The publisher's final edited version of this article is available at Figure 1A shows the 15N–1H HSQC spectrum of an in-cell NMR experiment on αSN. The spectrum is consistent with that from previous studies.9,16 Figure 1B shows the spectrum from the supernatant collected immediately after sample preparation. Only metabolite signals17 are observed. Figure 1C shows the spectrum from the supernatant recovered after the in-cell NMR experiment. Again, only metabolites are observed. The data demonstrate that the αSN spectrum in panel A comes from αSN in the cell. We have obtained similar results with the intrinsically disordered protein FlgM.8 We performed the same experiments with CI2 expressing cells. In contrast to αSN, all three spectra are nearly identical (Figure 1E–G) (and typical of a CI2 spectrum18 in dilute solution). These data suggest that CI2 leaks from the cells. SDS-PAGE confirms that ~20% of the CI2 is lost from cells.
 | Figure 1 1H–15N HSQC spectra of αSN (left panels) and CI2 (right panels): (A, E) in-cell spectra; (B, F) spectra of supernatants collected immediately after preparing the cells; (C, G) spectra of supernatants collected immediately after completing (more ...) |
Encapsulation in alginate microcapsules19 stabilizes cells20 and may prevent leakage. To test if encapsulation might be useful for in-cell NMR, we first tried αSN-expressing cells. The encapsulated cells yield a typical αSN spectrum (Figure 1D), proving that encapsulated cells can provide useful in-cell spectra.
We repeated the experiment with CI2-expressing cells. No CI2 signal was observed even though we increased the sensitivity by accumulating the data for a longer time compared to the other samples (Figure 1H). However, a typical CI2 spectrum was recovered after dissolving the encapsulates with EDTA (data not shown). These observations suggest that the signal from the intracellular CI2, which we know is present in detectable amounts, is too broad to observe. We reasoned that the broadening arises from an alteration in the dynamics of CI2, either from binding a larger species in cells or from the higher viscosity of E. coli cytoplasm, which can be 10–11 times that of water.21,22
Why would the intrinsically disordered proteins αSN and FlgM react differently compared with a globular protein CI2 to the increased viscosity in cells such that we detect αSN and FlgM, but not CI2? The ability to detect a protein by high-resolution NMR depends on its dynamics, which are affected by viscosity. In terms of NMR, dynamics are reflected in the relaxation rates, R1 and R2 of the observed nuclei.23 If R1 is too small, the nuclei do not relax between pulses, lowering the sensitivity of the experiment. If R2 is too large, the resonances are too broad to detect. In general, smaller proteins, flexible proteins, and proteins in less viscous solutions exhibit larger R1 values and smaller R2 values than do larger proteins, ordered proteins, and proteins in viscous solutions.
To test the idea that R1 and R2 for αSN and CI2 react differently, we studied the response of the proteins to viscosity increases induced by the macromolecular crowding agent poly(vinylpyrrolidone) (PVP, 300 g/L, 40 kDa average molecular weight). We used PVP because it is soluble, has protein-like properties, and does not interact strongly with proteins.24 Figure 2 shows the relaxation rates of backbone 15N nuclei of αSN and CI2 in dilute buffer and in PVP solution.
 | Figure 2 R1 (A) and R2 values (B) of αSN and CI2 in 300 g/L PVP and in dilute solution (25 °C). The red lines indicate a unitary slope. |
Figure 2A shows the R1 values in buffer and in PVP. For most positions, the values for αSN changed little in buffer compared to 300 g/L PVP, even though the viscosity of the PVP solution is >50 times that of the dilute solution. Values from CI2, however, decrease 3–4 fold in PVP compared to dilute solution. The differential viscosity-induced decrease in R1 values for CI2 compared to αSN would make it more difficult to detect CI2 in cells, consistent with our observations (Figure 1).
Figure 2B shows the R2 values. For αSN, R2 increased between 1.5- and 6-fold in PVP compared to buffer, while the values for CI2 increases between 3- and 40-fold. The increases for CI2 compared to αSN under crowded conditions would also make it more difficult to detect CI2 in cells, again consistent with our observations (Figure 1). These changes in R1 and R2 for CI2 are not caused by aggregation of the protein in PVP because NMR-detected diffusion experiments are consistent with a monomeric protein (Supporting Information). In summary, our data show that the ordered globular protein CI2 is more sensitive to viscosity than the intrinsically disordered protein αSN and that this increased sensitivity is expected to degrade spectra for ordered proteins in cells.
The atomic-level explanation of these differential effects lies in differences in global and local motions for ordered and disordered proteins. Because of their rigidity, the relaxation rates for globular proteins are most sensitive to global motion, which is described by a single rotational correlation time.23 Disordered proteins, on the other hand, are flexible. Their motions are best described by considering an ensemble of interconverting conformers where every residue has a different effective correlation time.25 In essence, the flexibility of disordered proteins lessens the deleterious effect of viscosity on their spectra.
To the best of our knowledge, this is the first report of a differential dynamical response of disordered and ordered proteins to macromolecular crowding. Our data suggest that it will be easier to detect in-cell signals from disordered proteins compared to ordered proteins and that a focus on flexible side chains will be advantageous for in-cell NMR of ordered proteins.26 Because the cytoplasm of eukaryotic cells is less viscous than that of E. coli cells,27 our observations imply that high resolution in-cell protein NMR data may be easier to acquire in eukaryotic cells.