Dissolution Dynamic Nuclear Polarization of Polypeptides

Nuclear Magnetic Resonance (NMR) spectroscopy provides remarkable site resolution, but often requires signal averaging because of low sensitivity. Dissolution dynamic nuclear polarization (DNP), which offers large signal enhancements, has been used to follow reactions involving small molecules that typically have long spin-lattice relaxation times. This thesis presents work in the development of dissolution DNP to directly hyper-polarize and observe polypeptides, which can subsequently be used for the study of a time dependent process, such as folding.

Dissolution DNP involves hyperpolarizing samples in the solid state, dissolving the samples with a stream of hot solvent and rapid transfer of the sample into an NMR tube for measurement in the solution state. Since protein samples are prone to foam under conditions for rapid sample injection, solvent systems were optimized. Solvents such as water/acetonitrile and water/methanol mixtures were utilized. An unlabeled peptide, bacitracin, was hyperpolarized on ^(1)H nuclei and enhancements of 30, 45 and 180 were obtained for amide, aliphatic and aromatic protons respectively. Although these enhancements are already significant, loss of hyperpolarization during sample injection was further alleviated by the use of isotopically enriched polypeptides. In [^(13)C, 50% ^(2)H] labeled samples of denatured L23, a 96 amino acid long ribosomal protein, signal enhancements of more than 500 times were obtained on the ^(13)C nuclei. This signal enhancement was then exploited to follow the protein folding process, using L23 as a model. Time resolved spectra of hyperpolarized L23 were measured after a pH jump and protein folding was monitored by observing changes in the carbonyl region of the spectra, which are indicative of the formation of secondary structures. Despite signal overlap in the protein spectra, using the statistical distribution of ^(13)C chemical shifts, the fractions of secondary structure elements were estimated for each scan of the DNP-NMR experiment. Additionally, individual resonances for methyl groups upfield of other protein resonances became resolved in the later transients. An option for the improvement of such site resolution by NMR experiments using coherence selection is discussed.

While DNP-NMR offers the capability to observe transient species, identification of such species is difficult in cases where not all chemical shifts are known. Here, a new strategy for the analysis of DNP-NMR data is proposed based on non-negative matrix factorization (NNMF). NNMF enables identification of various sources that contribute to an observed signal. This capability is demonstrated using a series of spectra measured from an enzymatic conversion reaction of oxaloacetic acid to malic acid. Simulations were carried out to evaluate the performance of NNMF under different experimental constraints, and the strengths and limitations of the method are discussed based on the simulations.