Nuclear magnetic resonance (NMR) spectroscopy has been widely adopted since its invention. What was once a cumbersome technique can now reveal the most cryptic details of sophisticated molecular systems. It is one of the most information-rich analytical techniques. The latest machines can place very small samples in a magnetic field with a strength of more than 21 T and detect radio-frequency signals of almost 1 GHz. Systems capable of automated and high-throughput sampling are poised to push NMR into the mainstream, not just as the analytical tool of choice but as a key component of the drug discovery process.

NMR speeds up
Several research teams are working on bringing NMR spectrometers into drug discovery laboratories and using them to further accelerate the rate of pharmaceutical R&D. According to researchers at Varian (Palo Alto, CA), one of the serious drawbacks in getting the best results from a combinatorial array is the inability to obtain a complete sample analysis.

In pioneering work on LC-NMR carried out by Jeremy Nicholson and John Lindon at Imperial College (London), in collaboration with Manfred Spraul of Bruker GmbH, Nicholson’s team separated and assigned a randomly synthesized collection of 27 tripeptides—all the combinations of Ala, Tyr, and Met—using one chromatographic run that took about 30 min (1). In Nicholson’s words, “Not a bad first attempt!” Varian scientists recently extended Nicholson’s research to other areas of combinatorial chemistry by devising an automated approach to NMR that allows combinatorial chemists to quickly and easily obtain the ¹H-NMR spectra of solution-phase samples.

The Varian team worked on obtaining the NMR spectra of compounds bound to solid supports and was rewarded with the rapid adoption of its techniques throughout the combinatorial community. Unfortunately, the teams’ solid-state NMR approach is confined to analyzing small numbers of samples and lacks the high-throughput capability needed for efficient analysis of vast compound libraries. A flow technique coupled with automated sample analysis using liquid-phase NMR would help the analyst rein in combinatorial libraries.

While developing HPLC-NMR techniques, the Varian team realized that the LC-NMR approach could be refined as a useful tool for combinatorial applications. Combinatorial chemistry not only traditionally generates large numbers of compounds in small quantities, but also tends to do away with the use of conventional glassware, replacing it with the increasingly familiar multiple-welled microtiter plates. To address these issues, Varian scientists built and tested a flow-NMR sample changer. “The system reduces the cost, time, and effort of sample handling, allows inexpensive sample containers to be used, and uses smaller quantities of sample than traditional automated NMR systems,” according to Varian.

The flow-NMR approach precludes the need for transferring samples from the microtiter plates to NMR tubes, which would be the biggest cost in high-resolution NMR of a large library, for which not only precision glass tubes and deuterated solvents are required for each sample from each cell, but also a drying (solvent removal) process. Instead, the team at Varian used an automated liquid-handling device, such as the Gilson Model 215 Liquids Handler, which takes a sample solution stored in a microtiter plate and injects it directly into an NMR flow probe.

With each step of the protocol controlled by a computer, the system first rinses the NMR flow cell with a solvent and disposes the waste solvent. The liquid handler then moves a controlled volume of the appropriate sample into the NMR probe, at which point the spectrometer is signaled to begin gathering data. The process can be repeated automatically with any number of NMR experiments on each sample. The team refers to the approach as direct injection (DI) NMR; and the liquid handler is referred to as the versatile automated sample transport (VAST).

Figure 1. Seeing how things develop. Using automated systems developed by companies such as Bruker and Varian, researchers can quickly generate 1H-NMR spectra of compounds synthesized in a 96-well plate. (Adapted from Reference 2.)

The DI VAST approach can quickly gather one-dimensional ¹H-NMR spectra for each member of a combinatorial library, an approach that the team says is almost routine at Varian and elsewhere (Figure 1). For example, at Monsanto (St. Louis) Bruce Hamper and his team used the VAST system to characterize a 96-member substituted methylene malonamic acid library (2).

“This only works in libraries that have one compound per well,” points out Lenore Martin, assistant professor in the department of biochemistry, microbiology, and molecular genetics at the University of Rhode Island. The standard in the industry is to have groups of compounds in each well, so there is still a requirement to couple the flow cell to a separation technique such as LC. “Another very promising technique is capillary electrophoresis (CE)-NMR,” adds Martin, “which is being developed by a group in the department of chemistry at the University of Illinois, Champaign-Urbana.”

Drug design by NMR
NMR is ideal for screening fragments of potential drug molecules, according to the work of Stephen Fesik of Abbott Laboratories (Abbott Park, IL). Recently, he and his colleagues devised a strategy for designing high-affinity ligands to create drugs that inhibit kinases (3). Fesik says that finding leads of sufficient specificity, bioavailability, and safety is “still an arduous process” and usually has a failure rate of 50% in the initial stages of drug discovery. A method to bump up successes without added synthetic effort would be useful. Fesik’s “fragment” approach fits the bill and involves screening a range of fragments that could be incorporated into an inhibitor without reducing potency but improving characteristics, such as solubility or reduced toxicity.

The first step is to fragment an existing lead molecule, identify a range of suitable replacements for the fragments, and build these into the original molecular skeleton. The problems arise in trying to identify suitable fragments. The fragments bind weakly to the target receptor or enzyme, so conventional screening methods cannot reliably detect their binding, because high concentrations are required to generate a detectable response. Moreover, standard assays indicate nothing about binding orientation or site, and so they offer no clues about optimal positioning of the fragment on the skeleton.

Fesik and his colleagues found a way to screen such fragments successfully by using NMR based on a Bruker system. The affinity and binding site location of the chosen fragment are determined by watching how the ¹⁵N–¹H heteronuclear single quantum coherence (HSQC) spectra of the ¹⁵N-labeled protein change when the test molecule is added. The next step involves using NMR to identify molecules that bind to the same site as the chosen fragment. The fragments identified can then be incorporated into the skeleton for further study.

“This approach is a valuable strategy for modifying existing leads to improve their potency, bioavailability, or toxicity profile, and thus represents a useful technique for lead optimization,” says Fesik. Moreover, he emphasizes that the use of NMR in this manner means that thousands of potential mimetics with a range of functionality can be quickly analyzed without the need for multiple synthetic routes to be implemented and thousands of putative leads prepared. Indeed, the Fesik team previously demonstrated high-throughput NMR that could investigate potential ligands for unknown proteins at a rate of 200,000 per month (4).

Toward proteomics
If NMR is going to respond to the postgenomic challenge of addressing thousands of new drug targets, innovations are needed to remove two key limitations. First, NMR structural studies cannot be performed for proteins much larger than 35 kD. Second, to attack thousands of proteins, a proteomically leveraged, highly parallel strategy to drug design is needed; but current strategies attack one target at a time. Triad Therapeutics in San Diego is removing both of these barriers, thus extending NMR drug discovery efforts in a proteome-wide manner.

Triad developed a suite of NMR technologies that allow for the characterization of protein–ligand interactions with unprecedented speed (days as opposed to months). These tools, combined with bioinformatics strategies, allow the systematic gathering of information that describes protein–ligand interactions across large gene families of proteins such as kinases and dehydrogenases. The term “enzyme mechanomics” describes this newly enabled gene-family-wide characterization of structure–function correlations.

“Triad makes use of a technology called NMR SOLVE—structurally oriented library valency engineering—to guide the design of combinatorial libraries tailored to entire gene families of proteins, using the enzyme mechanomic data,” says Daniel Sem, Triad’s vice president of biophysics. He and colleague Maurizio Pellecchia point out that NMR is intrinsically a noninvasive technique and thus is ideally suited to observing the dynamics of a molecular system, as well as acting as an analytical tool.

“Any NMR method that provides structural information on large proteins must provide a way to simplify NMR spectra—to focus in on that part of a spectrum corresponding to atoms that are in a protein’s binding site,” explains Sem. As such, Sem, Pellecchia, and colleagues at the University of Wisconsin have devised a technique that can reduce overlap in protein spectra and allow these complex biomolecules to be investigated in their native state with much greater clarity (5). This method, called solvent-exposed amides with transverse relaxation-optimized spectroscopy (SEA-TROSY), is combined with other experiments to look at very large protein structures, their backbone dynamics, and how ligands or inhibitors bind to them.

“NMR is now poised to tackle the postgenomic challenge of attacking large numbers of new drug targets with greater speed, in a highly parallel manner, and without the usual limitation to low-molecular-weight proteins,” adds Sem.

The metabolic end point
One approach to drug research closely considers the end product of the drug cycle. Jeremy Nicholson uses high-resolution NMR to screen body fluids and magic-angle spinning NMR to screen tissues for metabolic byproducts of drugs and to detect perturbations in endogenous metabolic profiles in disease processes (6, 7). Nicholson and his colleagues have spent the past two decades looking into metabonomics, a field driven mainly by NMR spectroscopy. Nicholson describes metabonomics, a term he coined about six years ago, as the “quantitative measurement of the dynamic multiparametric metabolic response of living systems to pathophysiological stimuli or genetic modification.”

Rather than focusing on single analytes as might be the case in a clinical diagnostics approach, Nicholson’s team has used ¹H-NMR to build up expertise in the multicomponent metabolic composition of cells, tissues, and biological fluids (saliva, blood, urine, semen, and even sweat). The team uses pattern recognition, expert systems, and related bioinformatic tools to interpret and classify the complex data sets generated by one- and two-dimensional NMR analysis of such samples. They can now spot telltale metabolic fingerprints in NMR spectra. NMR, in particular, gives a very complex fingerprint of a large number of metabolite signatures—thousands in the case of a urine sample (Figure 2).

“The quantitative analysis of such profiles gives insight into sites and mechanisms of toxicity according to the characteristic perturbations in the metabolic profile,” explains Nicholson. “Biomarker information can be statistically extracted from spectra and, as NMR is a structural organic chemistry tool, novel metabolic markers can be structurally characterized.

“The recovery of high-density metabolic information from complex spectra is facilitated by the use of an array of multivariate statistical and pattern recognition tools that classify toxicity or disease state according to spectral profile and identify critical regions of the NMR spectral fingerprints that are modified by the pathological process,” says Nicholson. Exact biomarker identification is then achieved or confirmed by judicious use of multidimensional NMR spectroscopy (e.g., ¹H-¹³C HSQC or heteronuclear multiple-bond correlation spectroscopy) combined with HPLC–NMR–mass spectrometry methods (8).

A holistic picture
The London team also recently introduced the concept of “integrated metabonomics”. This, Nicholson says, is the parallel NMR investigation of multiple biological fluids, and sometimes selected tissue samples, using magic-angle spinning NMR methods at various time points after drug exposure to gain a holistic picture of a series of metabolic events in the whole body.

Nicholson and his colleagues are now involved in cross-correlating integrated metabonomic data with those generated by genomics and proteomics (what he terms “integrated bionomics”) to describe the biochemical consequences of pathological processes at multiple levels of biomolecular organization and to learn about silent gene function.

From humble beginnings as a simple spectroscopic tool for working out molecular structures, NMR has raced to the front of the drug discovery arsenal, providing pharma researchers with a powerful weapon with which to hack through the molecular jungle.

References

1. Lindon, J. C.; et al. Magn. Reson. Chem. 1995, 33, 857–863.
2. Hamper, B. C.; et al. J. Comb. Chem. 1999, 1, 140–150.
3. Hadjuk, P. J.; et al. J. Med. Chem. 2000, 43, 4781–4786.
4. Hadjuk, P. J.; et al. J. Med. Chem. 1999, 42, 2315–2317.
5. Pellechia, M.; et al. J. Am. Chem. Soc. 2001, 123, 4633–4634.
6. Nicholson, J. K.; Lindon, J. C.; Holmes, E. Xenobiotica 1999, 29, 1181–1189.
7. Lindon, J. C.; et al. Concepts Magn. Reson. 2000, 12, 289–320.
8. Lindon, J. C.; Holmes, E.; Nicholson, J. K. Prog. Nucl. Magn. Reson. Spectrosc. 2001, 39, 1–40.
9. Holmes, E.; et al. Chem. Res. Toxicol. 2000, 13, 471–478.

David Bradley is a freelance writer living in Cambridge, UK. Send your comments or questions regarding this article to mdd@acs.org or the Editorial Office by fax at 202-776-8166 or by post at 1155 16th Street, NW; Washington, DC 20036.