Hyperdimensional Protein NMR Spectroscopy in Peptide-Sequence Space

摘要

To extract useful information on protein structure and dynamics from NMR spectra, the observed NMR frequencies need first to be assigned to individual nuclear spins in the protein. Correlating all nuclear spins sequentially along the peptide sequence in a single experiment would be the ideal way of performing NMR resonance assignment. In practice however, even in the most favorable cases spin relaxation-induced signal loss imposes a limit of less than ～10 nuclear spins that can be correlated by NMR pulse sequences. In addition, standard multidimensional (nD) NMR methods require measurement times that increase by about 2 orders of magnitude per additional spectral dimension. Therefore, nD NMR beyond n = 3 or 4 is generally impractical using standard methods. Each of the K cross-peaks detected in a 3D (or 4D) spectrum provides correlated frequency information of 3 (or 4) nuclear spins. Such a spectrum can thus be considered as the sum of K projections of the unique correlation peak detected in a virtual hyperdimensional protein NMR spectrum. Protein resonance assignment refers to the problem of reconstructing this hyperdimensional spectrum from a limited number of summed projection spectra. Here, we show that large parts of the spectral space formed by the amide ~1H and ~(15)N backbone nuclei of the protein can be reconstructed from a small set of 3D correlation experiments using a correlation-based reconstruction algorithm (COBRA). The COBRA method, closely related to the recently introduced concepts of covariance and hyperdimensional NMR spectroscopy, provides a powerful new tool for automated fast protein resonance assignment.