Dedicated to Professor Horst Kessler on the occasion of his 70th birthday and to Professor Christian Griesinger on the occasion of his 50th birthday Eukaryotic proteins typically have a modular architecture, characterized by multiple structural domains that are connected by flexible linkers. Regulation of cellular processes depends on an interaction network between these individual modules and the formation of a quaternary structure. Dynamic rearrangement, often coupled to ligand binding, is a common feature of these multicomponent systems. While compact and rigid complexes can be efficiently studied using X-ray crystallography, protein complexes or multidomain proteins that involve weak and transient domain interactions should be preferably investigated using solution techniques. A number of NMR spectroscopy studies of protein complexes have been reported in recent years. [1][2][3][4][5][6][7][8][9][10][11] However, a general protocol is not available, and many applications still rely on NOE-based interdomain distance restraints, which are difficult to obtain and to assign in high-molecular-weight systems.
Herein, we present a general and robust protocol for the structural analysis of multidomain proteins and protein complexes in solution. We rely primarily on the type of NMR spectroscopy data that can be obtained for complexes with molecular weights well above 50 kDa. Starting from the three-dimensional structures of isolated domains, residual dipolar coupling (RDC) data and paramagnetic relaxation enhancements (PREs) are combined to define the relative domain orientation and interdomain distances, respectively.
A salient feature of our calculation protocol is the semirigid-body assembly of structural domains. The reasons for this are: 1) domains are typically compact, predominantly rigid, and of a small size that is amenable to standard structure determination by X-ray crystallography or NMR spectroscopy; 2) atomic coordinates of isolated domains or subunits from a complex of interest are often available in the Protein Data Bank; 3) in the absence of a precalculated domain structure, homology modeling, combined with chemical shift information, [12,13] may be used to generate a reasonable initial domain structure. The generalized protocol consists of four steps (Figure 1): 1) local refinement of the template domains using RDCs and backbone dihedral angles derived from 13 C secondary chemical shifts; 2) randomization of regions for which no prior structural knowledge is available (e.g. linkers) and global domain orientation from RDC data; 3) application of domain-domain distance restraints from PREs; 4) structure calculation using the RDC and PRE data.
Our approach is demonstrated with the structure determination of the tandem RNA recognition motif (RRM) domains (RRM12) of the human splicing factor U2AF65 bound to a nine-uridine (U9) RNA oligonucleotide. U2AF65, together with U2AF35 and Splicing Factor 1, recognizes the 3' splice site (Figure 1) during the assembly of the spliceosome. [14] NMR 15 N relaxation data of the U2AF65 RRM12/U9 RNA complex indicate that the two RRMs and the RNA tumble together. Thus, the two domains interact even though they are connected by a 20-residue flexible linker (Mackereth et al., in preparation).
The first step of the protocol is to evaluate whether the input domain structures available in the PDB [15,16] resemble those in the complex. This comparison is efficiently achieved by comparing secondary chemical shifts and experimental RDC data measured for the complex with those predicted from the available three-dimensional domain structures. If small conformational differences are detected, a local refinement of the backbone conformation of the template structures can be performed using torsion-angle restraints from secondary chemical shifts and the RDC data. [7,[17][18][19] For the RRM12/U9 complex, we recorded H N -N and N-C' RDCs in two alignment media, Pf1-Phage [20] and PEG-hexanol (PEG = poly(ethylene glycol)). [21] The RDC data, together with backbone torsion angle restraints obtained from TALOS [22] and hydrogen-bond-derived distance restraints, were used for local refinement of the RRM1 and RRM2
[*] Dr.