Zooming in on a small multidrug transporter reveals details of asymmetric protonation

Drug resistance poses a major threat to human health. A principal mechanism of multidrug resistance is the active transport of chemically unrelated compounds out of the cell by integral membrane proteins known as multidrug transporters (1). Multidrug transporters are represented by four superfamilies: ABC (ATP-binding cassette), MFS (major facilitator superfamily), RND (resistance-nodulation-division), and SMR (small multidrug resistance) proteins. The SMR family of transporters is the smallest of all, and yet high-resolution atomic structures are lacking and detailed mechanistic understanding remains incomplete. Over the past decade, much progress has been made in characterizing the prototypical SMR member, EmrE from Escherichia coli, which extrudes a wide range of cationic compounds out of cytosol by using the proton gradient across the bacterial inner membrane (2). Cryoelectron microscopy (3), X-ray crystallography (4), NMR spectroscopy (5), and other experimental data established that EmrE is a homodimer, composed of two conformationally distinct, antiparallel subunits, each of which contains four transmembrane helices. The transport cycle of EmrE is believed to follow the alternating access model, in which the substrate/proton binding site, harboring two conserved glutamic acid residues, Glu14^A and Glu^B, are alternately accessible to the periplasm for proton binding or cytoplasm for drug loading, while the dimer is interconverting between the outward- and inward-facing states (Fig. 1) (5, 6). However, the details of the inward–outward conformational exchange and the coupling between proton binding and drug efflux remain unclear. Earlier experiments by Rotem and Schuldiner (7) based on transport assays established a 2:1 stoichiometry for proton:drug antiport; however, recent NMR data by Henzler-Wildman and coworkers challenged this view and raised the possibility that a one-proton stoichiometry is also possible (8, 9).

Fig. 1.

Thermodynamic linkage for the proton-coupled drug transport cycle of EmrE. A 3D diagram is used to describe proton–drug antiport coupled to the conformational exchange between inward- and outward-facing states (indicated by circles and squares). Conformational exchange, protonation, and drug binding are represented by the vertical, horizontal, and out-of-paper lines, respectively. The singly protonated states, which are not included in the traditional two-proton model but have been recently suggested (8, 9), are supported by the computational study of Vermaas et al. (10).

In PNAS, Vermaas et al. (10) report a computational study that combines a battery of state-of-the-art modeling and simulation tools to shed new light on the perplexing drug–proton antiport mechanism of EmrE at the atomic level. To enable the study, the authors first constructed a complete atomic model based on the 3.8-Å resolution, C_α-only crystal structure of EmrE in complex with tetraphenylphosphonium (TPP⁺) (4). The atomic model was then refined to the so-called fully atomic refined model (faRM) using molecular dynamics (MD) flexible fitting to the cryoelectron microscopy density map of EmrE at 7.5-Å resolution (11). The faRM was validated by running 500-ns all-atom MD simulations of EmrE embedded in a fully explicit dimyristoylphosphatidylcholine (DMPC) bilayer, whereby EmrE was in one of the five substrate/proton loading states: A⁻B⁻, A⁻B⁻TPP⁺, A^HB^H, A^HB⁻, and A⁻B^H, which are denoted apo, TPP⁺, A⁺B⁺, A⁺, and B⁺, respectively, in the paper by Vermaas et al. (10). While A⁻B⁻, A⁻B⁻TPP⁺, and A^HB^H are the only possible states in the two-proton model, A^HB⁻ and A⁻B^H are also possible in the one-proton model (Fig. 1). Note that the singly protonated drug-bound state, A^HB⁻TPP⁺ or A⁻B^HTPP⁺, was not studied by Vermaas et al. (10).

Having verified that the refined model is stable in the five loading states, Vermaas et al. (10) examined the hydration of the interior cavity (lumen) of EmrE. As expected from the inward-facing starting structure, A⁻B⁻, A^HB⁻, and A⁻B^H displayed water channels that connect the lumen to the cytoplasmic side. However, in A^HB^H, the lumen spontaneously dehydrated, consistent with the zero net charge. Although a large conformational change was absent, likely due to the limited simulation length, an occluded state is consistent with the recent NMR data of Traaseth and coworkers (12), which suggest that the conformation of EmrE is substantially different at low- and high-pH conditions. Perhaps the most intriguing finding here is that, in the only drug-bound state, A⁻B⁻TPP⁺, a water wire that connects both sides of the membrane transiently formed. Although the lifetime of the water wire may be too short to be relevant, the existence of a leaky state may also cast doubt on the validity of the doubly deprotonated EmrE upon drug binding (discussed below).

To circumvent the limitation in the simulation length, which makes the direct simulation of the inward–outward conformational transition prohibitive, Vermaas et al. (10) applied steered MD to the different loading states, using the rmsd between the initial inward- and target outward-facing structures as the biasing force. The target structure was built by taking advantage of the symmetry of the antisymmetric homodimer. Although rmsd-based biasing may sometimes lead to unrealistic pathways, steered MD allowed the authors to calculate the nonequilibrium work, which serves as a proxy for the transition barrier, and compare it to the conformational exchange rate measured by NMR. The calculated nonequilibrium work for A⁻B⁻TPP⁺ is the highest among all of the loading states, and the nonequilibrium work for A⁻B^H is higher than A^HB⁻. The former finding is consistent with the recent NMR data, which showed that the exchange rate of the TPP⁺-bound state is 50 times slower than the native state at 37 °C and a pH of 6.9 (13).

Simulations also allowed Vermaas et al. (10) to closely examine the interactions in the EmrE structure. Most significantly, they found that the hydrogen-bonding pattern of Glu14 is different in the two subunits. When Glu14^A is deprotonated, it accepts a water-mediated hydrogen bond from the hydroxyl group of Tyr60^B; however, this interaction does not exist between Glu14^B and Tyr60^A. This structural asymmetry is consistent with the two distinct sets of NMR chemical shifts shown in an earlier experiment (14). Considering that Tyr60 is absolutely conserved in the SMR family and mutation Y60F abolishes the transport activity of EmrE (15), the authors went on to hypothesize that the asymmetric hydrogen bond locks the two subunits in place, hindering the inward–outward transition, and protonation of Glu14 unlocks the electrostatic lock. The higher nonequilibrium work for A⁻B^H relative to A^HB⁻ is consistent with this hypothesis.

The presence of the asymmetric hydrogen bond suggests that the deprotonated Glu14^A may be more stabilized than Glu14^B. As such, Glu14^B may be the preferred proton binding site in the singly protonated state (Fig. 1). To test this hypothesis, Vermaas et al. (10) calculated the free energy difference between A⁻B^H and A^HB⁻ using replica-exchange thermodynamic integration. Indeed, A⁻B^H is more stable than A^HB⁻ by 1.5 kcal/mol, which means that the proton prefers to bind subunit B. Although the calculation does not inform the pK_as of the two glutamic acids, it is remarkable in that it substantiates the possibility that drug efflux requires only one proton. The most recent NMR data of Henzler-Wildman and coworkers (8, 9) showed that the drug-free EmrE exhibits more than two protonation states, A⁻B⁻ and A^HB^H, and the chemical shifts can be better fit with a two-step model, in which the first protonation step, A⁻B⁻ → A⁻B^H/A^HB⁻, has a pK_a of 8.5 $\pm$ 0.2 at 25 °C and the

In PNAS, Vermaas et al. report a computational study that combines a battery of state-of-the-art modeling and simulation tools to shed new light on the perplexing drug–proton antiport mechanism of EmrE at the atomic level.

second protonation step, A⁻B^H/A^HB⁻ → A^HB^H, has a pK_a of 6.8 $\pm$ 0.1 at 25 °C. Thus, given the peri- and cytoplasmic pH conditions of $\sim$6.5 and $\sim$7.5, respectively, the NMR data appear to rule out the biological relevancy of the doubly deprotonated state and suggest that EmrE alternates between the singly protonated, inward-facing and doubly protonated, outward-facing states. The computational data of Vermaas et al. (10) are consistent with this one-proton hypothesis and suggest that, due to an asymmetric, intersubunit hydrogen bond, only one subunit in EmrE can undergo protonation/deprotonation, A⁻B^H $\leftrightarrow$ A^HB^H. In light of the two pK_a values (8), existing experimental data (12, 13, 16) may need to be reinterpretated. Nonetheless, the two-proton model is supported by the previous (6, 7) and new transport assay data (9) and cannot be excluded by the recent NMR experiments (8, 12).

Nature is efficient; why would it require two protons if one is enough to do the job? A one-proton model is operative for the MFS transporter LacY (1), and it has been recently suggested for the RND transporter AcrB (17, 18); EmrE may be no exception. Asymmetric protonation of a carboxylic dyad is a common phenomenon in biology. In AcrB, a proton binding site harbors two aspartic acids but increasing evidence suggests that drug extrusion is coupled to a single proton (17–19). Many enzymes perform catalytic functions by using one carboxylic group as proton donor and the other one as nucleophile (20); this remarkable discrimination may be attributed to the asymmetric hydrogen-bonding environment of the dyad (20). Efficiency aside, a mixture of one- and two-proton models may offer flexibility for an extremely promiscuous transporter such as EmrE.

The work of Vermaas et al. (10) demonstrates the power of state-of-the-art modeling and molecular simulations in advancing our understanding of the structure–function relationships of membrane proteins that remain highly challenging to delineate with wet-laboratory experiments. Recently, continuous constant pH MD with pH replica exchange have been applied to tease out proton-binding residues in ion–proton and drug–proton antiporters (18, 21) and report on the thermodynamics of proton-coupled conformational transition of a small proton channel (22). With the development of new simulation tools and exponential growth of computing hardware speed, a full description of the proton-coupled drug transport cycle of EmrE (Fig. 1) may be within reach in the future.