Abstract
Membrane proteins are one of the main targets of medicinal research with over 50% of all medicines targeting these proteins. In particular, GQprotein coupled receptors (GPCRs) are a major target due to their ubiquitous and essential role in transferring external signals to cells in the human body. Unfortunately, uncovering the structure and signaling pathways of GPCRs is very challenging due to low expression yields and intrinsic conformational exchange that result in low protein stability. For NMR spectroscopy the additional challenge is the large size, which reduces the spectral quality and restricts the use of standard procedures. Moreover, the fact that the hydrophobic environment of the membrane needs to be imitated in order to use solution NMR further increases the size. In the first chapter the issues that are associated with using NMR spectroscopy on large membrane proteins are discussed, together with commonly used solutions that allow measurement, assignment of the spectral data and dynamics for large molecules. In addition an overview of automated assignment procedures is given as the FLYA automated assignment procedure is extensively used in this work.
In this thesis approaches to further improve measuring large proteins with NMR spectroscopy are investigated. For this purpose bacteriorhodopsin (bR), which is a 7QTM alpha helical membrane protein that transports protons through the cell membrane and is powered by light, is used as a model protein. The overall structure of bR is similar to the GPCR fold, but has a higher stability and more can be expressed more easily. The ligand, retinal, is covalently bound inside the helix bundle and is responsible for changing the conformation after light absorption. bR is incorporated into nanodiscs, which significantly increases the stability (at the expense of additional size of the assembly). The resonance assignment in this thesis is, to the best of our knowledge, the first in nanodiscs without prior assignment in detergent micelles.
Applying the standard procedure for obtaining assignment information showed its limitations when applying to bR. The more complex spectra HN(CO)CACB and HN(CA)CO had greatly reduced spectral quality and many peaks were not observed. The 4D NOESY spectra, which resolve spatial connections between amide and methyl or between methyl groups, offered good spectral quality although long measuring times were needed. To supplement these data additional data was gathered with spectra that are still feasible for large proteins, i.e. 2 dimensional spectra or sensitive 3 dimensional spectra such as HNCO/HNCA. The additional data measured are single amino acid 15N labeling, proximity to water/lipid, chemical shifts from assignments of protein fragments and partial 12C labeling. The initial assignment was assisted by the FLYA automated assignment procedure, part of the CYANA program, to handle the complexity of the many different types of input data. The assignments were validated manually by reviewing the data that led to the assignments and even slightly more assignments were found in the process of validation. Most of the additional assignments came from manual interpretation of the structure data as the crystal structure only shows a single rigid protein state.
With the protein assigned the influence of individual data sources was investigated by changing the input data for the automated assignment procedure and monitoring the amount of correct assignments. Either single data sources were removed from a full dataset or added to limited data scenarios where data that is difficult to measure for large proteins were removed. From this analysis it became clear that proximity data and fragment analysis did not contribute to the assignment, and that the use of proximity data had even a negative influence for the methyl assignment. The proximity data was probably not limiting the assignment possibilities sufficiently, and chemical shifts from fragments were deviating too much from the fullQlength protein to be useful. The single amino acid 15N labeling and the partial 12C labeling were redundant in a full data background, but when removing the 15NQNOESY spectra or the correlations between the amide and CβQ1/C’, these additional data types contributed to the assignment. For methyl assignments the 4D NOESY spectra proved crucial while the 3D NOESY counterparts were not able to yield the same amount of assignments. Other strong contributions came from the backbone data and the addition of correlations between the methyl group and Cα/Cβ atoms within the same residue.
With most of the data assigned it became clear that around 17% of the backbone residues in bR were not observed. The residues that remained unassigned are clustered on the extracellular side of bR, as well as in helix E. A common cause of missing peaks is conformational exchange that broadens the peaks, indicating the presence of a dynamic process on the extracellular side of bR. Some speculation about its origin is offered in this thesis. The occurrence of conformational exchange is also common in GPCRs, and therefore bR offers a unique model system in which techniques can be tested to improve NMR measurement of GPCRs.
Two types of dynamics were observed for bR. The first type was found in the presence of peak doubling in all spectra, which indicates two states that interchange slowly. The peak intensity ratio and the fact that the process is slow implied that the doubling comes from the thermal isomerization of retinal, which occurs in the dark over minutes. Using the mutation M145A, which almost completely removes the darkQadaptated state, confirmed this hypothesis. The second type of dynamics is much faster and can be described by the S2 order parameter, which was determined for the methyl groups. This parameter describes side chain motions and can be classified into different motional regimes. In bR the S2 order parameter mostly behaves as expected with rigid methyl groups inside the helix bundle and flexible on the outside. However, the opposite also occurs, and no clear correlation to parameters such as bQfactors or signal intensities could be found. Similar observations were made for soluble proteins previously, and the results presented here offer a valuable contribution for further investigations.
In addition to removing the darkQadaptated state the M145A mutant also disables the protonQpumping activity of bR and changes the color from purple to orange. The mutation is in the retinal binding pocket and therefore indicates what can happen to the protein when the binding pocket is perturbed. On the intracellular side within the helix bundle relatively strong chemical shift perturbations are observed as well as increased flexibility as shown by corresponding changes in the S2 order parameter. In this region the uptake of a proton takes place as one of the later steps in the proton pumping cycle. The disturbance of dynamics could be one of the reasons for the reduced activity. Overall the darkQstate equilibrium dynamics indicate that there is more to bR than has been observed so far and additional investigations would be worthwhile. We feel our data may be useful to understand how the protein is structurally prepared in the dark for entering the photocycle upon illumination.
Kooijman, Laurens