Computational protocol: Computational Study of Correlated Domain Motions in the AcrB Efflux Transporter

Similar protocols

Protocol publication

[…] On the basis of the crystallographic data for AcrB [, , ] it was postulated that the functional rotation is associated with a mechanical transduction of the energy stored in the transmembrane domain because of the proton flux toward the porter domain. In particular, Sennhauser et al. [] presented a scheme, which highlighted the essential conformational changes including those of helix tmH8 related to the efflux process. The helix tmH8 is one of two extended helices (the other is helix 2, tmH2) protruding from the transmembrane domain farther toward the porter domain. Analyzing the conformational differences between the states of the functional rotation (derived from asymmetric crystal structures), tmH8 moves more prominently while tmH2 translates only slightly []. Hence, the behaviour of tmH8 was addressed in more detail. At the same time, tmH8 together with tmH9 is speculated to form a possible entry pathway in the L monomer for substrates that are partitioned in the outer leaflet of the inner membrane [].To quantify structural changes in the protein we calculated the root mean square deviations (RMSDs) of the porter subdomains with respect to the TMD target state T-O-L (L-T-O is the starting state) and the obtained curves are collected in . The decreasing trend of the curves is due to the fact that the reference state is the final state of the TMD simulations. For freeDyn RMSDs (black curves in the panels) of PC1, PN1, and PN2 mainly fluctuated around the starting value, as expected. Only for PC2 we observed a clear decrease, which might indicate that the PC2 domain was not in its equilibrated conformation at the beginning of the simulations (see also ). For the fully steered simulation, fullTMD, the RMSDs (red curves) approached the zero value at the end of the simulation time. This value is usually not reached in TMD simulations due to the finite spring constant employed in the TMD approach. The flexibility of the spring allows for differences between the target and the actually reached conformations. If only the transmembrane domain is steered in the tmDom setup (green curves), PC1, PN1, and PN2 displayed RMSDs similar to those extracted from the unbiased trajectories. PC2 departed initially from the behaviour observed in freeDyn with an increasing RMSD. However, at the end of the simulation the curve for tmDom obtains a value close to the one for the unbiased trajectory.In freeMon (blue curves) only the monomers L and O and not T were forced to undergo the conformational change of the functional rotation. The blue curves in indicated that structural changes in the T monomer were significant but not as large as those observed in the fully biased simulations. Interestingly, by adding the transmembrane domain of the T monomer to the steered portions of the system (freePP simulation, magenta curves) no remarkable changes in the RMSD curves were observed with respect to freeMon. The domains PN1 and PC2 show slightly smaller RMSD values than those for the freeMon simulation while the values for the domains PN2 and PC1 are very similar. As a major finding we state that it is obviously not enough to only steer the transmembrane domain in short TMD simulations to observe differences, as demonstrated by the fact that results of freePP and freeMon simulations behave similarly. At the same time, steering the residues of the neighbouring monomers leads to significant conformational changes toward the target state different from the unbiased case.To quantitatively describe the actual displacement of the individual sections of AcrB in the simulations we evaluated the center-of-mass (CoM) displacements of the subdomains, which are reported in . For three out of the four subdomains no distinct direction of transition can be observed. Only PC2 shows a significant deviation of more than 1 Å for all simulations. To clarify that this is not simply due to the fact that PC2 could be the subdomain with the smallest mass, the number of atoms and the corresponding masses were determined. Subdomain PC1 contains 837 atoms, PC2 608 atoms, PN1 671 atoms, and PN2 579 atoms with the total masses of these subdomains being roughly proportional to the number of atoms. Hence, this measure does not seem to be very informative to quantify the subdomain motions. As alternative, we analyzed the orientations of the subdomains by evaluating the rotational movement of the major principal axis (PA) of the regions. The PAs are defined as the major PAs of the moment of inertia tensor of the respective subdomains and calculated using VMD []. The angle Φ is determined by projecting the movement of the major PA onto the membrane plane. Moreover, the angle with respect to the membrane normal is called Θ. Both angles are defined with respect to the initial conformation (for a graphical representation see Figure S5 in []). Using these angles in addition to the COM motion of the subdomains one can describe their overall movement more accurately.While, in the fullTMD simulation, the protein was driven along a rather distinct path by the TMD forces, the other simulations are characterized by more enhanced flexibility as shown in . Despite the less pronounced transitions of some of the subdomains compared to the fullTMD reference trajectory, major conformational changes especially in subdomains PC2 and PN2 were observed. While, for PC2, a distinct CoM displacement and rotation of the Θ angle was measured, PN2 in addition undergoes a rotational movement in both angles. Note that the latter subdomain did not show any significant CoM displacement during the fullTMD simulation. The Φ angle of PC2 did change by 20° in simulation fullTMD compared to a 10° change in the tmDom trajectory independently of the simulation time. Furthermore, the results concerning the Θ angle of PC2 did not vary substantially with the trajectory length. Other than minor changes in the Φ angle, significant conformational changes did not occur in the PN1 domain during the tmDom simulation. Apart from PN2, only the Φ angle of PC2 shows a distinct direction of change in all steered trajectories, which is probably due to the extended tmH8 helix and to the spatial proximity of the PC2 and transmembrane domains. The observed transitions of PN2 have smaller amplitudes than those of PC2 but do not seem to be dependent on the specific TMD selection. The comparison of the steering schemes tmDom and fullTMD indicated a correlation between the extrusion of the substrate from DP and the transition of PN2 (data not shown). [...] Since the simulation protocol is the same as that in our previous studies [–] we only list some major features here. Both of the unbiased MD and the TMD simulations were performed using the parallel MD code NAMD 2.7b1 []. For all amino acids, their standard protonation states were considered, that is, the states as for pH 7. After an equilibration procedure [–] the MD simulations were performed with a 1 fs time step in an NpT ensemble at 310 K and 1.013 bar. The functional rotation was enforced using TMD [] (built-in module of NAMD) which allows inducing conformation changes between two known states. In the present investigation different parts of the protein were steered using this approach. The force constant per atom was chose to be k = 3 kcal/(mol Å2). The setup, the analyses, and the atomic-level figures were performed using VMD [].To investigate the intra- and intermonomeric interactions, correlation matrices have been calculated using the program g_covar from the Gromacs package []. This tool computes the Pearson correlation of a set of atoms, in this case of all Cα atoms of the protein. Reference [] describes this approach as inapplicable to study three-dimensional protein systems since the Pearson correlation does only consider colinearly correlated motions of two atoms. Hence, more elaborated correlations cannot be estimated using this method. Therefore, Lange and Grubmüller [] developed a new method which they called “generalized correlation” and which is supposed to be able to cover these correlations as well and has been applied in the present study. […]

Pipeline specifications

Software tools VMD, NAMD, GROMACS
Application Protein structure analysis
Organisms Escherichia coli