NMR: Dealing with Deuterons
Metabolic tricks and isotopic treats have allowed spectroscopists to push the molecular-weight envelope.
The structural analysis of simple compounds by nuclear magnetic resonance (NMR) has a long history, but the ability to study biological macromolecules such as proteins in a similar way is more recent. The development of homonuclear 1H (proton) two-dimensional experiments allowed the detailed analysis of small proteins (up to 10 kDa), but beyond this point, line broadening and overlapping proton chemical shifts complicated the subsequent analysis. In part, this problem was addressed by isotopically labeling the proteins with other NMR-active nuclei such as 15N, 13C, and 19F. Combined with the development of heteronuclear multidimensional NMR techniques, the isotopic labeling strategy allowed structural biologists to push the molecular-weight envelope to 2025 kDa.
This process suffers, however, from efficient protonproton dipolar spin relaxation and heteronuclear spin relaxation. As the nuclei do not exist in a vacuum, their rapid relaxation results in decreased magnetization transfer between the nuclei, and therefore, in reduced sensitivity and resolution. This relaxation process can be slowed somewhat by replacing protons with deuterons (2H), which have a 6.7-fold lower gyromagnetic ratio. This processdeuterationeliminates many of the relaxation pathways available in an otherwise protonated sample, and thus increases the final signal.
The use of deuterons to simplify protein spectra has been around since the late 1960s, but the technique was complicated by the signal broadening of the carbon nuclei adjacent to the deuterons. In 1973, Douglas Browne of the University of California at Berkeley suggested that deuteron decouplingthe introduction of a second magnetic field to eliminate the effect of the deuterons spin on the carbonswould eliminate the relaxation problem. This would require, however, the development of higher field strengths than were available at the time. Two decades later, using sufficiently high field strengths, Ad Bax and colleagues at the National Institutes of Health (Bethesda, MD) generated carbon line widths that were actually narrower than those of protonated carbons. Since that time, experiments combining the analyses of partially deuterated (perdeuterated) and fully protonated 13C- and/or 15N-labeled samples have gone a long way to pushing the molecular-weight envelope toward 40 kDa. In fact, one group used a similar protocol to determine the structure of a 64-kDa complex of a dimerized protein bound to its DNA target.
But even perdeuteration will only reveal so much structural information, because the exchange of deuterons for solution protons typically limits the data to those describing local secondary structure elements (backbone assignments). To glean higher-order structural information, distances between amino acid sidechain atoms are required, especially those in the proteins hydrophobic core. This information, which relies on nuclear Overhauser effect spectroscopy (NOESY)the measurement of the transfer of magnetization between atoms separated in spaceis largely lost when sidechain carbons are deuterated. Although some researchers have tried to get around this problem by introducing specific protonated amino acids into the bacterial growth media used to produce the proteins of interest, this method can be prohibitively expensive. With this in mind, a group of Toronto researchers decided to try some metabolic sleight of hand to accomplish their goals.
Typically, when proteins are produced in bacterial cultures, the bacteria are grown in a rich mixture of salts, sugars, and an extract of beef or yeast, which supplies many of the nutrients ready-made for the microbes. For labeling purposes, however, this rich medium is replaced by a minimal mixture of salts and sugar (typically glycerol or glucose), so that the bacteria are required to synthesize most of the compounds necessary for life. Because 13C-glucose and 15N-ammonium chloride or sulfate are used, the bacteria produce proteins consisting of similarly labeled amino acids. To generate perdeuterated proteins, the water in the medium is replaced with 2H2O, and the glucose is labeled with 2H. This results in a protein with deuterated methyl, methylene, and methine groups, which provide much of the NOE spectra.
To get around this, Lewis Kay and colleagues (including the author) replaced 13C, 2H-glucose with 13C, 1H-pyruvate in a background of 2H2O. They relied on the bacterial metabolism to exchange many of the carbon-bound protons for solvent deuterons during amino acid biosynthesis, thus creating largely deuterated proteins. But, for those amino acids for which pyruvate acts as an immediate biosynthesis precursor, the methyl protons are exchanged much more slowly. This occurs for alanine (-methyl), leucine (-methyl), valine (-methyl), and isoleucine (2-methyl)residues highly enriched in hydrophobic cores of proteins (Figure 1A).
To test the concept, the researchers examined a small protein fragment, the solution structure of which had already been solved in the same lab, and they determined what a theoretical structure of this fragment would look like if based on amideproton interactions and dihedral angles, plus or minus interactions with pyruvate-supplied methyl protons. The researchers found that although the procedure is not perfect, including the data from the methyl protons allowed them to come to within a 2-Å root-mean-squared deviation (rmsd) of the final structure, as compared to an almost 7-Å rmsd without the methyl group information.
Unfortunately, although the experiment resulted in proteins where the aforementioned groups were predominantly protonated, a high percentage of methyl groups still bore a combination of protons and deuterons, complicating the resulting spectra with secondary and tertiary peaks. The researchers called these mixed groups isotopomers. The researchers also dealt with the problem of not being able to measure the 1-methyl groups of isoleucine, the signals of which are significantly better resolved than those of other methyl groups and thus provide more information about the protein cores.
The researchers discovered that isotopomers were created by proton deuteron shuffling during the biosynthesis. Thus, the more steps involved between precursor and product, the greater the likelihood of shuffling. Adding [3-2H] -ketoisovalerate, the immediate precursor to valine and four steps away from leucine, to a growth using deuterated glucose led to ~90% protonation of the leucine and valine methyl groups (Figure 1B) with almost no isotopomers. Similarly supplementing [3,3-2H] -ketobutyrate, which combines with pyruvate to form isoleucine four steps later, generated isoleucine residues highly protonated at the 1-methyl position (Figure 1C).
Recently, the Toronto group used this labeling strategy on the multidomain 370-residue maltose or maltodextrin-binding protein to generate a series of NOE data. When combined with information about dihedral angles, the data allowed the researchers to generate a global foldthe overall shapefor the 42-kDa protein at a resolution of 2.4 Å, with resolutions of 3.8 Å for its N- and C-terminal domains, and 5.5 Å overall. When the data was refined with information on dipolar coupling, the structures precision was improved further from 5.5 Å to 2.2 Å.
But what good can come from medium- to low-resolution structures? While we wait for the development of stronger magnets, better solvents, and increasingly complicated pulse sequences to push the molecular-weight envelope further, the ability to generate relatively clear structures of large proteins has its usesmost clearly in the area of drug development. The identification of potential ligand-binding or protein-interaction sites will facilitate the development of compounds that might enhance or negate the activity of proteins of interest.
Similarly, it is becoming increasingly evident that protein homologies exist on two levels: one based on sequence, the other based on tertiary structure. The structures derived from perdeuterated samples will allow researchers to determine whether proteins that have similar functions also have similar structures, even though little sequence similarity may exist. Using compounds that are effective against a related protein as a template or model, researchers will be able to develop new drugs against the proteins of interest.
- Gardner, K. H.; Kay, L. E. Ann. Rev. Biophys. Biomol. Struct. 1998, 27, 357406.
- Goto, N. K.; Gardner, K. H.; Mueller, G. A.; Willis, R. C.; Kay, L. E. J. Biomol. NMR 1999, 13, 369374.
- Goto, N. K.; Kay, L. E. Curr. Opin. Struc. Biol. 2000, 10, 585592.
- Grzesiek, S.; Anglister, J.; Ren, H.; Bax, A. J. Am. Chem. Soc. 1993, 115, 43694370.
- Mueller, G. A.; Choy, W. Y.; Yang, D.; Forman-Kay, J. D.; Venters, R. A.; Kay, L. E. J. Mol. Biol. 2000, 300, 197212.
- Rosen, M. K.; Gardner, K. H.; Willis, R. C.; Parris, W. E.; Pawson, T.; Kay, L. E. J. Mol. Biol. 1996, 263, 627636.
Randall C. Willis is an assistant editor with Todays Chemist at Work Send your comments or questions regarding this article to email@example.com or the Editorial Office 1155 16th St N.W., Washington, DC 20036.