Genomics and Proteomics - NMR Is Ready, so Bring on the Macromolecules

Genomics & Proteomics
Reed Business Information
Morris Plains, NJ, 07950

NMR Is Ready, so Bring on the Macromolecules

Nuclear magnetic resonance now rivals X-ray crystallography for protein and RNA structure studies. NMR's capabilities enable the investigation of dynamic properties of molecules, particularly protein-small molecule interactions, that cannot be observed by other methods. By Angelo DePalma, PhD

If the marriage of computing, miniaturization, and advanced electronics has benefited analytical instrumentation in general, nuclear magnetic resonance (NMR) spectroscopy has profited doubly. Although falling somewhat short of exponential advances in capability, such as those predicted for computer power by Moore's law, NMR's growing application base has in other ways surpassed expectations. Once firmly ensconced within the world of organic small molecules, NMR now serves life science explorations into macromolecular structure and dynamics.

Four-dimensional (4D) HCCH-TOCSY NMR spectrum of 13C, 15N labeled ubiquitin. The color bar encodes the 4D chemical shift. The experiment correlates protein signals from protons bonded to 13C atoms that are bonded to other 13C atoms and then another 1H. The experiment was collected with 15N decoupling. The chemical shift axes are x = 1H. y = 1H, z = 13C, and a = 13C (a is the color bar). The colored signals are NMR peaks in 3D space. The center position and color of each peak represent the chemical shift coordinates of the 1H or 13C NMR signals. Size of the peak represents the intensity of the NMR signal. This information is used to help determine the connectivity of the atoms in the protein and ultimately the 3D protein structure. (Source: JEOL)

Anyone acquainted with NMR recognizes the complexity of spectra with increasing carbon number. Proteins, with hundreds of carbon and thousands of non-exchangeable proton resonances, are immeasurably more difficult to parse than most small organic molecules.

In the early days of protein NMR, a single structure took a year or more to deduce [reminiscent of the early days of X-ray crystallography (XRC)], and protein size was severely limited to a few kilodaltons (kDa). By the late 1990s, structures could be solved much more rapidly, in as few as three to four months. Today, some proteins are solved in a few days, but many still take several weeks, says Iain Green, PhD, senior manager of product marketing for NMR systems, Varian Inc., Palo Alto, Calif.

As with so many life science research initiatives, NMR received a huge boost from the Human Genome Project. "There was a big push to solve protein structures," says Gaetano Montelione, PhD, professor at the Center for Advanced Biotechnology, Rutgers University, New Brunswick, N.J. "Large numbers of genes and proteins were being discovered and nobody knew what they did or [what they] looked like."

As a tool for protein structure elucidation, NMR is still in its infancy compared with XRC. Just 15% of the three-dimensional (3D) protein structures on deposit at the Research Collaboratory for Structural Biology (RCSB) Protein Data Bank were acquired by NMR.

NMR and XRC are often viewed as competing technologies, but they are actually complementary, says Montelione. "Proteins that do not crystallize or do so only after weeks of effort are good candidates for NMR structure determination, provided the molecular weight is below about 40 kD," he says.

NMR is for Genes, Too
While NMR structure studies on DNA have been described as “trivial,” the same isn’t true for RNA. Like proteins, RNA molecules possess secondary and tertiary structure that enable them to behave in some instances like genes and in others like proteins.
RNA molecules are difficult to crystallize, so NMR becomes the technique of last resort for obtaining 3D structures. The bad news is that like proteins, RNA molecules tend to be large and appear to suffer from upper size analysis barriers comparable to those of proteins.
Mirko Hennig, PhD, at the Scripps Research Institute, La Jolla, Calif., investigates RNA and RNA-protein complex structures in solution by NMR, using residual dipolar couplings and cross-correlated relaxation rate experiments. Hennig is particularly interested in the therapeutic potential of RNA. One project focuses on Rev-RRE, a small RNA-binding protein which mediates export of mRNA from HIV. With Shana Kelley, Boston College, Hennig investigates structural fragility of human mitochondrial tRNA-Ile, which is implicated in disease-related mutations, using multi-dimensional NMR.
Arthur Pardi, PhD, at the University of Colorado, Boulder, Colo., studies RNA enzymes (ribozymes) through NMR, particularly the well-characterized hammerhead ribozyme. Structures are obtained mostly through proton-proton experiments, but Pardi has isotopically labeled RNAs with 13C and/or 15N for 2D and 3D hetronuclear structure studies on various RNAs, particularly in vitro-selected RNA. Selected RNA shows high affinity for proteins and small molecule ligands and can discriminate between small molecules up to 100-fold better than monoclonal antibodies.
Hashim Al-Hashimi, PhD, at the University of Michigan, Ann Arbor, Mich., uses solution-state NMR to investigate RNA function during gene expression and virion functioning. This work centers on excited-state RNA, which differs structurally from ground-state molecules normally studied. Al-Hashimi therefore examines RNA as a function of time and other reaction coordinates using isotopically-labeled RNA.

When generation of complete 3D structures are impossible due to a protein's size, scientists can still tweak binding information using TROSY (transverse relaxation optimized spectroscopy). Invented about three years ago by Kurt Wüthrich, PhD, professor of biophysics at Eidgenössische Technische Hochschule Zürich (ETH), Switzerland, TROSY probes dynamic structure-related effects (but not 3D structures) of proteins with a size up to about 900 kDa.

For proteins that neither crystallize nor dissolve, solid-phase NMR may be the answer. In the past, solid-phase spectra suffered from unacceptable peak broadening. The combination of new experimental techniques—tools such as BioSolids probes from Bruker BioSpin Corp., Billerica, Mass.; digital electronics with rapid phase- and frequency-switching; and availability of fields up to 900 MHz—have made solid-state NMR practical. Other techniques, in which molecules are constrained by gels or scaffolds at specific angles and spun very quickly to simulate solution behavior, can also uncover interatomic connectivity and distance in proteins.

Not your daddy's NMR
Detailed structural NMR work would be impossible without multidimensional experiments. Two-dimensional spectra have been used for at least 25 years on organic molecules, yet proteins require 3D and even higher-dimensional techniques. NMR protein structures are obtained in three steps. A unique NMR resonance is assigned to each relevant atom, then nuclear distances are calculated. Finally, the 3D structure is assembled by molecular modeling software. Assigning individual resonances is complicated by overlapping signals from the large number of atoms in similar chemical environments. Higher fields resolve overlaps somewhat, but only multidimensional techniques can differentiate atoms with nearly identical chemical shifts.

Two experiments are particularly applicable. HNCa, a 3D experiment, transfers magnetization from an amide proton to the nitrogen bearing it (first dimension), then to an alpha carbon (second dimension), and back to the amide proton (third dimension). The related HNCaCo experiment transfers magnetization from amide proton to nitrogen, alpha carbon, carbonyl carbon, and back to the proton. Backbone and sidechain resonances are assigned through these experiments or similar ones.

Additional techniques detect nearby spins separated through space by up to 0.6 nm. These spatial proximities lead to a set of constraints which, with the primary structure, allows modeling software to reconstruct the 3D representation.

Almost all atom-proximity methods involve the nuclear Overhauser effect (NOE), which measures interaction between two nuclei and falls off the sixth power of the internuclear distance. During an NOE experiment, spectroscopists perturb the resonance of one proton while observing effects of that perturbation on neighboring protons. NOE protein investigations require instruments with fields of 400 MHz and higher.

click the image to enlarge
3D structure of an E. coli protein superimposed on an 800 MHz NMR spectrum NMR spectrum was performed by Lewis Kay,University of Toronto. (Source: Varian Inc.)

"Interpreting protein NOEs is computationally demanding," says James Prestegard, PhD, professor of chemistry and biochemistry, University of Georgia, Athens, Ga. In his work, Prestegard relies not only on proton NOEs but on 15N-1H interactions that yield information on the orientation of backbone structures that may be remote in space.

Still other information may be obtained from analysis of other NMR-active nuclei, such as 31P and 13C, which underscores the necessity for multiple-nuclei (channel) capability. "Anything less than four channels is inappropriate for state-of-the-art NMR experiments," says J. Douglas Meinhart, PhD, national laboratory manager at JEOL USA Inc., Peabody, Mass.

Knowing every interatomic distance for every pair of atoms reduces 3D structure calculations to a trivial (for a computer) distance matrix calculation. But because only proton-proton NOEs are useful, and these are not very accurate, spectroscopists must resort to a bit of legerdemain. "The solution is to use a redundant network of proton-proton distance constraints, calculate the structure many times, and superimpose those structures to obtain a statistically reliable overlap," says Rutgers' Montelione.

Spectroscopists use other tools to obtain clearer spectra. Lower sample temperatures slow internal molecular dynamics, making certain interactions easier to observe. Internuclear interactions such as J-couplings provide valuable structural data; residual dipolar couplings (arising from bonded-atom interactions) provide information about relative orientations of chemical bonds.

The frontier of structural NMR work involves higher-order experiments. Complex spectra are resolved through 4D and even 5D methods. G-matrix Fourier Transform NMR generates 4D and 5D spectra in better-resolved (and comprehended) 2D and 3D representations, and are acquired more rapidly than 4D or 5D spectra. These new methods, together with automated software analysis, reduce structure elucidation times to several days, if all goes right.

Dynamic capabilities

Academic NMR Groups to Watch
While NMR is hot in academic labs, drug developers are lagging behind. Cost is one reason, as very high-field instruments can cost $5 million. The principal reason is probably the novelty of high-quality NMR/protein work and the relative scarcity of NMR protein structure expertise. The bulk of protein/NMR expertise currently resides at universities. Some noteworthy groups working with industry include:
• The NMR Center at the University of California, San Francisco. An interdisciplinary (chemistry, biology, biophysics, bioengineering) that includes Thomas James, PhD, chair, department of pharmaceutical chemistry, who uses multidimensional NMR to study dynamic structures of proteins and nucleic acids, small molecule-macromolecule interactions, and RNA/protein systems.
• David Wemmer, PhD, at the University of California, Berkeley, heads the NMR effort at the Berkeley Structural Genomics Center. Wemmer’s principal areas of interest include multinuclear, multidimensional NMR to screen proteins for folding, aggregation state, presence of flexible tails, and biochemical activity in addition to full structure determination for smaller (The University of Florida’s and Florida State University’s NMR Spectroscopy and Imaging Program at the University of Florida maintains a number of high field instruments dedicated to both solution and solid-state NMR). The program’s researchers include Timothy Logan, PhD, who studies the relationship between protein structure/dynamics and biological activity.
• The Scripps Institute, La Jolla, Calif. In addition to sharing nobelist Kurt Wüthrich with Zurich’s ETH, Scripps faculty include H. Jane Dyson, PhD, who uses NMR to study proteins, and Peter Wright, PhD, who uses multi-dimensional hetero-nuclear NMR to study protein and enzyme dynamics, protein folding, and molecular recognition.
• RIKEN, Japan’s seven-center institute for advanced technologies includes a protein structure effort. RIKEN is believed to possess the highest concentration of high-field NMR instruments in the world.
• The Northeast Structural Genomics Consortium (NESG), which comprises eight universities (including Rutgers) and Pacific Northwest Laboratory, relies on both NMR and XRC. NESG’s approach to protein structures is based on homology among proteins within a particular class or family.

NMR's big advantage over X-ray crystallography lies in NMR's dynamic capabilities over broad time scales, says Gordon Rule, PhD, professor, Department of Biological Sciences, Carnegie Mellon University. Fluorescence, the more commonly-used technique for studying protein dynamics, measures nanosecond-scale events, whereas NMR widens the observation window from milliseconds to picoseconds.

Resonance frequency of nuclear spins is extremely sensitive to the immediate chemical environment, which makes NMR suitable for studying small-molecule binding for drug development. Shifts in resonance frequencies for either the protein or small molecule are an indicator of binding. Protein-drug interactions are also possible through XRC, but subtle effects that occur in a protein's natural milieu are lost in the solid phase. "X-ray crystallography provides structure, but molecules are frozen in time-space," says Varian's Green. "NMR examines structure, function, and dynamics in an environment that better mimics a protein's natural environment."

"Changes in the protein's or drug's NMR spectrum indicate that an association is occurring, but it doesn't tell you a priori where the drug binds or whether the drug will be an inhibitor," says Rule. "But, it's a good start. More detailed NMR studies can pinpoint the drug's binding site and measure protein structural changes that result."

This strategy has been exploited by Jeremy Nicholson, PhD, professor of biological chemistry, Imperial College, London. Nicholson, who is an active NMR researcher and a founder of startup firm Metabometrix Ltd., London, uses NMR for "metabometric" studies of protein-small molecule interactions. Metabometrics is defined as the quantitative, dynamic response of living systems to stimulation. Metabometrics has applications in drug discovery, medical diagnostics, and broader genome-phenotype studies.

Proton magnetism
Each succeeding level of NMR field strength requires new types of magnet materials. For example, niobium-titanium magnets worked well at fields up to about 360 MHz, whereas niobium-tin works up to 920 MHz or so.

NMR employs superconducting magnets with stringent homogeneity and stability requirements. Resolution required to distinguish between protons in slightly different chemical environments may be less than 0.1 Hz maintained over entire sample length, at a magnet field strength corresponding to a frequency of, say, 700 MHz. Homogeneity must therefore be better than 1 in 7 billion Hz. Stability of magnet and electronics must last for the length of the experiment, which is often many hours.

Progress in magnet technology and achievable NMR field strength are linked to the availability of suitable superconductor materials. Today's spectrometers primarily use low-temperature metallic superconductors. Lowering operating temperature to 2 K (through its UltraStabilized technology) enabled Bruker BioSpin to reach a field strength of 900 MHz in 2000. Bruker officials say that the company has thus far sold 10 such magnets.

Because magnetic fields quench current flow through present-day superconducting magnetic coils, existing magnets possess an inherent limitation. Future gigahertz-strength fields will require high-temperature superconducting (HTS) ceramic wires which offer higher current-carrying capability and possess lower dependence of critical current on magnetic field strength. Because NMR magnets typically employ many miles of superconducting wire, the challenge is to fabricate brittle HTS materials in long enough sections so that they can be used in NMR magnets.

Werner Maas, PhD, vice president of research and development at Bruker, is optimistic that even higher-field magnets will emerge from HTS work. "I expect that we will see a 1 GHz magnet in the next three years or so."

Sensitivity training
In the early 1970s, a 90 MHz NMR system had a signal-to-noise ratio (S/N) of about 16:1. The introduction of cryogenically cooled probes, such as Bruker's original CryoProbe, was a major upgrade. In these devices, radio-frequency coils and electronic detection circuitry are cooled to around 20 K, which reduces noise and thus provides a three- to four-fold improvement in S/N. Today, a 900 MHz instrument with a CryoProbe exceeds a S/N of 8000:1. Higher S/N, with higher magnetic field strengths, advanced NMR probe technologies, and better NMR electronics, have largely been responsible for improvements in NMR sensitivity.

"It takes roughly 100 times as long to obtain the same quality spectrum using a conventional probe at 500 MHz compared with a CryoProbe at 800 MHz," says Montelione. For the same field strength, the enhancement is "only" 10-fold.

NMR probes are crucial to analytical sensitivity, because these devices detect weak signals from nuclear spins. NMR probe technology has progressed gradually, yielding better signal-to-noise ratios through improved components and detection schemes. "CryoProbes caused a paradigm shift for NMR," said Maas, whose company enjoys an installed base of more than 350 CryoProbes.

To date, NMR's progress has been more evolutionary than revolutionary. But with 1 GHz (and higher) instruments on the way, and new pulse sequences under development at academic labs, NMR can only improve its status among protein-characterizing techniques.

Angelo DePalma is a freelance writer based in Newton, N.J.