Magnetic resonance spectroscopy plays a growing role in pharmaceuticals research.

Thirty-five years have gone by since NMR (nuclear magnetic resonance) spectroscopy first hit the analytical scene, and yet its capabilities and applications continue to evolve. Originally designed as a way to verify the structure of relatively small compounds, the technology of NMR has exploded and become a valuable means for studying protein structure. Now NMR has entered new arenas—drug discovery and structural genomics. With these latest advances and applications, NMR is poised to assume a greater role in the pharmaceutical and biotech industries.

Structure–Function Studies of Proteins
Traditionally, X-ray crystallography has been used for solving the structure of proteins; however, it is useful only for those that can be coaxed into a crystalline state. The development of multidimensional NMR and more powerful instruments (e.g., the 500-MHz NMR) opened the door for solving the structure of proteins and peptides in an aqueous environment, as they exist in biological systems. NMR allows you to observe the physical flexibility of proteins and the dynamics of their interactions with other molecules—a huge advantage when studying a protein’s biochemical function.

By viewing the three-dimensional (3-D) structure of a protein on a computer screen, you can see how ligand fits into the protein’s active site, much like the pieces of a jigsaw puzzle. The 3-D structure of a protein provides information about its biochemical function, including the catalysis of reactions and binding of molecules, such as DNA, RNA, and other proteins. The correlation between structure and biological function is known as the structure–activity relationship (SAR).

To observe ligand binding in solution, the two-dimensional (2-D) NMR spectrum of ligand-free protein is compared with that of ligand-bound protein. When ligand binds to a protein that has been either selectively or uniformly isotopically labeled, the signals corresponding to the protein may move, broaden, or disappear. Measuring changes in chemical shifts and hydrogen exchange rates serves to elucidate the dynamics of ligand binding and determine critical residues for ligand–protein interactions (1).

In drug design, ligand binding by a target protein induces the ultimate effect of interest, such as cell growth or cell death. By studying the structure of a target protein of a particular disease and learning something about its active or ligand-binding sites, you should be able to design inhibitors or activators to fit the architectural structure and chemical nature of that site and manipulate the elicited response. One approach for discovering ligands that bind to proteins has been recently described by Stephen Fesik and his colleagues at Abbott Laboratories and referred to as “structure–activity relationships by nuclear magnetic resonance” (SAR by NMR; 2).

Only the initial experiments of SAR by NMR are exploratory in nature: screening for weakly bound, low molecular weight inhibitors of the target protein (the dissociation constant, a measure of a drug’s potency, is in the millimolar range). The Abbott group uses ¹⁵N- and ¹H-2-D heteronuclear single quantum correlation (2-D HSQC) NMR to test the binding of these compounds to ¹⁵N-labeled protein (Figure 1). Unlike many other assays, this type of NMR-based screening of ligand–protein complexes can detect weakly bound molecules. In addition, NMR locates the binding site for ligand on a protein’s surface. Comparing the structures of compounds that bind to the same site on a protein provides information about the functional groups involved in ligand binding and can guide the synthesis of lead compounds by combinatorial chemistry.

After finding small molecules that bind at distinctly different yet proximal sites of the protein, the two molecules are linked together in the proper orientation to produce a ligand that binds tightly to the protein, and thus is more potent. The distance between the two binding sites and the structural information of the protein obtained from the NMR experiments define the design of the link between the two inhibitors.

“Using SAR by NMR, novel lead compounds are constructed that do not exist in corporate libraries and cannot be found using conventional methods,” says Fesik, and he’s proven it. He and his colleagues found inhibitors for both stromelysin, an important enzyme for tissue repair, and the DNA-binding domain of the human papillomavirus E2 protein, a molecule required for viral replication (3, 4). Although conventional high-throughput screening assays of large libraries of compounds (on the order of 100,000) did not identify inhibitors with potencies better than the micromolar range, SAR by NMR produced inhibitors with dissociation constants in the low nanomolar range. Arthritis and tumor metastases have been associated with the overexpression and unregulated activity of stromelysin and other matrix metalloproteinases, lending to their importance as a target for drug design. Because the E2 protein regulates gene transcription, compounds that inhibit its binding to DNA represent potential leads as antiviral agents against the human papillomavirus–DNA tumor viruses capable of infecting a wide variety of mammalian cells.

Not only has SAR by NMR proven successful where other routes have failed, but it also avoids the cost and time associated with synthesizing large numbers of complex molecules.

Enhancing HTS Assays
Companies and research institutes often generate thousands of compounds using combinatorial chemistry, in hopes of finding one lead compound. Whether the compounds are just off the shelf, purchased, or produced by synthetic means, all of these potential candidates must be screened. The development of high-throughput screening assays (HTS) gave scientists the ability to test large numbers of molecules for a desired biochemical activity, including binding and enzymatic catalysis. Designing such a specific assay isn’t always trivial; the assays can be complicated and may require several components that can interfere in the interaction of the drug candidate and the target protein. These assays also lend themselves to other problems such as low sensitivity, in the case of a weak signal or high background, and false positives.

The simplistic nature of SAR by NMR makes it ideal for implementing as an assay—requiring only the target protein and the test compound—and the assay detects differences in the ligand-binding sites of the target protein. Unfortunately, because of its low sensitivity, NMR generally requires high concentrations of protein, and with increased concentrations of protein and lead compounds, you normally face a solubility problem. In addition, you can screen only a few lead compounds at a time.

Fesik’s group at Abbott has gotten around the concentration problem by increasing the number of compounds screened in a given assay. As a result, they can decrease the concentration of the small molecule and the protein and maximize the data collection from one assay. They also avoid the longer acquisition times typically associated with low protein concentrations by increasing the mixture sizes, and they overcome low-sensitivity problems by using cryogenic NMR technology, or cryoprobes. “Cryoprobes offer a dramatic increase in sensitivity, which allows larger libraries to be rapidly screened using smaller amounts of protein,” says Fesik.

Cryogenic NMR technology, the helium-cooling of the preamplifier and radio frequency coils of the probe to about 20 K, enhances high-resolution NMR and increases the signal-to-noise ratio about fourfold. With cryoprobes, screening 100 different compounds at concentrations as low as 50 µM results in a satisfactory concentration of 5 mM and increases the data output of the assay by 10-fold when compared to the usual SAR by NMR assay.

The Abbott group used NMR-based HTS assays to find novel, weak inhibitors to the Erm family of methyltransferases, which are ultimately responsible for the erythromycin resistance in bacteria. These enzymes methylate RNA, preventing antibiotics from being able to bind the RNA and carry out their antibacterial properties (5, 6).

Two-Dimensional 15N-Heteronuclear Single-Quantum Correlation–NMR Spectroscopy When a molecule binds a protein, the chemical environment of the protein’s binding site changes and results in the chemical shift perturbation of nuclei at that site. Two-dimensional 1H -15N heteronuclear single-quantum correlation (HSQC)-NMR spectroscopy screens for ligand binding by detecting only the amide signals of 15N-labeled protein. The chemical shifts of 1H and 15N of each amide of the protein backbone are presented as a contour map. Because there is only one amide proton per amino acid, each HSQC signal represents one single amino acid. The spectrum of HSQC does not have a diagonal like a homonuclear spectrum, because different nuclei (1H and 15N) are observed during t1 and t2 of the pulse sequence. Comparing the 2-D 1H-15N HSQC spectra of 15N-labeled target protein in the absence and presence of ligand provides information about protein–ligand interactions. First, the chemical shifts of the spectrum move if ligand binds, and second, those shifts that move correlate to the amino acid residues of the ligand-binding site. The specificity of this type of NMR eliminates interference from all other components of the assay.

The Next Step: Structural Genomics
While massive amounts of genetic sequence spew out of the human genome project, scientists confront the daunting task of finding ways to manage and interpret all of these data. With support from the pharmaceutical industry and government, Gaetano Montelione and Stephen Anderson at Rutgers University are spearheading the New Jersey Commission on Science and Technology Initiative in Structural Genomics and Bioinformatics, which will bring protein NMR to the forefront of structural genomics and drug discovery (7). By developing technology that connects gene sequence to protein function via structure determination, predicting the biochemical function and structure of novel gene products will become possible. “Structure determination provides a way of getting clues to biochemical function, which can then be tested with specific assays,” says Montelione, “and you get the added bonus of having the structure.”

Montelione says that part of the project entails determining resonance assignments and 3-D structures for many proteins. These assignment structure databases will then be used to facilitate the structure determinations of homologous proteins—those proteins believed to share a common ancestor based on the similarity of their amino acid sequence.

Because the information technology exists, Montelione also envisions that scientists will systematically organize all of the genetic data from the human genome project into families of genes, based on their sequence similarity. A representative protein with a known biochemical function from that family will be expressed, and its structure solved. The structure of this representative protein serves as a model from which the structures of all of the other proteins of that gene family can be determined. By applying structure-based functional genomic analysis, the biochemical functions of novel gene products can then be predicted.

Why might this methodology work? Because genetic variation occurs so spontaneously and frequently, you’re likely to see many differences among proteins from the same family. However, “Structure [similarity] is more well preserved over evolution compared to genetic sequence,” explains Montelione, “. . . and by seeing the three-dimensional structure of the protein, you can gain testable hypotheses regarding possible biochemical functions.” This initiative also proposes solving the structures of all of the proteins of a particular organism, so that the protein–protein interactions can be simulated on an atomic level. Montelione admits that this is “a very ambitious goal” but makes the point that “we are developing the technology to rapidly obtain structures of protein” as techniques are being developed in X-ray crystallography and NMR spectroscopy to speed up the data collection and structure determination process.

Of course, the initiative has its application to drug discovery. Genomic sequence information provides a means for identifying gene products involved in human disease or unique to specific pathogens. Because these gene products constitute possible protein therapeutics or targets for drug design, Montelione wants to use structural genomics to “choose genetic targets without knowing the protein function or what it [the protein] is, then determine the structure of the protein, and later determine the function.” Montelione says, “We now have thousands and thousands of targets, because we have all of these genes. . . . The next step of the genome project is to determine the structures and function of the corresponding proteins.”

References

(1) Stockman, B. J. Prog. Nucl. Magn. Reson. Spectrosc. 1998, 33, 109–151.

(2) Shuker, S. B. et al. Science 1996, 274, 1531–1534.

(3) Hajduk, P. J. et al. J. Am. Chem. Soc. 1997, 119, 5818–5827.

(4) Hajduk, P. J. et al. J. Med. Chem. 1997, 40, 3144–3150.

(5) Hajduk, P. J. et al. J. Med. Chem. 1999, 42, 2315–2317.

(6) Hajduk, P. J. et al. J. Med. Chem. 1999, 42, 3852–3859.

(7) Montelione, G. T.; Anderson, S. Nature Struct. Biol. 1999, 6, 11–12.

Jennifer B. Miller is an assistant editor of Today’s Chemist at Work. Comments and questions for the author may be addressed to the Editorial Office.