Is in-cell NMR feasible?

As I mentioned in my last post, it is sometimes difficult to tell if a protein structural model matches the physiologically-active form. One way to address this problem would be to take structural data under physiological conditions, i.e. in the cell. While this obviously isn't a viable approach for crystallography, it seems possible with other techniques such as FRET or NMR. Existing NMR structures are solved under solution conditions, and the cytoplasm is a kind of solution. So we would expect to be able to solve structures, or at least get spectra, in vivo. However, the success of NMR strongly depends on what kind of solution you have, and a forthcoming paper in JACS suggests that the bacterial cytoplasm may not be amenable to NMR.

This paper has a bit of history behind it. A few years ago, Gary Pielak's lab attempted to determine the dynamic behavior of chymotrypsin inhibitor 2 (CI2) and apocytochrome-b5 in living E. coli (1),(2). They found in these experiments that the dynamics of these proteins did not differ significantly between in vivo and dilute solution conditions. A later experiment, however, showed that most or all of the signals that had been observed arose from protein that had leaked out of the cells and into the surrounding medium in which they were suspended. This wasn't true for all proteins: cells expressing α-synuclein (αSN) and FlgM did not leak protein. Subsequently the first two papers were retracted (3), while Gary and his folks tried to figure out just what was going on.

Li et al. encapsulated E. coli cells that were producing either CI2 or αSN in alginate microcapsules to prevent protein leakage and took some basic NMR spectra. Samples of the surrounding fluid proved that neither protein was leaking into the surrounding medium. Nonetheless, although normal spectra of αSN were observed, the cells expressing CI2 did not produce any CI2 spectral signals at all (normal metabolite signals were observed). Experiments showed that these cells did contain CI2 (4). What happened to the signal?

Li et al. hypothesize that this is an effect of the viscosity of the cytoplasm. The intensity of an NMR signal depends on the concentration of the atoms giving rise to the signal, and also the relaxation properties of these atoms, i.e. the rate at which the polarization induced by radio-frequency pulses returns to equilibrium levels. For larger molecules and viscous solutions, the rate at which detectable polarization decays (R₂) is higher, and the rate at which polarization returns to equilibrium (R₁) is lower. The behavior of R₂ is one of the chief obstacles to performing NMR on proteins or complexes that have high molecular weight.

In order to establish that this cytoplasmic viscosity is a reasonable cause of the invisibility of CI2, Li et al. measure the NMR relaxation of αSN and CI2 in a solution doped with poly(vinylpyrrolidone) (PVP), which increases the viscosity significantly. They find that the R₁ of αSN is essentially unchanged, while that of CI2 decreases significantly. The R₂ of αSN resonances increases 2-6 fold relative to dilute solution, but for CI2 this increase is much more significant, up to 40-fold. You can see these features in the graph at right. These results indicate that the relaxation behavior of resonances in a stably folded protein like CI2 is dominated by the global rotational diffusion, while the relaxation of resonances in a disordered protein like αSN is dominated by the significant local fluctuations. Because these latter motions involve less surface area and hence less drag (as compared to the folded protein), the viscosity of the solution has relatively less of an effect. Thus, the vanishing CI2 peaks could conceivably be the result of differential relaxation responses to viscosity.

However, other groups have managed to acquire NMR spectra of folded proteins in E. coli before. These experiments may have experienced the same leakage problem that affected the previous CI2 results, and some control experiments addressing this possibility are likely to be forthcoming. At the same time, the PVP solutions here are supposedly 5 times as viscous as E. coli cytoplasm, and yet CI2 resonances are apparently still visible. This suggests that the complete obliteration of these resonances in cells cannot be chalked up entirely to viscosity. One possibility is that CI2 overexpresses so significantly that it massively increases the effective viscosity of the cell. It may also be possible that the vast quantities of CI2 expressed in these experiments are crammed into a small volume somehow, perhaps as relatively loose inclusions in the cytoplasm or loaded in the periplasmic space, where interactions with peptidoglycan may rotationally lock them. It is also possible that PVP does not capture some important aspect of cellular crowding. For instance, the washing-out of the CI2 signal may be an effect of transient protein-protein interactions. Fluorescence and electron microscopy experiments may be able to address this concern.

If the observations of CI2 are really a result of some unique localization or a consequence of the significant overexpression then it is possible that NMR of different proteins on somewhat less active promoters may yet be viable. In addition, in vivo studies of proteins in systems such as X. laevis oocytes or cultured eukaryotic cells may not be subject to the same difficulties, due to differences in cytoplasmic viscosity. The major implication of this work, however, is that in vivo NMR results cannot be taken at face value. Careful controls will be required in order to ensure that the experiments actually measure protein within the cell, and in the absence of such controls the results should be interpreted cautiously.

1. Bryant, J., Lecomte, J., Lee, A., Young, G., Pielak, G. (2005). Protein Dynamics in Living Cells. Biochemistry, 44(26), 9275-9279. DOI: 10.1021/bi050786j

2. Bryant, J., Lecomte, J., Lee, A., Young, G., Pielak, G. (2006). Cytosol has a small effect on protein dynamics. Biochemistry, 45(33), 10085-10091. DOI: 10.1021/bi060547b

3. Bryant, J., Lecomte, J., Lee, A., Young, G., Pielak, G. (2007). Retraction. Biochemistry, 46(27), 8206-8206. DOI: 10.1021/bi700744h

4. Li, C., Charlton, L.M., Lakkavaram, A., Seagle, C., Wang, G., Young, G.B., Macdonald, J.M., Pielak, G.J. (2008). Differential Dynamical Effects of Macromolecular Crowding on an Intrinsically Disordered Protein and a Globular Protein: Implications for In-Cell NMR Spectroscopy. Journal of the American Chemical Society DOI: 10.1021/ja801020z