A murder-mystery villain, if she were resourceful and patient enough, could bring about her victim’s demise by replacing all his drinking water with heavy, or deuterated, water. That’s because D2O, although superficially similar to H2O, behaves differently enough in enzymatic reactions that in large quantities it’s detrimental to living tissues. Rats die within a week when given nothing but D2O to drink.
Escherichia coli bacteria, on the other hand, can survive in D2O. That’s fortunate for protein researchers, who genetically engineer the bacteria to manufacture deuterated proteins to use in nuclear magnetic resonance spectroscopy: Replacing most of a protein’s abundant magnetic hydrogen-1 atoms with deuterium, which is invisible to NMR tuned to 1H, helps to simplify the spectrum. To ensure uniform isotopic substitution, researchers first use bacteria to create entirely deuterated proteins, then back-substitute some of the D atoms with 1H at predetermined chemical locations.
Although D2O doesn’t kill the bacteria, it slows their metabolism, and they produce protein only in minute amounts. Furthermore, the back-protonation step isn’t easy. As a result, partially deuterated samples are difficult and expensive to make—and for many proteins, not even possible.
Now Guido Pintacuda and colleagues at the European Center for High-Field NMR in Lyon, France, have used 1H NMR to find the structures of two nondeuterated proteins, including a virus coat protein whose structure was previously unknown.1 To narrow and separate the spectral peaks, the researchers were aided by some of the most advanced equipment in the world, including an NMR spectrometer with a magnetic field of more than 23 T. More importantly, they took a decades-old technique called magic-angle spinning (MAS) and boosted its effectiveness by rotating their samples at speeds of more than 100 kHz. Such high spinning frequencies became possible only a few years ago.
Magnetic dynamics
NMR is widely used to elucidate the structures of molecules both large and small. When placed in a magnetic field, a molecule’s magnetic nuclei precess at a field-dependent frequency. But the local field felt by each nucleus isn’t exactly equal to the applied field: It’s slightly modified, for example, by the swirling cloud of electrons surrounding the nucleus and by proximity to other magnetic nuclei. The spectrum of those deviations, called chemical shifts and measured in parts per million, provides information about the specific environment of each nucleus and thus about the molecule’s structure.
For a small, simple molecule, a one-dimensional spectrum of chemical shifts may contain enough information to determine the structure. But for larger molecules, researchers typically require multidimensional spectra: Applying specially designed sequences of RF pulses, they probe spin excitations that coherently hop from one atom to another, with each chemical shift recorded on a separate, orthogonal axis. Such a spectrum offers invaluable clues about which of a molecule’s atoms—or which of a protein’s amino acids—are spatially close enough to transfer a spin excitation between them.
But that’s not all. The pulse sequences for multidimensional NMR spectra can be tailored to probe and follow changes in the molecule on time scales of picoseconds to milliseconds—not just changes in spin states, but also movements of atoms. That capability gives NMR one of its biggest advantages over other techniques for investigating protein structures. X-ray crystallography and cryoelectron microscopy, for example, can resolve structures with exquisite resolution. (For more on cryo-EM, see Physics Today, August 2016, page 13.) But both techniques are limited to molecules in static environments—immobilized in a crystal or frozen in a sheet of ice. NMR can probe not just a protein’s structure but also the dynamic processes that are central to its biological function.
Conventional solution-phase NMR relies on molecules’ rapid random tumbling to average out anisotropic dipolar couplings between nuclei. But that averaging doesn’t work for solids and solid-like samples, such as viscous, sticky agglomerates of bulky biomolecules. In those cases, each spectral line is broadened into a thicket of unresolved lines arising from molecules with different orientations. Fortunately, researchers can recover the benefits of rapid tumbling by spinning the sample about an axis oriented at a particular “magic” angle, 54.7°, with respect to the field. At that angle, nuclear couplings fall to zero. (See the article by Clare Grey and Robert Tycko, Physics Today, September 2009, page 44.)
The higher the spinning frequency, the more effective MAS is at narrowing and resolving spectral lines, but frequency is limited by the size of the cylindrical rotor that houses the sample. Spinning a too-large rotor too quickly can destroy the sample through frictional heating, destabilize the gas-lubricated rotation, or even disintegrate the rotor itself. Figure 1 shows some examples of rotors and their maximum spinning speeds. The smallest one, at 0.7 mm in diameter, was designed for the new experiments by Pintacuda’s collaborators at Bruker Biospin in Germany.
Figure 1. In magic-angle spinning NMR, a sample is loaded into a cylindrical rotor that is spun rapidly to mimic the effects of molecules tumbling in a fluid. As shown here, the maximum spinning frequency depends on the rotor diameter. The smallest rotor, 0.7 mm in diameter, was recently debuted. (Courtesy of Guido Pintacuda.)
Figure 1. In magic-angle spinning NMR, a sample is loaded into a cylindrical rotor that is spun rapidly to mimic the effects of molecules tumbling in a fluid. As shown here, the maximum spinning frequency depends on the rotor diameter. The smallest rotor, 0.7 mm in diameter, was recently debuted. (Courtesy of Guido Pintacuda.)
The march to higher MAS frequencies has been a decades-long technical challenge. Building and operating a smaller rotor requires miniaturizing everything from the cooling system to the seals that hold the container closed. It also entails inserting a gooey protein sample into a container the size of a mechanical pencil lead. A further challenge: Shrinking the sample itself, all else being equal, lowers the measurement sensitivity.
Isotope exchange
Although the spin precession of any magnetic isotope can be the basis for an NMR spectrum, solution-phase experiments are most often tuned to detect 1H. Its large gyromagnetic ratio leads to strong signals, and its high natural abundance means that samples can be prepared without any special isotopic labeling. But in solid-state NMR, dipole–dipole coupling between 1H nuclei is a major source of anisotropic line broadening that MAS at modest frequencies can’t adequately reverse.
As a result, nearly all solid-state NMR experiments were, until recently, based on isotopes other than 1H; for proteins and other biomolecules, that usually meant carbon-13 and nitrogen-15. Because of the low natural abundances of those isotopes, samples had to be isotopically enriched. The process is similar to deuteration: Insert the gene for the protein into bacteria that are fed with isotopically enriched nutrients. Fortunately, bacterial metabolism isn’t slowed by 13C and 15N the way it is by D2O.
Both 13C and 15N have low gyromagnetic ratios. To obtain reasonable signals from those nuclei, researchers needed relatively large samples, which limited MAS frequencies. For example, in 2002 when Hartmut Oschkinat and colleagues obtained the first complete structure of a protein2 from 13C and 15N MAS NMR, they used several protein samples of 6–10 mg each, packed into 4-mm-diameter rotors spun at just 8–13 kHz. Although faster spinning is possible with a rotor that size, the need for many-milligram samples was thought to place an effective cap on the frequencies that would ever be feasible.
Around the same time, several groups were experimenting with 1H NMR on heavily deuterated samples; in 2007 Chad Rienstra and colleagues used the method to resolve the structure of GB1, a model protein often used to test solid-state NMR techniques.3 The switch from 13C and 15N to 1H boosted the signal by almost an order of magnitude and motivated the development of smaller rotors and higher MAS frequencies. Much of the progress was spearheaded by Ago Samoson, of the Tallinn University of Technology in Estonia, and his colleagues, who crossed the 100 kHz threshold with a protein sample in 2012.
In 2014 Samoson collaborated with Matthias Ernst, Beat Meier (both at ETH Zürich), and Anja Böckmann (at the University of Lyon) to resolve a protein structure with 100-kHz MAS NMR from two samples of just 0.5 mg mass, each placed in a 0.8 mm rotor.4 Pintacuda’s new work, in comparison, used a single 0.5 mg sample of each protein.
The expense of growing bacteria in D2O is less daunting when all that’s needed is a milligram of protein, but the challenge of back-protonation remains a limitation. Each of a protein’s amino acids contains one H site—the so-called amide H, bound to the N atom that puts the “amino” in “amino acid”—that’s more easily displaced than the rest. When a deuterated protein comes in contact with H2O, amide D atoms readily change places with H atoms in the water, while the rest of the protein’s D atoms remain in place.
But that process can swap all the amide D atoms only if the protein is not yet folded, or if it’s prone to quick, stochastic folding and unfolding. For proteins that fold tightly and remain folded, the only amide D atoms that are exchanged are the ones near the surface of the folded structure. Large parts of the protein remain invisible to the NMR study. And the class of tightly folded proteins includes many of particular medical interest, including membrane proteins (which are often the targets of drugs) and virus capsids. Misfolded proteins known as amyloid fibrils, which are associated with Alzheimer’s and Parkinson’s diseases (see the article by Tuomas Knowles, Michele Vendruscolo, and Christopher Dobson, Physics Today, March 2015, page 36), are also best addressed by solid-state NMR in fully protonated form.
Virus frontiers
Pintacuda and colleagues’ initial plan was for a proof-of-principle study of the fully protonated form of GB1, the model protein used by Rienstra and colleagues. But on the strength of those results, the researchers decided also to look at the virus protein AP205CP (short for Acinetobacter phage 205 coat protein), whose structure had yet to be resolved by any technique. The protein is a dimer consisting of two copies of a sequence of 130 amino acids; 90 copies of the dimer, in turn, make up the virus capsid. And it’s an illustrative example of how back-protonation doesn’t always work.
Figure 2 shows two versions of a 2D spectrum of AP205CP’s amide H atoms and their associated N atoms. The spectrum of a deuterated, back-protonated sample is shown in black. The spectrum of the fully protonated sample, shown in red, has several peaks that the black spectrum lacks; those additional peaks correspond to amide sites on the deuterated protein that couldn’t be exchanged. Also shown is the NMR protein structure, with the locations of the unexchangeable amides shaded in red.
Figure 2. Two-dimensional NMR spectra of the virus coat protein AP205CP, partially deuterated (black) and fully protonated (red). The labeled amino-acid spectral peaks in the fully protonated spectrum are missing from the partially deuterated one; for example, “V82” indicates the 82nd amino acid in the protein, which is valine. The missing amino acids’ positions are shaded in red in the structural diagram at the top. (Adapted from ref. 1.)
Figure 2. Two-dimensional NMR spectra of the virus coat protein AP205CP, partially deuterated (black) and fully protonated (red). The labeled amino-acid spectral peaks in the fully protonated spectrum are missing from the partially deuterated one; for example, “V82” indicates the 82nd amino acid in the protein, which is valine. The missing amino acids’ positions are shaded in red in the structural diagram at the top. (Adapted from ref. 1.)
Pintacuda and colleagues didn’t need to deuterate their sample, but they did use 13C and 15N labeling to get information about the positions of C and N atoms. So their method is still limited to proteins that can be grown in laboratory cell cultures—they can’t, for example, harvest a protein sample directly from a human patient. Pintacuda speculates that it could one day be possible to find a solid-state NMR structure of a protein using no isotopic labeling at all, merely the naturally present magnetic nuclei. “But resolving the proton resonances would be much harder,” he says, “because we wouldn’t have the information from nearby carbons and nitrogens.”
Although GB1 and AP205CP are both small proteins, high-frequency MAS NMR can potentially be used on much larger molecules. The main challenge is measurement sensitivity: The larger the protein, the fewer copies of it there are in a 0.5 mg sample, and the weaker the resulting signals. Doubling the size of the molecule under study would require an experiment four times as long. But each of Pintacuda and colleagues’ structures was found with less than two weeks of data collection. There is plenty of room to expand.
