Why should we care about protein complexes with a molecular mass greater than 1 mega-dalton? Many of the most important biological processes in human health are mediated by large protein complexes. To make matters worse, many of the most powerful tools available to chemists and biochemists, such as NMR, are not suitable for such systems. What are the challenges in applying NMR spectroscopy to such large complexes? In traditional protein NMR (aka solution state NMR) there are two issues: #1) There are too many signals in large complexes. The resulting overlap makes it difficult to assign resonances and interpret data; #2) As the molecular mass increases, the spin-spin relaxation rate constant, R2 (= 1/T2) increases. The resulting broad lines diminish sensitivity and resolution. I'll note that many of the most important developments in NMR methodologies over the past 15 years have focused on addressing these limitations. Examples include TROSY (along with perdeuteration) which fosters backbone resonance assignment of proteins with molecular mass of ~ 80 kDa and methyl-based labeling and methyl-TROSY experiments for side chain dynamics of very large systems (> 500 kDa).
In contrast to solution state NMR, for immobilized rigid solids spun rapidly at the "magic angle", the line width of resonances do not depend on molecular mass. No matter how big the molecule, it is possible to acquire spectra with narrow lines (line width ~ 50 Hz) as long as the magic angle spinning (MAS) rate is greater than the anisotropic interactions that broaden lines in solids (such as dipolar coupling). Fortunately, there has been impressive technical developments to miniaturize rotors and maximize spin rates. The problem is that MAS solid state NMR (ssNMR) requires rigid solids, usually meaning crystals of protein complexes.
The authors have developed a technique that they call FROSTY MAS to take advantage of the fact that for solids the line width is not a function of molecular mass, while avoiding the inconvenience of working with solid samples. In FROSTY MAS ~40 uL of ~3 mM protein (40x10^-6 * 3x10^-3 * 1x10^6 = .12 g!) is dissolved in a buffer with 30-40% glycerol and loaded into a 4 mm rotor. The protein is fully deuterated at non-exchangeable sites and only ~20% protonated at exchangeable sites. Also, there is Cu(II)-EDTA in the solution. This mixture is loaded into a solids probe and spun at 22 kHz. Under these conditions rotational reorientation is impeded and the molecule behaves just as if it were a rigid solid lattice, allowing the authors to get narrow lines regardless of the protein molecular mass! The objective of this paper is to demonstrate that the FROSTY MAS technique can be used to acquire 1H-detected backbone-based ssNMR experiments for resonance assignments of large protein complexes.
The protein that Mainz et al. apply their novel technique to is the 20S proteasome with 11S activation lids. The 20S proteasome is a multi-subunit complex consisting of 4 heptameric rings. This protein plays a vital role in maintaining cellular function via selective degradation of proteins. Figure 1 explains the modular architecture of this protein (along with the molecular mass and rotational correlation time of each complex).
One trick the authors use to reduce the number of signals is that only the alpha subunit is isotopically enriched. So overall, their spectrum will have as many signals as a 26 kDa protein (233 residues), even though it tumbles like a 1.1 MDa complex!
What are the authors results? First, they record a proton-detected MAS spectrum for three complexes: the double heptameric ring (a7a7, 360 kDa), the full 20S proteasome (a7b7b7a7, 670 kDa) and the 20S-11S complex (1.1 MDa). Figure 3 shows these results.
The y-axis of these plots are normalized to the concentration of alpha subunit in the samples. The observation is that the signal increases with molecular mass. The authors interpretation is that the sensitivity increases due to "the reduced rotational mobility of the larger assemblies in the sedimented state." In other words, the larger complex is more solid-like so the ssNMR tricks work better. An alternative explanation that the authors address is that chemical exchange is responsible for the difference in signal intensity. One could imagine that the a7a7 or a7b7b7a7 samples are in slow or intermediate exchange between two conformations and the addition of the 11S cap stabilized one state. The traditional tools to address chemical exchange, at least for small molecules, are temperature and field. The authors do not use these tools here. To be honest, I do not really follow their argument (which takes ~1 paragraph and jumps from CP to TROSY), but at the end the authors dismiss the possibility that dynamics on the us-ms or ns-us timescale could be responsible for the sensitivity increase.
The other major result is the backbone resonance assignment of the alpha subunit of the 1.1 MDa complex. Kay and co-workers have assigned the single ring a7 complex (180 kDa) in solution. Supplemental figure S3 overlays the 1H-15N TROSY of this molecule (red) with the FROSTY-MAS 1H-15N correlation spectrum of the 1.1 MDa complex.
This figure certainly highlights the impressive sensitivity and resolution of the FROSTY-MAS experiment. The authors also record hCAhNH and hCOhNH (ssNMR equivalent of the HNCA and HNCO) of the 1.1 MDa complex. These experiments, in combination with the assignments from the solution state, enable backbone resonance assignment of 108 or 227 non-proline amino acids in the alpha subunit. The less optimistic description is that 119 residues are not assigned. Of course, Kay and co-workers could not assign (some of) these residues in the solution state either, because of ms-us timescale dynamics. At any rate, using the assignments they have, the authors can map the interface between the 11S activator and the alpha subunit using chemical shift perturbations. Also they can use a program like TALOS+ to assess the secondary structure from the Ca chemical shift.
Overall, this paper is a very exciting demonstration of NMR spectroscopy on protein complexes that are so large that it would have been unthinkable to try to study a few years ago. My imagination could run wild dreaming up uses for this technique. One example is the crowded cellular environment. I am a bit concerned about the mass and solubility requirements for this technique. It is not cheap to make > 100 g of triple labeled protein. Also, as I understand it, any aggregation or amorphous solids in the sample will undermine FROSTY-MAS. These concerns are minor, though. This paper is outstanding and paradigm shifting! For very large protein complexes we could see solid state NMR surpass solution in the very near future, particular when taken in combination with other emerging solid techniques, such as DNP.