The initial step in understanding molecular function is to determine the structure of the molecule of interest. Subsequently, this structural data often leads to further mechanistic hypothesis that can be tested with additional experiments that elucidate function. Accomplishing either of these steps requires a method that provides resolution at the atomic level and therefore X-ray diffraction and/or solution NMR are the methods of choice. Both provide data at a length scale that permits bond distances and angles and therefore molecular structures to be constrained.
However, there are many systems to which these approaches are not applicable because the molecules either do not crystallize or are insoluble or both. For example, amyloid proteins associated with a variety of age related diseases such as Alzheimer’s, Parkinson’s, Type-II diabetes, etc. form insoluble fibrillar aggregates that have thus far defied crystallization. Similarly, studies of membrane proteins usually require crystallization or solublization in detergents, as opposed to native membranes, for diffraction or solution NMR structural studies.
Magic Angle Spinning and Dipolar Recoupling
To address these challenges we have developed new approaches to determine structure that rely on a combination of solid state NMR techniques. To provide high resolution we employ magic angle spinning (MAS) an approach in which the sample is placed in a rotor that rotates rapidly (approximately 10-70 kHz) about an axis inclined at 54.7º (the magic angle) with respect to the magnet field, B0 (illustrated in Figure ic). Internuclear distances and torsion angles are available via dipolar recoupling experiments and high frequency dynamic nuclear polarization (DNP). MAS provides site specific high resolution required for atomic level structures. Dipolar recoupling permits spectral assignments and reintroduces the 13C-13C and 13C-15N distance and angular information into the MAS NMR spectra by utilizing a train of rotor synchronized rf pulses. The result is a series of two- (or higher) dimensional spectra (Figure ic) that permits a structure to be determined of an amyloid or membrane protein (Figure i). DNP, discussed further below, provides much increased sensitivity. Most of the instrumentation and methods to perform these experiments were developed at the Francis Bitter Magnet Laboratory, MIT.
The 1990’s were occupied with developing these magnetic resonance techniques and testing them on small model systems that permitted us to verify the validity of the approaches. The result was the creation of a repertoire of methods for applications that functioned well at lower fields ≤500 MHz 1H NMR frequencies. In 2002 these efforts culminated in the first atomic-resolution structure of a biomolecule using MAS dipolar recoupling NMR. Although the structure determination was of a mere tripeptide (N-f-MLF-OH) shown in Figure ia, it demonstrated that solid-state NMR would be able to produce structures as detailed as those obtained by X-ray crystallography. In the case of N-f-MLF-OH the RMSD, a measure of the quality of the structure, was an astonishing 0.02 Å! Furthermore, with these same techniques we were able to build on this success and determine the structure of a monomeric amyloid peptide embedded in an amyloid fibril. This structure, a β-strand that is microscopically well ordered, is illustrated in Figure ib superimposed on an electron micrograph showing the amyloid fibrils.
A key to addressing structural problems in larger proteins with MAS NMR has been the availability of high-field magnets – 700, 800, and 900 MHz 1H frequencies – that became available in the decade beginning in 2000. In particular, they provide the chemical shift resolution that is essential for assigning the spectra and determining protein structures. Thus, in 2010 by performing dipole recoupling on high field spectrometers, we determined the structure of matrix metalloproteinase 12, a protein of 159 amino acids (17.6 kDa molecular weight). Finally, we have combined all of these techniques together with high field spectra and DNP (vide infra) and very recently completed the first structure of an amyloid fibril illustrated in Figure id. On the left of this figure we illustrate and electron micrograph of the fibril showing its characteristic twisted ribbon structure and the middle and right parts of the figure shows the six protofilaments (composed of 12 of the monomers in Figure ib) that pack into the fibril that for the twisted structure. The atomic details of this structure – the interstrand alignment, the intersheet alignment and the head to tail arrangement of the protofibrils – emerged from MAS NMR measurements of distances and torson angles. It is not only the first structure of an amyloid fibril, but the first example of a three dimensional structure determined by MAS NMR.
Since the publication of our initial papers, many other laboratories have begun to work actively on the techniques, with a goal of solving the structures not amenable to crystallization. Today, some of the most exciting results are emerging from studies of membrane proteins, amyloid fibrils, or large protein aggregates, such as bacterial secretion needles or the nascent chain of the ribosome. As the methods continue to be refined and improved it will likely evolve to standard tool for structural investigations of biological systems, particularly amyloid and membrane proteins.
Dynamic Nuclear Polarization
Although we are now able to determine structures of proteins with MAS NMR, it remains a time consuming process requiring extended periods of signal averaging and multiple experiments. Thus, NMR would profit from a significant upgrade in sensitivity allowing structural studies to become more routine and accessible to the chemical and biological communities. To address this challenge we developed high frequency dynamic nuclear polarization (DNP), a technique that increases the NMR signal intensities of a sample by transferring the high polarization present in an electron spin system to nuclei in the sample. This is illustrated in Figure iia where the 1H polarization at 700 MHz 1H frequency and at 90 K the polarization (the number of spins up vs. down) is 0.0186%. In contrast, the electrons polarization is 12.24%, approximately 660 times larger. To perform the experiments the electrons are introduced into the sample under investigation in the form of a stable paramagnetic radical. Figure iib shows some biradical polarizing agents, developed in collaboration with Professor Tim Swager of the Chemistry Department, which are the preferred source of electrons. Finally, the sample-radical combination is irradiated with high frequency microwaves (150-600 GHz) that transfers the high polarization of the radical’s electron spin to the nuclear spins of a protein or other target molecule of interest. At the moment the preferred sources of high frequency microwaves are gyrotrons (a.k.a. cyclotron resonance maser) developed with our colleague Dr. Richard Temkin at the MIT Physics Department.
This approach has been very successful and it is now possible to routinely achieve signal enhancements of ≥100 in amyloid and membrane and protein samples. An example showing a DNP enhanced signal of ~290 on a model system of urea is shown in Figure iic. These signal enhancements are very important since a factor of 102 translates into an experimental time savings of 104. More directly, this means that an experiment which would takes one day sans DNP would require about approximately 30 years avec DNP! Thus, high frequency DNP has attracted the interest of instrument companies who have are now producing gyrotrons and the associated hardware for DNP/NMR commercially. Biradical polarizing agents are available from a startup company DyNuPol, founded as a direct result of these research efforts.
The above discussion makes it clear that the optimal approach to performing structural studies with MAS NMR would be to incorporate DNP into the experimental protocol. Thus, the time consuming experiments mentioned above could be shortened considerably and the precision of the structural data improved dramatically. The first complete example of this integrated approach was used to determine the structure of the amyloid fibrils mentioned above. Although the monomeric peptide structure was determine without DNP, the measurement of the longer distances (4-6 Å) involved in the assembly was done largely with DNP experiments. For example, the interstand distance required to determine the alignment within the β-sheet were partially determined with DNP and the head to tail distances. In the future we will likely see DNP and MAS NMR become an integral part of the protocol use for the determination of the structures of complex biological systems.