Selective isotope labeling strategy and computational interpretation of spectra for protein NMR analyses

Signal assignment is a mandatory step for the site-specific analyses of proteins by NMR. In challenging cases, such as low solubility target proteins or in-cell NMR, amino-acid selective stable isotope labeling facilitates the main-chain assignment, since it can determine the amino-acid type of each signal, thus providing complementary information to that obtained by standard three-dimensional triple resonance experiments. Inspired by information science methodologies, we developed a combinatorial selective labeling strategy, named stable isotope encoding (SiCode), to enable amino-acid typing even from noisy NMR spectra, with small numbers of labeled samples. The high noise-resistance of SiCode is achieved by both optimization of the isotope labeling pattern according to the information distance between amino acids and model-based computational retrieval of the amino-acid information from spectra, using replica exchange Monte Carlo computation. We demonstrate amino-acid typing with simulated low signal-to-noise-ratio spectra, with sample concentrations as low as micromolar order. Since the main-chain signal assignment is often a bottleneck process for various amide-signal-based NMR analyses of challenging target proteins, SiCode will facilitate and accelerate such analyses. (paper)