Simplifying Solid-State NMR Spectra for Biophysical Studies on Membrane Proteins: Selective Targeting of Sites and Interactions

Membrane proteins are arguably the most interesting molecules for biophysicists and structural biologists. Besides being of compelling interest to basic researchers, they represent important pharmacological targets with the majority of available drugs acting on membrane proteins. Traditionally, x-ray crystallography remains the most powerful technique to study membrane proteins and both the first membrane protein structures as well. By far, most structures have been solved by this technique. However, crystallography is restricted to study the molecules in a nonnatural environment, typically involving cubic phases, antibody binding, thermostabilization, or major mutagenesis.

Solid-state nuclear magnetic resonance (NMR) spectroscopy has also been applied to investigate membrane proteins for a long time. This technique allows studying these molecules in a natural liquid crystalline bilayer environment in the presence of physiological buffers and at biological temperatures. This approach provides access not only to the structure of membrane proteins but also allows studying their comprehensive molecular dynamics, which reveals an essential part of the function of these fascinating molecules. For a long time, solid-state NMR has predominantly been applied to membrane-bound peptides (1), but recent developments have allowed us to study the structure (2) and dynamics (3) of transmembrane proteins as large as G protein-coupled receptors.

However, solid-state NMR spectroscopy on large membrane proteins is far from being a standard technique that can always provide answers to biophysical questions. Although some large proteins can be prepared and reconstituted in high structural homogeneity (4), the majority of the membrane proteins exhibit NMR spectra that are characterized by structural heterogeneity, unfavorable intermediate timescale motions, and exchange processes—all of which result in line broadening and consequently severe signal superposition, rendering the analysis of solid-state NMR spectra difficult.

Over the last decade, researchers have worked on overcoming these difficulties and a popular solution to the problem is the application of spectroscopic filters that simplify the complicated and crowded solid-state NMR spectra. In a seminal article in this issue of Biophysical Journal, Huang et al. (5) have applied such filters to their complicated ¹⁵N NMR spectra to provide domain assignments and structural information of their membrane protein under study. They investigated the FMN-binding domain of cytochrome P450 reductase, which is involved in the function of important catabolic enzymes that metabolize the majority of the drugs currently in use, in a nativelike membrane environment (5).

A simple and straightforward filter is the application of selective isotopic labeling (4). Complicated and crowded NMR spectra can be simplified by decreasing the number of labeled residues in the protein, which reduces the number of signals in the NMR spectra leading to assignment of broadened signals. Several techniques allow selective labeling of isolated atoms or residues in recombinantly produced membrane proteins, and application of complementary labeling schemes have provided the full assignment of many membrane proteins (4). Also, cell-free expression provides unique possibilities for selective labeling without isotope scrambling (6). Furthermore, exchangeable amide protons can partially be replaced by deuterons (D), and these sites would not be detectable in ¹⁵N NMR spectra that receive their polarization from the attached ¹H nucleus. Amide hydrogens in membrane-buried secondary structures would resist exchange for D because of their strong hydrogen bonds, while non-hydrogen-bonded amide protons would exchange. Therefore, application of this filter discards the signals from the unstructured segments of a membrane protein and retains the resonances from stable secondary structures. Although several samples have to be prepared for the application of isotopically filtered NMR, this technique has been instrumental for the recent accomplishments in this field (2,5).

Another filter technique takes advantage of the fact that membrane proteins are dynamic molecules and that the dynamics is heterogeneously distributed in large membrane proteins. Typically, transmembrane segments are less mobile than the soluble domains or the loops and termini of a membrane protein. One can therefore apply dynamics filters that would allow us to exclusively detecting the rigid and the mobile segments of a membrane protein. In solid-state NMR, this is accomplished by varying the polarization transfer scheme that transfers the polarization from ¹H to either ¹³C or ¹⁵N, which are the nuclei that are traditionally detected in solid-state NMR of membrane proteins. Polarization transfer can be accomplished by using either dipolar or scalar couplings. Dipolar couplings are strong for rigid sites and decrease if the motional amplitude of a given segment is increased. For isotropically mobile sites, the dipolar couplings vanish. Scalar couplings, on the other hand, do not depend on molecular mobility, but can only be used for polarization transfer when a given segment shows sufficient mobility to reduce the dipolar coupling and provide a long T₂ relaxation time. Thus, complementary solid-state NMR experiments can be carried out, which solely excite the rigid sites either by using dipolar couplings in cross-polarization (CP) experiments or the mobile residues by using scalar couplings, for instance, in insensitive-nuclei-enhanced-by-polarization transfer (INEPT) experiments. Thus, a mobility filter can be applied to the NMR spectra that detect the NMR signals of a membrane protein based on its molecular dynamics (3). In CP-based NMR spectra, predominantly the rigid sites show intensity, whereas INEPT-based spectra only detect the highly mobile sites. Systematically increasing the CP contact time also allows detection of the more mobile sites of a large membrane protein (3).

A third way of applying filters to NMR spectra is to selectively detect signals from given secondary structures. To this end, pulse sequences have been developed that exclusively detect resonances from either α-helical (7) or β-sheet structures (8), thus simplifying crowded NMR spectra.

Finally, paramagnetic tags have also become popular in solid-state NMR spectroscopy because they can quench the NMR signals of sites that are in close proximity to the paramagnetic center (4), and thus simplify the complicated NMR spectra of membrane proteins.

In the article by Huang et al. (5), the authors decided to uniformly label the FMN-binding domain of cytochrome P450 reductase with ¹⁵N. To this end, expression of this large 239-amino-acid membrane protein in Escherichia coli and subsequent reconstitution into DLPC/DHPC bicelles was successfully accomplished. As expected, the NMR spectra of the large 27-kDa molecule were very crowded. However, by application of several of the above-mentioned filters, the group was able to identify the transmembrane segment of the FMN-binding domain and determine its orientation with respect to the membrane normal.

In particular, application of dynamics filters allowed us to spectroscopically separate the signals from the transmembrane and the soluble domains. By varying the CP contact time, i.e., the time during which the polarization transfer from ¹H to ¹⁵N is active, the authors could control which sites were excited. A short contact time excites only the rigid residues, which are localized in the transmembrane segment of the FMN-binding domain. Complementarily, an INEPT-based experiment only excited the mobile segments of the protein from the soluble domain. Thus, assignments of residues from the extracellular and the transmembrane segments were achieved.

To confirm this assignment of the peaks in the ¹⁵N NMR spectra, the authors also applied the hydrogen/deuterium (H/D) exchange filter and unambiguously identified the transmembrane segment of the FMN-binding domain. After 5 h of H/D exchange, the NMR spectra only showed the resonances that had previously been assigned to the transmembrane region of the spectrum, which is solvent-inaccessible. In contrast, the NMR signals from the residues that reside in the soluble domain of the protein showed a drastic reduction in intensity due to H/D exchange. With this important information, the authors could go ahead and carry out a separated local field experiment that allowed characterizing the structure and membrane orientation of the transmembrane segment of the FMN-binding domain. The authors give unambiguous evidence that the transmembrane segment of the protein represents an α-helix. With this information, they developed the first model of the structure of the large membrane protein.

The application of filters in solid-state NMR helps in studying large membrane proteins in native environments that show neither perfect structure homogeneity nor provide narrow NMR signals. These filters can be applied to both static and magic-angle spinning NMR experiments, and a combination of all these techniques appears to be the most promising strategy. Together with other biophysical methods such as molecular-dynamics simulations (9), solid-state NMR will play a crucial role in the investigation of important biophysical questions of membrane proteins that have eluded investigation so far (3,5,6,10).