Backbone-independent NMR resonance assignments of methyl probes in large proteins

Santrupti Nerli,Andrew C Mcshan,Nikolaos G Sgourakis,Viviane S De Paula

doi:10.1038/s41467-021-20984-0

Abstract

Methyl-specific isotope labeling is a powerful tool to study the structure, dynamics and interactions of large proteins and protein complexes by solution-state NMR. However, widespread applications of this methodology have been limited by challenges in obtaining confident resonance assignments. Here, we present Methyl Assignments Using Satisfiability (MAUS), leveraging Nuclear Overhauser Effect cross-peak data, peak residue type classification and a known 3D structure or structural model to provide robust resonance assignments consistent with all the experimental inputs. Using data recorded for targets with known assignments in the 10–45 kDa size range, MAUS outperforms existing methods by up to 25,000 times in speed while maintaining 100% accuracy. We derive de novo assignments for multiple Cas9 nuclease domains, demonstrating that the methyl resonances of multi-domain proteins can be assigned accurately in a matter of days, while reducing biases introduced by manual pre-processing of the raw NOE data. MAUS is available through an online web-server.

Highlights

Methyl-specific isotope labeling is a powerful tool to study the structure, dynamics and interactions of large proteins and protein complexes by solution-state nuclear magnetic resonance (NMR)
This strategy was first outlined in the method MAGMA13, which treats methyl assignments as a maximum subgraph matching problem, and invokes a graph theory algorithm[14] to enumerate all assignments, which satisfy the maximum number of nuclear Overhauser effect (NOE) connectivities
Methyl Assignments Using Satisfiability (MAUS) explicitly considers (i) all possible mappings of 3D NOE cross-peaks to two-dimensional (2D) reference peaks, or clusters (Fig. 1c), and (ii) all possible matchings between upper-diagonal and lowerdiagonal NOE cross-peaks, formally analyzed as connected components of a bipartite symmetrization graph (Fig. 1d)

Summary

Results and discussion

Rather than treating the methyl assignments as a maximum subgraph matching problem[13], MAUS models the NOE data as a sparse sample of all possible connectivities present in the input structure. MAUS explicitly considers (i) all possible mappings of 3D NOE cross-peaks to two-dimensional (2D) reference peaks, or clusters (Fig. 1c), and (ii) all possible matchings between upper-diagonal and lowerdiagonal NOE cross-peaks, formally analyzed as connected components of a bipartite symmetrization graph (Fig. 1d) This approach relieves the user from the burden of interpreting the raw data (3D or 4D NOE peaks), through an explicit and exhaustive consideration of all possible data graphs consistent with the input NOE peaks (Fig. 1e). The X-ray structures and ground truth assignments for these targets were

G Structure graph

Methods

Code availability