To model the possible dynamic fluctuations that might be associated with this putative allosteric pathway, we carried out Anisotropic Network Modeling (ANM) (40, 41). ANM is based on a Gaussian network that considers protein structures as elastic networks in which the nodes correspond to the Cα atoms, connected by identical spring constants and in which a Kirchhoff matrix is used to represent the topology of internal contacts. ANM has been used to successfully model the known dynamical modes of the allosteric transition in hemoglobin (49). The three most prominent normal modes obtained from the ANM calculation onA. fulgidus DsrAB reveal significant motion of one heterodimer with respect to the other, in addition to internal modes within monomers and heterodimers (Figure 7, Movies S1-S4). The red squares along the diagonal in the correlation matrix (Figure S5) indicate motions within each domain of each subunit. Within subunits, individual domains exhibit anti-correlated motions (blue). Off diagonal correlations (red) are observed between subunits within a heterodimer and between heterodimers. For example, motions of domain 1 of the DsrA1 subunit are correlated with the ferredoxin domain of the DsrB1 subunit (Figure S5, yellow circles) and with the domain 1 of the DsrB2 subunit (Figure S5, green circle). The largest motions as evaluated by the calculated B-factors (Figure S6), were observed in the slowest mode for the ferredoxin domains of the DsrA1/2 and DsrB1/2 subunits, and in a region of the DsrB1/2 subunits that is near the structural heme and adjacent to one of the FPEC heme road residues, N180B. The point of contact between the two heterodimers is the FPEC heme road residue, T351. It corresponds to a pivot point (low B-factor) for the most significant modes. Interestingly, the shifts in the backbone observed between the two models of the MV2-Eury sequence, particularly apparent in the ferredoxin domains (Figure 6), mimic the rocking like conformational changes associated with the first two normal modes (Figure 7).
In parallel, we resumed the crystallographic refinement using current Phenix software to evaluate the impact of simulating more numerous but smaller segments for TLS refinement of the Archeoglobus fulgidusDsrAB crystal structure, PDB3MM5 (20). Changing from two to twenty-one TLS groups did not affect significantly Rwork/Rfree factors (shifting from 15.8 to 16.1 and from 18.8 to 18.5, respectively) and did not yield a much nicer electron density for one highly flexible heterodimer while the other one is very well resolved and mainly rigid. This result suggested that the overall flexibility relates mainly to rigid body movement of the whole heterodimers. Our ensemble refinement based on 34 conformational states slightly improves the agreement with diffraction data (Rwork: 15.3 from 16.1 and Rfree: 18.5 from 18.5) and suggested some flexibility does occur in the crystal state. Consistent with the notion of dynamic displacements of one heterodimer relative to the other, ensemble refinement of the 3mm5 crystal structure yielded one highly resolved heterodimer (but for only small loops actually flexible), while the second appeared more flexible as a whole with some external segments appearing poorly resolved (see Fig S7). This asymmetric behavior relates to the odd crystal packing that is almost absent for the flexible heterodimer while crystal organization relies on contacts involving the other apparently more rigid heterodimer. The B-factors recapitulate the same picture with one ‘cold’ heterodimer and the second heterodimer getting ‘hotter’ from their common interface to its outward surface..