methyl group signal of (3).

[0033] FIG. 16. Pulse sequences for the multi-selective COSY and refocused COSY experiments. The 180° Gaussian SLP is positioned between two gradients of the same strength and sign. The length of the multi-selective pulse is typically in the range of 25 to 50 ms in duration. The τ value can be properly adjusted for different binding constants of the spy molecules. Weak affinity reference molecules necessitate a long τ period. An alternative to this selective excitation scheme is the excitation sculpting sequence (K. Stott et al., J. Am. Chem. Soc., 117, 4199 (1995)). The narrow and broad bars represent 90° and 180° pulses, respectively. The delay Δ and Δ′ are optimized for maximum COSY transfer according to the coupling constants and spin topology. N is an odd number and typically is set to 1. However for a two-spin system it can be set to larger values if the reference molecule binds with very weak affinity to the receptor. Water suppression is achieved with the double spin-echo scheme.

[0034] FIG. 17. 1H ligand-based competition binding NMR screening with one-dimensional multi-selective COSY. The protein concentration was 2 μM and the concentration of the reference molecule (3) and control molecule (2) was 50 and 25 μM, respectively. The spectra were recorded in the presence of 50 μM of the competing molecule (4). The COSY spectrum (top) was obtained using the pulse sequence of FIG. 16 with double selective excitation of the C3,5 H2 resonance (indicated by an arrow) of tolbutamide and the methyl group resonance of the Trp derivative. A total of 256 scans were recorded with 2.83 s repetition time. The double-selective pulse was a 180° Gaussian SLP of 25 ms in duration. The delay Δ was 30 ms in duration (optimum value for a scalar coupling of 8.3 Hz) and n was 1. (Bottom) Reference spectrum recorded with 128 scans and 3.83 s repetition time. The C3,5 H2 resonance of tolbutamide partially overlaps with some signals of (4).

[0035] FIG. 18. 1H ligand-based competition binding NMR screening with one-dimensional multi-selective COSY (a), refocused COSY (b) and TOCSY (c) experiments. The concentration of the reference molecule (tolbutamide) and control molecule (5-CH3 Trp) was 50 and 25 μM, respectively. The COSY and refocused COSY spectra were recorded with the pulse sequences of FIG. 16 with double selective excitation of the C3,5-H2 resonance of tolbutamide and the methyl group resonance of the Trp derivative. A total of 256 scans were recorded with 2.83 s repetition time and n set to 1. The double-selective pulse was a 180° Gaussian SLP of 25 ms in duration. The delay Δ was 30 ms (a) and 15 ms (b), and Δ was 30 ms (b). For the TOCSY experiment the same multi-selective excitation scheme was used. The homonuclear Hartmann-Hahn step was performed with a WALTZ-16 scheme (A. J. Shaka et al, J. Magn. Reson., 52, 335 (1983)) of 35 ms in duration. The expanded region contains the signal of the C3,5-H2 resonance of Tolbutamide and the methyl group of the Trp derivative. The spectra were recorded in the presence of 2 μM of HSA and in the absence (left) and presence (middle) of a mixture containing 50 μM of the competing molecule (4). The corresponding spectra in the absence of both the protein and the competing molecule are on the right. The signal of the control molecule does not change in intensity due to the lack of interactions with the protein. The intensity ratio of the two resonances is from left to right on the top 9.6, 2.8, 2.3 in the middle 9.8, 3.5, 2.9 and on the bottom 5.6, 2.6, 2.1.

[0036] FIG. 19. Structures of Compounds 1-4.

DETAILED DESCRIPTION OF ILLUSTRATIVE EMBODIMENTS OF THE INVENTION

[0037] The present invention is directed to the use of 1H NMR competition binding experiments. Competition binding experiments involve the displacement of a reference compound in the presence of a competitive molecule. Preferably, the reference compound binds to the target molecule with a binding affinity in the micromolar range. Preferably, the test compound binds to the target molecule with a binding affinity stronger than (i.e., less than) 1 micromolar (e.g., in the nanomolar range), although compounds binding with a binding affinity weaker than (i.e., more than) 1 micromolar can also be evaluated using the methods of the present invention.

[0038] Although this method is particularly useful for identifying ligands that are relatively strong binders to the target molecule, it can be used for identifying ligands of a wide range of binding affinities. The relatively strong binders are typically defined as those having a dissociation binding constant KD of less than about 1 micromolar, preferably less than about 500 nM, more preferably less than about 100 nM.

[0039] The present invention provides a variety of methods of identifying a ligand that interacts with a target molecule that have the following commonly applied steps: providing a reference compound that interacts with the target molecule; collecting a 1 D 1H nuclear magnetic resonance spectrum of the reference compound in the presence of the target molecule; providing a test sample (preferably a plurality of test samples), each test sample comprising at least one test compound; collecting a 1 D 1H nuclear magnetic resonance spectrum of the reference compound in the presence of each test sample and the target molecule; comparing the spectrum of the reference compound in the presence of the target molecule to the spectrum of the reference compound in the presence of each test sample and the target molecule to determine a change in one or more of the reference compound resonances; and identifying at least one test compound that interacts with the target molecule, wherein the test compound displaces the reference compound (typically, this results because the test compound has a binding affinity at least as tight as that of the reference compound).

[0040] Typically, a change in one or more of the reference compound resonances involves an increase in signal intensity in at least one reference resonance. Preferably, a change in one or more of the reference compound resonances involves a sharpening of at least one reference resonance.

[0041] The optimum experimental conditions for any of the methods described herein can be determined as described in the Examples Section. Specifically, this typically involves the following steps being carried out prior to collecting a 1 D 1H nuclear magnetic resonance spectrum of the reference compound in the presence of the target molecule for use in the comparing step: collecting 1 D 1H nuclear magnetic resonance spectra of the reference compound in the presence of the target molecule at different concentrations of the target molecule or at different concentrations of the reference compound. The information collected is used to determine the optimum experimental conditions for identifying at least one test compound that interacts with the target molecule as described in the Examples Section.

[0042] A wide variety of pulse sequences can be used when collecting the 1 D 1H NMR spectrum of the reference compound in the presence of each test sample and the target molecule. For effective comparison of spectra, it is desirable to have the same experimental conditions; however, target compound and reference molecule concentrations can be varied as long as the graphs with the titration experiments have been generated before the screening. Generally, the temperature and buffer conditions are the same, because a change in these experimental conditions can affect the binding constant of the reference compound.

[0043] In the generalized method described above for mixtures of two or more test compounds, identifying at least one test compound may preferably involve recording separate 1 D 1H nuclear magnetic resonance spectra of the reference compound in the presence of each test compound and the target molecule. This is followed by comparing the spectrum of the reference compound in the presence of the target molecule to the spectrum of the reference compound in the presence of each test compound and the target molecule to determine a change in the selected reference compound resonance. The pulse sequences of these experiments are generally the same. Such experiments are typically referred to by those of skill in the art as deconvolution experiments.

[0044] The dissociation constant (i.e., binding affinity) of a test compound and/or a reference compound can be determined using NMR techniques if desired, although other well-known techniques can be used as well (e.g., isothermal titration calorimetry). Preferably, the reference compound binding affinity is evaluated using isothermal titration calorimetry or fluorescence spectroscopy, the specific details of which are well-known to one of skill in the art and are described in the Examples Section.

[0045] For example, in one NMR-based method, in addition to the above-listed steps in the generalized method, 1 D 1H nuclear magnetic resonance spectra of the reference compound in the presence of the target molecule at different concentrations of the reference compound can be collected. Alternatively or additionally, 1 D 1H nuclear magnetic resonance spectra of the reference compound in the presence of the target molecule at different concentrations of the target molecule can be collected. This information can be used to determine the dissociation constant of the test compound as described in the examples.

[0046] For certain embodiments of the methods of the present invention, collecting a 1 D 1H nuclear magnetic resonance spectrum includes collecting a 1 D 1H selective or multi-selective spectrum. “Selective” refers only to one resonance (or frequency), “multi-selective” refers to more than one resonance (or frequencies). Exemplary pulse sequences for selective and multi-selective longitudinal relaxation experiments are provided in the Examples Section.

[0047] In a multi-selective experiment, two compounds are used: a control molecule that does not bind to the receptor of interest and a compound that interacts with a weak to medium binding affinity to the receptor. The latter is referred to as the reference or ‘spy’ molecule. In the simplest case, control and spy molecules are chosen that have singlet resonances in isolated spectral regions where spectral overlap is unlikely. In the specific case of HSA, many drugs bind to one of two primary binding sites: Sudlow site I (located in subdomain IIA) and Sudlow site II (located in subdomain IIIA). Sudlow site I accommodates bulky heterocyclic anions with a centralized charge, such as bilirubin, warfarin, and cyclic eicosanoids. Sudlow site II binds to hydrophobic aromatic moieties like those present in diazepam, ibuprofen, and L-tryptophan (T. Peters, Jr. in All about Albumin Biochemistry, Genetics, and Medical Applications, Academic Press, San Diego, U.S.A., pages 109-114 (1996)). Recently the x-ray structure of warfarin bound to recombinant HSA has been solved at high resolution. Warfarin (1) is a good spy molecule to monitor binding at Sudlow site I because of its excellent solubility in acqueous buffers and because it contains a methyl group. The methyl group NMR resonance is a sharp singlet signal at 2.21 ppm. Likewise, 5-CH3 D,L Trp (2) was chosen as the control molecule because it also contains a methyl group that gives rise to a sharp singlet resonance at 2.43 ppm. Previous ITC and WaterLOGSY experiments have clearly demonstrated that up to concentration of several hundreds μM this Trp derivative does not interact with HSA. Selection of molecules containing methyl groups allows for a reduction of the measuring time due to the high intensity of the methyl group signals. In addition, this permits one to lower the concentration of the reference and control molecules thus avoiding problems arising from aggregation, and non-specific binding.

[0048] For certain embodiments of the methods of the present invention, collecting a 1 D 1H nuclear magnetic resonance spectrum includes collecting a selective T2 weighted TOCSY spectrum, a multi-selective T2 weighted TOCSY spectrum, a selective T2 weighted COSY spectrum, or a multi-selective T2 weighted COSY spectrum.

[0049] In these experiments a multiplet resonance of the reference molecule is selectively excited and its magnetization is transferred via scalar couplings to other resonances where it is then detected during the acquisition time.

[0050] It is possible that problems of overlap will be encountered even when the reference and control molecules are selected according to the criteria described above. This may occur when complex chemical mixtures are screened and/or when large NMR signals arising from buffer or detergents are present. A convenient approach for circumventing this problem is the selection of molecules according to the presence in their NMR spectra of multiplet resonances. The spectra should contain at least one doublet resonance or, in the best case, a weakly scalar coupled two spin system for improved sensitivity. If the monitored resonance overlaps with other signals of the screened mixture, it is now possible to extract its relaxation properties by relaying it through scalar coupling mechanism to another resonance. Thee relaying process can be achieved with either COSY or TOCSY coherent magnetization transfer.

[0051] For increasing the precision of any one method of the present invention, various techniques can be used. Typically, an internal control can be used, which can be a non-interacting compound or an electronically generated ERETIC signal.

[0052] An alternative to the use of a non-interacting molecule is the use of the ERETIC method (S. Akoka et al., Anal. Chem., 71, 2554-2557 (1999); and V. Silvestre et al., S. Anal. Chem., 73, 1862-1868 (2001). This technique relies on the electronic generation of a signal of a defined frequency, linewidth and amplitude. A pseudo-FID is acquired with the FID originating from the sample. The amplitude of this artificial signal is adjusted to become comparable to the intensity of the signal of the reference compound recorded in the absence of the protein. This amplitude value is then used for the titration and NMR-screening experiments and the signal intensity ratio of the real and artificial signal is measured. Adding an ERETIC signal is like adding a fake signal to normalize the signals.

[0053] For certain embodiments of the methods of the present invention, the reference compound is provided in combination with an ERETIC signal with defined linewidth, amplitude, and frequency. For these methods, collecting a 1 D 1H nuclear magnetic resonance spectrum of the reference compound in the presence of the target molecule includes collecting a spectrum of the reference compound with the ERETIC signal in the presence of the target molecule; and collecting a 1 D 1H nuclear magnetic resonance spectrum of the reference compound in the presence of a test sample and the target molecule includes collecting a spectrum of the reference compound with the ERETIC signal in the presence of a test sample and the target molecule.

[0054] For certain embodiments of the methods of the present invention, the reference compound is provided in combination with a non-interacting compound. For these methods, collecting a 1 D 1H nuclear magnetic resonance spectrum of the reference compound in the presence of the target molecule includes collecting a spectrum of the reference compound and the non-interacting compound in the presence of the target molecule; and collecting a 1 D 1H nuclear magnetic resonance spectrum of the reference compound in the presence of a test sample and the target molecule includes collecting a spectrum of the reference compound and the non-interacting compound in the presence of a test sample and the target molecule. Such non-interacting compounds act as controls in that they do not bind to the target molecule at the concentrations evaluated.

[0055] In combination with the competition binding experiments of the present invention, the WaterLOGSY method can be used to identify the reference compound, as well as other methods such as spectroscopic or biochemical assays, which are well known to one of skill in the art. Preferably, the reference compound can be identified by the following steps: collecting a WaterLOGSY nuclear magnetic resonance spectrum of a potential reference compound in the absence of the target molecule; collecting a WaterLOGSY nuclear magnetic resonance spectrum of the potential reference compound in the presence of the target molecule; and comparing the WaterLOGSY spectra to identify whether the potential reference compound interacts with the target molecule.

[0056] The WaterLOGSY method (also referred to as the Water-Ligand Observed via Gradient Spectroscopy Y) is based on the transfer of magnetization from the protons of bulk water to the protons of compounds that interact with target molecules (e.g., proteins). Using WaterLOGSY techniques, binding compounds are distinguished from nonbinders by the opposite sign of their water-ligand nuclear Overhauser effects (NOEs). The WaterLOGSY method is described in greater detail in International Publication No. WO 01/23330 (published Apr. 5, 2001), in C. Dalvit et al., J. Biomol. NMR, 18, 65-68 (2000), in Applicants' Representatives copending U.S. application Ser. No. 60/386,896, filed on Jun. 5, 2002 (Attorney Docket No. 01168.PRO1).

[0057] The target molecules that can be used in the methods of the present invention include a wide variety of molecules, particularly macromolecules, such as polypeptides (preferably, proteins), polynucleotides, organic polymers, and the like.

[0058] “Polynucleotide” as used herein refers to a polymeric form of nucleotides of any length, either ribonucleotides or deoxynucleotides, and includes both double- and single-stranded DNA and RNA. A polynucleotide may include both coding and non-coding regions, and can be obtained directly from a natural source (e.g., a microbe), or can be prepared with the aid of recombinant, enzymatic, or chemical techniques. A polynucleotide can be linear or circular in topology. A polynucleotide can be, for example, a portion of a vector, such as an expression or cloning vector, or a fragment.

[0059] “Polypeptide” as used herein refers to a polymer of amino acids and does not refer to a specific length of a polymer of amino acids. Thus, for example, the terms peptide, oligopeptide, protein, and enzyme are included within the definition of polypeptide. This term also includes post-expression modifications of the polypeptide, for example, glycosylations, acetylations, phosphorylations, and the like.

[0060] The reference compound is one that interacts with the selected target molecule with a binding affinity sufficiently low so that it gives rise to a readily observed, positive-intensity signal in the WaterLOGSY recorded in the presence of the target molecule. Preferably, a weakly interacting reference compound is used. Relatively weakly interacting reference compounds are typically defined as those having a dissociation binding constant KD in the micromolar range.

[0061] The reference compound may preferably include methyl groups, particularly when WaterLOGSY is used to identify the reference compound because methyl groups typically provide a strong WaterLOGSY signal. Such methyl groups often are less hydrated, resulting in a smaller WaterLOGSY signal for the free ligands.

[0062] The reference compound preferably displays a singlet NMR signal for the methods described herein to reduce the occurrence of overlapping signals, except for methods involving TOCSY and COSY experiments. For the latter experiments, as discussed above, reference compounds that produce multiplet NMR signals are required.

[0063] The test compounds that can be evaluated can be any of a wide variety of compounds, which potentially have a wide variety of binding affinities to the target. Significantly, the method of the present invention has the ability to detect compounds that are relatively strong binders. The relatively strong binders are typically defined as those having a dissociation binding constant KD of less than about 1 micromolar. Compounds that can be screened using the method of the present invention include, for example, plant extracts, fungi extracts, other natural products, and libraries of small organic molecules.

[0064] The present invention can screen for ligands from a library of compounds that have a broad range of solubilities (the methods are particularly amendable to compounds having very low solubilities). Significantly and advantageously, for certain embodiments, the present invention preferably involves carrying out a binding assay at relatively low concentrations of target (i.e., target molecule) and high ratios of ligand to target. Thus, preferred embodiments of the present invention allow for the detection of compounds that are only marginally soluble. Typically the marginally soluble compounds are those that have a solubility in water of no greater than about 10 μM.

[0065] Preferably, the concentration of each test compound in each sample is no greater than about 100 μM, although higher concentrations can be used if desired. However, a significant advantage of the method of the present invention is that very low ligand concentrations (e.g., no greater than about 10 μM) can be used. More preferably, the concentration of target molecule is about 100 nM to about 10 μM.

[0066] The exact concentrations and ratios of test compound to target compound used can vary depending on the size of the target molecule, the amount of target molecule available, the desired binding affinity detection limit, and the desired speed of data collection.

[0067] The solvents used for the test mixtures can be any of a wide variety as long as they do not degrade (e.g., denature) the target. Typically water and DMSO are used. Protonated solvents can be used with the appropriate pulse sequence for suppressing the solvent signals. Such suppression sequences are well known to those of skill in the art.

[0068] If desired other components (e.g., buffers) can be added to the test mixtures for certain advantage, as is well known to one of skill in the art.

[0069] The present invention could also find useful applications for rapid screening of chemical mixtures (i.e., mixtures of two or more test compounds). Rapid screening techniques typically involve providing a plurality of test samples, each test sample comprising a mixture of two or more test compounds.

[0070] Once a ligand (preferably a high affinity ligand) has been identified and confirmed, its structure is used to identify available compounds with similar structures to be assayed for activity or affinity, or to direct the synthesis of structurally related compounds to be assayed for activity or affinity. These compounds are then either obtained from inventory or synthesized. Most often, they are then assayed for activity using enzyme assays. In the case of molecular targets that are not enzymes or that do not have an enzyme assay available, these compounds can be assayed for affinity using NMR techniques similar to those described above, or by other physical methods such as isothermal denaturation calorimetry. Compounds identified in this step with affinities for the molecular target of about 1.0×10−6 M or better are typically considered lead chemical templates.

[0071] Cryoprobe technology could further enhance the throughput of this screening process. In this case, the limiting factor will be the time required to change the sample, equilibrate the sample temperature, and shim the sample.

[0072] In some instances, ligand binding is further studied using more complex NMR experiments or other physical methods such as calorimetry or X-ray crystallography.

EXAMPLES

[0073] Objects and advantages of this invention are further illustrated by the following examples, but the particular materials and amounts thereof recited in these examples, as well as other conditions and details, should not be construed to unduly limit this invention.

Example I

Experimental Protocol

[0074] Fatty acid free human serum albumin (A-3782) was purchased from Sigma and used without further purification. The kinase domain MW˜34000 of a Serine/Threonine p21-activated kinase was expressed as a GST fusion protein in E. Coli and purified to homogeneity after removal of GST tag. Sucrose (S7903) and L-Trp (T0254) were purchased from Sigma, 5-CH3-D,L-Trp (69560) was purchased from Fluka. Diazepam was purchased from Carlo Erba, Italy.

[0075] NMR samples were in phosphate buffered saline (PBS) pH 7.4. D2O was added to the solutions (8% final concentration) for the lock signal. The small molecules were prepared in concentrated stock solutions in deuterated DMSO and stored at 253 K.

[0076] NMR Measurements. All spectra were recorded at 293 K with a Varian Inova 600 MHz NMR spectrometer equipped with a 5 mm triple-resonance inverse probe and a Sample Management System (SMS) autosampler. Water suppression in all experiments was achieved with the excitation sculpting sequence (T. -L. Hwang et al., J. Magn. Reson. A, 112, 275-279 (1995)). The two water selective 180° square pulses and the four PFGs of the scheme were 2.6 and 1 ms in duration, respectively. The gradient recovery time was 0.25 millisecond (ms). The data were collected with a sweep width of 7407 Hz, an acquisition time of 0.82 second (s), and a relaxation delay of 2.82 s. For the R1,s experiments, the relaxation delay ranged beween 5.82 to 10.82 s in order to achieve complete relaxation of the magnetization. Resonance selective inversion was achieved with a 24 ms duration 180° Gaussian pulse.

[0077] ITC Experiments. Calorimetric measurements were carried out with VP-ITC titration calorimeter (MicroCal). Heats of dilution were measured in blank titrations by injecting the protein into the buffer used in the particular experiment and were subtracted from the binding heats. Thermodynamic parameters were determined by non-linear least squares methods using routines included in the Origin software package (MicroCal, USA).

Results and Discussion

[0078] The first step is to identify a reference compound with medium to low affinity for the target. To accomplish this, a small library containing a few hundred very soluble and well characterized molecules is screened using the WaterLOGSY method (C. Dalvit et al., J. Biomol. NMR, 18, 65-68 (2000); and C. Dalvit et al., J. Biomol. NMR, 21, 349-359 (2001)). The identified binders are subsequently studied with isothermal titration calorimetry (ITC) experiments in order to determine their binding constants. One of these compounds, based on its binding constant, is selected as the reference compound for subsequent competition binding experiments. NMR-based HTS can then be carried out using either transverse or longitudinal relaxation experiments.

Transverse Relaxation

[0079] FIG. 1 shows an expanded spectral region of a selected compound in the absence and in the presence of a Serine/Threonine p21-activated Kinase. The molecule identified with WaterLOGSY has a KD of 7.1 μM as calculated from ITC measurements (see FIG. 2). The observed transverse relaxation rate constant (R2obs=FWHH *π) for a resonance of the ligand in the presence of the protein is provided by the equation (L. Y. Lian et al. in NMR of Macromolecules (G. C. K. Roberts, Ed.) Oxford University Press, pp. 153-182 (1993): 1$\begin{array}{cc}{R}_{2,\mathrm{obs}}=\frac{\left[\mathrm{EL}\right]}{\left[L_{\mathrm{TOT}}\right]}{R}_{2,\mathrm{bound}}+\left(1-\frac{\left[\mathrm{EL}\right]}{\left[L_{\mathrm{TOT}}\right]}\right)R_{2,\mathrm{free}}+\frac{\left[\mathrm{EL}\right]}{\left[L_{\mathrm{TOT}}\right]}{\left(1-\frac{\left[\mathrm{EL}\right]}{\left[L_{\mathrm{TOT}}\right]}\right)}^2\frac{4{{\pi }^2\left({\delta }_{\mathrm{free}}-{\delta }_{\mathrm{bound}}\right)}^2}{K_{-1}}& \left(1\right)\end{array}$

[0080] where [EL] is the concentration of bound ligand, [LTOT] is the total ligand concentration, R2,bound and R2,free are the transverse relaxation rate constant for the ligand in the bound and free state, respectively, δbound and δfree are the chemical shifts for the resonance of the ligand in the bound and free state, respectively and 1/K−1 is the residence time of the ligand bound to the protein. An increased broadening of the ligand resonances is observed in the spectrum of the ligand in the presence of the protein (see FIG. 1a). The resonance at 7.13 ppm, representing the sharpest resonance in the spectrum of the ligand in the absence of the protein (FIG. 1b), displays a significant broadening in the presence of the protein (FIG. 1a). The extensive broadening results from the contribution of the third term in equation (1). The compound binds in the ATP binding pocket of the kinase. The X-ray structure of the complex reveals the close proximity of this ligand proton to a phenylalanine. Therefore it is expected that the chemical shift difference δfree−δbound is large for this proton resonance. The exchange term does not contribute significantly to the R2,obs of the other resonances.

[0081] The next step is the acquisition of NMR spectra at different ligand concentrations and fixed protein concentration or different protein concentrations and fixed ligand concentration. The latter is preferred when the reference compound is not soluble at high concentrations and when the molecule has a second low-affinity binding site that will start to be partially populated at high concentrations. FIG. 3 shows the spectra of the ligand as a function of the total ligand concentration and ratio [EL]/[LTOT]. This ratio can be calculated from the knowledge of the ligand KD derived from ITC and the total protein [ETOT] and total ligand [LTOT] concentration used in the experiments according to the equation (L. Y. Lian et al. in NMR of Macromolecules (G. C. K. Roberts, Ed.) Oxford University Press, pp. 153-182 (1993); and A. Fersht, Enzyme Structure and Mechanism W. H. Freeman and Company New York, pages 98-120 (1985)): 2$\begin{array}{cc}\frac{\left[\mathrm{EL}\right]}{\left[L_{\mathrm{TOT}}\right]}=\frac{\left[E_{\mathrm{TOT}}\right]+\left[L_{\mathrm{TOT}}\right]+K_D-\sqrt{{\left(\left[E_{\mathrm{TOT}}\right]+\left[L_{\mathrm{TOT}}\right]+K_D\right)}^2-4\left[E_{\mathrm{TOT}}\right]\left[L_{\mathrm{TOT}}\right]}}{2\left[L_{\mathrm{TOT}}\right]}& \left(2\right)\end{array}$

[0082] At low [EL]/[LTOT] the NMR spectrum is approaching the spectrum of the ligand in the absence of the protein. The measured linewidth of the resonance at 7.13 ppm, or a more precise measurement, as shown in FIG. 4, the signal intensity ratio of the resonance at 7.13 ppm divided by another ligand resonance, are plotted as a function of [EL]/[LTOT]. Fitting functions are then used to extract an approximate value of the dissociation binding constant of the identified NMR hits.

[0083] The presence in a chemical mixture of a molecule competing with the reference molecule will result in a decrease in [EL]. This can be appreciated in FIG. 5, which shows the NMR spectrum of the reference molecule in the presence of a seven compound chemical mixture. The significant sharpening of the resonance at 7.13 ppm reveals the presence of a high-affinity ligand in the mixture. Deconvolution performed in the presence of the reference compound allows, as shown in FIG. 5, for the identification of the high affinity ligand. Despite the fivefold excess of the reference molecule concentration, the NMR-identified high affinity molecule causes almost complete displacement of the reference molecule from the receptor. Therefore, in order to obtain an approximate estimate of the dissociation binding constant for the high affinity ligand and not simply a lower limit, it is necessary to record an additional experiment with an even lower concentration of the NMR hit. The signal intensity ratio of the two reference molecule resonances, or the linewidth of the resonance at 7.13 ppm measured from these spectra, are used as input to calculate the value [EL]/[LTOT] from the fitting functions of FIG. 4. The knowledge of [LTOT] used in the NMR screening competition binding experiments permits one to calculate [EL] for the reference compound in the presence of the competing molecule. With [LTOT], [EL] and [ETOT] known it is possible to determine the apparent dissociation binding constant KDapp of the reference compound according to the equation: 3$\begin{array}{cc}{K}_D^{\mathrm{app}}=\frac{\left[E_{\mathrm{TOT}}\right]\left[L_{\mathrm{TOT}}\right]-\left[E_{\mathrm{TOT}}\right]\left[\mathrm{EL}\right]+{\left[\mathrm{EL}\right]}^2-\left[L_{\mathrm{TOT}}\right]\left[\mathrm{EL}\right]}{\left[\mathrm{EL}\right]}& \left(3\right)\end{array}$

[0084] In the approximation of a simple competitive mechanism the KDapp is then used to extract the binding constant K1 of the NMR-identified ligand according to the equation: 4$\begin{array}{cc}{K}_1=\frac{\left[I\right]K_D}{K_D^{\mathrm{app}}-K_D}& \left(4\right)\end{array}$

[0085] where [I] is the concentration of the competing molecule. Table 1 shows the K1 obtained using this approach for two different compounds with binding constants ranging from nM to μM.

TABLE 1
Single point NMR-derived binding constants for two inhibitors of the kinase and their
comparison with measured ITC values.
Inhibitor	I7.13/I6.65	I7.13/I7.52	KD	[I]	[ETOT]	[EL]/[LTOT]	[EL]	KDapp	KiNMR	KiITC
Cpd. E	1.143		5.6	40	1.5	0.0081	0.41	133.9	2.5 ± 0.7	2.5(a)
	1.143		10.0	40	1.5	0.0079	0.39	140.0
		0.827	5.6	40	1.5	0.0099	0.50	100.2	3.3 ± 0.9
		0.827	10.0	40	1.5	0.0097	0.48	104.3
Cpd. E	1.428		5.6	60	1.1	0.0041	0.21	217.4	2.2 ± 0.6	2.5(a)
	1.428		10.0	60	1.1	0.0039	0.20	230.4
		1.020	5.6	60	1.1	0.0059	0.30	135.6	3.7 ± 1.1
		1.020	10.0	60	1.1	0.0058	0.29	138.8
Cpd. B	0.995		5.6	3	2.5	0.0108	0.54	180.2	0.14 ± 0.04	0.25(b)
	0.995		10.0	3	2.5	0.0104	0.52	187.5
		0.714	5.6	3	2.5	0.0123	0.61	152.2	0.16 ± 0.05
		0.714	10.0	3	2.5	0.0118	0.59	159.8
[LToT] is 50 μM.
The KD values of 5.6 and 10 μM of the reference compound Compound A correspond to the lower and upper limits determined by the error in KD derived from the ITC measurement.
The same curves as described in FIG. 4 were derived for the two KD values.
KiNMR is the binding constant in μM measured with equation (4), KiITC is the binding constant in μM derived from ITC measurements (a) value measured at 283 K, (b) value measured at 298 K.
The binding thermodynamics parameters were KB = 4.0 ± 2.0 105 M−1 (a), 0.4 ± 0.02 105 M−1(b), ΔHobs = +1.5 ± 0.4 Kcal/mol (a), −5.5 ± 0.1 Kcal/mol (b) TΔS = 9.1 Kcal/mol (a), 3.3 Kcal/mol (b), N = 1.1 ± 0.1 (a), 0.9 ± 0.1 (b).
For Compound E two experiments were recorded at different ligand and protein concentration.
With a ±5% error in the signal intensity ratio measurement the KiNMR are, from top to bottom, 2.6 ± 0.9, 3.5 ± 1.4, 2.4 ± 1.2, 4.0 ± 2.1, 0.15 ± 0.05 and 0.16 ± 0.06.

[0086] The good qualitative agreement between the NMR-derived and ITC-measured binding constants is very promising. It should be pointed out that the NMR-derived values were obtained from a single point measurement. An advantage of the method proposed here is that it permits the determination of the concentration of [I]. Such task is achieved simply by comparing the integral of a resonance of the reference molecule for which the precise concentration is known and the integral of a resonance of the competing molecule. This is important for obtaining a meaningful value of the binding constant. An error in the effective concentration of the competing molecule either due to poor solubility of the molecule, or low purity would result in a large error in the determination of the binding constant

[0087] The example described here may represent a fortunate case since at 600 MHz one ligand resonance is in the so-called intermediate exchange regime. Working at stronger magnetic fields may be beneficial since the third term of equation (1) becomes larger and therefore many potential reference molecules may display at least one resonance in the intermediate exchange regime. However, even when the exchange term does not contribute significantly to the linewidth it is possible to perform similar competition binding experiments. In this case the R2, obs of the reference compound is measured with a Carr Purcell Meibom Gill (CPMG) pulse sequence (S. Meibom et al., Rev. Sci. Instrum., 29, 688 (1958)) at different [LTOT] or [ETOT]. The measured R2 values or the signal intensity ratio of a resonance in two experiments one recorded without R2 filter and the other recorded with a long CPMG are then plotted against [EL]/[LTOT] in a similar way as the graph of FIG. 4. The extent of reduction of R2 or the change of the signal intensity ratio for the reference compound in the presence of a competing molecule permits one to extract an approximate value of the binding constant of the inhibitor. This method described here should be very sensitive in the SLAPSTIC experiment (W. Jahnke et al., J. Am. Chem. Soc. 123, 3149 (2001)) where the difference in R2 for the bound and free state of the ligand is large. Other methods that can be used with the competition binding experiments are the single or multiple-selective R2 experiment (90°-τ-G-180°sel-G-τ-Water Suppression) and the 1 D selective TQCSY. G are gradients of same strength and 180°sel is a single or multiple selective inversion pulse applied at the chemical shift of the reference compound resonances. Typically, two spectra are recorded, one with τ=0 another with a long τ period. The intensity ratios of the signals extracted from the two spectra are used for the titration measurements and for the screening process as described above. For the system used in this work it was sufficient to record a selective R2 experiment with a single spin-echo period and double-selective inversion of the two resonances at 7.13 and 6.65 ppm (data not shown). In cases of severe overlap the 1 D selective TOCSY can also be used. However, it is applicable only if the reference compound contains scalar coupled spin systems. The selective excitation is achieved with the same scheme used in the selective R2 experiment. Two experiments are recorded as described above and with a fixed spin-lock period.

[0088] A similar approach, but using selective R1 is shown in the next example.

Longitutinal Relaxation

[0089] The observed longitudinal relaxation rate R1,obs of a molecule interacting weakly with a macromolecule is given by the equation (G. Valensin et al., J. Magn. Reson., 46, 23-29 (1982)): 5$\begin{array}{cc}{R}_{1,\mathrm{obs}}=\frac{\left[\mathrm{EL}\right]}{\left[L_{\mathrm{TOT}}\right]}{R}_{1,\mathrm{bound}}+\left(1-\frac{\left[\mathrm{EL}\right]}{\left[L_{\mathrm{TOT}}\right]}\right)R_{1,\mathrm{free}}& \left(5\right)\end{array}$

[0090] where R1,bound and R1,free are the longitudinal relaxation rate constants for the ligand in the bound and free state, respectively.

[0091] Non-selective R1 lacks the direct dependence on τc (i.e. tumbling correlation time) and therefore is not suited for monitoring binding events. Selective R1 (R1,s) contains the direct τc dependence necessary for identifying small molecules interacting with macromolecules (G. Valensin et al. in “NMR Spectroscopy in Drug Research”, J. W. Jaroszewski, K. Schaumburg, H. Kofod, Eds., p. 409, Munksgaard, Copenhagen (1988); E. Gaggelli et al., Magn. Reson. Chem., 30, 461 (1992); C. Rossi et al., Chem. Phys. Lett., 264, 205-209 (1997); C. Rossi et al., Chem. Phys. Lett., 310, 495-500 (1999); C. Rossi et al., Magn. Reson. Chem., 39, 457-462 (2001); and G. Veglia et al., J. Magn. Reson., 130, 281-286 (1998)).

[0092] R1,s experiments require the selective inversion of one resonance of the ligand. This technique has been used to characterize the binding of known ligands. Although these experiments have not been used in NMR screening because of the problems of achieving selectivity for a large library of chemically diverse compounds, they are particularly suited for performing HTS NMR screening with competition binding experiments. In these experiments it is sufficient to selectively invert always the same resonance of the reference molecule. Owing to the fact that the reference molecule is in excess compared to the protein and the molecule is a weak to medium affinity ligand, the observed chemical shift of the ligand resonances corresponds to the chemical shift of the free ligand. This permits the acquisition of the experiments in automation mode using the same excitation frequency for the selective inversion. For maximum sensitivity of the experiment the resonance to be selectively inverted should be chosen from among the ligand resonances that display the largest difference of R1,s in the free and bound state. When possible, a singlet is preferred to obtain greater intensity and reduced problems of overlap. It should be pointed out the even in the presence of overlap it is possible to detect ligands both in the R2 and R1,s experiments. In these particular cases it is necessary to record the same experiments also for the mixture in the absence of the reference compound. Subtraction of the spectra in the presence and absence of the reference compound permits one to ascertain the presence or not of a ligand.

[0093] The titration experiments are performed either by keeping the protein concentration fixed and varying the reference compound concentration or by keeping the ligand concentration fixed and varying the protein concentration.

[0094] The system used in this example is human serum albumin (HSA) and the reference compound is tryptophan (Trp). The endogenous Trp and other Trp derivatives bind on Sudlow site II of HSA (T. Peters, Jr. in All about Albumin Biochemistry, Genetics, and Medical Applications, Academic Press, San Diego, U.S.A., pages 109-111 (1996)). Therapeutic agents such as naproxen, diazepam and ibuprofen also bind to HSA on Sudlow site II (T. Peters, Jr. in All about Albumin Biochemistry, Genetics, and Medical Applications, Academic Press, San Diego, U.S.A., pages 113-114 (1996); and U. Kragh-Hansen, Biochem. J., 273, 641-644 (1991)).

[0095] An R1,s measurement for the C2-H resonance of Trp in the absence and presence of HSA is shown in FIG. 6. In the presence of the protein the R1,s becomes larger.

[0096] Experimental observables that can be plotted as a function of [EL]/[LTOT] are (i) the intensity ratio of an inverted resonance with a non inverted resonance in a R1,s experiment recorded with a single τ value (the time between the selective 180° pulse and the detection hard 90° pulse), or (ii) the intensity ratio of an inverted resonance in two R1,s experiments recorded with two different τ values as shown in FIG. 7. For better sensitivity the magnetization has to be negative with the first τ value and positive with the second τ value. In addition these τ values should be chosen in the time region where the largest intensity difference is observed between the spectra (a) and (b) of FIG. 6. This time region can be derived from a simple simulation as shown in FIG. 8. The advantage of (i) and (ii) that are defined as the R1,s filtered experiments is the rapid data acquisition since it is not necessary to measure the R1,s values.

[0097] FIG. 7 shows the titration experiments performed with tryptophan at a fixed 100 μM concentration and different HSA concentrations ranging from 10 to 0 μM. The diagrams describe the signal intensity ratio of the C2-H resonance in two experiments recorded with τ=0.48 and 1.91 s (upper trace of FIG. 7) and the signal intensity ratio of the inverted C2-H resonance and the non-inverted C4-H resonance in an experiment recorded with τ=1.91 s (lower trace of FIG. 7) as a function of the fraction of bound ligand.

[0098] It would also be possible to measure the R1,s rates by acquiring many experimental points (G. Valensin et al., J. Magn. Reson., 46, 23-29 (1982); and C. Rossi et al., Chem. Phys. Lett., 264, 205-209 (1997)). However, these measurements are rather lengthy and therefore not suitable as a rapid method for screening. In cases of overlap it is sufficient to selectively invert a multiplet resonance of the spy molecule and then add to the R1,s filter a selective TOCSY step. This step is performed with selective excitation of the multiplet resonance after the delay τ followed by a non-selective or doubly-selective (B. Boulat et al., J. Am. Chem. Soc., 114, 5412-5414 (1992)) homonuclear Hartmann-Hahn transfer.

[0099] Screening is performed using an R1,s filtered experiment recorded with a single τ value or two τ values as described above. For maximum sensitivity this delay should correspond to the τ value or values at which the largest intensity difference is observed, as shown in FIG. 8. For rapid visual inspection the τ value corresponding to the null point (i.e., 1-2exp(−τ R1,s)=0) can also be selected as shown in FIG. 9. The presence in the chemical mixture of a competing molecule will result in an R1,s filtered spectrum with a residual negative signal (FIG. 9b) (i.e., the R1,s of the reference compound is smaller because of partial displacement of the reference compound from the protein).

[0100] Deconvolution of the mixture carried out with the same R1,s filtered experiment in the presence of the reference molecule identified diazepam as the molecule competing with Trp (FIG. 9d).

[0101] It is possible to reduce the complexity of the NMR spectrum of the screened mixture by subtracting the R1,s spectrum of the mixture from the spectrum containing only the reference spectrum. The resulting spectrum (not shown) contains only the signals of the molecules comprising the mixture and a negative signal for the C2-H resonance of Trp. The other signals of the reference compound and the signals of the protein are absent. This strategy can be applied also in the transverse relaxation experiments described above.

[0102] With the graphs of FIG. 7 and following the same procedure described for the transverse relaxation experiments, it is possible to estimate the binding constant of the competing molecule as reported in Table 2.

TABLE 2
Single point NMR-derived binding constant for diazepam bound to HSA and its
comparison with the equilibrium dialysis measured value.
[I]	KD(a)	[ETOT]	[LTOT]	I4H/I2H	1/R 1,S	[EL]/[LTOT]	[EL]	KDapp	KiNMR	Ki(b)
30	23	8	100	1.282		0.0344	3.44	127.8	6.6	2.6
30	23	8	100		1.765	0.0260	2.60	202.9	3.9
[I] is the diazepam concentration in μM, 1/R1,s is the selective longitudinal relaxation of C2-H Trp in sec, I4H/I2H is the signal intensity ratio for the C4-H and C2-H Trp resonances in an R1,s experiment recorded with selective inversion of the C2-H resonance and a single τ value (τ = 1.905s).
(a)Binding constant of L-Trp and
(b)binding constant of diazepam in μM measured with equilibrium dialysis experiments (U. Kragh-Hansen, Biochem. J., 273, 641-644 (1991)) at 20° C.
Only the average value of KD was used due to the small reported error (U. Kragh-Hansen, Biochem. J., 273, 641-644 (1991)).
KiNMR is the binding constant of diazepam in μM, measured with equation (4).
With a ±5% error in the experimental measurement the KiNMR are, from top to bottom, 7.0 ± 1.8 and 4.2 ± 1.6.

[0103] The binding constant derived from NMR for diazepam although very similar in magnitude to the value measured by equilibrium dialysis reported in the literature (U. Kragh-Hansen, Biochem. J., 273, 641-644 (1991)) is somewhat larger. This small difference can probably be ascribed to the presence of a low affinity second binding site for Trp. The affinity for this site at 20° C. is of the order of several hundreds μM as measured previously by NMR. Similar findings were observed for Indole-3-acetic acid (data not shown). Despite this complication, the values that are rapidly derived from the NMR screening experiments serve for most of the purposes in the early stages of drug discovery.

[0104] Other techniques more suitable for precisely measuring binding constants can then be used for detailed characterization of the binding kinetics of the NMR-identified ligands. A drawback of the R1,s competition binding experiments is the detection of only molecules competing with the reference molecule for the same protein binding site. By contrast R2 and competition-WaterLOGSY experiments also permit the detection of weak to medium affinity ligands that bind at different sites. This is clearly valid only if the concentration of the compounds comprising the mixture is comparable to the concentration of the reference molecule.

[0105] The technique can also be used for rapidly ranking lead molecules for their affinity to HSA. In a pharmaceutical project aimed at identifying inhibitors of matrix metalloproteinases (MMPs), it was discovered by NMR that the presence in plasma of one lead molecule of a defined chemical class resulted in the displacement of the endogenous tryptophan from HSA. This is strong evidence that the compounds of this chemical class were binding on Sudlow site II of HSA. Prompted by these finding selective longitudinal (R1,s) competition binding experiments using tryptophan as the spy molecule were performed. It should be pointed out that the experiments could be performed also with the use of two reference compounds binding to two different sites on HSA. In this case it will be sufficient to record multi selective R1,s competition binding experiments.

[0106] The experimental conditions selected for the R1,s filtered competition binding experiments for HSA corresponded to the right penultimate point on the diagrams of FIG. 7, namely a concentration for Trp and HSA of 100 and 8 μM, respectively. FIG. 10 (b,c) shows two R1,s filtered competition binding experimens performed for two different molecules of the same chemical class. The same spectra recorded only for Trp are shown in FIG. 10a. The two molecules have similar inhibition potency for the MMP receptor and are both lead molecules. It is evident from these spectra that the two molecules have substantial different affinity for the Sudlow site II on HSA. This is inferred from the different amount of Trp displaced from HSA. At τ=0.96 s the C2-H signal is at the null point for Trp alone, is slightly negative in the presence of Compound C and becomes very negative in the presence of Compound D. The lengthening of T1 reflects, according to the diagram of FIG. 7, a reduced fraction of Trp bound to HSA. The diagrams of FIG. 7 are used for deriving the fraction of bound Trp in the presence of the competing molecule. With [LTOT], [EL] and [ETOT] known it is possible to determine the apparent dissociation binding constant KDapp of Trp according to equation (3) above.

[0107] In the approximation of a simple competitive mechanism the KDapp is then used to extract the binding constant K1 of the NMR-identified ligand according to equation (4) above.

[0108] Typically, the binding constants were measured at two different concentrations for the competing molecule. In a simple competitive mechanism the measured K1 at the two different concentrations will be the same.

[0109] FIG. 11 shows the R1,s filtered experiment recorded with τ=0.96 s for seven lead molecules of the MMP project. The compounds are ranked according to their binding affinity. The first experiment on the left corresponds to the control experiment with Trp alone. When the concentration of the molecules is the same in all the experiments, the absolute intensity of the negative signal of the C2-H Trp is directly proportional to the affinity of the competing molecule for HSA. The most potent ligand for HSA is on the most right of the figure. The binding constants derived with this method and the percentage of the ligand bound to HSA in the assumption of a HSA and compound concentration in plasma of 0.7 mM and 100 μM, respectively, are reported.

[0110] Recently, an NMR method has been presented for obtaining binding data and structural information on HSA/ligand complexes (H. Mao et al., J. Am. Chem. Soc., 43, 10429-10435 (2001)). The technique uses isotopically labeled HSA-III domain and detects the binding events with two-dimensional 15N/1H HSQC spectra. The binding constants are measured by monitoring the chemical shift changes of the protein backbone amide signals as a function of the ligand concentration. Both methods provide reliable values for the binding constant. This approach does not require isotopically labeled HSA necessitates only a small amount of HSA and it is applied to the full-length protein.

Example II

Experimental Protocol

[0111] Fatty acid free human serum albumin (A-3782) was purchased from Sigma and used without further purification. NMR samples were in phosphate-buffered saline (PBS, code: P-3813, Lot 100K8211 from Sigma) pH 7.4 in the presence of 5 μM EDTA. D2O was added to the solution (8% final concentration) for the lock signal. The small molecules were prepared in concentrated stock solutions in either deuterated DMSO or water and stored at 253 K.

[0112] NMR. All NMR spectra were recorded at 300 K with a Bruker Avance 600 MHz NMR spectrometer equipped with a three channel inverse probe. Water suppression in all the experiments was achieved with the excitation sculpting sequence (T. -L. Hwang et al., J. Magn. Reson. A, 112, 275-279 (1995)). The two water selective 180° square pulses and the four pulsed field gradients of the water suppression scheme were 2.7 and 1 ms in duration, respectively. The gradient recovery time was 200 μs long. The multi-selective experiments were performed with Gaussian 180° shifted laminar pulses (SLP) (J. Boyd et al., J. Magn. Reson., 85, 406 (1989); and S. L. Patt, J. Magn. Reson., 96, 94 (1992)) of 25 ms in duration. The data were collected with a sweep width of 11.97 ppm and an acquisition time of 0.84 s.

[0113] Fluorescence. Albumin affinities were determined using a previously described method (D. E. Epps et al., Anal. Biochem., 227, 342-350 (1995)). Fluorescence measurements were acquired on a Jasco J-715 spectropolarimeter using an auxiliary photomultiplier tube positioned perpendicular to the excitation beam. The excitation wavelength was 310 nm (with a 5 nm bandwidth) and a 385 nm cut-off filter was employed. Affinity measurements were made using the same source of fatty acid free HSA as used for NMR experiments. Analyte and HSA solutions were prepared in phosphate-buffered saline (PBS, code: P-3813, Lot 100K8211 from Sigma) pH 7.4 in the presence of 5 μM EDTA. The buffer was filtered through a 0.2 μm filter prior to use. Albumin affinity was determined by aliquoting 2.0 mL of a 3 μM solution of analyte into a quartz cuvette, pathlength of 1.0 cm, and titrating the solution with HSA (stock concentration of 250 μM). The change in fluorescence was recorded with each injection and the resulting binding isotherm fit according to the following equation,

F= (αW0+(β−α)W0B0/(Kd+B0)   (6)

[0114] Where, F is the measured fluorescence, α and β are proportionality constants, W0 is the initial ligand concentration, B0 is the albumin concentration and Kd is the equilibrium dissociation constant. Data were fit using a nonlinear least-squares method available in Origin 7.0 software (OriginLab Co., Northampton, Mass.). All parameters were permitted to float during the fitting procedure, except the concentration of analyte was held at 3.0 μM. Titrations were repeated and the reported affinities are the average of the two titrations.

[0115] ITC. Isothermal titration calorimetry experiments were performed using an OMEGA titrating microcalorimeter from Microcal, Inc. (Northampton, Mass.). The titrating microcalorimeter consisted of a sample and reference cell held in an adiabatic enclosure. The reference cell was filled with PBS. A 30 μM solution of HSA in PBS+5 μM EDTA was placed in the 1.37 mL sample cell. Analyte at 0.7 mM in the same buffer was held in a 250 μL syringe. Thirty injections (6 μL each) of analyte were made by a computer controlled stepper motor into the sample cell held at 37° C. The syringe stir rate was 400 rpm. Heat adsorbed or released with each injection was measured by a thermoelectric device connected to a Microcal nanovolt preamplifier. Titration isotherms for the binding interactions were comprised of the differential heat flow for each injection. Heat of dilution obtained by injecting analyte into PBS was negligible. Binding isotherms were fit to a single binding site model (B. Loun et al., Anal. Chem., 66, 3814-3822 (1994)) using an iterative nonlinear least-squares algorithm included with the instrument.

Results and Discussion

Selection of Control and Spy Molecules

[0116] Two molecules are detected in the multi-selective experiments: a control molecule that does not bind to the receptor of interest and a compound that interacts with a weak to medium binding affinity to the receptor (B. J. Stockman et al., Progr. Nucl. Magn. Res. Spec., 41, 187-231 (2002); A. H. Siriwardena et al., Angew. Chem. Int. Ed., 41, 3454-3457 (2002); and C. Dalvit et al., Comb. Chem. HTS, 5, 605-611 (2002)). The latter are referred to as the reference or ‘spy’ molecule. In the simplest case, control and spy molecules are chosen that have singlet resonances in isolated spectral regions where spectral overlap is unlikely.

[0117] In the specific case of HSA, many drugs bind to one of two primary binding sites: Sudlow site I (located in subdomain IIA) and Sudlow site II (located in subdomain IIIA) (T. Peters, Jr., All about Albumin Biochemistry, Genetics, and Medical Applications Academic Press, San Diego, U.S.A. 1996). Sudlow site I accommodates bulky heterocyclic anions with a centralized charge, such as bilirubin, warfarin, and cyclic eicosanoids. Sudlow site II binds to hydrophobic aromatic moieties like those present in diazepam, ibuprofen, and L-tryptophan. Recently the X-ray structure of warfarin bound to recombinant HSA has been solved at high resolution (I. Petitpas et al., J. Biol. Chem., 276, 22804-22809 (2001)). In this work, warfarin (1) was selected as the spy molecule to monitor binding at Sudlow site I because of its excellent solubility in acqueous buffers and because it contains a methyl group. The methyl group NMR resonance is a sharp singlet signal at 2.21 ppm. Likewise, 5-CH3 D,L Trp (2) was chosen as the control molecule because it also contains a methyl group that gives rise to a sharp singlet resonance at 2.43 ppm. Previous ITC and WaterLOGSY experiments have clearly demonstrated that up to concentration of several hundred μM this Trp derivative does not interact with HSA (C. Dalvit et al., J. Med. Chem., 45, 2610-2614 (2002)). Selection of molecules containing methyl groups allows for a reduction of the measuring time due to the high intensity of the methyl group signals. In addition, this permits one to lower the concentration of the reference and control molecules thus avoiding problems arising from aggregation, and non-specific binding.

Multi-Selective NMR Experiments

[0118] After selecting the control and spy molecules, multi-selective 1H NMR spectra are performed as a function of the protein concentration (C. Dalvit et al., J. Am. Chem. Soc., 124, 7702-7709 (2002); C. Dalvit et al., Comb. Chem. HTS, 5, 645-650 (2002); and in Example (I) above). This titration is necessary for identifying the optimal conditions for subsequent NMR-based screening experiments and for deriving the binding constants of the NMR hits. FIG. 12 shows the titration experiments obtained with the doubly selective R1 filtered experiments. The spectra have been recorded with simultaneous selective inversion of the two molecules' methyl resonances and with a filter delay of 0.4 s. Although the concentration of the two molecules is the same (25 μM) the signal of warfarin appears more negative in the spectrum recorded in the absence of the protein (FIG. 12 left). This is simply caused by the longer selective R1 of the warfarin methyl group (0.50 s−1) compared to the Trp derivative methyl group (0.77 s−1). This difference in relaxation can easily be explained by the different proton densities in proximity of the two molecules' methyl groups. As shown in FIG. 12, the 5-CH3 Trp signal does not change in intensity as a function of the protein concentration because it does not bind to the protein and therefore it represents an internal standard. By contrast, signal attenuation is observed for the warfarin resonance with increasing protein concentration.

[0119] The next step in this approach requires the measurement of the binding constant of the spy molecule (C. Dalvit et al., J. Am. Chem. Soc., 124, 7702-7709 (2002); C. Dalvit et al., Comb. Chem. HTS, 5, 645-650 (2002); and in Example (I) above). This is carried out with either ITC or fluorescence spectroscopy. Warfarin has an intrinsic fluorescence in polar solvents that is enhanced several fold in non-polar environments. Such an environmental change occurs when warfarin binds to the relatively hydrophobic pocket of albumin. FIG. 13 shows the binding isotherms obtained for HSA titrations of sodium warfarin and its analogue 4-hydroxy-3-[1-(p-iodophenyl)-3-oxobutyl] coumarin (4) followed by the change in fluorescence. Initially, there was a large increase in fluorescence as the analyte bound to albumin. This rise gradually plateaued as the binding site was saturated. The data were well fit to the binding equation described in the Material and Methods section as evidenced by the good agreement between the experimental points and the theoretical fit represented by the solid line. These titrations were repeated and the average equilibrium binding constants were determined to be 6.1±0.3 μM for sodium warfarin and 3.3±0.3 μM for compound (4).

[0120] A plot of the NMR signal intensity ratio of the two molecules methyl resonances in the R1 filtered experiments as a function of the fraction of bound warfarin is then constructed as shown in FIG. 14(a). The fraction of bound warfarin is calculated from the KD value obtained from the fluorescence measurements as previously explained (C. Dalvit et al., J. Am. Chem. Soc., 124, 7702-7709 (2002); C. Dalvit et al., Comb. Chem. HTS, 5, 645-650 (2002); and above in Example (I)). FIG. 14(b) illustrates the same signal intensity ratio measured in a spin-echo filter experiment recorded with a total period of 256 ms (4 cycles of 64 ms). The two graphs display the same monotonic trend as a function of the fraction of bound ligand. However, differences are observed in the experiments recorded at higher protein concentration (2.5 μM). This arises from the fact that at high protein concentration the warfarin resonance signal in the spin-echo experiment approaches the asymptote 0 value.

[0121] Screening is then performed by selecting the optimal experimental conditions according to the plots of FIG. 14. This is shown in the spectrum of FIG. 15 where the two resonances in the spin-echo and the doubly-selective R1 filtered experiments are monitored for the presence in chemical mixtures of compounds that compete with Warfarin for the same binding site on HSA. Tolbutamide (3), a molecule known to bind to HSA (K. J. Fehske et al., Mol. Pharmacology, 21, 387-393 (1982)), partially displaces warfarin as shown in FIG. 15 (middle). By following the same procedure described previously and using the graphs of FIG. 14 it is possible to determine the binding constant of the competing molecule. It should be pointed out that ligands often display multiple binding sites on HSA with different binding constants, typically one with high binding affinity and another with low binding affinity. Therefore the NMR derived binding constant will represent an approximate value. Despite this uncertainty, the method provides a rapid and reliable tool for ranking compounds of a defined chemical class for their binding affinity to a specific site on HSA. The KD for two molecules (3) and (4) measured with this method are reported in Table 3 together with the corresponding values obtained from ITC or fluorescence measurements. Tolbutamide (3) showed an intrinsic fluorescence that increased as it bound to albumin, but the binding isotherm showed only a small degree of curvature with a ten fold excess of HSA added. Therefore, the error in the binding constant determination was large. Isothermal titrating calorimetry was then used to determine binding affinity of tolbutamide to HSA and resulted in a KD of 26±13 μM. The KD values derived from the R1 filter and the spin-echo competition experiments are very similar. In addition there is a good agreement between the single-point NMR derived KD values with the KD values obtained from full-titration ITC and fluorescence measurements. This agreement is obtained clearly only for a pure competition binding mechanism provided that there are no allosteric effects.

TABLE 3
Single-Point NMR-Derived Binding constants for two ligands of HSA and their comparison with measured
fluorescence values.
	I(5-CH3 W)/I(CH3
	Warfarin)
	double-R1,s
Compound	filter	spin-echo	KD	[I]	[ETOT]	[EL]/[LTOT]	[EL]	KDapp	KINMR	KD
Tolbutamide	1.085		6.1a	150	1.5	0.023	0.575	39.3	27.6	26.0b
		1.559	6.1a	150	1.5	0.024	0.6	36.6	30.0
4-hydroxy-3-[1-(p-	1.275 6.1a	25	2.5	0.03	0.75	56.6	3.0	3.3a
iodophenyl)-3-oxobutyl]
		1.863	6.1a	25	2.5	0.029	0.725	59.4	2.9
The values of the binding constants are expressed in μM.
aValues determined by fluorescence,
bvalues determined by ITC.

Multi-Selective One-Dimensional COSY and TOCSY Experiments

[0122] It is possible that problems of overlap will be encountered even when the reference and control molecules are selected according to the criteria described above. This may occur when complex chemical mixtures are screened and/or when large NMR signals arising from buffer or detergents are present. A convenient approach for circumventing this problem is the selection of molecules according to the presence in their NMR spectra of multiplet resonances. The spectra should contain at least one doublet resonance or, in the best case, a weakly scalar coupled two spin system for improved sensitivity. If the monitored resonance overlaps with other signals of the screened mixture, it is now possible to extract its relaxation properties by relaying it through scalar coupling mechanism to another resonance (B. Boulat et al., J. Am. Chem. Soc., 114, 5412-5414 (1992)). The relaying process can be achieved with either COSY or TOCSY coherent magnetization transfer. The pulse sequences for the 1 D multi-selective COSY and refocused COSY experiments used in this work are described in FIG. 16. As with all ligand-based competition binding screening methods, the resonances of the actual compounds to be screened are not evaluated. Instead, the τ period in both pulse sequences is optimized for the spy molecule's resonances based on its known binding constant for the receptor. The C3,5-H2 resonance (7.35 ppm) of (3) (reference molecule) undergoes significant broadening in the presence of HSA and therefore its intensity is monitored in the competition binding experiments. Unfortunately, the signal resonates at a chemical shift where significant overlap is present as observed in the example of FIG. 17 (bottom). Therefore its intensity cannot be measured precisely. The use of selective COSY permits one to monitor its intensity indirectly via the scalar coupled C2,6-H2 resonance at 7.69 ppm (FIG. 17 (top)).

[0123] Ligand-based competition binding NMR screening performed with this approach is shown in FIG. 18. Both the methyl group resonance of 5-CH3 Trp and the C3,5-H2 resonances of Tolbutamide are simultaneously excited in three different types of 1 D multi-selective experiments. For the control molecule the singlet signal was excited because it does not overlap with other signals. If problems of overlaps are encountered for this control molecule resonance, it would be just as convenient to selectively excite either the C6-H or C7-H doublet resonance. The COSY signal for Tolbutamide is weak in the spectrum in the presence of HSA and absence of the competing molecule (FIG. 18 (upper left)). This results from the rapid relaxation of its antiphase magnetization. In addition the antiphase appearance of the multiplet results in some cancellation of the positive and negative signals. All of these effects contribute to the difference observed for the spy molecule's signal in the spectra recorded in the presence and absence of the protein as shown in FIG. 18 (left and right, respectively). The observed difference is smaller in the TOCSY spectra (FIG. 18 bottom) because of the shorter spin-locking period used and because of the slower relaxation of in-phase magnetization. The signal of the control molecule remains constant throughout the experiments and it is used as an internal marker. Prior to the screening process, titration 1 D multi-selective COSY or refocused COSY or TOCSY experiments are recorded as a function of the protein concentration. The COSY or TOCSY signal intensity ratio of the control molecule and the reference molecule is plotted as a function of the fraction of bound ligand in the same way as described above. The presence of (4) in the mixture results in significant displacement of tolbutamide as shown in FIG. 18 (middle). These spectra are similar to the spectra of tolbutamide in the absence of the protein as shown in FIG. 18 (right). The extensive displacement is in agreement with the NMR derived binding constants for the two molecules reported in Table 3.

[0124] The presence of molecules that have mutually scalar coupled resonances at exactly the same frequencies as the two doublet signals of the spy molecule would interfere with the measurements, but this occurrence would be rare.

Conclusion

[0125] It has been shown that with properly designed competition binding experiments it is possible to perform HTS with NMR for the detection of high to medium affinity ligands. The initial step of the approach typically involves the identification of a medium to low affinity ligand and the determination of its binding constant with ITC measurements. Transverse or R1,s experiments are recorded at different [LTOT] or [ETOT] and different experimental observables are plotted as a function of [EL]/[LTOT]. These graphs are used for designing the proper set-up of the NMR experiments and for extracting an approximate value for the binding constant of the detected ligand. With this approach it is possible to rapidly screen thousands of compounds against protein or DNA and RNA fragments. The methodology can be extended also to the screening of plant and fungi extracts.

[0126] Another useful application of these experiments is in the screening of a class of compounds for HSA binding affinity. Compounds that retain good activity for the desired receptor but have reduced HSA binding can be rapidly identified with this technology. This information is highly valuable in drug development.

[0127] The use of one-dimensional multi-selective 1H experiments in combination with the competition binding experiments permits rapid screening of large chemical mixtures against protein, DNA or RNA fragments. This method overcomes most of the problems associated with the severe proton spectral overlap encountered with the screening of complex chemical mixtures. The use of two molecules, a reference compound and a control molecule permits one to perform the screening by simply monitoring two signals recorded in multi-selective 1H COSY or TOCSY experiments. Finally, the precise measurement of the relative intensity of the two resonances permits determination of the binding constant of the NMR-hit. For a simple competition mechanism process the NMR derived binding constants are in excellent agreement with the values derived from other techniques.

[0128] The complete disclosures of the patents, patent documents, and publications cited herein are incorporated by reference in their entirety as if each were individually incorporated. Various modifications and alterations to this invention will become apparent to those skilled in the art without departing from the scope and spirit of this invention. It should be understood that this invention is not intended to be unduly limited by the illustrative embodiments and examples set forth herein. Such examples and embodiments are presented by way of example only with the scope of the invention intended to be limited only by the claims set forth herein as follows.

Claims

1. A method of identifying a ligand to a target molecule, the method comprising: providing a reference compound that interacts with the target molecule; collecting a 1 D 1H nuclear magnetic resonance spectrum of the reference compound in the presence of the target molecule; providing a test sample comprising at least one test compound; collecting a 1 D 1H nuclear magnetic resonance spectrum of the reference compound in the presence of each test sample and the target molecule; comparing the spectrum of the reference compound in the presence of the target molecule to the spectrum of the reference compound in the presence of each test sample and the target molecule to determine a change in one or more of the reference compound resonances; and identifying at least one test compound that interacts with the target molecule, wherein the test compound displaces the reference compound.

2. The method of claim 1 wherein the test compound has a binding affinity at least as tight as that of the reference compound.

3. The method of claim 1 wherein a change in one or more of the reference compound resonances comprises an increase in signal intensity in at least one reference resonance.

4. The method of claim 1 wherein identifying at least one test compound comprises recording separate 1 D 1H nuclear magnetic resonance spectra of the reference compound in the presence of each test compound and the target molecule.

5. The method of claim 1 further comprising: collecting 1 D 1H nuclear magnetic resonance spectra of the reference compound in the presence of the target molecule at different concentrations of the reference compound; and determining the dissociation constant of the test compound.

6. The method of claim 1 further comprising: collecting 1 D 1H nuclear magnetic resonance spectra of the reference compound in the presence of the target molecule at different concentrations of the target molecule; and determining the dissociation constant of the test compound.

7. The method of claim 1 wherein prior to collecting a 1 D 1H nuclear magnetic resonance spectrum of the reference compound in the presence of the target molecule for use in the comparing step, the method comprises: collecting 1 D 1H nuclear magnetic resonance spectra of the reference compound in the presence of the target molecule at different concentrations of the target molecule or at different concentrations of the reference compound; and determining the optimum experimental conditions for identifying at least one test compound that interacts with the target molecule.

8. The method of claim 1 wherein the target molecule is a macromolecule.

9. The method of claim 8 wherein the macromolecule is a polypeptide or polynucleotide.

10. The method of claim 8 wherein the macromolecule is a protein.

11. The method of claim 1 wherein the test compound has a solubility in water of no greater than about 10 micromolar.

12. The method of claim 1 wherein the reference compound binds to the target molecule with a binding affinity in the micromolar range.

13. The method of claim 12 wherein the binding affinity of the reference compound is determined by isothermal titration calorimetry or fluorescence spectroscopy.

14. The method of claim 1 wherein the test compound binds to the target molecule with a binding affinity stronger than 1 micromolar.

15. The method of claim 1 further comprising a step of identifying the reference compound comprising: collecting a WaterLOGSY nuclear magnetic resonance spectrum of a potential reference compound in the absence of the target molecule; collecting a WaterLOGSY nuclear magnetic resonance spectrum of the potential reference compound in the presence of the target molecule; and comparing the WaterLOGSY spectra to identify whether the potential reference compound interacts with the target molecule.

16. The method of claim 1 wherein the test sample comprises a mixture of two or more test compounds.

17. The method of claim 16 further comprising: collecting a 1 D 1H nuclear magnetic resonance spectrum of the reference compound in the presence of a test compound and the target molecule; and comparing the spectrum of the reference compound in the presence of the target molecule to the spectrum of the reference compound in the presence of a test compound and the target molecule to determine a change in the selected reference compound resonance.

18. The method of claim 1 wherein the test compound has a binding affinity tighter than that of the reference compound.

19. The method of claim 1 wherein collecting a 1 D 1H nuclear magnetic resonance spectrum comprises collecting a 1 D 1H selective or multi-selective T2 weighted spectrum.

20. The method of claim 1 wherein collecting a 1 D 1H nuclear magnetic resonance spectrum comprises collecting a 1 D 1H nuclear magnetic resonance selective or multi-selective longitudinal relaxation spectrum.

21. The method of claim 1 wherein collecting a 1 D 1H nuclear magnetic resonance spectrum comprises collecting a selective T2 weighted TOCSY spectrum, or a multi-selective T2 weighted TOCSY spectrum, a selective T2 weighted COSY spectrum, or a multi-selective T2 weighted COSY spectrum.

22. The method of claim 1 wherein: providing a reference compound comprises providing a reference compound and a non-interacting compound; collecting a 1 D 1H nuclear magnetic resonance spectrum of the reference compound in the presence of the target molecule comprises collecting a spectrum of the reference compound and the non-interacting compound in the presence of the target molecule; and collecting a 1 D 1H nuclear magnetic resonance spectrum of the reference compound in the presence of a test sample and the target molecule comprises collecting a spectrum of the reference compound and the non-interacting compound in the presence of a test sample and the target molecule.

23. The method of claim 1 wherein: providing a reference compound comprises providing a reference compound and an ERETIC signal with defined linewidth, amplitude, and frequency; collecting a 1 D 1H nuclear magnetic resonance spectrum of the reference compound in the presence of the target molecule comprises collecting a spectrum of the reference compound with the ERETIC signal in the presence of the target molecule; and collecting a 1 D 1H nuclear magnetic resonance spectrum of the reference compound in the presence of a test sample and the target molecule comprises collecting a spectrum of the reference compound with the ERETIC signal in the presence of a test sample and the target molecule.

24. The method of claim 1 wherein providing a test sample comprises providing a plurality of test samples, each test sample comprising at least one test compound.